First Principles Studies of the Surface
Chemistry of NiFe2O4 and NixCo3-xO4 Spinel
Oxides
XIAO SHI
A DISSERTATION
PRESENTED TO THE FACULTY
OF PRINCETON UNIVERSITY
IN CANDIDACY FOR THE DEGREE
OF DOCTOR OF PHILOSOPHY
RECOMMENDED FOR ACCEPTANCE
BY THE DEPARTMENT OF
CHEMISTRY
ADVISERS ANNABELLA SELLONI
STEVEN BERNASEK
JUNE 2018
copy Copyright by Xiao Shi 2018
All rights reserved
iii
Abstract Nickel cobaltite NiCo2O4 and nickel ferrite NiFe2O4 are spinel oxides with interesting
catalytic properties Nickel cobaltite oxidizes carbon monoxide and methane while
nickel ferrite is an electrocatalyst for water oxidation These materials have been recently
the focus of intense research aimed at modifying their activities and improving their
performances This thesis describes our theoretical studies of the structural and electronic
properties of nickel cobaltite and nickel ferrite their surfaces and their interactions with
probe molecules
The inverse spinel nickel cobaltite is a promising technological material with complex
electronic and magnetic properties Understanding these properties is important for the
development of novel electronic devices and as a basis for the study of their surface and
catalytic properties We have investigated the bulk electronic and magnetic properties of
nickel cobaltite using Density Functional Theory (DFT) calculations augmented with on-
site Hubbard U repulsion on 3d electrons (DFT+U) Starting from an analysis of nickel
doped cobalt oxides we found that nickel acts as a p-type dopant in Co3O4 NiCo2O4 has
a ferrimagnetic half-metallic ground state with fractional valence on Ni and Co cations at
tetrahedral sites (Td) caused by the partial occupancy of Ni and Co(Td)rsquos eg states We
also determined the formation energies of two relevant defects namely NiharrCo(Td)
exchanges and oxygen vacancies as a function of the values of the U terms Facile
NiharrCo(Td) exchange as observed experimentally was obtained using U values that are
significantly smaller than those predicted by linear response theory Our computed bulk
O-vacancy formation energies suggest that NiCo2O4 is an active oxidant similar to
Co3O4
We next extend our study to NiCo2O4 (NCO) surfaces focusing on the structure defects
and reactivity of (001) surfaces Our results suggest that the formation of surface oxygen
vacancies (VO) on the NCO (001) surface is strongly affected by the neighboring cation
in the 3rd layer In particular Ni in the 3rd layer significantly reduces the VO formation
energy As a result VO formation is generally much easier on NCO (001) than on Co3O4
(001) surfaces suggesting that NCO may be a better catalyst than Co3O4 for oxidation
iv
reactions based on the Mars Van Krevelen mechanism VOs on reduced NCO surfaces
can be healed through dissociative water adsorption at room temperature In contrast
adsorption of molecular oxygen at VOs is energetically unfavorable under ambient
conditions suggesting that O2 adsorption may represent the thermodynamic limiting step
for oxidation reactions on NCO (001) surfaces
We again use DFT+U calculations to investigate the mechanism of the low temperature
CO oxidation reaction (COOR) on Co3O4(110)(001) and NiCo2O4(001) as well as
methane oxidation on NiCo2O4(001) Our results indicate that the COOR is controlled by
the thermodynamics of surface re-oxidation on (001) surfaces and by the kinetic barrier
for CO2 formation on the on Co3O4 (110) surface The COOR is inhibited by water
adsorption at surface oxygen vacancies For methane oxidation the computed barrier of
the first C-H bond agrees well with experimental observations
Nickel ferrite NiFe2O4 is another spinel oxide with interesting properties and
applications particularly as a catalyst for water oxidation We have used DFT+U
calculations to study the structure electronic properties and energetics of the
NiFe2O4(001) surface and its interaction with water both in the absence and in the
presence of surface oxygen vacancies In a humid environment water adsorbs
dissociatively on the surface oxygen vacancies leading to the formation of surface
hydroxyls At high temperature water desorbs leaving a surface containing oxygen
vacancies These defects could represent useful reactive sites for various catalytic
reactions CO and methane oxidation on NiFe2O4 are slightly less favorable in
comparison to NiCo2O4 even though the reaction pathways are similar
v
Acknowledgement
I would like to start by thanking my advisors Prof Annabella Selloni who offered me
the opportunity to work in theoretical and computational chemistry her genial and
kindness helped me a lot through my graduate school life and her altitude toward science
helped me a lot in building a solid and rigorous research altitude Prof Steven Bernasek
who offered me the opportunity to work on this project on spinel oxides his gentle nature
helped me a lot in practicing experiment and have a broader wide in research Without
their support and guidance this dissertation would not be possible
I thank the members in my thesis and general exam committees Prof Robert Cava Prof
Andrew Bocarsly and Prof Carnet Chan for their support on the general exam on the
dissertation and over the years
I would also like to thank Prof Zhenhua Li my mentor at Fudan University who brought
me to the field of computational chemistry
Irsquom also very grateful to the members of Selloni Car and Bernasek Groups especially
Dr Yefei Li Dr Sencer Selcuk Hsinyu Ko Jiming Sun Lukas Muechler Matt Vallon
and Matt Frith for their insightful discussions and helps I also enjoyed the friendship
with them Irsquom also thankful to Duyu Chen and Boxiao Zheng outside my group for
sharing and discussion their research and for their friendship
This list would be incomplete without Robert LrsquoEsperance who has been very helpful and
encouraging through and after my teaching career and Meredith LaSalle-Tarantin and
Meghan Krause who are always very optimistic and kind graduate administrators who
helped me a lot through the whole graduate school
Outside Frick I am also grateful to Ruixiang Zhang my old roommate to Xue Wu who
has almost identical driver license photo with mine to Chen Zhao to Jintao Zhang and
shared a lot of happy moment with them They made my life at graduate school more
wonderful than I could imagine
vi
In the end I would like to thank my family who are always with me and give me full
support regardless of the distance which is almost halfway round the world
vii
Table of Contents Chapter I Introduction 1
1 Nickel Cobaltite NiCo2O4 2
2 Surface Oxidation Reactions on Spinel Cobaltite (MCo2O4) 4
21 CO Oxidation 4
22 Methane Oxidation5
3 Nickel Ferrite NiFe2O4 5
4 Organization of this Thesis 7
5 References 9
Chapter II Method 14
1 Basics of Quantum Chemistry 14
2 Density Functional Theory 16
21 Basis Sets and Pseudopotentials 18
22 Self-Interaction Correction 19
23 Structural Optimization 21
3 Nudged Elastic Band Method 22
4 References 24
Chapter III Formation electronic structure and defects of Ni substituted spinel
cobalt oxide 26
1 Introduction 26
2 Computational details 27
3 Results and Discussion 28
31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1) 28
32 Structure bonding and electronic properties of NCO 31
33 NiharrCo exchanges 36
34 Oxygen vacancies 40
4 Conclusions 43
5 References 45
viii
Chapter IV Oxygen deficiency and reactivity of spinel NiCo2O4 (001) surfaces 49
1 Introduction 49
2 Methods and Models 50
3 Results and Discussion 51
31 Pristine (001) (100) surfaces 51
32 Surface oxygen vacancies 56
33 Water adsorption 65
34 Oxygen adsorption 69
4 Conclusions 73
5 References 75
Chapter V Mechanism and activity of the oxidation reactions (CO and methane)
cobaltite spinels (NCO and Co3O4) 78
1 Introduction 78
2 Methods and Models 79
3 Results and Discussion 81
31 Co3O4 (001) and (110) surfaces 81
311 CO adsorption 81
312 CO oxidation 83
32 NCO (001)(100) surfaces 87
33 Methane oxidation on NCO (100)mix surface90
331 Methane adsorption and first C-H bond breaking 90
332 Second C-H bond breaking 92
4 Conclusions 92
5 References 94
Chapter VI Structure of the NiFe2O4(001) surface in contact with gaseous O2 and
water vapour and oxidation reactions for CO methane 97
1 Introduction 97
2 Methods and Models 98
3 Results and Discussion 101
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
copy Copyright by Xiao Shi 2018
All rights reserved
iii
Abstract Nickel cobaltite NiCo2O4 and nickel ferrite NiFe2O4 are spinel oxides with interesting
catalytic properties Nickel cobaltite oxidizes carbon monoxide and methane while
nickel ferrite is an electrocatalyst for water oxidation These materials have been recently
the focus of intense research aimed at modifying their activities and improving their
performances This thesis describes our theoretical studies of the structural and electronic
properties of nickel cobaltite and nickel ferrite their surfaces and their interactions with
probe molecules
The inverse spinel nickel cobaltite is a promising technological material with complex
electronic and magnetic properties Understanding these properties is important for the
development of novel electronic devices and as a basis for the study of their surface and
catalytic properties We have investigated the bulk electronic and magnetic properties of
nickel cobaltite using Density Functional Theory (DFT) calculations augmented with on-
site Hubbard U repulsion on 3d electrons (DFT+U) Starting from an analysis of nickel
doped cobalt oxides we found that nickel acts as a p-type dopant in Co3O4 NiCo2O4 has
a ferrimagnetic half-metallic ground state with fractional valence on Ni and Co cations at
tetrahedral sites (Td) caused by the partial occupancy of Ni and Co(Td)rsquos eg states We
also determined the formation energies of two relevant defects namely NiharrCo(Td)
exchanges and oxygen vacancies as a function of the values of the U terms Facile
NiharrCo(Td) exchange as observed experimentally was obtained using U values that are
significantly smaller than those predicted by linear response theory Our computed bulk
O-vacancy formation energies suggest that NiCo2O4 is an active oxidant similar to
Co3O4
We next extend our study to NiCo2O4 (NCO) surfaces focusing on the structure defects
and reactivity of (001) surfaces Our results suggest that the formation of surface oxygen
vacancies (VO) on the NCO (001) surface is strongly affected by the neighboring cation
in the 3rd layer In particular Ni in the 3rd layer significantly reduces the VO formation
energy As a result VO formation is generally much easier on NCO (001) than on Co3O4
(001) surfaces suggesting that NCO may be a better catalyst than Co3O4 for oxidation
iv
reactions based on the Mars Van Krevelen mechanism VOs on reduced NCO surfaces
can be healed through dissociative water adsorption at room temperature In contrast
adsorption of molecular oxygen at VOs is energetically unfavorable under ambient
conditions suggesting that O2 adsorption may represent the thermodynamic limiting step
for oxidation reactions on NCO (001) surfaces
We again use DFT+U calculations to investigate the mechanism of the low temperature
CO oxidation reaction (COOR) on Co3O4(110)(001) and NiCo2O4(001) as well as
methane oxidation on NiCo2O4(001) Our results indicate that the COOR is controlled by
the thermodynamics of surface re-oxidation on (001) surfaces and by the kinetic barrier
for CO2 formation on the on Co3O4 (110) surface The COOR is inhibited by water
adsorption at surface oxygen vacancies For methane oxidation the computed barrier of
the first C-H bond agrees well with experimental observations
Nickel ferrite NiFe2O4 is another spinel oxide with interesting properties and
applications particularly as a catalyst for water oxidation We have used DFT+U
calculations to study the structure electronic properties and energetics of the
NiFe2O4(001) surface and its interaction with water both in the absence and in the
presence of surface oxygen vacancies In a humid environment water adsorbs
dissociatively on the surface oxygen vacancies leading to the formation of surface
hydroxyls At high temperature water desorbs leaving a surface containing oxygen
vacancies These defects could represent useful reactive sites for various catalytic
reactions CO and methane oxidation on NiFe2O4 are slightly less favorable in
comparison to NiCo2O4 even though the reaction pathways are similar
v
Acknowledgement
I would like to start by thanking my advisors Prof Annabella Selloni who offered me
the opportunity to work in theoretical and computational chemistry her genial and
kindness helped me a lot through my graduate school life and her altitude toward science
helped me a lot in building a solid and rigorous research altitude Prof Steven Bernasek
who offered me the opportunity to work on this project on spinel oxides his gentle nature
helped me a lot in practicing experiment and have a broader wide in research Without
their support and guidance this dissertation would not be possible
I thank the members in my thesis and general exam committees Prof Robert Cava Prof
Andrew Bocarsly and Prof Carnet Chan for their support on the general exam on the
dissertation and over the years
I would also like to thank Prof Zhenhua Li my mentor at Fudan University who brought
me to the field of computational chemistry
Irsquom also very grateful to the members of Selloni Car and Bernasek Groups especially
Dr Yefei Li Dr Sencer Selcuk Hsinyu Ko Jiming Sun Lukas Muechler Matt Vallon
and Matt Frith for their insightful discussions and helps I also enjoyed the friendship
with them Irsquom also thankful to Duyu Chen and Boxiao Zheng outside my group for
sharing and discussion their research and for their friendship
This list would be incomplete without Robert LrsquoEsperance who has been very helpful and
encouraging through and after my teaching career and Meredith LaSalle-Tarantin and
Meghan Krause who are always very optimistic and kind graduate administrators who
helped me a lot through the whole graduate school
Outside Frick I am also grateful to Ruixiang Zhang my old roommate to Xue Wu who
has almost identical driver license photo with mine to Chen Zhao to Jintao Zhang and
shared a lot of happy moment with them They made my life at graduate school more
wonderful than I could imagine
vi
In the end I would like to thank my family who are always with me and give me full
support regardless of the distance which is almost halfway round the world
vii
Table of Contents Chapter I Introduction 1
1 Nickel Cobaltite NiCo2O4 2
2 Surface Oxidation Reactions on Spinel Cobaltite (MCo2O4) 4
21 CO Oxidation 4
22 Methane Oxidation5
3 Nickel Ferrite NiFe2O4 5
4 Organization of this Thesis 7
5 References 9
Chapter II Method 14
1 Basics of Quantum Chemistry 14
2 Density Functional Theory 16
21 Basis Sets and Pseudopotentials 18
22 Self-Interaction Correction 19
23 Structural Optimization 21
3 Nudged Elastic Band Method 22
4 References 24
Chapter III Formation electronic structure and defects of Ni substituted spinel
cobalt oxide 26
1 Introduction 26
2 Computational details 27
3 Results and Discussion 28
31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1) 28
32 Structure bonding and electronic properties of NCO 31
33 NiharrCo exchanges 36
34 Oxygen vacancies 40
4 Conclusions 43
5 References 45
viii
Chapter IV Oxygen deficiency and reactivity of spinel NiCo2O4 (001) surfaces 49
1 Introduction 49
2 Methods and Models 50
3 Results and Discussion 51
31 Pristine (001) (100) surfaces 51
32 Surface oxygen vacancies 56
33 Water adsorption 65
34 Oxygen adsorption 69
4 Conclusions 73
5 References 75
Chapter V Mechanism and activity of the oxidation reactions (CO and methane)
cobaltite spinels (NCO and Co3O4) 78
1 Introduction 78
2 Methods and Models 79
3 Results and Discussion 81
31 Co3O4 (001) and (110) surfaces 81
311 CO adsorption 81
312 CO oxidation 83
32 NCO (001)(100) surfaces 87
33 Methane oxidation on NCO (100)mix surface90
331 Methane adsorption and first C-H bond breaking 90
332 Second C-H bond breaking 92
4 Conclusions 92
5 References 94
Chapter VI Structure of the NiFe2O4(001) surface in contact with gaseous O2 and
water vapour and oxidation reactions for CO methane 97
1 Introduction 97
2 Methods and Models 98
3 Results and Discussion 101
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
iii
Abstract Nickel cobaltite NiCo2O4 and nickel ferrite NiFe2O4 are spinel oxides with interesting
catalytic properties Nickel cobaltite oxidizes carbon monoxide and methane while
nickel ferrite is an electrocatalyst for water oxidation These materials have been recently
the focus of intense research aimed at modifying their activities and improving their
performances This thesis describes our theoretical studies of the structural and electronic
properties of nickel cobaltite and nickel ferrite their surfaces and their interactions with
probe molecules
The inverse spinel nickel cobaltite is a promising technological material with complex
electronic and magnetic properties Understanding these properties is important for the
development of novel electronic devices and as a basis for the study of their surface and
catalytic properties We have investigated the bulk electronic and magnetic properties of
nickel cobaltite using Density Functional Theory (DFT) calculations augmented with on-
site Hubbard U repulsion on 3d electrons (DFT+U) Starting from an analysis of nickel
doped cobalt oxides we found that nickel acts as a p-type dopant in Co3O4 NiCo2O4 has
a ferrimagnetic half-metallic ground state with fractional valence on Ni and Co cations at
tetrahedral sites (Td) caused by the partial occupancy of Ni and Co(Td)rsquos eg states We
also determined the formation energies of two relevant defects namely NiharrCo(Td)
exchanges and oxygen vacancies as a function of the values of the U terms Facile
NiharrCo(Td) exchange as observed experimentally was obtained using U values that are
significantly smaller than those predicted by linear response theory Our computed bulk
O-vacancy formation energies suggest that NiCo2O4 is an active oxidant similar to
Co3O4
We next extend our study to NiCo2O4 (NCO) surfaces focusing on the structure defects
and reactivity of (001) surfaces Our results suggest that the formation of surface oxygen
vacancies (VO) on the NCO (001) surface is strongly affected by the neighboring cation
in the 3rd layer In particular Ni in the 3rd layer significantly reduces the VO formation
energy As a result VO formation is generally much easier on NCO (001) than on Co3O4
(001) surfaces suggesting that NCO may be a better catalyst than Co3O4 for oxidation
iv
reactions based on the Mars Van Krevelen mechanism VOs on reduced NCO surfaces
can be healed through dissociative water adsorption at room temperature In contrast
adsorption of molecular oxygen at VOs is energetically unfavorable under ambient
conditions suggesting that O2 adsorption may represent the thermodynamic limiting step
for oxidation reactions on NCO (001) surfaces
We again use DFT+U calculations to investigate the mechanism of the low temperature
CO oxidation reaction (COOR) on Co3O4(110)(001) and NiCo2O4(001) as well as
methane oxidation on NiCo2O4(001) Our results indicate that the COOR is controlled by
the thermodynamics of surface re-oxidation on (001) surfaces and by the kinetic barrier
for CO2 formation on the on Co3O4 (110) surface The COOR is inhibited by water
adsorption at surface oxygen vacancies For methane oxidation the computed barrier of
the first C-H bond agrees well with experimental observations
Nickel ferrite NiFe2O4 is another spinel oxide with interesting properties and
applications particularly as a catalyst for water oxidation We have used DFT+U
calculations to study the structure electronic properties and energetics of the
NiFe2O4(001) surface and its interaction with water both in the absence and in the
presence of surface oxygen vacancies In a humid environment water adsorbs
dissociatively on the surface oxygen vacancies leading to the formation of surface
hydroxyls At high temperature water desorbs leaving a surface containing oxygen
vacancies These defects could represent useful reactive sites for various catalytic
reactions CO and methane oxidation on NiFe2O4 are slightly less favorable in
comparison to NiCo2O4 even though the reaction pathways are similar
v
Acknowledgement
I would like to start by thanking my advisors Prof Annabella Selloni who offered me
the opportunity to work in theoretical and computational chemistry her genial and
kindness helped me a lot through my graduate school life and her altitude toward science
helped me a lot in building a solid and rigorous research altitude Prof Steven Bernasek
who offered me the opportunity to work on this project on spinel oxides his gentle nature
helped me a lot in practicing experiment and have a broader wide in research Without
their support and guidance this dissertation would not be possible
I thank the members in my thesis and general exam committees Prof Robert Cava Prof
Andrew Bocarsly and Prof Carnet Chan for their support on the general exam on the
dissertation and over the years
I would also like to thank Prof Zhenhua Li my mentor at Fudan University who brought
me to the field of computational chemistry
Irsquom also very grateful to the members of Selloni Car and Bernasek Groups especially
Dr Yefei Li Dr Sencer Selcuk Hsinyu Ko Jiming Sun Lukas Muechler Matt Vallon
and Matt Frith for their insightful discussions and helps I also enjoyed the friendship
with them Irsquom also thankful to Duyu Chen and Boxiao Zheng outside my group for
sharing and discussion their research and for their friendship
This list would be incomplete without Robert LrsquoEsperance who has been very helpful and
encouraging through and after my teaching career and Meredith LaSalle-Tarantin and
Meghan Krause who are always very optimistic and kind graduate administrators who
helped me a lot through the whole graduate school
Outside Frick I am also grateful to Ruixiang Zhang my old roommate to Xue Wu who
has almost identical driver license photo with mine to Chen Zhao to Jintao Zhang and
shared a lot of happy moment with them They made my life at graduate school more
wonderful than I could imagine
vi
In the end I would like to thank my family who are always with me and give me full
support regardless of the distance which is almost halfway round the world
vii
Table of Contents Chapter I Introduction 1
1 Nickel Cobaltite NiCo2O4 2
2 Surface Oxidation Reactions on Spinel Cobaltite (MCo2O4) 4
21 CO Oxidation 4
22 Methane Oxidation5
3 Nickel Ferrite NiFe2O4 5
4 Organization of this Thesis 7
5 References 9
Chapter II Method 14
1 Basics of Quantum Chemistry 14
2 Density Functional Theory 16
21 Basis Sets and Pseudopotentials 18
22 Self-Interaction Correction 19
23 Structural Optimization 21
3 Nudged Elastic Band Method 22
4 References 24
Chapter III Formation electronic structure and defects of Ni substituted spinel
cobalt oxide 26
1 Introduction 26
2 Computational details 27
3 Results and Discussion 28
31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1) 28
32 Structure bonding and electronic properties of NCO 31
33 NiharrCo exchanges 36
34 Oxygen vacancies 40
4 Conclusions 43
5 References 45
viii
Chapter IV Oxygen deficiency and reactivity of spinel NiCo2O4 (001) surfaces 49
1 Introduction 49
2 Methods and Models 50
3 Results and Discussion 51
31 Pristine (001) (100) surfaces 51
32 Surface oxygen vacancies 56
33 Water adsorption 65
34 Oxygen adsorption 69
4 Conclusions 73
5 References 75
Chapter V Mechanism and activity of the oxidation reactions (CO and methane)
cobaltite spinels (NCO and Co3O4) 78
1 Introduction 78
2 Methods and Models 79
3 Results and Discussion 81
31 Co3O4 (001) and (110) surfaces 81
311 CO adsorption 81
312 CO oxidation 83
32 NCO (001)(100) surfaces 87
33 Methane oxidation on NCO (100)mix surface90
331 Methane adsorption and first C-H bond breaking 90
332 Second C-H bond breaking 92
4 Conclusions 92
5 References 94
Chapter VI Structure of the NiFe2O4(001) surface in contact with gaseous O2 and
water vapour and oxidation reactions for CO methane 97
1 Introduction 97
2 Methods and Models 98
3 Results and Discussion 101
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
iv
reactions based on the Mars Van Krevelen mechanism VOs on reduced NCO surfaces
can be healed through dissociative water adsorption at room temperature In contrast
adsorption of molecular oxygen at VOs is energetically unfavorable under ambient
conditions suggesting that O2 adsorption may represent the thermodynamic limiting step
for oxidation reactions on NCO (001) surfaces
We again use DFT+U calculations to investigate the mechanism of the low temperature
CO oxidation reaction (COOR) on Co3O4(110)(001) and NiCo2O4(001) as well as
methane oxidation on NiCo2O4(001) Our results indicate that the COOR is controlled by
the thermodynamics of surface re-oxidation on (001) surfaces and by the kinetic barrier
for CO2 formation on the on Co3O4 (110) surface The COOR is inhibited by water
adsorption at surface oxygen vacancies For methane oxidation the computed barrier of
the first C-H bond agrees well with experimental observations
Nickel ferrite NiFe2O4 is another spinel oxide with interesting properties and
applications particularly as a catalyst for water oxidation We have used DFT+U
calculations to study the structure electronic properties and energetics of the
NiFe2O4(001) surface and its interaction with water both in the absence and in the
presence of surface oxygen vacancies In a humid environment water adsorbs
dissociatively on the surface oxygen vacancies leading to the formation of surface
hydroxyls At high temperature water desorbs leaving a surface containing oxygen
vacancies These defects could represent useful reactive sites for various catalytic
reactions CO and methane oxidation on NiFe2O4 are slightly less favorable in
comparison to NiCo2O4 even though the reaction pathways are similar
v
Acknowledgement
I would like to start by thanking my advisors Prof Annabella Selloni who offered me
the opportunity to work in theoretical and computational chemistry her genial and
kindness helped me a lot through my graduate school life and her altitude toward science
helped me a lot in building a solid and rigorous research altitude Prof Steven Bernasek
who offered me the opportunity to work on this project on spinel oxides his gentle nature
helped me a lot in practicing experiment and have a broader wide in research Without
their support and guidance this dissertation would not be possible
I thank the members in my thesis and general exam committees Prof Robert Cava Prof
Andrew Bocarsly and Prof Carnet Chan for their support on the general exam on the
dissertation and over the years
I would also like to thank Prof Zhenhua Li my mentor at Fudan University who brought
me to the field of computational chemistry
Irsquom also very grateful to the members of Selloni Car and Bernasek Groups especially
Dr Yefei Li Dr Sencer Selcuk Hsinyu Ko Jiming Sun Lukas Muechler Matt Vallon
and Matt Frith for their insightful discussions and helps I also enjoyed the friendship
with them Irsquom also thankful to Duyu Chen and Boxiao Zheng outside my group for
sharing and discussion their research and for their friendship
This list would be incomplete without Robert LrsquoEsperance who has been very helpful and
encouraging through and after my teaching career and Meredith LaSalle-Tarantin and
Meghan Krause who are always very optimistic and kind graduate administrators who
helped me a lot through the whole graduate school
Outside Frick I am also grateful to Ruixiang Zhang my old roommate to Xue Wu who
has almost identical driver license photo with mine to Chen Zhao to Jintao Zhang and
shared a lot of happy moment with them They made my life at graduate school more
wonderful than I could imagine
vi
In the end I would like to thank my family who are always with me and give me full
support regardless of the distance which is almost halfway round the world
vii
Table of Contents Chapter I Introduction 1
1 Nickel Cobaltite NiCo2O4 2
2 Surface Oxidation Reactions on Spinel Cobaltite (MCo2O4) 4
21 CO Oxidation 4
22 Methane Oxidation5
3 Nickel Ferrite NiFe2O4 5
4 Organization of this Thesis 7
5 References 9
Chapter II Method 14
1 Basics of Quantum Chemistry 14
2 Density Functional Theory 16
21 Basis Sets and Pseudopotentials 18
22 Self-Interaction Correction 19
23 Structural Optimization 21
3 Nudged Elastic Band Method 22
4 References 24
Chapter III Formation electronic structure and defects of Ni substituted spinel
cobalt oxide 26
1 Introduction 26
2 Computational details 27
3 Results and Discussion 28
31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1) 28
32 Structure bonding and electronic properties of NCO 31
33 NiharrCo exchanges 36
34 Oxygen vacancies 40
4 Conclusions 43
5 References 45
viii
Chapter IV Oxygen deficiency and reactivity of spinel NiCo2O4 (001) surfaces 49
1 Introduction 49
2 Methods and Models 50
3 Results and Discussion 51
31 Pristine (001) (100) surfaces 51
32 Surface oxygen vacancies 56
33 Water adsorption 65
34 Oxygen adsorption 69
4 Conclusions 73
5 References 75
Chapter V Mechanism and activity of the oxidation reactions (CO and methane)
cobaltite spinels (NCO and Co3O4) 78
1 Introduction 78
2 Methods and Models 79
3 Results and Discussion 81
31 Co3O4 (001) and (110) surfaces 81
311 CO adsorption 81
312 CO oxidation 83
32 NCO (001)(100) surfaces 87
33 Methane oxidation on NCO (100)mix surface90
331 Methane adsorption and first C-H bond breaking 90
332 Second C-H bond breaking 92
4 Conclusions 92
5 References 94
Chapter VI Structure of the NiFe2O4(001) surface in contact with gaseous O2 and
water vapour and oxidation reactions for CO methane 97
1 Introduction 97
2 Methods and Models 98
3 Results and Discussion 101
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
v
Acknowledgement
I would like to start by thanking my advisors Prof Annabella Selloni who offered me
the opportunity to work in theoretical and computational chemistry her genial and
kindness helped me a lot through my graduate school life and her altitude toward science
helped me a lot in building a solid and rigorous research altitude Prof Steven Bernasek
who offered me the opportunity to work on this project on spinel oxides his gentle nature
helped me a lot in practicing experiment and have a broader wide in research Without
their support and guidance this dissertation would not be possible
I thank the members in my thesis and general exam committees Prof Robert Cava Prof
Andrew Bocarsly and Prof Carnet Chan for their support on the general exam on the
dissertation and over the years
I would also like to thank Prof Zhenhua Li my mentor at Fudan University who brought
me to the field of computational chemistry
Irsquom also very grateful to the members of Selloni Car and Bernasek Groups especially
Dr Yefei Li Dr Sencer Selcuk Hsinyu Ko Jiming Sun Lukas Muechler Matt Vallon
and Matt Frith for their insightful discussions and helps I also enjoyed the friendship
with them Irsquom also thankful to Duyu Chen and Boxiao Zheng outside my group for
sharing and discussion their research and for their friendship
This list would be incomplete without Robert LrsquoEsperance who has been very helpful and
encouraging through and after my teaching career and Meredith LaSalle-Tarantin and
Meghan Krause who are always very optimistic and kind graduate administrators who
helped me a lot through the whole graduate school
Outside Frick I am also grateful to Ruixiang Zhang my old roommate to Xue Wu who
has almost identical driver license photo with mine to Chen Zhao to Jintao Zhang and
shared a lot of happy moment with them They made my life at graduate school more
wonderful than I could imagine
vi
In the end I would like to thank my family who are always with me and give me full
support regardless of the distance which is almost halfway round the world
vii
Table of Contents Chapter I Introduction 1
1 Nickel Cobaltite NiCo2O4 2
2 Surface Oxidation Reactions on Spinel Cobaltite (MCo2O4) 4
21 CO Oxidation 4
22 Methane Oxidation5
3 Nickel Ferrite NiFe2O4 5
4 Organization of this Thesis 7
5 References 9
Chapter II Method 14
1 Basics of Quantum Chemistry 14
2 Density Functional Theory 16
21 Basis Sets and Pseudopotentials 18
22 Self-Interaction Correction 19
23 Structural Optimization 21
3 Nudged Elastic Band Method 22
4 References 24
Chapter III Formation electronic structure and defects of Ni substituted spinel
cobalt oxide 26
1 Introduction 26
2 Computational details 27
3 Results and Discussion 28
31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1) 28
32 Structure bonding and electronic properties of NCO 31
33 NiharrCo exchanges 36
34 Oxygen vacancies 40
4 Conclusions 43
5 References 45
viii
Chapter IV Oxygen deficiency and reactivity of spinel NiCo2O4 (001) surfaces 49
1 Introduction 49
2 Methods and Models 50
3 Results and Discussion 51
31 Pristine (001) (100) surfaces 51
32 Surface oxygen vacancies 56
33 Water adsorption 65
34 Oxygen adsorption 69
4 Conclusions 73
5 References 75
Chapter V Mechanism and activity of the oxidation reactions (CO and methane)
cobaltite spinels (NCO and Co3O4) 78
1 Introduction 78
2 Methods and Models 79
3 Results and Discussion 81
31 Co3O4 (001) and (110) surfaces 81
311 CO adsorption 81
312 CO oxidation 83
32 NCO (001)(100) surfaces 87
33 Methane oxidation on NCO (100)mix surface90
331 Methane adsorption and first C-H bond breaking 90
332 Second C-H bond breaking 92
4 Conclusions 92
5 References 94
Chapter VI Structure of the NiFe2O4(001) surface in contact with gaseous O2 and
water vapour and oxidation reactions for CO methane 97
1 Introduction 97
2 Methods and Models 98
3 Results and Discussion 101
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
vi
In the end I would like to thank my family who are always with me and give me full
support regardless of the distance which is almost halfway round the world
vii
Table of Contents Chapter I Introduction 1
1 Nickel Cobaltite NiCo2O4 2
2 Surface Oxidation Reactions on Spinel Cobaltite (MCo2O4) 4
21 CO Oxidation 4
22 Methane Oxidation5
3 Nickel Ferrite NiFe2O4 5
4 Organization of this Thesis 7
5 References 9
Chapter II Method 14
1 Basics of Quantum Chemistry 14
2 Density Functional Theory 16
21 Basis Sets and Pseudopotentials 18
22 Self-Interaction Correction 19
23 Structural Optimization 21
3 Nudged Elastic Band Method 22
4 References 24
Chapter III Formation electronic structure and defects of Ni substituted spinel
cobalt oxide 26
1 Introduction 26
2 Computational details 27
3 Results and Discussion 28
31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1) 28
32 Structure bonding and electronic properties of NCO 31
33 NiharrCo exchanges 36
34 Oxygen vacancies 40
4 Conclusions 43
5 References 45
viii
Chapter IV Oxygen deficiency and reactivity of spinel NiCo2O4 (001) surfaces 49
1 Introduction 49
2 Methods and Models 50
3 Results and Discussion 51
31 Pristine (001) (100) surfaces 51
32 Surface oxygen vacancies 56
33 Water adsorption 65
34 Oxygen adsorption 69
4 Conclusions 73
5 References 75
Chapter V Mechanism and activity of the oxidation reactions (CO and methane)
cobaltite spinels (NCO and Co3O4) 78
1 Introduction 78
2 Methods and Models 79
3 Results and Discussion 81
31 Co3O4 (001) and (110) surfaces 81
311 CO adsorption 81
312 CO oxidation 83
32 NCO (001)(100) surfaces 87
33 Methane oxidation on NCO (100)mix surface90
331 Methane adsorption and first C-H bond breaking 90
332 Second C-H bond breaking 92
4 Conclusions 92
5 References 94
Chapter VI Structure of the NiFe2O4(001) surface in contact with gaseous O2 and
water vapour and oxidation reactions for CO methane 97
1 Introduction 97
2 Methods and Models 98
3 Results and Discussion 101
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
vii
Table of Contents Chapter I Introduction 1
1 Nickel Cobaltite NiCo2O4 2
2 Surface Oxidation Reactions on Spinel Cobaltite (MCo2O4) 4
21 CO Oxidation 4
22 Methane Oxidation5
3 Nickel Ferrite NiFe2O4 5
4 Organization of this Thesis 7
5 References 9
Chapter II Method 14
1 Basics of Quantum Chemistry 14
2 Density Functional Theory 16
21 Basis Sets and Pseudopotentials 18
22 Self-Interaction Correction 19
23 Structural Optimization 21
3 Nudged Elastic Band Method 22
4 References 24
Chapter III Formation electronic structure and defects of Ni substituted spinel
cobalt oxide 26
1 Introduction 26
2 Computational details 27
3 Results and Discussion 28
31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1) 28
32 Structure bonding and electronic properties of NCO 31
33 NiharrCo exchanges 36
34 Oxygen vacancies 40
4 Conclusions 43
5 References 45
viii
Chapter IV Oxygen deficiency and reactivity of spinel NiCo2O4 (001) surfaces 49
1 Introduction 49
2 Methods and Models 50
3 Results and Discussion 51
31 Pristine (001) (100) surfaces 51
32 Surface oxygen vacancies 56
33 Water adsorption 65
34 Oxygen adsorption 69
4 Conclusions 73
5 References 75
Chapter V Mechanism and activity of the oxidation reactions (CO and methane)
cobaltite spinels (NCO and Co3O4) 78
1 Introduction 78
2 Methods and Models 79
3 Results and Discussion 81
31 Co3O4 (001) and (110) surfaces 81
311 CO adsorption 81
312 CO oxidation 83
32 NCO (001)(100) surfaces 87
33 Methane oxidation on NCO (100)mix surface90
331 Methane adsorption and first C-H bond breaking 90
332 Second C-H bond breaking 92
4 Conclusions 92
5 References 94
Chapter VI Structure of the NiFe2O4(001) surface in contact with gaseous O2 and
water vapour and oxidation reactions for CO methane 97
1 Introduction 97
2 Methods and Models 98
3 Results and Discussion 101
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
viii
Chapter IV Oxygen deficiency and reactivity of spinel NiCo2O4 (001) surfaces 49
1 Introduction 49
2 Methods and Models 50
3 Results and Discussion 51
31 Pristine (001) (100) surfaces 51
32 Surface oxygen vacancies 56
33 Water adsorption 65
34 Oxygen adsorption 69
4 Conclusions 73
5 References 75
Chapter V Mechanism and activity of the oxidation reactions (CO and methane)
cobaltite spinels (NCO and Co3O4) 78
1 Introduction 78
2 Methods and Models 79
3 Results and Discussion 81
31 Co3O4 (001) and (110) surfaces 81
311 CO adsorption 81
312 CO oxidation 83
32 NCO (001)(100) surfaces 87
33 Methane oxidation on NCO (100)mix surface90
331 Methane adsorption and first C-H bond breaking 90
332 Second C-H bond breaking 92
4 Conclusions 92
5 References 94
Chapter VI Structure of the NiFe2O4(001) surface in contact with gaseous O2 and
water vapour and oxidation reactions for CO methane 97
1 Introduction 97
2 Methods and Models 98
3 Results and Discussion 101
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
ix
31 Bulk properties 101
32 NiFe2O4 (001) surface 104
321 Defect-free surface 104
322 Surface O vacancy 106
33 Water adsorption 108
331 Water adsorption on the defect-free surface 108
332 Water adsorption on the defected surface 109
34 Phase diagram 112
35 CO oxidation 113
36 Methane oxidation 115
4 Summary and Conclusions 116
5 References 117
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83
1
Chapter I
Introduction Spinel oxides with formula AB2O4 form a class of compounds that crystallize in the cubic
crystal system and include a variety of divalent trivalent and tetravalent cations Cations
of the iron group (Fe Co and Ni) - being earth abundant similar in chemical properties
and able to form magnetic materials - have an important place within this family For
example magnetite (Fe3O4) which is known as the most magnetic naturally-occurring
mineral on earth is widely used in audio recording Magnetite is also widely used as a
catalyst in ammonia synthesis
Recently there has been a lot of interest in the catalytic activity of Fe Co and Ni spinel
oxides After the discovery of low temperature CO oxidation on cobalt oxide Co3O4 by
Haruta and Shen1 numerous studies have focused on understanding and modifying the
properties of this material For example Gao et al studied the synthesis and activity of
doped and substituted cobalt oxides for low temperature methane oxidation2 Iron based
spinel oxides have also been investigated extensively For example the prediction that
Fe3O4 partially dissociates water3 has motivated several studies of both water-ferrite
interactions and the catalytic activity iron based spinel oxides Potential catalytic
applications range from harvesting solar energy to water gas shift reactions to reduce
exhaust gases 4-5
CO and methane oxidation on doped cobalt oxides is of special interest 6-7 Among the
different spinel cobaltites nickel cobaltite NiCo2O4 is the most promising material for
this purpose8 On the other hand nickel ferrite NiFe2O4 though not active for CO
oxidation was found to be quite active for water oxidation4 In the following sections we
will give a brief introduction to the properties of NiCo2O4 (denoted NCO in this thesis)
and NiFe2O4 (denoted NFO in this thesis) and their catalytic activities
2
1 Nickel Cobaltite NiCo2O4
Nickel cobaltite (NCO) crystallizes in the spinel structure The oxygen anions form a face
centered cubic lattice while octahedral and tetrahedral sites are partially occupied (12 of
octahedral and 18 of tetrahedral) by cations A B Spinel oxides can have two types of
structure normal spinel and inverse spinel Normal spinels (Figure 1a) have all the
tetrahedral sites occupied by cation A and octahedral sites occupied by cation B Inverse
spinels have all the tetrahedral sites occupied by cation B and octahedral sites equally
occupied by both A and B The symmetry on octahedral sites can further lead to either α
or β type inverse spinel (Figure 1b c) NCO is an inverse spinel and prefers the β type
structure The experimental lattice constant is 8115 Aring9 NCO decomposes into NixCo3-
xO4 and NiO at around 400~600 degC10-12 depending on the synthetic procedure used
Figure 1 Spinel AB2O4 a normal spinel where A cations occupy tetrahedral sites and B
cations occupy octahedral sites b α inverse spinel with B cations occupying tetrahedral
sites and A B occupying octahedral sites c β inverse spinel where octahedral sites
along [001] direction are occupied by either A or B
Being an inverse spinel NCO has Co at tetrahedral sites (Co(Td)) and a mixture of Ni
and Co at octahedral sites (Ni(Oh) and Co(Oh)) with both Co(Td) and Ni(Oh) showing
mixed 2+ and 3+ oxidation states13-14 However NiharrCo exchanges at tetrahedral sites
are frequent10 which leads to a reduced degree of inversion of the structure Co(Td) and
Ni(Oh) are in the high and low spin states respectively and contribute to NCOrsquos
magnetic properties NCO is usually found to be ferrimagnetic15 with Co(Td) and Ni(Oh)
having anti-parallel spins However the synthetic procedure can influence NCOrsquos
magnetic ordering16-17 for example high temperature growth may result in diamagnetic
3
order NCOrsquos electronic properties are also interesting as this material shows high
conductivity as well as an optical band gap of 197 eV Single crystal studies10 18-19 could
provide more detailed information on NCOrsquos electronic and magnetic properties but are
rare due to the difficulty of growing good NCO crystals
The surface properties of NCO are important for understanding its oxidation activity The
most stable surfaces of NCO are (001) (111) and (110) for which surface energies of
129 Jm-2 142 Jm-2 and 160 Jm-2 respectively have been computed20 The (001)
surface is the most common surface with predicted abundance of 548 while the
abundance is only 70 for the (110) surface In the following we shall thus focus on
NCO (001) surfaces to learn about active sites and reaction mechanisms Since NCO is β
type inverse spinel 16 of NCO (001) surfaces contain only Co(Oh) cations 16 contain
only Ni(Oh) cations and the remaining 23 contain 50 Ni(Oh) and 50 Co(Oh)
(Figure 2)
Figure 2 Side views of NCO (001)(100) surfaces a pure Ni(Oh) terminated b pure
Co(Oh) terminated and c mix of Ni(Oh) and Co(Oh) termination
2 Surface Oxidation Reactions on Spinel Cobaltites (MCo2O4) 21 CO Oxidation
CO oxidation on spinel Co3O4 was first discovered over 15 years ago and subsequently
similar studies were performed on other doped cobaltites as well1 21 The reaction was
first observed at room temperature for pre-oxidized Co3O422-23 when cobalt oxide was
heated and cooled in oxygen rich environment before being treated with CO the reaction
was found to occur at temperatures as low as 20 degC However the catalyst was found to
4
slowly deactivate over time and the mechanism of the deactivation remained
controversial hindering further development
As the techniques of growing Co3O4 crystals improved it was easier for researchers to
control their shape and exposed surfaces Co3O4 nanorods were later found to steadily
oxidize CO at temperature as low as -77 degC without significant deactivation1 The
nanorods expose both (110) and (001) surfaces (Figure 3) Co3+ ions originating from
octahedral sites are believed to have a key role in the CO oxidation reaction on the
surface The (110) surface was found to be about three times more active than the (001)
surface21
Figure 3 Side views of the Co3O4 (110) and (001) surfaces Both surfaces expose
Co(Oh) with 3+ charge and two types of oxygen sites On the (110) surface an oxygen
atom bonded to 2 (3) Co cations is denoted O2f (O3f) on the (001) surface oxygen atoms
bonded to 3 Co(Oh) are denoted O1 while oxygen atoms bonded to 2 Co(Oh) and 1
Co(Td) are denoted O2
DFT calculations of CO oxidation on the Co3O4(110) surface24-25 show a strong
dependence on the value of Hubbard U term (see also Ref 26) When U = 0 Co3+ is the
preferred adsorption site for CO27 When Hubbard U is applied to achieve a better
description of the thermodynamic and electronic properties CO tends to be adsorbed on
O2f while pointing toward Co3+ 28 Both O2f and Co3+ sites are crucial for the CO
oxidation reaction on the (110) surface In contrast to the numerous studies for the (110)
5
surface the mechanism of CO oxidation has not been studied for the most abundant (001)
surface
22 Methane Oxidation
In the case of methane oxidation Co3O4 is often used as a support for another catalyst29-30
that is more active at breaking the first C-H bond which is usually the rate determining
step of this reaction31 Co3O4 starts to oxidize methane at around 400 degC30 With Ni
doping notably for Ni05Co25O4 the activation barrier was found to be reduced by 35
KJmol2 Further doping leads to NCO which starts to oxidize methane from 200 degC7
Lattice oxygen atoms are involved in the formation of CO2 during the oxidation process
The activation barrier involving Ni on the (110) surface is lower and thus consistent with
the observation that doping Ni would make methane oxidation easier
A recent DFT study examined the methane oxidation activities of Co3O4 (110) and (001)
surfaces32 The (110) surface was found to be more active than the (001) one though the
first C-H bond breaking appears to be easier on (001) However the low activity of O2
sites on the (001) surface (Figure 3) hinders or even prevents further oxidation Methane
oxidation on the (001) surfaces of Co3O4 and NCO could therefore be of great interest for
achieving partial oxidation of methane
3 Nickel Ferrite NiFe2O4
Nickel ferrite (NFO) is an α-type inverse spinel33 with lattice constant of 833 Aring34
Similar to NCO NFO is also ferrimagnetic35 with the spin of Fe(Td) anti-parallel to
Ni(Oh) and Fe(Oh)36 according to DFT calculations The Neacuteel temperature is about 850
K37 An indirect band gap of 16 eV is observed possibly involving d-d and p-d charge
transfer transitions between the correlated bands of Ni2+ and Fe3+ sites Although NFO is
usually insulating tuning the growth conditions to oxygen free environment would
greatly increase its conductivity38
6
Studies of NFO surfaces are still scarce in comparison to those of bulk NFO The growth
direction of NFO crystals is controlled by the substrate39 eg films grown on MgAl2O4
(001) expose the (001) surface Other synthesis conditions like hydrothermal synthesis
lead to crystals preferentially exposing (111) surfaces NFOrsquos chemical properties are
altogether similar to those of Fe3O4 which is known to partially dissociate water on the
(001) surface and fully dissociate water on (111) surface40 However doping Ni into
Fe3O4 was found to substantially enhance the activity of two step water splitting
reactions41-42 suggesting that NFO might be generally more active than Fe3O4 for
reactions involving water For example this might be the case for the water gas shift
(WGS) reaction which was found to be controlled by water adsorption and dissociation
on the ferrite surface5 Recently NFO was shown to photocatalytically oxidize water with
the help of a photosensitizer and an oxidant4 Possible formation of high valence Ni
(Ni3+Ni4+) might be crucial for this photocatalytic reaction
Theoretical calculations can help understand the fundamentals of NFO interaction with
water and interpret how Ni increases the reactivity of Fe3O4 for catalytic reactions
involving water Previous DFT calculations for the NFO (111) surface showed that the
interaction of water with octahedral Fe sites is enhanced by the presence of Ni relative to
tetrahedral Fe and becomes favored thermodynamically43 Water dissociation on Fe(Oh)
is also a barrier free reaction The activity is believed to come from the strong interaction
between the OH σ orbitals and Fe d orbitals The NFO (001) surface also exposes
Fe(Oh) sites with one of its octahedral ligand site empty and able to bind a water
molecule or an OH group Our theoretical study of water adsorption on the NFO (001)
surface (Figure 4) is discussed in detail in Chapt VI
7
Figure 4 Water adsorption and dissociation on NFO (001) surface with water adsorbed
on oxygen vacancies and on top of Fe and Ni
4 Organization of this Thesis
In the next chapter we review the methods used for this work and will give a brief
summary of the relevant theoretical background
In chapter 3 we investigate the thermodynamic and electronic properties of bulk NCO
As mentioned above NCO exhibits interesting electronic and magnetic properties eg
coexistence of high conductivity with a large band gap whose origin was not well
understood Our results show the existence of fractional valence states for Co(Td) and
Ni(Oh) which largely determine NCOrsquos properties
In chapter 4 we extend our study to NCO (001)(100) surfaces which were predicted to
be the most abundant surfaces of this material20 A recent study of methane oxidation on
NCO7 suggests that the formation of surface oxygen vacancies might be a key to
understand NCOrsquos surface reactivity Focusing on the thermodynamics of (001)(100)
surfaces we show that Ni prefers to segregate to the surface The Ni (Oh) close to the
surface will significantly influence the formation of oxygen vacancies (VO) and thus
8
possibly improve the catalytic performance Furthermore re-oxidation of the reduced
surface is difficult and could possibly limit the activity on (001)(100) surfaces
Chapter 5 examines the mechanism of CO and methane oxidation on NCO (001)(100)
surfaces in comparison to Co3O4 Since experiments show that CO oxidation on Co3O4
is very efficient on the (110) surface whereas the (001) surface is less active21 previous
theoretical studies focused on the Co3O4 (110) surface27 and did not examine the
difference between (110) and (001) surfaces Our results indicate that CO is oxidized
more easily on Co3O4NCO (001) than on the (110) surface due to easier VO formation
On (001) however it is more difficult to recover the active surface after the first
oxidation Our additional study of methane oxidation on NCO (100) shows that the
kinetic barrier for the first C-H bond breaking (the well-known rate determining step of
methane oxidation) is smaller than that reported for doped CeO2 a well know low
temperature methane oxidation catalyst This result thus confirms the strong NCOrsquos
activity for methane activation
In chapter 6 we investigate the structure of the NFO (001) surface and its interaction
with water We study the thermodynamics of water adsorption including the effect of
surface oxygen vacancies Our results show that water adsorbs preferentially on VO sites
where it tends to dissociate Results for CO and methane oxidation on NFO (001) are also
presented
9
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
9 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
10 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10
11 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
12 Lapham D P Tseung A C C The Effect of Firing Temperature Preparation
Technique and Composition on the Electrical Properties of the Nickel Cobalt Oxide
Series Nixco1 - Xoy Journal of Materials Science 2004 39 251-264
13 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
14 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
15 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
16 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
17 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
18 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
19 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
11
21 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
22 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
23 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
24 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
25 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
26 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
27 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
28 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
29 Liotta L F Di Carlo G Pantaleo G Deganello G Catalytic Performance of
Co3o4Ceo2 and Co3o4Ceo2ndashZro2 Composite Oxides for Methane Combustion
Influence of Catalyst Pretreatment Temperature and Oxygen Concentration in the
Reaction Mixture Applied Catalysis B Environmental 2007 70 314-322
30 Hoflund G B Li Z Surface Characterization Study of a PdCo3o4 Methane
Oxidation Catalyst Applied Surface Science 2006 253 2830-2834
31 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
32 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
12
33 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
34 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
35 Hutlova A Niznansky D Plocek J Bursik J Rehspringer J-L Journal of
Sol-Gel Science and Technology 2003 26 473-477
36 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
37 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
38 Luumlders U Bartheacuteleacutemy A Bibes M Bouzehouane K Fusil S Jacquet E
Contour J P Bobo J F Fontcuberta J Fert A Nife2o4 A Versatile Spinel Material
Brings New Opportunities for Spintronics Advanced Materials 2006 18 1733-1736
39 Luumlders U Bibes M Bobo J F Fontcuberta J Tuning the Growth Orientation
of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A 2005 80 427-
431
40 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
41 Gokon N Murayama H Nagasaki A Kodama T Thermochemical Two-Step
Water Splitting Cycles by Monoclinic Zro2-Supported Nife2o4 and Fe3o4 Powders and
Ceramic Foam Devices Solar Energy 2009 83 527-537
42 Gokon N Mataga T Kondo N Kodama T Thermochemical Two-Step
Water Splitting by Internally Circulating Fluidized Bed of Nife2o4 Particles Successive
Reaction of Thermal-Reduction and Water-Decomposition Steps International Journal
of Hydrogen Energy 2011 36 4757-4767
13
43 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14
Chapter II
Methods The development of computational methods has progressed enormously within the
scientific and engineering communities over the last few decades The increase in
computational power makes solving previous unpractical problems and questions viable
Nowadays computation has become an indispensable research tool alongside with
experiment
In this chapter we shall briefly introduce the computational methods used to obtain the
results presented in this thesis Our basic tool is Density Functional Theory (DFT)
currently one of the most important tools in the field of computational chemistryphysics
and materials science Our discussion is limited to the scope of this thesis and one should
refer to many other text books available for a more comprehensive presentation
1 Basics of Quantum Chemistry
Just as Newtonrsquos laws of motions offered fundamental principles to describe the
movement of classical objects like earth and moon the motion of quantum particles like
single atoms or molecules is described by Schroumldinger equation expressed as the
following in non-relativistic form
minus119894119894ħ120597120597120597120597120597120597120627120627 = Ĥ120627120627
Here Ψ is the wave function a mathematical description of the quantum state of an
isolated system which contains all the information about this system and Ĥ is the
Hamiltonian operator which corresponds to the total energy of the system described by
Ψ
Without loss of generality Ψ can be written as a linear combination of wave functions
Ψn where Ψns are eigenvectors for Ĥ and satisfy
15
Ĥ120569120569119899119899 = 119864119864119899119899120569120569119899119899
Thus Ĥ contains all the system specific information For the systems of interest in
chemistry Ĥ is usually composed of the kinetic energies of nuclei and electrons and the
potential energies of the electrostatic interactions among nuclei between nuclei and
electrons and among electrons themselves
The first basic approximation used to reduce the complexity of chemical systems is the
BornndashOppenheimer approximation The success of this approximation is due to the huge
difference between the mass of the electron and that of the nuclei for example the
lightest nucleus the proton is 1835 times heavier than an electron while the O and Co
nuclei are over 20000 times heavier than electron As a result nuclei move at a much
smaller rate in comparison to electrons and their velocity is negligible Thus one can
separate the pure nuclear energy (nuclear kinetic energy and nuclei-nuclei electrostatic
interaction) in an adiabatic process where Ψ is decomposed into the product of an
electron Ψel and a nuclear Ψnu wavefunction As a consequence one only needs to
solve the electronic Schroumldinger equation as the first step The solution of Schroumldinger
equation can be used to construct the potential for the nuclei Schroumldinger equation In
most cases the nuclei can be treated classically and further reduce the computational cost
for the system
Solving the electronic Schroumldinger equation is the key to study chemical systems Except
for the simple case of a hydrogen atom one still needs to consider a many electron
system where an analytical solution is hard to achieve A common practice to solve the
problem is by mean field theory where an electron is treated as subjected to the mean
field created by all other particles Thus the current wave function Ψel determines the
electronic Hamiltonian Ĥel from which a new wave function Ψel can be calculated When
starting with a reasonable guess for Ψel one can iteratively update Ψel and Ĥel until
convergence is achieved This is known as the self-consistent field method Hartree-Fock
theory as one of the earliest realizations of this approach
16
2 Density Functional Theory
Even within Hartree-Fock theory calculating the electronic wave function can be still
difficult and computationally intensive For example a typical NCO model investigated
in this thesis involves around 600 electrons and around 1500 basis functions
Building on the ideas in the Thomas-Fermi model Walter Kohn and Pierre Hohenberg
rigorously demonstrated that the ground state properties of a many-electron system eg
interacting electron gas with external field like fixed nuclei from BornndashOppenheimer
approximation are uniquely determined by the 3-dimensional charge density as long as
the ground state is non-degenerate and there is no external magnetic field1 This is known
as the first Hohenberg-Kohn theorem They subsequently demonstrated the second
Hohenberg-Kohn theorem which states that the true ground state electron density
minimizes the systemrsquos total energy These theorems provide the foundation for Density
Functional Theory (DFT) where the number of degrees of freedom is reduced from 3N
(for N electrons) to 3 Soon afterward Hohenberg-Kohnrsquos results were generalized to
degenerate systems2 and systems in the presence of an external magnetic field3-4 Time
dependent DFT was also proposed in the attempt to describe excited states5
The two Hohenberg-Kohn theorems suggest that for a given number of electrons N
defined by 119873119873 = int119899119899(119903119903)119889119889119903119903 and charge density 119899119899(119903119903) = 119873119873int1198891198891199031199032 ∙∙∙ int 119889119889119903119903119899119899 120569120569lowast(119903119903 1199031199032∙∙∙
119903119903119899119899)120569120569(119903119903 1199031199032∙∙∙ 119903119903119899119899) there are universal functionals 119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)] that represent the
kinetic energy of the electrons and the potential energy of their interaction in the ground
state On the other hand the external field influence is described by a non-universal
functional 119881119881119890119890119890119890119890119890[119899119899(119903119903)] As a result the total energy can be expressed in the following
form
119864119864[119899119899(119903119903)] = 119879119879[119899119899(119903119903)] + 119880119880[119899119899(119903119903)] + 119881119881119890119890119890119890119890119890[119899119899(119903119903)] ∙ 119899119899(119903119903) ∙ 119889119889119903119903
This equation can be solved variationally provided one is able to find proper forms of
119879119879[119899119899(119903119903)] and 119880119880[119899119899(119903119903)]
An effective method to construct such functional was proposed by Kohn and Sham in
19656 They separated the electron-electron interaction 119880119880[119899119899(119903119903)] into two parts The first
17
part treats the classical Coulomb interaction of the electron density to account for
electron-electron repulsion The second part accounts for the exchange energy and
correlation effects that are ignored when electrons are treated as electron density Thus
119880119880[119899119899(119903119903)] can be expressed as
119880119880[119899119899(119903119903)] =12
119899119899(119903119903) ∙ 119899119899(119903119903prime)|119903119903 minus 119903119903prime|
119889119889119903119903119889119889119903119903prime + 119864119864119890119890119909119909[119899119899(119903119903)]
Here the first term is the Coulomb interaction and the second term is the exchange-
correlation energy
The kinetic energy 119879119879[119899119899(119903119903)] being a unique functional of 119899119899(119903119903) can be solved exactly for
non-interacting electrons with the same density as the system of interest Thus only the
exchange-correlation term is unknown The original solution suggested by Kohn and
Sham is to borrow this term from the homogeneous electron gas (HEG) with the same
local density of the system of interest (note that HEGrsquos exchange energy can be
expressed analytically while its correlation part can be calculated numerically) This leads
to the following expression for the exchange-correlation energy
119864119864119890119890119909119909119871119871119871119871119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903)]119889119889119903119903
This expression is known as the local density approximation (LDA) and is still widely
used However the LDA fails in systems where rapid changes of density occur such as
in many small molecules In our work we applied another widely used approximation
the generalized gradient approximation (GGA) which improves the LDA exchange-
correlation functional by accounting for the spatial variation of the density and usually
offers better result
119864119864119890119890119909119909119867119867119867119867119871119871[119899119899(119903119903)] = 119899119899(119903119903) ∙ 119907119907119890119890119909119909119867119867119867119867119867119867[119899119899(119903119903) |nabla119899119899(119903119903)|2]119889119889119903119903
More detailed discussion of the GGA can be found in the article by Perdew Burke and
Ernzerhof7 where the so-called PBE functional is introduced
18
21 Basis Sets and Pseudopotentials
The analytical form of electronic wave function Ψel is not known for most chemical
systems Hence it is often convenient to use some basis set of know functions that can be
linearly combined to approximate the real wave function For periodic systems as
considered in this thesis plane waves 119890119890119894119894119896119896 119903119903 are a good choice according to Blochrsquos
theorem as any square-integrable continuous function can be expanded into an infinite
series of plane waves In practice however truncation of the plane waves is always
necessary to make the calculations feasible Usually good accuracy can be achieved with
careful selection of the truncation threshold
Figure 5 scheme of pseudopotential (PS) its corresponding wave function and compared
with all electron potential and wave functions where they converge at cutoff radius
19
For some system the like hydrogen atom the wave function is smooth and can be
described quite well with relatively few basis functions However when systems become
more complicated eg transition metals like Fe and Ni some wave functions start to
change rapidly and sharply One needs to include a huge amount of basis functions to
better describe the wave function Luckily for most chemical systems the region where
the wave function is rapidly changing is always close to the core electrons which form a
closed shell structure In the chemical relevant region for valence electrons the wave
functions are usually relatively smooth Thus the core electrons together with the nuclei
can be regarded as a pseudo core that can be conveniently described via pseudo potentials
for the valence electrons thus saving computational time and achieving smoother pseudo
wave functions (Figure 5)
The first type of pseudopotentials that were proposed for use within DFT were called
norm-conserving pseudo potentials (NCPPs)8 Three main criteria exist for constructing a
NCPP First the energy eigenvalues of the valence orbitals should be the same given by
all electron calculations second the wave function should replicate the all electron wave
function beyond a cut off radius as this is the key region where bonds form between
different atoms The last criterion for NCPP is that the total charge is preserved for the
valence electron density A drawback of NCPPs is that they often require a large basis set
to represent the wavefunction especially for 2 p and 3 d series elements Thus this thesis
will adopt another type of pseudo potentials the so-called ultrasoft pseudopotentials
(USPPs)9 With these pseudopotentials the shape of pseudo wave functions in softened
in the core region while conservation of the total charge is dealt with using some
reshaping operator As a consequence USPPs though more complicated to generate with
good transferability usually reduce the computational cost substantially by allowing the
use of significantly smaller basis sets
22 Self-Interaction Correction
The self-interaction error (SIE) is a common type of issue occurring in DFT and makes
many predictions less meaningful quantitatively It originates from the Hartree term
20
electrostatic potential that comes from the charge distribution for multiple electrons
system The expression
119907119907119867119867[119903119903119899119899(119903119903)] = 119890119890119899119899(119903119903prime)
|119903119903 minus 119903119903prime|119889119889119903119903prime
represents the potential energy of an electron moving in the field generated by the
electronic charge density 119899119899(119903119903) However 119899119899(119903119903) also includes the electron itself thus
leading to unphysical repulsion between the electron and itself As a consequence this
repulsion would usually result in wrongly delocalized charge The effect of SIE is
significant in many strongly correlated materials like transition metal oxides that have d
electrons and results in smaller band gap than expected and also results in inaccurate
thermodynamic and kinetic properties
This thesis focuses on iron group spinel oxides where transition metals Fe Co and Ni are
studied in their 2+ and 3+ charge states All these cations contain 3 d electrons and thus
SIE will be a big issue Taking a step backward Hartree-Fock theory offers an exchange
term which exactly cancels the SIE influence Inspired by Hartree-Fock theory
researchers mixed the exact exchange with the DFT exchange and correlation in order to
handle the SIE error This is known as the hybrid functional DFT method where the
exact exchange usually accounts for 20-25 of the exchange energy as in the well-
known B3LYP and PBE0 functionals10-11 However due to the inclusion of exact
exchange hybrid functionals are quite computationally demanding for many systems For
the systems of interest in this thesis like NCO another method of correction called
DFT+U12 is more commonly used DFT+U is more than ten times computationally
cheaper than hybrid DFT while keeping sufficient accuracy and is thus more
convenient The idea of DFT+U comes from the Hubbard model and introduces on-site
Coulomb interaction for localized electrons especially d and f electrons This new
potential helps reduce SIE and is given by the following expression
119864119864119880119880[119899119899119897119897119897119897] =11988011988021205821205821198941198941198971198971198971198971 minus 120582120582119894119894119897119897119897119897
119894119894119897119897119897119897
21
Here λ is the occupation number for the orbital and spin lσ and varies between 0 and 1 l
is the index for orbital angular momentum σ represents the spin channel and i is the
atomic index 119899119899119897119897119897119897 is the charge density projected onto the specific lσ orbital As seen by
the formula when λ=10 meaning the orbital is either occupied or empty the Hubbard U
correction will have no contribution When λ is close to 05 the correction is maximized
As a consequence it will penalize those electrons especially d and f electrons which
become delocalized U can be determined ab initio by application of linear response
theory13-14 In practice however U is often determined empirically by fitting to
experiment typically to thermodynamic results15
23 Structural Optimization
In section 21 and 22 we introduced the general procedures of how to determine the
electronic structure within the BornndashOppenheimer approximation The problem of the
ionic motion such as the determination of the optimal atomic geometry still needs to be
addressed As mentioned earlier the ionic motion is usually regarded as a classical
problem That is the goal is to find the ionic coordinates 119877119877 that minimize the potential
energy defined by 119877119877 and the electron density 120569120569119890119890119897119897119877119877 2 This defines an optimization
problem Thus the minimized structure will be in some local minimum where nabla 119881119881119877119877 =
0 and nabla 2119881119881119877119877 gt 0 The first criteria equivalent to classic force equal to zero indicates
no tendency toward moving when in equilibrium and the second criteria indicates the
structure is in a stable state Optimization problems are widely seen in different fields in
science and engineering and are very well studied The simplest method to solve
optimization problem is by the steepest descent algorithm where the optimization goes
into the opposite direction of the gradient nabla 119881119881119877119877 However when the gradient is small
as it always is when close to the local minimum steepest descent is very slow and takes a
lot of time to converge This problem can be overcome by applying momentum or
Newtonrsquos method In Newtonrsquos method the second order derivative nabla 2119881119881119877119877 also known
as Hessian matrix is calculated to help determine the optimization step length and thus
greatly reduce the optimization steps However in many systems such as the NCO in this
22
thesis calculating the Hessian is too expensive and should be avoided Instead we
applied a quasi-Newton method called Broyden ndash Fletcher ndash Goldfarb ndash Shanno (BFGS)
algorithm16 where only the initial Hessian is calculated accurately and is successively
updated with previous step information
3 Nudged Elastic Band Method
Aside from structural optimization which explores thermodynamic properties it is often
desirable to determine the systemrsquos kinetic properties eg finding the reaction pathways
and activation energies These can be obtained from the total energies of transition states
which can be determined by DFT alongside the total energies of the reactants and
products A popular method for this purpose is the nudged elastic band (NEB) method an
improved algorithm derived from the elastic band method17 The elastic band method
starts with creating a number of intermediate structures also known as images linearly
interpolated and in most cases evenly distributed between the reactant and product Each
of these images is assumed to be connected to the closest images by springs thus
preventing them to relax into the same local minimum Thus as a result the springs add
additional forces onto the total force where the force of atom j in image i can be
expressed as shown below
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894 + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1
However forces added by springs may shift the structure away from the minimum energy
path In the actual minimum energy path the force minusnabla 119881119881119877119877 119894119894119894119894 should be on the pathway
and the force perpendicular to the pathway should be 0 Thus in the NEB the force is
decomposed into two parts first the true force perpendicular to the pathway to optimize
the structure onto the pathway and second the spring force projected onto the direction
or tangent of the reaction pathway (Figure 6)
119865119894119894119894119894 = minusnabla 119881119881119877119877 119894119894119894119894perp + 119896119896119894119894+1119877119877 119894119894+1 minus 119877119877 119894119894 minus 119896119896119894119894119877119877 119894119894 minus 119877119877 119894119894minus1∥
23
Figure 6 NEB method showing the force and optimization path Forces on the 5th image
are shown in the enlarged region where the black arrow indicates the true force
However the force projected onto the path (blue arrow) is not used whereas the force
coming from the spring (red arrow) is used on the path
Though NEB constrains all the images onto the reaction path it still doesnrsquot give the
structure of the saddle point To determine the latter the climbing image technique is
applied to a single image to move it to the saddle point18 The force part perpendicular to
the path remains the same as in the NEB method while the parallel force is the inverse
of the true force projected onto the path Thus the image is climbing uphill toward the
saddle point and stopped there
24
4 References
1 Hohenberg P Kohn W Inhomogeneous Electron Gas Physical Review 1964
136 B864-B871
2 Levy M Universal Variational Functionals of Electron Densities First-Order
Density Matrices and Natural Spin-Orbitals and Solution of the V-Representability
Problem Proceedings of the National Academy of Sciences 1979 76 6062-6065
3 Vignale G Rasolt M Density-Functional Theory in Strong Magnetic Fields
Physical Review Letters 1987 59 2360-2363
4 Grayce C J Harris R A Magnetic-Field Density-Functional Theory Physical
Review A 1994 50 3089-3095
5 Runge E Gross E K U Density-Functional Theory for Time-Dependent
Systems Physical Review Letters 1984 52 997-1000
6 Kohn W Sham L J Self-Consistent Equations Including Exchange and
Correlation Effects Physical Review 1965 140 A1133-A1138
7 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Physical Review Letters 1996 77 3865-3868
8 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Physical Review Letters 1979 43 1494-1497
9 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Physical Review B 1990 41 7892-7895
10 Stephens P J Devlin F J Chabalowski C F Frisch M J Ab Initio
Calculation of Vibrational Absorption and Circular Dichroism Spectra Using Density
Functional Force Fields The Journal of Physical Chemistry 1994 98 11623-11627
11 Perdew J P Ernzerhof M Burke K Rationale for Mixing Exact Exchange
with Density Functional Approximations The Journal of Chemical Physics 1996 105
9982-9985
12 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Physical Review B 1991 44 943-954
13 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Physical Review B 2005 71
25
14 Kulik H J Cococcioni M Scherlis D A Marzari N Density Functional
Theory in Transition-Metal Chemistry A Self-Consistent Hubbarduapproach Physical
Review Letters 2006 97
15 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Physical Review B 2006 73
16 Liu D C Nocedal J On the Limited Memory Bfgs Method for Large Scale
Optimization Mathematical Programming 1989 45 503-528
17 JOacuteNsson H Mills G Jacobsen K W Nudged Elastic Band Method for
Finding Minimum Energy Paths of Transitions 1998 385-404
18 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
26
Chapter III
Formation electronic structure and
defects of Ni substituted spinel cobalt
oxide
1 Introduction
Nickel cobaltite NiCo2O4 (NCO) is an emerging technological material with a variety of
promising applications ranging from supercapacitors1-4 to catalysts for CH4 and low
temperature CO oxidation5-7 For instance a recent study has shown that NCO can
completely oxidize methane to CO2 and water in the temperature range of 350ndash550 C7
Considering that NCO is also a cheap material made of earth-abundant elements this result
suggests that NCO may be a better methane oxidation catalyst than typical precious-metal-
based catalysts8
NCO is generally considered to have an inverse spinel structure with mixed valence
typically expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ [1198621198621198621198623+1198731198731198941198941minus1198901198902+ 1198731198731198941198941198901198903+]1198741198744 where tetrahedral (Td) sites are
occupied by both Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+ Ni3+
and Co3+ ions9-11 However evidence for Ni(Td) ions is also reported indicating that
Ni(Oh)harr Co(Td) exchanges can take place rather easily9 Co(Oh) ions are non-magnetic
whereas Co(Td) and Ni(Oh) tend to have anti-parallel spins (see Figure 7c) so that the
material is usually found to be ferrimagnetic10 12 However synthesis conditions can
influence the magnetic order which has led to some contradictory results12-14 Only a few
studies on NCO single crystals have been conducted9 15-16 these confirmed that NCO is
ferrimagnetic and metallic and the electronic and magnetic behaviors are strongly
correlated with the concentration of Ni3+(Oh) ions17 NCO was also reported to become
unstable around 600 degC in vacuum and to partially decompose into NiO and NixCo3-xO49
The loss of oxygen suggests that oxygen vacancies would form in this process Oxygen
27
vacancies are common and important defects in metal oxides18 but their effects on NCO
are not clearly understood Theoretical studies on NCO are also scarce various bulk13 19
and surface7 20 properties have been investigated but many aspects of the behavior of this
material eg the origin of the half-metallic ferrimagnetic structure and the influence of
oxygen vacancies are still largely unexplored
With the growing interest in using NCO for various applications a more detailed and
complete understanding of the structure and electronic properties of this material would be
desirable This has motivated us to carry out first principles calculations on defect-free and
defected bulk of NCO using Density Functional Theory (DFT) with on-site Coulomb
repulsion (DFT+U) As a first step we have studied the evolution from Co3O4 to NCO by
determining the substitution energies and the atomic and electronic structures of NixCo3-
xO4 as a function of the doping ratio x from x=0 to x=1 Our results show that Ni acts as a
p-type dopant in Co3O4 and is fractionally occupied This leads to a half-metallic
ferrimagnetic structure for NCO consistent with previous calculations13 19 and recent
experiments14 17 Typical bulk defects notably Ni(Oh)harr Co(Td) exchanges and oxygen
vacancies have been studied using different U values Facile NiharrCo(Td) exchange as
observed experimentally can be reproduced using U values smaller than those determined
from first principles linear response theory Oxygen vacancies occur preferentially at sites
coordinated to a larger number of Ni ions with a computed formation energy similar to
that found for pure spinel cobalt oxide
2 Computational details
Spin-polarized DFT calculations were performed using the plane-wave-pseudopotential
scheme as implemented in the Quantum Espresso package21 Exchange and correlation
terms were described using the Perdew-Burke-Ernzerhof (PBE)22 functional with on-site
Coulomb repulsion U term on Co and Ni 3 d electrons Unless otherwise specified we used
the U values determined from first principles linear response theory23 denoted as ULR
notably ULR(Ni) = 66 eV for nickel and ULR(Co) = 44 eV and 67 eV for Co ions at Td
site and Oh sites respectively Ultrasoft pseudopotentials24 were generally employed
28
where valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Selected calculations using norm-conserving pseudopotentials25 were also
performed in this case a kinetic energy cutoff of 100 Ryd was used Structural
optimizations were carried out by relaxing all atoms until forces were smaller than 1 times 10-
3 au Calculations were performed using the 56-atom conventional cubic cell containing 8
formula units (Figure 7) with a 3 times 3 times 3 Monkhorst-Pack k-point mesh to sample the
Brillouin zone All results eg total energies or magnetic moment that are reported in the
following will be referred to this conventional unit cell Wannier functions were calculated
using the wannier 90 code26
Figure 7 Conventional cell of (a) α type and (b) β type inverse spinel The β-type inverse
spinel structure was used in most calculations (c) Spin occupations of the 3d states of Ni
and Co ions in the β-type inverse spinel structure as obtained from our calculations (sect
32) red arrows for Ni and Co(Td) indicate fractional occupation of the corresponding
orbitals
3 Results and Discussion 31 Ni substitution in Co3O4 NixCo3-xO4 (x = 0 -1)
We studied the formation of NCO starting from the normal spinel Co3O4 and successively
substituting 8 Co at the octahedral sites of a 56 atom cubic supercell with Ni atoms thus
gradually increasing the doping ratio from x=0 to x=1 by steps Δx=18 By this procedure
the resulting NCOrsquos structure is an inverse spinel with tetrahedral sites occupied by Co
only For each x the lattice constant was determined from the Birch-Murnagham equation
29
of state and found to increase linearly with increasing doping ratio x by ~001 Aring per 18
change in x (see Table 1) For Co3O4 we determined a lattice constant of 8149 Aring which
is 09 larger than the experimental value 8085 Aring27 while the computed lattice constant
for NCO is 8237 Aring which is about 15 larger than the experimental value of 8115 Aring27
Doping energies were determined from the expression
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 = 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909120583120583119873119873119894119894 + 119909119909120583120583119862119862119889119889119909119909
Table 1 Lattice constant bulk modulus and doping energy of NixCo3-xO4 for different
values of x calculated using DFT+ULR The doping energies are determined for
different orderings of the spins on Co(Td) ions The preferred ordering is anti-parallel in
Co3O4 and parallel in NCO
Doping ratio x Lattice
constant a0 Aring
Bulk modulus B
GPa
Doping energy eV
Spin parallel Spin anti-
parallel
0 8149 203 - -
18 8162 190 0028 0031
14 8173 179 -0009 -0019
38 8184 175 -0056 -0067
12 8196 171 -0062 -0073
58 8199 136 -0086 -0093
34 8216 167 -0095 -0103
78 8227 170 -0113 -0113
1 8237 163 -0125 -0119
Whereas for the above formula 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 and 11986411986411986211986211988911988931198741198744 are the total energies of NixCo3-
xO4 and Co3O4 respectively and μCo and μNi are the chemical potentials of Co and Ni at
Oh sites The latter were evaluated as 120583120583119872119872 = 119864119864119872119872119874119874 minus121205831205831198741198742 where 119864119864119872119872119874119874 is the computed
30
total energy per formula unit of the MO binary oxide (M= Co or Ni) and 1205831205831198741198742 is the
chemical potential of an oxygen molecule As a result 119864119864119889119889119889119889119889119889119894119894119899119899119889119889 can be rewritten as
119864119864119889119889119889119889119889119889119894119894119899119899119889119889 asymp 1198641198641198731198731198941198941199091199091198621198621198891198893minus1199091199091198741198744 minus 11986411986411986211986211988911988931198741198744 minus 119909119909119864119864119873119873119894119894119874119874 + 119909119909119864119864119862119862119889119889119874119874119909119909
Figure 8 Electronic density of states (DOS) of NixCo3-xO4 for different values of the doping
ratio x calculated using DFT+ULR The DOS is not affect by the spin ordering of Co(Td)
ions
The computed doping energies for different x values are reported in Table 1 We can see
that the doping energy is positive at x=18 indicating that doping Co3O4 with Ni is not
favorable at low concentration For xgt 14 however the doping energy becomes
increasingly negative with increasing doping ratio indicating that further doping is
favorable For xlt 34 the spins of Co ions at neighboring Td sites prefer to be anti-parallel
Both anti-parallel and parallel spins are present at x=78 while the spin ordering of Co(Td)
31
ions changes to parallel at x=1 (the spins of Ni(Oh) and Co(Td) are always antiparallel)
Together with the decrease of doping energies Table 1 shows a small increase of the lattice
constant with increasing x The bulk modulus decreases significantly with increasing x
reaches a minimum at x=58 and slightly increases afterwards
Figure 8 shows the evolution of the Density of States (DOS) of NixCo3-xO4 as a function
of x We can see that Ni doping gradually transforms Co3O4 an insulating material into
a half metal in which the majority spin channel remains insulating while the minority one
becomes conducting The valence bands for both spin channels thus shift toward the
Fermi level in comparison to Co3O4 These results also indicate that Ni (which prefers a
charge state Ni2+) substituting a Co3+ at an Oh site acts as p-type dopant in Co3O4
consistent with recent experimental findings14 19 The presence of Ni2+ at an octahedral
site causes indeed the formation of a Co(Td)3+ which has only one occupied e orbital in
the minority spin channel (see Figure 7) The other empty e orbital on Co(Td) gives
then rise to a hole (acceptor) state at the top of the valence band (note that the e states of
Co(Td) are part of the valence band in Co3O4) Similarly a Ni3+ at an octahedral site
(resulting in Co(Td) 2+) has an empty eg state which also leads to a hole state at the top of
or just above the valence band Hence in both cases empty states above or at the top of
the minority spin valence band are formed indicating that Ni acts as a p-dopant in Co3O4
32 Structure bonding and electronic properties of NCO
We considered two different symmetries known as α-type and β-type of the inverse spinel
crystal structure (Figure 7a-b) The β-type structure has alternating layers of -Co(Oh)-O-
and -Ni-O- along the z direction in Figure 7b whereas layers containing both Co(Oh) and
Ni ions are present in the α-type structure Total energy calculations are performed for
many possible magnetic configurations with these symmetries The β-type inverse spinel
is found to be slightly more stable by 005 eV per conventional cell in comparison to the
α-type structure For both symmetries the most stable state is half-metallic and
ferrimagnetic with anti-parallel spins on Ni(Oh) and Co(Td) ions and a total magnetization
of 16 microB per conventional unit cell (note that the Co(Oh) ions have always fully occupied
32
t2 states and no spin polarization) However the ferromagnetic state (total magnetization
of 32 microB) is only 007 eV higher in energy and also half metallic By contrast the
ferrimagnetic configuration with insulating behavior in both spin channels has much higher
energy (117 eV) The Densities of States of NCOrsquos ground state and excited ferromagnetic
and insulating ferromagnetic states are compared in Figure 9
Figure 9 DOS calculated for the insulating ferrimagnetic (I) (top panel) metallic
ferromagnetic (M) (middle panel) and metallic ferrimagnetic (M) (bottom panel)
configurations of NCO
Maximally Localized Wannier functions (MLWFs)28-29 are a useful tool for studying
bonding properties their centers provide information on the atomic oxidation states while
their shapes help understand the character of the bonds Calculation of the MLWFs for
NCOrsquos half-metallic ferrimagnetic ground state shows that there are 8 7 and 6 d-type
Wannier functions centered on the Ni Co (Td) and Co (Oh) ions respectively suggesting
33
oxidation state +3 for Co(Oh) and +2 for Ni and Co (Td) ions However this would
introduce a charge imbalance in NCO More detailed analysis of the band structure shows
indeed that the eeg states of Ni and Co(Td) ions in the minority spin channel cross the
Fermi energy and are not fully occupied (Figure 10a) The partial occupation of these states
gives rise to both the fractional valence and the metallic behavior of the minority spin states
consistent with recent XPS data showing that both +2 and +3 oxidation states are present
for Ni and Co in NCO12 17 As shown by the projected band structure (Figure 11) however
also the oxygen orbitals contribute to the states around the Fermi level suggesting the
presence of covalent interactions with the Ni and Corsquos ege states Due to the partial
occupation of the oxygen orbitals the fractions of Co2+ and Ni3+ states can be different and
therefore NCOrsquos composition can be better expressed as 1198621198621198621198621198901198902+1198621198621198621198621minus1198901198903+ 1198621198621198621198623+1198731198731198941198941minus1199101199102+ 1198731198731198941198941199101199103+1198741198744
For comparison we also calculated the MLWFs of the ferrimagnetic insulating state and
found that in this case one e orbital of Co(Td) is only singly occupied (see Figure 7 for
comparison) consistent with a perfect Co3+[Ni2+Co3+]O4 structure as reported in Ref9
Figure 10 (a) Band structure for the minority spin states and (b) Projected density of
state (PDOS) of NCO calculated using DFT+ULR The different colors in (a) reflect the
contribution of the e states of Co (Td)
34
Figure 11 Projected DOS calculated for the eg states of (a) Co (Td) and (b) Ni The
fractional valence calculated from this PDOS is ~21+ for both Co(Td) and Ni (c) Band
structure around the Fermi level different colors indicate the contribution of the oxygen
orbitals
The projected Density of States (PDOS) for NCO in its ferrimagnetic ground state is shown
in Figure 10b The main contribution to the states around the Fermi energy in the minority
spin channel comes from Ni and Co(Td) as well as O ions consistent with the presence of
a significant hybridization between cation 3d and oxygen 2p orbitals The computed atomic
magnetic moments are 239 and -112 μB for Co(Td) and Ni respectively in good
agreement with the corresponding experimental values of 218 and -149 μB 10 On the other
hand the contributions of Co(Td) and Ni to the conduction bands are very different Co(Td)
ions do not contribute to the majority spin conduction states whereas Ni ions do not
contribute to the minority spin states The detailed band structure plotted along several
symmetry directions is shown in Figure 12a and 12b There is an indirect band gap of about
20 eV along the Γ X direction for the majority spin states However the valence bands are
35
quite flat so that the direct gap found near X is very similar 21 eV The minority spin
bands show a stronger dispersion especially across the Fermi level
Figure 12 Band structure of NCO calculated using DFT+ULR (a) majority and (b)
minority spin states
To obtain insight into the absorption spectrum the Joint Density of States (JDOS) was also
calculated (Figure 13) The norm-conserving pseudopotentials used for these calculations
yield a band structure similar to the ultrasoft pseudopotential results of Figure 12a and 12b
with a somewhat larger bandgap of 24 eV for the majority spins The shapes of the JDOS
curves are similar for the majority and minority spin states except for the longer low energy
tail of the latter This is consistent with recent experiments indicating that there is optical
adsorption for NCO at relatively large wavelength (gt700nm) together with an optical band
gap around 26 eV17
36
Figure 13 Electronic structure of NCO calculated using norm conserving
pseudopotentials with U terms from linear response band structure for (a) majority and
(b) minority spins (c) Projected DOS and (d) Joint DOS The JDOS suggests an optical
band gap of ~ 25 eV for the majority spins and absorption at long wavelengths for the
minority spins
33 Ni harr Co exchanges
Motivated by the experimental evidence of Ni(Td) ions19 we have investigated the
stability of NCO with respect to Ni(Oh) harr Co(Td) exchanges by calculating the
formation energy (119864119864119904119904119904119904119904119904)
119864119864119904119904119904119904119904119904 = (119864119864119890119890119890119890119909119909ℎ minus 1198641198640) 119899119899frasl
where 119864119864119890119890119890119890119909119909ℎ is the total energy of the crystal after exchanging Ni(Oh) and Co(Td) 1198641198640 is
the total energy of the defect free bulk in the inverse spinel structure and n is the number
of Ni sites exchanged Results with our standard DFT+ ULR setup predict however a
37
rather large (~ 13 eV) energy cost for the Ni harr Co(Td) exchange which seems at
variance with experiment10 19 Since it is known that the choice of U in DFT+U
calculations is not unique and different U values may perform better in different
situations30-31 we then decided to examine how the formation energy 119864119864119904119904119904119904119904119904 is affected by
the choice of U Results for different nrsquos and different choices of the Hubbard U terms
are reported in Table 2 Specifically we compare 119864119864119904119904119904119904119904119904 computed using our standard
linear response ULR values with results obtained using (i) pure PBE (U = 0) and (ii) U
values (Ueff) suggested by previous studies ie Ueff = 55 and 30 eV for Ni32 and both
types of Co ions30 33 respectively As a reference the DOS of NCO calculated with these
different choices of U are shown in Figure 14 We notice that all U values predict the
material to be ferrimagnetic with total magnetization of 16 μB per conventional unit cell
and the minority spin states to be conducting The shapes of the DOS obtained with
different U values are also similar but pure PBE shows no gap for the minority spin
states whereas a gap occurs in both DFT+ULR and DFT+Ueff calculations
Figure 14 DOS of pristine NCO calculated using different U values as indicated
38
As shown in Table 2 our computed substitution energies do not change monotonically
with increasing U value With PBE and intermediate U values (Ueff) the substitution
energies are very small which seems to provide the best agreement with the experimental
observation that intermediate structures between inverse spinel and normal spinel occur
frequently10 19 By contrast the Ni harr Co(Td) exchange is energetically very costly with
large U values (ULR) indicating that ULR does not predict well the thermodynamic
properties To make closer connection with experiment we also determined the formation
temperature 119879119879119904119904119904119904119904119904 = 120549120549119864119864119904119904119904119904119904119904∆119878119878 where 120549120549119864119864119904119904119904119904119904119904 is the formation energy difference due to the
exchange and ∆119878119878 is the corresponding change of configurational entropy (see Table 2)
This was determined from the expression ∆119878119878 = 11989611989611990411990411989711989711989911989911988211988211198821198820
where W0 and W1 are the number
of configurations corresponding to the initial and final values of the exchange ratio PBE
and PBE+Ueff predict moderate exchange formation temperatures whereas PBE+ULR
strongly overestimates the temperature Although PBE appears to well reproduce the
experimental observations9 one should notice that the results in Table 2 do not include the
changes of vibrational entropy which may give comparable contribution compared with
configurational entropy34 This and other sources of inaccuracy could lead to an error of
about 01 eV for 119864119864119904119904119904119904119904119904 and thus a ~ 300 K difference in 119879119879119904119904119904119904119904119904 Increase of the substitution
ratio R from 18 to 14 Ni gives slightly lower 119864119864119904119904119904119904119904119904 but higher 119879119879119904119904119904119904119904119904 due to a smaller ΔS
After the Ni harr Co(Td) exchange the Ni ions at the Td sites tend to be spin parallel to the
Co at the other Td sites The total magnetization does not change and the DOS changes are
also minor using both Ueff and ULR (Figure 15) Finally complete Ni harr Co(Td) exchange
(R=1 in Table 2) leads to NCO in the normal spinel structure for which the values of
119864119864119904119904119904119904119904119904 are similar to those at smaller exchange ratios
39
Table 2 Computed Ni(Oh)harrCo(Td) exchange energies 119864119864119904119904119904119904119904119904 and formation
temperatures Tsub (between parentheses) for different exchange ratios R and different U
values For example R = 18 corresponds to one Co(Td)harr Ni exchange per
conventional (56 atoms) cell
Esub eV (Tsub K)
Method R = 18 R= 14 R = 1
PBE 0157 (438) 0151 (672) 0113
DFT+Ueff 0044 (123) 0042 (185) 0062
DFT+ULR 1305 (3541) 1288 (5887) 1314
Figure 15 DOS calculated using Ueff (left) and ULR (right) with (ad) no substitution
(be) 18 substitution and (cf) 14 substitution
40
34 Oxygen vacancies
NCO is unstable and forms NiO and NixCo3-xO4 above 650 C9 The decomposition is
associated to a loss of oxygen and thus to creation of oxygen vacancies The formation
energy of oxygen vacancies is thus an important quantity for evaluating the thermal
stability of NCO
We considered 1 vacancy per conventional (56 atoms) unit cell and determined the
formation energy from the expression
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 121205831205831198741198742 minus 1198641198640 minus 119896119896119904119904119879119879119897119897119899119899119879119879
Table 3 Oxygen vacancy formation energies (Eform) for NCO calculated using DFT+U
with various choices of U (see text) Two situations were considered (a) T = 0 K and
standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen pressure (shown in
bracket) For comparison the O-vacancy formation energies in Co3O4 are also shown
For each U the corresponding optimized lattice constant was used For Co3O4 we were
not able to obtain a well converged ground state for the defected crystal using Ueff and
therefore the corresponding value of Eform is not reported
Material U (eV) Lattice
constant Aring
Eform (eV)
site A
Eform(eV)
site B
NCO 0 8141 317 (076) 302 (061)
Ueff 8209 235 (-006) 226 (-017)
ULR 8237 180 (-061) 163 (-077)
Co3O4 0 8088 334 (093)
ULR 8149 159 (-082)
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies of the defected and pristine crystal
respectively 1205831205831198741198742 is the chemical potential of O2 and 119896119896119904119904119897119897119899119899119879119879 is the configurational entropy
Two inequivalent sites for oxygen vacancies exist in NCO site A which has 3 Co and 1
41
Ni neighbors and site B with 2 Co and 2 Ni neighbors to test the possible dependence of
their relative stabilities on the U term calculations for different choices of U as described
in Sec IIIC were performed Results are reported in Table 3 for two different conditions
namely (a) T= 0 K and standard oxygen pressure and (b) T = 1000 K and 10-9 atm oxygen
pressure (decomposition conditions) This shows that the O-vacancy formation energy at
site B is somewhat smaller than that at site A (independent of U) both being however
similar to the O-vacancy formation energy in Co3O4 Moreover both PBE+ULR and
PBE+Ueff predict negative O-vacancy formation energy for case (b) indicating that oxygen
vacancies form spontaneously under such conditions This is consistent with experiments
which indicate thermal decomposition of NCO under similar conditions (900 ndash 1000 K in
vacuum)19
Oxygen vacancies are known to have an important influence on the electrical properties of
metal oxides often causing the material to become a n-type semiconductor18 It is thus
interesting to study how this defect affects the structural and electronic properties of NCO
The atomic structure of an O-vacancy at site B is displayed in Figure 16b Analysis of the
atomic relaxations around the vacancy shows a large displacement of 026 Aring away from
the vacancy site for the Co(Td) bonding with the oxygen that was removed (Co1 in Figure
16b) On the other hand the displacements of the neighboring Co(Oh) Ni and O atoms are
moderate around 005 Aring with O1O2 and O3 moving toward the vacancy site As for the
electronic structure the formation of an oxygen vacancy gives rise to several low lying
states with different total magnetizations making the identification of the ground state
quite delicate The electronic Densities of States for defect free and defected NCO with A
and B-type oxygen vacancies are compared in Figure 16a For the majority spin states both
vacancy types result in the formation of additional states in the band gap For the minority
spin states the states at the Fermi level split in the case of an A-type vacancy leading to
the opening of a small band gap of ~ 03 eV here the low energy empty states originate
from Ni ions which do not contribute to the conduction bands of defect free bulk (Figure
17) In the case of a B-type vacancy the minority spin states remain conducting and the
overall changes in the electronic structure are minor Since B-type vacancies are favored
over A-type ones this suggests that at variance with other oxides oxygen vacancies have
only a minor influence on the conductivity of NCO The occupation of the d orbitals on the
42
Co(Td) Co(Oh) and Ni ions close to the vacancy are illustrated in Figure 16c Interestingly
although Co(Td) has a large displacement its occupation remains unchanged A change
occurs only for the Co(Oh) ion which becomes 2+ and assumes an anti-parallel spin with
respect to Ni (we note that the same results are obtained with DFT+ULR and DFT+Ueff)
Moreover the total magnetization remains unchanged in the presence of the vacancy
possibly because the hybridization between metal and oxygen states leads to compensation
of the spin polarization
Figure 16 (a) DOS of defect free (top) and reduced NCO containing an oxygen vacancy
either at site A (middle) or at site B (bottom) (b) Atomic structure of a B-type oxygen
vacancy the dark gray ball is the vacancy site (c) Occupation of Ni and Co d orbitals in
the presence of an oxygen vacancy the red arrow indicates a newly occupied orbital in
comparison to those in defect free NCO
43
Figure 17 PDOS of defect free bulk (a) oxygen vacancy on Site A (b) and oxygen vacancy
on Site B (c) where the PDOS changes a lot by creating oxygen vacancy on Site A
4 Conclusions
In this work we have used DFT+U calculations to investigate the electronic and
thermodynamic properties of spinel NiCo2O4 (NCO) and obtain insight into the origin of
the high conductivity and ferrimagnetic properties of this material We have studied the
formation of NCO from Co3O4 focusing on the electronic structure of NixCo3-xO4 as a
function of the doping ratio x Our results show that Ni acts a p-type dopant in Co3O4 and
gradually makes the minority spin channel metallic Of the two possible symmetries α-
type and β-type of inverse spinels NCO favors the latter but the energy difference
between the two structures is quite small Investigation of different possible magnetic
structures indicates that NCO has a ferrimagnetic ground state with a low lying
ferromagnetic excited state which may explain why different conditions of synthesis often
44
cause major differences in the observed magnetic properties7 Further analysis of the
ferrimagnetic ground state indicates that the ege states of Ni and Co(Td) are partially
occupied resulting in fractional valence and metallic behavior consistent with recent
experimental results17
We also studied the influence of two frequently observed defects Ni harr Co(Td) exchanges
and oxygen vacancies on the structural and electronic properties The computed energy
cost of Ni harr Co(Td) exchanges is largely independent of the value of the exchange ratio
consistent with the fact that significant cation disorder is usually observed in spinel oxides19
Oxygen vacancies are predicted to occur more frequently at sites coordinated to a larger
number of Ni ions and to form spontaneously under conditions of high temperature and
low oxygen pressure where thermal decomposition of NCO is actually observed to take
place On the other hand oxygen vacancies are found to have only a minor effect on the
magnetic and electronic properties and therefore do not modify the p-type character of the
conductivity
45
5 References
1 Zhang G Lou X W Controlled Growth of Nico2o4 Nanorods and Ultrathin
Nanosheets on Carbon Nanofibers for High-Performance Supercapacitors Scientific
Reports 2013 3 1470
2 Liu S Hu L Xu X Al-Ghamdi A A Fang X Nickel Cobaltite
Nanostructures for Photoelectric and Catalytic Applications Small 2015 11 4267-4283
3 Yu L Zhang G Yuan C Lou X W Hierarchical Nico2o4Mno2corendash
Shell Heterostructured Nanowire Arrays on Ni Foam as High-Performance
Supercapacitor Electrodes Chem Commun 2013 49 137-139
4 Liu X Shi S Xiong Q Li L Zhang Y Tang H Gu C Wang X Tu J
Hierarchical Nico2o4Nico2o4coreShell Nanoflake Arrays as High-Performance
Supercapacitor Materials ACS Appl Mater Interfaces 2013 5 8790-8795
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 53 7223-7227
6 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
7 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
46
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
13 Dileep K Loukya B Silwal P Gupta A Datta R Probing Optical Band
Gaps at Nanoscale from Tetrahedral Cation Vacancy Defects and Variation of Cation
Ordering in Nico2o4epitaxial Thin Films J Phys D Appl Phys 2014 47 405001
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Silwal P Miao L Hu J Spinu L Ho Kim D Talbayev D Thickness
Dependent Structural Magnetic and Electronic Properties of the Epitaxial Films of
Transparent Conducting Oxide Nico2o4 J Appl Phys 2013 114 103704
17 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
18 Deml A M Holder A M OrsquoHayre R P Musgrave C B Stevanović V
Intrinsic Material Properties Dictating Oxygen Vacancy Formation Energetics in Metal
Oxides The Journal of Physical Chemistry Letters 2015 6 1948-1953
19 Ndione P F Shi Y Stevanovic V Lany S Zakutayev A Parilla P A
Perkins J D Berry J J Ginley D S Toney M F Control of the Electrical
Properties in Spinel Oxides by Manipulating the Cation Disorder Adv Funct Mater
2014 24 610-618
20 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
47
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
21 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
22 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
23 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in Thelda+Umethod Phys Rev B 2005 71 035105-1
-035105-16
24 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
25 Hamann D R Schluumlter M Chiang C Norm-Conserving Pseudopotentials
Phys Rev Lett 1979 43 1494-1497
26 Mostofi A A Yates J R Lee Y-S Souza I Vanderbilt D Marzari N
Wannier90 A Tool for Obtaining Maximally-Localised Wannier Functions Comput
Phys Commun 2008 178 685-699
27 Nkeng P Characterization of Spinel-Type Cobalt and Nickel Oxide Thin Films
by X-Ray near Grazing Diffraction Transmission and Reflectance Spectroscopies and
Cyclic Voltammetry J Electrochem Soc 1995 142 1777
28 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
29 Marzari N Vanderbilt D Maximally Localized Generalized Wannier Functions
for Composite Energy Bands Phys Rev B 1997 56 12847-12865
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
31 Garciacutea-Mota M Bajdich M Viswanathan V Vojvodic A Bell A T
Noslashrskov J K Importance of Correlation in Determining Electrocatalytic Oxygen
Evolution Activity on Cobalt Oxides J Phys Chem C 2012 116 21077-21082
32 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
48
33 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
34 Fultz B Vibrational Thermodynamics of Materials Prog Mater Sci 2010 55
247-352
49
Chapter IV
Oxygen deficiency and reactivity of spinel
NiCo2O4 (001) surfaces
1 Introduction
Spinel cobalt oxide (Co3O4) has recently attracted attention as a highly active catalyst for
various oxidation reactions1-3 Interest in this material has also generated efforts aimed at
tuning its catalytic activity through doping or substitution with selected transition metals
Among such substituted cobaltites NiCo2O4 (NCO) has emerged as a particularly
promising catalyst for low temperature methane and CO oxidation4-5 as well as the
oxygen evolution reaction6-7 For instance recent experiments have shown that NCO can
completely oxidize methane at 350-550 degC suggesting that in some cases NCOrsquos activity
could be higher than that of precious-metal-based catalysts8
NCO is a material with complex structural and electronic properties It is generally
considered to have an inverse spinel structure with mixed valence where tetrahedral (Td)
sites are occupied by Co2+ and Co3+ ions and octahedral (Oh) sites are occupied by Ni2+
Ni3+ and Co3+ ions9-13 However Ni(Oh)harr Co(Td) exchanges can take place rather
easily9 resulting in considerable cation disorder NCO is also generally described as
ferrimagnetic and metallic with much higher conductivity compared to other
cobaltites12 14-15 It was indeed suggested that NCO is a more efficient water oxidation
catalyst compared to pure Co3O4 due to its higher conductivity16
Despite the growing interest in the use of NCO in catalysis understanding of its
fundamental surface properties is still limited So far only few experimental and
theoretical studies on well-defined NCO surfaces have been reported5 17-19 The aim of
this work is to obtain insight into NCOrsquos surface structure and reactivity through Density
Functional Theory (DFT) calculations on the (001) surface which is one of the most
common surfaces of spinel materials20 Using DFT with the addition of on-site Coulomb
50
repulsion U terms on Co and Ni 3d shells (DFT+U)21 we investigate surfaces with
various CoNi ratios focusing on the formation of surface oxygen vacancies (VOs) which
have been proposed to play a key role in the oxidation of CO and methane on NCO4 We
also investigate the adsorption of two typical probe molecules water and O2 which are
important for characterizing the surface structure under ambient conditions and the
surface reoxidation process during catalytic reactions respectively Our results clearly
show that Ni has a major influence on the formation of surface oxygen vacancies leading
to VO formation energies significantly lower than those found for Co3O4 On the other
hand O2 adsorption is more difficult and is likely to represent the thermodynamic
limiting step of oxidation reactions on NCO(001)
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package22 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)23
functional with on-site Coulomb repulsion U term on Co and Ni 3 d electrons We used
the values U(Co)=30 eV and U(Ni)=55 eV which we recently found to provide a better
description of NCOrsquos thermodynamic properties compared to the U values determined
from Linear Response theory24-25 Ultrasoft pseudopotentials26 were employed and the
valence electrons include O 2 s 2 p Co 3 d 4 s and Ni 3 d 4 s states Kinetic energy
cutoffs used were 50 Ryd and 500 Ryd for wave function and augmented density
respectively Structural optimizations were carried out by relaxing all atoms until forces
were smaller than 10 times 10-3 au
Of the two possible symmetries α-type and β-type of inverse spinels the latter is slightly
more stable20 and was thus chosen for studying the surface properties In this structure
the (100) and (001) surfaces are inequivalent the former exposing both Ni and Co
cations and the latter either Co or Ni only see Figure 18 These surfaces were modeled
using slabs of 11 layers terminated by oxygen and octahedral Co and Ni ions and a
vacuum region of 20 Aring was used to separate adjacent slabs We kept the same chemical
51
composition for the (100) and (001) slabs resulting in two different terminations for the
latter nonetheless the net polarization is zero in all cases because the system is metallic
(see below) We used theoretical lattice parameters20 and considered a square surface
unit cell of dimensions 8209 x 8209 Aring2 exposing 4 octahedral cations and 8 oxygens in
the outer layer and 2 Co(Td) in the second layer just below We sampled the surface
Brillouin zone using 2 times 2 times 1 k-point grid Oxygen vacancies and adsorbed molecule
were introduced only on one of the surfaces of the slab Dipole corrections were found to
have only minor effects (about 001 eV) and were not included in our standard setup
3 Results and Discussion 31 Pristine (001) (100) surfaces
Experimental control of the NCO stoichiometry is very challenging and in practice NCO
samples are always non-stoichiometric14 17 Nonetheless the perfectly stoichiometric
NCO surface remains a convenient reference system for the study of surface oxygen
vacancies which is the reason why we consider it here The structures of the defect-free
NCO (001) and (100) slabs are shown in Figure 18 We found an energy difference of
less than 001 eV between the two slabs which is consistent with the fact that they have
the same overall stoichiometries (Ni12Co22O48) All investigated surfaces are terminated
by oxygen anions and metal cations that are at Oh sites in the bulk both Ni and Co(Oh)
are present on the (100) surface (denoted noted as (100)mix Figure 18be) while either Ni
or Co(Oh) are present on the (001) surface (indicated as (001)Ni and (001)Co Figure
18acd) In addition two different types of surface oxygen anions exist O1 bonded to
two 1st layer metal cations and a 3rd layer cation at an Oh site and O2 bonded to two 1st
layer metals and a 2nd layer Co(Td) On the (100)mix surface O1 atoms can be further
distinguished in O1a and O1b bonded to 3rd layer Co and Ni cations respectively We
further calculate the surface energy difference of those surfaces by using the formula of
calculating surface energy from
120574120574 =(119864119864119904119904119897119897119904119904119904119904 minus 119899119899119864119864119904119904119904119904119897119897119896119896 minus 119909119909120583120583119862119862119889119889 minus 119910119910120583120583119873119873119894119894 minus 1199111199111205831205831198741198742)
119860119860
52
where E is the total energy calculated for slab and bulk respectively and μ is the chemical
potential of corresponding compounds We get
∆120574120574 = (∆119864119864119904119904119897119897119904119904119904119904 + 119899119899(120583120583119873119873119894119894 minus 120583120583119862119862119889119889))119860119860
where the chemical potential of Ni and Co are refer to our previous paper20 ΔEslab were
calculated from extrapolating energies of different layered surface to have a more
accurate energy difference We found (001)Ni surface is more stable and (100)mix and
(001)Co surface has a surface energy of 0229 Jm2 and 0515Jm2 relative to (001)Ni
surface indicating NCO surface tends to be Ni rich in agreement with experiment that
when decomposed NiO will form on the surface9 27
Figure 18 Side views of (a) NCO(001) and (b) NCO(100) slabs blue gray and small red
spheres represent Co Ni and oxygen atoms respectively Top views of (c) (001)Ni (d)
(001)Co and (e) (100)mix surfaces only the atoms of the first and second layers are
represented by spheres O1 and O2 are defined in the text
53
Table 4 Average displacements of surface Ni and Co(Oh) and second layer Co(Td) from
their ideal bulk positions Positive (negative) out of plane displacements correspond to
outward (inward) displacements
Surface Type Atom type In-plane
displacementAring
out of plane
displacementAring
(001)Ni
Ni
Co(Td)
O1
O2
0029
0051
0133
0091
-0087
0146
-0118
-0046
(001)Co
Co(Oh)
Co(Td)
O1
O2
0026
0039
0170
0122
-0084
0093
0000
-0049
(100)mix
Ni
Co(Oh)
Co(Td)
O1
O2
0010
0035
0018
0137
0090
-0080
-0118
0106
-0073
-0084
Average displacements of surface and second layer atoms from their ideal bulk positions
are reported in Table 4 while in-plane and out of plane Co-O and Ni-O distances are
compared to computed and experimental10-11 17 bulk anion-cation bond lengths in Table
5 Surface oxygen atoms tend to have large in-plane and smaller out-of-plane
displacements especially on the (001)Co surface and O1 shows somewhat larger
displacements compared to O2 The metallic cations (Ni Co(Oh) and Co(Td)) exhibit
54
smaller in-plane and larger out-of-plane displacements compared to oxygen ions Co(Td)
in the 2nd layer relaxes outward whereas surface atoms relax inward ie toward the bulk
As a result the distance between 1st and 2nd layer is reduced by ~02 Aring
Table 5 Anion-cation bond lengths calculated for NCO bulk and (001)(100) surfaces
where computational value of O-Co(Td) was found to be falls within experimental results
between 188 and 1979 Aring However the increasing bond length in the out of plane
direction dues to the in-plane distortion when forming surface whereas bond length of O-
Co(Td) decreases in general indicating a closer top and 2nd layer distance
Structure type O-M cation type In-plane direction
Aring
Out of plane
direction Aring
bulk
Ni 2007
Co(Oh) 1944
Co(Td) 1933
(001)Ni Ni 1925 2022
Co(Td) 1882
(001)Co Co(Oh) 1892 1949
Co(Td) 1931
(100)mix Ni 1925 2024
Co(Oh) 1907 1993
Co(Td) 1882
In Figure 19 we compare the Densities of States (DOS) of the (100) and (001) slabs to
the DOS of bulk NCO Differences between the bulk and slab DOS are larger for the
minority-spin states than for the majority-spin ones As shown by the layer resolved DOS
55
(Figure 20) these differences originate mainly from the atoms in the surface layers For
the majority-spin states a band gap of 14 eV comparable to the bulk band gap is still
present on the (100) slab whereas the majority spin states of the (001) slab are
conducting due to surface states around the valence bands maximum These surface
states as well as those near the conduction band are mainly contributed by surface
Co(Oh) ions on the (001)Co surface On both (100)mix and (001)Co surfaces Co(Oh) ions
while remaining 3+ become spin polarized with their spin parallel to Ni cations and
anti-parallel to Co(Td) At the same time the 2nd layer Co(Td) ions change their bulk
fractional valence state to a 3+ high spin state
Figure 19 Computed projected DOS for (a) NCO (001) slab (b) NCO(100) slab and (c)
bulk NCO The majority spin states of the (001) slab is conducting due to surface states
on the (001)Co surface
56
Figure 20 layer resolved DOS for a (001)Ni b (001)Co and c (100)mix surfaces Each
layer includes a layer of Oxygen and Oh site cations and an inner layer of Td site cations
32 Surface oxygen vacancies
The formation of surface oxygen vacancies (VOs) has a key role in the oxidation activity
of metal oxide materials often based on the Mars-van Krevelen (MvK) mechanism4 28
We considered 1 and 2 VOs per surface unit cell corresponding to a coverage of 18 and
14 monolayer (ML) respectively The formation energies are calculated as
119864119864119891119891119889119889119903119903119891119891 = 119864119864119889119889119890119890119891119891 + 11989911989921205831205831198741198742 minus 1198641198640 119899119899frasl
where 119864119864119889119889119890119890119891119891 and 1198641198640 are the total energies for the defected and pristine surfaces
respectively n is the number of surface oxygen vacancies 1205831205831198741198742 is the chemical potential
of O2 Our results are summarized in Table 6 (computed VOs formation energies) Figure
57
21 (geometries and electronic structures) and Figure 22 (surface stability diagrams as a
function of the oxygen chemical potential)
From Table 6 we can see that VOs at O1 sites (Eform ~ 11 eV on average) are about 06
eV more favorable than at O2 sites (Eform ~ 17 eV) in the case of one vacancy per unit
cell (18 ML) This can be explained by the fact that a VO at O2 would result in a 3-
cordinated Co(Td) which is quite unfavorable Formation of an O1 vacancy has a
particularly low energy cost on the (001)Co surface (040 eV) followed by O1b vacancies
on the (100)mix surface (082 eV) For comparison the computed VO formation energy is
much larger for bulk NCO (~ 23 eV)20 and for the Co3O4(100) surface (156 eV)
whereas a somewhat smaller VO formation energy 034 eV was obtained for the (100)
surface of NiFe2O4 a spinel catalyst with some similarity to NCO24 From the structural
point of view O1 vacancies on the NCO(001)Co surface are characterized by a strong
local relaxation such that the closest O1 oxygen moves to the middle between two metal
cations resulting in the formation of a so-called ldquosplit-vacancyrdquo The same effect is
observed also for O1a vacancies on the (100)mix surface (Figure 21c)
Table 6 (shown on next page) Average surface oxygen vacancy (VO) formation energies
(in eV) for different surfaces and different VO concentrations Both values at T= 0 K and
under ambient conditions (T=300K and p(O2) = 02 atm corresponding to the oxygen
partial pressure in air) are reported For the (100)mix and (100)exch surfaces inequivalent
O1 sites are denoted as lsquoabcdrsquo in the case of a single vacancy and lsquoaaabbbrsquo in the case
of two vacancies per surface cell (see Figs19e and 21d)
58
Surface Type Vacancy site Formation energy
(0 K)
Formation energy
(300 K 02 atm)
(001)Ni
O1
O2
O1-O1
138
173
141
110
144
112
(001)Co
O1
O2
O1-O1
040
191
079
011
162
051
(100)mix
O1a
O1b
O2
O1a-O1a
O1a-O1b
O1b-O1b
118
082
170
131
113
088
090
054
142
102
085
060
(100)exch
O1a
O1b
O1c
O1d
111
087
125
081
082
059
096
053
Figures 21a and 21b show the electronic structure changes induced by the formation of an
oxygen vacancy on the (100)mix surface Both O1 and to a smaller extent O2 vacancies
generate new empty states in the majority-spin band gap which are primarily contributed
by Co(Oh) and both Ni and Co(Oh) ions respectively with a significant contribution by
oxygen The spin states of Co(Oh) and Ni on the defected surface remain the same as on
59
the pristine surface except for the case of a split-vacancy where Co(Oh) acquires a high
spin state
Figure 21 Projected DOS for the defected (100)mix surface with a VO at (a) O1 and (b)
O2 (c) Top view of the O1a split-vacancy on the (100)mix surface (d) (100)exch surface
obtained after exchanging the Co(Oh) and Ni metal sites on (100)mix inequivalent O1
sites are indicated
The exchange of surface Co(Oh) and Ni ions has a computed energy cost of only 003 eV
and is thus expected to occur quite frequently on NCO(100)mix On the resulting (100)exch
surface (see Figure 21d) four different types of O1 oxygen sites are present
characterized by 0123 Ni neighbors and denoted as O1abcd respectively Our
calculations show that the VO formation energy at O1b and O1d is similar to that at O1b
on the regular (100)mix surface whereas O1a and O1c have formation energies similar to
60
O1a on the (100)mix surface (Table 6) This indicates that 3rd layer ions have larger
influence on the formation of surface oxygen vacancies than surface cations In
particular the formation energy appears to be lower when the surface O1 is bonded to a
3rd layer Ni ion
To better understand the role of 3rd layer cations on O-vacancy formation energies we
performed calculations on (100) surfaces obtained by interchanging Co and Ni ions in the
surface and 3rd layer The resulting surfaces with 25 and 75 Ni cations are shown in
Figure 23 Interestingly the 75 Ni surface is found to be 02-03 eV more stable in
comparison to the (100)mix surface while the 25 Ni surface is 02 eV less stable This
suggests that it is thermodynamically favorable for Ni to segregate at the surface
consistent with the experimental observation that during high temperature decomposition
Ni moves to the surface and forms NiO on top of spinel NCO9 27 The average VO
formation energies and standard deviations on the 75 and 25 Ni surfaces are reported
in Table 7 As shown in Table 7 standard deviations are very small when formation
energies are grouped according to the 3rd layer cations neighboring the vacancy thus
confirming that 3rd layer cations have a major impact on VO formation This result can be
rationalized by considering that upon VO creation the less coordinated cations that are
formed on the surface can undergo significant displacements to partially compensate the
reduced coordination whereas displacements are more difficult for the 3rd layer cations
In particular our computed average VO formation energies are 126 088 and 050 eV for
O-vacancies coordinated to 3rd layer Co Ni and Ni neighboring another Ni ion
respectively It is clearly more favorable to create a VO coordinated to a 3rd layer Ni
compared to a vacancy coordinated to a Co(Oh)3+ because the excess electrons associated
with the vacancy can form a stable Ni2+ species from the original Ni fractional valence
between +2 and +3
61
Table 7 Average surface oxygen vacancy formation energies and corresponding standard
deviations on NCO(100) with 25 and 75 Ni surface concentrations (Figure 23) The
results refer to the different surfaces shown in Figure 23 (first four rows) or to the
different surface oxygen types (last three rows) where O1 is a surface oxygen bound to a
3rd layer Co O2 a surface oxygen bound to a 3rd layer Ni O3 a surface oxygen bound to
a 3rd layer Ni that has a neighboring Ni The standard deviations are significantly smaller
when formation energies are grouped according to the oxygen type indicating that the VO
formation energies depend mostly on the 3rd layer cation
Surface type Oxygen type Average VO
formation energy
eV
Standard deviation
eV
Figure 23a
Figure 23b
Figure 23c
Figure 23d
078 033
075 022
119 027
121 020
O1 126 018
O2 088 006
O3 050 010
62
Figure 22 Phase diagram for VO formation on a (001)Ni b (001)Co and c (100)mix surfaces
as a function of the oxygen chemical potential (referred to an isolated O2 molecule at
T=0K) In all cases the black horizontal line represents the pristine surface Shaded
regions indicate ambient conditions (p(O2) = 02 atm T= 300 K ndash 350 K blue) and
typical conditions for CO and methane oxidation (02 atm at 600 K ndash 800 K yellow)
63
Figure 23 NCO (100) surfaces with different Ni cation concentrations (ab) 25 in the
surface and 75 in the 3rd layer (cd) 75 in the surface and 25 in the 3rd layer VO
sites are indicated as O1 O2 and O3 depending on the 3rd layer cation arrangement O1
oxygen bonding with 3rd layer Co O2 oxygen bonding with 3rd layer Ni O3 oxygen
bonding with 3rd layer Ni and a neighboring Ni similar to oxygen on (001)Co surface
We further investigated the possibility of stronger surface reduction and examined
models containing two O1 vacancies per unit cell As sites for creating the second
vacancy we choose O1 sites not bonding to 4-coordinated Co(Oh) and Ni cations
generated by the first vacancy in order to avoid the formation of 3-coordinated Co(Oh)
64
and Ni ions From Table 6 we can see that the first and second VO formation energies are
similar on the (100)mix and (001)Ni surfaces whereas Eform is much higher (119 eV) for
the second vacancy than for the first one (04 eV) on the (001)Co surface
Finally we combined the results in Table 6 with the temperature and pressure
dependence of the oxygen chemical potential to determine the surface stability diagrams
shown in Figure 22 Note that these diagrams account only for the relative energies of the
structures with 18 and 14 ML VOs and do not take into account the possibility that a
lower energy 18 ML concentration could be created by phase separating into surface
regions with no VOs and regions with frac14 ML VOs Moreover the range of variation of
Δμ(O2) (the oxygen chemical potential referred to an isolated O2 molecule at T=0K) has
been extended beyond the narrow range -081 eV le_Δμ(O2) le 0 that is determined
from the conditions of thermodynamic equilibrium of bulk NCO with O2 gas Co3O4 and
NiO (Table 8) with corrections on O2 binding energy29-30 This is done because NCO
nanoparticles are actually observed to be stable up to rather high temperatures8 As shown
in Figure 22 the behaviors of the three investigated surfaces are clearly quite different
VOs (18 ML) can form easily slightly above room temperature at ambient O2 pressure
on the (001)Co surface whereas VO formation is unlikely on the Ni terminated (001)Ni
surface even under typical oxidation conditions On the (100)mix surface formation of
18 and 14 ML VOs becomes favorable around 700 and 800 K respectively suggesting
that this surface should be quite active in high temperature oxidation reactions
65
Table 8 heat of formation (ΔH) was calculated for NCO Co3O4 and NiO to find the
lower bond of O2 chemical potential for stable NCO crystal structures by
3∆119867119867119873119873119862119862119874119874 minus 2∆11986711986711986211986211988911988931198741198744 minus 3∆119867119867119873119873119894119894119874119874 le12∆1205831205831198741198742 le 0
Where ΔHMO is the heat of formation for metal oxides MO and ∆1205831205831198741198742is the chemical
potential of O2 Calculations are done with no binding energy correction for O2 and
binding energy corrections used by a Jia of 101 eV and b Ceder of 136 eV Results are
compared with experimental heat of formation for Co3O4 and NiO in combination with
O2 binding energy correction of 101 eV Where with corrections on O2 bind energy we
found the NCO will be stable around room temperature or some high temperature
situation
∆HNCO ∆HCo3O4 ∆HNiO Lower bond of
∆microO2
No correction -699 -812 -169 062
Correction a -901 -1014 -219 -039
Correction b -971 -1084 -237 -074
Experimental -929 -253
33 Water adsorption
Water adsorption free energies 119864119864119904119904119889119889119904119904 on pristine and reduced NCO(001)(100) containing
one oxygen vacancy per unit cell (18 ML) are reported in Table 9 Values at T = 0 K
and under ambient conditions (T = 300 K and water pressure pH2O = 002 atm) were
determined using 119864119864119904119904119889119889119904119904 = 1198641198641198671198672119874119874lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 1205831205831198671198672119874119874 + 120549120549120549120549120549120549119864119864 where 1198641198641198671198672119874119874lowast and 119864119864119904119904119904119904119903119903119891119891 are the
computed total energies of the surface with adsorbed water and without water
respectively 1205831205831198671198672119874119874 is the water chemical potential and 120549120549120549120549120549120549119864119864 is the zero point energy
difference between adsorbed water and an isolated water molecule
66
As shown in Table 9 the values of 119864119864119904119904119889119889119904119904 are similar for the Ni and Co sites of pristine
surfaces However water adsorbs in molecular form on Ni sites whereas the adsorption
is dissociative on Co Moreover Nirsquos electronic structure is barely influenced by the
adsorption whereas Co is oxidized from 3+ to 4+ and its magnetization is reduced Water
adsorption on Ni and Co sites is not favorable (119864119864119904119904119889119889119904119904 gt 0) under ambient conditions
indicating that the pristine surface is very little affected by water The introduction of
Van der Waals interactions (not included in our calculations) is not expected to change
this conclusion even though it may slightly strengthen the adsorption at T = 0K
On oxygen-deficient surfaces water adsorption is easier at VO sites where it is
thermodynamically favorable (119864119864119904119904119889119889119904119904 lt 0) also at room temperature For water at a VO the
hydrogen atoms tend to form H-bonds of length 146-167Aring with neighboring lattice
oxygens (Figure 24b) Proton transfer to one of these oxygens is facile making
dissociative adsorption at VOs energetically more stable by about 040 eV relative to
molecular adsorption at the same site Water adsorption is obviously less favorable at the
reconstructed split-vacancy sites For instance even though water tends to remove the VO
reconstruction and dissociate on the (001)Co surface its adsorption energy is very small (-
005 eV) under ambient conditions By combining the results for VO formation (Table 6)
and water adsorption (Table 9) we can also estimate the energetics of hydroxyl formation
on NCO(100)(001) For example the formation energy of an O1b vacancy on the
(100)mix surface is +054 (+082) eV under ambient conditions (at T=0K) while the
dissociative water adsorption energy at VO1b is -049 (-115) eV under the same
conditions This indicates that the surface is likely to be partially hydroxylated at low T
and become pristine under ambient conditions (after desorption of the hydroxyl
hydrogens to form H2) A slightly different picture was obtained for the NiFe2O4(100)
surface24 for which the lower energy cost of VO formation makes hydroxylation highly
favorable also under ambient conditions
67
Figure 24 Water and O2 adsorption structures on the (100)mix surface (top views) (a)
water dissociatively adsorbed at a Co site (b) molecular water at a VO site (c) O2 at a
VO site (d) O2 adsorbed on a surface with two VOs per unit cell Oxygen atoms of
adsorbed molecules are shown in orange oxygen vacancy sites are indicated by a cyan
dotted line
68
Table 9 Computed water adsorption free energies at Co and Ni sites on pristine
(100)(001) and at VO sites on reduced surfaces Both results at T=0K and under ambient
conditions (T = 300 K and pH2O = 002 atm corresponding to the water partial pressure in
air) are reported All values refer to 025 monolayer coverage (ie one adsorbed molecule
per surface unit cell) and positive values indicate that adsorption is thermodynamically
unfavorable For water adsorbed at a VO M and D indicate molecular and dissociative
adsorption respectively VO1b denotes a vacancy at an O1b site Figures showing the
various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
001)Ni
Ni
VO (M)
VO (D)
-042
-094
-157
024
-028
-091
Figure 25a
Figure 25b
Figure 25c
(001)Co Co
VO (M)
VO (D)
-043
-063
-071
022
003
-005
Figure 25d
Figure 25e
Figure 25f
(100)mix Ni
Co
VO1b (M)
VO1b (D)
-044
-033
-079
-115
022
033
-013
-049
Figure 25g
Figure 24a
Figure 24b
Figure 25h
69
Figure 25 Water adsorption structures on different (001)(100) surfaces with and without
oxygen vacancies (only non-split VOs are considered) a water adsorbed at a Ni site on
the pristine (001)Ni surface b water adsorbed at a VO site on the reduced (001)Ni surface
with 1 VOcell c water dissociatively adsorbed at VO on (001)Ni surface with 1 VOcell
d water dissociatively adsorbed at a Co site on the (001)Co pristine surface e water
adsorbed at VO on (001)Co surface with 1 VOcell f water dissociatively adsorbed at VO
site on (001)Co surface with 1 VOcell g water adsorbed at a Ni site on the (100)mix
pristine surface h water dissociatively adsorbed at VO on (100)mix surface with 1 VOcell
Adsorption energies for these structures are given in Table 9 of main text
34 Oxygen adsorption
O2 adsorption is the first step of surface reoxidation in catalytic processes based on the
MvK mechanism Computed O2 adsorption free energies on reduced NCO(100)(001)
surfaces are reported in Table 10 These were determined using 119864119864119904119904119889119889119904119904 = 1198641198641198741198742lowast minus 119864119864119904119904119904119904119903119903119891119891 minus121205831205831198741198742 where 1198641198641198741198742lowast and 119864119864119904119904119904119904119903119903119891119891 are the total energies of the surface with and without
adsorbed O2 and 1205831205831198741198742 is the chemical potential of O2 From Table 10 it appears that O2
can adsorb at a VO but not at a surface metal site at low T whereas adsorption is always
unfavorable under ambient conditions This suggests that surface re-oxidation may be the
70
thermodynamic limiting step for MvK oxidation reactions on NCO(100)(001) For O2
adsorbed at a Vo (Figure 24c) one oxygen atom of the molecule binds to a top layer
metal atom while the other binds to both a metal atom at the surface and a metal of the 3rd
layer The O-O bond length is 136 Aring suggesting the formation of a superoxide O2minus We
also examined whether the adsorbed molecule could dissociate (Figure 26) but found the
dissociated configuration to be less stable than the molecular one by 027 eV Unlike
water O2 does not adsorb on split vacancies notably on the (001)Co surface
Figure 26 O2 adsorption on reduced NCO (001)(100) surfaces a O2 adsorbed at a Ni site
on the (001)Ni surface with 1 VOcell b O2 adsorbed at VO on (001)Ni surface with
1VOcell c O2 adsorbed on (001)Ni surface with 2 VOscell d O2 adsorbed at a Co site
71
on (001)Co surface with 1 split VOcell e O2 adsorbed at non-split VO on (001)Co surface
with 1 split VO and 1 non-split VOcell f O2 adsorbed at Ni on (100)mix surface with 1
VO1bcell g O2 adsorbed at Co on (100)mix surface with 1 VO1bcell h O2 adsorbed at
VO1b on (100)mix surface with 2 VO1bcell i Oxygen transfer onto surface Co from Figure
24c If not specified VO (oxygen vacancies) is the non-split vacancy Adsorption energies
for these structures are given in Table 10 of main text
The adsorption of an O2 molecule on a surface with higher oxygen deficiency (2 VOs per
surface unit cell) is slightly more favorable than on a surface with only 1 VO per cell
However O2 adsorption remains unfavorable at room temperature and atmospheric
pressure When the 2 vacancies are close to each other (eg on the (100)mix surface with
O1a-O1b vacancies or on the (001)Co surface) O2 takes a different adsorption structure
(Figure 24d) where only one of the two oxygens binds to a neighboring surface cation
whereas the other oxygen points toward the vacancy site with an O-O bond length of
137 Aring Breaking of the O-O bond to recover the pristine surface is highly favored
thermodynamically and has a small activation barrier of only 025 eV with an O-O
distance of 139 Aring at the transition state (Figure 27) This suggests that the mechanism of
surface reoxidation by O2 would involve the diffusion of oxygen vacancies to form a
close pair as rate limiting step
72
Table 10 Computed O2 adsorption free energies at Co and Ni sites on pristine
(100)(001) and (non-split) VO sites on reduced (100)(001) surfaces Both values at
T=0K and under ambient conditions (T=300 K and p(O2) = 02 atm corresponding to the
O2 partial pressure in air) are reported All values refer to 025 monolayer coverage (ie
one adsorbed molecule per surface unit cell) and positive values indicate that adsorption
is not favorable VO (M) and VO(M) + VO indicate O2 adsorption at a vacancy site of a
reduced surface with 1 VO and 2 VOs per surface unit cell respectively Figures showing
the various structures are listed in the last column
Surface Type Adsorption site Eads (0 K)eV Eads (air)eV Figure
(001)Ni Ni
VO (M)
VO (M) + VO
017
-040
-062
082
025
003
Figure 26a
Figure 26b
Figure 26c
(001)Co Co
VO (M) + VO
-001
-043
063
022
Figure 26d
Figure 26e
(100)mix Ni
Co
VO1b (M)
VO1a (M) + VO1b
VO1b (M) + VO1b
-003
005
-046
-055
-047
062
070
019
010
017
Figure 26f
Figure 26g
Figure 16c
Figure 16d
Figure 26h
73
Figure 27 Selected structures along the reoxidation pathway of a (100)mix surface with
2VOscell (a) initial state with O2 adsorbed at a VO and pointing toward the other VO
(b) transition state with slightly increased O-O distance (c) final state showing the
reoxidized (100)mix surface Relative energies are listed below the structures Calculations
were performed via constrained minimizations
4 Conclusions
In this work we have studied the structure and chemistry of NCO(100)(001) surfaces
with different CoNi terminations using DFT+U calculations Our results show that there
is a thermodynamic driving force for Ni to segregate to the surface which is consistent
with the experimental observation of NiO formation on the surface during thermal
decomposition27 On the other hand oxygen vacancy formation is considerably more
difficult on the purely Ni-terminated NCO(001)Ni surface than on the Co-terminated
(001)Co and mixed Ni and Co-terminated (100)mix surfaces (Figure 22) The latter are thus
expected to represent the active surfaces in oxidation reactions On these surfaces VO
formation is easiest at O1 sites which are not bound to 2nd layer Co(Td) and especially at
O1 sites that are bound to 3rd layer Ni atoms while VOs at O1 sites with more Co(Oh)
than Ni neighbors tend to reconstruct to form split-vacancies The computed formation
energy of a regular (ie non reconstructed) VO at O1 is approximately 08 ndash 09 eV at T =
0 K (Table 6) which is essentially half the value (156 eV) that we find for a VO on the
Co3O4(100) surface Easier VO formation on NCO suggests that this material may be a
74
better oxidation catalyst than Co3O4 under mild conditions (ie at temperatures below Ni
segregation to the surface takes place)
NCOrsquos surface reactivity has been further characterized by studying the adsorption of two
typical probe molecules water and O2 Both molecules preferentially adsorb at oxygen
vacancy sites at low temperature Under ambient conditions however VOs can be easily
healed via dissociative water adsorption whereas adsorption of O2 is not favorable These
results suggest that O2 adsorption is likely to represent the thermodynamic limiting step
for oxidation reactions on NCO(001)(100) surfaces
75
5 References
1 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
2 Ma C Y Mu Z Li J J Jin Y G Cheng J Lu G Q Hao Z P Qiao S
Z Mesoporous Co3o4and AuCo3o4catalysts for Low-Temperature Oxidation of Trace
Ethylene J Am Chem Soc 2010 132 2608-2613
3 Hu L Peng Q Li Y Selective Synthesis of Co3o4nanocrystal with Different
Shape and Crystal Plane Effect on Catalytic Property for Methane Combustion J Am
Chem Soc 2008 130 16136-16137
4 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
5 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
6 Chen S Qiao S-Z Hierarchically Porous Nitrogen-Doped Graphenendash
Nico2o4hybrid Paper as an Advanced Electrocatalytic Water-Splitting Material ACS
Nano 2013 7 10190-10196
7 Shi H Zhao G Water Oxidation on Spinel Nico2o4nanoneedles Anode
Microstructures Specific Surface Character and the Enhanced Electrocatalytic
Performance J Phys Chem C 2014 118 25939-25946
8 Farrauto R J Low-Temperature Oxidation of Methane Science 2012 337 659-
660
9 Iliev M N Silwal P Loukya B Datta R Kim D H Todorov N D
Pachauri N Gupta A Raman Studies of Cation Distribution and Thermal Stability of
Epitaxial Spinel Nico2o4 Films J Appl Phys 2013 114 033514
10 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E I Palmer H M
Greaves C Berry F J Journal of Materials Chemistry 2001 11 3087-3093
76
11 Marco J F Gancedo J R Gracia M Gautier J L Riacuteos E Berry F J
Characterization of the Nickel Cobaltite Nico2o4 Prepared by Several Methods An
Xrd Xanes Exafs and Xps Study J Solid State Chem 2000 153 74-81
12 Bitla Y et al Origin of Metallic Behavior in Nico2o4 Ferrimagnet Scientific
Reports 2015 5 15201
13 Cui B Lin H Liu Y-z Li J-b Sun P Zhao X-c Liu C-j
Photophysical and Photocatalytic Properties of Core-Ring Structured
Nico2o4nanoplatelets J Phys Chem C 2009 113 14083-14087
14 Zakutayev A Paudel T R Ndione P F Perkins J D Lany S Zunger A
Ginley D S Cation Off-Stoichiometry Leads to Highp-Type Conductivity and
Enhanced Transparency in Co2zno4and Co2nio4thin Films Phys Rev B 2012 85
085204-1 - 085204-8
15 Silwal P Miao L Stern I Zhou X Hu J Ho Kim D Metal Insulator
Transition with Ferrimagnetic Order in Epitaxial Thin Films of Spinel Nico2o4 Appl
Phys Lett 2012 100 032102
16 Lee D U Kim B J Chen Z One-Pot Synthesis of a Mesoporous Nico2o4
Nanoplatelet and Graphene Hybrid and Its Oxygen Reduction and Evolution Activities as
an Efficient Bi-Functional Electrocatalyst J Mater Chem A 2013 1 4754
17 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
18 Kim J G Pugmire D L Battaglia D Langell M A Analysis of the Nico2o4
Spinel Surface with Auger and X-Ray Photoelectron Spectroscopy Appl Surf Sci 2000
165 70-84
19 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
20 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
77
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
23 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
24 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
27 Cabo M s Pellicer E Rossinyol E Castell O Surinach S Baro M D
Mesoporous Nico2o4spinel Influence of Calcination Temperature over Phase Purity and
Thermal Stability Cryst Growth Des 2009 9 4814-4821
28 McFarland E W Metiu H Catalysis by Doped Oxides Chem Rev 2013 113
4391-4427
29 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Phys Rev B 2011 83 245204-1 -245204-7
30 Wang L Maxisch T Ceder G Oxidation Energies of Transition Metal Oxides
within Thegga+Uframework Phys Rev B 2006 73 195107-1 - 109107-6
78
Chapter V
Mechanism and activity of the oxidation
of CO and methane on spinel Co3O4 and
NiCo2O4
1 Introduction
Various oxidation reactions are found to proceed very efficiently on Fe Co and Ni based
spinel oxides1-2 For example ferrites shows high catalytic activity for oxygen evolution3-
4 and is a potential catalyst for the water gas shift reaction5 (though limited by water
desorption6) while spinel cobaltites especially Co3O4 and NiCo2O4 (NCO) are very
active for CO oxidation7-8 and hold promise as catalysts for low temperature methane
oxidation9-10
Among oxidation reactions the oxidation of CO to CO2 is of great interest both as the
reaction typically used for reducing the amount of CO in waste gases and also as a
benchmark reaction for oxidation catalysts Following the work of Haruta et al8
numerous experimental studies have examined CO oxidation on Co3O48 11-13
The (110)
surface has been found to be the most active low-index surface of this material CO can
be converted to CO2 even at -70 degC and complete conversion takes place at around
150 degC8-9 On the theoretical side studies of CO oxidation on Co3O4 (110)14-16 have
predicted a Mars-van Krevelen mechanism with a major role of Co ions at octahedral
sites However room temperature CO oxidation on Co3O4 (110) gradually deactivates
during the reaction11 17 which has been attributed to surface reconstruction and water
adsorption Moreover (110) surfaces are not very common for the spinel structure (111)
and (001) surfaces have lower surface energies and thus generally make larger
contributions to the surface area18 For Co3O4 however the (111) and (001) surfaces
79
have been reported to be inactive for CO oxidation8 The origin of these different surface
activities is difficult to understand on the basis of simple structural differences
To obtain better insight we have carried out a comparative theoretical study of the
COOR 2CO + O2 2CO2 over pristine and defected Co3O4(110) Co3O4(001) and
NiCo2O4(001) surfaces using DFT calculations with the addition of on-site Coulomb
repulsion U terms on Co Ni and Fe 3d shells (DFT+U)21 Our results suggest different
rate determining steps for the different surfaces of COOR In particular surface
reoxidation is found to be rate limiting on (001) surfaces while CO2 formation is the
most difficult step on Co3O4 (110)
Methane is an abundant energy resource and there is currently enormous interest in
developing more efficient and environmentally benign strategies for its utilization A
better understanding of methane oxidation is essential for such development Numerous
studies have examined methane oxidation on Co3O4 It was found that methane is
completely oxidized on the Co3O4 (110) surface and lattice oxygen atoms are involved in
the oxidation process10 Recent theoretical studies indicate the possibility of incomplete
methane oxidation on the Co3O4 (001) surface19 which may lead to interesting
applications It was also reported that NCO is a better catalyst for methane oxidation
compared to Co3O49
Motivated by this finding we have investigated the first two steps
of methane oxidation on NiCo2O4(001) notably the first C-H bond breaking that is
commonly believed rate determining20 as well as the second C-H bond breaking An
interesting result is that methane oxidation on NiCo2O4(001) has a pseudo barrier
comparable to that found on doped CeO222 a well-known catalyst for low temperature
methane oxidation
2 Methods and Models
Spin polarized DFT+U calculations were performed using the plane-wave-
pseudopotential scheme as implemented in the Quantum Espresso package23 Exchange
and correlation terms were described using the Perdew-Burke-Ernzerhof (PBE)24
functional with on-site Coulomb repulsion U term on Co Ni and Fe 3 d electrons We
80
used the values U(Co)=30 eV U(Ni)=55 eV and U(Fe)=35 and 34 eV for Fe at Oh and
Td sites which we found to provide a better description of Co3O4 and NCO and NFOrsquos
thermodynamic properties compared to the U values determined from Linear Response
theory25-26 Ultrasoft pseudopotentials27 were employed and the valence electrons include
O 2 s 2 p Ni 3 d 4 s Co 3 d 4 s and Fe 3 d 4 s states Kinetic energy cutoffs used were
35 and 350 Ryd for wave functions and augmented density on Co3O4 and NCO and 50
and 500 Ryd for wave functions and augmented density on NFO Structural optimizations
were carried out by relaxing all atoms until forces were smaller than 10 times 10-3 au
Co3O4 NCO and NFO (001) surfaces were modeled using slabs of 11 layers terminated
by oxygen and metal cations occupying octahedral sites in the bulk material The Co3O4
(110) surface was modeled using slabs of 7 layers with oxygen rich termination A
vacuum region of 20 Aring was used to separate adjacent slabs We kept similar chemical
composition for the different (001) slabs with 10 metal atoms at tetrahedral (Td) sites 24
metal atoms at octahedral (Oh) sites and 48 oxygens The Co3O4 (110) slab included 14
(6) Co atoms at Oh (Td) sites and 28 oxygens We employed theoretically optimized
lattice parameters and considered square (2 x 2) surface unit cells of dimensions 8136 times
8136 Aring2 8209 times 8209 Aring2 and 8447 times 8447 Aring2 for Co3O4 NCO and NFO (001)
surfaces respectively25-26 28 the unit cell dimensions for Co3O4 (110) were 8149 times
5762 Aring2 We sampled the surface Brillouin zone using a 2 times 2 times 1 k-point grid for both
(001) and (110) surfaces CO and O2 molecules were adsorbed only on one of the two
surfaces of each slab previous calculations showed that dipole corrections have minor
influence (001 eV) and were not included in our setup
Kinetic barriers for selected reaction steps were determined using the Climbing Image
Nudged Elastic Band (CI-NEB) method29 We typically used 9 images that we optimized
until forces were smaller than 005 eVAring
Adsorption energies of various species X (eg CO O2 etc) were calculated using
119864119864119904119904119889119889119904119904 = 119864119864119883119883lowast minus 119864119864119904119904119904119904119903119903119891119891 minus 120583120583119883119883
81
where 119864119864119883119883lowast and 119864119864119904119904119904119904119903119903119891119891 are the computed total energies of the surfaces with and without
adsorbed X respectively and 120583120583119883119883 denotes the chemical potential of X calculated from
JANAF Tables30
3 Results and Discussion 31 CO adsorption and oxidation on Co3O4 (001) and (110) surfaces
311 CO adsorption
CO adsorption energies on Co3O4(001) and (110) surfaces are reported in Table 11 For
Co3O4 (001) we considered both the pristine surface and the surface with one oxygen
vacancy (VO) per unit cell (Figure 28ab) as the VO concentration may influence further
VO formation and therefore also the energetics of CO oxidation through the Mars-van
Krevelen (MvK) mechanism The computed CO adsorption energy at Co(Oh) is Eads = -
092 eV (at T= 0K) The adsorbed CO is perpendicular to the surface with C-Co distance
of 175 Aring suggesting a rather strong interaction CO adsorption is less favorable at a VO
site (Eads = -065 eV) while it is much more favorable at the two nonequivalent O1 (Eads
= -186 eV) and O2 (Eads = -143 eV) oxygen sites31 (Figure 28a) When CO adsorbs at
the latter sites it actually forms a CO2 molecule without any apparent intermediate or
barrier The resulting adsorption configuration can be described as a CO2 at a surface VO
(see structure II in Figure 29) consistent with a MvK mechanism
82
Figure 28 a Top view of the (2x2) surface unit cell and b side view of the top few layers
of the Co3O4 (001) slab model used in our calculations Only the atoms of the top and
second layer are shown as balls The dashed orange circle in a indicates a surface oxygen
vacancy (VO ) other labels indicate Co cobalt at a surface octahedral site O1 surface
oxygen bonding with 3rd layer Co(Oh) O2 surface lattice oxygen bonding with 2nd layer
Co(Td) c Top view of the surface unit cell and d side view of the first few layers of the
pristine Co3O4 (110) slab model used in our study Labels indicate Co cobalt at a surface
octahedral site O1 surface oxygen bonding with 2 inner layer Co(Oh) and 1 surface
Co(Oh) O2 surface oxygen bonding with 1 inner layer Co(Td) and 1 surface Co(Oh)
For Co3O4 (110) we considered only the pristine surface (Figure 28cd) as the defected
(110) surface is found to reconstruct25 The computed adsorption energy for CO at the 4-
coordinated surface Co(Oh) is Eads = -180 eV thus more negative (favorable) than on
the (001) surface where Co(Oh) is 5-fold coordinated CO points toward one of the
Co(Oh)-O bonds forming an angle of 45 degrees with the surface In addition to
Co(Oh) two types of oxygen sites exist for CO adsorption (Figure 28c) notably O1
bound to three Co(Oh) where Eads = -149 eV and O2 bound to one Co(Oh) and one
Co(Td) where Eads = -191 eV thus slightly larger than on Co(Oh) At both O1 and O2
sites CO forms bridging bonds with the oxygen and a surface Co(Oh) with bond lengths
83
of 130 and 195 Aring respectively (see structure II in Figure 30 where all CO adsorption
structure on lattice oxygen is shown in Figure 31)
Table 11 Computed CO adsorption energies (in eV) at various surface sites on pristine
and defected Co3O4 (001) and (110) surfaces Adsorption energies are reported at both T
= 0 K and T = 300 K (in parentheses) with CO pressure of 1 atm
Co O1 O2 VO
Pristine 001 -095(-042) -186(-133) -143(-090)
Defected
001 -090(-038) -175(-122) -104(-052) -065(-013)
Pristine 110 -180(-128) -149(-097) -191(-139)
312 CO oxidation
Our results for the intermediates and energetics of the COOR on Co3O4 (001) and
Co3O4(110) are summarized in Figure 29 Figure 30 and Table 12 The first three
structures in Figure 29 correspond to the bare (001) surface (I) the adsorption of a CO
molecule and formation of CO2 by abstraction of a surface oxygen atom (II) and the
surface with an oxygen vacancy after the desorption of CO2 (III) In the subsequent step
an O2 molecule adsorbs at the VO site with one of the oxygen atoms taking the position
of the missing lattice oxygen while the other oxygen stretches outwards (IV) the O-O
bond length is 1344 Aring suggesting a superoxide species O2minus The computed O2
adsorption energy is -042 eV at T = 0K and + 017 eV ie endothermic under ambient
conditions (Table 12) These results indicate that re-oxidation of the surface is difficult at
room temperature The last step in Figure 29 is the adsorption of a second CO molecule
which reacts with the protruding oxygen of O2 and leaves as CO2 (V) so that the surface
can recover its initial state The energy release for this process is large - 444 (-392) eV
at T=0 (300) K The presence of an additional VO on the (001) surface has only minor
84
influence on the various intermediates (Table 12) the main effect being to slightly
increase the likelihood of O2 adsorption
Figure 29 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on pristine Co3O4 (001) pristine surface (I) adsorbed CO forming a
CO2 molecule with a surface O atom (II) defective surface following CO2 desorption
(III) re-oxidized surface with an O2 molecule at the vacancy site (IV) pristine surface
with physisorbed CO2 formed upon adsorption of another CO (V) Free energy changes
are given at both T=0 K and T=300 K (between parentheses) with CO O2 and CO2
partial pressures of 1 atm 02 atm and 00004 atm respectively
The intermediates of the COOR on Co3O4(110) are shown for the case of adsorption at
the O1 site (Figure 30) After CO adsorption and CO2 formation desorption of CO2 to
form a surface VO has an energy cost of +013 eV at 0K (-064 eV at RT) close to the
analogous cost on the (001) surface At variance with the (001) surface however the
subsequent adsorption of O2 at the vacancy site on Co3O4 (110) is very favorable being
exothermic at both T = 0K (-140 eV) and RT (-081 eV) The O2 adsorption
configuration is characterized by one of the oxygen atoms occupying the missing lattice
oxygen site while the other oxygen stretches out toward a neighboring Co(Oh) with an
85
O-O bond length of 137 Aring This oxygen is very reactive when another CO molecule
arrives CO2 forms readily without barrier and with an energy release of -458 eV In the
case of CO adsorption at the O2 site CO2 formation does not involve the CO bridging
structure in Figure 30 and the COOR mechanism is more similar to the one on the (001)
surface As shown in Table 12 the energy release is -197 (-144) eV compared to -113
(-060) eV for the O1 site while CO2 desorption has an energy cost of +040 (-038) eV at
0 K (RT) which is higher than the value found for the O1 site Adsorption of O2 is also
very favorable -105 (-046) eV at 0 K (RT) and further CO oxidation occurs with an
energy release of -399 (-346) eV
Figure 30 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on Co3O4 (110) with initial CO adsorption at the O1 site The
sequence of intermediates is the same as in Figure 29 Free energy changes are given at
both T=0 K and T=300 K (between parentheses) with CO O2 and CO2 partial pressures
of 1 atm 02 atm and 00004 atm respectively
By comparing the free energy changes for the various steps of the COOR on Co3O4 (110)
and (001) surfaces (Table 12) we can see that the thermodynamic limiting step of the
COOR on Co3O4(001) is the adsorption of O2 which becomes unfavorable around room
86
temperature On Co3O4 (110) on the other hand the COOR is more likely controlled by
the kinetics of CO2 desorption which requires a significant structural rearrangement
whereas O2 adsorption is more favorable
Figure 31 CO adsorption structure on different spinel surfaces with a on Co3O4 (110)
pristine surface b on Co3O4 (001) pristine surface c on NCO (001)Ni pristine surface
d on NCO (001)Co surface with one splitted oxygen vacancy e on NCO (100)mix pristine
surface and f on NFO (001) pristine surface
87
Table 12 Computed free energy changes (in eV) for the various steps of the COOR on
the Co3O4 (001) and (110) surfaces The labels PD indicate pristine and defected (with a
VO) surface respectively Results are reported at both T= 0 K and T = 300K (between
parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
O1 site (001)
P
-186(-133) 015(-062) -042(017) -444(-392) 004(-073)
O1 site (001)
D
-175(-122) 014(-063) -053(006) -458(-406) 019(-058)
O1 site (110)
P
-113(-060) 013(-064) -134(-075) -428(-376) 010(-068)
O2 site (110)
P
-197(-144) 040(-038) -105(-046) -399(-346) 009(-068)
32 CO oxidation on NCO (001)(100) surfaces
Different from normal spinel Co3O4 NCO has an inverse spinel structure where
tetrahedral (Td) sites are occupied primarily by Co2+ and Co3+ ions and octahedral (Oh)
sites by Ni2+ Ni3+ and Co3+ ions27 NCO (001)(100) surfaces can expose both Co(Oh)
and Ni cations or only one type of metal cation following Ref31 we denote (100)mix a
surface exposing both Co(Oh) and Ni while (001)Ni and (001)Co expose only Ni and
Co(Oh) respectively
Figure 32 and Table 13 show the computed intermediates and free energy changes for the
COOR on pristine and defected NCO(001)(100) By comparing these results to those in
88
Table 12 for Co3O4 we can see that the free energy release for CO adsorption and CO2
formation (step I II) on NCO is larger than on Co3O4 This free energy release is
indeed strongly correlated with the VO formation energy for which cations in the 3rd
layer play a key role Ni cations causing smaller VO formation energies31 On the other
hand computed O2 adsorption energies at VO sites (step IIIIV) on reduced
NCO(001)(100) with 18 and frac14 ML VO coverages are similar to those on Co3O4 (001)
ie are typically positive at RT (Table 13) This suggests that reoxidation is difficult also
on NCO On the Co-terminated NCO(001)Co surface in particular VOs tend to
reconstruct at low coverage31 which prevents O2 adsorption Thus the COOR cannot
continue on this surface except at high VO concentrations where non-reconstructed VOs
are also present Reduced surfaces always tend to adsorb O2 more easily than pristine
ones as expected
Table 13 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NCO (001)(100) surfaces Results are reported at both T= 0
K and T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm
respectively
I II II III III IV IV V V I
P - (001)Ni -217(-164) 028(-049) -039(019) -423(-370) -003(-080)
D - (001)Ni -203(-151) 020(-057) -062(-004) -322(-270) -085(-163)
D - (001)Co -217(-165) 009(-068) -043(016) -373(-320) -029(-107)
P - (100)mix -250(-197) 006(-072) -046(013) -342(-290) -021(-098)
D - (100)mix -210(-157) 027(-050) -055(004) -425(-372) 009(-068)
89
Figure 32 Intermediates (side views) and free energy changes (in eV) for the various
steps of the COOR on the NCO (100)mix pristine surface The sequence of intermediates
is the same as in Figure 29 Free energy changes are given at both T=0 K and T=300 K
(between parentheses) with CO O2 and CO2 partial pressures of 1 atm 02 atm and
00004 atm respectively
To further characterize the COOR on NCO we studied the kinetic pathways of two key
steps of the reaction the formation of the 1st CO2 (I II) and of O2 adsorption (III
IV) (Figure 33) For the 1st CO adsorptionCO2 formation NEB calculations give a
barrier of 011 eV (Figure 33a) The C=O bond length at the transition state is 112 Aring
while the distances of the C atom to surface oxygen and Ni atoms are 135 and 217 Aring
respectively For O2 adsorption on defected NCO (100)mix NEB calculations show a
smooth potential energy surface with a very small barrier of about 001 eV (Figure 33b)
The O-O bond length at the transition state is 1239 Aring and the distance of the lower
oxygen to the surface is 1403 Aring indicating that interaction with the surface is still weak
These results suggest that the COOR on NCO is controlled by the thermodynamics of O2
adsorption
90
Figure 33 pathways of a formation of CO2 (I rarr II) and b O2 adsorption (III rarr IV) with
their transition barrier shown in the middle of the panel The relative energy (in eV) with
respect to initial structure is display below the structures CO2 formation is accomplished
by CO adsorption to surface lattice oxygen and then leave as CO2 after itrsquos adsorbed
33 Methane oxidation on the NCO (100)mix surface 331 Methane adsorption and first C-H bond breaking
Our calculation indicate that methane can only physisorb on the NCO (100)mix surface
CH4 remaining more than 28 Aring away from the surface To determine the energy of C-H
bond breaking we calculated the adsorption energies of several configurations with
hydrogen and methyl groups at different adsorption sites The methyl group favors
adsorption at the surface O1 site (Figure 34a) with hydrogen adsorbed on an O1 site
away from the methyl the computed adsorption energy is -222 eV Possible intermediate
states were also studied (Figure 34) When the methyl group is adsorbed on a surface Ni
(Co) site and H on a neighboring O1 a formation energy of -040 (-051) eV is obtained
91
(Figure 34bc) for the intermediate state The C-M bond length is around 197 Aring
Intermediate states with a methanol like structure involving methyl and hydrogen
adsorbed on the same surface O1 and O2 sites were also studied (Figure 34d e) We
obtained formation energies of -128 and -046 eV at O1 and O2 respectively with a C-O
bond length of ~ 146 Aring The adsorption barrier was estimated by considering the
configuration shown in Figure 34f with hydrogen adsorbed on the surface and the methyl
radical in gas phase for which an adsorption energy of 048 eV is found Although the
true barrier may be slightly larger than our estimated (pseudo) barrier22 it would still be
comparable to that on doped CeO2 a well-known low temperature methane oxidation
catalyst indicating the ability to break C-H bond at relative low temperature After
adsorption of the methyl radical onto the surface hydrogen diffusion will be favored
thermodynamically over desorption of methanol and further oxidationdehydrogenation
should occur
Figure 34 first C-H bond breaking structures on NCO (100)mix surface a most stable
structure with methyl and hydrogen adsorbed on different O1 sites b methyl adsorbed
on Ni c methyl adsorbed on Co d methyl and hydrogen adsorbed on the same O1 site
to form methanol e methyl adsorbed and hydrogen adsorbed on same O2 site f gas
phase methyl radical and hydrogen adsorbed surface to approximate transition barrier
92
332 Second C-H bond breaking
To verify the viability of further methane oxidation we studied the energetics of the
second C-H bond breaking from the stable methyl adsorption structure After C-H bond
breaking the resulting methylene group (CH2) adsorbs forming a bridge between O1 and
neighboring surface sites When CH2 is bridging O1 with Ni and Co the total energy
decreases by 038 eV and 058 eV respectively the C-O bond is reduced to around 136 Aring
and C-M bond length is around 197 Aring CH2 bridging O1 with another O1 and O2 sites is
favored with formation energy of -142 eV and -132 eV respectively The average C-O
bond length is slightly reduced to 142 Aring for the O1-C-O1 bridge and is almost the same
for the O1-C-O2 bridge Unlike in the cases of CO oxidation and first C-H bond
breaking the reactivity of the O2 site in the 2nd C-H bond breaking is comparable to that
of the O1 site It is possible to create O2 site vacancies with the 2nd C-H bond breaking
which may facilitate the surface reoxidation
4 Conclusions
Our results show that the (001) surfaces of spinel Co3O4 NCO and NFO have all the
same mechanism for the CO oxidation reaction CO molecules adsorb strongly at lattice
oxygen sites and form CO2 readily while re-oxidation through O2 adsorption at VO sites
is thermodynamically unfavorable at room temperature or higher This may be the reason
why the Co3O4(100) surface has been reported to be rather inert in experiment On the
other hand our results show that the rate-limiting step of CO oxidation on the Co3O4
(110) surface which is found to be quite reactive in experiment is not the re-oxidation
but the formation of CO2 from adsorbed CO If the temperature is relatively low CO
oxidation on (001) surfaces is also limited by the hydroxylation of the surface upon
water adsorption and dissociation at a VO site surface OH groups are formed that hinder
the interaction of CO with lattice oxygens These results suggest that the COORrsquos activity
of Co3O4 and NCO (001) surfaces is determined by the weak interaction of O2 with the
surface and competing molecules like water can adsorb at the VO site and block the active
site Our calculations also suggest that the NCO (001) surface has potential catalytic
93
activity for low temperature methane oxidation with a transition barrier comparable to
various doped CeO2 for the first C-H bond breaking rate determining step
94
5 References
1 Liang Y Li Y Wang H Zhou J Wang J Regier T Dai H Co3o4
Nanocrystals on Graphene as a Synergistic Catalyst for Oxygen Reduction Reaction
Nature Materials 2011 10 780-786
2 Ma N Selective Oxidation of Styrene over Nanosized Spinel-Type
Mgxfe3minusXo4 Complex Oxide Catalysts Applied Catalysis A General 2003 251 39-47
3 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
4 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
5 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
6 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
7 Zhu J Gao Q Mesoporous Mco2o4 (M=Cu Mn and Ni) Spinels Structural
Replication Characterization and Catalytic Application in Co Oxidation Microporous
Mesoporous Mater 2009 124 144-152
8 Xie X Li Y Liu Z-Q Haruta M Shen W Low-Temperature Oxidation of
Co Catalysed by Co3o4 Nanorods Nature 2009 458 746-749
9 Ren Z Botu V Wang S Meng Y Song W Guo Y Ramprasad R Suib
S L Gao P-X Monolithically Integrated Spinel Mxco3minusXo4(M=Co Ni Zn)
Nanoarray Catalysts Scalable Synthesis and Cation Manipulation for Tunable Low-
Temperature Ch4and Co Oxidation Angew Chem Int Ed 2014 126 7351-7355
10 Tao F F Shan J-j Nguyen L Wang Z Zhang S Zhang L Wu Z
Huang W Zeng S Hu P Understanding Complete Oxidation of Methane on Spinel
Oxides at a Molecular Level Nat Commun 2015 6 7798
95
11 Jansson J Palmqvist A E C Fridell E Skoglundh M Oumlsterlund L
Thormaumlhlen P Langer V On the Catalytic Activity of Co3o4 in Low-Temperature Co
Oxidation Journal of Catalysis 2002 211 387-397
12 Hu L Sun K Peng Q Xu B Li Y Surface Active Sites on Co3o4 Nanobelt
and Nanocube Model Catalysts for Co Oxidation Nano Research 2010 3 363-368
13 Jansson J Low-Temperature Co Oxidation over Co3o4Al2o3 Journal of
Catalysis 2000 194 55-60
14 Broqvist P A Dft Study on Co Oxidation over Co3o4 Journal of Catalysis
2002 210 198-206
15 Pang X-Y Liu C Li D-C Lv C-Q Wangthinsp G-C Structure Sensitivity of
Co Oxidation on Co3o4 A Dft Study ChemPhysChem 2013 14 204-212
16 Xu X-L Yang E Li J-Q Li Y Chen W-K A Dft Study of Co Catalytic
Oxidation by N2o or O2on the Co3o4(110) Surface ChemCatChem 2009 1 384-392
17 Xu X L Li J Q Dft Studies on H2o Adsorption and Its Effect on Co
Oxidation over Spinel Co3o4 (110) Surface Surface Science 2011 605 1962-1967
18 Zasada F Gryboś J Indyka P Piskorz W Kaczmarczyk J Sojka Z
Surface Structure and Morphology of M[Comprime]O4(M = Mg Zn Fe Co and Mprime = Ni Al
Mn Co) Spinel NanocrystalsmdashDft+U and Tem Screening Investigations J Phys Chem
C 2014 118 19085-19097
19 Hu W Lan J Guo Y Cao X-M Hu P Origin of Efficient Catalytic
Combustion of Methane over Co3o4(110) Active Low-Coordination Lattice Oxygen and
Cooperation of Multiple Active Sites ACS Catalysis 2016 6 5508-5519
20 Arndtsen B A Bergman R G Mobley T A Peterson T H Selective
Intermolecular Carbon-Hydrogen Bond Activation by Synthetic Metal Complexes in
Homogeneous Solution Accounts of Chemical Research 1995 28 154-162
21 Anisimov V I Zaanen J Andersen O K Band Theory and Mott Insulators
Hubbarduinstead of Stoneri Phys Rev B 1991 44 943-954
22 Kumar G Lau S L J Krcha M D Janik M J Correlation of Methane
Activation and Oxide Catalyst Reducibility and Its Implications for Oxidative Coupling
ACS Catalysis 2016 6 1812-1821
96
23 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
24 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
25 Selcuk S Selloni A Dft+U Study of the Surface Structure and Stability of
Co3o4(110) Dependence on U J Phys Chem C 2015 119 9973-9979
26 Shi X Li Y-F Bernasek S L Selloni A Structure of the Nife2o4(001)
Surface in Contact with Gaseous O2 and Water Vapor Surface Science 2015 640 73-79
27 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
28 Shi X Bernasek S L Selloni A Formation Electronic Structure and Defects
of Ni Substituted Spinel Cobalt Oxide A Dft+U Study J Phys Chem C 2016 120
14892-14898
29 Henkelman G Uberuaga B P Joacutensson H A Climbing Image Nudged Elastic
Band Method for Finding Saddle Points and Minimum Energy Paths The Journal of
Chemical Physics 2000 113 9901-9904
30 Malcolm W Chase Jr Nist-Janaf Thermochemical Tables Fourth edition
Washington DC American Chemical Society New York American Institute of
Physics for the National Institute of Standards and Technology 1998 1998
31 Shi X Bernasek S L Selloni A Oxygen Deficiency and Reactivity of Spinel
Nico2o4 (001) Surfaces The Journal of Physical Chemistry C 2017 121 3929-3937
97
Chapter VI
Surface chemistry of NiFe2O4(001)
surface structure in contact with O2 and
water vapor and reactivity toward CO
and methane
1 Introduction
The spinel ferrites with general formula AFe2O4 are materials of both fundamental and
technological interest1 In particular NiFe2O4 is a promising material for magnetic
storage systems2 magnetic-resonance imaging3 spintronics4-5 etc Recently NiFe2O4 has
also attracted significant attention for its catalytic activity for the oxygen evolution
reaction6 and as a potential catalyst for the Water Gas Shift (WGS) reaction7 where the
reactivity is limited by water desorption8 The interaction of NiFe2O4 surfaces with water
plays a key role in all the applications of this material in catalysis Understanding this
interaction is thus essential for the design of more efficient NiFe2O4 catalysts for the
WGS and other oxidation reactions
While numerous theoretical9 and experimental10-12 investigations of the bulk electronic
and magnetic properties of NiFe2O4 have been reported studies on the surfaces of
NiFe2O4 are still scarce NiFe2O4 exposes different surfaces depending on the growth and
preparation conditions11-12 the (111) and (001) surfaces being the most frequent ones For
instance hydrothermal synthesis of NiFe2O4 nanoparticles often leads to faceted
octahedra enclosed by (111) planes13 while films grown on MgAl2O4 or SrTiO3 expose
the (001) surface11-12 Recently DFT calculations have shown that H2O undergoes strong
dissociative adsorption on the metal terminated (111) surface13 Instead the structure of
the (001) surface and its interaction with water have not yet been studied
98
To help fill this gap we present here a computational study of the structure of the
NiFe2O4(001) surface exposed to molecular oxygen and water vapour the most common
gases with which a surface can be in contact Using the DFT+U method we start by
studying the electronic density of states (DOS) of bulk NiFe2O4 both in the absence and
in the presence of an oxygen vacancy These results are used as a reference in order to
understand the effects of the (001) termination on the electronic structure We next study
the adsorption of water on both the defect-free and defected NiFe2O4(001) surfaces Our
computed surface stability diagram predicts that the NiFe2O4(001) surface is
hydroxylated at ambient conditions while it exhibits surface oxygen vacancies in the
temperature range of 600-900 K that is often used for catalytic reactions
Another important aspect of the surface chemistry of NiFe2O4(001) that is discussed in
this chapter is the surface reactivity toward CO and methane The oxidation reactions of
CO and methane on NiFe2O4(001) are found to go through reaction paths similar to those
found for NiCo2O4 CO oxidation is limited by O2 adsorption on vacancies which is more
difficult than on NiCo2O4 Methane oxidation is found to have a slightly higher pseudo
transition barrier compared to NiCo2O4
2 Methods and Models
DFT calculations were performed within the plane-wavendashpseudopotential scheme as
implemented in the Quantum Espresso package14 Spin polarization was always included
and exchange and correlation terms were described using the gradient corrected Perdew-
Burke-Ernzerhof (PBE)15 functional with the on-site Coulomb repulsion U term on the Fe
and Ni 3 d states We used the values U(Ni) =55 eV and U(Fe)=35 eV and 34 eV for
different Fe sites which were determined from linear response16 Ultrasoft
pseudopotentials17 were employed and the valence electrons included O 2 s 2 p Fe 3 d 4
s and Ni 3 d 4 s states Kinetic energy cutoffs of 50 Ry and 500 Ry were chosen for the
wave functions and augmented density respectively Structural optimizations were
carried out by relaxing all atomic positions until all forces were smaller than 1 times 10 minus3
au
99
Bulk calculations were performed using different unit cells The lattice constant was
determined using the 28-atom primitive cell (Figure 35a) with a 4 times 4 times 3 Monkhorst-
Pack k-point grid to sample the Brillouin zone To model a bulk oxygen vacancy we
used the conventional cubic cell (Figure 35b) with one oxygen atom removed and
sampled the Brillouin zone with a 3 times 3 times 3 k-point grid
As in previous studies of the surfaces of spinel oxides18-20 we modeled the NiFe2O4(001)
surface using symmetric slabs of 11 layers terminated by layers exposing oxygen and
octahedral Fe and Ni sites The slabs were separated by a vacuum region 20 Aring wide To
check the convergence of the slab thickness we calculated the surface oxygen vacancy
formation energy (see definition below) for slabs of different thicknesses we found the
formation energy difference between 11-layer and 13-layer slabs to be less than 001 eV
We considered 1 times 1 square unit cell which corresponds to the conventional cell We
sampled the surface Brillouin zone using a 3 times 3 times 1 k-point grid Adsorption calculations
and defected surface calculations were performed with adsorbed species and surface
defects present on one side only of the slab (Model I) To validate this approach we
performed test calculations with adsorbed species and defects symmetrically present on
both sides of the slab (Model II) We found that the two models predict very similar
results For instance the computed surface oxygen vacancy formation energy at T = 0K is
0342 eV with Model I and 0366 eV with Model II Moreover the value obtained for
Model I changed by less than 001 eV when dipole corrections were included Similarly
water dissociation at the surface vacancy yields an energy gain of 1026 eV with Model I
and 1025 eV with Model II
In order to study the surface phase diagram we computed the formation energy
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) of the surface in contact with various gases at temperature T as
follows21
119864119864119865119865119867119867(119879119879 119901119901119894119894 119899119899119894119894) = 119864119864119890119890119889119889119890119890(119899119899119894119894) minus 119864119864119875119875 minus sum 119899119899119894119894 times 119906119906119894119894(119879119879119901119901119894119894)119889119889119904119904119904119904 (1)
Here 119901119901119894119894 are the partial pressures of the gaseous species (oxygen and water) 119899119899119894119894 are
the corresponding numbers of adsorbeddesorbed molecules onfrom the surface
119864119864119890119890119889119889119890119890(119899119899119894119894) is the total energy of the slab with the adsorbeddesorbed species at 0 K (the
100
T-dependence of the chemical potential of the slab can be considered negligible) 119864119864119875119875 is
the total energy of the clean pristine (defect-free) surface and 119906119906119894119894(119879119879119901119901119894119894) is the chemical
potential of gas species i at temperature T and pressure 119901119901119894119894 For the latter the expression
given in Ref21 was used The zero point energy (ZPE) contribution of molecular water
and adsorbed water and hydroxide groups was included in the corresponding chemical
potential and total energies
As a special case of Eq (1) the formation energy of an oxygen vacancy at 0 K and
oxygen partial pressure of 1 atm was calculated using the expression 119864119864119865119865119867119867 = 119864119864119907119907119904119904119909119909 minus
119864119864119899119899119889119889minus119889119889119890119890119891119891 minus 121198641198641198741198742 where 119864119864119907119907119904119904119909119909 and 119864119864119899119899119889119889minus119889119889119890119890119891119891 denote the total energies of the systems with
and without vacancy respectively
Figure 35 Primitive cell (a) and conventional cell (b) of the α-type inverse spinel crystal
structure of NiFe2O4 (c) Local structure of an oxygen vacancy (black sphere) in the
conventional cell with different atoms labeled as in Table 14 Red gray and yellow
spheres represent oxygen Ni and Fe atoms respectively
101
3 Results and Discussion 31 Bulk properties
NiFe2O4 crystallizes in the α type inverse spinel (Figure 35a) which has a tetragonal
P4122P4322 symmetry22 Equal numbers of Ni and Fe atoms occupy octahedral sites
while the remaining Fe atoms occupy tetrahedral sites The conventional cell (Figure
35b) is a radic2 times radic2 times 1 primitive cell having 8 formula units The lattice constant was
determined by fitting the computed total energies to the Birch-Burnagham equation of
state We obtained a value of 845 Aring which is about 14 larger than the experimental
value of 833 Aring23
The computed DOS for the defect-free crystal (Figure 36a) shows an overall band gap of
10 eV The band gap is 15 eV for the majority bands (spin up) and 16 eV for the
minority bands in good agreement with the experimental value of the optical band gap of
16 eV24 Also in agreement with experiment10 24 we find NiFe2O4 to be ferrimagnetic
the octahedral (Oh) and tetrahedral (Td) sites being occupied predominantly by majority
and minority spins respectively The oxidation states of the metal ions computed using
the method in Ref25 are 2+ for Ni and 3+ for Fe
Figure 36 Density of states for (a) defect-free bulk NiFe2O4 (b) defective crystal with
an oxygen vacancy (Figure 35c)
102
As most oxidation reactions on metal oxides occur through a Mars-van Krevelen
mechanism the oxygen vacancy formation energy is usually considered a good
descriptor of the reactivity of these materials26 NiFe2O4 contains two inequivalent
oxygen anions OI bound to 2 Fe and 1 Ni at Oh sites and 1 Fe at Td and OII bound to 1
Fe and 2 Ni at Oh sites and 1 Fe at Td We found that an OI vacancy is slightly preferred
(less costly) with respect to an OII one Still the computed formation energy 287 eV is
rather high indicating that one may need high vacuum and high temperatures to create
oxygen vacancies in bulk NiFe2O4 For comparison we also computed the oxygen
vacancy formation energy in Co3O4 another well-studied spinel oxide using DFT+U
with U = 44 and 66 eV for Co ions at Td and Oh sites respectively27 The resulting
value 288 eV is very similar to that found for NiFe2O4
Table 14 Displacements of the atoms close to a bulk O-vacancy (Figure 235c) with
respect to their positions in the defect-free crystal (Figure 35b) Atoms are labelled as in
Figure 35c Only the atoms closest to the vacancy are considered
Displacement (Aring)
x y Z
O2 0071 -0044 -0048
O11 0008 0044 0045
O13 -0041 0005 0053
O14 0051 0012 -0047
O18 0057 0054 -0001
O19 -0033 -0011 -0004
O25 0077 0055 0075
O31 -0025 -0052 0056
Fe2 0281 -0242 0273
103
Fe10 -0029 0021 -0009
Fe13 -0036 0021 -0033
Ni6 -0010 0060 -0017
The displacements of several atoms around the oxygen vacancy are reported in Table 14
Fe2 the Fe(Td) cation closest to the oxygen vacancy (which was originally bonded to the
removed O atom) undergoes a large displacement 046 Aring with respect to its position in
the defect-free crystal Large displacements of about 01 Aring are present also for a few
oxygens (O2 O25 and O31) bonded to Fe2 The electronic DOS for the defective crystal
is shown in Figure 36b We can see that the DOS for the majority spin states is little
affected by the O-vacancy whereas additional bands are present for the minority spin
states in particular a new occupied state above the original minority band valence band
maximum This results in a smaller minority band gap of ~ 10 eV At the same time the
Fermi levels moves up in the band gap consistent with the fact that the O-vacancy is an
electron donor Analysis of the spin and charge distribution further shows that one of the
two excess electrons donated by the O-vacancy reduces the oxidation state of the
neighboring Fe2(Td) cation from 3+ to 2+ and at the same time reduces also its
magnetic moment The other excess electron is shared by the three neighboring Fe10
Fe13 and Ni6 cations at Oh sites and slightly reduces their magnetic moments Since Td
and Oh sites have different spin states the total magnetization remains unchanged in the
presence of the O-vacancy
104
32 NiFe2O4 (001) surface 321 Defect-free surface
The optimized structure of the NiFe2O4(001) surface is shown in Figure 37 while the
atomic displacements relative to the positions of the bulk-terminated surface are reported
in Table 15 We can see significant outward displacements of the atoms in the first three
layers in comparison to the positions of the bulk-terminated surface Particularly large
(gt01Aring) displacements along the [001] direction are present for Fe2(Td) in the second
layer and for O2 O3 O4 in the top three layers Large in plane displacements are also
present for O2 and O3 the surface oxygens that are not bonded to Fe2
Figure 37 Structure of the NiFe2O4 (001) surface (a) top view of the top three layers
and (b) side view Various O Fe and Ni atoms are indicated
105
Table 15 Displacements of the atoms in the first three layers of the relaxed (001) surface
relative to their positions at the bulk-terminated surface
Displacement Aring in plane (001) direction
O1 0098 0026
O2 0185 0126
O3 0185 0080
O4 0085 0108
Ni1 0026 0024
Ni2 0017 0062
Fe1 0049 0036
Fe2 0042 0196
Fe3 0062 0082
The DOS (Figure 38a) for the defect-free NiFe2O4(001) surface shows the formation of
surface states in the band gap of the majority spin DOS which make the surface metallic
An analogous result was found for the Co3O4(110) surface18 By contrast the minority
spin band is almost unaffected by the presence of the surface and remains very similar to
the minority spin band in the bulk The computed work function was determined from
the analysis of the electrostatic potential profile and found to have a value of about 6
eV
106
Figure 38 DOS for (a) clean defect-free NiFe2O4(001) (denoted as P) (b) clean surface
with 1 oxygen vacancyunit cell (P + 1VO) (c) hydroxylated surface resulting from the
adsorption of 1 water moleculeunitcell on the surface in (b) (P + 1VO + 1H2O) (d) fully
water-covered surface resulting from the adsorption of 6 water moleculesunit cell on the
surface with 25 concentration of oxygen vacancies (P + 2VO + 6H2O)
322 Surface O vacancy
To determine the preferred structure of NiFe2O4(001) we studied the formation of 1 and
2 surface oxygen vacancies (VOrsquos) per unit cell corresponding to surface VO
concentrations of 18 (Figure 39a) and frac14 (Figure 39b) respectively The formation of 1
VO unit cell has an energy cost of 034 eV at 0 K and 1atm O2 pressure The most
favorable site for VO formation is the O3 site ie the oxygen that bonds to two Ni cations
(Ni1 and Ni2) and one Fe (Fe1) By comparing to the bulk formation energy of 287 eV
it is clear that creating an oxygen vacancy at the surface is much easier than in the bulk
This remains true also at higher VO concentrations even though the VO formation energy
107
increases significantly with increasing concentration The formation of 2 oxygen
vacancies per unit cell (Figure 39b) has indeed an energy cost of 143 eV which
corresponds to an average formation energy of 071 eV per vacancy The two oxygen
vacancies prefer to form both at O3 sites and all O3 anions are removed by forming 2
oxygen vacancies unit cell Formation of the second oxygen vacancy at O2 which binds
to 2 Fe(Oh) site and 1 Ni(Oh) has a slightly higher energy cost than at the O3 site
whereas O1 is much harder to remove Since O1 binds to Fe2(Td) removing O1 would
indeed result in a undercoordinated Fe(Td) which is much more unstable
Figure 39 Top views of the first three layers of the NiFe2O4(001) surface in the presence
of (a) 1 O-vacancy (P+ 1VO) (b) 2 O-vacancies (P+ 2VO) The oxygen vacancies are
indicated by the black spheres Atoms are labelled as in Figure 37
The DOS for the surface with 18 VO coverage is shown in Figure 38b We can see a
change in the majority spin surface state band which results in the opening of a narrow
band gap of ~01 eV at the Fermi level On the other hand the minority spin band does
not change significantly except for a small increase by 01 eV of the band gap The
108
computed work function 59 eV remains almost unchanged relative to the one for the
pristine surface
33 Water adsorption 331 Water adsorption on the defect-free surface
The adsorption structures of water on the pristine NiFe2O4(001) surface are found to vary
significantly with coverage At frac14 mono-layer (ML) coverage corresponding to 1 water
molecule per surface cell (Figure 40a) water adsorbs in molecular form on a Ni cation
whereas it dissociates on Fe resulting in an OH on top of the Fe ion and an H donated to
a surface oxygen The adsorption energy on Ni is 053 eV which is about 01 eV more
favorable that on Fe At frac12 ML coverage (Figure 40b) the two water molecules prefer to
adsorb both in molecular form one on a Ni site and one on a Fe site with an adsorption
energy of 044 eVH2O At frac34 ML coverage (Figure 40c) the most stable configuration
corresponds to two water molecules adsorbed in molecular form on Ni sites and one
dissociated water on a Fe site and the adsorption energy is 058 eV H2O Finally the
preferred configuration at full water coverage has all the 4 water molecules adsorbed in
molecular form (Figure 40d) two on Ni one on Fe and one forming an H-bond with an
O site with average adsorption energy of 052 eV H2O A mixed molecular-dissociated
structure with two intact and two dissociated water molecules adsorbed on Ni and Fe
sites respectively was found to be slightly higher in energy
109
Figure 40 Phase diagram of pristine NiFe2O4 (001) exposed to water vapour The two
vertical lines in cyan indicate the region of water chemical potential corresponding to
liquid water (300 K ndash 400 K) The side panels show the structures of adsorbed water at
different coverages (top views) (a) frac14 ML (b) frac12 ML (c) frac34 ML (d) 1 ML Reported
adsorption energies include the change of zero point energies
The phase diagram for the pristine surface exposed to water is shown in Figure 40 This
was obtained by considering the dependence of the computed surface formation energies
on the water chemical potential 12054912054911990611990611986711986721198741198741198791198791199011199011198671198672119874119874 = 1199061199061198671198672119874119874119879119879 1199011199011198671198672119874119874 minus 1198641198641198671198672119874119874 where EH2O
denotes the total energy (including ZPE) of a water molecule at T=0K This diagram
predicts that all water desorbs from the pristine surface at temperatures above ~ 300K
332 Water adsorption on the defected surface
Figure 41 shows some adsorption structures of water on defected NiFe2O4(001) surfaces
at various coverages From the reported adsorption energies (Eads) we can see that water
adsorption is much more favorable on the defected surface than on the pristine surface
110
(Figure 40) In the presence of a surface oxygen vacancy (Figure 39a structure P+1Vo)
a water molecule dissociates on the VO giving rise to two surface hydroxyls with Eads =
090 eV (Figure 41a structure P+1Vo+1H2O) By adsorbing three additional water
molecules to this structure the water molecule adsorbed on the Fe site dissociates into
OH groups while the other two molecules at Ni sites remain intact (Figure 41b
P+1Vo+4H2O) The average adsorption energy per molecule is 070 eV which is smaller
than the value for a single water molecule in Figure 41a On the surface with two oxygen
vacancies (P+2Vo) the configuration with two water molecules adsorbed dissociatively
on the two VOrsquos gives the highest adsorption energy 123eV molecule (structure
P+2Vo+2H2O Figure 41c) When two additional water molecules are adsorbed on this
surface one prefers to adsorb in molecular form on a Ni site while the other is
dissociatively adsorbed on a Fe site(Figure 41d P+2Vo+4H2O) The adsorption energy
086 eVmolecule is lower compared to Figure 41c but still quite higher than water
adsorbed on the pristine surface Finally the addition of two further water molecules
leads to a configuration where all metal sites are covered by adsorbed water (Figure 41e
P+2Vo+6H2O) The two added molecules dissociate on Fe1 sites and the average
adsorption energy is 072 eVmolecule Altogether it appears that mixed molecular-
dissociated configurations are favored at high coverages Water dissociation takes place
both at oxygen vacancies and at Fe sites whereas adsorption in molecular form is
preferred at Ni sites
111
Figure 41 Adsorption structures (top views) of water on defected NiFe2O4(001) surfaces
at different coverages (a) one water molecule adsorbed to a surface with 1 oxygen
vacancy (P+1VO+1H2O) (b) four water molecules adsorbed to a surface containing 1
oxygen vacancy (P+1VO+4H2O) (c) two water molecules adsorbed to two oxygen
vacancies (P+2VO+2H2O) (d) four water molecules adsorbed to a surface containing 2
oxygen vacancies (P+2VO+4H2O) (e) six water molecules adsorbed to a surface with two
oxygen vacancies (P+2VO+6H2O) All structures shown refer to a single surface cell
Reported adsorption energies per molecule include ZPE corrections
The DOS for the surface with an adsorbed water molecule at an oxygen vacancy site
(Figure 38c) is very similar to that of the bare surface with the VO (Figure 38b) Analysis
of the surface metal oxidation states shows that also these oxidation states remain
unchanged upon water adsorption Altogether this indicates that the influence of
adsorbed water on the surface electronic structure is rather limited Support for this
conclusion is also provided by the DOS for the fully hydrated surface (Figure 41e) which
is shown in Figure 38d Comparison to the DOS in Figure 38b and 38c for the clean
defective and hydroxylated surfaces indicates that the main effect of the adsorbed water
112
is the presence of additional bands for both spin states in the band gap near the Fermi
level
Figure 42 Phase diagram of NiFe2O4(001) exposed to H2O and O2 as a function of the
relative chemical potentials 120549120549119906119906119894119894(119879119879119901119901119894119894) = 119906119906119894119894(119879119879119901119901119894119894) minus 119864119864119894119894 referred to the their values at
T=0K The two horizontal lines in cyan indicate the region of water chemical potential
corresponding to liquid water P in the phase diagram indicates the pristine (001) surface
34 Phase diagram
In order to characterize the stability of the (001) surface in O2 and water vapour
environment we examined about 30 surface configurations without and with adsorbed
water and used them to determine the stability diagram of the NiFe2O4 (001) surface in
O2 and water vapor environment The resulting diagram is shown in Figure 42 We
113
identified eight favored structures in the relevant range of water and oxygen chemical
potentials Among these two structures ie P+1VO and P+2VO+2H2O are more
prominent In a wide range of conditions including ambient conditions the P+2VO+2H2O
structure corresponding to a surface with ~ 25 of hydroxyls is predicted to occur At
higher temperatures water desorbs and a dry surface with oxygen vacancies (~ 12
concentration) becomes more favorable The latter structure may be the one which is
typically present during catalytic reactions like the WGS and the CO oxidation reactions
35 CO oxidation
The interaction of CO with the (001) surface of NiFe2O4 (NFO) is weaker in comparison
to that with Co3O4 and NCO (001) surfaces We found indeed an average CO adsorption
energy of -011 and -042 eV on Fe and Ni sites of NFO (001) to be compared to -031
and -088 eV on Ni and Co sites of NCO (100)(001) and -092 eV on Co sites of
Co3O4(001) thus NFO tends to have inert cation-CO interaction On the other hand
NFO(001) has the lowest computed VO formation energy among these materials which
has important consequences on the COOR as identical reaction path is found compared
with NCO As shown in Table 16 the free energy change for the step of CO2 formation (I
II) of the COOR is in fact significantly more favorable on NFO(001) than on Co3O4
(100) and NCO (001)(100) surfaces with similar reaction pathway (Figure 43) while the
step of O2 adsorption at a VO (III IV) is less favorable on NFO(001) in comparison to
Co3O4(100) and NCO (001)(100) This could be a reason why NFO is a less efficient
COOR catalyst than Co3O4 and NCO
114
Table 166 Computed free energy changes (in eV) for the various steps of the COOR on
pristine (P) and defected (D) NFO (001) surfaces Results are reported at both T= 0 K and
T = 300K (between parentheses) with CO and O2 pressure of 1 and 02 atm respectively
I II II III III IV IV V V I
P ndash (001) -302(-250) 010(-067) -022(037) -354(-302) 015(-062)
D ndash (001) -228(-175) 010(-068) -040(019) -390(-337) -006(-083)
Figure 43 Pathway of formation of CO2 (I rarr II) on NFO (001) surface from initial state
(IS structure I) to transition state (TS) and final state (FS structure II) The relative
energy (in eV) with respect to initial structure is display below the structures It shows
identical mechanism compared with NCO (001) surfaces with minor different being the
energy of transition barrier
As the NFO (001) surface tends to be hydroxylated in humid environment we also
investigated the influence of surface hydroxyls on CO adsorption We used the
P+2VO+2H2O structure where all O1 sites are changed into OH groups which represents
the stable surface in a wide range of temperatures and found that CO tends to remain at
115
least 25 Aring away from the surface This indicates only weak interaction between the
hydroxylated surface and gaseous CO which may be another reason of the low COOR
activity of NFO
36 Methane oxidation
As on the NCO (100)mix surface CH4 can only physisorb on the NFO (001)surface as
the molecule remains more than 28 Aring away from the surface in fact the computed
adsorption energy is only -009 eV The barrier of the first C-H bond breaking was first
estimated by the energy of the configuration where hydrogen is adsorbed on the surface
and the methyl radical is in gas phase In this way we obtained a pseudo barrier of 058
eV slightly larger than on NCO (100)mix indicating that the NFO (001) surface is slightly
less active compared with NCO (100)mix surface even though oxygen vacancies are easier
to form on NFO(001) We also determined the barrier with more accurate NEB
calculations which always gave transition states with a gas phase methyl radical like
structure (as used for the preliminary estimate) However the transition barrier obtained
with NEB is much larger 113 eV indicating an under estimation by around 055 eV
with the pseudo barrier approximation To determine the energy of the first C-H bond
breaking we examined configurations where both the methyl and hydrogen were
adsorbed on the surface Results are comparable to those for the NCO (100)mix surface
Methyl adsorbed on surface Fe and Ni results in a formation energy of -027 eV and -031
eV respectively Methyl on surface O1 sites results in a formation energy of -074 eV and
-244 eV respectively for hydrogen adsorbed on the same and different O1 sites NCO
(100)mix surface is likely to be more active when compared with NFO (001) surfaces due
to slightly lower C-H bond breaking barrier
As found for NCO (100)mix the second C-H bond breaking leads to an adsorbed CH2
bridging O1 and another surface site In the most favorable configuration with CH2
bridging two O1 sites the formation energy is -119 eV whereas CH2 bridging O1 and
O2 sites is much less favored with a formation energy of -032 eV All other structures
116
including CH2 bridging metal sites or forming an adsorbed CH2O species on VO are
energetically unfavored
4 Summary and Conclusions
We have studied the atomic structure electronic properties and reactivity of the bulk and
(001) surface of NiFe2O4 using the PBE+U method Our results show that unlike in the
bulk oxygen vacancies form quite easily on the nickel ferrite surface especially at
oxygen sites that are coordinated mainly to Ni ions Our results also indicate that
dissociative adsorption of water at vacancy sites is much more favorable than adsorption
at regular surface sites thus suggesting that a humid environment may help the creation of
oxygen vacancies From our computed surface phase diagram we infer that the
NiFe2O4(001) is hydroxylated at ambient conditions while water desorption should lead
to a defective surface containing a significant fraction of oxygen vacancies at higher
temperature
It is also interesting to notice that our computed phase diagram for NiFe2O4(001) in
Figure 42 is significantly different from that for the Fe3O4 (001) surface20 exposed to
water and oxygen despite the similarity of the two surfaces In comparison to
NiFe2O4(001) the Fe3O4 (001) surface shows a stronger tendency to adsorb water
whereas formation of an oxygen vacancy appears to be much more difficult CO
oxidation through the Mars-Van Krevelen mechanism and breaking of the methane C-H
bond on NiFe2O4 (001) are found to proceed through reaction pathways that are
qualitatively similar but energetically less favorable than those on the NCO (100)mix
surface
117
5 References
1 Brabers V A M Chapter 3 Progress in Spinel Ferrite Research 1995 8 189-
324
2 Han D-H Luo H-L Yang Z Remanent and Anisotropic Switching Field
Distribution of Platelike Ba-Ferrite and Acicular Particulate Recording Media Journal of
Magnetism and Magnetic Materials 1996 161 376-378
3 Cunningham C H Arai T Yang P C McConnell M V Pauly J M
Conolly S M Positive Contrast Magnetic Resonance Imaging of Cells Labeled with
Magnetic Nanoparticles Magnetic Resonance in Medicine 2005 53 999-1005
4 Worledge D C Geballe T H Magnetoresistive Double Spin Filter Tunnel
Junction Journal of Applied Physics 2000 88 5277
5 Hu G Suzuki Y Negative Spin Polarization of Fe3o4 in MagnetiteManganite-
Based Junctions Physical Review Letters 2002 89
6 Hong D Yamada Y Nagatomi T Takai Y Fukuzumi S Catalysis of
Nickel Ferrite for Photocatalytic Water Oxidation Using [Ru(Bpy)3]2+and S2o82ndash
Journal of the American Chemical Society 2012 134 19572-19575
7 Reddy G K Gunasekera K Boolchand P Dong J Smirniotis P G High
Temperature Water Gas Shift Reaction over Nanocrystalline Copper Codoped-Modified
Ferrites The Journal of Physical Chemistry C 2011 115 7586-7595
8 Zhou C Zhang Q Chen L Han B Ni G Wu J Garg D Cheng H
Density Functional Theory Study of Water Dissociative Chemisorption on the
Fe3o4(111) Surface The Journal of Physical Chemistry C 2010 114 21405-21410
9 Sun Q-C Sims H Mazumdar D Ma J X Holinsworth B S OrsquoNeal K
R Kim G Butler W H Gupta A Musfeldt J L Optical Band Gap Hierarchy in a
Magnetic Oxide Electronic Structure of Nife_2O_4 Physical Review B 2012 86
10 Šepelaacutek V Baabe D Mienert D Schultze D Krumeich F Litterst F J
Becker K D Evolution of Structure and Magnetic Properties with Annealing
Temperature in Nanoscale High-Energy-Milled Nickel Ferrite Journal of Magnetism and
Magnetic Materials 2003 257 377-386
118
11 Lders U Bibes M Bobo J F Fontcuberta J Tuning the Growth
Orientation of Nife2o4 Films by Appropriate Underlayer Selection Applied Physics A
2004 80 427-431
12 Klewe C Meinert M Boehnke A Kuepper K Arenholz E Gupta A
Schmalhorst J M Kuschel T Reiss G Physical Characteristics and Cation
Distribution of Nife2o4 Thin Films with High Resistivity Prepared by Reactive Co-
Sputtering Journal of Applied Physics 2014 115 123903
13 Kumar P V Short M P Yip S Yildiz B Grossman J C High Surface
Reactivity and Water Adsorption on Nife2o4(111) Surfaces The Journal of Physical
Chemistry C 2013 117 5678-5683
14 Giannozzi P et al Quantum Espresso A Modular and Open-Source Software
Project for Quantum Simulations of Materials J Phys Condens Matter 2009 21
395502
15 Perdew J P Burke K Ernzerhof M Generalized Gradient Approximation
Made Simple Phys Rev Lett 1996 77 3865-3868
16 Cococcioni M de Gironcoli S Linear Response Approach to the Calculation of
the Effective Interaction Parameters in the Lda+U Method Physical Review B 2005 71
17 Vanderbilt D Soft Self-Consistent Pseudopotentials in a Generalized Eigenvalue
Formalism Phys Rev B 1990 41 7892-7895
18 Chen J Selloni A Electronic States and Magnetic Structure at the Co3o4(110)
Surface A First-Principles Study Physical Review B 2012 85
19 Mulakaluri N Pentcheva R Scheffler M Coverage-Dependent Adsorption
Mode of Water on Fe3o4(001) Insights from First Principles Calculations The Journal
of Physical Chemistry C 2010 114 11148-11156
20 Mulakaluri N Pentcheva R Wieland M Moritz W Scheffler M Partial
Dissociation of Water on Fe_3O_4(001) Adsorbate Induced Charge and Orbital
Order Physical Review Letters 2009 103
21 Reuter K Scheffler M Composition and Structure of Theruo2(110)Surface in
Ano2and Co Environment Implications for the Catalytic Formation Ofco2 Physical
Review B 2003 68
119
22 Ivanov V G Abrashev M V Iliev M N Gospodinov M M Meen J
Aroyo M I Short-Range B-Site Ordering in the Inverse Spinel Ferrite Nife_2O_4
Physical Review B 2010 82
23 Baltzer P K White J G Crystallographic and Magnetic Studies of the System
(Nife2o4)1minusX + (Nimn2o4)X Journal of Applied Physics 1958 29 445
24 Perron H Mellier T Domain C Roques J Simoni E Drot R Catalette
H Structural Investigation and Electronic Properties of the Nickel Ferrite Nife2o4 A
Periodic Density Functional Theory Approach Journal of Physics Condensed Matter
2007 19 346219
25 Sit P H L Car R Cohen M H Selloni A Simple Unambiguous
Theoretical Approach to Oxidation State Determination Via First-Principles Calculations
Inorganic Chemistry 2011 50 10259-10267
26 McFarland E W Metiu H Catalysis by Doped Oxides Chemical Reviews
2013 113 4391-4427
27 Chen J Wu X Selloni A Electronic Structure and Bonding Properties of
Cobalt Oxide in the Spinel Structure Physical Review B 2011 83