Post on 03-Jul-2020
transcript
UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
Solid ‘oxygen reservoirs’ for selective hydrogen oxidation
Beckers, J.
Link to publication
Citation for published version (APA):Beckers, J. (2009). Solid ‘oxygen reservoirs’ for selective hydrogen oxidation.
General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.
Download date: 21 Jul 2020
ABOUT THE AUTHORBefore starting his PhD at the University of Amsterdam in 2005, Jurriaan Beckers (Reimerswaal, 1972) obtained his bachelor's degree at the Hogeschool Zeeland in Vlissingen and worked at the Agrotechnologisch Onderzoeksinstituut (ATO-DLO) in Wageningen, Lyckeby Starch in Kristianstad, Sweden, GE Plastics in Bergen op Zoom and at the University of Amsterdam. Jurriaan currently lives in Amsterdam with his wife and nine guitars. His hobbies are music and writing. On Sunday afternoon you can often fi nd him at the Crea Café in Amsterdam.
Solid 'oxygen reservoirs' for selective hydrogen oxidation
Jurriaan BeckersISBN/EAN:978-90-9024476-1Cover design: S.J. de Vet
Solid 'oxygen reservoirs' for selective hydrogen oxidation Jurriaan Beckers 2009
Solid 'oxygen reservoirs' for
selective hydrogen oxidation
UITNODIGING
Voor het bijwonen van de openbare verdediging van
mijn proefschrift:
op dinsdag 22 septemberom 12.00 uur
in de Agnietenkapel van deUniversiteit van AmsterdamOudezijds Voorburgwal 231
1012 EZ Amsterdam
na afl oop bent u van hartewelkom bij de receptie
ter plaatse
Paranimfen:Arjen Boogaard
a.boogaard@ewi.utwente.nl
Lars van der Zandelars.vanderzande@shell.com
Jurriaan BeckersMuntendamstraat 421091DV Amsterdam
020-4122624jurriaanbeckers@wanadoo.nl
Solid ‘oxygen reservoirs’ for
selective hydrogen oxidation
Jurriaan Beckers
Solid ‘oxygen reservoirs’ for
selective hydrogen oxidation
ACADEMISCH PROEFSCHRIFT
ter verkrijging van de graad van doctor
aan de Universiteit van Amsterdam,
op gezag van de Rector Magnificus
prof. dr. D.C. van den Boom
ten overstaan van een door het college voor promoties ingestelde commissie,
in het openbaar verdedigd in de Agnietenkapel
op dinsdag 22 september 2009 om 12 uur
door
Jurriaan Beckers
geboren te Reimerswaal
Promotiecommissie
Promotor: Prof. dr. G. Rothenberg
Promotor: Prof. dr. C.J. Elsevier
Overige leden: Prof. dr. K.J. Hellingwerf
Prof. dr. F. Kapteijn
Prof. dr. ir. B.M. Weckhuysen
Dr. A.F. Lee
Dr. M. Ruitenbeek
Dr. G. Zwanenburg
Faculteit der Natuurwetenschappen, Wiskunde en informatica
The research reported in this thesis was carried out at the Van 't Hoff Institute for
Molecular Sciences, Faculty of Science, University of Amsterdam (Nieuwe
Achtergracht 166, 1018 WV Amsterdam), with financial support of the Advanced
Sustainable Processes by Engaging Catalytic Technologies (ASPECT) programme,
part of the Advanced Chemical Technologies for Sustainability (ACTS) platform
of the Netherlands Organisation for Scientific Research (NWO).
Opgedragen aan mijn grootvader Drs. Hubert Maria Beckers
Contents
Chapter 1 Introduction
1.1 General introduction 11
1.2 What are we dealing with? Some properties and pitfalls 12
1.3 Oxidative dehydrogenation (ODH) using ceria based materials 16
1.3.1 Oxidative dehydrogenation of ethane 17
1.3.2 Oxidative dehydrogenation of propane 21
1.3.3 Oxidative dehydrogenation of other hydrocarbons 24
1.3.4 Combined dehydrogenation and selective hydrogen combustion 26
Chapter 2 Selective hydrogen oxidation reactions using solid ‘oxygen reservoirs’
2.1 Ceria-based selective hydrogen oxidation catalysts via genetic
algorithms 37
2.2 Perovskites as solid oxygen reservoirs for selective hydrogen oxidation 77
2.3 Lead-containing solid oxygen reservoirs 99
Chapter 3 Characterisation of solid oxygen reservoirs
3.1 Redox kinetics of ceria-based catalysts 119
3.2 Redox properties of doped and supported copper-ceria catalysts 145
3.3 The optimisation and characterisation of bismuth doped ceria catalysts 173
3.4 Particle size and dopant concentration effects on the catalytic
properties of ceria based solid oxygen reservoirs
199
Summary 237
Samenvatting 241
List of publications 245
List of abbreviations 249
Dankwoord 251
Appendix I Success of doping 257
Appendix II Catalyst activity 267
9
Chapter 1
Introduction
This work has been submitted to a peer-reviewed journal.
10
Chapter 1 Introduction
11
1.1 General introduction Cerium was discovered in 1803, by both Jöns Jakob Berzelius and
Wilhelm Hisinger in Sweden, and Martin Heinrich Klaproth in Germany.[1] It was
named after the dwarf planet Ceres, discovered in 1801, which was in turn named
after the Roman goddess of agriculture (particularly the growth of cereals).[2]
Cerium is part of the ceric or light rare-earth elements, and is the main component
of the Bastnasite (USA, China), Monazite (Australia) and Loparite (Russia) rare-
earth minerals.[3] Cerium or cerium-oxide (‘ceria’) is used in various applications,
such as the removal of free oxygen and sulphur from the melt in the casting of iron,
as a polishing agent for glass and for the decolourisation of glass.[3, 4] Due to their
good ionic conduction, ceria-based materials can also be applied as electrolytes in
solid oxide fuel cells.[5] One of the most successful industrial applications of ceria
is as oxygen storage material in automotive three way (catalytic) converters
(TWC).[6-9] Plain ceria has been used from the eighties onwards, and, in 2003,
TWC sales accounted for one quarter of the global catalytic market.[10] The
successful application of ceria in TWCs is due to its temperature stability and its
facile Ce3+ Ce4+ + e– redox reaction, allowing the ceria to easily store and
release oxygen.[11] In the catalytic converter, this aids the hydrocarbon combustion
in the fuel-rich mode, and the NO reduction in the fuel-lean mode. The application
of ceria-based materials in TWCs has been reviewed by several authors.[7, 10, 12] An
excellent review on the physical and (electro-) chemical properties of ceria-based
oxides was published by Mogensen et al. in 2001.[13] The group of Trovarelli has
performed a lot of work on the (redox) chemistry and catalysis of ceria and ceria-
based materials,[11, 14, 15] and a book on the subject was published in 2002.[3]
The TWC is related to combustion engines running on conventional fuel,
with the ceria aiding full combustion and NO reduction. In the last ten years,
however, the research focus has shifted to alternative power sources, such as fuel
cells, and alternative fuels, such as hydrogen. Furthermore, alternatives for crude
oil for the production of (fine) chemicals are sought, such as the production of
syngas from (bio-)methane, and the conversion of these to larger molecules by the
Fischer-Tropsch process. Interestingly, in this field ceria-based materials have also
gained a lot of attention, but contrary to the TWC, most recent literature focuses on
selective oxidation applications rather than on full combustion. Many of these
Chapter 1 Introduction
12
recent papers deal with selective oxidation of CO from CO/hydrogen (preferential
oxidation, PROX), often using copper-ceria, to clean up the hydrogen used in fuel
cells.[16-18] The CO removal is needed since it poisons the Pt-catalyst at the low
operating temperatures used. Secondly, in the search for alternative fuels and
chemical building blocks, the generation of syngas by partial oxidation of methane
(POM), or methane steam reforming has gained a lot of attention.[19-22] Ceria-based
materials have also been used for selective oxidations of small molecules such as
H2S to S or H2 to H2O2, and for selective oxidations of various hydrocarbons.[23-28]
In view of the topic of this thesis, I will discuss the performance of ceria-based
materials in one type of selective oxidation, namely oxidative dehydrogenation
(ODH). This yields valuable small alkenes such as ethene and propene. These can
be used either as building blocks for various chemicals or high purity monomers
for plastics such as polypropene and polyethene.
Ceria-based materials come in many forms. Ceria can act as support for
‘active metals’, but at low concentrations these metals can also be doped in the
ceria's fluorite lattice, forming ‘solid solutions’. Ceria can also form ‘mixed oxides’
with other metal oxides, drastically changing its catalytic properties. Importantly,
the ceria's facile redox cycle often results in a strong metal-support interaction
(SMSI), which can have a pronounced effect on the catalysis. I therefore start by
explaining this interaction, as well as outlining some basic properties relevant to
catalysis, before discussing ceria's role in ODH.
1.2 What are we dealing with? Some properties and pitfalls There are various routes for synthesising ceria,[3] but the simplest is by
calcining ceriumnitrate, Ce(NO3)3.[29, 30] At about 65 °C the ceriumnitrate melts,
followed by dehydration and, from about 200 °C onwards, nitrate
decomposition.[31, 32] No further weight loss occurs above 400 °C, and this
temperature is sufficient to form the ceria fluorite structure.[32] At these low
calcination temperatures, the ceria's crystallite size is small, and the surface area
high. Ceria catalysts are often used at higher temperatures, however, and increasing
the calcination temperature will increase the crystallite size, decreasing the specific
surface area. Typical values obtained when preparing ceria by calcining
ceriumnitrate are a crystallite size of 10 nm and surface area of 85 m2/g when
Chapter 1 Introduction
13
calcining at 550 °C, and a crystallite size of 25 nm and surface area of 30-50 m2/g
when calcining at 700 °C.[30, 33] High surface area cerias can be obtained by
applying sol-gel or surfactant assisted synthesis methods.[34-36] With these
techniques, surface areas ranging from about 125 - 230 m2/g can be achieved, at
calcination temperatures of 800 °C and 450 °C, respectively. The sintering
behaviour of ceria is dependent on the gas atmosphere. Practically, this means that
when a ceria catalyst is used in a reducing atmosphere, it can still sinter, even at
temperatures below its calcination temperature. The sintering behaviour of ceria
under various atmospheres is shown in Figure 1. All data was obtained starting
from the same batch of ceria.[37]
0
20
40
60
80
100
120
140
400 500 600 700 800 900
Temperature (°C)
Sur
face
are
a (m
2 /g)
H2
Vacuum
CO2H2O
CO
0
20
40
60
80
100
120
140
400 500 600 700 800 900
Temperature (°C)
Sur
face
are
a (m
2 /g)
H2
Vacuum
CO2H2O
CO
Figure 1. The specific surface area of plain ceria when heated in various gasses (at
atmospheric pressure) and under vacuum.[37] All data is obtained starting from the same
batch of ceria. Data of the treatment in air is not added for clarity, but their trend is similar
to those obtained in vacuum. Reproduced with permission of the author and the publisher.
The high temperatures encountered in automotive catalysis, typically 1000-
1100 °C, sparked the search for ceria-based materials with a higher temperature
stability. Ceria-zirconia mixed oxides have higher thermal stability and excellent
redox behaviour.[10, 33, 38-40] Indeed, ceria-based materials are versatile since the
ceria can be used not only as support for active metals, but these metals can also be
doped into the ceria lattice itself, or form ceria containing mixed oxides.
Incorporation of dopant atoms in the ceria bulk allows for tuning the oxygen
Chapter 1 Introduction
14
conduction, the electronic conduction, and with them the catalytic properties.[13, 41,
42] Importantly, the distinction between an active metal supported on ceria or doped
into the ceria lattice is not always clear, especially at elevated temperatures and/or
in the presence of reducing gasses. The facile redox of the ceria can result in strong
metal-support interaction when ceria is used as support.[43-45] Phase segregation or a
change in phase composition can occur for ceria-based solid solutions and mixed
oxides.[39, 46] Thus, the active site can change during catalysis. For example, when
nickel supported on ceria is reduced at 750 °C, the nickel crystallites can spread
over the reduced ceria support (see Scheme 1).[43] Conversely, at too high doping
levels or temperatures, a dopant can segregate from the ceria as a separate oxide, or
the catalyst's surface can be enriched in dopant atoms.[47-50] Importantly, the
spreading of nickel crystallites over the ceria surface in case of a ceria support, and
the surface enrichment in case of the doped ceria could lead to similar surface
structures. These effects also complicate catalyst characterisation. In case of XRD,
for example, the absence of dopant oxide crystals does not prove dopant
incorporation in the bulk. Indeed, several groups observed that when impregnating
copper on ceria supports, no copper oxide phases were detected, provided that
copper loading and calcination temperatures were kept low.[51] Lattice doping can
be demonstrated by using XRD, EXAFS or EPR.[48, 52-55] This does not exclude,
however, that the surface of the material, where catalysis takes place, is enriched
with the dopant. A small amount of dopant or dopant oxide clusters, or a dopant
enriched surface phase will not be detected by bulk techniques. Surface sensitive
techniques such as LEIS and XPS can detect surface enrichment, and in case of
XPS, the oxidation states of surface components.[49] Their signal however, is still
the average of the entire catalyst surface, which can complicate things if multiple
types of surface species are present.
Chapter 1 Introduction
15
O Ce NiO Ce NiO Ce Ni
Scheme 1. Proposed spreading of nickel supported on ceria upon reduction to
750 °C, drawn after Gonzalez-DelaCruz.[43]
Another type of strong metal-support interaction is the so-called
‘decoration’ of Pd, Pt and Rh by ceria observed upon reduction to 600-700 °C.
Contrary to nickel, the supported metal crystals stay intact, but are decorated with a
layer of ceria upon the high-temperature reduction, shielding the noble metal's
surface and thereby affecting the catalysis (see Figure 2).[44] Note that this shielding
effect can not be observed by, for example, XRD.
Chapter 1 Introduction
16
Figure 2. Metal decoration of 4 wt% Pt supported on ceria upon reduction in
hydrogen at 700 °C.[44] Reproduced with permission of the author and the publisher.
1.3 Oxidative dehydrogenation (ODH) using ceria-based materials The demand for small alkenes is high. Propene demand, for example, is
expected to rise to 80 million tonnes in 2010 worldwide.[56-58] The main routes to
propene are steam cracking, fluid catalytic cracking, and catalytic dehydrogenation.
All these processes are endothermic. Advantageous on-demand production of
alkenes is achieved by catalytic dehydrogenation, but it is equilibrium limited and
deactivation of the catalyst occurs due to coking (the formation of a carbon-rich
solid on the catalyst's surface).[59-61] Oxidative dehydrogenation (ODH), where
oxygen or an oxygen containing molecule such as N2O or CO2 is added to the gas
feed, can overcome these limitations, allowing exothermic, non equilibrium-
limited, and on-demand production of the alkenes (see Scheme 2).[61] Furthermore,
the addition of oxygen limits catalyst deactivation due to coking. It can, however,
result in over-oxidation of the hydrocarbons to CO and CO2.[62] This is a big
challenge, since the alkene product is more reactive than the alkane starting
material. Thus, specific ODH catalysts have to be developed, and up till now,
(supported) vanadium or molybdenum oxides have gained most attention.[61] For
ethane ODH, yields comparable to those of steam cracking are obtained, but
propane ODH yields are still far from being interesting for industry. In both cases,
little is reported on catalyst lifetime.[61] In the search for better ODH catalysts,
Chapter 1 Introduction
17
various ceria-based materials have also been investigated. The results are
summarised in the following sections.
Dehydrogenation C3H8 ↔ C3H6 + H2
ΔHo460 °C = +130 kJ/mol
Oxidative dehydrogenation C3H8 + ½ O2 ↔ C3H6 + H2O
ΔHo420 °C = –117 kJ/mol
Scheme 2. Propane dehydrogenation (top)[63] and oxidative propane
dehydrogenation (bottom)[64]
1.3.1 Oxidative dehydrogenation of ethane Table 1 gives the catalyst composition and catalytic performance of ceria-
based materials as catalysts for ethane ODH (note that some data is taken from
figures, not tables). Doped ceria's (or solid solutions, where the crystal structure of
the ceria remains unchanged by the addition of the dopant), or mixed oxides
containing metal M and ceria are denoted as ‘Ce–M–O’. Metal M supported on
ceria is denoted as ‘M/CeO2’. The catalysts generally give a high selectivity
towards ethene at low conversion, and lower selectivities at high conversion. I have
therefore incorporated the data of the highest selectivity, highest activity, and
highest overall performance in the table, where available. The activity and
selectivity data in the table is presented graphically in Figure 3.
Chapter 1 Introduction
18
V 4
Ca 8
V 5
Sr 1
V 6
Ca 9
CeO2 10
Sr 2
CeO2 7
Sr 3
0
25
50
75
100
25 50 75 100
Activity (% ethane conversion)
Sel
ectiv
ity to
war
ds e
then
e (%
)
11
V 4
Ca 8
V 5
Sr 1
V 6
Ca 9
CeO2 10
Sr 2
CeO2 7
Sr 3
0
25
50
75
100
25 50 75 100
Activity (% ethane conversion)
Sel
ectiv
ity to
war
ds e
then
e (%
)
11
Figure 3. Activity and selectivity in ethane ODH. The labels show the type of
metal added to the ceria, and the catalyst number. A Mo-V-Te-Nb-mixed oxide catalysts
(11) is added for reference.
The data show that the best catalysts, achieving highest selectivity and
activity, are the calcium doped 9, the supported strontium 2 and 3 and the plain
ceria 10. Interestingly, most of these catalysts were tested under special conditions:
steam was added in case of 2, increasing the activity by dilution and the selectivity
by prevention of coking, and the ODH was performed with CO2 instead of oxygen
in case of 9 and 10.[65-67] The ODH with CO2 was performed at rather high
temperatures (≥750 °C), which is higher than the temperature needed for the
(endothermic) catalytic dehydrogenation.[68] The vanadium-containing catalysts 4–
6 run below 600 °C, but also with lower activity and selectivity. The supported Sr 3
has rather high selectivity and activity at 660 °C, running without addition of
steam, nor using CO2 as the oxidant.[69] Their performance falls short, however,
compared to the best reference catalyst 11, a V and Mo containing mixed oxide,
which also operates at lower temperature (400 °C) and without steam and/or CO2.
(see Table 1 and Figure 3). Note however, that catalyst 11 is one of three catalysts,
out of 70 tested, which displays this good performance,[61] and that the other 67
catalysts have substantially lower activity and selectivity. Far less research has
been performed on ceria-based materials. No data is available, for example, for
molybdenum-containing ceria-based materials, although the data on Ce-V-O shows
Chapter 1 Introduction
19
that using components of the benchmark reference catalysts does not guarantee
good performance in ceria-based materials (vide infra). Indeed, in a pre-screening
for a combinatorial catalysis approach, Ni-Ce-Nb and Ni-Ce-Ta mixed oxides were
found to be the most interesting leads, outperforming the best Mo-V-Nb oxide in
ethane ODH.[70] Up till now, no Ce-Ni-O or Ce-(Ni, Nb, Ta)-O catalyst was tested
for the ethane ODH.
Interestingly, operando studies with vanadia supported on ceria show that
this system is highly interactive, as was seen for nickel and noble metals supported
on ceria.[71-73] Starting from vanadia supported on ceria, the vanadia interacts
strongly with the ceria, eventually forming a CeVO4 phase. Surprisingly, this
process does not affect the selectivity and activation energy in the ethane ODH.[73]
Possibly, the active phase consists of Ce3+-O-V5+, which is present in both
supported vanadia and the CeVO4. This also explains why the catalytic properties
of Ce-V-O materials differ from other vanadium containing ODH catalysts (which
are the ODH benchmark).
Cha
pter
1
Int
rodu
ctio
n
20
T
able
1. C
eria
-bas
ed m
ater
ials
use
d fo
r et
hane
OD
H.
Cat
alys
t
num
ber
Cat
alys
t com
p.
Con
cent
ratio
n
adde
d m
etal
A
lkan
e:O
2[a]
Spa
ce
velo
city
(ml/
g.h)
[b]
Tem
pera
ture
(°C
)
Eth
ane
conv
ersi
on
(%)
Sel
ectiv
ity
tow
ards
eth
ene
(%)
Ref
eren
ce
1 S
r/C
eO2
10 m
ol%
6:
1 (m
olar
) 10
200
700
18
56[c
] [7
4]
2
80
0 50
88
[c]
3 S
rCl 2
/CeO
2 30
mol
%
2:1
6000
66
0 73
69
[7
5]
4 V
/CeO
2 3
wt%
1:
2 90
000
510
1 69
[7
1-73
]
5
3 w
t%
590
9 38
6
1 w
t%
590
19
20
7 C
eO2 re
f
550[d
] 66
4
Usi
ng C
O2
inst
ead
of o
xyge
n:
8 C
e-C
a-O
10
mol
%
1:2
(CO
2)
1200
0 65
0 3
98[e
] [6
8]
9
75
0 25
90
10
CeO
2 re
f.
75
0 41
71
Non
-cer
ia-b
ased
ref
eren
ce c
atal
yst:
11
Mo-
V-T
e-N
b 1:
0.2:
0.17
:0.1
7 9:
6 10
0 40
0 87
84
[7
6]
[a] C
O2
whe
n ap
plic
able
. [b] R
eact
ions
wer
e pe
rfor
med
at a
tmos
pher
ic p
ress
ure.
[c] S
team
was
add
ed a
t H2O
:C2H
6 =
1:1
mol
ar r
atio
. [d
] Sam
e pe
rfor
man
ce a
t 51
0 an
d 59
0 °C
. [e] T
his
is t
he v
alue
obt
aine
d af
ter
pre-
trea
ting
the
cata
lyst
at
750
°C u
nder
the
rea
ctio
n co
nditi
ons.
The
fre
sh c
atal
yst h
as a
sel
ecti
vity
of
~55%
, and
the
incr
ease
in s
elec
tivi
ty is
irre
vers
ible
.
Chapter 1 Introduction
21
1.3.2 Oxidative dehydrogenation of propane The success of nickel-containing catalysts in propane and isobutane ODH,
has led to the testing of nickel-containing ceria-based catalysts in propane ODH.[77,
78] The group of Barbaux showed that using nickel-containing ceria-based catalysts,
ODH can be performed at lower temperatures, as compared to when the other
nickel catalysts are used (300 °C).[77] Upon comparing ceria-nickel mixed oxides
with nickel supported on ceria, it was found that the supported catalysts gave the
highest selectivity, but low conversion (see Table 2, 1). The mixed oxides give
higher conversion, but lower selectivity (2 and 3, all data taken at 300 °C). Nickel-
containing mixed oxides were found to be superior in yield as compared to mixed
oxides with either Cr, Co, Cu or Zn (catalysts 4, 5, and 6).[79] Note that the mixed
oxides from both studies, at the same composition and reaction temperature, differ
strongly in activity and selectivity (compare 2 and 5, note that both space velocity
and conversion of 5 are higher). The preparation methods of the catalysts are very
similar, except for the calcination temperature (700 °C for 2, 500 °C for 5). As was
the case for ethane ODH, the supported vanadium catalysts perform less well than
other ceria-based mixed oxides (catalysts 7-9, Table 2).[78]
Contrary to the ethane ODH experiments shown in Table 1, the propane
ODH is performed at temperatures well below that of the catalytic propane
dehydrogenation, albeit at rather low conversion. No data is available on the use of
other oxidants, such as CO2, but the selectivity of plain ceria is seen to increase
substantially when adding trichloromethane gas (10-12).[80] It is worthwhile to
investigate the effect of halogen addition on the catalyst performance in propane
ODH via a more practically applicable route, such as using supported metal-
halogens, as was done for the ethane ODH.[75] As was the case in ethane ODH, the
non-ceria-based reference catalyst (13) outperforms the ceria-based ones. But
again, much more non-ceria-based catalysts have been tested, and the large
majority of these perform less well than 13.
Chapter 1 Introduction
22
Ni 3
Ni 4
Ni 5 Ni 6
V 7
V 8
V 9CeO2 10
CeO2 TCM 12
13
Ni(K) 1
Ni 2CeO2 TCM 11
25
50
75
100
0 25 50 75
Activity (% propane conversion)
Sel
ectiv
ity to
war
ds p
rope
ne (
%)
Ni 3
Ni 4
Ni 5 Ni 6
V 7
V 8
V 9CeO2 10
CeO2 TCM 12
13
Ni(K) 1
Ni 2CeO2 TCM 11
25
50
75
100
0 25 50 75
Activity (% propane conversion)
Sel
ectiv
ity to
war
ds p
rope
ne (
%)
Figure 4. Activity and selectivity in propane ODH. The labels show the type of
metal added to the ceria, and the catalyst number. A supported vanadia catalyst (13) is
added for reference.
Cha
pter
1
Int
rodu
ctio
n
23
T
able
2. C
eria
-bas
ed m
ater
ials
use
d fo
r pr
opan
e O
DH
.
Cat
alys
t
num
ber
Cat
alys
t
com
p.
Con
cent
rati
on
adde
d m
etal
A
lkan
e:O
2
Alk
ane,
O2
conc
.
(%v/
v)
Spac
e
velo
city
(ml/
g.h)
[a]
Tem
pera
ture
(°C
)
Pro
pane
conv
ersi
on
(%)
Sel
ecti
vity
tow
ards
prop
ene
(%)
Ref
.
1 N
i-K
/CeO
2 N
i:C
e =
1,
K:N
i =0.
05[b
] 1:
2 4,
8
545
300
8 72
[7
7]
2 C
e-N
i-O
N
i:C
e =
0.5[b
]
15
58
3 C
e-N
i-O
N
i:C
e =
1[b]
19
60
4 C
e-N
i-O
[c,d
] N
i:C
e =
0.5[b
] 1:
3 5,
15
3000
0 20
0 2
50
[79]
5
300
25
12
6
375
62
10
7 V
/CeO
2 12
wt%
1:
3 5,
15
5000
30
0 2
85
[78]
8
6 w
t%
30
0 14
34
9
6 w
t%
40
0 24
20
10
CeO
2
1:1
14, 1
3[e]
3600
45
0 17
5
[80]
11
CeO
2+T
CM
17
% v
/v T
CM
[f]
1:1
14, 1
3[e]
23
52
12
CeO
2+T
CM
17
% v
/v T
CM
3.
5:1
14, 4
[e]
17
70[g
]
Non
-cer
ia-b
ased
ref
eren
ce c
atal
yst:
13
V/M
CF
[h]
4.2
wt %
1:
1 10
, 10
7200
0 55
0 31
84
[8
1]
[a] R
eact
ions
wer
e pe
rfor
med
at
atm
osph
eric
pre
ssur
e. [
b] T
hese
are
ato
mic
rat
ios.
[c]
Ref
eren
ce m
easu
rem
ents
wer
e pe
rfor
med
on
ceri
a an
d ni
ckel
oxi
de.
Cer
ia:
3% c
onve
rsio
n, 2
% s
elec
tivi
ty (
300
°C),
10%
con
vers
ion,
6%
, se
lect
ivit
y (4
00 °
C,
note
: qu
ite
clos
e to
ent
ry 1
0).
NiO
10%
con
vers
ion,
17%
,
sele
ctiv
ity
(350
°C
). [
d] N
i ou
tper
form
s si
mil
ar c
atal
ysts
con
tain
ing
Cr,
Co,
Cu
or Z
n. [
e] T
he c
once
ntra
tion
is
in k
Pa
inst
ead
of %
v/v.
[f] T
CM
sta
nds
for
tric
hlor
omet
hane
. [g]
At t
his
oxyg
en p
ress
ure
the
valu
es f
or p
lain
cer
ia a
re: 7
% c
onve
rsio
n, 1
0% s
elec
tivi
ty. [
h] M
CF
stan
ds f
or M
isoc
ello
us S
ilic
a Fo
ams.
Chapter 1 Introduction
24
1.3.3 Oxidative dehydrogenation of other hydrocarbons Besides ethane and propane, ceria-based materials have been applied in
isobutane and ethylbenzene ODH (see Table 3). High selectivity and conversion, at
temperatures lower than those at which commercial catalysts are used, were
obtained for ethylbenzene ODH over plain ceria, using N2O as oxidant (1).[82] The
high activity was attributed to a high concentration of Ce4+-O--Ce3+ defect sites.
Both doped and supported chromium-ceria catalysts were applied in the isobutane
ODH (see Figure 5).[83, 84] The chromium containing catalysts show better results
than plain ceria. The activity of chromium supported on ceria is higher than plain
chromium oxide and the chromium ceria mixed oxide (this is the case at both
270 °C and 300 °C). Conversely, the selectivity of the chromium supported on
ceria is somewhat lower as compared to the ceria chromium mixed oxides and
plain Cr2O3. In case of the chromium-ceria systems, well dispersed Cr6+-Ox, and
not Cr2O3 aggregates, was proposed as the active site, which was poisoned by the
presence of potassium.[84]
Cr/CeO2 2
Cr/CeO2 3
Ce-Cr-O 4
Ce-Cr-O 5
CeO2 6 CeO2 7
Cr2O3 8
Cr2O3 9
25
50
75
100
0 5 10 15 20 25
Activity (% isobutane conversion)
Sel
ectiv
ity to
war
ds is
obut
ene
(%)
Cr/CeO2 2
Cr/CeO2 3
Ce-Cr-O 4
Ce-Cr-O 5
CeO2 6 CeO2 7
Cr2O3 8
Cr2O3 9
25
50
75
100
0 5 10 15 20 25
Activity (% isobutane conversion)
Sel
ectiv
ity to
war
ds is
obut
ene
(%)
Figure 5. Isobutane ODH over chromium-ceria catalysts at 270 °C (open circles)
and 300 °C (full circles). The labels show the type of metal added to the ceria, and the
catalyst number. Plain ceria and Cr2O3 are added for reference.
Cha
pter
1
Int
rodu
ctio
n
25
T
able
3. C
eria
-bas
ed m
ater
ials
use
d fo
r et
hylb
enze
ne a
nd is
obut
ane
OD
H.
Cat
alys
t
num
ber
Cat
alys
t
com
p.
Con
cent
ratio
n ad
ded
met
al
Rea
ctan
t
Spa
ce
velo
city
(ml/
g.h)
Tem
pera
ture
(°C
)
Alk
ane
conv
ersi
on
(%)
Sel
ectiv
ity
tow
ards
alke
ne (
%)
Ref
.
1 C
eO2
E
thyl
benz
ene[a
] -
325
45
94
[82]
2 C
r/C
eO2
5-40
Cr
atom
s / n
m2
Isob
utan
e[b]
- 27
0 10
54
[8
3]
3
30
0 20
36
4 C
e-C
r-O
C
r:C
e =
0.2
-1.8
[c]
270
5 60
5
Cr:
Ce
= 0
.2-1
.8[c
]
30
0 10
45
6 C
eO2
27
0 5
20
7
30
0 10
20
8 C
r 2O
3
270
5 68
9
30
0 15
40
[a] N
2O w
as u
sed
inst
ead
of O
2. N
o da
ta o
n th
e re
acta
nt c
once
ntra
tion
s is
giv
en. [b
] The
isob
utan
e : O
2 ra
tio is
1 :
1, a
t 6.5
% v
/v.
[c] T
hese
are
(bu
lk)
atom
ic r
atio
s.
Chapter 1 Introduction
26
1.3.4 Combined dehydrogenation and selective hydrogen
combustion
Another type of ODH has been industrially implemented. Here, the
dehydrogenation is performed over a conventional dehydrogenation catalyst, and a
second catalyst is added to selectively combust part of the hydrogen formed. The
process may be therefore viewed as ‘two-step ODH’ (Scheme 3). The selective
hydrogen combustion generates heat and shifts the equilibrium to the products side,
yielding the same benefits as the ‘conventional’ ODH. The use of two catalysts or
two reactors, however, allows for separate tuning of the dehydrogenation and the
hydrogen oxidation reactions, and the advantage of this over conventional ODH is
proven by the industrial implementation of the process. For example, the STAR
(Steam Active Reforming) oxydehydrogenation process is implemented in two
plants for the ODH of isobutane to MTBE. In this process, an oxydehydrogenation
reactor is placed after a STAR-dehydrogenation reactor, both using the same
catalyst.[85] The SMART process (Styrene Monomer Advanced Reheat
Technology) is in operation in five plants for the ODH of ethylbenzene to styrene.
This process uses two catalysts in one reactor (the selective hydrogen combustion
catalyst is Pt-based).[86, 87] In both processes, steam is added to minimise coking
and dilute the feed. These processes use a co-fed approach, where small amounts of
oxygen are added to the gas feed (Scheme 4, left). The mixing of gaseous oxygen
with hydrogen and hydrocarbons at elevated temperatures is, however, a safety
risk, which is avoided in the redox-mode (Scheme 4, right). Here, no gaseous
oxygen is added, but the lattice oxygen of the selective hydrogen combustion
catalyst is used. At one point in time, however, the lattice oxygen is depleted and
has to be refilled. That is, the reactor has to be purged and an oxygen containing
feed has to be applied to the catalyst bed, resulting in a cyclic process. Note that the
conventional DH catalyst also has to be regenerated periodically to burn off the
coke accumulated on its surface. In redox mode ODH of ethane and propane, high
selectivities towards hydrogen combustion can be obtained for several supported
oxides (e.g. Sb2O4, In2O3, WO3, PbO and Bi2O3).[88-91] These are, however, unstable
under the high temperature redox cycling. The melting point of most of these
metals lies below the operating temperature, and when the supported metal oxide is
reduced to metal(0), it liquefies, causing sintering and deactivation. Conversely,
Chapter 1 Introduction
27
ceria is stable under the redox cycling conditions, and has a good oxygen storage
capacity. The selectivity of plain ceria is low, but in a screening experiment,
working with hydrogen/ethane/ethene mixtures, we showed that doping the ceria
lattice can overcome both the problems of low selectivity and low stability.[92] In
this thesis, we investigated the fundaments of this oxidative dehydrogenation
process.
Energy H2
H2O + Ce2O3 2 CeO2
Dehydrogenationcatalyst
Propane Propene
Energy H2
H2O + Ce2O3 2 CeO2
Dehydrogenationcatalyst
Propane Propene
Scheme 3. Catalytic cycle for redox mode oxidative dehydrogenation using ceria
as solid oxygen reservoir.
Chapter 1 Introduction
28
CnH(2n+2)
A
N2O2N2
B C D
N2O2COx
N2CnH2nH2O
Fresh SHC
Fresh DH
Spent SHC
Spent DH
ReoxidationReduction
CnH(2n+2) + O2
CnH2nH2O
Co-fed process Redox process
Purge Purge
CnH(2n+2)
A
N2O2N2
B C D
N2O2COx
N2CnH2nH2O
Fresh SHC
Fresh DH
Spent SHC
Spent DH
ReoxidationReduction
CnH(2n+2) + O2
CnH2nH2O
Co-fed process Redox process
Purge Purge
Scheme 4. Left: scheme of a co-fed mode oxidative dehydrogenation process.
Right: scheme of a redox-mode dehydrogenation process (SHC: selective hydrogen
combustion catalyst). After the dehydrogenation step A, the bed is flushed with nitrogen
(B), and the catalysts are regenerated through reoxidation (C). This burns coke from the
dehydrogenation catalyst and restores the lattice oxygen of the selective hydrogen
combustion catalyst. After another nitrogen flush (D) the reactor is ready for the next redox
cycle.
Chapter 1 Introduction
29
References [1] J. Emsley, The Elements, Oxford University Press, Oxford, 1989. [2] G. Houtzager, Geïllustreerde Griekse Mythologie Encyclopedie, Rebo productions
b.v., Lisse, 2003. [3] A. Trovarelli, Catalysis by ceria and related materials, Imperial College Press,
London, 2002. [4] T. C. Schutt, Ceram. Bull., 1972, 51, 155. [5] B. Zhu and M. D. Mat, Int. J. Electrochem. Sci., 2006, 1, 383. [6] S. Bernal, G. Blanco, J. J. Calvino, J. A. P. Omil and J. M. Pintado, J. Alloy.
Compd., 2006, 408, 496. [7] J. Kašpar, P. Fornasiero and M. Graziani, Catal. Today, 1999, 50, 285. [8] J. Kašpar, P. Fornasiero and N. Hickey, Catal. Today, 2003, 77, 419. [9] R. Di Monte and J. Kašpar, Top. Catal., 2004, 28, 47. [10] J. Kašpar and P. Fornasiero, J. Solid State Chem., 2003, 171, 19. [11] A. Trovarelli, C. de Leitenburg, M. Boaro and G. Dolcetti, Catal. Today, 1999, 50,
353. [12] M. Boaro, M. Vicario, C. de Leitenburg, G. Dolcetti and A. Trovarelli, Catal.
Today, 2003, 77, 407. [13] M. Mogensen, N. M. Sammes and G. A. Tompsett, Solid State Ionics, 2000, 129,
63. [14] E. Aneggi, M. Boaro, C. de Leitenburg, G. Dolcetti and A. Trovarelli, J. Alloy
Compd., 2006, 408, 1096. [15] E. Aneggi, C. de Leitenburg, G. Dolcetti and A. Trovarelli, Catal. Today, 2006,
114, 40. [16] A. Gómez-Cortés, Y. Márquez, J. Arenas-Alatorre and G. Díaz, Catal. Today,
2008, 133, 743. [17] D. Gamarra, C. Belver, M. Fernández-Garcia and A. Martínez-Arias, J. Am. Chem.
Soc., 2007, 129, 12064. [18] E. Y. Ko, E. D. Park, K. W. Seo, H. C. Lee, D. Lee and S. Kim, Catal. Today,
2006, 116, 377. [19] S. A. Larrondo, A. Kodjaian, I. Fabregas, M. G. Zimicz, D. G. Lamas, B. E. W. de
Reca and N. E. Amadeo, Int. J. Hydrogen Energ., 2008, 33, 3607. [20] M. Salazar, D. A. Berry, T. H. Gardner, D. Shekhawat and D. Floyd, Appl. Catal.
A: Gen., 2006, 310, 54. [21] T. J. Huang and M. C. Huang, Chem. Eng. J., 2008, 145, 149. [22] J. H. Xu, C. M. Y. Yeung, J. Ni, F. Meunier, N. Acerbi, M. Fowles and S. C.
Tsang, Appl. Catal. A: Gen., 2008, 345, 119. [23] S. Yasyerli, G. Dogu and T. Dogu, Catal. Today, 2006, 117, 271. [24] V. R. Choudhary, A. G. Gaikwad and S. D. Sansare, Catal. Lett., 2002, 83, 235.
Chapter 1 Introduction
30
[25] D. I. Enache, D. W. Knight and G. J. Hutchings, Catal. Lett., 2005, 103, 43. [26] J. Barbier Jr., L. Oliviero, B. Renard and D. Duprez, Top. Catal., 2005, 33, 77. [27] S. Demirel, P. Kern, M. Lucas and P. Claus, Catal. Today, 2007, 122, 292. [28] M. W. Xue, X. D. Gu, J. P. Chen, H. L. Zhang and J. Y. Shen, Thermochim. Acta,
2005, 434, 50. [29] N. Milinski, B. Ribar and M. Sataric, Cryst. Struct. Commun., 1980, 9, 473. [30] J. Beckers, F. Clerc, J. H. Blank and G. Rothenberg, Adv. Synth. Catal., 2008, 350,
2237. [31] P. F. Campbell, M. H. Ortner and C. J. Anderson, Anal. Chem., 1961, 33, 58. [32] B. Zhu, X. G. Liu, M. T. Sun, S. J. Ji and J. C. Sun, Solid State Sci., 2003, 5, 1127. [33] C. de Leitenburg, A. Trovarelli, J. Llorca, F. Cavani and G. Bini, Appl. Catal. A:
Gen., 1996, 139, 161. [34] S. H. Overbury, D. R. Huntley, D. R. Mullins and G. N. Glavee, Catal. Lett., 1998,
51, 133. [35] D. Terribile, A. Trovarelli, J. Llorca, C. de Leitenburg and G. Dolcetti, Catal.
Today, 1998, 43, 79. [36] D. Terribile, A. Trovarelli, J. Llorca, C. de Leitenburg and G. Dolcetti, J. Catal.,
1998, 178, 299. [37] V. Perrichon, A. Laachir, S. Abouarnadasse, O. Touret and G. Blanchard, Appl.
Catal. A: Gen., 1995, 129, 69. [38] S. Bernal, G. Blanco, J. J. Calvino, J. C. Hernández, J. A. P. Omil, J. M. Pintado
and M. R. Yeste, J. Alloy. Compd., 2008, 451, 521. [39] R. T. Baker, S. Bernal, G. Blanco, A. M. Cordón, J. M. Pintado, J. M. Rodríguez-
Izquierdo, F. Fally and V. Perrichon, Chem. Commun., 1999, 149. [40] R. Di Monte and J. Kašpar, J. Mater. Chem., 2005, 15, 633. [41] A. E. C. Palmqvist, M. Wirde, U. Gelius and M. Muhammed, Nanostruct. Mater.,
1999, 11, 995. [42] D. A. Andersson, S. I. Simak, N. V. Skorodumova, I. A. Abrikosov and B.
Johansson, Proc. Nat. Acad. Sci. USA, 2006, 103, 3518. [43] V. M. Gonzalez-DelaCruz, J. P. Holgado, R. Pereñíguez and A. Caballero, J.
Catal., 2008, 257, 307. [44] S. Bernal, J. J. Calvino, M. A. Cauqui, J. M. Gatica, C. Larese, J. A. P. Omil and J.
M. Pintado, Catal. Today, 1999, 50, 175. [45] J. Beckers and G. Rothenberg, Dalton Trans., 2008, 6573. [46] X. C. Zheng, X. L. Zhang, X. Y. Wang, S. R. Wang and S. H. Wu, Appl. Catal. A:
Gen., 2005, 295, 142. [47] M. Jobbágy, F. Mariño, B. Schöbrod, G. Baronetti and M. Laborde, Chem. Mater.,
2006, 18, 1945. [48] P. Bera, K. R. Priolkar, P. R. Sarode, M. S. Hegde, S. Emura, R. Kumashiro and
N. P. Lalla, Chem. Mater., 2002, 14, 3591.
Chapter 1 Introduction
31
[49] P. J. Scanlon, R. A. M. Bink, F. P. F. van Berkel, G. M. Christie, L. J. van IJzendoorn, H. H. Brongersma and R. G. van Welzenis, Solid State Ionics, 1998, 112, 123.
[50] A. Y. C. Tschöpe, Mater. Sci. Eng., A, 1995, 204, 267. [51] Tang et al. observed no CuO phase for a sample of 5 wt% Cu on ceria, calcined at
500 °C, with a final surface area of 80 m2/g (X. L. Tang, B. C. Zhang, Y. Li, Y. D. Xu, Q. Xin and W.J. Shen, Appl. Catal. A: Gen., 2005, 288, 116). The loading of 5 wt% Cu corresponds to 12.5 mol% Cu, if a doped ceria would have been prepared. This gives an indication of the capacity of ceria to support copper species which cannot be detected by XRD. Other groups reported similar results, see: X.C. Zheng, X. L. Zhang, X. Y. Wang, S. R. Wang and S. H. Wu, Appl. Catal. A: Gen., 2005, 295, 142; G. Jacobs, E. Chenu, P.M. Patterson, L. Williams, D. Sparks, G. Thomas and B.H. Davis, Appl. Catal. A: Gen., 2004, 258, 203; X.Y. Jiang, L.P. Lou, Y.X. Chen and X.M. Zheng, J. Mol. Catal. A: Chem., 2003, 197, 193).
[52] T. S. Zhang, J. Ma, L. H. Luo and S. H. Chan, J. Alloy. Compd., 2006, 422, 46. [53] S. Hočevar, U. O. Krašovec, B. Orel, A. S. Aricó and H. Kim, Appl. Catal. B:
Environ., 2000, 28, 113. [54] C. Lamonier, A. Bennani, A. D'Huysser, A. Aboukais and G. Wrobel, J. Chem.
Soc., Faraday Trans., 1996, 92, 131. [55] B. Murugan, A. V. Ramaswamy, D. Srinivas, C. S. Gopinath and V. Ramaswamy,
Chem. Mater., 2005, 17, 3983. [56] J. Plotkin and E. Glatzer, Eur. Chem. News, 2005, 82, 20. [57] N. Alperowicz, Chem. Week, 2007, 169, 27. [58] N. Alperowicz, Chem. Week, 2006, 168, 17. [59] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 1. [60] T. A. Nijhuis, S. J. Tinnemans, T. Visser and B. M. Weckhuysen, Chem. Eng. Sci.,
2004, 59, 5487. [61] F. Cavani, N. Ballarini and A. Cericola, Catal. Today, 2007, 127, 113. [62] R. K. Grasselli, Top. Catal., 2002, 21, 79. [63] H. Weyten, K. Keizer, A. Kinoo, J. Luyten and R. Leysen, Aiche J., 1997, 43,
1819. [64] R. Grabowski, Catal. Rev. - Sci. Eng., 2006, 48, 199. [65] R. X. Valenzuela, G. Bueno, V. C. Corberán, Y. D. Xu and C. L. Chen, Catal.
Today, 2000, 61, 43. [66] R. X. Valenzuela, G. Bueno, A. Solbes, F. Sapiña, E. Martínez and V. C.
Corberán, Top. Catal., 2001, 15, 181. [67] M. Guío, J. Prieto and V. C. Corberán, Catal. Today, 2006, 112, 148. [68] V. C. Corberán, Catal. Today, 2005, 99, 33.
Chapter 1 Introduction
32
[69] Note that the increased activity, as compared to the Sr-containing catalysts 1 and 2, could originate from the higher Sr loading (30 mol% instead of 10 mol% for catalysts 1 and 2).
[70] S. Bergh, P. J. Cong, B. Ehnebuske, S. H. Guan, A. Hagemeyer, H. Lin, Y. M. Liu, C. G. Lugmair, H. W. Turner, A. F. Volpe, W. H. Weinberg, L. Woo and J. Zysk, Top. Catal., 2003, 23, 65.
[71] M. V. Martínez-Huerta, J. M. Coronado, M. Fernández-García, A. Iglesias-Juez, G. Deo, J. L. G. Fierro and M. A. Bañares, J. Catal., 2004, 225, 240.
[72] M. V. Martínez-Huerta, G. Deo, J. L. G. Fierro and M. A. Bañares, J. Phys. Chem. C, 2007, 111, 18708.
[73] M. V. Martínez-Huerta, G. Deo, J. L. G. Fierro and M. A. Bañares, J. Phys. Chem. C, 2008, 112, 11441.
[74] V. R. Choudhary, S. A. R. Mulla and V. H. Rane, J. Chem. Technol. Biot., 1998, 71, 167.
[75] H. X. Dai, C. F. Ng and C. T. Au, J. Catal., 2001, 199, 177. [76] P. Botella, E. García-González, A. Dejoz, J. M. L. Nieto, M. I. Vázquez and J.
González-Calbet, J. Catal., 2004, 225, 428. [77] P. Boizumault-Moriceau, A. Pennequin, B. Grzybowska and Y. Barbaux, Appl.
Catal. A: Gen., 2003, 245, 55. [78] W. Daniell, A. Ponchel, S. Kuba, F. Anderle, T. Weingand, D. H. Gregory and H.
Knozinger, Top. Catal., 2002, 20, 65. [79] L. Jalowiecki-Duhamel, A. Ponchel, C. Lamonier, A. D'Huysser and Y. Barbaux,
Langmuir, 2001, 17, 1511. [80] S. Sugiyama, Y. Iizuka, E. Nitta, H. Hayashi and J. B. Moffat, J. Catal., 2000,
189, 233. [81] Y. M. Liu, W. L. Feng, T. C. Li, H. Y. He, W. L. Dai, W. Huang, Y. Cao and K.
N. Fan, J. Catal., 2006, 239, 125. [82] B. Murugan and A. V. Ramaswamy, J. Am. Chem. Soc., 2007, 129, 3062. [83] P. Moriceau, B. Grzybowska, Y. Barbaux, G. Wrobel and G. Hecquet, Appl.
Catal. A: Gen., 1998, 168, 269. [84] P. Moriceau, B. Grzybowska, L. Gengembre and Y. Barbaux, Appl. Catal. A:
Gen., 2000, 199, 73. [85] www.uhde.eu [86] F. M. Dautzenberg, Catal. Rev., 2004, 46, 335. [87] F. M. Dautzenberg and P. J. Angevine, Catal. Today, 2004, 93-95, 3. [88] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 9. [89] L. M. van der Zande, E. A. de Graaf and G. Rothenberg, Adv. Synth. Catal., 2002,
344, 884.
Chapter 1 Introduction
33
[90] E. A. de Graaf, A. Andreini, E. J. M. Hensen and A. Bliek, Appl. Catal. A: Gen., 2004, 262, 201.
[91] E. A. de Graaf, G. Zwanenburg, G. Rothenberg and A. Bliek, Org. Process. Res. Dev., 2005, 9, 397.
[92] G. Rothenberg, E. A. de Graaf and A. Bliek, Angew. Chem., Int. Ed., 2003, 42, 3366.
34
35
Chapter 2
Selective hydrogen oxidation reactions using solid
‘oxygen reservoirs’
Experimental setup used to screen the catalysts performance in the selective hydrogen
oxidation reaction.
36
37
2.1
Ceria-based selective hydrogen oxidation catalysts
via genetic algorithms
Mg
Ca Cr MnTiK Fe Cu
Al
Pd SnRuZrYSr
Ta W Pt Pb Bi
YbSmNdPrLa Ce
Energy H2
H2O Ce1-x-yM1xM2yO2
Dehydrogenationcatalyst
Propane Propene
Gd
Mg
Ca Cr MnTiK Fe Cu
Al
Pd SnRuZrYSr
Ta W Pt Pb Bi
YbSmNdPrLa Ce
Energy H2
H2O Ce1-x-yM1xM2yO2
Dehydrogenationcatalyst
Propane Propene
Gd
Part of this work has been published as:
- 'A “green route” to propene through selective hydrogen oxidation', Jan Hendrik
Blank, Jurriaan Beckers, Paul F. Collignon, Frédéric Clerc, and Gadi Rothenberg,
Chem. Eur. J. 2007, 13, 5121.
- 'Selective Hydrogen Oxidation Catalysts via Genetic Algorithms', Jurriaan
Beckers, Frédéric Clerc, Jan Hendrik Blank, and Gadi Rothenberg, Adv. Synth.
Catal. 2008, 350, 2237.
Chapter 2.1 Catalysis
38
Abstract
Solid ‘oxygen reservoirs’, such as doped ceria, can be successfully applied in a
novel process for propane oxidative dehydrogenation. The ceria lattice oxygen
selectively burns hydrogen from the dehydrogenation mixture at 550 °C. This gives
three key advantages: it shifts the dehydrogenation equilibrium to the desired
products side, generates heat in situ, which aids the endothermic dehydrogenation,
and simplifies product separation. We have applied a genetic algorithm to screen
doped cerias for their performance in the selective hydrogen oxidation. Five
generations of doped ceria catalysts (97 catalysts in total), were synthesised.
Dopants were chosen from a set of 26 elements, and with a maximum of two
dopants per catalyst, at five different concentrations. The catalyst performance
(activity and selectivity), is expressed by a fitness value. The average fitness value
increases from generation one to three, and then stabilises. That is, the system
converges after three generations. The dopant type has a large effect on the catalyst
fitness. We identified six dopant atoms which lead to selective hydrogen
combustion catalysts, namely Bi, Cr, Cu, K, Mn, Pb and Sn (‘good’ dopants).
Analysis of the effect of electronegativity, ionic radius and dopant concentration
shows that most elements yielding a high fitness have an electronegativity in the
range of 1.5–1.9. Generally, the properties of catalysts containing two dopants can
be predicted from the behaviour of singly doped ones. Synergy does occur for
certain copper and iron containing catalysts. The addition of Ca or Mg to Cu doped
catalysts doubles the activity, and the selectivity of iron doped catalysts can be
improved by adding Cr, Mn or Zr. Importantly, the doped cerias show a high
stability in the redox cycling, much higher than that of supported oxides. A Cr and
Zr doped catalyst (Ce0.90Cr0.05Zr0.05O2) was highly selective and active over 250
redox cycles (a total of 148 hours on stream), with no phase segregation or change
in particle size.
Chapter 2.1 Catalysis
39
Introduction Selective oxidation is applied in the production of many important bulk
chemicals and intermediates, such as acrolein, acrylic acid, MTBE, and maleic
anhydride.[1] In these processes, the commercial value of the hydrocarbons is
increased by selective addition of oxygen atoms, and the greatest challenge is to
prevent over-oxidation. Selective oxidation can also, however, remove certain
species by combusting them. This is the case in oxidative dehydrogenation, which
can be used to obtain propene from propane (Scheme 1).[2-9] Propane
dehydrogenation is an endothermic reaction, but oxidative dehydrogenation can
overcome this limitation. Selectively combusting the formed hydrogen into water
generates heat in situ and shifts the equilibrium towards the products side.
Oxidation
Latticerecharged
Propane
Propene + H2O
O2
CO2
Energy H2
H2O + Ce2O3 2 CeO2
Dehydrogenationcatalyst
Propane Propene Oxidation
Latticerecharged
Propane
Propene + H2O
O2
CO2
Energy H2
H2O + Ce2O3 2 CeO2
Dehydrogenationcatalyst
Propane Propene
Scheme 1. Left: the reactions occurring during the combined propane
dehydrogenation and selective hydrogen combustion. The dehydrogenation consumes
energy and yields hydrogen. The hydrogen combustion consumes hydrogen and yields
energy. Right: Cartoon showing a proposed reactor configuration for the redox process,
enabling continuous production of high purity propene. Whilst the propane
dehydrogenation and selective hydrogen combustion are performed in the left hand reactor,
the catalysts in the right hand reactor are being regenerated (coke combustion and refilling
of the lattice oxygen for the solid oxygen reservoir).
Chapter 2.1 Catalysis
40
We recently introduced a new type of oxidative dehydrogenation system,
employing doped cerias as solid oxygen reservoirs (SORs).[10, 11] The
dehydrogenation step is performed over a conventional Pt-Sn-Al2O3 catalyst, and
the hydrogen is combusted using the oxygen of the ceria lattice (Scheme 1, left).[12]
After the ceria is reduced, the lattice oxygen vacancies are re-filled using air,
creating a cyclic redox process (Scheme 1, right) and simultaneously burning off
any coke. This is safer than mixing gaseous O2 and H2 at high temperatures
(typically 500–600 °C). Furthermore, the use of two catalysts allows for separate
tuning of the hydrogen combustion and the dehydrogenation. Supported metal
oxides can also perform this selective hydrogenation, but they sinter under redox
cycling.[13-17] Ceria has high temperature stability and a facile
Ce3+ Ce4+ + e– reaction, making it a good SOR.[18] CeO2 itself, however, is
not selective, but we showed that selectivity, activity and stability can be tuned by
doping the ceria lattice with different cations.[10] In a preliminary screening, we
tested ten catalysts for their selectivity towards hydrogen combustion from a mix
with ethane and ethene, with the dopant type as the only variable. A tungsten-
doped catalyst showed excellent selectivity and stability for this reaction.[10] In this
study, we screen doped ceria catalysts for hydrogen combustion from a mix with
propane and propene. Twenty-six different dopants are used, at five possible
concentrations, with a maximum of two dopants per catalysts. This yields a huge
amount of catalyst candidates. Synthesising and testing all of the combinations is
not practical. Instead, we employ a genetic algorithm (GA), to find the optimal
catalyst using an iterative approach. GAs mimic evolutionary biology in silico.[19-21]
Compared with the almost ubiquitous application of GAs in other scientific fields,
few researchers have used an evolutionary approach for screening heterogeneous
catalysts. Baerns et al. have studied propane oxidative dehydrogenation, using
gaseous oxygen,[22-24] total propane oxidation,[25] and the production of hydrocyanic
acid.[26] Yamada et al. have investigated methanol synthesis,[27-29] and others have
performed studies on (selective) oxidation,[30-33] reduction,[34] methane
reforming,[35] and isomerisation.[36] Most of the catalysts used in these studies are
mixed oxides containing four to five different metals. Kim et al. have used doped
cerias for the reforming of methane.[35]
Chapter 2.1 Catalysis
41
In this paper, we present the results of a genetic algorithm-based catalyst
optimisation. A total of 97 doped-cerias are evaluated for their performance in
selective hydrogen combustion. The evolution of the fitness value over five
generations is evaluated, and the performance of the catalysts is correlated to their
composition.
Results and Discussion Catalyst preparation and characterisation. Catalysts were prepared by
co–melting a mixture of the metal nitrate hydrate precursors (chlorides or
ammonium metallates were used when nitrates were not available).[11, 37] After the
precursor has liquefied, the pressure was lowered and a solid mixed metal nitrate
formed. This was converted into the mixed oxide by calcining in static air at
700 °C for 5 h. The following notation is used: Gn–m, where n is the generation
number and m is the catalyst number. The activity, selectivity and fitness value of
all 97 catalysts are given in Table 1, together with the catalyst composition and
characterisation data. The activity is determined as the percentage of the hydrogen
feed combusted by each catalyst (labelled ‘hydrogen activity’). The fitness function
is defined as: F = [selectivity + (0.2 × activity)]/120 × 100, and ranges from 0–100.
Note that the ‘hydrogen activity’ presented in Table 1 is converted into a 0 – 100
scale prior to the calculation of the fitness function. An activity-value of either 0,
33, 66 or 100 is given to catalysts with hydrogen activities ranging from 0%,
0.1–7%, 7.1–21.2%, or are higher than 21.3%, respectively (these values appear
strange since they are based on a different unit of activity, which we have replaced
with the more informative hydrogen activity). Specific attributes of some of the
catalysts are presented in detail thereafter.
Figure 1 shows pictures of three catalysts after heating at reduced pressure
(top, mixed nitrates), and after calcination (bottom, mixed oxides). Whereas pure
cerium nitrate is white, and CeO2 is pale yellow, the mixed oxides show a variety
of colours. These ceria based mixed oxides have the general catalyst formula
Ce1-x-yM1xM
2yO2. The metals M1 and M2 are added at zero, two, five, eight or ten
mol%, and chosen from 26 candidates (vide infra). Each catalyst was characterised
using powder X-ray diffraction, to ensure it consists of a uniform phase. That is,
Chapter 2.1 Catalysis
42
only the diffraction lines of the cerias fluorite structure are observed, and no
separate oxides of the dopants are present. Importantly, the catalysts were not
prepared by impregnating CeO2 supports. The co-melting of the cerium nitrate with
the nitrates of the appropriate metals yields a well mixed liquid catalyst precursor.
This allows for incorporation of the dopants into the ceria fluorite structure after
calcination. However, dopant enriched surface phases can occur for these type of
catalysts, and cannot be detected by XRD.[38-40] Indeed, in case of a copper-ceria
mixed oxide, Bera et al. observed both surface enrichment and bulk incorporation
of the copper.[40]
Figure 1. Photos of three catalysts (G3–09 Ce0.96Mn0.02Cu0.02O2, G3–10
Ce0.96W0.02Sn0.02O2 and G3–15 Ce0.90Bi0.08Cu0.02O2) after heating under reduced pressure
(top, mixed nitrates) and calcining (bottom, mixed oxides).
Cha
pter
2.1
Cat
alys
is
43
T
able
1. C
ompo
sitio
n, c
hara
cter
isat
ion
data
and
cat
alyt
ic p
erfo
rman
ce o
f do
ped
ceri
as 1
–97.
Cat
alys
t C
ompo
sitio
n[a]
Sur
face
are
a
(m2 /g
)
Cry
stal
lite
siz
e
(nm
)[b]
Lat
tice
para
met
er (
Å)[c
]
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
H2
oxid
atio
n
sele
ctiv
ity (
%)
Fit
ness
valu
e[d]
G1–
01
Ce 0
.87A
l 0.0
8Ta 0
.05O
2 58
n.
d.[e
] n.
d.
0 0
0
G1–
02
Ce 0
.96C
a 0.0
2Sr 0
.02O
2 42
n.
d.
n.d.
0.
5 18
21
G1–
03
Ce 0
.89C
r 0.0
2Fe 0
.09O
2 28
n.
d.
n.d.
9
85
82
G1–
04
Ce 0
.96P
d 0.0
4O2
53
n.d.
n.
d.
0 0
0
G1–
05
Ce 0
.98S
n 0.0
2O2
67
12
5.40
8 8
77
75
G1–
06
Ce 0
.89P
t 0.0
2Mn 0
.09O
2 57
n.
d.
n.d.
0
0 0
G1–
07
Ce 0
.90T
a 0.0
5Ti 0
.05O
2 46
n.
d.
n.d.
0
0 0
G1–
08
Ce 0
.92Z
r 0.0
2Mg 0
.08O
2 26
n.
d.
n.d.
0
0 0
G1–
09
Ce 0
.90N
d 0.1
0O2
61
14
5.43
0 0
0 0
G1–
10
Ce 0
.90Y
b 0.0
8Gd 0
.02O
2 24
n.
d.
n.d.
1
40
39
G1–
11
Ce 0
.90R
u 0.0
5Sm
0.05
O2
65
n.d.
n.
d.
0 0
0
G1–
12
Ce 0
.90Y
0.05
Sr 0
.05O
2 27
n.
d.
n.d.
0.
4 44
42
G1–
13
Ce 0
.90B
i 0.0
5K0.
05O
2 17
n.
d.
n.d.
18
[f]
91
87
G1–
14
Ce 0
.91L
a 0.0
9O2
49
n.d.
n.
d.
0 0
0
G1–
15
Ce 0
.90W
0.10
O2
25
26
5.41
1 0
0 0
G1–
16
Ce 0
.92C
r 0.0
8O2
24
26
5.41
4 15
86
83
G1–
17
Ce 0
.90F
e 0.1
0O2
50
14
5.40
4 0
0 0
G1–
18
Ce 0
.90C
u 0.1
0O2
47
15
5.41
1 7
89
85
G1–
19
Ce 0
.90B
i 0.1
0O2
33
18
5.41
6 33
[f]
77
81
G1–
20
Ce 0
.91M
n 0.0
9O2
56
11
5.40
7 5
93
83
Cha
pter
2.1
Cat
alys
is
44
T
able
1, c
ontin
ued.
Cat
alys
t C
ompo
sitio
n[a]
Sur
face
are
a
(m2 /g
)
Cry
stal
lite
siz
e
(nm
)[b]
Lat
tice
para
met
er (
Å)[c
]
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
H2
oxid
atio
n
sele
ctiv
ity (
%)
Fit
ness
valu
e[d]
G1–
21
Ce 0
.91C
a 0.0
9O2
22
28
5.41
6 0
0 0
G1–
22
Ce 0
.92P
b 0.0
8O2
56
13
5.41
1 46
[f]
92
93
G1–
23
Ce 0
.90P
d 0.1
0O2
72
13
5.41
1 0
0 0
G1–
24
CeO
2 36
[g]
25[g
] 5.
409[g
] 0
0 0
G1–
25
Ce 0
.90Z
r 0.1
0O2
71
n.d.
n.
d.
1 36
36
G2–
01
Ce 0
.90Y
b 0.1
0O2
n.d.
18
5.
406
0 0
0
G2–
02
Ce 0
.86C
a 0.0
9Cu 0
.05O
2 n.
d.
24
5.41
1 18
87
84
G2–
03
Ce 0
.90C
r 0.0
5Bi 0
.05O
2 31
28
5.
412
38[f
] 84
87
G2–
04
Ce 0
.87M
g 0.0
5Cu 0
.08O
2 n.
d.
18
5.40
9 17
87
84
G2–
05
Ce 0
.90M
n 0.0
2Fe 0
.08O
2 n.
d.
16
5.40
5 4
87
78
G2–
06
Ce 0
.84Z
r 0.0
8Cu 0
.08O
2 54
17
5.
411
13
92
88
G2–
07
Ce 0
.87B
i 0.0
8Sn 0
.05O
2 55
14
5.
411
45[f
] 84
87
G2–
08
Ce 0
.96G
d 0.0
2Bi 0
.02O
2 n.
d.
19
5.41
2 7
68
68
G2–
09
Ce 0
.88C
r 0.0
2W0.
10O
2 n.
d.
21
5.40
9 1
36
36
G2–
10
Ce 0
.90Z
r 0.0
2Fe 0
.08O
2 n.
d.
14
5.40
2 2
78
71
G2–
11
Ce 0
.90A
l 0.1
0O2
n.d.
13
5.
408
0 0
0
G2–
12
Ce 0
.90A
l 0.0
2Cu 0
.05O
2 52
14
5.
409
11
89
85
G2–
13
Ce 0
.90K
0.10
O2
n.d.
93
5.
411
3 94
84
G2–
14
Ce 0
.96P
r 0.0
2Gd 0
.02O
2 n.
d.
17
5.41
3 0
0 0
G2–
15
Ce 0
.94M
n 0.0
4Sr 0
.02O
2 n.
d.
13
5.40
9 0
0 0
Cha
pter
2.1
Cat
alys
is
45
T
able
1, c
ontin
ued.
Cat
alys
t C
ompo
sitio
n[a]
Sur
face
are
a
(m2 /g
)
Cry
stal
lite
siz
e
(nm
)[b]
Lat
tice
para
met
er (
Å)[c
]
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
H2
oxid
atio
n
sele
ctiv
ity (
%)
Fit
ness
valu
e[d]
G2–
16
Ce 0
.92T
i 0.0
8O2
n.d.
16
5.
408
0 0
0
G2–
17
Ce 0
.98L
a 0.0
2O2
n.d.
17
5.
415
0 0
0
G2–
18
Ce 0
.96P
r 0.0
2W0.
02O
2 n.
d.
17
5.41
2 0
0 0
G3–
01
Ce 0
.98K
0.02
O2
n.d.
53
5.
411
2 88
79
G3–
02
Ce 0
.98Z
r 0.0
2O2
n.d.
21
5.
414
1 65
60
G3–
03
Ce 0
.98P
r 0.0
2O2
n.d.
18
5.
411
0.3
50
47
G3–
04
Ce 0
.96M
n 0.0
4O2
59
13
5.40
8 4
85
76
G3–
05
Ce 0
.98A
l 0.0
2O2
n.d.
13
5.
407
1 35
35
G3–
06
Ce 0
.93A
l 0.0
2Yb 0
.05O
2 n.
d.
12
5.40
8 1
41
40
G3–
07
Ce 0
.96Z
r 0.0
2Cu 0
.02O
2 56
16
5.
409
6 95
85
G3–
08
Ce 0
.96L
a 0.0
2Bi 0
.02O
2 n.
d.
17
5.41
8 9
87
84
G3–
09
Ce 0
.96M
n 0.0
2Cu 0
.02O
2 n.
d.
13
5.40
7 5
75
68
G3–
10
Ce 0
.96W
0.02
Sn 0
.02O
2 n.
d.
17
5.40
2 6
92
82
G3–
11
Ce 0
.96G
d 0.0
5O2
n.d.
20
5.
414
1 66
61
G3–
12
Ce 0
.98M
n 0.0
2O2
60
14
5.40
8 3
95
85
G3–
13
Ce 0
.98R
u 0.0
2O2
n.d.
16
5.
408
0 0
0
G3–
14
Ce 0
.90Y
0.05
Fe 0
.05O
2 n.
d.
14
5.40
7 0
0 0
G3–
15
Ce 0
.90B
i 0.0
8Cu 0
.02O
2 28
25
5.
415
29[f
] 83
86
G3–
16
Ce 0
.98P
t 0.0
2O2
n.d.
16
5.
411
0 0
0
G3–
17
Ce 0
.90C
r 0.0
5Zr 0
.05O
2 29
22
5.
405
9 95
90
Cha
pter
2.1
Cat
alys
is
46
T
able
1, c
ontin
ued.
Cat
alys
t C
ompo
sitio
n[a]
Sur
face
are
a
(m2 /g
)
Cry
stal
lite
siz
e
(nm
)[b]
Lat
tice
para
met
er (
Å)[c
]
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
H2
oxid
atio
n
sele
ctiv
ity (
%)
Fit
ness
valu
e[d]
G3–
18
Ce 0
.96S
n 0.0
2Pd 0
.02O
2 n.
d.
16
5.41
1 0
0 0
G4–
01
Ce 0
.98N
d 0.0
2O2
n.d.
19
5.
414
1 74
67
G4–
02
Ce 0
.98Y
0.02
O2
n.d.
22
5.
410
2 72
66
G4–
03
Ce 0
.98S
r 0.0
2O2
n.d.
20
5.
412
1 87
78
G4–
04
Ce 0
.98B
i 0.0
2O2
n.d.
18
5.
411
10
94
89
G4–
05
Ce 0
.93K
0.02
Yb 0
.05O
2 n.
d.
36
5.41
0 3
81
73
G4–
06
Ce 0
.98W
0.02
O2
n.d.
19
5.
411
0 0
0
G4–
07
Ce 0
.96A
l 0.0
2Cu 0
.02O
2 n.
d.
13
5.40
8 7
90
81
G4–
08
Ce 0
.96Z
r 0.0
2Fe 0
.02O
2 n.
d.
13
5.40
5 0
0 0
G4–
09
Ce 0
.95A
l 0.0
5O2
n.d.
11
5.
409
2 56
52
G4–
10
Ce 0
.98C
a 0.0
2O2
n.d.
23
5.
413
1 93
83
G4–
11
Ce 0
.95R
u 0.0
5O2
n.d.
14
5.
410
0 0
0
G4–
12
Ce 0
.93B
i 0.0
7O2
n.d.
17
5.
416
26[f
] 82
85
G4–
13
Ce 0
.93A
l 0.0
2La 0
.05O
2 n.
d.
12
5.42
5 2
62
57
G4–
14
Ce 0
.93G
d 0.0
2Yb 0
.05O
2 n.
d.
17
5.40
9 2
100
89
G4–
15
Ce 0
.92C
u 0.0
8O2
n.d.
14
5.
411
6 90
81
G4–
16
Ce 0
.88M
n 0.0
2Cu 0
.10O
2 n.
d.
14
5.41
0 7
95
85
G4–
17
Ce 0
.90N
d 0.0
8Fe 0
.02O
2 n.
d.
14
5.42
2 0
0 0
G4–
18
Ce 0
.90B
i 0.0
7Al 0
.02O
2 n.
d.
13
5.41
6 25
[f]
85
88
G5–
01
Ce 0
.98G
d 0.0
2O2
n.d.
20
5.
411
1 65
60
Cha
pter
2.1
Cat
alys
is
47
T
able
1, c
ontin
ued.
Cat
alys
t C
ompo
sitio
n[a]
Sur
face
are
a
(m2 /g
)
Cry
stal
lite
siz
e
(nm
)[b]
Lat
tice
para
met
er (
Å)[c
]
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
H2
oxid
atio
n
sele
ctiv
ity (
%)
Fit
ness
valu
e[d]
G5–
02
Ce 0
.96A
l 0.0
2Pt 0
.02O
2 n.
d.
15
5.41
0 0
0 0
G5–
03
Ce 0
.98T
i 0.0
2O2
n.d.
19
5.
409
0 0
0
G5–
04
Ce 0
.96K
0.02
Cu 0
.02O
2 n.
d.
45
5.41
0 2
100
89
G5–
05
Ce 0
.96N
d 0.0
2Sn 0
.02O
2 n.
d.
13
5.40
9 8
89
85
G5–
06
Ce 0
.96C
r 0.0
2Al 0
.02O
2 n.
d.
n.d.
n.
d.
1 76
69
G5–
07
Ce 0
.93M
n 0.0
2Cu 0
.05O
2 n.
d.
n.d.
n.
d.
7 86
77
G5–
08
Ce 0
.95Y
0.05
O2
n.d.
n.
d.
n.d.
1
54
51
G5–
09
Ce 0
.93G
d 0.0
2Mn 0
.05O
2 n.
d.
14
5.40
8 4
74
67
G5–
10
Ce 0
.98Y
b 0.0
2O2
n.d.
18
5.
408
2 55
51
G5–
11
Ce 0
.93P
r 0.0
2Zr 0
.05O
2 n.
d.
15
5.41
0 2
50
47
G5–
12
Ce 0
.92Z
r 0.0
8O
n.d.
13
5.
405
2 52
49
G5–
13
Ce 0
.91M
n 0.0
4Sr 0
.05O
2 n.
d.
13
5.41
1 0
0 0
G5–
14
Ce 0
.88C
r 0.0
8Bi 0
.04O
2 n.
d.
25
5.41
0 36
[f]
83
86
G5–
15
Ce 0
.86G
d 0.0
8Cu 0
.06O
2 n.
d.
18
5.41
7 5
93
83
G5–
16
Ce 0
.89S
n 0.0
4La 0
.07O
2 n.
d.
12
5.43
1 16
92
88
G5–
17
Ce 0
.95C
a 0.0
3Pt 0
.02O
2 n.
d.
17
5.41
2 0
0 0
G5–
18
Ce 0
.96M
n 0.0
2Bi 0
.02O
2 n.
d.
13
5.40
9 13
83
80
[a
] Not
e th
at in
the
GA
, con
cent
rati
ons
of 2
, 5, 8
, and
10
mol
% a
re u
sed.
[b] C
eria
has
an
aver
age
crys
tall
ite
size
of
25 n
m (
stan
dard
dev
iati
on =
4, n
= 4
). D
opin
g w
ith p
otas
sium
yie
lds
larg
er c
ryst
alli
tes
(93
nm f
or G
2–13
, Ce 0
.90K
0.10
O2,
and
53
nm f
or G
3–01
, Ce 0
.98K
0.02
O2)
. In
gene
ral,
how
ever
, dop
ing
decr
ease
s th
e cr
ysta
llite
siz
e. T
he a
vera
ge c
ryst
alli
te s
ize
of t
he r
est
of t
he d
oped
cat
alys
ts i
s 17
nm
(s
tand
ard
devi
atio
n =
6, n
= 7
8). [c
] Cer
ia h
as a
n av
erag
e la
ttice
par
amet
er o
f 5.
4094
Å (
stan
dard
dev
iatio
n =
0.0
008,
n =
4).
Dop
ing
with
neo
dym
ium
yie
lds
a la
rger
latt
ice
para
met
er (
5.43
0 Å
for
G1–
09,
Ce 0
.90N
d 0.1
0O2,
and
5.4
22 f
or G
4–17
, Ce 0
.90N
d 0.0
8Fe 0
.02O
2). A
lso,
cat
alys
t G4–
13, C
e 0.9
3Al 0
.02L
a 0.0
5O2,
has
a v
alue
of 5
.425
. The
re a
re n
o tr
ends
, how
ever
, upo
n do
ping
, the
ave
rage
latt
ice
para
met
er o
f th
e re
st o
f th
e do
ped
cata
lyst
s is
5.4
104
nm (
stan
dard
dev
iati
on =
0.0
040,
n =
77)
. [d
] The
fit
ness
val
ue i
s de
fine
d as
: F
= [
sele
ctiv
ity
+ (
0.2
× a
ctiv
ity)
]/12
0 ×
100
, an
d ra
nges
fro
m 0
– 1
00.
Not
e th
at t
he
‘hyd
roge
n ac
tivi
ty’
is c
onve
rted
into
a 0
– 1
00 s
cale
pri
or to
the
calc
ulat
ion
of th
e fi
tnes
s fu
ncti
on. A
n ac
tivi
ty v
alue
of
eith
er 0
, 33,
66
or 1
00 is
giv
en to
cat
alys
ts w
ith
hydr
ogen
act
ivit
ies
rang
ing
from
0%
, 0.1
–7%
, 7.1
–21.
2%, o
r ar
e hi
gher
than
21.
3%, r
espe
ctiv
ely.
[e] N
ot d
eter
min
ed. [f
] T
hese
cat
alys
ts c
onve
rt 1
00%
of
the
hydr
ogen
fee
d at
the
begi
nnin
g of
the
redu
ctiv
e cy
cle.
Thi
s do
es n
ot a
ffec
t the
to
tal a
ctiv
ity,
how
ever
, sin
ce a
ll o
f th
ese
cata
lyst
s ar
e de
plet
ed b
efor
e th
e en
d of
the
redu
ctio
n cy
cle.
[g] A
vera
ge o
f 4
sam
ples
.
Chapter 2.1 Catalysis
48
Selectivity towards hydrogen oxidation. In a typical reaction, (Scheme 1)
250 mg of SOR catalyst was placed on a quartz wool plug in a quartz reactor and
heated to 550 °C in 1% v/v O2 in Ar. The selectivity and activity were assessed
over nine redox cycles, each consisting of an 18 min oxidation step (1% v/v O2 in
Ar), a 4 min purge in pure Ar, a 10 min reduction step (4:1:1% v/v C3H8:C3H6:H2
in Ar), and a second 3 min purge in pure Ar. The reductive gas feed simulates the
effluent from industrial propane dehydrogenation.[14] The selectivity and activity
are assessed during this step using the data of 15 GC measurements, spread over
the 10 min interval. The selectivity is determined as the ratio 1002 total
H
conversion
conversion.
The ‘total conversion’ is the conversion of hydrogen, propane and propene. A
selective catalyst will convert only hydrogen, yielding a selectivity of 100%.
Conversion of propene and/or propane will lower the selectivity. Note that several
interactions between these hydrocarbons and the catalyst can occur which result in
hydrocarbon conversion (see Scheme 2). The activity of the catalyst is determined
as the percentage of the hydrogen feed which is combusted during the reduction
cycle, and is labelled ‘hydrogen activity’. Note that the oxygen source for this
combustion is the catalyst's lattice oxygen, which has to be refilled once depleted,
hence the redox cycling. The lattice oxygen of all of the catalysts tested was
depleted before the end of the reduction cycle (i.e. within 10 minutes).
H2
SOR-O
(De)hydrogenation
C CC
SOR-O
Coking
COx
SOR-
Hydrocarboncombustion
H2O
SOR-
Hydrogencombustion
H2H2 CH3
Cracking
SOR-O
H2
SOR-O
(De)hydrogenation
C CC
SOR-O
Coking
COx
SOR-
Hydrocarboncombustion
H2O
SOR-
Hydrogencombustion
H2COx
SOR-SOR-
Hydrocarboncombustion
H2O
SOR-
Hydrogencombustion
H2H2 CH3
Cracking
SOR-O
CH3
Cracking
SOR-O
Scheme 2. Cartoon showing possible interactions between the dehydrogenation
gas mixture and the SOR catalyst. The so-called oxygen demand is the total amount of
catalyst oxygen used by the processes.
Chapter 2.1 Catalysis
49
Figure 2 shows a scheme of the proposed redox process, where the reactor
contains both the SOR (black) and dehydrogenation catalyst (white).[41] The alkane
is fed over the reactor bed and is dehydrogenated by the dehydrogenation catalyst.
The formed H2 is selectively burned from the gas mixture by the SOR (Figure 2A).
Since the colour of ceria changes from yellow to black when reduced, the process
can be actually seen. This is shown in the top picture, which is taken in our reaction
setup, after opening the reactor (catalyst used: Ce0.90Zr0.10O2, G1–25). Note that in
our screening reaction, the reactor only contains the SOR catalyst, over which a
mixture of 4/1/1% v/v propane/propene/hydrogen is fed. The pictures were taken
with a two-second time interval during the reduction step (A). The quick colour
change is mainly caused by coking (the initial selectivity of all catalysts is low,
probably due to the presence of highly reactive adsorbed oxygen). After this initial
quick colour change, the bed gets darker and darker during the remainder of the
reduction cycle, due to reduction of the ceria. Just before the entire SOR is spent,
the bed is flushed with nitrogen to remove the reductive gases (Figure 2 B). Then,
oxygen is fed to the reactor, reoxidising the SOR and burning off coke from the
dehydrogenation catalysts (C). The bottom pictures, taken at ten second intervals,
show this step for catalyst G1–25. When the bed is reoxidised, the oxidative gas
mix is flushed out and ready for another redox cycle (Figure 2 D).
Chapter 2.1 Catalysis
50
CnH(2n+2)
A
N2O2N2
B C D
N2O2COx
N2CnH2nH2O
Fresh SOR
Fresh DH
Spent SOR
Spent DH
Time (2 s steps)
Time (10 s steps)
Reoxidation
Reduction
CnH(2n+2)
A
N2O2N2
B C D
N2O2COx
N2CnH2nH2O
Fresh SOR
Fresh DH
Spent SOR
Spent DH
Time (2 s steps)Time (2 s steps)
Time (10 s steps)Time (10 s steps)
Reoxidation
Reduction
Figure 2. Top: pictures showing the colour change of the SOR catalyst
Ce0.90Zr0.10O2, G1–25, during reduction, taken at two-second time intervals. Middle:
scheme of the proposed industrial redox dehydrogenation process. After the
dehydrogenation step A, the bed is flushed with nitrogen (B), and the catalysts are
regenerated through reoxidation (C). This burns coke from the dehydrogenation catalyst
and restores the lattice oxygen of the SOR catalyst. After another nitrogen flush (D) the
reactor is ready for the next redox cycle. Bottom: pictures showing the colour changes of
the SOR G1–25 during reoxidation, taken at ten-second time intervals. The difference in
colour between the reduced catalyst in the top and the bottom rows of pictures is due to the
different time scales (the top pictures show only the initial 18 seconds of the reduction).
Chapter 2.1 Catalysis
51
Reproducibility. Table 2 shows characterisation and catalytic data of three
Mn-doped catalysts, with undoped ceria added as a reference. The three Mn-doped
catalysts show comparable surface areas, crystallite sizes, lattice-parameters,
selectivities and fitness values. The catalyst with the highest level of Mn-doping
(G1–20), does show the highest activity.
The crystallite size of the Mn-doped catalysts is smaller than that of the
undoped ceria, resulting in a larger surface area. This decrease in crystallite size
occurs for most dopant types. Ceria has an average crystallite size of 25 nm
(standard deviation = 4, n = 4). Doping with potassium yields larger crystallites (93
nm for G2–13, Ce0.90K0.10O2, and 53 nm for G3–01, Ce0.98K0.02O2, see Table 1).
Doping with Ca, W and Cr yields catalysts with a crystallite size comparable to
that of the undoped ceria. In general, however, doping decreases the crystallite size,
the average crystallite size of the rest of the doped catalysts is 17 nm (standard
deviation = 6, n = 78). Table 2 shows that the lattice parameter of the Mn-doped catalysts is
comparable to that of undoped ceria. The average lattice parameter of undoped
ceria is 5.4094 Å (standard deviation = 0.0008, n = 4). Doping with neodymium
yields a larger lattice parameter (5.430 Å for G1–09, Ce0.90Nd0.10O2, and 5.422 for
G4–17, Ce0.90Nd0.08Fe0.02O2). Also, catalyst G4–13, Ce0.93Al0.02La0.05O2, has a value
of 5.425. There are no trends, however, upon doping. The average lattice parameter
of the rest of the doped catalysts is 5.4104 nm (standard deviation = 0.0040, n =
77).
The average fitness value of the three Mn-doped catalysts shown in Table
2 is 81, with a standard deviation of 5 (note that the doping level of these catalysts
varies). A standard deviation of 3 of the fitness value was determined by
synthesising and testing duplo samples of catalyst G1–03, with fitness value 82,
G1–05, with fitness value 75, and G1–13, with fitness value 87 (note that the
activity of these was determined as the total oxygen consumption, and not as the
amount of oxygen used specifically for hydrogen combustion, the ‘hydrogen
activity’). Measurements of two fresh batches of a perovskite-type catalyst gave a
comparable standard deviation of 2 (La0.9Sr0.1MnO3, at a average fitness value of
95, see Chapter 2.2). Standard deviations of 5 and 6 of the fitness value are
obtained for the fourteen copper-containing doped ceria catalyst of the set, and the
Chapter 2.1 Catalysis
52
twelve bismuth-containing ones, respectively (see Table 1). In these cases, the
standard deviation can be higher due to the varied catalyst composition (the
presence of a second dopant besides the copper or bismuth, and different dopant
concentrations). The standard deviation is still relatively low, however, due to the
lack of synergy between dopants and the lower weight of activity in the fitness
value.
Cha
pter
2.1
Cat
alys
is
53
T
able
2. C
atal
ytic
and
cha
ract
eris
atio
n da
ta f
or th
e m
agne
sium
-cer
ia m
ixed
oxi
des.
Cat
alys
t C
ompo
sitio
n S
urfa
ce a
rea
(m2 /g
)
Cry
stal
lite
size
(nm
)
Lat
tice
para
met
er (
Å)
Hyd
roge
n ac
tivity
(% H
2 co
mb.
)
H2
oxid
atio
n
sele
ctiv
ity (
%)
Fit
ness
valu
e
G1–
24
CeO
2 36
[a]
25[a
] 5.
409[a
] 0
0 0
G3–
12
Ce 0
.98M
n 0.0
2O2
60
14
5.40
8 3
95
85
G3–
04
Ce 0
.96M
n 0.0
4O2
59
13
5.40
8 4
85
76
G1–
20
Ce 0
.91M
n 0.0
9O2
56
11
5.40
7 5
93
83
[a] A
vera
ge o
f fo
ur s
ampl
es.
Chapter 2.1 Catalysis
54
Setting up and running the algorithm. The dopant metals (M1 and M2 in
Ce1-x-yM1xM
2yO2) are chosen from 26 candidates, marked green in Figure 3. These
were selected to cover a wide range of the periodic table. Elements that are non-
solids, highly toxic, or are only artificially prepared were excluded (red squares in
Figure 3). The white squares denote elements that are possible dopants, but were
not included in this study. Using these 26 dopant candidates in five concentrations
(0, 2, 5, 8 and 10 mol%), and with a maximum of two dopants per catalyst, already
yields a huge catalyst space – over 17000 combinations. Synthesising and testing
all of these is not an option, and we therefore used a genetic algorithm to explore
this catalyst space. This method is based on operators inspired by evolutionary
biology, such as mutation, inheritance, natural selection and recombination.[21, 42]
The catalysts are treated as a group of organisms, the best of which are allowed to
breed (exchange their ‘genetic material’, i.e. the dopant type and the dopant
concentration), producing a new generation (a new set of catalysts). The algorithm
follows an iterative process, wherein several generations of catalysts are
synthesised and tested. In each iteration, the fitness value F of each catalyst is
calculated from its activity and selectivity, and used for selecting the catalysts for
the next generation.
The algorithm we use does not support multi-objective optimisation
(optimising several parameters at once). Therefore, the fitness value must be a
single parameter, representing both the activity and selectivity of the catalysts. As
noted above, the fitness function is defined as: F = [selectivity + (0.2 ×
activity)]/120 × 100, and ranges from 0 – 100. We decided to give more weight to
the selectivity, since once a selective catalyst is discovered, its activity may still be
improved, for example by increasing its surface area (the latter can be achieved by
reducing the crystallite size, and results in a higher oxygen release at lower
temperatures, that is, in the temperature range where the selective hydrogen
combustion is performed. See also Figure 11 in Chapter 3.4). Moreover, the
catalyst may be applied in other processes where the intrinsic activity is not
relevant, such as co-fed oxidative dehydrogenation, where a small amount of
gaseous oxygen is added to the feed,[13, 14] or by applying the catalysts on an
oxygen permeable membrane, which continuously restores the lattice oxygen.[43]
Chapter 2.1 Catalysis
55
The left hand side of Scheme 3 shows a classic flowchart of an GA, the
right hand side shows the algorithm that includes a data-analysis step (meta
modeling algorithm). The dashed boxes pertain to experimental steps, the rest is
performed in silico.
Mg
Ca Cr MnTiK Fe Cu
Al
Pd SnRuZrYSr
Ta W Pt Pb Bi
YbSmNdPrLa Ce Gd
Mg
Ca Cr MnTiK Fe Cu
Al
Pd SnRuZrYSr
Ta W Pt Pb Bi
YbSmNdPrLa Ce Gd
Figure 3. The Periodic Table showing the 26 dopant metals used in this research.
Elements which are non-solids, highly toxic, or are artificially prepared were not
considered as candidate (marked grey). The white boxes denote dopant candidates not used
as yet.
Chapter 2.1 Catalysis
56
YES
NO
Standard GA Meta modeling algorithm
START
Generate nvirtual catalysts
Synthesise ncatalysts
Evaluate fitness
Criteria met?NO
END
Generate nvirtual catalysts
Generate n x 10virtual catalysts
Determinepredicted fitness
Select ncatalysts
YES
NO
Standard GA Meta modeling algorithm
START
Generate nvirtual catalysts
Synthesise ncatalysts
Evaluate fitness
Criteria met?NO
END
Generate nvirtual catalysts
Generate n x 10virtual catalysts
Determinepredicted fitness
Select ncatalysts
Scheme 3. General flowchart for performing an optimisation using a genetic
algorithm, with and without a data-analysis step (meta modeling algorithm). Dashed boxes
denote experimental steps.
Our first generation consists of 18 randomly generated catalyst
formulations plus seven catalysts from a kinetic study.[11, 37] The formulations are
synthesised and tested, and the fitness value is calculated from the obtained activity
and selectivity values. From this data, a set of new catalyst formulations (‘virtual
catalysts’), are generated, using three steps: selection (based on the fitness value),
cross over (the exchange of genes) and mutation (a random alteration of a small
amount of genes). We use the classic GA settings of tournament selection, 50%
exchange crossover and 10% mutation. In a classic GA, n virtual catalysts are
generated, where n is the size of the next generation of ‘real catalysts’ that will be
synthesised and tested (Scheme 3, left). In the ‘meta-modeling’ algorithm we apply
(Scheme 3, right), n × 10 virtual catalysts are generated. The predicted fitness of
these virtual catalysts is calculated, and based on this, eighteen of the virtual
catalysts are selected for synthesis and testing (real catalysts, G2). This pre-
screening can decrease the time needed for finding the optimal catalyst.
Chapter 2.1 Catalysis
57
The predicted fitness value is obtained by analysing the experimental data
of the first generation of real catalysts. A simple regression is performed between
the fitness values and certain physical properties (descriptors) of the catalysts
which may attribute to the selectivity and activity. We chose the electronegativity,
ionic radius and concentration of the dopant as descriptors (vide infra). The
program then calculates the predicted fitness value for the 250 virtual catalysts,
using the regression data of the real catalysts and the descriptors of the virtual ones.
For example, if there is a positive correlation between fitness value and the
concentration of dopant, the virtual catalysts with high dopant concentration will
get a high predicted fitness value. Finally, eighteen virtual catalysts are selected,
based on their predicted fitness value, for synthesis and testing (G2). This meta
modeling combines the advantages of a genetic algorithm (efficient mapping of the
catalyst space), and data mining (rough pre-screening of the catalysts
candidates).[44-47] For G3, the genetic algorithm uses the fitness data of G1 and G2
(43 real catalysts) to create 430 virtual catalysts. Again, the predicted fitness value
of the virtual catalysts is determined, and 18 virtual catalysts are selected for
synthesis based on this predicted fitness value.
Choosing the catalyst descriptors. Since our reaction involves oxidation
using lattice oxygen, we chose as descriptors the dopant electronegativity
(influencing the strength and polarity of the metal oxygen bond),[1, 48, 49] its ionic
radius (which can affect the oxygen bond strength and oxygen flux by influencing
the amount of stress in the ceria lattice),[50, 51] the dopant concentration, and
combinations of these three. A full list of the descriptors is given in the
experimental section. Note that no experimental catalytic data was available at the
start of the algorithm.
The data show little or no correlation between the fitness value and
descriptors containing concentration or ionic radius. Note that the initial amount of
data is low (25 catalysts after G1, 43 after G2 and G3). Furthermore, most
catalysts contain two dopants. A positive effect on the fitness value of dopant 1
may be clouded by the presence of an unselective dopant 2. Because of this, we
analysed separately a set of single-dopant catalysts. This set also does not show a
correlation between fitness value and either ionic (or atomic) radius, or dopant
Chapter 2.1 Catalysis
58
concentration (note that at our maximum concentration of 10 mol%, phase
segregation between the dopant and ceria starts to occur). Clearly, the type of
dopant has more effect on the fitness value than its concentration, and the ionic
radius and fitness are unrelated.
There is a correlation between catalyst fitness value and dopant electro-
negativity (R2 = 0.2, after analysis of the 43 catalysts of G1 and G2). Indeed, a plot
of the fitness value of the singly doped catalysts against the electronegativity of the
dopants shows that most catalysts with high fitness values contain dopants with
electronegativities ranging from 1.5 to 1.9 (Figure 4). The correlation is not
perfect: there are also bad catalysts in this electronegativity range, and good
catalysts with lower or higher electronegativities.
Fitn
ess
valu
e
K
Ca
Sr
La
Ce
Pr
Nd
Gd
Yb
Zr
Al
Ti
MnCr
W
Fe
Sn
CuBi
Ru
Pd
Pt
Pb
2.2 - 2.31.5 - 1.9EN:
0
25
50
75
100
0.8 - 1.4
Fitn
ess
valu
e
K
Ca
Sr
La
Ce
Pr
Nd
Gd
Yb
Zr
Al
Ti
MnCr
W
Fe
Sn
CuBi
Ru
Pd
Pt
Pb
2.2 - 2.31.5 - 1.9EN: 2.2 - 2.31.5 - 1.9EN:
0
25
50
75
100
0.8 - 1.4
Figure 4. Fitness value against Pauling's electronegativity for the singly doped
cerias. Note that the scale is not linearly increasing, e.g. Ca, Sr, and La all have electro-
negativity of 1. Data for undoped ceria is added. For clarity, the fitness value of catalyst
with a very low activity is set to zero.
The algorithm uses the regression data between the descriptors and the
fitness of the real catalysts to calculate the predicted fitness of the virtual catalyst
set. Clearly, the reliability of this predicted fitness is linked to the correlation
coefficients of the regression. Since these are low, there is little correlation between
the predicted and real fitness of the catalyst. Because of this, it is important to not
Chapter 2.1 Catalysis
59
simply choose the virtual catalysts with best predicted fitness, but to select
catalysts over the entire predicted fitness range. This is also important for model
validation. Furthermore, we limited the amount of noble metal dopants Pd, Pt and
Ru to a maximum of three per generation. Because of their high electronegativity,
noble metals have a high predicted fitness value, but they generally result in very
unselective catalysts, with high levels of coking of the hydrocarbons.
Catalyst fitness value evolution over five generations. Figure 5 shows
the average fitness value (black bars) and the percentage of catalyst with zero
fitness value (hatched bars) of generations G1–G5. The data shows that the
average fitness value increases, and that the amount of bad catalysts decreases from
G1 to G3, and then stabilises. It follows that, in this case, the optimum has been
reached after three generations of catalysts.
Generation
1 2 3 4 50
25
50
75
Ave
rage
fitn
ess
va
lue
(so
lid)
0
25
50
75
% W
ith a
fitn
ess
valu
eof
ze
ro (
hatc
hed
)
Generation
1 2 3 4 50
25
50
75
Ave
rage
fitn
ess
va
lue
(so
lid)
0
25
50
75
% W
ith a
fitn
ess
valu
eof
ze
ro (
hatc
hed
)
Figure 5. The average fitness value (black) and the percentage of catalysts with
zero fitness value (hatched) of G1–G5. A standard deviation of 3 of the fitness value was
determined by synthesising and testing duplo samples of catalyst G1–03, G1–05 and G1–
13. A standard deviation of 2 (at fitness value 95) was found for two fresh batches of a
perovskite-type catalyst (La0.9Sr0.1MnO3, see Chapter 2.2). Standard deviations of 5 and 6
are obtained for the fourteen copper-containing doped ceria catalyst, and the twelve
bismuth-containing ones of the set, respectively. Note that some of these catalyst are bi-
doped, and that the doping levels vary, see Table 1.
Chapter 2.1 Catalysis
60
Figure 6 shows the fitness value of the individual catalysts of G1–G3,
ordered by increasing fitness (G5 and G6 show little change and are not displayed
for clarity). In accordance with the average values shown in Figure 5, the number
of catalysts with F > 0 increases per generation, from G1 to G3.
25
50
75
100
0 5 10 15 20 25Catalyst
Fitn
ess
va
lue
G3G2G1
25
50
75
100
0 5 10 15 20 25Catalyst
Fitn
ess
va
lue
G3G2G1
Figure 6. The fitness values of generations 1 (grey), 2 (white) and three (black)
ordered for increasing fitness value. Note that there is no substantial improvement in the top
catalysts between generations.
The data set, however, contains catalysts with a very low activity
(hydrogen combustion < 2%). Including these catalysts allows for the discovery of
a set of selective catalysts, for which the activity may still be increased, or which
may be used in processes where the intrinsic activity is not relevant. For the redox
process, however, the intrinsic activity is of importance, and the low activity of
these catalysts renders them unsuitable. These catalysts typically have an ‘average’
fitness value, ranging from 25 to 60, and when their fitness value is set to zero
(‘inactive’), the catalysts set consist of two distinct groups, one with fitness zero,
the second with the fitness values ranging from about 75 to 95. This clustering of
the catalysts is also seen in Figure 7, where the selectivity is plotted against activity
for all 97 catalysts (the activity and selectivity of the ‘low active’ catalysts set to
zero). The figure shows that the selectivity of the catalysts is either ‘low’ or ‘high’,
similar to the fitness values. There are few catalysts with an ‘average’ selectivity.
This is because the unselective catalysts tend to have a high activity for
Chapter 2.1 Catalysis
61
hydrocarbon conversion. The degree of coking, combustion or fragmentation of the
hydrocarbons does vary, but these catalysts all produce hydrogen through
hydrocarbon coking. This lack of net hydrogen combustion results in zero activity
(which is defined as the percentage of hydrogen combusted) and zero selectivity
(the ratio of hydrogen combustion over total combustion). Therefore, these
catalysts all have a fitness value of zero. The selectivity and activity of the inactive
catalysts is zero as well, and they are grouped together with the unselective
catalysts at the ordinate of Figure 7. The majority of the second group of catalysts
show selectivities ranging from about 70% to 95%. Note that this is still a
significant difference, since high selectivities are required in the presence of the
valuable hydrocarbons.
50 catalysts
0
25
50
75
100
0 10 20 30 40 50
Hydrogen activity (% H2 combusted)
Se
lect
ivity
(%
)
50 catalysts
0
25
50
75
100
0 10 20 30 40 50
Hydrogen activity (% H2 combusted)
Se
lect
ivity
(%
)
Figure 7. Selectivity versus hydrogen activity for all 97 catalysts. Fifty of the
catalysts have zero values for both selectivity and activity. The ten best catalysts, based on
fitness, shown in Table 3 are marked white. The activity and selectivity of catalyst with
very low activity has been set to zero.
Figure 6 shows that there is a plateau in the fitness values of the best
catalysts. The new combinations of ‘good’ elements in the consecutive generations
do not lead to catalysts with a fitness value above this plateau, i.e. there is little
synergy between the dopants (vide infra). Secondly, the spread in selectivity of the
‘good’ catalysts is much smaller than the spread in activity (Figure 7). Since the
Chapter 2.1 Catalysis
62
weight of the selectivity in the fitness function is much higher than that of the
activity, this also contributes to the plateau in fitness.
Table 3 shows the ten best catalysts overall, selected on fitness value. These
catalysts are marked white in Figure 7. The Pb-doped sample G1–22 shows the
best performance, combining high activity with high selectivity. However, during
synthesis a separate PbO phase is easily formed. This was the case for samples of
Ce0.90Ca0.05Pb0.05O2, Ce0.87Pb0.05Sr0.08O2, and Ce0.93Pb0.05Zr0.02O2.[52] Even
monodoped Pb-Ce samples are difficult to prepare without formation of PbO.
Because of this, no lead containing samples are present in G2 to G5. It is
interesting to note that the selectivity of the catalysts containing a separate PbO
phase is as high as for the Pb-doped ceria. However, previous studies with alumina
and silica supported lead oxides have shown that these catalysts are not stable
under the high temperature redox cycling.[17]
High selectivities are obtained also with the Cr/Zr doped, Sn/La doped, and
Zr/Cu doped catalysts (G3–17, G5–16, and G2–06, respectively), albeit at lower
activity. The rest of the ten best catalysts all contain Bi, and show high activities.[53]
Indeed, analysis of the five most active catalysts shows that these contain either
lead (G1–22) or bismuth (G2–07, G2–03, G5–14, and G1–19). All mixed oxides
of the set, with a Bi concentration of 5 mol% or higher, are amongst the most
active catalysts. However, adding Bi results in some combustion of the propene
feed, giving lower selectivities (<85%).
The most active catalysts are so active that they combust all of the
hydrogen during part of the reductive cycle (100% conversion). It is known that the
presence of hydrogen can limit hydrocarbon combustion. Indeed, increasing the
amount of hydrogen for G1–19, Ce0.90Bi0.10O2, decreases the amount of propene
combustion, and so increases the selectivity. This strategy is not practical,
however, for the oxidative dehydrogenation process, since adding hydrogen shifts
the equilibrium away from the desired products. Interestingly, the Pb-doped ceria
G1–22 combusts all of the hydrogen as well, but without burning any of the
hydrocarbons. Apparently, the lead has and intrinsically lower affinity for
converting propene as compared to bismuth, under these reaction conditions.
Cha
pter
2.1
Cat
alys
is
63
T
able
3. C
atal
ysts
with
the
high
est f
itnes
s va
lue.
Cat
alys
t C
ompo
sitio
n[a]
Sur
face
are
a
(m2 /g
)
Cry
stal
lite
size
(nm
)
Lat
tice
para
met
er (
Å)
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
H2
oxid
atio
n
sele
ctiv
ity (
%)
Fit
ness
valu
e
G1–
22
Ce 0
.92P
b 0.0
8O2
56
13
5.41
1 46
[b]
92
93
G3–
17
Ce 0
.90C
r 0.0
5Zr 0
.05O
2 29
22
5.
405
9 95
90
G4–
04
Ce 0
.98B
i 0.0
2O2
n.d.
[c]
18
5.41
1 10
94
89
G5–
16[d
] C
e 0.8
9Sn 0
.04L
a 0.0
7O2
n.d.
12
5.
431
16
92
88
G2–
06
Ce 0
.84Z
r 0.0
8Cu 0
.08O
2 54
17
5.
411
13
92
88
G4–
18
Ce 0
.90B
i 0.0
7Al 0
.02O
2 n.
d.
13
5.41
6 25
[b]
85
88
G1–
13
Ce 0
.90B
i 0.0
5K0.
05O
2 17
n.
d.
n.d.
18
[b]
91
87
G2–
07
Ce 0
.87B
i 0.0
8Sn 0
.05O
2 55
14
5.
411
45[b
] 84
87
G2–
03
Ce 0
.90C
r 0.0
5Bi 0
.05O
2 31
28
5.
412
38[b
] 84
87
G5–
14
Ce 0
.88C
r 0.0
8Bi 0
.04O
2 n.
d.
25
5.41
0 36
[b]
83
86
G1–
24
CeO
2[e]
36[f
] 25
[f]
5.40
9[f]
0 0
0 [a
] Not
e th
at in
the
GA
, con
cent
ratio
ns o
f 2,
5, 8
, and
10
mol
% a
re u
sed.
[b] T
hese
cat
alys
ts c
onve
rt 1
00%
of
the
hydr
ogen
fee
d at
the
begi
nnin
g of
the
redu
ctiv
e cy
cle.
Thi
s do
es n
ot a
ffec
t the
tota
l act
ivity
, how
ever
, sin
ce a
ll of
thes
e ca
taly
sts
are
depl
eted
bef
ore
the
end
of th
e
redu
ctio
n cy
cle.
[c] N
ot d
eter
min
ed. [d
] Cat
alys
t G5-
4 an
d G
4-14
hav
e hi
gh s
elec
tivi
ties
, but
low
act
iviti
es, a
nd a
re th
eref
ore
not i
nclu
ded
in
the
tabl
e. [e
] Add
ed f
or r
efer
ence
. [f] T
he v
alue
s ar
e th
e av
erag
e of
4 s
ampl
es.
Chapter 2.1 Catalysis
64
Parameters influencing activity and selectivity. Table 3 shows that
generally, the ten best catalysts have a somewhat lower crystallite size as compared
to undoped ceria, their surface area is higher (these two are correlated), and their
lattice parameter is about the same. This, however, is also the case for most low
fitness value catalysts (see also Table 1). For example, G1–23, Ce0.90Pd0.10O2, has a
small crystallite size (13 nm), a high surface area (72 m2/g) and a lattice parameter
of 5.411. Still, this catalyst shows high degree of hydrocarbon combustion and
coking. This shows that the dopant type is more important for selectivity than
general properties such as the crystallite size or lattice parameter. Concerning
activity, studies of undoped ceria showed that a smaller crystallite size and larger
surface area increases the amount of oxygen released below 600 °C.[54, 55] For the
activity in the selective hydrogen combustion, however, the type of dopant added
has a larger effect. For example, the crystallite size of the Cr/Bi doped ceria G2–03
is larger than that of the Zr/Cu doped ceria G2–06, and its surface area is smaller.
Still, G2–03 is three times as active as G2–06.
The activity of copper containing catalysts is increased by the addition of certain
dopants. The monodoped Cu-Ce catalysts only show average activities: G1–18,
containing 10 mol% Cu, combusts 7% of the hydrogen feed, samples containing
7 mol% and 3 mol% of Cu show comparable values (see Table 1). However,
addition of calcium (G2–02) or magnesium (G2–04) increases this activity to 18%
and 17%, respectively. The addition of calcium also increases the time in which the
catalyst is active. This prevents the typical coking of the hydrocarbons, which is
normally seen for Cu–Ce catalysts after their oxygen has been depleted.
Interestingly, doping ceria with Ca or Mg alone, or in combinations with
elements other than copper, results in catalysts with no or only very low activity
(see G1–02, G1–21, G1–08, and G4–10 in Table 1). To assess if the increased
activity stems from an increased surface concentration of copper, X-ray
Photoelectron Spectroscopy (XPS) has been performed on catalysts G2–04
(Ce0.87Mg0.05Cu0.08O2) and a reference sample doped with an equivalent amount of
copper (7–Cu, see Table 4). The data show that the copper surface concentrations
of the two samples are comparable. Furthermore, the oxidation state of the copper
ions is equal for both samples (not shown). It follows that the increased activity of
Chapter 2.1 Catalysis
65
G2–04 is related to the presence of the magnesium, and not merely to a variation in
copper surface concentration.
Table 4. Surface concentrations of doped ceria catalysts components as
determined by XPS.
Catalyst Composition Surface composition (at. %)
Ce O Cu Mg C[a]
G2–04 Ce0.87Mg0.05Cu0.08O2 17.8 55.9 3.6 3.8 18.8
7–Cu Ce0.93Cu0.07O2 24.4 54.6 4.1 - 16.9 [a] This carbon does not stem from hydrocarbon coking during the selective hydrogen
combustion experiments, since the XPS measurements were performed on fresh samples.
Catalyst stability. Catalyst G3–17 (Ce0.90Cr0.05Zr0.05O2) was subjected to a
total of 250 redox cycles.[56] Figure 8 shows that there is some variation in activity,
but selectivity is uncompromised (indeed, the selectivity increases). Furthermore,
XRD analysis of the spent catalyst shows no phase segregation, and no noteworthy
increase in crystallite size (i.e. no sintering). Also, XPS analysis of the fresh and
spent catalyst shows comparable dopant surface concentrations of 2.9 and 3.4 at. %
in case of Cr, and 1.0 and 1.1 at. % in case of Zr (fresh and spent catalysts,
respectively).
Chapter 2.1 Catalysis
66
Selectivity
Activity25
50
75
100
0 50 100 150 200 250
Redox cycle
Se
lect
ivity
(%
)
0
2
4
6
8
10
Hyd
roge
n a
ctiv
ity
(% H
2co
mbu
sted
)
Selectivity
Activity25
50
75
100
0 50 100 150 200 250
Redox cycle
Se
lect
ivity
(%
)
0
2
4
6
8
10
Hyd
roge
n a
ctiv
ity
(% H
2co
mbu
sted
)
Figure 8. Selectivity (●) and hydrogen activity (○) for G3–17 (Ce0.90Cr0.05Zr0.05O2)
for a total of 250 redox cycles (148 hours on stream). Measurements were performed in two
batches of 125 cycles, the catalyst was stored under air at room temperature in between the
batches. Note that the activity (% hydrogen conversion) is lower than presented in Table 1
and 3, since fewer GC measurements were taken each cycle.
The relationship between dopant type and fitness value. Table 3 shows
that the best catalysts often contain Bi, Cr, Sn and Zr. Most of these catalysts,
however, contain two dopants. Analysis of monodoped catalysts shows that indeed
most of the aforementioned metals yield catalysts with a high fitness value (Figure
4). Generally, the type of dopant added results in catalysts with three types of
behaviour: ‘good’, ‘bad’, and ‘inactive’ (Figure 9). ‘Bad’ dopants, such as Pd, Ru,
and Fe, result in unselective catalysts, which coke and combust part of the
hydrocarbon feed, and therefore have a fitness value of zero. ‘Inactive’ dopants,
such as Y, Ti or Pr, yield inactive catalysts, or catalysts with a very low activity,
and are also given a fitness value of zero. ‘Good’ dopants, such as Pb, Cu, or Bi,
yield catalysts with a relatively high selectivity and activity, with fitness values
ranging from 75 – 95. For example, the fourteen catalyst of the set containing
copper, either mono- or bi-doped, have an average fitness value of 83, with a
standard deviation of 5. The twelve bismuth-containing catalyst have an average
fitness value of 84, with a standard deviation of 6.
Chapter 2.1 Catalysis
67
'Inactive'
'Bad'
KCr
Mn
Cu
Sn PbBi
'Good'
Fe
Ta
RuPt
W La
Pd
Yb
Al
Nd
PrZr
CaSr
Gd
TiY
Mg
'Inactive'
'Bad'
KCr
Mn
Cu
Sn PbBi
'Good'
KCr
Mn
Cu
Sn PbBi
'Good'
Fe
Ta
RuPt
W La
Pd
Yb
Al
Nd
PrZr
CaSr
Gd
TiY
Mg
Figure 9. The dopants classified according to their catalytic behaviour, based on
the performance of single-dopant catalysts. Some distinction can be made for the bad
metals: addition of Fe, Pd, Ru, and Pt results in very high amounts of coking and
combustion of the hydrocarbons, where this is much lower for the metals La, Al and W.
Also, when La, Al and W are added at lower concentration, the distinction between ‘bad’
and ‘inactive’ becomes less clear. Catalyst with very low activity, such as Gd and Yb, are
placed in the inactive group. The metals Sm, Ta, and Mg have has not been tested as single
dopant. The behaviour of Ta and Mg is derived from its combination with an inactive
metal.
The behaviour of catalysts containing two dopants can be predicted from
that of the monodoped ones, as shown in Table 5. For example, combining a bad
dopant with a ‘good’, ‘inactive’, or ‘bad’ one, yields a ‘bad’ catalyst. This is
because the ‘bad’ dopant still results in coking of the hydrocarbons, yielding a low
selectivity. The combination of a ‘good’ and ‘inactive’ dopant yields a ‘good’
catalyst, and so does the combination of two ‘good’ dopants. That is, there is not
much synergy, and because of this, a plateau is present around fitness 80 (Figure
6). This is also caused by the high weight of selectivity in the fitness function
(comprising 80% of the fitness value). The combination of two ‘good’ (selective)
dopants yields a catalyst with comparable selectivity to that of the separate
dopants, resulting in the plateau in fitness value.
Chapter 2.1 Catalysis
68
Table 5. Rules for the performance of catalysts containing combinations of
‘good’, ‘bad’ and ‘inactive’ metals.
Rule
number
Combination of dopant
type
Catalyst
performance
Typical fitness
value
1 ‘good’ ‘good’ 75 –95
2 ‘good’ + ‘inactive’ ‘good’ 75 –95
3 ‘good’ + ‘good’ ‘good’ 75 –95
4 ‘inactive’ ‘bad’ 0
5 ‘inactive’ + ‘inactive’ ‘bad’ 0
6 ‘bad’ + ‘inactive’ ‘bad’ 0
7 ‘bad’ + ‘good’ ‘bad’ 0
8 ‘bad’ + ‘bad’ ‘bad’ 0
9 ‘bad’ ‘bad’ 0
Out of the 97 catalysts tested, 91 comply with the rules given in Table 5.
The exceptions are catalysts G2–15 and G5–13, where the combination of ‘good’
Mn and ‘inactive’ Sr yields a ‘bad’ catalyst, catalyst G2–09, where the
combination of ‘good’ Cr and ‘bad’ W yields an ‘inactive’ catalyst, and three iron
containing catalysts G1–03, G2–05 and G2–10. Note that synergy also occurs in
case of Pt–Bi doped ceria, but this catalyst is bi–phasic, and is therefore not
included here (see Chapter 3.3). The fitness value of iron doped ceria itself is low
(G1–17, Ce0.90Fe0.10O2), but the addition of some Cr yields a highly selective
catalyst (G1–03 Ce0.89Cr0.02Fe0.09O2). Interestingly, monodoped Cr-catalysts are
selective as well, but only when doped > 2 mol% (not shown). We performed XPS
to assess whether the high selectivity of G1–03 is related to the surface
concentration Cr. Monodoped Fe (G1–17), and two monodoped chromium
catalysts (2–Cr and G1–16), were analysed as reference. The 2–Cr sample has the
same bulk concentration chromium as G1–03, but is not selective. The data shown
in Table 6 show that indeed, the surface concentration chromium in G1–03 is much
higher than in the (unselective) reference catalyst 2–Cr. Since the Cr and Fe
oxidation states do not vary between samples, and the concentration of iron in
reference G1–17 and G1–03 is equal, the selectivity of G1–03 can be related to the
surface concentration Cr. In this view, note the increase in surface Cr concentration
Chapter 2.1 Catalysis
69
and selectivity of catalyst G3–17 (Ce0.90Cr0.05Zr0.05O2), shown in Table 4 and
Figure 8, respectively.
Besides Cr, addition of 2 mol % Mn or Zr to Ce0.92Fe0.08O2 is beneficial as
well, albeit to a lesser extent (catalysts G2–05 and G2–10, respectively). Addition
of Y (G3–14) or Nd (G4–17) has no effect on catalyst performance.
Table 6. Surface concentrations of doped ceria catalysts components as
determined by XPS.
Catalyst Composition Surface composition (at. %)
Ce O Fe Cr C
G1–17 Ce0.90Fe0.10O2 23.6 52.0 6.2 - 18.1
G1–03 Ce0.89Cr0.02Fe0.09O2 20.6 61.1 5.8 2.3 10.1
2–Cr Ce0.98Cr0.02O2 24.9 56.6 - 1.3 17.3
G1–16 Ce0.92Cr0.08O2 20.5 57.0 - 4.7 17.7
Chapter 2.1 Catalysis
70
Conclusions We have successfully applied a genetic algorithm for screening doped ceria
catalysts for the selective combustion of hydrogen from a mixture with propane
and propene. Five generations, with a total of 97 doped ceria catalysts, have been
synthesised and tested. An increase in average fitness and a lower amount of
catalysts with a fitness value of zero is seen for generations 1 to 3, and no further
improvement is seen for generations 4 and 5. The dopant type has a large effect on
both the activity and selectivity of the catalyst. Three types of catalytic behaviour
are identified, depending on the dopant added: the catalysts are either selective,
inactive or unselective. The best results are obtained when doping with lead,
chromium, copper, manganese and tin (‘good’ dopants). Interestingly, most of
these metals have electronegativities ranging from 1.5 – 1.9.
Generally, the properties of catalysts containing two dopants can be predicted from
the behaviour of singly doped ones. Incorporation of a ‘bad’ dopant yields an
unselective catalysts, regardless of the second dopant added, and the combination
of two ‘good’ dopants yields a catalysts with a fitness comparable to catalysts
doped with a single ‘good’ dopant. Synergy does occur for certain copper and iron
containing catalysts. The activity of copper doped catalysts doubles when adding
calcium or magnesium, and the selectivity of iron doped catalysts is improved by
adding chromium, manganese or zirconium. The stability under the high
temperature redox cycling is excellent. A Cr and Zr doped catalyst was selective
and active over 250 redox cycles (a total of 148 hours on stream), with no phase
segregation or increase in particle size.
Chapter 2.1 Catalysis
71
Experimental Section Materials and instrumentation. Chemicals were purchased from Sigma-
Aldrich, Merck, The British Drug Houses Ltd. or Koch-Light Laboratories Ltd and
used as received. Gases had a purity of 99.5% or higher and were purchased from
Praxair. The O2, He, Ar and N2 streams were purified further over molsieves and/or
BTS columns. All gas flows were controlled by Bronkhorst mass flow controllers.
The specific surface areas were measured by N2 adsorption at 77 K on a
Sorptomatic 99 (CE Instruments) and evaluated using the BET equation. Powder
X-ray diffraction measurements were performed using a Philips PW-series X-ray
diffractometer with a Cu tube radiation source (λ = 1.54 Å), a vertical axis
goniometer and a proportional detector. The 2θ detection measurement range was
10 ° – 93 ° with a 0.02 ° step size and a 5 second dwell time. Lattice constants and
crystallite sizes were obtained after Rietveld refinement (structure fit) using
PANalytical's X'pert software package. Inductive Coupled Plasma (ICP)
measurements were performed on a Perkin Optima 3000XL ICP instrument. ICP
samples were prepared using a Perkin-Elmer Micro Wave Sample Preparation
System. The software used is OptiCat,[57] allowing for designing and exploiting
evolutionary algorithms, and Statistica V6.1 for the data mining.[58] X-ray
photoelectron spectra were recorded on a Kratos HSi spectrometer equipped with a
charge neutraliser and monochromated Al K X-ray source (1486.61 eV) operating
at 144 W. Spectra were recorded with a pass energy of 40 eV at normal emission,
and energy referenced to the valence band and adventitious carbon. Analysis was
conducted using CasaXPS Version 2.3.15.
Procedure for catalyst synthesis. The metal nitrate precursors (or
chlorides or ammonium metallates, where nitrates were not available) were
weighed into a crucible and placed on a heater. When liquefied, they were mixed
with a spatula. If necessary, 2–6 drops of water were added to aid the solution of
the precursors. After about 5 minutes, the crucible was placed in a 140 °C vacuum
oven. Pressure was reduced to < 10 mbar in about 10 minutes. The latter was
performed carefully to prevent vigorous boiling. After 4h, the crucible was placed
in a muffle oven and calcined for 5h at 700 °C in static air (ramp rate: 300 °C/h).
The resulting solid was pulverized, ground and sieved in fractions of 125–212 µm
(selectivity assessment) and < 125 µm (XRD and BET measurements). The final
Chapter 2.1 Catalysis
72
metal concentration was calculated from the amount of precursor weighed in,
corrected for the water content as determined on catalysts G1–01 to G1–18 by
ICP.[11]
Procedure for selective hydrogen combustion experiments. Activity and
selectivity were determined on a fully automated system built in-house, which was
described in detail previously.[11] In a typical experiment, about 250 mg of sample
(125–212 μm) was placed on a quartz wool plug in a 4 mm id quartz reactor. The
reactor was placed in a water cooled oven and heated to 550 °C at 1200 °C/h, under
oxygen flow. At this temperature, redox cycling was started. The selectivity was
determined by GC during the 10 min reduction in 4:1:1% v/v C3H8:C3H6:H2 in Ar
(total flow 50 mL/min), with 5 mL/min of N2 added as internal standard. The 4:1:1
ratio of reductive gases is chosen since this is the equilibrium mixture of a
conventional dehydrogenation catalyst.[14] The gas hourly space velocity (GHSV) is
13200 / h (at the typical bed volume of 0.25 cm3 and the reduction cycle's total
flow of 55 mL/min). The weight hourly space velocity (WHSV) is 1.2 / h, and is
calculated from the weight of C3H8 + C3H6 + H2 per h per the weight of the
catalyst. After a 4 min purge step (pure Ar), the sample was reoxidised for 18
minutes in 1% v/v O2 in Ar (50 mL/min total flow). The redox cycle is completed
by another purge step in pure Ar. The selectivity is determined as the ratio H2
conversion:total conversion (The ‘total conversion’ is the conversion of hydrogen,
propane and propene). Activity is determined as the percentage of the hydrogen
feed which is combusted during the reduction step (hydrogen activity). Both
selectivity and activity are averaged over eight redox cycles.
Procedure for ICP analysis. ICP was performed on catalysts G1–01 to
G1–18, containing all dopant types used in this study. The metals were brought in
solution by heating approximately 50 mg sample in 6 mL aqua regia to 170–
200 °C using a microwave oven. These temperatures were held for 25 min, during
which the pressure typically rose to 40–55 bar. After cooling, the volume was
brought to 100 mL with demineralised water. Before analysis this sample was
diluted 100 times. Cerium recovery from a pure ceria sample using this method
was 98.3% (average of 6 measurements). W, K, and Ta could not be determined
using this method, probably due to limited solubility of the oxides. An alternative
method involving gentle heating in a mixture of 5 mL concentrated HF and 2 mL
Chapter 2.1 Catalysis
73
2M H2SO4 for 2 h did not work either. Therefore the concentration of these
elements was calculated from the amount of precursors weighed.
GA programming, descriptors and implementation. The GA part uses
binary encoding of the variables; the dopants are encoded on 9 bits, while their
quantity is coded on 6 bits. The statistical model used for the meta modeling is a
classic linear regression. The descriptors are the total dopant concentration (mol%),
the ionic radius of dopant 1, the ionic radius of dopant 2, the electronegativity of
dopant 1 (Pauling's scale), the electronegativity of dopant 2, the concentration
ionic radius of dopant 1, the concentration ionic radius of dopant 2, the
concentration electronegativity of dopant 1, and the concentration electro-
negativity of dopant 2.
Procedure for XPS experiments. XPS was performed on 50 mg sample.
The electron analyser pass energy was 160 eV for wide scans and 40 eV for high
resolution spectra. Compositions were corrected using the appropriate elemental
response factors on spectra following a Shirley background-subtraction.
Acknowledgements We thank Dr. M.C. Mittelmeijer–Hazeleger the BET surface area
measurements, Dr. A.F. Lee of the University of York for performing and
analysing the XPS measurements, A.C. Moleman and W.F. Moolhuijzen for help
with the XRD measurements, A.J. van Wijk and L. Hoitinga for performing the
ICP measurements, and NWO–ASPECT for financial support and feedback.
Chapter 2.1 Catalysis
74
References [1] R. K. Grasselli, Top. Catal., 2002, 21, 79. [2] N. Alperowicz, Chem. Week, 2006, 168, 17. [3] G. Parkinson, Chem. Eng. Prog., 2004, 100, 8. [4] J. Plotkin and E. Glatzer, Eur. Chem. News, 2005, 82, 20. [5] A. Scott and M. Bryner, Chem. Week, 2007, 169, 14. [6] F. Cavani, N. Ballarini and A. Cericola, Catal. Today, 2007, 127, 113. [7] R. Grabowski, Catal. Rev. - Sci. Eng., 2006, 48, 199. [8] L. Låte, J. I. Rundereim and E. A. Blekkan, Appl. Catal. A: Gen., 2004, 262, 53. [9] L. Låte, W. Thelin and E. A. Blekkan, Appl. Catal. A: Gen., 2004, 262, 63. [10] G. Rothenberg, E. A. de Graaf and A. Bliek, Angew. Chem., Int. Ed., 2003, 42,
3366. [11] J. H. Blank, J. Beckers, P. F. Collignon, F. Clerc and G. Rothenberg, Chem. Eur.
J., 2007, 13, 5121. [12] E. A. de Graaf, G. Rothenberg, P. J. Kooyman, A. Andreini and A. Bliek, Appl.
Catal. A: Gen., 2005, 278, 187. [13] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 9. [14] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 1. [15] J. G. Tsikoyiannis, D. L. Stern and R. K. Grasselli, J. Catal., 1999, 184, 77. [16] C. H. Lin, K. C. Lee and B. Z. Wan, Appl. Catal. A: Gen., 1997, 164, 59. [17] L. M. van der Zande, E. A. de Graaf and G. Rothenberg, Adv. Synth. Catal., 2002,
344, 884. [18] A. Trovarelli, C. de Leitenburg, M. Boaro and G. Dolcetti, Catal. Today, 1999, 50,
353. [19] J. H. Holland, Adaptation in Natural and Artificial Systems, The University Press
of Michigan, Ann Arbor, MI., 1975. [20] D. Farrusseng and F. Clerc, Appl. Surf. Sci., 2007, 254, 772. [21] D. E. Goldberg, Genetic Algorithms in Search, Optimization and Machine
Learning, Addison-Wesley, Reading, MA, 1989. [22] D. Wolf, O. V. Buyevskaya and M. Baerns, Appl. Catal. A: Gen., 2000, 200, 63. [23] O. V. Buyevskaya, D. Wolf and M. Baerns, Catal. Today, 2000, 62, 91. [24] O. V. Buyevskaya, A. Brückner, E. V. Kondratenko, D. Wolf and M. Baerns,
Catal. Today, 2001, 67, 369. [25] U. Rodemerck, D. Wolf, O. V. Buyevskaya, P. Claus, S. Senkan and M. Baerns,
Chem. Eng. J., 2001, 82, 3.
Chapter 2.1 Catalysis
75
[26] S. Moehmel, N. Steinfeldt, S. Engelschalt, M. Holena, S. Kolf, A. Baerns, U. Dingerdissen, D. Wolf, R. Weber and M. Bewersdorf, Appl. Catal. A: Gen., 2008, 334, 73.
[27] Y. Watanabe, T. Umegaki, M. Hashimoto, K. Omata and M. Yamada, Catal. Today, 2004, 89, 455.
[28] K. Omata, Y. Watanabe, M. Hashimoto, T. Umegaki and M. Yamada, Ind. Eng. Chem. Res., 2004, 43, 3282.
[29] K. Omata, M. Hashimoto, Y. Watanabe, T. Umegaki, S. Wagatsuma, G. Ishiguro and M. Yamada, Appl. Catal. A: Gen., 2004, 262, 207.
[30] F. Clerc, M. Lengliz, D. Farrusseng, C. Mirodatos, S. R. M. Pereira and R. Rakotomalala, Rev. Sci. Instrum., 2005, 76, 062208.
[31] Y. Yamada and T. Kobayashi, J. Jpn. Petrol. Inst., 2006, 49, 157. [32] G. Kirsten and W. F. Maier, Appl. Surf. Sci., 2004, 223, 87. [33] C. Breuer, M. Lucas, F. W. Schutze and P. Claus, Comb. Chem. High T. Scr.,
2007, 10, 59. [34] O. C. Gobin, A. M. Joaristi and F. Schuth, J. Catal., 2007, 252, 205. [35] D. K. Kim and W. F. Maier, J. Catal., 2006, 238, 142. [36] A. Corma, J. M. Serra and A. Chica, Catal. Today, 2003, 81, 495. [37] J. H. Blank, J. Beckers, P. F. Collignon and G. Rothenberg, ChemPhysChem,
2007, 8, 2490. [38] P. J. Scanlon, R. A. M. Bink, F. P. F. van Berkel, G. M. Christie, L. J. van
IJzendoorn, H. H. Brongersma and R. G. van Welzenis, Solid State Ionics, 1998, 112, 123.
[39] G. Rothenberg, E. A. de Graaf, J. Beckers and A. Bliek, Catal. Org. React., 2005, 104, 201.
[40] P. Bera, K. R. Priolkar, P. R. Sarode, M. S. Hegde, S. Emura, R. Kumashiro and N. P. Lalla, Chem. Mater., 2002, 14, 3591.
[41] E. A. de Graaf, G. Zwanenburg, G. Rothenberg and A. Bliek, Org. Process. Res. Dev., 2005, 9, 397.
[42] D. E. Sadava, Life: The Science of Biology, Sinauer Associates, Inc., Sunderland, MA, 2008.
[43] J. A. Dalmon, A. Cruz-Lopez, D. Farrusseng, N. Guilhaume, E. Iojoiu, J. C. Jalibert, S. Miachon, C. Mirodatos, A. Pantazidis, M. Rebeilleau-Dassonneville, Y. Schuurman and A. C. van Veen, Appl. Catal. A: Gen., 2007, 325, 198.
[44] Y. Yin, Soft Computing, 2003, 9, 3. [45] F. Clerc, Optimization and datamining for catalysts library design, Claude Bernard
University, Lyon, 2006. [46] G. Rothenberg, Catal. Today, 2008, 137, 2. [47] V. Prasad and D. G. Vlachos, Ind. Eng. Chem. Res., 2008, 47, 6555. [48] R. T. Sanderson, J. Am. Chem. Soc., 1983, 105, 2259.
Chapter 2.1 Catalysis
76
[49] R. Burch, D. J. Crittle and M. J. Hayes, Catal. Today, 1999, 47, 229. [50] M. Mogensen, N. M. Sammes and G. A. Tompsett, Solid State Ionics, 2000, 129,
63. [51] V. Butler, C. R. A. Catlow, B. E. F. Fender and J. H. Harding, Solid State Ionics,
1983, 8, 109. [52] For some of these catalysts, PbO is not detected by XRD, but the sample is
inhomogeneous in color. The PbO clusters may be below the XRD detection limit (~3 nm).
[53] Due to its low doping level of 2 mol%, the activity of G4-04 is somehwat lower. [54] E. Aneggi, M. Boaro, C. de Leitenburg, G. Dolcetti and A. Trovarelli, J. Alloy
Compd., 2006, 408, 1096. [55] F. Giordano, A. Trovarelli, C. de Leitenburg and M. Giona, J. Catal., 2000, 193,
273. [56] Two batches of 125 cycles (73 hours on stream) were performed on the same
catalyst. [57] F. Clerc, OptiCat - A Combinatorial Optimisation Software. OptiCat can be
downloaded free of charge from http://eric.univ-lyon2.fr/~fclerc/. [58] Statistica 6.1 is commercially avaliable from StatSoft, Inc., 1984-2008, 2300 East
14th Street, Tulsa, OK 74104.
77
2.2
Perovskites as solid ‘oxygen reservoirs’ for
selective hydrogen oxidation
B3+
H2 H2O
B2+
OA B
B3+
H2 H2O
B2+
OA B OA B
Solid ‘oxygen reservoirs’: Perovskite-type oxides, ABO3, can be successfully
applied as solid ‘oxygen reservoirs’ in redox reactions, such as selective hydrogen
combustion. The high selectivity towards hydrogen combustion, from a mixture
with propane and propene, makes them attractive catalysts for a novel propane
oxidative dehydrogenation process.
This work has been published as:
'Selective hydrogen oxidation in presence of C3 hydrocarbons using perovskite
oxygen reservoirs,’ Jurriaan Beckers, Ruben Drost, Ilona van Zandvoort, Paul F.
Collignon and Gadi Rothenberg, ChemPhysChem 2008, 9, 1062.
Chapter 2.2 Catalysis
78
Abstract
Perovskite-type oxides, ABO3, can be successfully applied as solid ‘oxygen
reservoirs’ in redox reactions, such as selective hydrogen combustion. This
reaction is part of a novel process for propane oxidative dehydrogenation, wherein
the lattice oxygen of the perovskite is used to selectively combust hydrogen from
the dehydrogenation mixture at 550 °C. This gives three key advantages: it shifts
the dehydrogenation equilibrium to the desired products side, generates heat aiding
the endothermic dehydrogenation, and simplifies product separation (H2O vs. H2).
Furthermore, the process is safer since it uses the catalysts' lattice oxygen instead
of gaseous O2. We screened fourteen perovskites for activity, selectivity and
stability in selective hydrogen combustion. The catalytic properties depend
strongly on the composition. Changing the B atom in a series of LaBO3 perovskites
shows that Mn and Co give a higher selectivity than Fe and Cr. Replacing part of
the La-atoms with Sr or Ca also affects the catalytic properties. Doping with Sr
increases the selectivity of the LaFeO3 perovskite, but yields a catalyst with low
selectivity in case of LaCrO3. Conversely, doping LaCrO3 with Ca increases
selectivity. The best results are achieved with Sr-doped LaMnO3, converting about
35% of the hydrogen feed with selectivities up to 92%. This catalyst,
La0.9Sr0.1MnO3, shows excellent stability, even after 125 redox cycles at 550 °C
(70 h on stream). Notably, the activity per unit surface area of the perovskite
catalysts is higher than that of doped cerias, the current benchmark of solid oxygen
reservoirs.
Chapter 2.2 Catalysis
79
Introduction Catalyst producers are enjoying a strong growth in demand for petroleum
catalysts.[1] In case of fluid catalytic cracking (FCC), a growth of 2.7% per year is
expected. This mainly stems from the construction of new FCC-units, and the use
of heavier feedstock. Another factor is that more FCC catalysts and additives are
needed in order to increase the production of valuable light olefins, mainly
propene.[1] The price of this monomer has increased to the point where it seriously
affects the polypropene margin.[2] About 95% of the propene production stems
from crackers and oil refineries. Nevertheless, dedicated processes, such as
metathesis and propane dehydrogenation, are gaining ground. The problem is that
propane dehydrogenation is an endothermic, equilibrium–limited process. One way
of solving this problem is applying oxidative dehydrogenation (ODH). Here, the
hydrogen formed by the dehydrogenation is combusted into water. This generates
heat in situ, and shifts the equilibrium to the products side (Scheme 1).
Energy H2
H2O + SOR SOR-O
Dehydrogenationcatalyst
Propane Propene
CnH(2n+2)
A
N2O2N2
B C D
N2O2
COx
N2CnH2n
H2O
Fresh SOR
Fresh DH
Spent SOR
Spent DH
Reduction ReoxidationPurge Purge
Energy H2
H2O + SOR SOR-O
Dehydrogenationcatalyst
Propane Propene
CnH(2n+2)
A
N2O2N2
B C D
N2O2
COx
N2CnH2n
H2O
Fresh SOR
Fresh DH
Spent SOR
Spent DH
Reduction ReoxidationPurge Purge
Scheme 1. Dehydrogenation combined with selective hydrogen combustion (left), and
the reactor configuration (right). A fixed bed is filled with dehydrogenation catalyst (DH)
and solid oxygen reservoirs (SOR), and propane is fed to the bed (A). Before breakthrough
the bed is purged with nitrogen (B), and regenerated with diluted oxygen, which burns off
coke and replenishes the solid oxygen reservoir (C). After purging with inert the bed is
ready for dehydrogenation again (D).
Chapter 2.2 Catalysis
80
Recently, we introduced a new type of ODH system, employing doped
cerias as solid oxygen reservoirs.[3-6] Instead of using gaseous O2, the hydrogen is
combusted using the oxygen of the ceria lattice. The dehydrogenation step is
performed over a conventional dehydrogenation catalyst, such as Pt/Sn/Al2O3 (see
Scheme 1). The use of two catalysts allows for separate tuning of the hydrogen
combustion and the dehydrogenation. The overall process is safer, as it avoids
mixing gaseous O2 and H2 at high temperatures (typically 500–600 °C). After the
ceria is reduced, the oxygen vacancies are re-filled using air, creating a cyclic
redox process. The facile redox and high temperature stability of ceria make it a
good solid oxygen carrier. Note that although ceria itself is not selective, its
activity and selectivity can be improved by doping.[3, 7] Another class of materials
which, depending on their composition, can feature good redox properties and
thermal stability are perovskite oxides.[8] The general formula of these is ABO3,
where the cation A is larger than B. The fact that nearly all the metallic elements
can form perovskites, and that both the A and B cations can be partially substituted,
offers a variety of choices and properties. Perovskites can be insulators (SrTiO3),
metallic conductors (LaCrO3), or even superconductors (layered copper oxides).
They can have extremely high melting points, (Ba3MgTa2O9), be ionic conductors
(LaxSr1-xGayMg1-yO3), or be piezoelectric (PbZrxTi1-xO3).[9-12] Perovskites are used
as catalysts in many applications, including oxidation and hydrogenation-
reactions,[8, 13-15] solid oxygen fuel cells,[8, 16, 17] and pollution abatement.[8, 15, 18] One
elegant application is using perovskites as Diesel exhaust soot filters.[19, 20] By
coating a low loss ceramic monolith with a micro-wave susceptible perovskite
catalyst, fast and efficient heating is achieved, since all energy is supplied where it
is needed, i.e. on the combustion catalyst itself.[21-24]
The qualities of oxygen mobility, redox capacity and temperature stability
make perovskites promising candidates for solid oxygen reservoirs in propane
ODH. In this paper, we screen fourteen perovskites for selectivity, activity and
stability in the selective hydrogen combustion from a mixture of propane and
propene. We find that the catalytic properties are strongly related to the catalyst
composition, with Sr-doped LaMnO3 perovskites showing high selectivity, activity
and stability.
Chapter 2.2 Catalysis
81
Results and Discussion Selectivity towards hydrogen oxidation. The catalysts are prepared using
spray pyrolysis, low temperature thermal decomposition, high temperature solid
state reaction and co-precipitation. In a typical reaction, 250 mg of sample is
placed on a quartz wool plug in a quartz reactor and heated to 550 °C in 1% v/v
O2/Ar. The catalytic activity and product selectivity are determined by GC and MS
over sixteen redox cycles. Each cycle consists of a 10 min reduction step in 4:1:1%
v/v C3H8:C3H6:H2 in Ar, and an 18 min oxidation step in 1%v/v O2/Ar, separated
by 4 min purge cycles. The 4:1:1 ratio of the reductive gasses simulates the effluent
stream from industrial propane dehydrogenation.[25]
Table 1 shows the composition of catalysts 1–17, together with the physical
data and the catalytic performance. The selectivity is expressed as the ratio
1002 total
H
conversion
conversion , and therefore represents the competitive process between H2
oxidation and hydrocarbon conversion. Hydrocarbon conversion is a complicated
process. Combustion into CO or CO2, coking, (de)hydrogenation and
fragmentation into smaller hydrocarbons often occur simultaneously. A selective
catalyst will oxidise only H2, leaving the hydrocarbons unaffected. The activity of
the catalysts is expressed as the ‘oxygen demand’, and as the ‘hydrogen activity’,
which is the percentage hydrogen combusted. The oxygen demand is the amount of
oxygen the catalyst uses during reoxidation. This oxygen is used to refill the lattice
oxygen and to burn off coke. Since unselective catalysts show high coking levels,
their oxygen demand is high (see, for example, catalyst 14 in Table 1). We
therefore also use the hydrogen activity to compare the activity. This is the
percentage of the hydrogen feed the catalysts combusts in a reduction cycle. Since
unselective catalysts show a net formation of hydrogen, due to coking, their
hydrogen activity cannot be determined.
Importantly, the selective hydrogen oxidation reaction is fundamentally
different from the widely published CO and hydrocarbon oxidation using gaseous
O2.[8] First, it is a cyclic process instead of a continuous one. Second, the lattice
oxygen of the catalyst is used as the oxygen source. This means that no suprafacial
processes involving the co-adsorption of oxygen and other reagents occur.
Chapter 2.2 Catalysis
82
The results in Table 1 show that perovskites can be both active and selective in
hydrogen combustion. The catalytic properties depend strongly on the catalyst
composition. For example, SrTiO3 is inactive, LaFeO3 is active but has a low
selectivity, and La0.9Sr0.1MnO3 is both active and selective. The highest selectivities
are obtained using the LaMnO3-based catalysts 1–6. These type of catalysts are
also very stable: 3 shows high selectivity and activity over 70 h on stream (125
redox cycles, see Figure 1).
0
25
50
75
100
0 25 50 75 100 125
Cycle
Se
lect
ivity
(%
)
0
25
50
75
100
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)
Selectivity
Activity
0
25
50
75
100
0 25 50 75 100 125
Cycle
Se
lect
ivity
(%
)
0
25
50
75
100
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)
Selectivity
Activity
Figure 1. Selectivity (♦) and hydrogen activity (◊) of catalyst 3 (La0.9Sr0.1MnO3)
during 73 h on stream at 550 ºC (125 redox cycles).
Undoped perovskites. Catalysts 1, 2, 8, 10, 11 and 14, consist of LaBO3-
type perovskites with either Mn, Co, Fe or Cr as the B-atom (Table 1). In the CO
and hydrocarbon oxidation with gaseous O2, it was found that the catalytic activity
depends mainly on the 3d metal, with Mn and Co showing the highest activities.[22,
26] In the selective hydrogen combustion, the Mn-based (1, 2) and Co-based (8)
catalysts are the most selective. This seems contradictory, since a high hydrocarbon
combustion rate will lower the selectivity towards H2 combustion. However, we
measure here the competitive H2 and hydrocarbon combustion. The relatively high
selectivity of the Mn- and Co-based catalysts results from their high H2 combustion
rate. This is clearly shown for catalyst 8 (Figure 2). This catalyst combusts propene
Chapter 2.2 Catalysis
83
throughout the reductive cycle, but the rate of H2 combustion is higher, yielding a
selectivity of 72%. The LaMnO3 catalysts 1 and 2 convert some hydrocarbons in
the beginning of the reduction cycle, but mainly combust H2 (see Figure 2).
The Co-based catalysts are unstable under redox cycling. The interaction
with the hydrocarbons, both coking and combustion, increases during each cycle.
XRD of the spent catalysts 8 shows formation of Co2O3 and La2O3. Since catalyst
17, La2O3 is inactive, it follows that the decreased selectivity of 8 stems from the
Co2O3. Overall, the Mn-based catalysts show the best performance.
0 200 400 600
Time (s)
20
40
60
80
100
Con
vers
ion
(%) LaMnO3 1
20
40
60
80
100
0 200 400 600
Time (s)
Propene
Hydrogen
Propane
LaCoO3 8
Con
vers
ion
(%
)0 200 400 600
Time (s)
20
40
60
80
100
Con
vers
ion
(%) LaMnO3 1
20
40
60
80
100
0 200 400 600
Time (s)
Propene
Hydrogen
Propane
LaCoO3 8
Con
vers
ion
(%
)
Figure 2. Time resolved conversion profiles of LaMnO3 1 (left) and LaCoO3 8
(right), showing the H2 (▲), C3H6 (○) and C3H8 (●) conversion during a reduction cycle.
Catalyst 8 is unstable after 7 redox cycles.
Tanaka et al.[13] and Tejuca et al.[26] showed that, in the presence of
gaseous O2, the activity in CO and hydrocarbon combustion of Cr- and Fe-based
perovskites is lower than that of Mn- and Co-based ones. In the selective hydrogen
oxidation, the Cr-based catalysts 10 and 11 show similar behaviour (no CO or CO2
is produced, vide infra). Conversely, LaFeO3 14 does combust the hydrocarbons
during the first 50 s of the reduction cycle (CO2 is formed). During the remainder
of the cycle, the hydrocarbons are converted together with the generation of H2,
which is typical for coking. Clearly, this catalyst is unselective in both the oxidised
and reduced form. The short time-period in which CO2 is formed shows that the
catalyst has little oxygen to spare, which is in agreement with our TPR data (Figure
4).
Cha
pter
2.2
Cat
alys
is
84
T
able
1. P
hysi
cal d
ata
and
acti
vity
of
the
pero
vski
te c
atal
ysts
1–1
7.[a
]
Cat
alys
t, co
mpo
siti
on
Surf
ace
area
[m
2 /g]
Synt
hesi
s
met
hod[b
]
Sel
ecti
vity
(%)[c
]
Oxy
gen
dem
and
(mol
O /k
g)
Hyd
roge
n ac
tivi
ty
(% H
2 co
mbu
sted
)[d]
Hyd
roge
n ac
tivi
ty
(% H
2 co
mbu
sted
/ m
2 )[e]
1, L
aMnO
3 3.
8 c.
86
0.
49
35
37
2. L
aMnO
3 13
.2
t.d.
77
0.61
31
9
3, L
a 0.9Sr
0.1M
nO3
3.1
s.
92
0.49
31
40
4, (
La 0
.85S
r 0.1
5)0.
98M
nO3
3.4
s.
89
0.48
36
43
5, L
a 0.8Sr
0.2M
nO3
5.1
s.
92
0.57
44
34
6, (
La 0
.7Sr
0.3)
0.98
MnO
3 5.
8 s.
85
1.
05
72
49
7, L
a 0.8C
e 0.2M
nO3
10.7
s.
75
0.
88
33
12
8, L
aCoO
3 3.
2 s.
72
/ 0
[f]
2.2
/ 4.9
[f]
79[g
] 99
[g]
9, L
a 0.7Sr
0.3C
oO3
5.1
s.
70 /
50[f
] 2.
7 / 3
.6 [f
] 45
[g]
35[g
]
10, L
aCrO
3 3.
8 t.d
. d.
h.[h
] 0.
32
- -
11, L
aCrO
3 n.
d.[i
] h.
t. In
acti
ve
- -
-
12, (
La 0
.85S
r 0.1
5)1.
05C
rO3
2.4
s.
0 0.
36
- -
13, L
a 0.8C
a 0.2C
rO3
2.6
s.
79
0.36
10
16
14, L
aFeO
3 14
.3
t.d.
0 1.
19
- -
15, L
a 0.8Sr
0.2F
eO3
6.3
s.
69
0.48
7
5
16, S
rTiO
3 n.
d.
- In
acti
ve
- -
-
17, L
a 2O
3 n.
d.
- In
acti
ve
- -
-
[a]
Rea
ctio
n co
ndit
ions
: In
a t
ypic
al r
eact
ion,
250
mg
cata
lyst
sam
ple
was
pla
ced
on a
qua
rts
woo
l pl
ug i
n a
quar
ts r
eact
or a
nd h
eate
d in
1%
v/v
air
to
550
°C, a
fter
whi
ch 1
6 re
dox
cycl
es w
ere
perf
orm
ed.
The
oxi
dati
ve g
as f
eed
cons
iste
d of
1%
v/v
oxy
gen
and
1% v
/v H
e (t
race
r) i
n ar
gon,
the
red
ucti
ve g
as f
eed
cons
iste
d of
4:1
:1%
v/v
of
C3H
8:C
3H6:
H2
in A
rgon
at
a to
tal
flow
of
55 m
L/m
in. [
b] s
. = s
pray
pyr
olys
is, t
.d. =
low
tem
pera
ture
the
rmal
dec
ompo
siti
on, c
. = c
o-pr
ecip
itat
ion,
h.t.
= h
igh
tem
pera
ture
sol
id
stat
e re
acti
on. [
c] D
eter
min
ed b
y G
C d
urin
g th
e 10
min
red
ucti
on s
tep,
exp
ress
ed a
s H
2 co
nver
sion
: t
otal
con
vers
ion
* 10
0. T
he in
itia
l un
sele
ctiv
e co
mbu
stio
n is
not
take
n in
to
acco
unt
whe
n ca
lcul
atin
g th
e se
lect
ivit
y. [
d] P
er 2
50 m
g sa
mpl
e. T
his
valu
e ca
nnot
be
dete
rmin
ed f
or u
nsel
ecti
ve c
atal
ysts
sin
ce t
hey
gene
rate
hyd
roge
n vi
a co
king
. [e]
Thi
s is
th
e hy
drog
en a
ctiv
ity
* 4
(= 1
g s
ampl
e) d
ivid
ed b
y th
e su
rfac
e ar
ea (
m2 /g
). [
f] F
or t
he c
obal
t co
ntai
ning
sam
ples
, se
lect
ivit
y w
as f
ound
to
decr
ease
wit
h ev
ery
redo
x cy
cle
perf
orm
ed. S
imul
tane
ousl
y, r
eact
or b
ack
pres
sure
and
oxy
gen
upta
ke d
urin
g th
e re
oxid
atio
n st
ep in
crea
sed.
The
refo
re, s
elec
tivi
ty a
nd o
xyge
n de
man
d is
giv
en f
or th
e in
itia
l and
fi
nal r
edox
cyc
les.
[g]
Sin
ce th
e ca
taly
sts
are
not s
tabl
e, th
e in
itia
l val
ues
are
give
n. [
h] d
.h. =
deh
ydro
gena
tion
. Hyd
roge
n an
d pr
open
e ar
e fo
rmed
und
er c
onve
rsio
n of
pro
pane
. N
o C
O o
r C
O2
is d
etec
ted,
how
ever
, oxy
gen
is u
sed
in th
e re
oxid
atio
n st
ep in
dica
ting
OD
H o
ccur
s. [i
] Not
det
erm
ined
.
Chapter 2.2 Catalysis
85
The influence of doping. The properties of the active B atom in LaBO3-type
perovskites are easily modified by substitution of the La3+ ions.[27] Often, part of
the La-atoms are replaced by Sr. Although Sr2+ itself is inactive (16, Table 1), its
larger size and lower valence affects the crystal lattice.[9, 28] Ideal La-based
perovskites consist of La3+B3+O32-, with the positive and negative charges
cancelling out. To maintain charge neutrality, incorporation of Sr2+ will result in
the formation of B4+ (under oxygen rich conditions), or the creation of oxygen
vacancies (under oxygen poor conditions). Conversely, replacing La3+ by Ce4+ can
lead to the formation of B2+-ions.[29, 30] Both the formation of oxygen vacancies and
the change in oxidation state of the B-atom can affect the catalytic properties of the
perovskite.[8, 31, 32] Indeed, our data show that Sr doping affects the catalytic
behaviour of perovskites in selective hydrogen combustion. Doping with Sr
increases the selectivity of the Fe-based catalysts (14 and 15 in Table 1), and the
stability of the Co-based catalysts (8 and 9). The Sr-doping does not have a
beneficial effect per se: the selectivity of catalyst 12, Sr-doped LaCrO3, is low.
Furthermore, doping the LaCrO3 with another divalent ion (Ca2+, 13) does yield a
selective catalyst. Considering the number of possible A and B-elements and the
possibility of doping, predicting the effects of doping is difficult. Every unique set
of constituting atoms can yield unique catalytic behaviour. For example, in a
previous study of the oxidation of propane and CO with gaseous oxygen, the
La0.8Ca0.2CrO3 catalyst 13 combusted propane at lower temperatures than CO,
contrary to the other seventeen perovskite catalysts.[22] Note that the catalytic
behaviour is reproducible. The average selectivity determined on four fresh
portions of catalyst 3 was 92, with a standard deviation of 4. Duplo measurements
on fresh batches of catalysts 2, 6, 7, 8 and 15 gave standard deviations in the
determined selectivity of 2, 4, 4, 6 and 0, respectively. Note that the amount of
catalyst was varied. The standard deviation in the hydrogen activity of catalyst 3,
after correction for the amount of catalyst weighed in, was 5, at an average activity
of 34% H2 combusted (n = 4). Due to the high activity of this catalyst, however, the
lattice oxygen was not depleted in all cases at the end of the ten minute reduction
cycle.
The doped LaMnO3 perovskites are the best catalysts of the set. The addition of
Ce4+ has no beneficial effect (7), but addition of Sr yields the most selective and
active catalysts 3–6. Possibly, Mn4+ is more selective than Mn2+, since Sr2+ can
Chapter 2.2 Catalysis
86
increase the Mn4+-content and Ce4+ that of Mn2+. The increased activity may stem
from the facile formation of bulk oxygen vacancies in the presence of Sr2+,
resulting in an increased oxygen flux to the surface. There is no clear trend,
however, in selectivity and (normalised) activity with increased Sr content (2–6).
Perovskites vs. doped cerias. Previously, we have assessed the performance
of doped cerias as selective hydrogen combustion catalysts (see Chapter 2.1).[3, 6, 7]
We showed that whilst ceria itself is unselective, its catalytic properties can be
tuned both up and down by doping. A high selectivity can be achieved by doping
with elements such as Cr, Cu, Bi, Mn and Pb. Fe doped ceria, amongst others,
gives poor selectivity. Doping with lead or bismuth yields the most active catalysts.
Table 2 shows the doped ceria catalysts, out of 97 tested, with a hydrogen activity
of 30% or higher. The data show that the most active catalysts contain either Pb or
Bi, and that the selectivity of the Bi-doped catalysts is somewhat lower.
Table 2. The most active doped ceria solid oxygen reservoir catalysts.
Catalyst[a] Composition
Surface
area
(m2/g)
Sel.
(%)
Hydrogen act.
(% H2
combusted) [b]
Hydrogen act.
(% H2
combusted / m2)
G1–19 Ce0.90Bi0.10O2 33 77 30 3.6
G1–22 Ce0.92Pb0.08O2 56 92 45 3.2
G2–03 Ce0.90Cr0.05Bi0.05O2 31 84 37 4.8
G2–07 Ce0.87Bi0.08Sn0.05O2 55 84 47 3.4
G5–14 Ce0.88Cr0.08Bi0.04O2 n.d.[c] 83 35 - [a] These codes are identical to those used in Chapter 2.1. [b] Per 250 mg catalyst. All of these catalysts
convert 100% of the hydrogen feed at the beginning of the reductive cycle. This does not affect the
total activity, however, since all of these catalysts are depleted before the end of the reduction cycle. [c] Not determined.
The hydrogen activity of the doped cerias, normalised for mass, is similar to
that of the perovskites (about 30 – 50% H2 combustion). However, the spray
pyrolysis method, by which most perovskites are prepared, generally yields low
surface areas. The perovskite catalyst contain a surface area of about 4 m2/g,
compared to 30 – 60 m2/g for the doped cerias. As a result, the hydrogen activity of
the perovskite materials normalised for surface area is much higher than that of the
doped cerias. For selective catalysts, the hydrogen activity is about 40 – 50% H2
combusted / m2 for the perovskites, as compared to about 3 – 5% H2 combusted /
Chapter 2.2 Catalysis
87
m2 for the doped cerias (during the 10 min reduction cycle). Figure 3 shows the
hydrogen activity and selectivity of the best perovskites and the best doped-ceria
catalyst (Ce0.92Pb0.08O2). The preparation of high surface area perovskites is a
promising route to obtain highly active catalysts.[15, 33]
1
Catalyst
Dopedceria
3 4 5 60
20
40
60
80
100
Se
lect
ivity
(%
)
0
10
20
30
40
50
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed /
m2 )
1
Catalyst
Dopedceria
3 4 5 61
Catalyst
Dopedceria
3 4 5 60
20
40
60
80
100
Se
lect
ivity
(%
)
0
10
20
30
40
50
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed /
m2 )
Figure 3. Selectivity (full) and hydrogen activity (hatched, normalised for surface
area) of the most promising perovskite catalysts. The data of Ce0.92Pb0.08O2 is added as a
reference.
Activity and selectivity during a reduction cycle. Oxidative
dehydrogenation using solid oxygen carriers involves reaction with lattice oxygen.
Therefore, the nature of the catalyst changes during the reductive cycle – from
oxidised to reduced. The extend of the reduction depends on the activity of the
catalyst and the length of the reduction cycle. Furthermore, at the start of the cycle,
adsorbed oxygen species are likely present. Indeed, Figure 2 shows that the
selective catalyst 1 does combust hydrocarbons at the beginning of the reduction
cycle (0–50 s). This behaviour is seen for almost all catalysts, and also occurs for
the doped-cerias.[3] It may therefore be correlated to unselective reaction with
adsorbed oxygen species.[33] Only the inactive catalysts 11, 16 and 17 do not show
this phenomena.
Following this initial unselective reaction, the oxygen of the catalyst itself
is addressed. Here, the varied catalytic properties of the perovskites are displayed
(Table 1). Selective catalysts, such as the Mn based perovskites, combust H2 into
Chapter 2.2 Catalysis
88
H2O. No or little CO and CO2, from hydrocarbon combustion, is observed.
Unselective catalysts do combust hydrocarbons (8, 9, 12), sometimes together with
coking (resulting in a net H2 production, 12). The inactive catalysts 11, 16 and 17,
however, do not release their oxygen, nor coke the hydrocarbons.
Once the available lattice oxygen is spent, production of H2O and/or
CO/CO2 stops. Due to the high activity of some of the perovskites, not all catalysts
reach this point during the 10 min reduction cycle. For this set, it was only the case
for eight catalysts. The catalytic behaviour of these ‘reduced catalysts’ varies
depending on the catalyst composition (B-atom and type of dopant). The
La(Sr)CoO3 catalysts 8, 9 and LaFeO3 14 coke the hydrocarbons. The La(Ce)MnO3
2, 7, La(Sr)CrO3 10, 12, and LaSrFeO3 15 catalysts are inactive.
Activity and selectivity in view of oxygen binding energy. The oxygen
binding energy is an important property of the solid oxygen reservoirs. Activity can
suffer from a high binding energy, since no oxygen is then released, and selectivity
may suffer from a low binding energy, since this facilitates reaction with the
hydrocarbons. To assess the relationship between oxygen binding energy and our
catalytic data, we performed temperature programmed reduction (TPR) studies. In
a typical experiment, 100 mg of sample was placed on a quartz wool plug in a
quartz reactor, which was placed in a water-cooled oven. Samples were calcined in
situ to 300 ºC under 5% v/v O2/Ar prior to analysis. For the actual TPR
measurement, the sample was heated at 5 ºC/min to 800 ºC, under a 20 mL/min
flow of 67% H2 in Ar. H2-uptake was monitored using a thermal conductivity
detector (TCD). Figure 4 shows the TPR profiles of catalysts 1, 8, 11 and 14.
Catalyst 11, LaCrO3, does not release any oxygen in TPR, and shows no activity in
the selective H2-combustion. The same holds for catalysts 16 and 17. They are not
active in the selective H2-combustion, and do not show any TPR reduction features
≤ 550 °C. Conversely, the TPR-profiles of catalyst 1, 8 and 14 do show low
temperature reduction features, and these perovskites are catalytically active. The
small size of the TPR reduction feature of 14 is in accordance with the catalytic
data: the catalyst does combust hydrocarbons to CO2, but only in the beginning of
the reductive cycle (i.e. its oxygen is spent quickly).
Both LaMnO3 1 and LaCoO3 8 show substantial TPR reduction features at
similar temperatures (~ 350–400 °C). The behaviour in the selective H2-
combustion, however, is very different: LaCoO3 combusts propene throughout the
Chapter 2.2 Catalysis
89
reduction cycle, LaMnO3 mainly combusts hydrogen. This shows that that the TPR
reduction temperature (oxygen binding energy) is not the only factor that governs
the selectivity. The affinity for the hydrocarbons differs per B-type atom. This is
also seen in the different level of coking of the catalysts in the reduced form (see
previous section).
In the CO oxidation with gaseous O2 , activity was correlated to the
binding energy of the oxygen in the perovskite.[27] Cr-based catalysts show a low
activity and a high oxygen binding energy, while Co-based perovskites show a
high activity and a low oxygen binding energy. This agrees with our activity and
TPR results (compare 8 with 11). Note that when the LaCrO3 is doped with either
Sr or Ca (12, 13), the activity in selective hydrogen combustion is increased (Table
1), together with the appearance of substantial low temperature reduction features
in TPR (see Figure 5). This may reflect an increased oxygen flux through the
lattice as a result of the doping with divalent dopants.[34]
Chapter 2.2 Catalysis
90
0 200 400 600 800
Temperature (°C)
800
1600
2400
TC
D s
igna
l (a
u)
8 LaCoO3
1 LaMnO3T
CD
sig
nal (
au
)
11 LaCrO3
14 LaFeO3
200
400
600
0 200 400 600 800
Temperature (°C)
0 200 400 600 800
Temperature (°C)
800
1600
2400
TC
D s
igna
l (a
u)
8 LaCoO3
1 LaMnO3T
CD
sig
nal (
au
)
11 LaCrO3
14 LaFeO3
200
400
600
0 200 400 600 800
Temperature (°C)
Figure 4. TPR of LaMnO3 1, LaCoO3 8 (top), and LaCrO3 11, LaFeO3 14
(bottom). Note the different scaling of the two ordinates. Conditions: 5 °C/min to 800 °C in
67% v/v H2/Ar flowed at 20 mL/min. Samples were calcined in situ to 300 °C (5% v/v
O2/Ar, 30 min hold at 300 °C) prior to analysis.
Chapter 2.2 Catalysis
91
100
200
300
400
500
TC
D s
igna
l (a
u)
11 LaCrO3
10 LaCrO3
13 LaCaCrO3
12 LaSrCrO3
0 200 400 600 800
Temperature (°C)
100
200
300
400
500
TC
D s
igna
l (a
u)
11 LaCrO3
10 LaCrO3
13 LaCaCrO3
12 LaSrCrO3
11 LaCrO3
10 LaCrO3
13 LaCaCrO3
12 LaSrCrO3
0 200 400 600 800
Temperature (°C)
0 200 400 600 800
Temperature (°C)
Figure 5. TPR profiles of the Cr-based catalysts 10, 11, 12 and 13.
Chromium-based perovskites as dehydrogenation catalysts. The catalytic
properties of the two LaCrO3-catalysts 10 and 11 differ. Catalyst 11 is inactive,
where 10 performs dehydrogenation: it converts propane, together with the
production of propene and hydrogen. It is only active, however, during the first half
of the reduction cycle. Indeed, the TPR profiles of these catalysts show a small
reduction feature for 10, and no reduction features for the inactive 11 (Figure 5).
The small size of the reduction feature of 10 can explain the limited time during
which it is active in oxidative dehydrogenation. The XRD patterns of the two
catalysts are identical. Both show the presence of some chromium oxide, which is a
known dehydrogenation catalyst.[35] The different behaviour of the catalysts may be
linked to the different preparation methods used. Intrinsic defects, such as cation
vacancies, will not affect the XRD pattern, but can increase the oxygen flux
through the lattice. Furthermore, the amount of these defects can vary with the
preparation method. The volatile nature of the thermal decomposition method,
through which 10 is made, increases the chance of cation vacancy formation.
Finally, the limited oxygen release of 10 makes it unsuited as solid oxygen carrier.
It may however be put to use in co-fed ODH, where the hydrocarbon and gaseous
oxygen are fed simultaneously.
Chapter 2.2 Catalysis
92
Conclusions Perovskite oxides can be used as solid oxygen reservoirs in redox reactions. They
can be successfully applied in the selective combustion of hydrogen from a mixture
with propane and propene. The catalytic properties are strongly dependent on the
catalyst composition. Changing the B-atom of LaBO3 shows that Mn and Co have
higher selectivities than Fe and Cr. Doping part of the La changes the catalytic
properties, albeit unpredictably. Sr–doping increases the selectivity of LaFeO3, but
yields an unselective catalysts in case of LaCrO3. Doping LaCrO3 with Ca,
however, does yield a selective catalyst. TPR data show that the selectivity cannot
be correlated to oxygen binding energy. The best results are obtained by using Sr-
doped LaMnO3, with selectivities up to 92% and activities around 35% H2
combustion. These catalysts are also very stable, with La0.9Sr0.1MnO3 showing no
drop in activity and selectivity after 125 redox cycles (70 h on stream). Compared
to doped cerias, the perovskite type catalysts show a higher activity per unit surface
area. Thus high surface area perovskites are promising catalysts for selective
oxidative dehydrogenation.
Chapter 2.2 Catalysis
93
Experimental Section Materials and instrumentation. Chemicals were purchased from Sigma-
Aldrich or Merck and used as received. Gasses were purchased from Praxair and
had a purity of 99.5% or higher. The O2, He, Ar and N2 streams were purified
further over molsieves and/or BTS columns. All gas flows were controlled by
Bronkhorst mass flow controllers. The specific surface areas were measured by N2
adsorption at 77 K on a Sorptomatic 99 (CE Instruments) and evaluated using the
BET equation. Powder X-ray diffraction measurements were performed using a
Philips PW-series X-ray diffractometer with a Cu tube radiation source (λ =
1.54 Å), a vertical axis goniometer and a proportional detector. The 2θ detection
measurement range was 10 ° – 93 ° with a 0.02 ° step size and a 5 second dwell
time.
Procedure for catalyst synthesis. The preparation method and catalyst
notation are listed in Table 1. Except for 9, catalyst made by combustion spray
pyrolysis were obtained from Praxair Specialty Ceramics, Woodinville, USA.
Catalyst 9 was kindly provided by Prof. Dr. K. Wiik, (NTNU Trondheim,
Norway), and was prepared by feeding 1 M solutions of the metal nitrates into
ovens kept at 1000 °C, with a feeding speed of 1 L/h. The resulting powders were
calcined in air at 900 °C with cooling and heating rates of 200 °C/h. After
calcination the samples were ball milled overnight in ethanol using Si3N3 balls, to
obtain a small particle size and to break down agglomerates. The particle size of
the spray pyrolysis samples varies from 0.5 to 1.5 μm. Co-precipitation was
performed using the appropriate nitrate salts, with NaOH and H2O2 as precipitating
agents (50 °C, pH 9.1). The precipitate was filtered, washed with distilled water,
and dried overnight at 120 °C in air. Following the drying the precipitate was
crushed and ground to the desired particle size (<200 μm).[36] Low temperature
thermal decomposition was also performed using the corresponding nitrate
solutions.[37] The nitrates were dissolved in demineralised water and heated to the
boiling point. Then, 250 mol% of glycerol was added and the mixture was heated
to 200 °C and kept there for 30 min to complete the decomposition. (CAUTION!
This reaction can become vigorous). The solid was dried and ground to the desired
particle size (<200 μm). Catalysts 11, 16 and 17 were obtained from commercial
sources and used as received. The high temperature solid state reaction method
Chapter 2.2 Catalysis
94
(11) involves reacting lanthanum- and chromium oxide at high temperature and
milling down the resulting solid. Powder X-ray diffraction showed traces of CeO2
in 7, traces of chromium oxide in 10 and 11, and traces of unknown phases in 12.
Procedure for TPR experiments. TPR experiments were performed on a
standard TPR set up built in house, equipped with a TCD detector. In a typical
experiment, 100 mg sample was placed on top of a quartz wool plug in a 4 mm id
quartz reactor. The samples were calcined in situ to 300 °C (10 °C/min, 30 min
hold time), in 5% v/v O2/Ar at 50 mL/min total flow. The samples were allowed to
cool overnight, after which the detector is allowed to equilibrate for about 1.5 h in
a 67% v/v H2/Ar at 20 mL/min total flow. For the actual TPR measurement, the
sample is heated to 800 °C with a ramp of 5 °C/min. Data is collected with 12 s
intervals.
Procedure for testing catalytic activity. Activity and selectivity were
determined on a fully automated system built in house, which was described in
detail previously.[3] In a typical experiment, about 250 mg of sample was placed on
a quartz wool plug in a 4 mm id quartz reactor. The reactor was placed in a water
cooled oven and heated to 550 °C in 1 %v/v O2 in Ar. At this temperature, sixteen
consecutive redox cycles were performed, consisting of an 18 minute oxidation
step in 1 %v/v O2 in Ar at 50 mL/min total flow, 4 minute purge in pure Ar, a 10
minute reduction step in 4:1:1% v/v C3H8:C3H6:H2 in Ar at 50 mL/min total flow,
with 5 mL/min N2 added as internal standard, and a 3 min purge in pure Ar. The
gas hourly space velocity (GHSV) is 26400 / h (at the typical bed volume of
0.125 cm3 and the reduction cycle's total flow of 55 mL/min). The weight hourly
space velocity (WHSV) is 1.2 / h, and is calculated from the weight of C3H8 + C3H6
+ H2 per h per the weight of the catalyst. The oxygen demand is determined as the
amount of oxygen taken up during the oxidation step, as determined by MS. The
selectivity is determined in the reduction step as the relative conversions of H2,
C3H6, and C3H8, as determined by GC. The first data point of each reduction cycle
(after 25 s), is not included in the final selectivity calculation, since all catalyst,
good or bad, show conversion of the hydrocarbons at this point. Activity is
determined as the percentage hydrogen combusted during the reduction step
(labelled ‘hydrogen activity’). Both selectivity and activity are averaged over
fourteen redox cycles.
Chapter 2.2 Catalysis
95
Acknowledgements We thank Prof. Dr. K. Wiik (NTNU Trondheim), Dr. Y. Zhang-Steenwinkel
(ECN) and L. van der Zande for preparing perovskite samples, Dr. M. C.
Mittelmeijer-Hazeleger for the BET surface area measurements, and NWO-
ASPECT for financial support.
Chapter 2.2 Catalysis
96
References [1] A. Scott and M. Bryner, Chem. Week, 2007, 169, 14. [2] N. Alperowicz, Chem. Week, 2006, 168, 17. [3] J. H. Blank, J. Beckers, P. F. Collignon, F. Clerc and G. Rothenberg, Chem. Eur.
J., 2007, 13, 5121. [4] G. Rothenberg, E. A. de Graaf and A. Bliek, Angew. Chem., Int. Ed., 2003, 42,
3366. [5] E. A. de Graaf, G. Zwanenburg, G. Rothenberg and A. Bliek, Org. Process. Res.
Dev., 2005, 9, 397. [6] J. Beckers, F. Clerc, J. H. Blank and G. Rothenberg, Adv. Synth. Catal., 2008, 350,
2237. [7] J. H. Blank, J. Beckers, P. F. Collignon and G. Rothenberg, ChemPhysChem,
2007, 8, 2490. [8] M. A. Peña and J. L. G. Fierro, Chem. Rev., 2001, 101, 1981. [9] A. S. Bhalla, R. Y. Guo and R. Roy, Mater. Res. Innov., 2000, 4, 3. [10] M. A. Keane, J. Mater. Sci., 2003, 38, 4661. [11] J. A. Rodgers, A. J. Williams and J. P. Attfield, Z. Naturforsch. B., 2006, 61, 1515. [12] V. Thangadurai and W. Weppner, Ionics, 2006, 12, 81. [13] H. Tanaka and M. Misono, Curr. Opin. Solid St. M., 2001, 5, 381. [14] V. C. Corberán, Prog. Catal., 1997, 6, 113. [15] M. Alifanti, J. Kirchnerova, B. Delmon and D. Klvana, Appl. Catal. A: Gen.,
2004, 262, 167. [16] D. M. Bastidas, S. W. Tao and J. T. S. Irvine, J. Mater. Chem., 2006, 16, 1603. [17] J. W. Fergus, Solid State Ionics, 2006, 177, 1529. [18] J. A. Rodriguez, Catal. Today, 2003, 85, 177. [19] Y. Zhang-Steenwinkel, L. M. van der Zande, H. L. Castricum, A. Bliek, R. W. van
den Brink and G. D. Elzinga, Chem. Eng. Sci., 2005, 60, 797. [20] Y. Zhang-Steenwinkel, H. L. Castricum, J. Beckers, E. Eiser and A. Bliek, J.
Catal., 2004, 221, 523. [21] J. Beckers and G. Rothenberg, ChemPhysChem, 2005, 6, 223. [22] J. Beckers, L. M. van der Zande and G. Rothenberg, ChemPhysChem, 2006, 7,
747. [23] H. Will, P. Scholz and B. Ondruschka, Top. Catal., 2004, 29, 175. [24] H. Will, P. Scholz and B. Ondruschka, Chem. Eng. Technol., 2004, 27, 113. [25] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 1. [26] J. L. G. Fierro, in Properties and Applications of Perovskite Type Oxides, eds. L.
G. Tejuca and J. L. G. Fierro, Marcel Dekker, New York, 1992, pp. 195 [27] B. Viswanathan, in Properties and Applications of Perovskite Type Oxides, eds. L.
G. Tejuca and J. L. G. Fierro, Marcel Dekker, New York, 1992, pp. 271.
Chapter 2.2 Catalysis
97
[28] R. Ran, X. D. Wu, C. Z. Quan and D. Weng, Solid State Ionics, 2005, 176, 965. [29] T. Kuznetsova, V. Sadykov, L. Batuev, K. Larissa and S. Neophytides, React.
Kinet. Catal. Lett., 2005, 86, 257. [30] D. Weng, H. S. Zhao, X. D. Wu, L. H. Xu and M. Q. Shen, Mater. Sci. Eng., A,
2003, 361, 173. [31] H. Arai, T. Yamada, K. Eguchi and T. Seiyama, Appl. Catal., 1986, 26, 265. [32] A. A. Leontiou, A. K. Ladavos and P. J. Pomonis, Appl. Catal. A: Gen., 2003, 241,
133. [33] D. Fino, N. Russo, G. Saracco and V. Speechia, J. Catal., 2003, 217, 367. [34] T. Arakawa in Properties and Applications of Perovskite Type Oxides, (Eds.: L. G.
Tejuca, J.L.G. Fierro), Marcel Dekker, New York, 1992, pp. 361-377. [35] T. A. Nijhuis, S. J. Tinnemans, T. Visser and B. M. Weckhuysen, Chem. Eng. Sci.,
2004, 59, 5487. [36] Y. Zhang-Steenwinkel, J. Beckers and A. Bliek, Appl. Catal. A: Gen., 2002, 235,
79. [37] A. Delmastro, D. Mazza, S. Ronchetti, M. Vallino, R. Spinicci, P. Brovetto and M.
Salis, Mater. Sci. Eng., B, 2001, 79, 140.
98
99
2.3
Lead-containing solid ‘oxygen reservoirs’
Al2O3 Ce0.92Pb0.08O2 PbCrO4
PbO
H2O
C3H8
C3H6
H2
H2O
C3H8
C3H6
H2H2O
C3H8
C3H6
H2
Performance+
Al2O3 Ce0.92Pb0.08O2 PbCrO4
PbO
H2O
C3H8
C3H6
H2H2O
C3H8
C3H6
H2
H2O
C3H8
C3H6
H2H2O
C3H8
C3H6
H2H2O
C3H8
C3H6
H2H2O
C3H8
C3H6
H2
Performance+
Lead containing catalysts are generally highly selective and active in selective
hydrogen combustion from a mixture with C3-hydrocarbons. This makes them
good candidates for solid ‘oxygen reservoirs’ in a novel process for oxidative
dehydrogenation of propane. A comparison between three different types of lead-
containing catalysts shows that the best results are obtained with PbCrO4, giving
both high activity and selectivity.
This work will be published as:
'Lead-containing solid oxygen reservoirs for selective hydrogen combustion',
Jurriaan Beckers and Gadi Rothenberg, Green Chem. 2009, DOI:
10.1039/b913994j.
Chapter 2.3 Catalysis
100
Abstract
Lead-containing catalyst can be applied as solid ‘oxygen reservoirs’ in a novel
process for propane oxidative dehydrogenation. The catalyst's lattice oxygen
selectively burns hydrogen from the dehydrogenation mixture at 550 °C. This shifts
the dehydrogenation equilibrium to the desired products side and can generate heat,
aiding the endothermic dehydrogenation reaction. We compared the activity,
selectivity and stability of three types of lead-containing solid oxygen reservoirs:
alumina-supported lead oxide, lead-doped ceria, and lead chromate (PbCrO4). The
first is active and selective, but not stable: part of the lead evaporates during the
redox cycling. Stability studies of a biphasic catalyst, consisting of doped ceria
with a separate PbO phase, show that the PbO phase is not stabilised by the ceria.
Evaporation of lead and segregation of lead from the doped ceria occur during
prolonged redox cycling (125 redox cycles at 550 °C, 73 h on stream). The activity
of this catalyst does increase over time, which may be related to the segregation of
lead. Segregation of lead into a separate phase also occurs when starting from lead-
doped ceria (Ce0.92Pb0.08O2). The activity of this catalyst, however, does not
increase with time on stream. Lead chromate (PbCrO4) shows the highest
selectivity (~100%) and activity (2.8 mol O / kg) of all solid oxygen reservoirs
tested (doped cerias, perovskites, and supported metal oxides). The activity is
comparable to the theoretical maximum activity of CeO2 (2.9 mol O /kg). This
activity does drop, however, during the first 60 redox cycles, to about 25% of the
starting value. This is still higher than the best doped cerias, and these test were
carried out on ‘as received’ PbCrO4, of which the stability can possibly be
increased.
Chapter 2.3 Catalysis
101
Introduction Solid ‘oxygen reservoirs’ (SORs) can be successfully applied in a novel
process for propane oxidative dehydrogenation.[1] The catalysts lattice oxygen
selectively burns hydrogen from the dehydrogenation mixture at 550 °C, which can
generate heat and shifts the dehydrogenation equilibrium to the desired products
side. This selective hydrogen combustion can be performed by supported metal
oxides, but they sinter under redox cycling.[2-6] We found that doped cerias, in
which part of the cerium atoms are replaced with dopant atoms, are active,
selective ad stable catalysts in this selective hydrogen combustion. The catalytic
properties depend strongly on the type of dopant used.[7] Our screening study, using
26 different dopant atoms, showed that doping with lead results in high selectivity
and activity. Indeed, out of 97 catalysts tested, a 10 mol% Pb-doped catalyst
showed the highest activity and high selectivity. The lead, however, easily
segregates from the ceria, forming a separate PbO phase, which is not stable in the
redox cycling. The high activity and selectivity of the lead-doped cerias prompted
us to investigate the selectivity and stability of various lead-containing SOR
catalysts.
Results and Discussion We studied the activity, selectivity and stability in the selective hydrogen
combustion of three types of lead-containing SOR catalysts: alumina-supported
lead oxide, lead-doped ceria, and lead chromate (PbCrO4). In a typical reaction,
250 mg of catalyst was placed on a quartz wool plug in a quartz reactor and heated
to 550 °C in 1% v/v O2 in Ar. The selectivity and activity were assessed over 125
redox cycles, each consisting of an 18 min oxidation step (1% v/v O2 in Ar), a 4
min purge in pure Ar, a 10 min reduction step (4:1:1% v/v C3H8:C3H6:H2 in Ar, at
50 mL/min total flow), and a second 3 min purge in pure Ar. The selectivity and
activity are assessed during this step using the data of six GC measurements,
spread over the 10 min interval. The selectivity is determined as the ratio
1002 total
H
conversion
conversion, and the activity as the percentage of the hydrogen feed
combusted by each catalyst (labelled ‘hydrogen activity’). Note that the oxygen
Chapter 2.3 Catalysis
102
source for this combustion is the catalyst lattice oxygen, which has to be refilled
once depleted, hence the redox cycling.
Alumina-supported lead oxide. Both Grasselli and co-workers,[2, 4] and
we[6] showed that supported oxides such as PbO, Bi2O3 or In2O3 can catalyze the
selective oxidation of hydrogen in the presence of C2 and C3 hydrocarbons. We
showed that these catalysts give excellent selectivity (> 99.8%), and the active
oxide loading can be as high as 30–50%.[6] However, most of these metals melt
below or around 500–700 ºC, and thus sinter during the reduction step.[2] Part of the
metal evaporates and a metal deposit is observed on the reactor wall after the
catalytic tests. This deposit is typically found at the end of the quartz reactor, where
it exits the oven (cold spot), and sometimes near the reactor bed itself (see Scheme
1). To see whether this loss of active metal affects the activity and selectivity of the
catalyst, we have subjected catalyst 1, an alumina-supported PbO catalyst (1 mmol
PbO /g catalyst) to 125 redox cycles at 550 °C (73 h on stream, see Figure 1).
Indeed, a very clear metal deposit was observed after the reaction at the reactor
exit, and near the reactor bed (not shown). The data in Figure 1 show, however,
that this does not affect the selectivity and activity of the catalyst, within this time
frame (73 h).
Oven
Quartz reactor
Catalyst
Gas flow
Deposit atreactor exit
Deposit atcatalyst bed
Oven
Quartz reactor
Catalyst
Gas flow
Deposit atreactor exit
Deposit atcatalyst bed
Scheme 1. Schematic of the oven fitted with a quartz reactor and catalyst, and
photos showing a metal deposit at the reactor exit (Bi2O3/Al2O3), and at the reactor bed
(note that the catalyst is removed).
Chapter 2.3 Catalysis
103
0
25
50
75
100
0 25 50 75 100 125
Cycle
Se
lect
ivity
(%
)
0
10
20
30
40
50
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)Selectivity
Activity
0
25
50
75
100
0 25 50 75 100 125
Cycle
Se
lect
ivity
(%
)
0
10
20
30
40
50
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)Selectivity
Activity
Figure 1. Selectivity and activity for catalyst 1, PbO/Al2O3 (1 mmol PbO /g) for a
total of 125 redox cycles (73 hours on stream). The selectivity of the catalyst is 100%. A
slight negative conversion of propene is observed, possibly originating from to the
formation of propene via propane dehydrogenation. For clarity, part of the 125 selectivity
data-points have been removed.
Lead-doped ceria.
Catalyst preparation and characterisation. Doped ceria catalysts can be
easily prepared batch–wise. by co–melting the appropriate the metal nitrate hydrate
precursors (or chlorides or ammonium metallates, when nitrates are not available)
at 140 °C.[1, 8] After the precursors liquefy, the pressure is lowered to about 10 mbar
and a solid mixed metal nitrate forms. This is then converted into the doped ceria
by calcining in static air at 700 °C for 5 h. X-ray diffraction is performed to ensure
the catalysts consist of a uniform phase. This procedure works well when the
dopant precursors melt below 120 °C, or dissolve in to the molten cerium nitrate
(which has a melting point of ~65 °C), and at a maximum dopant concentration of
10 mol % (see Appendix I). The melting point of most nitrates used in our study
lies below 120 °C, but the melting point of lead nitrate is very high (470 °C,
decomposes). Furthermore, it does not dissolve easily in the molten cerium nitrate.
This results in formation of a separate PbO phase, when the standard experimental
is used. The lead nitrate does, however, dissolve well in water. We therefore
adjusted the experimental as follows: water is added drop wise to the lead nitrate,
Chapter 2.3 Catalysis
104
under continuous stirring, until it dissolves. The amount of water added should be
as little as possible, since the addition of too much water results in phase
segregation in the finished catalyst. When the lead nitrate has completely
dissolved, the cerium nitrate is added and mixed to a slurry. This mixture is gently
heated on a heating plate, under continuous stirring, until the cerium nitrate melts.
CAUTION! this step should be performed in a fume hood. The crucible is quickly
placed in a 120 °C vacuum oven, placed in the same fume hood, and the pressure is
lowered to about 10 mbar within about 10 min. After 4h, the sample is calcined in
static air at 700 °C for 5 h. This adjusted experimental increases the success rate of
the synthesis.
Stability in the selective hydrogen combustion reaction. We have tested
the selectivity, activity and stability of two types of lead-doped ceria catalysts: a
biphasic catalyst (2), where part of the lead is present as separate PbO, and
monophasic lead-doped ceria (3, Ce0.92Pb0.08O2). We analyse the biphasic 2 to
assess if the ceria can stabilise the lead oxide phase. Ceria reduces at fairly low
temperatures, starting from about 470 °C, which is below our reaction temperature
of 550 °C. Because of this, strong metal-support interaction (SMSI) can occur.
Metal oxides supported on ceria can spread out over the ceria surface, forming a
Ce–metal–O surface phase, and the ceria can ‘crawl over’ (or ‘decorate’) the
metal(oxide) particles during the redox cycling.[9, 10] Indeed, we observed the
disappearance of a separate CuO phase during the redox cycling for one of our
copper-ceria catalyst (see Chapter 3.2). Possibly, the stability of PbO supported on
ceria may be higher than that of alumina-supported PbO (i.e. less sintering and
metal evaporation). We therefore subjected catalyst 2, Pb-doped ceria (8 mol% Pb),
with part of the lead present as PbO, to 125 redox cycles at 550 °C. The selectivity
and activity data of this catalyst is shown in Figure 2. Interestingly, the activity
increases up until the hundredth cycle, and then stabilises. This change in activity
does not affect the selectivity. After the 73 h on stream, there is a clear metal
deposit on the quartz reactor near the reactor bed, indicating that part of the lead
has evaporated. The increase in activity could be related to the segregation of lead
during the long term redox cycling.
Chapter 2.3 Catalysis
105
0
25
50
75
100
0 25 50 75 100 125
Cycle
Se
lect
ivity
(%
)
0
10
20
30
40
50Selectivity
Activity
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)
0
25
50
75
100
0 25 50 75 100 125
Cycle
Se
lect
ivity
(%
)
0
10
20
30
40
50Selectivity
Activity
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)
Figure 2. Selectivity and activity of catalyst 2, Ce0.92Pb0.08O2 / PbO, for a total of
125 redox cycles (73 hours on stream).
The separate PbO phase of catalyst 2 had formed during the synthesis of
the catalyst. To see if phase separation can also occur during in the selective
hydrogen combustion reaction, 125 redox cycles at 550 °C were performed using
the monophasic catalyst 3 (Ce0.92Pb0.08O2). Figure 3 shows the activity and
selectivity of this catalyst in the selective hydrogen combustion. Contrary to 2, the
activity remains constant during the 125 redox cycles. Interestingly, the activity of
3 is comparable to final activity of 2. XRD analysis of fresh and spent catalyst 3
show, however, that part of the lead has segregated from the ceria into a separate
phase. The crystallite size has also increased from 14 nm (fresh) to 20 nm (spent,
note this was not analysed for 2). As was the case for the biphasic catalyst 2, a
clear band is seen at the reactor bed. A light band has also formed at the reactor
exit. It follows that the lead-doped ceria is not stable under the redox cycling at
550 °C.
Chapter 2.3 Catalysis
106
0
25
50
75
100
0 25 50 75 100 125
Cycle
Se
lect
ivity
(%
)
0
10
20
30
40
50Selectivity
Activity
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)
0
25
50
75
100
0 25 50 75 100 125
Cycle
Se
lect
ivity
(%
)
0
10
20
30
40
50Selectivity
Activity
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)
Figure 3. Selectivity and activity of catalyst 3, Ce0.92Pb0.08O2, for a total of 125
redox cycles (73 hours on stream).
Lead chromate. The high activity and selectivity of lead-doped ceria
prompted us to search for other lead-containing oxides which might be suitable
SORs. By change, we had already synthesised SrPbO3, which was formed when
attempting to make lead and strontium-doped ceria (Ce0.87Pb0.05Sr0.08O2). This
catalysts shows activity and selectivity values comparable to lead-doped ceria, but
is less stable: the activity drops already after 10 redox cycles (not shown). We then
studied lead chromate (PbCrO4), since it is a lead-containing oxide with a high
melting point (844 °C), and chromium is one of the ‘good’ dopant atoms (it is,
however, toxic). First, we made a PbCrO4 – ceria catalyst (4), by mixing about
5 mol% PbCrO4 with cerium nitrate, heating this in a 120 °C vacuum oven,
lowering the pressure to about 10 mbar in 10 min, and calcining for 5 h at 700 °C.
This did not result in ad Pb/Cr-doped ceria, but in a mixture of PbCrO4 and CeO2
(as determined by XRD). The catalyst is very active, combusting 100% of the
hydrogen feed at the start of the reduction cycle, and performing propane
dehydrogenation at the end of the reduction cycle (see Figure 4). The catalyst does
combust part of the propene, however, resulting in a lower selectivity than lead-
doped ceria. We then tested the PbCrO4 itself (catalyst 5). This consists of very fine
powder of low surface area (<1 m2/g). The powder tends to coagulate into small
lumps, even when dried at 110 °C, possibly due to static charging. It follows that
Chapter 2.3 Catalysis
107
the material in this form is not an ideal catalyst, indeed, we had to lower the
amount of sample from 250 mg to about 40 mg, to prevent too high reactor back
pressures. Still, the PbCrO4 (5) is very active and selective (see Figure 5). Both the
selectivity and activity of this catalyst are the highest of all catalysts tested. The
typical initial unselective conversion, which is observed for doped cerias,
perovskites and most supported metal oxides, is not present. Also, no CO or CO2 is
formed during reoxidation, which means that no coking has occurred. Some CO
and CO2 is formed during the reduction cycle, indicating hydrocarbon combustion
does occur, but this is not enough to result in a detectable propane or propene
conversion (see Figure 5). Note that the activity of the PbCrO4 almost equals the
theoretical maximum of CeO2 (an oxygen release of 2.8 and 2.9 mol O / kg
catalyst, respectively). Importantly, this activity is determined under the selective
hydrogen combustion reaction conditions, that is at elevated temperatures and
using the hydrocarbon/hydrogen gas mixture (the activity, usually expressed as the
percentage of H2 combusted, is converted into ‘mol O released /kg catalyst’ for
comparison). Also, the maximum activity of ceria based catalysts is determined
from full (surface and bulk) reduction of Ce4+ to Ce3+. This is hard to achieve at the
reaction temperature of 550 °C.
-100
-50
0
50
100
300
Time (s)
Con
vers
ion
(%
)
Hydrogen
Propene
Propane
-100
-50
0
50
100
300
Time (s)
Con
vers
ion
(%
)
Hydrogen
Propene
Propane
Figure 4. Time resolved conversion profile of PbCrO4 / CeO2 (4) at 550 °C,
showing the C3H8 (○), C3H6 (◊) and H2 (▲) conversion during a reduction cycle. At the end
of the reduction cycle, the propene and hydrogen conversions are negative, indicating they
are produced instead of converted.
Chapter 2.3 Catalysis
108
25
50
75
100
0 300 600
Time (s)
Con
vers
ion
(%
) HydrogenPropenePropane
25
50
75
100
0 300 600
Time (s)
Con
vers
ion
(%
) HydrogenPropenePropane
Figure 5. Time resolved conversion profile of PbCrO4 (5) at 550 °C, showing the
C3H8 (○), C3H6 (◊) and H2 (▲) conversion during a reduction cycle. Note that 40 mg
sample instead of 250 mg was used.
When the PbCrO4 catalyst is subjected to prolonger redox cycling,
however, the activity drops. Figure 6 shows the performance of the PbCrO4 catalyst
(5) over 125 redox cycles (73 h on stream). During the first 60 cycles, the activity
drops to about one quarter of the starting value. Interestingly, this coincides with a
drop in hydrocarbon combustion (the amount of CO2 and CO detected in the
reduction step). Importantly, at the end of the steep drop in activity (from cycle 40
onwards) no CO2 is observed anymore. The catalyst is now truly 100% selective
(this phenomena has been observed for two separate batches of catalyst). Note that
the ‘low’ activity of the PbCrO4 is still twice that of the best doped ceria. Also, we
used the ‘as received’ PbCrO4. The activity loss could be due to sintering of the
PbCrO4 crystallites (the catalyst bed had shrunk), but formation of a hard outer
shell of the coagulated powder will also result in loss of activity by preventing gas
flow through part of the bed. To achieve and maintain a better flow through the bed
we mixed 50 mg of PbCrO4 with 200 mg of inert SiC using a spatula. The two do
not mix well, however, the PbCrO4 still has a tendency to coagulate into separate
particles. Figure 7 shows that indeed, the drop in activity still occurs, even a bit
sooner as compared to the pure PbCrO4. Again, the amount of CO2 (hydrocarbon
combustion) drops to zero together with the drop in activity. No band was seen at
the reactor bed after the 125 cycles. Possibly, some metal deposit was present at the
reactor exit, but this was hard to tell.
Chapter 2.3 Catalysis
109
Future work should focus on increasing the stability of the PbCrO4 catalyst
without compromising its activity. This could be achieved by adding a small
amount of alumina, similar to the copper-zinc-alumina methanol synthesis
catalyst.[11, 12]
0
25
50
75
100
0 25 50 75 100 125
Cycle
Sel
ect
ivity
(%
)
0
10
20
30
40
50
Selectivity
Activity
Hyd
roge
n ac
tivity
(%H
2co
mbu
sted
)
0
25
50
75
100
0 25 50 75 100 125
Cycle
Sel
ect
ivity
(%
)
0
10
20
30
40
50
Selectivity
Activity
Hyd
roge
n ac
tivity
(%H
2co
mbu
sted
)
Figure 6. Selectivity and activity of catalyst 5, PbCrO4, for a total of 125 redox
cycles (73 hours on stream). Note that 40 mg samples was used, instead of 250 mg. For
clarity, part of the 125 selectivity data-points have been removed.
Chapter 2.3 Catalysis
110
0
25
50
75
100
0 25 50 75 100 125Cycle
Se
lect
ivity
(%
)
0
10
20
30
40
50Selectivity
Activity
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)
0
25
50
75
100
0 25 50 75 100 125Cycle
Se
lect
ivity
(%
)
0
10
20
30
40
50Selectivity
Activity
Hyd
roge
n a
ctiv
ity(%
H2
com
bust
ed)
Figure 7. Selectivity and activity of catalyst 5, PbCrO4, mixed with SiC (50 mg
and 200 mg, respectively) for a total of 125 redox cycles (73 hours on stream). For clarity,
part of the 125 selectivity data-points have been removed.
Besides PbCrO4, we have also tested nickel chromite (NiCr2O4, high
activity but low selectivity), copper chromite (Cu2Cr2O4, selectivity 85%, activity
60% H2 combustion). chromium molybdate (Cr2(MoO4)3, inactive) manganese
molybdate (MnMoO4, inactive).
Chapter 2.3 Catalysis
111
A comparison of the activity of the lead-containing catalysts. Table 1
shows the activity of catalysts 1, 2, 3, and 5 normalised for the catalysts' weight
and for the amount of lead present in the catalyst. Note that the values have no
physical meaning and are just used to compare the activity of the catalysts. The
data show that catalysts 1 (PbO/Al2O3) and 5 (PbCrO4) are the most active when
normalised for the catalysts' weight. The initial activity of 5 is much higher than
that of 1, but after 125 cycles the activities are comparable. When normalised for
the amount of lead present in the catalyst, the ceria based catalysts 2 (based on the
final activity) and 3 are most active.
Table 1. The normalised activity of the several lead-containing catalysts.
Catalyst Activity per
gram catalyst[a]
Activity per
gram Pb[b]
Activity per
cm3 [c]
Activity
based on[d]
1 PbO/Al2O3 250 1130 1010 -
2 PbO/Ce092Pb0.08O2 100 1100 790 I.A.
170 1780 1280 F.A.
3 Ce092Pb0.08O2 120 1260 900 -
5 PbCrO4 890 1390 5620 I.A.
240 380 1530 F.A.
5 PbCrO4 + SiC 930 1450 2990 I.A.
200 320 660 F.A. [a]
The numbers have no physical meaning but are used to compare the activity of the catalysts. They
are calculated by dividing the percentage of hydrogen combusted by the amount of catalyst weighed
in. The last digit has been rounded off for clarity. [b] The numbers have no physical meaning but are
used to compare the activity of the catalysts. They are calculated by dividing the percentage of
hydrogen combusted by the amount of lead present in the catalyst. The last digit has been rounded off
for clarity. [c] The numbers have no physical meaning but are used to compare the activity of the
catalysts. They are calculated by multiplying the activity per gram by the density of the catalyst (in
g/cm3). In case of catalyst 5, the density of the SiC is taken since this is the main component of the
catalyst bed. The last digit has been rounded off for clarity. [d] In case of the non-stable catalysts, the
data is calculated based on the initial activity (I.A.) and the final activity (F.A.).
Chapter 2.3 Catalysis
112
Conclusions We have compared the activity, selectivity and stability of three types of
lead-containing solid oxygen reservoirs: alumina-supported lead oxide, lead-doped
ceria, and lead chromate (PbCrO4). These solid oxygen reservoirs can be used for
the selective combustion of hydrogen from a mixture with propane and propene.
The alumina-supported lead oxide is active and selective, but not stable: part of the
lead evaporates during the redox cycling. Stability studies of a biphasic
Ce0.92Pb0.08O2 / PbO catalyst show that the separate PbO phase is not stabilised by
the ceria. During prolonged redox cycling (125 redox cycles at 550 °C, 73 h on
stream), the activity of the catalyst increases, which may be related to the
segregation of the lead.
Overall, lead chromate (PbCrO4) shows the highest selectivity (~100%)
and activity (2.8 mol O / kg). The activity is comparable to the theoretical
maximum activity of CeO2 (2.9 mol O /kg). Although it drops during the first 60
redox cycles to about 25% of the starting value, it is still higher than the best doped
cerias.
Chapter 2.3 Catalysis
113
Experimental Section Materials and instrumentation. Chemicals were purchased from Sigma-
Aldrich or Merck and used as received. Gasses were purchased from Praxair and
had a purity of 99.5% or higher. The O2, He, Ar and N2 streams were purified
further over molsieves and/or BTS columns. All gas flows were controlled by
Bronkhorst mass flow controllers. The specific surface areas were measured by N2
adsorption at 77 K on a Sorptomatic 99 (CE Instruments) and evaluated using the
BET equation. Powder X-ray diffraction measurements were performed using a
Philips PW-series X-ray diffractometer with a Cu tube radiation source (λ =
1.54 Å), a vertical axis goniometer and a proportional detector. The 2θ detection
measurement range was 10 ° – 93 ° with a 0.02 ° step size and a 5 second dwell
time.
Procedure for catalyst synthesis. Synthesis of alumina-supported lead
oxide: This catalyst was synthesised in a specially designed reactor, allowing for
the simultaneous impregnation of 6 supports, described in detail earlier.[6] The
appropriate amount of Pb(NO3)2 was dissolved in 20 mL demineralised water.
Alumina (1.00 g) was placed in the impregnation reactor. The reactor was
evacuated and 0.61 mL solution was injected using a syringe. The reactor was
vehemently shaken for 4 min using a vortex instrument. The material was dried
overnight at 120 °C and exposed to air for 24 h at 25 °C. Consecutive
impregnations were carried out to achieve the desired loading, after which the
material was dried overnight and then calcined in ceramic vessels at 650 °C for 5 h
(heating rate 300 °C/h) under a flow of dry air (125 mL/min).
Synthesis of the doped cerias: The metal nitrate precursors were weighed
into a crucible and just enough water was added to dissolve the metal nitrates
(usually 4–6 drops). The desired amount of cerium nitrate was added, mixed to a
slurry, and the crucible was placed on a heater under continuous stirring. After
about 5 minutes, the crucible was placed in a 140 °C vacuum oven. Pressure was
reduced to < 10 mbar in about 10 minutes. The latter was performed carefully to
prevent vigorous boiling. After 4h, the crucible was placed in a muffle oven and
calcined for 5h at 700 °C in static air (ramp rate: 300 °C/h). The resulting solid was
pulverized, ground and sieved in fractions of 125–212 µm (selectivity assessment)
and < 125 µm (XRD and BET measurements). The final metal concentration was
Chapter 2.3 Catalysis
114
calculated from the amount of precursor weighed in, corrected for the water
content as determined by ICP on a reference set of catalysts .[1]
Procedure for testing catalytic activity. Activity and selectivity were
determined on a fully automated system built in house, which was described in
detail previously.[1] In a typical experiment, about 250 mg of sample was placed on
a quartz wool plug in a 4 mm id quartz reactor. The reactor was placed in a water
cooled oven and heated to 550 °C in 1 %v/v O2 in Ar. At this temperature, sixteen
consecutive redox cycles were performed, consisting of an 18 minute oxidation
step in 1 %v/v O2 in Ar at 50 mL/min total flow, 4 minute purge in pure Ar, a 10
minute reduction step in 4:1:1% v/v C3H8:C3H6:H2 in Ar at 50 mL/min total flow,
with 5 mL/min N2 added as internal standard, and a 3 min purge in pure Ar. The
gas hourly space velocity (GHSV) is 13200 / h (at the typical bed volume of
0.25 cm3 and the reduction cycle's total flow of 55 mL/min). The weight hourly
space velocity (WHSV) is 1.2 / h, and is calculated from the weight of C3H8 + C3H6
+ H2 per h per the weight of the catalyst. The selectivity is determined in the
reduction step as the relative conversions of H2, C3H6, and C3H8, as determined by
GC. The first data point of each reduction cycle (after 25 s), is not included in the
final selectivity calculation, since all catalyst, good or bad, show conversion of the
hydrocarbons at this point. Activity is determined as the percentage hydrogen
combusted during the reduction step.
Acknowledgements We thank Dr. M. C. Mittelmeijer-Hazeleger for the BET surface area
measurements, L.M. van der Zande for synthesising the alumina-supported lead
oxide catalyst and NWO-ASPECT for financial support.
Chapter 2.3 Catalysis
115
References [1] J. H. Blank, J. Beckers, P. F. Collignon, F. Clerc and G. Rothenberg, Chem. Eur.
J., 2007, 13, 5121. [2] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 9. [3] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 1. [4] J. G. Tsikoyiannis, D. L. Stern and R. K. Grasselli, J. Catal., 1999, 184, 77. [5] C. H. Lin, K. C. Lee and B. Z. Wan, Appl. Catal. A: Gen., 1997, 164, 59. [6] L. M. van der Zande, E. A. de Graaf and G. Rothenberg, Adv. Synth. Catal., 2002,
344, 884. [7] J. Beckers, F. Clerc, J. H. Blank and G. Rothenberg, Adv. Synth. Catal., 2008, 350,
2237. [8] J. H. Blank, J. Beckers, P. F. Collignon and G. Rothenberg, ChemPhysChem,
2007, 8, 2490. [9] V. M. Gonzalez-DelaCruz, J. P. Holgado, R. Pereñíguez and A. Caballero, J.
Catal., 2008, 257, 307. [10] S. Bernal, J. J. Calvino, M. A. Cauqui, J. M. Gatica, C. Larese, J. A. P. Omil and J.
M. Pintado, Catal. Today, 1999, 50, 175. [11] G. Ghiotti and F. Boccuzzi, Catal. Rev. - Sci. Eng., 1987, 29, 151. [12] V. F. Anufrienko, T. M. Yurieva, F. S. Hadzhieva, T. P. Minyukova and S. Y.
Burylin, React. Kinet. Catal. Lett., 1985, 27, 201.
116
117
Chapter 3
Characterisation of solid ‘oxygen reservoirs’
The Temperature Programmed Reduction (TPR) setup.
118
119
3.1
Redox kinetics of ceria-based catalysts
Temperature
TC
D s
ign
alT
CD
sig
nal
Time
Red
uctio
n le
vel
Ce-Ca-O2
Red
uctio
n le
vel
CeO2
Ce-Cu-O2
Ce-Ca-O2
Ce-Cu-O2
H2-TPR H2-reduction rate at 550 °C
More!
110 °C
550 °C
Slow!
Fast!
TimeTemperature
Temperature
TC
D s
ign
alT
CD
sig
nal
Time
Red
uctio
n le
vel
Ce-Ca-O2
Red
uctio
n le
vel
CeO2
Ce-Cu-O2
Ce-Ca-O2
Ce-Cu-O2
H2-TPR H2-reduction rate at 550 °C
More!
110 °C
550 °C
Slow!
Fast!
TimeTemperature
Doping ceria generally increases the amount of oxygen released, and lowers the
TPR reduction temperature. This increases the reduction rate of the catalysts at
550 °C, the temperature at which the selective hydrogen combustion reaction is
performed.
This work has been published as:
'Redox kinetics of ceria-based mixed oxides in selective hydrogen combustion', Jan
Hendrik Blank, Jurriaan Beckers, Paul F. Collignon, and Gadi Rothenberg,
ChemPhysChem 2007, 8, 2490.
Chapter 3.1 Characterisation
120
Abstract
Doped cerias, in which about 10 mol% of the cerium is replaced by another metal,
catalyse the selective combustion of hydrogen from a mixture of hydrogen,
propane, and propene at 550 ºC. This makes them attractive catalysts for oxidative
dehydrogenation of propane. The hydrogen combustion shifts the equilibrium to
the products side, supplies energy for the endothermic dehydrogenation, and
simplifies product separation. The dopant type has a large effect on the catalytic
properties. To gain insight in the process, a set of six doped cerias were synthesised
and the catalytic properties and redox behaviour were tested. The doped cerias
generally release more oxygen compared to plain ceria. Cerias doped with Bi, Cu,
Fe, Pd or Ca release 1.6 to 2.0 mg oxygen per 100 mg sample, compared with only
1.2 mg for plain ceria. This is important for reactions where the catalyst acts as an
oxygen reservoir, such as the selective hydrogen combustion. The temperature
where the oxygen is released is generally lower for the doped cerias, and varies
from 110 °C (Cu-CeO2) to 550 °C (Ca-CeO2). This enables catalytic applications
over a wide temperature range. The reduction rate at 550 °C is correlated to the
reduction onset of the catalyst. Catalysts with a relatively low reduction
temperature, such as Cu-, Mn-, Bi- and Pb-CeO2, show a high reduction rate at
550 °C. Conversely, catalysts with a high reduction temperature, such as Fe-CeO2
and plain ceria, reduce more slowly. These catalysts also have a low selectivity
towards hydrogen combustion. The influence of the catalyst composition and
crystallite size on the activity and selectivity is discussed.
Chapter 3.1 Characterisation
121
Introduction Propene is a valuable bulk chemical, used mainly for producing polypropene
(PP).[1] The demand for propene is huge, and is expected to rise to 80 million
tonnes in 2010 worldwide.[2-4] About 95% of the propylene production stems from
crackers and oil refineries. Nevertheless, dedicated processes such as metathesis
and propane dehydrogenation are gaining ground. The dehydrogenation of propane
to propene is an endothermic, equilibrium limited reaction, but these drawbacks
can be overcome by using oxidative dehydrogenation. Here, hydrogen is
combusted into water, generating heat in situ, and shifting the equilibrium to the
products side (Scheme 1).
H2
Energy H2O
1/2 O2
dehydrogenationcatalyst
H2
Energy H2O
1/2 O2
dehydrogenationcatalyst
Scheme 1. Dehydrogenation combined with selective hydrogen combustion.
Recently, we introduced doped cerias as solid ‘oxygen reservoirs’ for oxidative
dehydrogenation.[5, 6] In this process, the dehydrogenation is performed over a
conventional catalyst, and the H2 by-product is burned by oxygen exchange with
the ceria. The addition of the ceria allows for separate tuning of the
dehydrogenation and the hydrogen combustion, and the process is safer since it
avoids mixing gaseous O2 and H2 at high temperatures (typically 500–600 °C).
After the reduction of the ceria, the oxygen vacancies are re-filled using air,
creating a cyclic redox process. Studies by Grasselli et al.,[7-9] Lin et al.[10] and us[11]
showed that supported metal oxides can also perform this selective oxidation.
However, these catalysts are not stable under the redox cycling. Most of the metals
melt below 550 ºC, and when the supported metal oxide is reduced to metal(0), it
liquefies, causing sintering and deactivation. Conversely, ceria is stable under the
redox cycling, and has good oxygen storage capacities. Unfortunately, its
selectivity is low. In our previous study, we showed that using doped cerias can
overcome both the problems of low selectivity and low stability (Chapter 2.1).[5, 12]
Chapter 3.1 Characterisation
122
The activity, selectivity and stability depends strongly on the dopant type. We
found that generally, doping results in catalysts with three types of behaviour:
‘good’ (active and selective), ‘bad’ (active, but unselective), and ‘inactive’. To gain
more insight into the selective hydrogen combustion, we synthesised and tested a
set of doped cerias with dopants from each of these groups, namely Mn, Bi, Cu and
Pb (‘good’), Fe, (‘bad’), and Ca (‘inactive’). This paper reports the results of the
catalyst synthesis, testing, and characterisation.
Results and Discussion Catalyst preparation and characterisation All catalysts were prepared by
co-melting a mixture of the metal nitrate hydrate precursors. After the precursor
has liquefied, the pressure was lowered and a solid mixed metal nitrate formed.
This was converted into the doped ceria by calcining in static air at 700 °C for 5 h.
Catalysts containing about 10 mol% of Mn, Bi, Cu, Fe, Pb and Ca were
successfully prepared (samples 1–6). Sample 7 consists of plain ceria. Figure 1
shows a Fe-CeO2 catalyst (10 mol% Fe) at various stages of preparation. Image A
shows the nitrate salts after weighing and mixing. These precursors become
transparent, often brightly coloured liquids, when heated to 115 °C (image B).
Image C shows the solid catalyst precursor that forms after reducing the pressure to
~ 10 mbar. Finally, image D shows the doped ceria after the calcination step.
Figure 2 shows a picture of catalysts 1–7 after calcination and grinding to particle
size < 212 μm.
Figure 1. A 10 mol% Fe-CeO2 catalyst at various stages of preparation. A: after
mixing the nitrates. B: after heating to 115 °C. C: after treatment in a vacuum oven
(115 °C, 10 mbar). D: after calcination (700 °C, 5h).
Chapter 3.1 Characterisation
123
Figure 2. Pictures of the doped ceria catalysts 1–7 (7 is plain ceria).
Importantly, the catalyst were not prepared by impregnating a cerium
oxide support. The co-melting of the cerium nitrate with the nitrate of the
appropriate metal yields a liquid precursor. This ensures that the ceria and the
dopant are ideally mixed prior to calcination. This allows for incorporation of the
dopant into the ceria's fluorite structure. The X-ray diffraction patterns of catalysts
1–7 exclusively show the ceria's fluorite structure (not shown). No oxides of the
dopant metals are observed. Still, amorphous, dopant enriched surface phases can
occur for these type of catalysts, and cannot be detected by XRD.[13-15] Indeed, Bera
et al. observed that, in case of copper-doped ceria, both surface enrichment and
bulk incorporation of the copper occurred.[15]
In addition to 1–7, we also prepared Ce0.90Sn0.10O2, Ce0.90In0.10O2 and
Ce0.90Cr0.10O2. However, X-ray diffraction patterns of these samples show
formation of separate dopant phases. The surface area, crystallite sizes and lattice
constants of samples 1–7 are summarized in Table 1. Generally, samples with a
larger crystallite size have a lower surface area. With the exception of 6, the doped
catalysts have a smaller crystallite size than plain ceria.[16, 17] Possibly, the dopant
influences the crystallite growth or the sintering behaviour.[18] The lattice parameter
of CeO2 7 is in good agreement with the accepted value of 5.411 Å.[19] Vegard's
rule states that, in case of doping, the lattice parameter varies linearly with the size
Chapter 3.1 Characterisation
124
of the dopant.[19] Indeed, the lattice parameter of the Fe-CeO2 catalyst is smaller,[20]
and that of the Ca- and Bi-CeO2 catalysts is bigger.[21, 22] The other catalysts show
no substantial changes, but such complex systems can deviate from Vegard's
rule.[13, 15, 23, 24]
Table 1. Catalyst composition, BET surface area, crystallite size and lattice constants
of samples 1–7.
Catalyst/
Composition
Surface area
(m2/g)
Crystallite size
(nm)[a]
Lattice parameter
(Å)
1 Ce0.91Mn0.09O2 56 11 5.407
2 Ce0.90Bi0.10O2 33 18 5.416
3 Ce0.90Cu0.10O2 47 15 5.411
4 Ce0.90Fe0.10O2 50 14 5.404
5 Ce0.92Pb0.08O2 56 13 5.411
6 Ce0.91Ca0.09O2 22 28 5.416
7 CeO2 38 26 5.409 [a] Derived from the peak broadening of the Ce(111) XRD peak using the Scherrer equation.
Reduction rates at 550 °C and temperature programmed reduction.
The dopant type influences the selectivity and activity of the ceria in hydrogen
combustion.[5] In a typical reaction, a mixture containing H2, C3H6 and C3H8 is fed
over the catalyst at 550 °C. The selectivity towards hydrogen combustion is
expressed as the ratio: 1002 total
H
conversion
conversion. We evaluated the rate of combustion of
pure H2 at 550 °C by thermogravimetric analysis (TGA), to gain insight in the
reduction kinetics. In a typical experiment, about 400 mg of sample is placed in a
quartz cup, with a frit bottom that enables gas flow through the sample. The cup is
hung in the thermobalance, heated to 550 °C in synthetic air and kept at this
temperature. The system is flushed with pure argon, after which 60% v/v of H2 is
added, and the weight loss is recorded against time. After 15 min, the system is
flushed again and the sample is reoxidised in 2.5% v/v O2/Ar for 15 min. This
cycle is repeated three times. Figure 3 shows the reduction level against time for
catalyst 1–7. Note that the reduction level is expressed as ‘mg O released per 100
mg sample’ throughout this chapter (‘mass over mass’). The data is presented in
Chapter 3.1 Characterisation
125
several other units (‘mass over volume’, ‘moles over volume’, etc) in Appendix II.
This appendix also contains an assessment of the amount of SOR catalyst which
needs to be added to the dehydrogenation catalyst in the proposed redox process,
using the catalysts activities determined in our studies.
Figures 3A and 3B show that, in general, doping increases the amount of
oxygen released from the catalyst at 550 °C. Furthermore, the reduction typically
proceeds in two stages: a fast initial reduction step and a slower secondary step
(catalysts 1, 2, 3 and 5 in Figure 3A). The catalysts shown in Figure 3B either lack
the fast step (6), the rate of the fast step is lower (7), or the reduction profile
consists of several steps (4).
Red
uct
ion
(mg
O /
100
mg
sam
ple
)
0.5
1.0
1.5
0 500 1000
Time (s)
4 Fe
7 Ce6 Ca
2 Bi
3 Cu5 Pb1 Mn
A
0.5
1.0
1.5
0 500 1000
Time (s)
B
Red
uct
ion
(mg
O /
100
mg
sam
ple
)
0.5
1.0
1.5
0 500 1000
Time (s)
4 Fe
7 Ce6 Ca
2 Bi
3 Cu5 Pb1 Mn
A
0.5
1.0
1.5
0 500 1000
Time (s)
B
Figure 3. Reduction rates of catalysts 1–7 (550 °C, 60% v/v H2/Ar at 200 mL/min
total flow).
The typical TPR profile of ceria also shows two stages.[25-28] The
temperature of 550 °C, where we have determined the reduction rates, lies between
these two TPR reduction stages. Scheme 2 shows the typical TPR profile of plain
ceria 7, together with a typical reduction rate profile (catalyst 3). Two proposed
models explaining the TPR profile are added, labelled ‘model 1’ and ‘model 2’. In
model 1, the two stage process is explained by reduction of surface oxygen (TPR
peak A), and bulk oxygen (peak B).[29] The higher reduction temperature of the
bulk oxygen is ascribed to a limited diffusion of oxygen through the lattice at lower
temperatures. Recently, Trovarelli and co-workers proposed an alternative model
(model 2 in Scheme 2). They noted that above approximately 400 °C, bulk oxygen
diffusion is already that fast that it cannot explain the occurrence of peak B.[26]
Chapter 3.1 Characterisation
126
Smaller ceria particles were shown to reduce at lower temperatures, and the TPR
peaks are correlated to the reduction of small and large ceria crystallites (peak A
and B, respectively). Since small particles are ‘mainly surface’, and large particles
‘mainly bulk’, the two models do not contradict each other. Both predict that a
sample containing small crystallites will result in a relatively large peak A, as little
bulk oxygen is present (model 1), or because of the low reduction enthalpy of the
small crystallites (model 2). Similarly, both models predict a relatively larger peak
B when the ceria consist of large crystallites, since they contain more bulk oxygen
or a higher reduction enthalpy. Indeed, peak A completely disappears for ceria
containing large crystallites.[30]
To assess if the ‘two stages’ of the TGA reduction rates are linked to the
‘two stages’ of the TPR profiles, TPR was performed on catalysts 1–7. In a typical
measurement, about 100 mg of sample was placed on a quartz wool plug in a
quartz reactor. The sample was calcined in situ to 300 °C in 5% v/v O2/Ar, after
which the TPR run was performed in 20 mL/min of 67% H2/Ar, heating to 800 °C
with a ramp of 5 °C/min. H2 uptake was monitored using a thermal conductivity
detector (TCD).
Figure 4 shows the TPR profiles of samples 1–6, with that of ceria 7 added
as a reference. The figure shows that the dopant has little effect on the position of
peak B.[31] Peak A, however, has disappeared completely, together with the
appearance of a peak at lower temperature, labelled peak C (catalyst 6 (Ca) is an
exception, its peak C lies at a higher temperature). We will use this nomenclature
throughout the paper: plain ceria contains peak A (470 °C) and peak B (700 °C),
the doped catalysts contain peak C (at various temperatures) and peak B (at about
700 °C). The labels are shown in Figure 4 for catalysts 7 and 3. Note that the
formation of a low-temperature reduction peak C, together with the disappearance
of peak A of ceria, was also observed for ceria-supported noble metals such as Rh,
Pd, and Pt, or Cu.[25, 32, 33] In those cases, peak C was attributed to reduction of both
the supported oxide and the ceria surface, since the amount of oxygen present in
peak C exceeds the amount present in the supported oxide, and peak A (of ceria),
has disappeared. The phenomena is explained by spill-over of hydrogen, activated
on the (noble) metal, which then reduces the ceria surface. The dissociative
adsorption of hydrogen has been proposed as the rate limiting step of the reduction
Chapter 3.1 Characterisation
127
of pure ceria at temperatures below that of peak A.[34] Peak B is not affected by the
addition of the metal, since at these temperatures the hydrogen dissociation is not
rate limiting anymore.
Our data show that by changing the dopant type, the reduction temperature
(peak C) can be varied from as low as 110 °C (Cu, 3) to as high as 550 °C (Ca, 6,
Figure 3). The Cu, Pb, and Ca doped catalysts show a narrow reduction region, but
Mn, Fe and Bi reduce over a broader temperature range.
For a better comparison with the TGA data, the TPR data was quantified.
Table 2 shows the amount of oxygen released by each catalyst in TPR peak C, and
the total oxygen release (peaks C + B). The data show that, similar to the TGA
data, the total amount of oxygen released from the catalyst is increased in case of
the doped cerias (Table 2). Furthermore, whereas the plain ceria releases most of
its oxygen at high temperatures (peak B, 700 °C), the doped samples (with the
exception of 6) release the majority of their oxygen at lower temperatures (peak C,
110–420 °C).
The data show a variation in the position and relative size of the TPR peaks
for catalysts 1–7. Both models, developed for plain ceria, ascribe these type of
differences to physical properties of the catalyst, such as crystallite size and surface
area.[26, 35] For example, according to model 2, smaller crystallites reduce at lower
temperature (position peak A), and in model 1, the size of peak A is proportional to
the surface area. The crystallite size and surface area vary for catalyst 1–7 as well
(Table 1). One wonders whether the differences in reduction temperature and
oxygen release stem from the dopant type, or from the physical properties of the
catalyst. The data show that the latter cannot explain all the results. The size of
peak C, and the degree of reduction at 550 °C of the doped catalysts 1–6 do not
correlate with the surface area (Table 1, Table 2). For example, the surface area of
catalyst 1 is higher than those of 2–4, but the size of peak C and the degree of
reduction at 550 °C are lower. Furthermore, both models predict that samples with
a large crystallite size have a higher reduction temperature (i.e. a relatively larger
peak B). Since peak B lies around 700 °C, this will result in a lower degree of
reduction at 550 °C. Indeed, the catalysts with the largest crystallite size (6 and 7),
do show the highest reduction temperature (Peak C), the lowest initial reduction
rates, and a lower degree of reduction at 550 °C (Figure 3 and 4). However, when 3
Chapter 3.1 Characterisation
128
was subjected to prolonged reduction, the crystals had sintered from ~15 nm to ~42
nm. This is well above the crystallite size of catalysts 6 and 7 (~30 nm), but still
peak C of the sintered catalyst 3 lies at a low temperature (~150 °C, not shown),
and the degree of reduction at 550 °C remained as high as that of fresh 3. Also, the
crystallite size of catalyst 2 (Bi) is larger than that of 5 (Pb) (~18 nm vs. ~13 nm,
respectively), but the reduction temperature is lower (~250 °C vs. ~300 °C), and
the degree of reduction at 550 °C higher. Clearly, the crystallite size is not the sole
cause of variation in reduction temperature (peak C) and degree of reduction at
550 °C between samples. The type of dopant is also important.
Both the TGA and the TPR data show that the catalyst reduction is a two
step process (Figures 3 and 4). The TGA reduction rates are determined at 550 °C,
i.e. in between peak C and peak B. The ‘fast step’ of the reduction profile may
therefore be correlated to TPR peak C, and the ‘slow step’ to peak B. In Table 2,
the amount of oxygen released in this ‘fast step’ is compared to the size of peak C.
The trends are only roughly comparable. However, the position of TPR peak C,
and the reduction rates as determined by TGA are correlated (Figures 3 and 4).
Peak C of catalysts 4, 6, and 7, lie at temperatures close to 550 °C, and show the
lowest initial reduction rate. Moreover, catalyst 6, with peak C positioned at
550 °C, lacks the fast reduction step. Finally, peak C of catalyst 4 is spread out
over a broad temperature range (~150–460 °C), and also exhibits several reduction
rates. The varying reduction temperature of peak C reflects the (total) energy
needed to reduce the catalysts. Therefore, the smaller the difference between peak
C and the TGA reduction temperature (550 °C), the slower the TGA reduction rate.
Indeed, peak C of catalysts 1–3 and 5 (Figure 3B) are farthest removed from
550 °C, and contain the highest TGA reduction rates (no discrimination can be
made between them). Furthermore, the broad TPR peaks of 1 and 2 do not translate
into multiple reduction rates, as seen for catalyst 4. The large temperature
difference between peak C and 550 °C obscures any difference in reduction rates.
Chapter 3.1 Characterisation
129
0 200 400 600 800Temperature (°C)
A
B
TPR data (7)
B
AReduction ratedata (3, 550 °C)
Time
Model 1:Surface/bulk O
B
B
Model 2:A
A
Crystallite size
0 200 400 600 800Temperature (°C)
A
B
TPR data (7)
B
AReduction ratedata (3, 550 °C)
Time
Model 1:Surface/bulk O
B
B
Model 2:A
A
Crystallite size
Scheme 2. Typical TPR and reduction rate profiles, together with the two models
proposed to explain the TPR data.
Chapter 3.1 Characterisation
130
800
0 200 400 600 800
2, Bi
TC
D s
igna
l (au
) 800
0 200 400 600 800
1, Mn
800
0 200 400 600 800
AB
C 3, Cu
Temperature (°C)
0 200 400 600 800
6, Ca800
0 200 400 600 800
800 4, Fe
0 200 400 600 800
800 5, Pb
Temperature (°C)
TC
D s
igna
l (au
)T
CD
sig
nal (
au)
800
0 200 400 600 8000 200 400 600 800
2, Bi
TC
D s
igna
l (au
) 800
0 200 400 600 800
1, Mn
800
0 200 400 600 800
AB
C 3, Cu
Temperature (°C)
0 200 400 600 800
6, Ca800
0 200 400 600 800
800 4, Fe
0 200 400 600 800
800 5, Pb
Temperature (°C)
TC
D s
igna
l (au
)T
CD
sig
nal (
au)
Figure 4. TPR of doped cerias 1–6, with CeO2 7 added as reference. Conditions:
5 °C/min to 800 °C in 67% v/v H2/Ar flowed at 20 mL/min. Samples were calcined in situ
to 300 °C (5% v/v O2/Ar, 30 min hold at 300 °C) prior to analysis (200 °C in case of the
CeO2 sample).
Chapter 3.1 Characterisation
131
Table 2. Quantitative TPR and TGA data of catalysts 1–7.
Catalyst/
Composition
Size peak C
(mg O/100 mg
sample)[a,b,c]
Total oxygen release
(mg O/100 mg
sample)[d]
Size ‘fast reduction part’
(mg O/100 mg sample,
TGA)[e]
1 Ce0.91Mn0.09O2 0.9 1.2 0.6
2 Ce0.90Bi0.10O2 1.2 2.0 1.2
3 Ce0.90Cu0.10O2 1.2 1.7 1.0
4 Ce0.90Fe0.10O2 1.2 1.9 1.3
5 Ce0.92Pb0.08O2 1.2 1.6 0.8
6 Ce0.91Ca0.09O2 0.5 1.7 0.3
7 CeO2 0.4 1.2 0.3 [a] Data obtained by calibrating the TCD detector using a CuO standard. The peak area of
this standard is integrated and the area is correlated to the amount of oxygen present in the
CuO. [b] Whether normalised for surface area or weight, the doped cerias release more
oxygen compared to the plain ceria (7). When normalised for surface area, the trends are
the same, except for a higher oxygen release for Bi (2) and Ca (6), since their surface are is
low. [c] Peak A in case of catalyst 7 (ceria). [d] When normalised for surface area instead of
weight, the Mn 1 and Pb 5 doped catalysts release less oxygen compared to plain ceria.
Note that catalysts 1 and 5 have the highest surface areas. Since the total oxygen release
includes the oxygen from the bulk of the sample, normalising for surface area may be less
valid. [e] Values are the average of three measurements, two in case of catalyst 6. Since
catalysts 4 and 6 do not contain a clear ‘fast’ and ‘slow’ part, the value given is the amount
of oxygen released after 1000 s.
Both the TGA and TPR data show that generally, the doped cerias release
more oxygen than the plain ceria (Figure 3, Table 2). Doping ceria with atoms of
lower valence, such as Ca2+, or Cu2+, can increase the oxygen flux through the
lattice.[19, 36] Another possible source for the extra oxygen are separate phases of the
dopant atoms. Supported crystalline oxides can be excluded, since these were not
detected by XRD. However, an amorphous surface phase, enriched with the
dopant, could be present.[13, 14, 33] Indeed, except for the Cu-doped catalyst 3,
peak C can be attributed solely to the reduction of the dopant oxide. That is, a
separate oxide of the dopant in its highest or most common oxidation state would
contain enough oxygen to account for the size of peak C. Therefore, the catalyst
could consist of a dopant oxide layer on top of the ceria. Indeed, Tang et al., and
Chapter 3.1 Characterisation
132
Zheng et al. showed that, when impregnating ceria supports, the ceria can facilitate
relatively high amounts (about 12 mol %) of amorphous dopant.[33, 37] However,
this was only the case when calcination temperatures were kept relatively low
(≤ 600 °C), and our catalysts were calcined at 700 °C, for 5 h. Furthermore, we did
not impregnate the ceria surface, but co-melted the metal nitrates. Because of this,
and since the metals we used are common ceria dopants, it is unlikely that all the
dopant atoms will be present at the surface.[16, 38, 39]
Besides the quantitative data, the TPR data allows some more evaluation of
the catalysts' uniformity. The reduction temperature (TPR peak C) can vary with
the oxidation state and crystallite size of the oxide.[30, 40, 41] The broad reduction
range of the Mn, Fe and Bi-doped catalysts indicates that they are more
heterogeneous concerning one or both of these factors. Indeed, X-ray photoelectron
spectroscopy (XPS) analysis of 4 shows that the iron is present as both Fe2+ and
Fe3+ (not shown). The spread in crystallite size can be assessed by transmission
electron microscopy (TEM). We have measured the catalyst 4 (Fe, broad peak C)
and plain ceria 7 (narrow peak A). The TEM images show that the average particle
size of 7 is larger than that of 4, which is in agreement with the XRD data (see
Figure 5, Table 1). Neither catalyst, however, is uniform in size. Both contain
relatively large clustered, and smaller unclustered crystallites. Due to the local
nature of the TEM measurements, the average size distribution cannot be
determined. We have, however, analysed two TEM images as an indication of the
spread in crystallite size. This showed clustered crystallites of about 12 nm to 18
nm for 4, and about 12 nm to 36 nm for 7, and unclustered crystallites of about
5-10 nm for both catalysts.
Chapter 3.1 Characterisation
133
7, Ce 4, Fe
100 nm 100 nm
7, Ce 4, Fe
100 nm 100 nm
Figure 5. TEM image of catalyst 7 (left) and 4 (right). Both images were taken at
x 22,000 magnification.
The Cu, Pb and Ca-doped catalysts show fairly narrow TPR reduction
regions, contrary to 1, 2 and 4. Moreover, peak A (of plain ceria), is not observed.
Still, plain (undoped) ceria crystallites may be present, and be reduced by hydrogen
spilled-over from the doped ceria crystallites.[42] This can be excluded however, for
the Cu- and Ca-doped cerias. When we performed TPR on a 1:1 physical mixture
of the Cu-CeO2 (3) and plain ceria, both peak C and peak A are observed (not
shown). Therefore, the Cu-doped catalyst will not reduce separate crystallites of
plain ceria by hydrogen spill-over, and the presence of pure CeO2 in 3 is unlikely.
For the Ca-CeO2 catalyst, peak C lies above the peak A of plain ceria (550 °C and
470 °C, respectively), and peak A is not present as well. This excludes the
reduction of plain ceria by hydrogen spill-over from Ca, and we conclude that no
plain ceria crystallites are present.
Still, we cannot discriminate between doped ceria, a ceria surface enriched
in dopant, or small metal oxide clusters below the XRD detection limit, as the
cause of TPR peak C. Whatever the origin, the TPR and TGA data show that the
amount of oxygen released from the catalyst is increased by doping, and that the
reduction temperature is shifted over a broad range (~ 400 °C), depending on the
dopant type. The latter affects the combustion rate at 550 °C. This tuning of
reduction rate and reduction degree is important for the catalysts' selectivity and
activity.
Chapter 3.1 Characterisation
134
Selectivity and activity. The average selectivity and activity of catalysts
1–7 was determined over 16 redox cycles at 550 °C (Table 3). Each cycle consists
of an oxidation step (18 min, 1% v/v O2 in Ar), a purge (4 min, pure Ar), a
reduction step (10 min, 4:1:1% v/v C3H8:C3H6:H2 in Ar), and a second purge (4
min, pure Ar). The activity of the catalysts is given by two parameters. The first is
the so-called ‘oxygen demand’, which is the amount of oxygen used in the
oxidation step to refill the reduced lattice vacancies and to combust coke, if
present. Thus, the oxygen demand represents both selective and unselective
processes. The second is the ‘hydrogen activity’, representing the percentage of the
hydrogen feed which is combusted by each catalyst.
Table 3 shows that catalysts 4 (Fe) and 7 (plain ceria) are the least selective.
These also show a relatively high reduction temperature (TPR peak C), and a
relatively slow reduction rate at 550 °C (Figures 3 and 4). Possibly, a pre-requisite
of a selective catalyst is that it has a high hydrogen combustion rate, leaving little
change for the hydrocarbons to be converted. Indeed, catalysts 1, 2, 3 and 5 show
the highest selectivities, and have the highest reduction rates at 550 °C, and the
lowest TPR reduction temperatures. Still, in the TPR and TGA experiments,
hydrogen is the sole reducing agent. The hydrocarbon conversion is part of the
selectivity equation, and so it should be taken into account. The interactions
between the catalysts and the hydrocarbons however, can be very complex. Several
processes can occur simultaneously, such as coking, combustion into CO or CO2,
dehydrogenation, hydrogenation, and cracking into smaller hydrocarbon
fragments.[5] Furthermore, we showed that certain metals can be active both in the
oxidised and in the reduced form.[5] The latter occurs at the end of the reductive
cycle, when all the lattice oxygen is spent, and the exposed metal atoms can
interact with the hydrocarbons. The Mn-doped catalyst 1 is a striking example: its
initial selectivity is high, but at the end of the reductive cycle, it starts to coke the
hydrocarbons (not shown).
There is little correlation between the catalyst activity and the degree of
reduction determined by TGA and quantitative TPR (‘hydrogen activity’, Table 3).
For example, the difference in hydrogen activity between the active catalysts (2
and 5) and the less active catalysts (1 and 3), is much larger than the differences in
the quantitative TPR or TGA data (Table 3). However, the Ca doped catalyst 6 is
Chapter 3.1 Characterisation
135
(virtually) inactive, and has a high TPR reduction temperature, which results in a
low TGA reduction rate and low degree of reduction at 550 °C .
The oxygen demand determined in the catalytic experiments is generally
higher than the degree of reduction as determined by the quantitative TPR and
TGA experiments (Table 3). Indeed, the TPR and TGA experiments are performed
using only hydrogen as the reducing agent. Conversely, the oxygen demand
reflects both the reduction by hydrogen and the coking and combustion of the
hydrocarbons. Because of this, catalyst 4 (Fe) has a high oxygen demand: its low
selectivity results in large amounts of coking, and a lot of oxygen is used in the
oxidation step to combust this coke. Catalyst 2 (Bi) is more selective, but does
combust part of the hydrocarbon feed, which increases its oxygen demand, and the
Mn-doped catalyst 1 has a high initial selectivity, but cokes the hydrocarbons at the
end of the reductive cycle. Catalyst 5 (Pb), is both active and selective. Indeed, its
oxygen demand is comparable to the degree of reduction determined by TPR and
TGA (Table 3).
Cha
pter
3.1
Cha
ract
eris
atio
n
136
T
able
3. C
atal
ytic
dat
a an
d de
gree
of
redu
ctio
n de
term
ined
by
TP
R a
nd T
GA
.
Cat
alys
t/
Com
posi
tion
Sel
ectiv
ity
(%)[a
]
Oxy
gen
dem
and
(mg
O /
100
mg)
Hyd
roge
n ac
tivity
(%H
2 co
mbu
sted
)
Are
a T
PR
Pea
k C
(mg
O /
100
mg)
Siz
e ‘f
ast r
educ
tion’
par
t TG
A
(mg
O /
100
mg)
[b]
1 C
e 0.9
1Mn 0
.09O
2 93
1.
9 5
0.9
0.6
2 C
e 0.9
0Bi 0
.10O
2 77
1.
7 33
[c]
1.2
1.2
3 C
e 0.9
0Cu 0
.10O
2 89
1.
4 7
1.2
1.0
4 C
e 0.9
0Fe 0
.10O
2 L
ow[d
] 3.
0 -
1.2
1.3
5 C
e 0.9
2Pb 0
.08O
2 92
1.
1 46
[c]
1.2
0.8
6 C
e 0.9
1Ca 0
.09O
2 n.
a.[e
] 0.
2 -
0.5
0.3
7 C
eO2
Low
[d]
0.6
- 0.
4[f]
0.3
[a] S
elec
tivity
for
hyd
roge
n co
mbu
stio
n fr
om a
mix
ture
with
pro
pene
and
pro
pane
(1:
1:4,
res
pect
ivel
y) a
t 550
°C
, ave
rage
of
15 r
edox
cyc
les.
[b
] Val
ues
are
the
aver
age
of t
hree
mea
sure
men
ts,
two
in c
ase
of c
atal
yst
6. S
ince
cat
alys
ts 4
and
6 d
o no
t co
ntai
n a
clea
r ‘f
ast’
and
‘sl
ow’
part
, th
e va
lue
is t
he a
mou
nt o
f ox
ygen
rel
ease
d af
ter
1000
s. [c
] T
hese
cat
alys
ts c
onve
rt 1
00%
of
the
hydr
ogen
fee
d at
the
beg
inni
ng o
f th
e
redu
ctiv
e cy
cle.
Thi
s do
es n
ot a
ffec
t the
tota
l act
ivity
, how
ever
, sin
ce a
ll of
thes
e ca
taly
sts
are
depl
eted
bef
ore
the
end
of th
e re
duct
ion
cycl
e.
[d] N
o va
lue
can
be c
alcu
late
d si
nce
hydr
ogen
is f
orm
ed f
rom
cok
ing
of th
e hy
droc
arbo
ns, r
esul
ting
in a
neg
ativ
e hy
drog
en c
onve
rsio
n. [e
] Not
appl
icab
le, t
he c
atal
yst i
s in
activ
e. [f
] Pea
k A
in c
ase
of th
e pl
ain
ceri
a 7.
Chapter 3.1 Characterisation
137
Conclusions The redox properties of ceria can be tuned by doping. A set of six different
dopant metals shows a variety in degree of reduction, temperature onset of
reduction and reduction rate. The doped cerias generally display an increased
degree of reduction. Doping with Bi, Cu, Fe, Pb or Ca results in and oxygen release
of 1.6 to 2.0 mg oxygen per 100 mg sample, whereas plain ceria releases
1.2 mg O/100 mg. This is important for catalytic reactions where the lattice oxygen
is used as a reagent, such as the selective combustion of hydrogen shown here.
Furthermore, the oxygen is generally released at lower temperatures. The onset of
reduction can be tuned from ~110 °C (Cu-CeO2) to ~550 °C (Ca-CeO2). Since the
onset of reduction for plain ceria is ~470 °C, it can be tuned both up and down. The
influence of crystallite size and surface area on the reduction onset and degree of
reduction of ceria crystallites are overruled by the influence of the dopant metal.
The large variation in reduction temperature of the catalysts results in different
hydrogen combustion rates at 550 °C. Catalysts with a relatively high reduction
temperature, such as Fe-CeO2 and plain ceria, show slower reduction rates. These
samples also show a low selectivity towards hydrogen combustion. Doping with
Mn, Bi, Cu, and Pb results in high hydrogen reduction rates, and high selectivities.
To fully address the selectivity however, the interactions with the hydrocarbons
have to be taken into account.
Chapter 3.1 Characterisation
138
Experimental Section Materials and instrumentation. Chemicals were purchased from Sigma-
Aldrich, Merck, The British Drug Houses Ltd. or Koch-Light Laboratories Ltd and
used as received. Gasses were purchased from Praxair and had a purity of 99.5% or
higher. The O2, He, Ar and N2 streams were purified further over molsieves and/or
BTS columns. All gas flows were controlled by Bronkhorst mass flow controllers.
The specific surface areas were measured by N2 adsorption at 77 K on a
Sorptomatic 99 (CE Instruments) and evaluated using the BET equation. Powder
X-ray diffraction measurements were performed using a Philips PW-series X-ray
diffractometer with a Cu tube radiation source (λ = 1.54 Å), a vertical axis
goniometer and a proportional detector. The 2θ detection measurement range was
10 ° – 93 ° with a 0.02 ° step size and a 5 second dwell time. Lattice constants and
crystallite sizes were obtained after Rietveld refinement (structure fit) using
PANalytical's X'pert software package. TEM was performed on a Philips CM12
microscope at 80 kV. The thermogravimetric measurements were performed on a
Setaram TG 85 thermobalance (accuracy 10 micrograms). X-ray photoelectron
spectra were recorded on a Kratos HSi spectrometer equipped with a charge
neutraliser and monochromated Al K X-ray source (1486.61 eV) operating at
144 W. Spectra were recorded with a pass energy of 40 eV at normal emission, and
energy referenced to the valence band and adventitious carbon. Analysis was
conducted using CasaXPS Version 2.3.15.
Procedure for catalyst synthesis. The synthesis procedure was described
in detail previously.[5] The appropriate metal nitrates are weighed into a crucible
and mixed with a spatula. A maximum of six crucibles are placed into a vacuum
oven set at 115 °C. The precursors were allowed to melt for about 5 min, after
which the pressure was lowered to < 10 mbar in about 15 minutes. The latter was
performed carefully to prevent vigorous boiling. After 4h, the precursors were
placed in a muffle oven and calcined for 5h at 700 °C in static air (ramp rate:
300 °C/h). The resulting solid was pulverized, ground and sieved in fractions of
125–212 µm (selectivity assessment) and < 125 µm (TPR, TGA, XRD, XPS and
BET measurements). The final metal concentration was calculated from the
amount of precursor weighed in, corrected for the water content by ICP.[5] The Bi
(2) and Pb (5) doped catalysts were prepared using an adjusted experimental, since
Chapter 3.1 Characterisation
139
their nitrate precursors demix easily. To prevent this, a few drops of water were
added to the precursors and they were heated on a heating plate under continuous
stirring. When the nitrates had melted, the crucible was placed in the vacuum oven
and the pressure was lowered immediately and quickly (in about 5 minutes). From
here on the standard synthesis procedure was followed.
Procedure for TGA measurements. In a typical experiment, 400 mg of
sample (<125 μm) was placed in a quartz cup with a quartz frit bottom and placed
inside a water-cooled oven. The sample was heated to 550 °C in a 120 mL/min
flow of synthetic air (ramp rate: 10 °C/min). When the set temperature was
reached, the thermobalance was evacuated to remove the oxygen and refilled with
Ar (80 mL/min). The mass of the sample was recorded every 0.2 s. When the mass
signal had stabilised, 120 mL/min of hydrogen was added, resulting in a 60% v/v
H2/Ar flow, and the mass of the sample was recorded during 15 minutes. The
optimal hydrogen concentration was determined by measuring the reduction rate of
Pd-CeO2 for H2/Ar mixtures with increasing H2 concentration (this sample has a
high reduction rate). From H2 concentrations of 30% v/v onward the reduction rate
did not increase anymore, and 60% v/v was chosen for the actual rate
determinations. The reduction rates of catalysts 1 and 2 were determined on 150
mg instead of 400 mg of sample. To check if the reduction profile is affected by
this change in space velocity, 150 and 400 mg portions of the Pd-CeO2 catalyst
were tested and compared. At the scaling used in Figure 3, these profiles were
identical.
The TGA system is designed to add the H2 gas as instantaneously as
possible. This is achieved by continuously flowing the H2 gas to a vent. To start the
measurement, this flow was directed to the thermobalance by switching a valve,
positioned close to the sample cup (see Scheme 3). Switching on the hydrogen
causes a change in buoyancy and drag which results in an apparent mass change of
the sample. Therefore, the total weight loss during the reduction was determined
from the weight of the sample just before the hydrogen gas is switched on, and just
after it was switched back to the vent. After the reduction step, the hydrogen was
flushed from the system (monitored by MS), and the sample was reoxidised for 15
min in 2.5% v/v O2/Ar. This process was repeated three times.
Chapter 3.1 Characterisation
140
Referencecup
Argon inlet
H2 inlet
Vent
ThermocoupleOven
Sample cup
Balance
Vent
Referencecup
Argon inlet
H2 inlet
Vent
ThermocoupleOven
Sample cup
Balance
Vent
Scheme 3. Schematic of the TGA setup. Arrows denote the direction of the gas
flow.
Procedure for TPR experiments. TPR experiments were performed on a
standard TPR set up built in house, equipped with a TCD detector. In a typical
experiment, 100 mg sample (<125 µm) was placed on top of a quartz wool plug in
a 4 mm id quartz reactor. The samples were calcined in-situ to 300 °C (10 °C/min,
30 min hold time), in 5% v/v O2/Ar at 50 mL/min total flow. The samples were
allowed to cool overnight, after which the detector is allowed to equilibrate for
about 1.5 h in a 67% v/v H2/Ar at 20 mL/min total flow. For the actual TPR
measurement, the sample is heated to 800 °C with a ramp of 5 °C/min. Data is
collected with 12s intervals.
Procedure for XPS experiments. XPS was performed on 50 mg sample.
The electron analyser pass energy was 160 eV for wide scans and 40 eV for high
resolution spectra. Compositions were corrected using the appropriate elemental
response factors on spectra following a Shirley background-subtraction.
Procedure for selective hydrogen combustion experiments. Activity and
selectivity were determined on a fully automated system built in house, which was
described in detail previously.[5] In a typical experiment, about 250 mg of sample
Chapter 3.1 Characterisation
141
(125–212 μm) was placed on a quartz wool plug in a 4 mm id quartz reactor. The
reactor was placed in a water cooled oven and heated to 550 °C at 1200 °C/h, under
oxygen flow. At this temperature, redox cycling was begun. The selectivity was
determined by GC during the 10 minute reduction in 4:1:1% v/v C3H8:C3H6:H2 in
Ar (total flow 50 mL/min), with 5 mL/min of N2 added as internal standard. The
4:1:1 ratio of reductive gases is chosen since this is the equilibrium mixture of a
conventional dehydrogenation catalyst.[8] The gas hourly space velocity (GHSV) is
13200 / h (at the typical bed volume of 0.25 cm3 and the reduction cycle's total
flow of 55 mL/min). The weight hourly space velocity (WHSV) is 1.2 / h, and is
calculated from the weight of C3H8 + C3H6 + H2 per h per the weight of the
catalyst. After a 4 min purge step (pure Ar), the sample was reoxidised for 18
minutes in 1% v/v O2 in Ar (50 mL/min total flow). The redox cycle is completed
by another purge step in pure Ar. The selectivity is determined as the ratio H2
conversion:total conversion. Activity is determined as the amount of oxygen taken
up during the oxidation step, determined by MS (‘oxygen demand’), and as the
percentage of the hydrogen feed combusted, determined by GC (‘hydrogen
activity’). Both selectivity and activity are averaged over fifteen redox cycles
Acknowledgements We thank Dr. J.W.M. van Lent from Wageningen University for the TEM
measurements, Dr. M.C. Mittelmeijer-Hazeleger the BET surface area
measurements, A.C. Moleman for allowing use of and giving instructions on the
XRD instrument, Dr. Adam Lee from the University of York for XPS analysis and
NWO-ASPECT for financial support and feedback.
Chapter 3.1 Characterisation
142
References [1] G. Parkinson, Chem. Eng. Prog., 2004, 100, 8. [2] J. Plotkin and E. Glatzer, Eur. Chem. News, 2005, 82, 20. [3] N. Alperowicz, Chem. Week, 2007, 169, 27. [4] N. Alperowicz, Chem. Week, 2006, 168, 17. [5] J. H. Blank, J. Beckers, P. F. Collignon, F. Clerc and G. Rothenberg, Chem. Eur.
J., 2007, 13, 5121. [6] G. Rothenberg, E. A. de Graaf and A. Bliek, Angew. Chem., Int. Ed., 2003, 42,
3366. [7] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 9. [8] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 1. [9] J. G. Tsikoyiannis, D. L. Stern and R. K. Grasselli, J. Catal., 1999, 184, 77. [10] C. H. Lin, K. C. Lee and B. Z. Wan, Appl. Catal. A: Gen., 1997, 164, 59. [11] L. M. van der Zande, E. A. de Graaf and G. Rothenberg, Adv. Synth. Catal., 2002,
344, 884. [12] J. Beckers, F. Clerc, J. H. Blank and G. Rothenberg, Adv. Synth. Catal., 2008, 350,
2237. [13] P. J. Scanlon, R. A. M. Bink, F. P. F. van Berkel, G. M. Christie, L. J. van
IJzendoorn, H. H. Brongersma and R. G. van Welzenis, Solid State Ionics, 1998, 112, 123.
[14] G. Rothenberg, E. A. de Graaf, J. Beckers and A. Bliek, Catal. Org. React., 2005, 104, 201.
[15] P. Bera, K. R. Priolkar, P. R. Sarode, M. S. Hegde, S. Emura, R. Kumashiro and N. P. Lalla, Chem. Mater., 2002, 14, 3591.
[16] M. Jobbágy, F. Mariño, B. Schöbrod, G. Baronetti and M. Laborde, Chem. Mater., 2006, 18, 1945.
[17] P. Bera, S. T. Aruna, K. C. Patil and M. S. Hegde, J. Catal., 1999, 186, 36. [18] R. T. Baker, S. Bernal, G. Blanco, A. M. Cordón, J. M. Pintado, J. M. Rodríguez-
Izquierdo, F. Fally and V. Perrichon, Chem. Commun., 1999, 149. [19] M. Mogensen, N. M. Sammes and G. A. Tompsett, Solid State Ionics, 2000, 129,
63. [20] E. Aneggi, C. de Leitenburg, G. Dolcetti and A. Trovarelli, Catal. Today, 2006,
114, 40. [21] A. E. C. Palmqvist, M. Wirde, U. Gelius and M. Muhammed, Nanostruct. Mater.,
1999, 11, 995. [22] V. A. Sadykov, T. G. Kumetsova, G. M. Alikina, Y. Frolova, A. I. Lukashevich,
Y. V. Potapova, V. S. Muzykantov, V. A. Rogov, V. V. Kriventsov, D. I. Kochubei, E. M. Moroz, D. I. Zyuzin, V. I. Zaikovskii, V. N. Kolomiichuk, E. A.
Chapter 3.1 Characterisation
143
Paukshtis, E. B. Burgina, V. V. Zyryanov, N. F. Uvarov, S. Neophytides and E. Kemnitz, Catal. Today, 2004, 93-95, 45.
[23] X. Q. Wang, J. A. Rodriguez, J. C. Hanson, D. Gamarra, A. Martínez-Arias and M. Fernández-García, J. Phys. Chem. B, 2005, 109, 19595.
[24] X. Q. Wang, J. A. Rodriguez, J. C. Hanson, D. Gamarra, A. Martínez-Arias and M. Fernández-García, J. Phys. Chem. B, 2006, 110, 428.
[25] M. Boaro, M. Vicario, C. de Leitenburg, G. Dolcetti and A. Trovarelli, Catal. Today, 2003, 77, 407.
[26] F. Giordano, A. Trovarelli, C. de Leitenburg and M. Giona, J. Catal., 2000, 193, 273.
[27] M. Salazar, D. A. Berry, T. H. Gardner, D. Shekhawat and D. Floyd, Appl. Catal. A: Gen., 2006, 310, 54.
[28] Y. Zhang, S. Andersson and M. Muhammed, Appl. Catal. B: Environ., 1995, 6, 325.
[29] H. C. Yao and Y. F. Yao, J. Catal., 1984, 86, 254. [30] E. Aneggi, M. Boaro, C. de Leitenburg, G. Dolcetti and A. Trovarelli, J. Alloy
Compd., 2006, 408, 1096. [31] Catalyst 6 Ca does feature a shift of peak B towards lower temperatures. However,
this also occurs for a physical mixture of Cu-CeO2 3, and plain ceria 7 (not shown). When these catalysts are measured separately, peak B does not shift. Therefore, the peak shift is not correlated to physical properties of the catalysts such as crystallite size or bulk incorporation of the dopant.
[32] C. de Leitenburg, A. Trovarelli and J. Kašpar, J. Catal., 1997, 166, 98. [33] X. L. Tang, B. C. Zhang, Y. Li, Y. D. Xu, Q. Xin and W. J. Shen, Appl. Catal. A:
Gen., 2005, 288, 116. [34] J. Elfallah, S. Boujana, H. Dexpert, A. Kiennemann, J. Majerus, O. Touret, F.
Villain and F. Lenormand, J. Phys. Chem., 1994, 98, 5522. [35] M. F. L. Johnson and J. Mooi, J. Catal., 1987, 103, 502. [36] G. S. Li, R. L. Smith and H. Inomata, J. Am. Chem. Soc., 2001, 123, 11091. [37] X. C. Zheng, X. L. Zhang, X. Y. Wang, S. R. Wang and S. H. Wu, Appl. Catal. A:
Gen., 2005, 295, 142. [38] A. E. C. Palmqvist, E. M. Johansson, S. G. Jaras and M. Muhammed, Catal. Lett.,
1998, 56, 69. [39] A. Tschöpe, W. Liu, M. Flytzani-Stephanopoulos and J. Y. Ying, J. Catal., 1995,
157, 42. [40] F. Kapteijn, L. Singoredjo, A. Andreini and J. A. Moulijn, Appl. Catal. B:
Environ., 1994, 3, 173. [41] F. Kapteijn, A. D. van Langeveld, J. A. Moulijn, A. Andreini, M. A. Vuurman, A.
M. Turek, J. M. Jehng and I. E. Wachs, J. Catal., 1994, 150, 94. [42] T. Inui, Stud. Surf. Sci. Catal., 1993, 77, 17.
Chapter 3.1 Characterisation
144
145
3.2
Redox properties of doped and supported copper-
ceria catalysts
H2
Heat
TPR: 110 °C 150 °C 190 °C
H2
HeatCuO/CeriaCe0.9Cu0.1O2
H2
Heat
TPR: 110 °C 150 °C 190 °C
H2
HeatCuO/CeriaCe0.9Cu0.1O2
Copper-doped ceria can be applied in selective hydrogen combustion, a novel
process for oxidative dehydrogenation. We describe the structural changes the
catalyst undergoes upon the inherent redox cycling, and show that highly dispersed
copper is the active phase.
Part of this work has been published as:
'Redox properties of doped and supported copper-ceria catalysts', Jurriaan Beckers
and Gadi Rothenberg, Dalton Trans. 2008, 6573.
Chapter 3.2 Characterisation
146
Abstract
Copper-doped ceria catalysts feature in a variety of catalytic reactions. One
important application is selective hydrogen combustion via oxygen exchange,
which forms the basis of cyclic oxidative dehydrogenation. This paper describes
the synthesis of monophasic (doped) and biphasic (supported) Cu/ceria catalysts,
that are then characterized using a combination of temperature programmed
reduction (TPR) and X-ray diffraction (XRD) methods. The catalysts are analysed
both as fresh samples and after redox cycling at 550 – 800 °C. TPR and XRD
characterisation clarify the role of the active sites on the catalyst surface and the
copper-ceria interactions. Depending on the catalyst type, reduction occurs at
~110 °C, ~150 °C, or ~190 °C. The reduction at 110 °C is ascribed to highly
dispersed copper species doped in the ceria lattice, and that at 190 °C to CuO
crystallites supported on ceria. Remarkably, both types converge to the 150 °C
feature after redox cycling. The reduction temperature of the doped catalyst
increases after redox cycling, indicating that stable Cu clusters form at the surface.
Conversely, the reduction temperature of the ‘supported’ catalyst decreases after
redox cycling, and the CuO crystallites disappear. With this knowledge, a copper–
doped ceria catalyst is analysed using TPR and XRD after application in selective
hydrogen combustion (16 consecutive redox cycles at 550 °C). No CuO crystallites
are observed, and the sample reduces at ~110 °C. This suggests that copper-doped
ceria is the active oxygen exchange phase in selective hydrogen combustion.
Calorimetric measurements show that the hydrogen combustion by doped cerias
can indeed be a net exothermic process. The enthalpy of reduction of a 7 mol%
Cu–doped ceria is -5 kJ/mol.
Chapter 3.2 Characterisation
147
Introduction Copper is essential to human life, as a cofactor of several redox enzymes
that control cell respiration, free-radical defence mechanisms, and blood
metabolism.[1, 2] Copper containing catalysts are also used in various industrial
processes, including methanol synthesis,[3] the water-gas shift reaction,[4] and the
production of acetaldehyde.[5] Notably, copper nanoclusters catalyze C–C coupling
reactions, forming an inexpensive alternative to noble metal catalysts.[6, 7] One
interesting application of copper is as a catalyst dopant or promoter. In this context,
cerium oxide, or ceria, is a suitable partner.[8-11] Ceria is often used in oxidation
reactions because of the facile Ce3+ Ce4+ + e– redox cycle.[12] The
incorporation of copper can increase both activity and selectivity, for example in
selective CO oxidation, methane combustion or NO reduction.[9, 13, 14]
Our interest in copper-doped ceria started within a project on selective
hydrogen combustion,[15] an important reaction that enables a new type of oxidative
dehydrogenation cycle (Figure 1).[16] In this approach, propane, for example, is
dehydrogenated over a conventional Pt/Sn or Cr catalyst, and the hydrogen by-
product is then selectively burned from the mixture at high temperatures, using the
ceria lattice oxygen. The hydrogen combustion shifts the equilibrium towards the
products side. It also generates heat, aiding the endothermic dehydrogenation
reaction. Following the reduction of the ceria lattice, the oxygen vacancies are re-
filled using air, creating a cyclic redox process. The use of ceria as a solid oxygen
reservoir enables separate optimisation of the dehydrogenation and selective
oxidation reactions, and prevents the mixing of molecular oxygen with the
reductive gases. Ceria itself, however, is not selective, but its properties can be
tuned by doping. We have shown that the activity and selectivity of the doped ceria
lattice depend strongly on the dopant type and amount.[17, 18] Supported transition
metal oxides also show high selectivities for hydrogen combustion, but they are not
stable under the redox cycling.[19, 20] Doped cerias are more stable, but under too
severe conditions, surface segregation and formation of copper-oxide can occur.
In this paper, we study the behaviour of copper-doped ceria under
simulated and real reaction conditions, to investigate the interactions on the
surface. We use X-ray diffraction (XRD) and temperature-programmed reduction
(TPR),[21, 22] to identify typical copper-ceria surface species. With this information,
Chapter 3.2 Characterisation
148
one can monitor the stability and predict the active phase of the copper-doped
catalysts under real reaction conditions.
Alkane
Alkene
Energy
H2
H2O + Ce2O3
2 CeO2
dehydrogenationcatalyst
Alkane
Alkene
Energy
H2
H2O + Ce2O3
2 CeO2
dehydrogenationcatalyst
dehydrogenationcatalyst
Figure 1. Catalytic cycle for oxidative dehydrogenation using ceria as a solid
oxygen reservoir.
Results and Discussion
Catalyst preparation and characterization. All samples were prepared by
melting mixtures of the metal nitrate hydrates under reduced pressure, followed by
calcination in static air at 700 °C for 5 h. The control sample 1 contained pure
CeO2, while samples 2 and 3 contained 10 mol% dopant (1:9 Cu:Ce molar ratio).
Figure 2 shows sample 3 at various stages of the preparation. Image A shows the
nitrate salts after weighing and mixing. These precursors form transparent, brightly
coloured liquids when heated to 120 °C (image B). When the pressure is lowered to
10 mbar, a solid forms (image C). During this treatment, the liquefied precursor
first starts to boil (at around 150 mbar), and then solidifies. Image D shows the
final catalyst after calcination for 5 h at 700 ºC. Note that after vacuum drying, the
catalyst precursor is still blue, as is the copper nitrate starting material (image C).
After calcination the catalyst is black (image D), indicating the presence of CuO
instead of Cu2O, which is yellow/red.[23]
Chapter 3.2 Characterisation
149
Figure 2. The copper-ceria sample 3 at various stages. A: the mixed nitrate precursors;
B: after heating to 120 ºC; C: after treatment in the vacuum oven (120 ºC, 10 mbar, 4 h); D:
after calcination in static air (700 ºC, 5h).
In the preparation of sample 2, the pressure was lowered quickly (~10 min),
giving no possibility for the formation of separate CuO and CeO2 phases.
Conversely, in sample 3 the pressure was lowered gradually over 1 h (from 1 bar to
<10 mbar, at a constant rate), allowing the formation of separate phases. The BET
surface area of each sample was determined by nitrogen physisorption. The
crystallite size, lattice constants and phase uniformity were determined by XRD
(see Table 1).
Table 1. Catalyst composition, BET surface area, crystallite size and lattice constant.
Sample/
composition Phase composition
Surface
area
(m2/g)
Ceria
crystallite
size (nm)[a]
Lattice constant
(Å)
1/ CeO2 monophasic
(fluorite) 38 26.4
5.409
(lit. 5.411[24])
2/ Ce0.90Cu0.10O2 monophasic 47 14.9 5.411
3/ Ce0.90Cu0.10O2 Biphasic
(CuO and fluorite) 9 33.3 5.411
[a] derived from the peak broadening of the Ce(111) XRD peak using the Scherrer equation.
Assessing successful doping by XRD analysis. X-ray diffraction is a
popular method for examining crystalline phases of doped ceria.[9, 25, 26] However,
the absence of a separate crystalline dopant phase does not necessarily mean that
the dopant is actually incorporated in the lattice. Indeed, several groups observed
that when impregnating copper on ceria supports, no copper oxide phases were
Chapter 3.2 Characterisation
150
detected, provided that copper loading and calcination temperatures were kept
low.[27] The copper can be present as an amorphous copper oxide, as CuxO
crystallites smaller than the XRD detection limit (~3 nm), or as another CexCuyOz
phase.[28-30]
Besides the identification of different crystal phases, XRD patterns give
information on the lattice parameter and the crystallite size. These may be affected
by the dopant. Our results show that the crystallite size of the copper-doped sample
2 is smaller than that of the undoped ceria sample 1 (see also Chapter 3.4).[13, 14, 31]
Still, this does not prove that the dopant is in the fluorite lattice.[14] Since it is not
clear how the dopant atom influences the crystallite size, one cannot discriminate
between bulk and surface effects.
Conversely, the lattice parameter does reflect properties of the bulk fluorite
lattice, but the information is of limited value. Doping creates a very complex
system where several factors affect the lattice parameter: Cu2+ is smaller than Ce4+,
and will reduce the lattice parameter, but Cu+ and Ce4+ are similar in size.[26, 31] The
lower valence of the copper ions induces oxygen vacancies, which are smaller than
the oxygen anions.[24, 32] That being said, the lattice may expand with the creation
of oxygen vacancies, because the cations are less shielded.[33] Finally, several
groups observed surface enrichment of the dopant atoms.[26, 34, 35] Even for catalysts
doped with the same amount of copper, this enrichment may vary with the
preparation method, and lead to different lattice parameters. Indeed, for copper-
doped ceria both a decrease, and no change in the lattice parameter were
observed.[25, 26, 31, 36, 37] Comparing the lattice parameters of samples 1–3, we see
little or no variation. To avoid these issues, we use XRD only for determining
whether crystalline CuO phases form or not. The nature of the catalysts is then
assessed using temperature programmed reduction (TPR) analysis.
Temperature programmed reduction (TPR). TPR is used extensively
for characterizing both ceria and copper-ceria catalysts.[21, 30, 38-40] Here, we used it
for determining the various Ce and Cu surface species, and for studying the active
sites in selective hydrogen combustion. First, we used TPR for studying freshly-
prepared catalyst samples, with the aim of understanding the effects that copper
doping has on the surface interactions. Figure 3 shows the TPR profiles of CeO2 1,
monophasic Ce0.90Cu0.10O2 2, and plain CuO. We observe the typical peaks of
Chapter 3.2 Characterisation
151
undoped ceria (1), at about 470 °C (peak A) and 700 °C (peak B). These peaks are
ascribed to the reduction of surface oxygen and bulk oxygen, respectively.[41]
Indeed, the catalyst’s BET surface area is correlated with the area of peak A, but
not sufficiently for using TPR for surface area determination.[39, 42] Recently,
Trovarelli’s group explained the two peaks in terms of reduction of smaller and
larger crystallites.[39] A higher reduction temperature for larger particles was also
shown for CeO2/Al2O3, while lower temperatures were observed for smaller
particles.[43-45] Since small particles are ‘mostly surface’, and large particles are
‘mostly bulk’, the surface/bulk model and the particle size model are equivalent.
0 200 400 600 800
Temperature (°C)
TC
D s
igna
l (a
u)
CuO CeO2 (1)
Ce0.90Cu0.10O2 (2)
AB
C
0 200 400 600 800
Temperature (°C)
TC
D s
igna
l (a
u)
CuO CeO2 (1)
Ce0.90Cu0.10O2 (2)
AB
C
Figure 3. TPR of CuO (profile scaled down), CeO2 (1), and Ce0.90Cu0.10O2 (2).
Conditions: 5 °C/min to 800 °C in 67% v/v H2/Ar flowed at 20 mL/min. Samples were
calcined in situ to 300 °C (5% v/v O2/Ar, 30 min hold at 300 °C) prior to analysis (200 °C
in case of the CeO2 sample).
Note that the position of peak B is unaffected by the presence of copper
(catalyst 2, Figure 3). However, peak A disappears, and a new peak, C, appears at
around 110 °C. This has been explained by H2 dissociation on the copper surface,
which then spills over and reduces the ceria surface oxygen.[21, 29] The dissociative
adsorption of hydrogen on pure ceria was proposed as the rate-limiting step below
470 °C.[46, 47] The fact that peak A is shifted to lower temperatures (~110 °C),
Chapter 3.2 Characterisation
152
indicates that this step is rate-limiting over a broad temperature range. The upper
limit of this range is 470 ºC, where pure ceria is reduced. This is why peak B is
unaffected by the addition of copper. The disappearance of peak A proves that (part
of) peak C reflects the reduction of the ceria surface – the CuO reduction is
insufficient to account for the area of peak C.[26, 48] A similar effect was observed
for copper and other metals supported on ceria (not doped). Here too, the support
reduction by spillover hydrogen was preceded by reduction of the metal oxide. A
lower onset of copper oxide reduction is generally ascribed to highly dispersed
species (small copper particles, clusters or atoms).[48-50] Indeed, Figure 3 shows that
the copper-doped sample 2 reduces at lower temperatures than pure CuO.
Table 2 shows quantitative TPR data of catalysts 1–3. The amount of
oxygen released in peak C (peak B in case of 1, ceria) is given per 100 mg sample.
The data show that the addition of copper increases the amount of oxygen released
from the catalysts. Note that the lower oxygen yield of the biphasic catalyst 3 can
be related to its low surface area.
Table 2. Quantitative TPR data of catalysts 1–3.
Sample/
composition Phase composition
Surface area
(m2/g)
Size of TPR peak C
(mg O/100 mg sample)[a,b]
1/ CeO2 monophasic
(fluorite) 38 0.38
2/ Ce0.90Cu0.10O2 monophasic 47 1.22
3/ Ce0.90Cu0.10O2 Biphasic
(CuO and fluorite) 9 0.73
[a] Data obtained by calibrating the TCD detector using a CuO standard. The peak area of
this standard is integrated and the area is correlated to the amount of oxygen present in the
CuO. [b] Peak A in case of catalyst 1 (ceria).
Following the measurements of the freshly–prepared catalysts, we repeated
the TPR analysis using a series of spent catalyst samples. Such studies are
important, especially in cases of cyclic redox processes, where the initial catalyst
often differs from the catalyst after some cycles. We ran five repetitions, but no
change was observed after the second cycle. Figure 4 shows the low-temperature
TPR section (peak C), for the fresh sample 2, and the spent sample 2A. The latter
Chapter 3.2 Characterisation
153
is sample 2 reduced at 800 °C in 67 % hydrogen. Both profiles show multiple
features. Sample 2 has three features at ~90 ºC, 100 ºC and 110 °C.[29, 48, 49] Here,
we will focus only on the reduction temperature range, and compare this between
samples.[51] The reason is that the fine-structure of the peaks depends on the
detection method, and thus any conclusions drawn based on this fine-structure are
problematic (a detailed technical explanation is given in the appendix of this
chapter). The reduction maximum of sample 2A is shifted about 40 °C upwards, to
~150 °C, compared to sample 2 (Figure 4). This indicates an aggregation of copper
atoms into clusters, and/or a sintering of copper clusters into larger particles. We
also measured the TPR of a portion of catalyst 2 that was subjected to severe redox
cycling (sample 2B, 12 h reduction at 550 °C, followed by three redox cycles at
550 °C). This sample showed a similar shift in the reduction temperature (data
included in the appendix of this chapter), and XRD analysis showed that some CuO
has formed (see Figure 5). Therefore, the TPR feature at 150 °C may indicate a
copper-enriched phase, possibly with CuO crystals.
200
400
600
800
0 50 100 150 200 250
Temperature (°C)
TC
D s
igna
l (a
u) Fresh
Spent
200
400
600
800
0 50 100 150 200 250
Temperature (°C)
TC
D s
igna
l (a
u) Fresh
Spent
Figure 4. TPR results for the fresh catalyst 2 (Ce0.90Cu0.10O2) and the spent
catalyst 2A. Spent sample 2A is catalyst 2 after reduction at 800 °C in 67% H2/Ar. Both
samples were calcined at 300 °C in 5% v/v O2/Ar prior to analysis.
Chapter 3.2 Characterisation
154
The fresh catalyst 3 also exhibits a separate CuO phase (Figure 5). We
therefore expect this catalyst to have a higher reduction maximum than the doped
catalyst 2. Indeed, the reduction temperature of 3 is considerably higher,
approaching that of pure CuO (cf. the TPR profiles in figures 6 and 3).
CuO CuO
3A32B
250
500
750
1000
20 25 30 35 40
2θ (°)
Inte
nsity
(co
unts
) CuO CuO
3A32B
250
500
750
1000
20 25 30 35 40
2θ (°)
Inte
nsity
(co
unts
)
Figure 5. XRD data of the spent catalyst 2B, fresh catalyst 3, and spent catalyst
3A.
2 3A
3
500
1000
1500
0 50 100 150 200 250
Temperature (°C)
TC
D s
igna
l (a
u)
2 3A
3
500
1000
1500
0 50 100 150 200 250
Temperature (°C)
TC
D s
igna
l (a
u)
Figure 6. TPR profile of the fresh catalyst 3, and spent catalyst 3A. Spent catalyst
3A is catalyst 3 after reduction at 200 °C in 67% H2/Ar. The profile of sample 2 is included
for comparison.
Chapter 3.2 Characterisation
155
Figure 6 shows that after reduction at 200 °C, part of the 190 °C feature of
catalyst 3 is replaced by the typical feature at ~150 °C. The lower reduction
temperature indicates that the copper has been dispersed. Indeed, the XRD data of
3A (Figure 5) shows that the CuO crystallites have disappeared. The TPR and
XRD data show that there is a strong interaction between ceria and copper.
Reduction of the doped monophasic sample 2 leads to clustering of the copper
atoms, and peak C rises to ~150 °C. When the biphasic sample 3 is reduced, the
crystalline CuO disappears, and peak C lowers to ~150 °C. The overall result is a
similar TPR feature, as in both cases a thermodynamically stable phase is formed.
Interestingly, when copper is supported on alumina, no reduction is observed
below 150 °C, contrary to the copper–ceria case.[49, 52, 53] Strong metal-ceria
interactions were also observed for Pt, Ni, Rh and Pd. The ceria support can ‘crawl
over’ the Pt particles upon reduction (a phenomenon known as decoration), and
ceria supported Ni–particles can spread to a Ni–monolayer upon reduction.[54, 55]
The active catalyst phase in selective hydrogen combustion. Both
copper-doped ceria and copper supported on ceria are used as catalysts in a variety
of processes, including (preferential) CO oxidation,[8, 9, 13, 56] methane oxidation,[9,
56] water gas shift reaction,[37] SO2 reduction,[9, 57] NO reduction,[14, 26] and phenol
oxidation.[31] This may reflect similar active sites in the doped and supported
catalysts. TPR and XRD measurements can shed some light on the nature of these
active sites. We applied these techniques on catalyst 2, a promising candidate for
burning hydrogen selectively from a hydrogen/propane/propene mixture at
550 °C.[17] The selectivity towards hydrogen combustion was measured over 16
redox cycles. Each cycle consisted of an oxidation step (18 min, 1% v/v O2 in Ar),
a purge (4 min, pure Ar), a reduction step (10 min, 4:1:1% v/v C3H8:C3H6:H2 in
Ar), and a second purge (4 min, pure Ar). The selectivity was determined during
the reduction step, as the ratio H2 conversion:total conversion The activity is
expressed as the percentage of the hydrogen feed which is combusted (labelled
‘hydrogen activity’). The catalyst gave 92% selectivity, and 7% H2 combustion
(values are averaged over 16-fold experiments). Control experiments with 3
showed that the presence of the separate CuO phase results in a higher hydrogen
activity (25% H2 combustion), but lower selectivity (83%). Indeed, alumina
Chapter 3.2 Characterisation
156
supported CuO also has a lower selectivity (72%, 13 wt% Cu/Al2O3, which
corresponds to about 13 mol% Cu). The activity of the CuO/Al2O3 catalyst is not as
high as for the CuO/ceria 3, which possibly related to a larger CuO crystallite size
in case of the CuO/Al2O3.
To assess the stability of the copper doped ceria 2, three consecutive TPR
profiles were measured after the catalyst was subjected to 16 redox cycles in the
reaction mixture (4:1:1% v/v C3H8:C3H6:H2 in Ar, 550 °C). The sample was
reoxidised before each TPR measurement. The TPR profiles (Figure 7) exhibit the
same low-temperature reduction region at ~110 °C, which is also present in the
fresh catalyst (cf. Figure 4). Moreover, the XRD pattern of the spent catalyst
showed no CuO peaks (data not shown). This means that the highly dispersed
copper species related to the TPR feature at 110 °C are the active species in the
selective hydrogen combustion. The first TPR was stopped at 300 °C, to preserve
the integrity of the sample. Note that the second TPR still shows the 110 °C feature
(the profiles overlap). However, the third TPR, carried out after reduction to
800 °C, shows the typical reduction peak at ~150 °C, similar to the spent samples
2A and 2B.
200
400
600
800
0 50 100 150 200 250
Temperature (°C)
TC
D s
igna
l (a
u) First
Second
Third
200
400
600
800
0 50 100 150 200 250
Temperature (°C)
TC
D s
igna
l (a
u) First
Second
Third
Figure 7. Three consecutive TPR measurements (left to right) of catalyst 2 after
measuring the selectivity and activity in selective hydrogen combustion (16 redox cycles at
550 °C). The sample was reoxidised before each measurement.
Chapter 3.2 Characterisation
157
The enthalpy of reduction of copper doped ceria. Besides shifting the
equilibrium to the products side, combusting the formed hydrogen also generates
heat, aiding the endothermic dehydrogenation reaction. The combustion of
hydrogen is very exothermic (-242 kJ/mol, when forming gaseous H2O).
Subtracting the enthalpy needed for the dehydrogenation reaction (130 kJ/mol at
460 °C) leaves -112 kJ/mol.[58-60] The oxygen is delivered, however, by the ceria,
and its reduction is endothermic. Full (surface and bulk) reduction of CeO2 to
Ce2O3 takes 371 kJ/mol, resulting in a net endothermic process (see Scheme 1).[61]
At the dehydrogenation reaction conditions (550 – 600 °C), however, we expect
only surface reduction to occur. Secondly, we use doped ceria instead of plain
CeO2. No data on the enthalpy of (surface) reduction of doped cerias is available,
but the enthalpy of formation of the dopant oxides can be used as a guideline.
Table 3 shows the enthalpy of formation of the oxides of several metals we use as
dopant in our study. The oxides with the lowest enthalpy of formation will
consume the least energy when reduced. Since the combustion of hydrogen
generates 242 kJ/mol, a net exothermic reduction is expected for the oxides with a
heat of formation below this value. The data in Table 3 show that this is the case
for CuO, Cu2O and PbO. Still, the enthalpy of formation of the dopant oxides is not
the same as the reduction of a doped ceria surface.
Chapter 3.2 Characterisation
158
Table 3. Enthalpy of formation of several metal oxides.[58]
Oxide Enthalpy of formation
(kJ/mol)
CuO -157
Cu2O -171
PbO -219
Pb2O -274
CaO -635
FeO -272
Fe2O3 -825
Fe3O4 -1121
Bi2O3 -568
MnO -385
MnO2 -520
SnO2 -578
To assess the enthalpy of reduction of copper doped ceria, in situ TG-DSC
was performed on catalyst 4, doped with 7 mol% Cu (Ce0.93Cu0.07O2). The
combined TG-DSC apparatus allows for the determination of both the mass change
and the enthalpy changes upon heating a sample in hydrogen.[62] In a typical
experiment, about 50 mg of sample was placed in a quartz cup, which was heated
to 300 °C in air (30 min hold), to remove any adsorbed species. The sample was
allowed to cool to room temperature under helium. For the actual experiment, the
sample was heated to 600 °C at 10 °C/min in 25% H2 in helium, at a 40 mL/min
total flow. Figure 8 shows the heat flow and weight loss of the sample during the
experiment. The data show that indeed, the surface reduction of the copper doped
ceria is net exothermic (a positive peak is present around 150 °C, the broad
negative peak around 100 °C stems from the desorption of water). This peak at
150 °C coincides with weight loss of the sample, due to the removal of oxygen.
Figure 9 shows the DSC data combined with the TPR data catalyst 4. Note that
these data are from two batches of catalyst 4, and measured on two different
apparatus. The exothermic DSC peak overlaps nicely with the TPR reduction peak,
associated with the surface reduction of the Cu doped ceria.
Chapter 3.2 Characterisation
159
-8
-4
0
4
8
200 400 600
Temperature (°C)
Hea
t flo
w (
mW
)
-0.4
-0.3
-0.2
-0.1
0.0
ΔW
eigh
t lo
ss (
mg)
Weight loss
Heat flow
-8
-4
0
4
8
200 400 600
Temperature (°C)
Hea
t flo
w (
mW
)
-0.4
-0.3
-0.2
-0.1
0.0
ΔW
eigh
t lo
ss (
mg)
Weight loss
Heat flow
Figure 8. Heat flow and weight loss of Ce0.93Cu0.07O2 (4) during reduction in
hydrogen. Conditions: 50 mg sample, heated to 600 °C at 10 °C/min in 25% H2 in helium,
at a 40 mL/min total flow.
-8
-4
0
4
8
200 400 600
Temperature (°C)
Hea
t flo
w (
mW
)
TC
D s
igna
l (A
U)
TPR
Heat flow
-8
-4
0
4
8
200 400 600
Temperature (°C)
Hea
t flo
w (
mW
)
TC
D s
igna
l (A
U)
TPR
Heat flow
Figure 9. The heat flow of Ce0.93Cu0.07O2 (4) during reduction in hydrogen, as
determined on the TG-DSC set up, and the TPR profile of 4. Note that these data were
determined on two different experimental set ups using two different batches of catalyst 4.
Chapter 3.2 Characterisation
160
Quantitative analysis of the DSC data shows that the enthalpy of reduction
of the copper doped ceria 4 is -5 kJ/mol. This is small compared to the enthalpy of
the dehydrogenation of 130 kJ/mol. It follows that the beneficial effect of the
hydrogen combustion on the propene yield must originate from the shifting of the
dehydrogenation equilibrium, instead of from the generation of heat. Note that the
positive effect of shifting the reaction equilibrium on the propene yield is
substantial. Previously, we have modelled the combined dehydrogenation and
selective hydrogen combustion process, using plain ceria as SOR, and assuming
the reduction of the ceria is endothermic.[63] The model was similar to the industrial
Catofin propane dehydrogenation process.[64] In this industrially implemented, a
chromium/alumina dehydrogenation catalyst is used in a fixed bed reactor. In the
cyclic redox process, 12 min dehydrogenation steps are alternated with 12 min
oxidation steps, with purge steps in between. The oxidation step is needed combust
the coke, built up on the catalyst surface during the dehydrogenation. Because this
is also a redox process, we can simply replace part of the dehydrogenation catalyst
with the SOR, and run the process under standard conditions (the only difference
with our model study is that the Catofin process runs under slight vacuum). Our
data showed that, at 550 °C, the addition of the SOR can increase the conversion
from about 40% to about 60%. The best results are obtained when adding ~10%
v/v of the SOR. At higher SOR concentrations, the propene yield suffers from the
decrease in concentration of the dehydrogenation catalyst. Note again that in this
model, the reduction of ceria was endothermic, and a positive effect on the propene
yield was still observed. The reduction kinetics, however, were not taken into
account.
Importantly, the selective hydrogen combustion will always generate heat,
regardless whether the reduction of ceria is endo- or exothermic. Indeed, the ceria
is reduced in the dehydrogenation step, but reoxidised in the reoxidation step, and
the net heat effect of these processes is always zero. The beneficial effect of the
hydrogen combustion is therefore, regardless of the reduction enthalpy of the ceria:
-242 kJ/mol (hydrogen combustion) + 130 kJ/mol (propane dehydrogenation) =
-112 kJ/mol. This is shown in Scheme 1, where the heat effects of the individual
steps of the redox process are given. In this example, plain ceria is used as SOR,
which is fully reduced from CeO2 to Ce2O3. This full reduction of the ceria is very
Chapter 3.2 Characterisation
161
endothermic (+371 kJ/mol), resulting in a net endothermic reduction step (see
Scheme 1). During the reoxidation however, the reoxidation of the ceria generates
the same value of -371 kJ/mol. Therefore, the net effect of the overall redox
process is -112 kJ/mol - y kJ/mol, where y is the enthalpy generated by the
combustion of the coke from the dehydrogenation catalyst (see Scheme 1).
Dehydrogenation step: Enthalpy (kJ/mol)
H2 + ½ O2 ↔ H2O (g) -242
2 CeO2 ↔ Ce2O3 + ½ O2 +371
C3H8 ↔ C3H6 + H2 +130 +
net heat effects of the dehydrogenation step: +259
Oxidation step:
Ce2O3 + ½ O2 ↔ 2 CeO2 -371
C + O2 → CO2 -y +
net heat effects of the oxidation step: -371 - y
overall: -112 - y
Scheme 1. Heat effects of the individual steps in the combined dehydrogenation
and selective hydrogen combustion process, using plain ceria as SOR, and assuming full
surface and bulk reduction of the ceria occurs.
Note that in the example shown in Scheme 1, the heat is generated in the
reoxidation step, where it is most needed in the dehydrogenation step. The heat
generated in the oxidation step is stored in the catalyst bed, however, and so
transported to the dehydrogenation step. This effect is used in the Catofin
dehydrogenation process, where the combustion of the coke heats the catalyst bed.
Still, cooling the reactor bed during the dehydrogenation, as a result of the
reduction of the SOR, is not ideal. Our measurements on the copper doped ceria 4
show that this does not need to occur for the doped ceria catalysts, since the
reduction of 4 is net exothermic. The actual value of the ceria's reduction enthalpy
governs if the heat is generated during the dehydrogenation or oxidation step. This
is shown in Scheme 2. Here, the oxidised SOR is denoted ‘SOR-O’, and the
reduced SOR as ‘SOR-R’. The enthalpy of reduction is the SOR is = +x kJ/mol.
Chapter 3.2 Characterisation
162
Scheme 2 shows that when x < 112 kJ/mol, heat will be released during both the
dehydrogenation step and the oxidation step, and when x > 112 kJ/mol, heat will be
released during the oxidation step only. The DSC measurements on the copper
doped ceria catalyst 4 show a net value of -5 kJ/mol for its reduction. It follows
that the value of x for this catalyst is: -242 + x = -5, x = 237 kJ/mol. This results in
a net heat effect during the dehydrogenation of -112 + 237 = +125 kJ/mol, instead
of the 130 kJ/mol, when the SOR would not be added to the dehydrogenation
catalyst. Again, it follows that the benefit of adding the SOR, as observed in our
model study, is the result of the shift in equilibrium. Our data do show that that the
reduction of a doped ceria surface can be net exothermic, so that the addition of the
SOR does not have to result in extra cooling of the reactor bed during the
dehydrogenation.
Dehydrogenation step: Enthalpy (kJ/mol)
H2 + ½ O2 ↔ H2O (g) -242
SOR-O ↔ SOR-R + ½ O2 +x
C3H8 ↔ C3H6 + H2 +130 +
net heat effects of the dehydrogenation step: -112 + x
Oxidation step:
SOR-R + ½ O2 ↔ SOR-O -x
C + O2 → CO2 -y +
net heat effects of the oxidation step: -x - y
overall: -112 - y
Scheme 2. Heat effects of the individual steps in the combined dehydrogenation
and selective hydrogen combustion process, using a SOR with reduction enthalpy of
x kJ/mol.
Chapter 3.2 Characterisation
163
Conclusions The combination of TPR and XRD is a simple and powerful tool for distinguishing
between surface phases of copper-ceria catalysts. Depending on the type of
catalyst, reduction occurs at ~110 °C, ~150 °C or ~190 °C. The reduction at 110 °C
is ascribed to highly dispersed copper atoms (doped ceria), and that at 190 °C to
CuO crystallites supported on ceria. After sufficient reduction, both doped and
supported catalysts show the same TPR feature, at 150 °C. The copper-ceria system
is highly interactive: copper atoms cluster in case of the doped catalyst, raising the
reduction temperature to 150 °C. Conversely, CuO crystals in the (biphasic)
supported catalysts mix into the ceria surface, lowering the reduction temperature
to 150 °C. Using this knowledge, we analyzed a copper-doped ceria catalyst used
in selective hydrogen combustion. The absence of CuO crystallites as determined
by XRD, and the reduction temperature of ~110 °C as determined by TPR, indicate
that copper doped ceria is the active phase in this reaction. Calorimetric
measurements show that the hydrogen combustion by doped cerias can indeed be a
net exothermic process. The enthalpy of reduction of a 7 mol% Cu–doped ceria is
-5 kJ/mol.
Chapter 3.2 Characterisation
164
Experimental Section
Materials and instrumentation. All chemicals were purchased from
commercial sources and used as received. Gases were purchased from Praxair
(>99.5% purity) and were further purified over BTS columns and/or molsieves.
Stable gas flows were obtained using Bronkhorst Mass Flow Controllers.
Temperature Programmed Reduction measurements were performed on a
conventional system built in house. The TCD detector was equipped with Rhenium
Tungsten filaments and powered by a model 40-202 GOW MAC Instrument Co.
power supply. Powder X-ray diffraction was performed using a Philips PW-series
X-ray diffractometer with a Cu tube radiation source (λ = 1.54 Å), a vertical axis
goniometer and a proportional detector. The measurement range was 10 ° – 93 °,
with a 0.02 ° step size and 5 s dwell time. Lattice constants and crystallite sizes
were obtained after Rietveld refinement (structure fit) using PANalytical's X'pert
software package. The surface areas of the catalysts were measured by N2
adsorption at 77 K on a Sorptomatic 99 (CE Instruments), and calculated using the
BET equation. GC analysis was performed on an Interscience CompactGC,
separating water, CO2 and C2 and C3 hydrocarbons on a Porabond Q column (He
carrier gas) and H2, CO, CH4, O2 and N2 on a 5 Å molsieve column (Ar carrier
gas). MS analysis was performed using a Pfeiffer QMS 200 mass spectrometer
(m/z range 0–200). Thermogravimetric analysis was performed on a Setaram
TG-85 thermobalance, and the TG-DSC analysis on a Setaram TG-DSC 111.
Procedure for catalyst synthesis. The catalysts were prepared by
sequential co–melting, drying, and calcining of the mixed metal nitrates, using
Ce(NO3)3.6H2O and Cu(NO3)2.3H2O.[17] The metal nitrates are weighed in an open
porcelain crucible and mixed with a spatula. The crucible is placed in a vacuum
oven pre–heated to 120 °C. After the nitrates have melted (5–10 min), the pressure
is carefully lowered from 1 bar to < 10 mbar in about 10 min, making sure no
vigorous boiling occurs. A brightly coloured solid is formed. After 4 h, the samples
are placed in a furnace and calcined under static air at 700 °C (ramp rate 300 °C/h,
5 h hold). The resulting solid is pulverized, ground and sieved in fractions of 125–
212 µm (selectivity assessment) and < 125 µm (TPR, XRD, TG-DSC and BET
measurements).
Chapter 3.2 Characterisation
165
Procedure for TPR measurements. In a typical measurement, 100 mg of
sample is placed on a quartz wool plug in a 4 mm i.d. quartz reactor. The sample is
calcined in situ to 300 °C (ramp rate 10 °C/min, 30 min hold time) in 5% v/v
oxygen in argon (50 mL/min total flow). After cooling to room temperature and
purging with pure argon, the system is allowed to equilibrate in 67 % hydrogen in
argon (20 mL/min total flow) for about 1 h. For the actual TPR measurement, the
sample is heated with a 5 °C/min heating rate to 800 °C (no hold). When the final
temperature is reached, the sample is allowed to cool to room temperature. When
subsequent measurements are performed, the sample is reoxidised in 5 %v/v
oxygen in argon (300 °C, 30 min hold time). The same procedure was followed for
the TPR measurements performed in the thermobalance, with the difference that
about 250 mg sample is placed in a quartz cup, and mass loss is determined instead
of hydrogen uptake. The thermobalance has an accuracy of 10 microgram.
Procedure for selective hydrogen combustion experiments. The activity
and selectivity were determined using an automated cyclic redox reactor system
built in-house, described in detail elsewhere.[17] About 250 mg of sample was
placed on a quartz wool plug in a quartz reactor and heated in 1% v/v oxygen in
argon to 550 °C at 1200 °C/h. At this temperature, the selectivity was determined
by GC during the 10 minute reduction in 4:1:1% v/v C3H8:C3H6:H2 in Ar (total
flow 50 mL/min), with 5 mL/min of N2 added as internal standard. The gas hourly
space velocity (GHSV) is 13200 / h (at the typical bed volume of 0.25 cm3 and the
reduction cycle's total flow of 55 mL/min). The weight hourly space velocity
(WHSV) is 1.2 / h, and is calculated from the weight of C3H8 + C3H6 + H2 per h per
the weight of the catalyst. After a 4 minute purge step (pure Ar), the sample was
reoxidised for 18 minutes in 1% v/v O2 in Ar (50 mL/min total flow). The redox
cycle is completed by another purge step in pure Ar. The selectivity is determined
as the ratio H2 conversion:total conversion, and is averaged over sixteen redox
cycles. The activity is determined as the percentage of the hydrogen feed which is
combusted (labelled ‘hydrogen activity’).
Procedure for TG-DSC experiments. In a typical experiment, about 50
mg of sample was placed in a quartz cup, which was heated to 300 °C in air (30
min hold), to remove any adsorbed species. The sample was allowed to cool to
Chapter 3.2 Characterisation
166
room temperature under helium. For the actual experiment, the sample was heated
to 600 °C at 10 °C/min in 25% H2 in helium, at a 40 mL/min total flow.
Acknowledgements We thank Dr. M. C. Mittelmeijer-Hazeleger for the BET surface area
measurements, A. C. Moleman and W. Moolhuijzen for allowing use of and giving
instructions on the XRD instrument, Dr. Vesna Rakić of the University of Belgrade
(Serbia) for performing the TG-DSC measurements, Dr. Aline Auroux of
IRCELYON-CNRS (Lyon, France) for allowing the use of the TG-DSC, and
NWO-ASPECT for financial support and feedback.
Chapter 3.2 Characterisation
167
Appendix
Interpreting fine-structure in TPR: Pros and cons. In this study, the
temperature range where reduction occurs is compared between samples. The fine
structure of each TPR trace is not analysed. The reason for this is that when using
TCD as a detection method, one measures the hydrogen consumption of the
sample. Not all of this consumption may be related to reduction of the phase of
interest, that is the copper- or cerium oxide. Interactions of hydrogen with
carbonaceous or nitrate like species, and hydrogen uptake by ceria, facilitated by
the presence of copper, will result in hydrogen uptake as well.[21, 51, 65-67]
Furthermore, any species desorbing during the TPR measurement will be detected
by the universal TCD. Most of these defects can be avoided by in situ calcination
prior to analysis, to remove adsorbed species. To test our experimental we
performed TPR of our undoped ceria (1) without in the situ calcination. No ghost
peaks or hydrogen uptake were observed below 300 °C (the onset of the first
reduction peak). To ensure we do not discard information by only looking at
reduction regions, we analyzed three portions of sample 2B. Two portions of the
sample were analyzed on the standard TPR–TCD set up, and one using thermo–
gravimetric analysis (TGA). With the latter, the mass loss of the sample is
measured during reduction, so the technique is less sensitive to hydrogen
absorption effects. Figure 10 shows the reduction profiles of the three
measurements of sample 2B. The feature at 120 °C is far less prominent when
using thermo–gravimetric analysis, and differs in size in the two TCD
measurements. The other two features at 100 °C and 150 °C are comparable
between the two techniques.[68] The size of the 120 °C feature seems to be related
to the detection method and confirms that care should be taken when interpreting
small features or shoulders when using TPR.
Chapter 3.2 Characterisation
168
200
400
600
800
50 100 150 200
Temperature (°C)
TC
D s
igna
l (a
u)
TCD 1TCD 2TGA
200
400
600
800
200
400
600
800
50 100 150 20050 100 150 200
Temperature (°C)
TC
D s
igna
l (a
u)
TCD 1TCD 2TGA
Figure 10. TPR of three portions of catalyst 2B, measured using TCD (TCD 1,
TCD 2) and gravimetric detection (TGA).
Chapter 3.2 Characterisation
169
References
[1] E. D. Harris, Crit. Rev. Cl. Lab. Sci., 2003, 40, 547. [2] R. Uauy, M. Olivares and M. Gonzalez, Am. J. Clin. Nutr., 1998, 67, 952s. [3] K. Klier, Adv. Catal., 1982, 30, 243. [4] M. J. L. Gines, N. Amadeo, M. Laborde and C. R. Apesteguia, Appl. Catal. A:
Gen., 1995, 131, 283. [5] B. Cornils, W. A. Herrmann, R. Schlögl and C. Wong, Catalysis from A to Z - A
Concise Encyclopedia, Wiley-VCH, Weinheim, 2003. [6] M. B. Thathagar, J. Beckers and G. Rothenberg, J. Am. Chem. Soc., 2002, 124,
11858. [7] M. B. Thathagar, J. Beckers and G. Rothenberg, Green Chem., 2004, 6, 215. [8] D. H. Kim and J. E. Cha, Catal. Lett., 2003, 86, 107. [9] A. Tschöpe, W. Liu, M. Flytzani-Stephanopoulos and J. Y. Ying, J. Catal., 1995,
157, 42. [10] G. Jacobs, E. Chenu, P. M. Patterson, L. Williams, D. Sparks, G. Thomas and B.
H. Davis, Appl. Catal. A: Gen., 2004, 258, 203. [11] J. Y. Chane-Ching, F. Cobo, D. Aubert, H. G. Harvey, M. Airiau and A. Corma,
Chem. Eur. J., 2005, 11, 979. [12] A. Trovarelli, C. de Leitenburg, M. Boaro and G. Dolcetti, Catal. Today, 1999, 50,
353. [13] M. Jobbágy, F. Mariño, B. Schöbrod, G. Baronetti and M. Laborde, Chem. Mater.,
2006, 18, 1945. [14] P. Bera, S. T. Aruna, K. C. Patil and M. S. Hegde, J. Catal., 1999, 186, 36. [15] G. Rothenberg, E. A. de Graaf and A. Bliek, Angew. Chem., Int. Ed., 2003, 42,
3366. [16] F. Cavani, N. Ballarini and A. Cericola, Catal. Today, 2007, 127, 113. [17] J. H. Blank, J. Beckers, P. F. Collignon, F. Clerc and G. Rothenberg, Chem. Eur.
J., 2007, 13, 5121. [18] J. H. Blank, J. Beckers, P. F. Collignon and G. Rothenberg, ChemPhysChem,
2007, 8, 2490. [19] L. M. van der Zande, E. A. de Graaf and G. Rothenberg, Adv. Synth. Catal., 2002,
344, 884. [20] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 9. [21] M. Boaro, M. Vicario, C. de Leitenburg, G. Dolcetti and A. Trovarelli, Catal.
Today, 2003, 77, 407. [22] I. E. Wachs, Catal. Today, 2005, 100, 79. [23] N. N. Greenwood and A. Earnshaw, Chemistry of the Elements, Pergamon Press,
Oxford, 1989.
Chapter 3.2 Characterisation
170
[24] M. Mogensen, N. M. Sammes and G. A. Tompsett, Solid State Ionics, 2000, 129, 63.
[25] Y. Zhang, S. Andersson and M. Muhammed, Appl. Catal. B: Environ., 1995, 6, 325.
[26] P. Bera, K. R. Priolkar, P. R. Sarode, M. S. Hegde, S. Emura, R. Kumashiro and N. P. Lalla, Chem. Mater., 2002, 14, 3591.
[27] Tang et al. observed no CuO phase for a sample of 5 wt% Cu on ceria, calcined at 500 °C, with a final surface area of 80 m2/g (X. L. Tang, B. C. Zhang, Y. Li, Y. D. Xu, Q. Xin and W.J. Shen, Appl. Catal. A: Gen., 2005, 288, 116). The loading of 5 wt% Cu corresponds to 12.5 mol% Cu, if a doped ceria would have been prepared. This gives an indication of the capacity of ceria to support copper species which cannot be detected by XRD. Other groups reported similar results, see: X.C. Zheng, X. L. Zhang, X. Y. Wang, S. R. Wang and S. H. Wu, Appl. Catal. A: Gen., 2005, 295, 142; G. Jacobs, E. Chenu, P.M. Patterson, L. Williams, D. Sparks, G. Thomas and B.H. Davis, Appl. Catal. A: Gen., 2004, 258, 203; X.Y. Jiang, L.P. Lou, Y.X. Chen and X.M. Zheng, J. Mol. Catal. A: Chem., 2003, 197, 193).
[28] C. Lamonier, A. Bennani, A. D'Huysser, A. Aboukais and G. Wrobel, J. Chem. Soc., Faraday Trans., 1996, 92, 131.
[29] X. C. Zheng, X. L. Zhang, X. Y. Wang, S. R. Wang and S. H. Wu, Appl. Catal. A: Gen., 2005, 295, 142.
[30] A. Martínez-Arias, R. Cataluna, J. C. Conesa and J. Soria, J. Phys. Chem. B, 1998, 102, 809.
[31] S. Hočevar, U. O. Krašovec, B. Orel, A. S. Aricó and H. Kim, Appl.Catal. B: Environ., 2000, 28, 113.
[32] D. P. Woodruff, Oxide Surfaces, Elsevier, Amsterdam, 2001. [33] X. Q. Wang, J. A. Rodriguez, J. C. Hanson, D. Gamarra, A. Martínez-Arias and
M. Fernández-García, J. Phys. Chem. B, 2005, 109, 19595. [34] G. Rothenberg, E. A. de Graaf, J. Beckers and A. Bliek, Catal. Org. React., 2005,
104, 201. [35] P. J. Scanlon, R. A. M. Bink, F. P. F. van Berkel, G. M. Christie, L. J. van
IJzendoorn, H. H. Brongersma and R. G. van Welzenis, Solid State Ionics, 1998, 112, 123.
[36] W. Shan, W. Shen and C. Li, Chem. Mater., 2003, 15, 4761. [37] X. Q. Wang, J. A. Rodriguez, J. C. Hanson, D. Gamarra, A. Martínez-Arias and
M. Fernández-García, J. Phys. Chem. B, 2006, 110, 428. [38] R. T. Baker, S. Bernal, G. Blanco, A. M. Cordón, J. M. Pintado, J. M. Rodríguez-
Izquierdo, F. Fally and V. Perrichon, Chem. Commun., 1999, 149. [39] F. Giordano, A. Trovarelli, C. de Leitenburg and M. Giona, J. Catal., 2000, 193,
273. [40] A. Pintar, J. Batista and S. Hočevar, J. Colloid Interf. Sci., 2005, 285, 218.
Chapter 3.2 Characterisation
171
[41] H. C. Yao and Y. F. Yao, J. Catal., 1984, 86, 254. [42] M. F. L. Johnson and J. Mooi, J. Catal., 1987, 103, 502. [43] A. Piras, S. Colussi, A. Trovarelli, V. Sergo, J. Llorca, R. Psaro and L. Sordelli, J.
Phys. Chem. B, 2005, 109, 11110. [44] H. Cordatos, D. Ford and R. J. Gorte, J. Phys. Chem., 1996, 100, 18128. [45] E. Aneggi, J. Llorca, M. Boaro and A. Trovarelli, J. Catal., 2005, 234, 88. [46] P. Fornasiero, J. Kašpar and M. Graziani, Appl. Catal. B: Environ., 1999, 22, L11. [47] J. Elfallah, S. Boujana, H. Dexpert, A. Kiennemann, J. Majerus, O. Touret, F.
Villain and F. Lenormand, J. Phys. Chem., 1994, 98, 5522. [48] X. L. Tang, B. C. Zhang, Y. Li, Y. D. Xu, Q. Xin and W. J. Shen, Appl. Catal. A:
Gen., 2005, 288, 116. [49] W. P. Dow, Y. P. Wang and T. J. Huang, Appl. Catal. A: Gen., 2000, 190, 25. [50] A. Pintar, J. Batista and S. Hočevar, J. Colloid Interf. Sci., 2007, 307, 145. [51] F. M. Z. Zotin, L. Tournayan, J. Varloud, V. Perrichon and R. Frety, Appl. Catal.
A: Gen., 1993, 98, 99. [52] J. M. Dumas, C. Geron, A. Kribii and J. Barbier, Appl. Catal., 1989, 47, L9. [53] M. Ferrandon and E. Bjornbom, J. Catal., 2001, 200, 148. [54] S. Bernal, J. J. Calvino, M. A. Cauqui, J. M. Gatica, C. Larese, J. A. P. Omil and J.
M. Pintado, Catal. Today, 1999, 50, 175. [55] V. M. Gonzalez-DelaCruz, J. P. Holgado, R. Pereñíguez and A. Caballero, J.
Catal., 2008, 257, 307. [56] W. Liu and M. Flytzani-Stephanopoulos, J. Catal., 1995, 153, 304. [57] A. Rodas-Grapaín, J. Arenas-Alatorre, A. Gómez-Cortés and G. Díaz, Catal.
Today, 2005, 107-08, 168. [58] http://webbook.nist.gov/chemistry [59] H. Weyten, K. Keizer, A. Kinoo, J. Luyten and R. Leysen, Aiche J., 1997, 43,
1819. [60] M. M. Bhasin, J. H. McCain, B. V. Vora, T. Imai and P. R. Pujado, Appl. Catal. A:
Gen., 2001, 221, 397. [61] M. E. Huntelaar, A. S. Booij, E. H. P. Cordfunke, R. R. van der Laan, A. C. G. van
Genderen and J. C. van Miltenburg, J. Chem. Thermodyn., 2000, 32, 465. [62] J. Lebars, J. C. Vedrine, A. Auroux, B. Pommier and G. M. Pajonk, J. Phys.
Chem.-US, 1992, 96, 2217. [63] E. A. de Graaf, G. Zwanenburg, G. Rothenberg and A. Bliek, Org. Process. Res.
Dev., 2005, 9, 397. [64] http://www.cbi.com/lummus/process-technology/pdfs/Dehyrogenation.pdf [65] R. A. van Santen, Catalysis: An Integrated Approach, Elsevier Science B.V.,
Amsterdam, 1999. [66] K. Sohlberg, S. T. Pantelides and S. J. Pennycook, J. Am. Chem. Soc., 2001, 123,
6609.
Chapter 3.2 Characterisation
172
[67] L. A. Bruce, M. Hoang, A. E. Hughes and T. W. Turner, Appl. Catal. A: Gen., 1996, 134, 351.
[68] In consecutive redox cycles, the feature at 150 °C was seen to grow on account of the feature at 100 °C, indicating some sintering of the copper occurs.
173
3.3
The optimisation and characterisation of bismuth
doped ceria catalysts
Ce1-xBixO2
+Pt +Sn
+H2ΔT
PerformancePerformance
Pe
rform
ance
Per
form
ance
Ce1-xBixO2
+Pt +Sn
+H2ΔT
PerformancePerformancePerformancePerformance
Pe
rform
anceP
erfo
rmance
Per
form
ance
Per
form
ance
Doped ceria catalysts can be applied in a new process for oxidative propane
dehydrogenation, selectively burning hydrogen from the dehydrogenation mixture.
Bismuth is a promising dopant, but its selectivity can be improved. We found four
ways to achieve this: by increasing the hydrogen content of the feed, by co-doping
with Pt or Sn (the latter prevents coking), or by adjusting the reaction temperature
(optimal performance at 400 °C).
This work has been published as:
'Bismuth-doped ceria, Ce0.90Bi0.10O2: A selective and stable catalyst for clean
hydrogen combustion', Jurriaan Beckers, Adam F. Lee and Gadi Rothenberg, Adv.
Synth. Catal. 2009, 351, 1557.
Chapter 3.3 Characterisation
174
Abstract
Bismuth doped cerias are successfully applied as solid ‘oxygen reservoirs’ in
propane oxidative dehydrogenation. The lattice oxygen of the ceria is used to
selectively combust hydrogen from the dehydrogenation mixture at 550 °C. This
process has three key advantages: it shifts the dehydrogenation equilibrium to the
desired products side, generates heat aiding the endothermic dehydrogenation, and
simplifies product separation (H2O vs. H2). Furthermore, the process is safer since
it uses the catalysts' lattice oxygen instead of gaseous O2. We show here that
bismuth doped cerias are highly active and stable towards hydrogen combustion,
and explore four different approaches for optimising their application in the
oxidative dehydrogenation of propane. First, the addition of extra hydrogen which
lowers hydrocarbon conversion by suppressing both combustion and coking.
Second, the addition of tin which completely inhibits coking. Third, the addition of
platinum which increases selectivity, but at the expense of lower activity. The best
results are obtained through tuning the reaction temperature. At 400 °C, high
activity and selectivity was obtained for Ce0.90Bi0.10O2. Here, 90% of the hydrogen
feed is converted at 98% selectivity. This optimal reaction temperature can be
rationalised from the hydrogen and propene TPR: 400 °C lies above the reduction
maximum of hydrogen, yet below that of propene. That is, this temperature is
sufficiently high to facilitate rapid hydrogen combustion, but low enough to
prevent hydrocarbon conversion.
Chapter 3.3 Characterisation
175
Introduction Bismuth, although less common than metals such as lead, copper and iron,
is widely used in everyday life. Bismuth subsalicylate, for example, is the active
ingredient of over the counter drugs such as ‘Bismatrol’ and ‘Pepto-Bismol’, used
for treating several gastrointestinal ailments.[1] The low melting point of many
bismuth alloys enables their use in sprinkler system heads, where the melting of the
alloy unplugs the sprinkler,[2] and as a replacement for lead in low temperature
solders.[3] In catalysis, bismuth salts can be applied as 'green catalysts' in various
organic syntheses.[4-6] Bismuth is also part of a mixed-oxide catalyst used for
industrial production of acrylonitrile (the so–called ‘SOHIO process’).[7] The high
oxygen conductivity of bismuth oxides enables their use as electrodes for solid
oxide fuel cells, or as low temperature oxygen permeable membranes.[8-12] These
membranes can also be applied as catalysts in selective oxidations, such as the
oxidation of propene.[13, 14] The BiMeVOX mixed oxides (Bi4V2O11, where part of
the vanadium atoms can be replaced with a variety of metals), are a well known
example.[12]
In the 1990's, supported bismuth-oxides have been applied as catalysts in
oxidative dehydrogenation (ODH).[15, 16] Propane ODH yields valuable propene, the
building block of polypropene. The demand for propene is huge, and is expected to
rise to 80 million tonnes in 2010 worldwide.[17-19] Currently, crackers and oil
refineries account for about 95% of the propene supply, but more advantageous
methods such as metathesis an catalytic dehydrogenation of propane are gaining
ground. These allow for on-demand production of high purity monomer.[20]
Propane ODH is typically performed over supported vanadium and molybdenum
oxide catalysts, with a small amount of oxygen added to the gas feed.[21, 22]
However, the mixing of hydrocarbons and gaseous oxygen at elevated
temperatures is potentially hazardous and limits selectivity.[21, 23] These drawbacks
can be overcome by using a redox process, where the dehydrogenation is combined
with selective hydrogen combustion.[16, 24-29] The dehydrogenation is performed
over conventional Pt-Sn or Cr catalysts (typically at 550–600 °C), with a solid
‘oxygen reservoir‘ (SOR) added. The latter metal-oxide catalyst selectively
combusts the hydrogen in-situ from the dehydrogenation mixture, using its lattice
oxygen (Scheme 1, left).[24, 25] This process generates heat, aiding the endothermic
Chapter 3.3 Characterisation
176
dehydrogenation. It also increases the yield, since it shifts the equilibrium to the
products side, and is safer, since no gaseous oxygen is used. Following the
reduction of the SOR lattice, the oxygen vacancies are replaced using air, creating
a cyclic redox process. The use of two catalysts allows for separate tuning of the
dehydrogenation and selective hydrogen combustion. Scheme 1, right, illustrates a
possible industrial redox dehydrogenation process.
Fresh SOR
Spent SOR
CnH(2n+2)
A
N2O2N2
B C D
N2O2COx
N2CnH2nH2O
Fresh DH
Spent DH
ReoxidationReduction Purge Purge
H2
2 CeO2
Dehydrogenationcatalyst
Energy
H2O + Ce2O3
Propane Propene
Fresh SOR
Spent SOR
CnH(2n+2)
A
N2O2N2
B C D
N2O2COx
N2CnH2nH2O
Fresh DH
Spent DH
ReoxidationReduction Purge Purge
H2
2 CeO2
Dehydrogenationcatalyst
Energy
H2O + Ce2O3
Propane Propene
Scheme 1. Left: combined propane dehydrogenation and selective hydrogen
combustion: the selective hydrogen combustion consumes part of the hydrogen formed
during the dehydrogenation step, shifting the equilibrium to the products side and
generating heat. Right: cartoon of the complete redox cycle. After the dehydrogenation step
A, the bed is flushed with nitrogen (B), and the catalysts are regenerated through
reoxidation (C). This burns coke from the dehydrogenation catalyst and restores the lattice
oxygen of the SOR catalyst. After another nitrogen flush (D) the reactor is ready for the
next redox cycle.
Supported metal oxides, such as bismuth oxide, have been applied as SOR,
exhibiting high selectivity towards hydrogen combustion. However, due to their
low melting points they are unstable under redox cycling.[25, 30, 31] We recently
discovered a new type of SOR, based on ceria, that overcomes this limitation.[32, 33]
Ceria is often used in redox reactions because of the facile Ce3+ Ce4+ + e–
redox cycle.[34] Although it possess greater stability, the activity and selectivity of
pure ceria are low.[33] Nevertheless, active, selective and stable SOR catalysts can
Chapter 3.3 Characterisation
177
be formed by replacing about 10 mol% of the cerium atoms.[35, 36] The most
promising dopants are Cu, Mn, K, Cr, Pb, Sn and Bi. Out of these, the highest
activities are achieved using Pb or Bi as dopant. Bismuth has a low toxicity, in
contrast to lead, but bismuth doped ceria catalysts are also less selective. Here, we
describe several methods for increasing the selectivity of bismuth doped ceria in
selective hydrogen combustion.
Results and Discussion Catalyst preparation and characterisation. All catalysts were prepared
by co-melting mixtures of the metal nitrate hydrate precursors (chlorides in case of
Pt and Sn), at 140 °C in a vacuum oven. After the precursor has liquefied, the
pressure was lowered and a solid mixed metal nitrate formed. This was converted
into the mixed oxide by calcining in static air at 700 °C for 5 h.[37] Importantly,
these catalyst were not prepared by impregnating a cerium oxide support. Rather,
the co–melting of the cerium nitrate with the nitrate or chloride of the appropriate
metal yields a liquid precursor. This ensures ideal mixing prior to calcination,
incorporating the dopant into the fluorite lattice.
The catalyst composition, characterisation data and catalytic performance
of all catalysts are given in Table 1. The selectivity of the catalysts is determined
as: 10083632
2 HCHCH
H
conversion
conversion, and the activity as the percentage of the hydrogen
feed combusted by each catalyst (labelled ‘hydrogen activity’). Except for 8, the X-
ray diffraction patterns of the catalysts exclusively exhibit ceria's fluorite structure
(not shown), and no oxides of the added metal are observed. Catalyst 8 contains
some metallic platinum.
Cha
pter
3.3
Cha
ract
eris
atio
n
178
T
able
1. C
atal
yst c
hara
cter
isat
ion
and
cata
lytic
per
form
ance
dat
a at
550
°C
.
Cat
alys
t / C
ompo
sitio
n
crys
talli
te
size
(nm
)[a]
Lat
tice
spac
ing
(Å)
Sel
ectiv
ity
(%)[b
]
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
Cok
ing
(mg
C /
10 m
in)[c
]
Com
bust
ion
(vol
% C
O2
/ 10
min
)[d]
1 / C
eO2
22
5.41
1 0
0 0.
09
1.3
2 / C
e 0.9
8Bi 0
.02O
2 18
5.
411
94
10
0.19
2.
3
3 / C
e 0.9
3Bi 0
.07O
2 17
5.
416
82
26
0.22
3.
5
4 / C
e 0.9
0Bi 0
.10O
2 18
5.
416
77
33
0.16
4.
0
5 / C
e 0.8
8Bi 0
.08S
n 0.0
4O2
14
5.41
2 82
45
0.
00
4.7
6 / C
e 0.9
8Sn 0
.02O
2 12
5.
408
77
8 0.
10
2.8
7 / C
e 0.9
3Sn 0
.07O
2 10
5.
407
89
20
0.00
4.
2
8 / C
e 0.9
1Bi 0
.05P
t 0.0
4O2/
Pt0
[e]
12
5.41
7 95
14
0.
10
5.9
9 / C
e 0.9
8Pt 0
.02O
2 16
5.
411
0 0
0.77
2.
3 [a
] Der
ived
fro
m t
he p
eak
broa
deni
ng o
f th
e C
e(11
1) X
RD
pea
k us
ing
the
Sch
erre
r eq
uatio
n. [b
] The
fir
st d
ata
poin
t (a
fter
25
s), i
s no
t ta
ken
into
acc
ount
whe
n ca
lcul
atin
g se
lect
ivity
, si
nce
all
cata
lyst
s sh
ow u
nsel
ecti
ve c
ombu
stio
n he
re,
prob
ably
due
to
unse
lect
ive
reac
tion
with
adso
rbed
oxy
gen.
[c] T
he l
evel
of
coki
ng i
s de
term
ined
fro
m t
he a
mou
nt o
f C
O a
nd C
O2
dete
cted
by
MS
dur
ing
the
reox
idat
ion
cycl
e. I
t
ther
efor
e re
pres
ents
the
tot
al a
mou
nt o
f ca
rbon
(m
g) d
epos
ited
on
the
cata
lyst
sur
face
dur
ing
one
10 m
in r
educ
tion
cyc
le. [d
] Sum
mat
ion
of
the
amou
nt o
f C
O2
dete
cted
in
15 G
C a
naly
sis
perf
orm
ed d
urin
g th
e 10
min
red
uctio
n cy
cle.
Not
e th
at m
ost
cata
lyst
s on
ly p
rodu
ce C
O2
duri
ng th
e fi
rst 7
5 se
cond
s of
this
cyc
le. [e
] XR
D a
naly
sis
show
s th
at a
sep
arat
e m
etal
lic p
latin
um p
hase
is p
rese
nt.
Chapter 3.3 Characterisation
179
Selectivity towards hydrogen oxidation. In a typical reaction, 250 mg of
sample is placed on a quartz wool plug in a quartz reactor and heated to 550 °C in
1% v/v O2/Ar. The catalytic activity and product selectivity are determined by GC
and MS over nine redox cycles. Each cycle consists of a 10 min reduction step in
4:1:1% v/v C3H8:C3H6:H2 in Ar, and an 18 min oxidation step in 1%v/v O2/Ar,
separated by 4 min purge cycles. The 4:1:1 ratio of the reductive gases simulates
the effluent stream from industrial propane dehydrogenation.[16] Figure 1 shows the
time resolved conversion profile of catalyst 3 (Ce0.93Bi0.07O2), which is typical for
bismuth doped ceria catalysts (2, 3, 4, Table 1). The figure shows the conversion of
hydrogen (▲), propene (◊) and propane (○), during the reduction cycle. The
Ce-Bi-O catalysts have a good activity towards hydrogen combustion. Indeed, from
a set of 97 doped ceria catalysts, containing 26 different dopant elements, the Bi-
doped catalyst were amongst the most active, second only to lead doped ceria.[33, 35]
Lead, however, easily segregates from the ceria, forming a separate lead oxide
phase, which is unstable under redox cycling.[30] The Bi-doped cerias are more
stable, but convert part of the propene feed (grey area in Figure 1). Thus, having
discovered that bismuth doped ceria is an active, stable and non toxic SOR
catalysts, we set out to improve its selectivity.
0
25
50
75
100
0 200 400 600Time (s)
Con
vers
ion
(%)
Hydrogen, 1% v/v
Propene, 1% v/v
Propane, 4% v/v
T = 550 °C
0
25
50
75
100
0 200 400 600Time (s)
Con
vers
ion
(%)
Hydrogen, 1% v/v
Propene, 1% v/v
Propane, 4% v/v
T = 550 °C
Hydrogen, 1% v/v
Propene, 1% v/v
Propane, 4% v/v
T = 550 °C
Figure 1. Time resolved conversion profile of Ce0.93Bi0.07O2 (3) at 550 °C,
showing the H2 (▲), C3H6 (◊) and C3H8 (○) conversion during a reduction cycle. The grey
area indicates the level of propene conversion.
Chapter 3.3 Characterisation
180
Increasing the hydrogen concentration. Studies in hydroprocessing of
light gas oil showed that an increased concentration of hydrogen can limit coking,
possibly by shielding the catalyst surface from the hydrocarbons.[38] To check
whether this also holds for our doped cerias, we assessed the level of propene
conversion of catalyst 3 (Ce0.93Bi0.07O2) at 1–8% v/v of H2 (the equilibrium mixture
of the dehydrogenation contains 1% v/v of H2, 1% v/v C3H6 and 4% v/v C3H8).[39]
Figure 2 shows that indeed propene conversion, coking, and hydrocarbon
combustion are all lowered at higher H2 concentrations.[40] This strategy, however,
is not advantageous in the selective hydrogen combustion, since adding hydrogen
shifts the equilibrium towards the reactants side (see Scheme 1).
0
1
2
3
4
1 2 4 8C
ombu
stio
n(%
v/v
CO
2/ 1
0 m
in)
0.0
0.5
1.0
1.5
1 2 4 8Pro
pene
con
vers
ion
(% v
/v /
10 m
in)
0.0
0.1
0.2
1 2 4 8
H2 concentration (% v/v)
Cok
ing
(mg
C /
10 m
in)
Propene conversion Coking Combustion
0
1
2
3
4
1 2 4 8C
ombu
stio
n(%
v/v
CO
2/ 1
0 m
in)
0.0
0.5
1.0
1.5
1 2 4 8Pro
pene
con
vers
ion
(% v
/v /
10 m
in)
0.0
0.1
0.2
1 2 4 8
H2 concentration (% v/v)
Cok
ing
(mg
C /
10 m
in)
Propene conversion Coking Combustion
Figure 2. Propene conversion, coking and hydrocarbon combustion at various
hydrogen concentrations using Ce0.93Bi0.07O2 (3) (550 °C, 1% v/v propene and 4% v/v
propane). Note: all measurements were performed on one sample. To check if the
subsequent measurements have not affected the sample, the measurement at 1% H2 was
repeated, and gave similar results to the first measurement. The coking data pertains to the
total amount of carbon (mg), which is deposited on the catalyst surface during a single
reduction cycle. The combustion data is a summation of the amount of CO2 detected in 15
GC analyses performed during the 10 min reduction cycle. Note that most catalysts only
produce CO2 during the first 75 seconds of this cycle.
Chapter 3.3 Characterisation
181
Addition of tin. Tin is an essential promoter in the Pt-based
dehydrogenation catalysts, since it limits catalyst deactivation due to coking.[41] We
found that it has the same beneficial effect when applied as a dopant in the ceria
SOR catalysts. Tin is the only element, out of 26 tested, that prevents coking.[35]
Figure 3 shows the level of coking of plain ceria (1, reference), the Bi-doped
catalysts 2–4 and Ce0.88Bi0.08Sn0.04O2 (5). We have seen that both the activity and
the level of hydrocarbon combustion increase with increasing Bi–doping (catalysts
2–4, Table 1). However, the amount of coking is not correlated to the amount of
bismuth dopant, yet it is higher compared to plain ceria. This coking is completely
inhibited by the addition of 4 mol% of tin (5, Table 1). At too low Sn–doping
levels, coking is not prevented (6, 7, Table 1). Possibly, the 2 mol% tin is not
sufficient to cover all of the surface. Similar results were obtained with Cu–Sn
doped ceria (not shown). The coking is inhibited in case of a Ce0.87Cu0.08Sn0.05O2 /
SnO2 / CuO catalyst, and lowered, but not completely prevented, in case of a
Ce0.93Cu0.05Sn0.02O2 / SnO2 catalyst (this is probably due to the low Sn doping level
of 2 mol%). Note that the presence of separate tin–oxide (SnO2) does not alter the
ability to prevent coking (no coke was observed for a Ce0.90Sn0.10O2 / SnO2 catalyst,
not shown).
Since tin itself is active in the selective hydrogen combustion (6, 7), the
activity of the bismuth catalyst is increased upon addition of tin. Catalysts 3 and 5
contain roughly equal amounts of Bi, but adding 4 mol% of tin to catalyst 5
increases its activity by about 40%. Tin also increases hydrocarbon conversion via
combustion. That is, tin increases the activity and limits coking, but does not
increase the total selectivity of the catalysts. Note that the Bi-concentration also
affects the selectivity. The low loaded catalyst 2 (Ce0.98Bi0.02O2) is more selective
than 3 and 4, since the propene conversion drops quicker than the hydrogen
conversion (Table 1). These low doping levels, however, also result in a drop in
activity.
Chapter 3.3 Characterisation
182
1CeO2
2Ce0.98Bi0.02O2
3Ce0.93Bi0.07O2
4Ce0.90Bi0.10O2
5Ce0.88Bi0.08Sn0.04O2
0.00
0.05
0.10
0.15
0.20
0.25
Cok
ing
(mg
C /
10 m
in r
ed c
ycle
)
1CeO2
2Ce0.98Bi0.02O2
3Ce0.93Bi0.07O2
4Ce0.90Bi0.10O2
5Ce0.88Bi0.08Sn0.04O2
0.00
0.05
0.10
0.15
0.20
0.25
Cok
ing
(mg
C /
10 m
in r
ed c
ycle
)
Figure 3. Effect of Bi concentration and Sn addition on coking. Ceria is added as a
reference. The data reflect the total amount of carbon which is deposited on the catalyst
surface during one 10 min reduction cycle. The coking level of catalyst 5 is zero.
Addition of platinum. Doping ceria with noble metals such as Pt, Pd
and Ru results in unselective catalysts, with high levels of coking and cracking of
the hydrocarbons.[33, 35, 37] Surprisingly, this is not the case for the Pt-Bi doped ceria
8.[42] Figure 4 shows the time resolved conversion profiles of catalyst 8 and catalyst
9, Ce0.98Pt0.02O2. The conversion profile of 9 is typical for noble metal doped cerias,
showing high levels of hydrocarbon conversion, and formation of hydrogen (a
negative conversion), as a result of coking. Figure 4 shows that the Pt-Bi doped 8 is
much more selective than 9. Surprisingly, propene conversion over catalyst 8 is
also limited to the first 25 s of the reduction cycle, unlike the typical long term
propene conversion seen for Bi doped cerias (see Figure 1). The initial propane and
propene conversions of 8 are still higher, however, than those of Bi doped ceria
(compare Figure 4 with Figure 1). Since these unselective reactions also use up
oxygen, the activity of the Pt-Bi doped 8 is lower than that expected from its Bi
content (compare with the Bi doped catalysts 2–4). Still, the selectivity of the Pt-Bi
doped ceria 8 is dramatically higher than any Pt-doped ceria catalyst tested (11 in
total).[35] Indeed, alumina supported Pt is used in a common dehydrogenation
catalysts, but Sn needs be added to limit coking.[43] Our results show that adding
Bismuth can have the same effect.
Chapter 3.3 Characterisation
183
-300
-200
-100
0
100
200 400 600
Con
vers
ion
(%)
Ce0.98Pt0.02O2 (9)
Ce0.88Bi0.05Pt0.04O2 / Pt0 (8)
0
25
50
75
100
0 200 400 600Time (s)
Con
vers
ion
(%)
Hydrogen
Propene
Propane
-300
-200
-100
0
100
200 400 600
Con
vers
ion
(%)
Ce0.98Pt0.02O2 (9)
Ce0.88Bi0.05Pt0.04O2 / Pt0 (8)
0
25
50
75
100
0 200 400 600Time (s)
Con
vers
ion
(%)
Hydrogen
Propene
Propane
Figure 4. Time resolved conversion profiles of Ce0.98Pt0.02O2 (9, top) and
Ce0.91Bi0.05Pt0.04O2, (8, bottom), showing the H2 (▲), C3H6 (◊) and C3H8 (○) conversion
during a reduction cycle. Note: for catalyst 8, some hydrogen formation is observed in the
last part of the reduction cycle.
Table 2 shows the surface concentrations and binding energies of catalysts
3 (Ce-Bi-O), 9 (Ce-Pt-O), and 8 (Ce-Bi-Pt-O), as determined by XPS.[44] The
surface concentrations of the dopants are lower than the expected bulk value (see
Table 2). However, the surface concentration ratios of Bi in 3 and 8, and Pt in 8
and 9 are in accordance with their bulk concentrations. Analysis of the oxidation
states of the dopants shows that the bismuth is present as Bi0 in the singly doped
catalyst 3 (see Table 2). Since no bismuth metal is detected by XRD, the catalyst
surface likely consists of Bi0-clusters with a size below the XRD detection limit
(< 3 nm). The same holds for the platinum doped 9: metallic Pt is detected by XPS,
but not by XRD. Analysis of the binding energies of the Bi-Pt doped catalyst 8
Chapter 3.3 Characterisation
184
indicates that alloy formation occurs between the bismuth and platinum. There is a
significant chemical shift of +0.8 eV for both Bi and Pt surface species, with the
resulting bismuth binding energy in good agreement with that reported for Bi–Pt
alloys.[45, 46] The mutual perturbation of both dopants likewise evidences Bi–Pt
alloying. Note that the typical chemical shifts between metallic and oxidic bismuth
(Bi2O3) are much larger, ranging from 1.4 to 4.4 eV.[47-49] Although the presence of
bismuth oxide cannot be entirely excluded, the small chemical shift between
catalysts 3 and 8 points to 8 containing a mixture of Bi0 particles and Bi-Pt alloy
particles, and not bismuth oxide.[45, 46] The increased selectivity of the Bi-Pt catalyst
8 as compared to Pt doped cerias most likely results from alloy formation, and the
associated electronic (in addition to geometric) influence of bismuth upon
platinum.[50]
Cha
pter
3.3
Cha
ract
eris
atio
n
185
Tab
le 2
. Sur
face
con
cent
ratio
ns o
f th
e ca
taly
sts
com
pone
nts
as d
eter
min
ed b
y X
PS
and
bin
ding
ene
rgie
s of
Bi a
nd P
t.
Cat
alys
t C
ompo
siti
on
Sur
face
com
posi
tion
(%
of
tota
l)
Bin
ding
ene
rgy
(eV
)
Ce
O
Bi
Pt
C[a
] B
i 4f 7
/2
Pt 4
f 7/2
3 C
e 0.9
3Bi 0
.07O
2 22
.4
55.1
4.
6 -
17.8
15
6.4
-
9 C
e 0.9
8Pt 0
.02O
2 25
.6
53.8
-
0.7
20.0
-
70.2
8 C
e 0.9
1Bi 0
.05P
t 0.0
4O2/
Pt0
25.4
57
.3
3.4
1.2
12.7
15
7.3
71
[a] A
naly
sis
of h
igh
reso
lutio
n ca
rbon
C 1
s sp
ectr
a sh
ows
iden
tica
l car
bon
spec
ies
for
all s
ampl
es. T
he d
etec
ted
carb
on, t
here
fore
, mos
t lik
ely
stem
s fr
om a
com
mon
con
tam
inan
t an
d no
t fr
om s
peci
fic
Pt-
CxO
y or
Bi-
CxO
y sp
ecie
s. T
here
fore
, th
is c
omm
on c
onta
min
ant
will
not
aff
ect
the
conc
entr
atio
n ra
tios
pres
ente
d he
re.
Chapter 3.3 Characterisation
186
Variation of temperature. The selectivity of an SOR is governed by its
relative reactivity towards hydrogen versus hydrocarbons. A selective catalyst will
have a high hydrogen reduction rate, and a low hydrocarbon reduction rate at the
reaction temperature (550 °C). Previously, we showed that the reduction rate at
550 °C is related to the reducibility of the catalyst, as determined by H2-TPR.[37]
Catalysts with a low TPR reduction temperature, have a high reduction rate at
550 °C. Conversely, catalysts with a high reduction temperature (close to 550 °C),
have a low reduction rate at 550 °C. To understand what governs the redox cycle,
we performed both hydrogen and propene TPR on selected catalysts. In a typical
experiment, 250 mg of catalyst is heated from room temperature to 600 °C in either
5% v/v H2/Ar or 1% v/v C3H6/Ar. The hydrogen consumption or CO2 production is
determined by MS. The CO2 formation is chosen as a measure of the propene
conversion, since the MS response for CO2 is much higher than for C3H6, and both
profiles are identical. For better comparison, the hydrogen conversion is inverted.
We chose propene as a measure of the SOR reactivity towards hydrocarbons,
because apart from the initial unselective part, the catalysts do not convert propane
below 550 °C.
Figure 5 shows the reduction profiles of pure ceria, Ce0.93Bi0.07O2 (3) and
Ce0.92Pb0.08O2. Doping with Pb yields a highly selective catalyst, however with
lower stability (vide supra). Note that these data represent two separate
experiments, where fresh sample is reduced in either hydrogen or in propene, but
not in a mixture of the two. The data show that, for the unselective CeO2, the
reduction maximum of propene occurs at lower temperatures than that of hydrogen
(Figure 5, top). That is, this catalyst has a higher affinity for propene combustion
than hydrogen combustion (the hydrocarbon combustion rate will be higher).[37] In
contrast, for the selective Ce0.93Bi0.07O2 (3) this order is reversed: the reduction
maximum for hydrogen occurs before that of propene. Thus, hydrogen is reduced
faster than propene at 550 °C, and indeed the Bi-doped catalysts are much more
selective than plain ceria. Note that for the Bi–doped catalyst, the bulk reduction
temperature of propene lies at 500 °C, that is, still below the reaction temperature
of 550 °C. Indeed, the Bi–doped catalysts do convert part of the propene feed at
550 °C (see also Figure 1). The reduction maximum of hydrogen of the selective
Pb–doped ceria also lies below that of propene (Figure 5, bottom), and for this
Chapter 3.3 Characterisation
187
catalyst, the bulk reduction temperature of propene is higher compared to the Bi-
doped catalysts. Indeed, Pb–doped cerias are even more selective, albeit less stable
than the Bi–doped ones.
Ce0.93Bi0.07O2 (3)
100 200 300 400 500 600
HydrogenPropene
Temperature (°C)
MS
Sig
nal
CeO2
100 200 300 400 500 600
Hydrogen Propene
MS
Sig
nal
Ce0.92Pb0.08O2
PropeneHydrogen
MS
Sig
nal
100 200 300 400 500 600
Ce0.93Bi0.07O2 (3)
100 200 300 400 500 600
HydrogenPropene
Temperature (°C)
MS
Sig
nal
CeO2
100 200 300 400 500 600
Hydrogen Propene
MS
Sig
nal
Ce0.92Pb0.08O2
PropeneHydrogen
MS
Sig
nal
100 200 300 400 500 600
Figure 5. H2–TPR and C3H6–TPR for ceria (top), Ce0.93Bi0.07O2 (3, middle) and
Ce0.92Pb0.08O2 (bottom). Fresh samples were heated from 25 °C to 600 °C in either 5% v/v
H2/Ar or 1% v/v C3H6/Ar. The CO2 evolution is taken as a measure for the propene
Chapter 3.3 Characterisation
188
conversion. The H2 profile is inverted for better comparison. Note that above 550 °C, 3
shows a net formation of hydrogen. Hydrogen adsorption by ceria below 400 °C, and its
subsequent release at higher temperatures has been reported by several other authors.[51, 52]
The hydrogen reduction peak of catalyst 3 between 200–400 °C is indeed caused by
hydrogen combustion, since an identical water peak is observed by MS (from 400 °C
onwards, no water is detected).
The reduction profiles of Ce0.90Bi0.10O2 (4) show that the selectivity of this
catalyst may be optimised by selecting the appropriate reaction temperature. At
temperatures below ~400 °C, the catalyst should be ineffective for propene
combustion, but still able to combust hydrogen, affording high selectivity. Above
400 °C, propene combustion should dominate, lowering selectivity to the desired
H2 combustion. To test this hypothesis, we ran five selective hydrogen combustion
experiments at 200–600 °C (see Figure 6, Table 3). At 200 °C, TPR shows neither
hydrogen or propene combustion, and the catalyst is inactive (Figure 6). At 300 °C,
the H2-TPR shows that Ce0.90Bi0.10O2 burns hydrogen, and indeed the catalyst is
active (bottom row). Note this temperature is still below the hydrogen combustion
maximum. At 400 °C the catalyst is more active, but still selective. At 550 °C and
600 °C, significant propene combustion and coking is visible from both the TPR
and catalytic experiments (numerical data is shown in Table 3).
Cha
pter
3.3
Cha
ract
eris
atio
n
189
100
300
400
500
600
°C
MS Signal
Pro
pene
H2
200
300
400
550
600
°C
Ce 0
.93B
i 0.0
7O2
(3)
200
300
°C40
0 °C
550
°C60
0 °C
255075100
020
040
060
0
Tim
e (s
)
Conversion (%)
255075100 0
200
400
600
Tim
e (s
)
255075100
020
040
060
0
Tim
e (s
)T
ime
(s)
Hyd
roge
n
Pro
pene
Pro
pane
255075100 0
200
400
600
Tim
e (s
)
200
400
600
0
255075100
200
°C
100
300
400
500
600
°C
MS Signal
Pro
pene
H2
200
300
400
550
600
°C
Ce 0
.93B
i 0.0
7O2
(3)
200
300
°C40
0 °C
550
°C60
0 °C
255075100
020
040
060
0
Tim
e (s
)
Conversion (%)
255075100 0
200
400
600
Tim
e (s
)
255075100
020
040
060
0
Tim
e (s
)T
ime
(s)
Hyd
roge
n
Pro
pene
Pro
pane
255075100 0
200
400
600
Tim
e (s
)
200
400
600
0
255075100
200
°C
F
igu
re 6
. T
op:
H2–
TP
R a
nd p
rope
ne–T
PR
of
Ce 0
.93B
i 0.0
7O2
3. B
otto
m:
sele
ctiv
e hy
drog
en c
ombu
stio
n da
ta o
f C
e 0.9
0Bi 0
.10O
2 (4
) at
vari
ous
tem
pera
ture
s. T
he T
PR
dat
a of
3 i
s sh
own,
sin
ce a
sep
arat
e B
i 2O
3 ph
ase
had
form
ed i
n ca
se o
f ca
taly
st 4
aft
er t
he c
atal
ytic
mea
sure
men
ts a
t 60
0 °C
. N
ote
that
the
re i
s no
t m
uch
diff
eren
ce b
etw
een
the
TP
R d
ata
of t
he f
resh
, m
onop
hasi
c 3
and
spen
t, bi
phas
ic 4
,
exce
pt f
or tw
o ex
tra
redu
ctio
n fe
atur
es a
t abo
ut 4
20 a
nd 5
80 °
C in
cas
e of
the
spen
t 4, s
ee F
igur
e 7.
Chapter 3.3 Characterisation
190
Table 3. Catalytic data of Ce0.90Bi0.10O2 (4) at various temperatures.
Reaction
temperature
(°C)[a]
Selectivity
(%)
Hydrogen activity
(%H2 combusted)
Coking
(mg C / 10 min)
Combustion
(vol% CO2 / 10
min)
200 n.a.[b] 0 0.00 0
300 100 45 0.03 0
400 98 89 0.06 0.5
550 77 33 0.16 4.0
600 56 14 0.21 5.7 [a] All measurements were performed on the same sample, in the order 550 °C, 200 °C,
300 °C, 400 °C, 600 °C. After the 600 °C measurement, a yellow band was observed at the
reactor exit, and XRD analysis of the sample showed that a separate Bi2O3-phase had
formed. [b] Not applicable.
Temperature (°C)
MS
Sig
nal
PropeneHydrogen
100 200 300 400 500 600
Hydrogen Propene
Ce0.90Bi0.10O2 (4)
Spent, Bi2O3 present
Ce0.93Bi0.07O2 (3)
FreshMS
Sig
nal
100 200 300 400 500 600
Temperature (°C)
MS
Sig
nal
PropeneHydrogen
100 200 300 400 500 600
Hydrogen Propene
Ce0.90Bi0.10O2 (4)
Spent, Bi2O3 present
Ce0.93Bi0.07O2 (3)
FreshMS
Sig
nal
100 200 300 400 500 600
Figure 7. TPR of fresh Bi–doped (3, top) and spent Bi–doped ceria (4, bottom),
which has a separate Bi2O3 phase. Interestingly, the hydrogen formation of 3 above 550 °C
is not observed for the spent catalyst 4.
Chapter 3.3 Characterisation
191
Conclusions Bismuth doped ceria (Bi concentration 2-10 mol%) has great potential as the solid
‘oxygen reservoir’ component in a novel catalytic process for propane
dehydrogenation. While hydrogen combustion activity rises with increasing Bi-
content from 2% to 10 mol%, this also results in higher hydrocarbon conversion.
We explore four strategies to increase the selectivity and/or activity of the
Ce–Bi–O catalysts. First, the addition of extra hydrogen decreases hydrocarbon
conversion by suppressing both coking and combustion pathways. However, this
strategy is not advantageous in a dehydrogenation process since it shifts the
equilibrium towards the reactants. Second, we found that addition of 4–7 mol% tin
completely inhibits hydrocarbon coking, and increases hydrogen combustion.
Unfortunately tin also boosts the level of hydrocarbon combustion. Third, adding
Pt increases selective H2 combustion, but lowers net activity. The best performance
can be obtained by controlling the reaction temperature. The optimal temperature
for Bi doped ceria is 400 °C, wherein Ce0.90Bi0.10O2 shows 98% selectivity and
converts 90% of the hydrogen feed. Indeed for the majority of the reduction cycle,
this catalyst converts all the hydrogen without converting any propane or propene.
That is, at 400 °C, the Bi doped catalyst does not require additional tin, platinum
nor extra hydrogen to achieve high selectivity and activity. The optimal reaction
temperature can be rationalised from hydrogen and propene TPR measurements:
400 °C lies in between the reduction maxima for hydrogen (< 400 °C) and propene
(> 400 °C), resulting in high activity and selectivity to hydrogen combustion.
Similar trends were observed for Ce0.92Pb0.08O2. For undoped ceria this phenomena
is reversed: the propene reduction maximum lies below that for hydrogen
reduction, resulting in a higher reactivity towards hydrocarbon versus hydrogen
combustion, and thus an unselective catalyst.
Chapter 3.3 Characterisation
192
Experimental Section Materials and instrumentation. Chemicals were purchased from Sigma-
Aldrich or Merck and used as received. Gases were purchased from Praxair and
had a purity of 99.5% or higher. The O2, He, Ar and N2 streams were purified
further over molsieves and/or BTS columns. Powder X-ray diffraction
measurements were performed using a Philips PW-series X-ray diffractometer with
a Cu tube radiation source (λ = 1.54 Å), a vertical axis goniometer and a
proportional detector. The 2θ detection measurement range was 10 ° – 93 ° with a
0.02 ° step size and a 5 second dwell time. Lattice constants and crystallite sizes
were obtained after Rietveld refinement (structure fit) using PANalytical's X'pert
software package. GC analysis was performed on an Interscience CompactGC
equipped with TCD detectors, separating water, CO2 and C2 and C3 hydrocarbons
on a Porabond Q column (He carrier gas) and H2, CO, CH4, O2 and N2 on a 5 Å
molsieve column (Ar carrier gas). MS analysis was performed using a Pfeiffer
QMS 200 mass spectrometer (m/z range 0–200). X-ray photoelectron spectra were
recorded on a Kratos HSi spectrometer equipped with a charge neutraliser and
monochromated Al K X-ray source (1486.61 eV) operating at 144 W. Spectra
were recorded with a pass energy of 40 eV at normal emission, and energy
referenced to the valence band and adventitious carbon. Analysis was conducted
using CasaXPS Version 2.3.15.
Procedure for catalyst synthesis. The procedure for catalyst preparation
was described in detail previously.[33, 35] The metal nitrates (chloride in case of Sn)
are weighed in a porcelain crucible and heated to about 100 °C, so that the cerium
nitrate melts. The mixture is stirred until all components are dissolved or have
melted, in case of Sn, 2–4 drops of water are added to aid the dissolution. The
crucible is placed in a vacuum oven set at 140 °C and the pressure is carefully
lowered to < 10 mbar (in 10–15 min), making sure no vigorous boiling occurs.
After 4 h, the samples are placed in a furnace and calcined under static air at
700 °C (ramp rate 300 °C/h, 5 h hold). The resulting solid is pulverized, ground
and sieved in fractions of 125–212 µm (selectivity assessment and TPR) and
< 125 µm (XRD measurements).
Procedure for testing catalytic activity. The activity and selectivity were
determined using an automated cyclic redox reactor system built in-house,
Chapter 3.3 Characterisation
193
described in detail elsewhere.[33] In a typical experiment, about 250 mg of sample
(125–212 μm) was placed on a quartz wool plug in a 4 mm id quartz reactor. The
reactor was placed in a water cooled oven and heated to 550 °C at 1200 °C/h, under
oxygen flow. At this temperature, redox cycling was started. The selectivity was
determined by GC during the 10 min reduction in 4:1:1% v/v C3H8:C3H6:H2 in Ar
(total flow 50 mL/min), with 5 mL/min of N2 added as internal standard. The gas
hourly space velocity (GHSV) is 13200 / h (at the typical bed volume of 0.25 cm3
and the reduction cycle's total flow of 55 mL/min). The weight hourly space
velocity (WHSV) is 1.2 / h, and is calculated from the weight of C3H8 + C3H6 + H2
per h per the weight of the catalyst. The 4:1:1 ratio of reductive gases is chosen
since this is the equilibrium mixture of a conventional dehydrogenation catalyst.[16]
After a 4 min purge step (pure Ar), the sample was reoxidised for 18 minutes in
1% v/v O2 in Ar (50 mL/min total flow). The redox cycle is completed by another
purge step in pure Ar. The selectivity is determined as the ratio H2 conversion:total
conversion. Activity is determined as the percentage hydrogen combusted during
the reduction step. Both selectivity and activity are averaged over eight redox
cycles. The amount of coking was assessed by the amount of CO and CO2 formed
in the reoxidation step, determined by MS. All signals were normalised using a
small amount of helium added to the gas feed (1 vol %), and were integrated using
CasaXPS v2.1.18 software.
Procedure for TPR experiments. TPR experiments were performed in the
same set up where catalytic activity was tested. In a typical experiment, 250 mg
sample (125–212 μm) was placed on top of a quartz wool plug in a 4 mm id quartz
reactor. Either 5% v/v H2/Ar or 1% v/v C3H6/Ar was fed over the reactor bed, and
the sample was heated from room temperature to 600 °C at 10 °C/min. The
hydrogen consumption (H2–TPR) or CO2 evolution (C3H6–TPR), as determined by
MS, were used to obtain the TPR profile.
Procedure for XPS experiments. XPS was performed on 50 mg sample.
The electron analyser pass energy was 160 eV for wide scans and 40 eV for high
resolution spectra. Compositions were corrected using the appropriate elemental
response factors on spectra following a Shirley background-subtraction.
Chapter 3.3 Characterisation
194
Acknowledgements We thank Dr. K. Wilson (University of York) for assistance with XPS
measurements, A.C. Moleman and W.F. Moolhuijzen for help with the XRD
measurements, and NWO–ASPECT for financial support and feedback.
Chapter 3.3 Characterisation
195
References [1] P. C. Andrews, G. B. Deacon, C. M. Forsyth, P. C. Junk, I. Kumar and M.
Maguire, Angew. Chem., Int. Edit., 2006, 45, 5638. [2] W. C. Smith, Metals and Alloys, 1945, 22, 387. [3] W. X. Dong, Y. W. Shi, Z. D. Xia, Y. P. Lei and F. Guo, J. Electron Mater., 2008,
37, 982. [4] R. M. Hua, Curr. Org. Synth., 2008, 5, 1. [5] M. Rueping, B. J. Nachtsheim and W. Ieawsuwan, Adv. Synth. Catal., 2006, 348,
1033. [6] H. Gaspard-Iloughmane and C. Le Roux, Eur. J. Org. Chem., 2004, 2517. [7] T. A. Hanna, Coordin. Chem. Rev., 2004, 248, 429. [8] E. Capoen, M. C. Steil, G. Nowogrocki, M. Malys, C. Pirovano, A. Lofberg, E.
Bordes-Richard, J. C. Bolvin, G. Mairesse and R. N. Vannier, Solid State Ionics, 2006, 177, 483.
[9] M. C. Steil, F. Ratajczak, E. Capoen, C. Pirovano, R. N. Vannier and G. Mairesse, Solid State Ionics, 2005, 176, 2305.
[10] E. Capoen, G. Nowogrocki, R. J. Chater, S. J. Skinner, J. A. Kilner, M. Mays, J. C. Boivin, G. Mairesse and R. N. Vannier, Solid State Ionics, 2006, 177, 489.
[11] J. Chmielowiec, G. Paściak and P. Bujło, J. Alloy. Compd., 2008, 451, 676. [12] V. V. Kharton, F. M. B. Marques and A. Atkinson, Solid State Ionics, 2004, 174,
135. [13] A. Lofberg, S. Boujmiai, E. Capoen, M. C. Steil, C. Pirovano, R. N. Vannier, G.
Mairesse and E. Bordes-Richard, Catal. Today, 2004, 91-92, 79. [14] A. Lofberg, H. Bodet, C. Pirovano, M. C. Steil, R. N. Vannier and E. Bordes-
Richard, Catal. Today, 2006, 118, 223. [15] L. Ji, J. S. Liu, C. Y. Liu and X. C. Chen, Appl. Catal. A: Gen., 1994, 114, 207. [16] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 1. [17] J. Plotkin and E. Glatzer, Eur. Chem. News, 2005, 82, 20. [18] N. Alperowicz, Chem. Week, 2007, 169, 27. [19] N. Alperowicz, Chem. Week, 2006, 168, 17. [20] G. Parkinson, Chem. Eng. Prog., 2004, 100, 8. [21] F. Cavani, N. Ballarini and A. Cericola, Catal. Today, 2007, 127, 113. [22] R. Grabowski, Catal. Rev. - Sci. Eng., 2006, 48, 199. [23] N. R. Shiju, M. Anilkumar, S. P. Mirajkar, C. S. Gopinath, B. S. Rao and C. V.
Satyanarayana, J. Catal., 2005, 230, 484. [24] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Abstr. Pap. Am. Chem. S.,
1999, 217, U687. [25] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 9.
Chapter 3.3 Characterisation
196
[26] J. G. Tsikoyiannis, D. L. Stern and R. K. Grasselli, J. Catal., 1999, 184, 77. [27] C. H. Lin, K. C. Lee and B. Z. Wan, Appl. Catal. A: Gen., 1997, 164, 59. [28] L. Låte, J. I. Rundereim and E. A. Blekkan, Appl. Catal. A: Gen., 2004, 262, 53. [29] L. Låte, W. Thelin and E. A. Blekkan, Appl. Catal. A: Gen., 2004, 262, 63. [30] L. M. van der Zande, E. A. de Graaf and G. Rothenberg, Adv. Synth. Catal., 2002,
344, 884. [31] E. A. de Graaf, A. Andreini, E. J. M. Hensen and A. Bliek, Appl. Catal. A: Gen.,
2004, 262, 201. [32] G. Rothenberg, E. A. de Graaf and A. Bliek, Angew. Chem., Int. Ed., 2003, 42,
3366. [33] J. H. Blank, J. Beckers, P. F. Collignon, F. Clerc and G. Rothenberg, Chem. Eur.
J., 2007, 13, 5121. [34] A. Trovarelli, C. de Leitenburg, M. Boaro and G. Dolcetti, Catal. Today, 1999, 50,
353. [35] J. Beckers, F. Clerc, J. H. Blank and G. Rothenberg, Adv. Synth. Catal., 2008, 350,
2237. [36] J. Beckers and G. Rothenberg, Dalton Trans., 2008, 6573. [37] J. H. Blank, J. Beckers, P. F. Collignon and G. Rothenberg, ChemPhysChem,
2007, 8, 2490. [38] B. M. Vogelaar, J. Gast, E. M. Douma, A. D. van Langeveld, S. Eijsbouts and J.
A. Moulijn, Ind. Eng. Chem. Res., 2007, 46, 421. [39] We measured the propene conversion since the level of propane conversion is
negligible. [40] The drop in carbon deposition with increased hydrogen concentration shown in
Figure 2 can also be the result of the reaction of the coke or coke precursors with the added hydrogen (J. Barbier, E. Churin and P. Marecot, J. Catal., 1990, 126, 228). The propene conversion, however, is independent of the coke formation, and also drops with increased hydrogen concentration.
[41] L. W. Lin, T. Zhang, J. L. Zang and Z. S. Xu, Appl. Catal., 1990, 67, 11. [42] Note that part of the Pt is present as Pt0. [43] E. A. de Graaf, G. Rothenberg, P. J. Kooyman, A. Andreini and A. Bliek, Appl.
Catal. A: Gen., 2005, 278, 187. [44] E. B. Fox, A. F. Lee, K. Wilson and C. S. Song, Top. Catal., 2008, 49, 89. [45] U. W. Hamm, D. Kramer, R. S. Zhai and D. M. Kolb, Electrochim. Acta, 1998, 43,
2969. [46] A. V. Tripković, K. D. Popović, R. M. Stevanović, R. Socha and A. Kowal,
Electrochem. Commun., 2006, 8, 1492. [47] V. S. Dharmadhikari, S. R. Sainkar, S. Badrinarayan and A. Goswami, J. Electron.
Spectrosc., 1982, 25, 181. [48] W. E. Morgan, W. J. Stec and J. R. V. Wazer, Inorg. Chem., 1973, 12, 953.
Chapter 3.3 Characterisation
197
[49] F. M. Ismail and Z. M. Hanafi, Z. Phys. Chem. -Leipzig, 1986, 267, 667. [50] N. F. Dummer, R. Jenkins, X. B. Li, S. M. Bawaked, P. McMorn, A. Burrows, C.
J. Kiely, R. P. K. Wells, D. J. Willock and G. J. Hutchings, J. Catal., 2006, 243, 165.
[51] F. M. Z. Zotin, L. Tournayan, J. Varloud, V. Perrichon and R. Frety, Appl. Catal. A: Gen., 1993, 98, 99.
[52] K. Sohlberg, S. T. Pantelides and S. J. Pennycook, J. Am. Chem. Soc., 2001, 123, 6609.
198
199
3.4
Particle size and dopant concentration effects on
the catalytic properties of ceria based solid
‘oxygen reservoirs’
CeO2
C CC
CO2 (g)
H2O (g)
H2O (g)
550 °CH2
CeO2
Ce-Cu-O2
Ce-Cr-O2
CeO2
C CC
CO2 (g)
H2O (g)
H2O (g)
550 °CH2
CeO2
Ce-Cu-O2
Ce-Cr-O2
C CC
CO2 (g)
H2O (g)
H2O (g)
550 °CH2
CeO2
Ce-Cu-O2
Ce-Cr-O2
Chromium- or copper-doped ceria can be applied in selective hydrogen
combustion, part of a novel process for oxidative dehydrogenation. Pure ceria is
unselective: small crystals mainly coke, and large crystals mainly combust the
hydrocarbons present. By doping with 3–5% of Cr or Cu, we can increase the
catalysts' selectivity, activity and stability.
This work has been published as:
'Ce0.95Cr0.05O2 and Ce0.97Cu0.03O2: Active, selective and stable catalysts for selective
hydrogen combustion', Jurriaan Beckers and Gadi Rothenberg. Dalton Trans. 2009,
5673.
Chapter 3.4 Characterisation
200
Abstract
Ceria based materials are promising solid ‘oxygen reservoirs’ for propane oxidative
dehydrogenation. The ceria lattice oxygen can selectively combust hydrogen from
the dehydrogenation mixture at 550 °C. This has three key advantages: it shifts the
dehydrogenation equilibrium to the desired products side, generates heat aiding the
endothermic dehydrogenation, and simplifies product separation (H2O vs. H2).
Furthermore, the process is safer since it does not require mixing gaseous O2 and
H2 at high temperatures.
Ceria itself is unselective, but its catalytic properties can be tuned by doping. Here
we report the effects of dopant type, concentration and the crystallite size on the
catalytic properties. Plain ceria has an optimal crystallite size of about 20 nm
(where both hydrocarbon coking and combustion are minimised). Doping with Cr
or Cu increases both the selectivity and activity of the ceria, albeit that propane
combustion also increases linearly with the Cu-concentration. The Cu-doped
catalysts give selectivities up to 95% and combust up to 8% of the hydrogen feed.
The best results are obtained with Cr-doped ceria, with selectivities up to 98%, and
combusting up to 15% of the hydrogen feed. The larger Cr-doped crystallites (up to
50 nm), show the least amount of coking, and the highest activity. Importantly, the
Cr-doped catalysts are stable in the reductive gas feed. No extra coke is formed
when the catalyst is subjected to an extra 10 min in the dehydrogenation mixture,
after the hydrogen combustion reaction has stopped. This robustness is essential for
any catalytic application in industrial dehydrogenation.
Besides the beneficial effects on the selectivity and activity, doping with Cr or Cu
also increases the sinter stability of the ceria. The sinter stability of the small
crystals is increased most notably by Cu-doping.
Chapter 3.4 Characterisation
201
Introduction Propene is an important bulk chemical. It is a building block for a variety
of compounds, but is primarily used for producing polypropene.[1] The annual
propene demand is expected to rise to 80 million tonnes in 2010 worldwide.[2-4]
About 95% of the current production of propene is supplied by crackers and oil
refineries, but more advantageous methods such as metathesis and catalytic
dehydrogenation of propane are gaining ground. These allow for on-demand
production of high purity monomer.[5, 6] The catalytic dehydrogenation of propane
usually employs a platinum or chromium oxide catalyst on alumina, typically at
550–600 °C.[7-9] Unfortunately, this is an endothermic, equilibrium limited reaction.
One way to overcome these problems is by selectively burning the hydrogen by-
product (Scheme 1, left).[10-17] This generates energy and shifts the equilibrium to
the desired products side. Gaseous oxygen can be used, but mixing oxygen,
propane, propene and hydrogen at high temperatures and in the presence of a
catalyst is dangerous. A better solution is applying a redox process, where a solid
‘oxygen reservoir’ (SOR) is added to the dehydrogenation catalyst. Now, instead of
using gaseous oxygen, the lattice oxygen of the SOR does the selective hydrogen
combustion. This is safer, and allows for separate tuning of the dehydrogenation
and selective hydrogen combustion processes using two catalysts. Following the
reduction of the SOR lattice, the oxygen vacancies are replaced using air, creating
a cyclic redox process (Scheme 1, right).
Chapter 3.4 Characterisation
202
Energy H2
H2O + SOR SOR-O
Dehydrogenationcatalyst
Propane Propene
CnH(2n+2)
A
N2O2N2
B C D
N2O2
COx
N2CnH2n
H2O
Fresh SOR
Fresh DH
Spent SOR
Spent DH
Reduction ReoxidationPurge Purge
Energy H2
H2O + SOR SOR-O
Dehydrogenationcatalyst
Propane Propene
CnH(2n+2)
A
N2O2N2
B C D
N2O2
COx
N2CnH2n
H2O
Fresh SOR
Fresh DH
Spent SOR
Spent DH
Reduction ReoxidationPurge Purge
Scheme 1. Left: dehydrogenation combined with selective hydrogen combustion. The
selective hydrogen combustion consumes part of the hydrogen formed during the
dehydrogenation step, shifting the equilibrium to the products side and generating heat.
Right: cartoon of the complete redox cycle. After the dehydrogenation step (A), the bed is
flushed with nitrogen (B), and the catalysts are regenerated through reoxidation (C). This
burns coke from the dehydrogenation catalyst and restores the lattice oxygen of the SOR
catalyst. After another nitrogen flush (D) the reactor is ready for the next redox cycle.
Selective hydrogen combustion can be performed by supported metal
oxides, such as Bi2O3/SiO2, but these sinter under redox cycling.[11-14, 18] Ceria has
high temperature stability and a facile Ce3+ Ce4+ + e– reaction, making it a
good solid oxygen reservoir.[19] CeO2 itself, however, is not selective, but we
showed that selectivity, activity and stability can be tuned by doping the ceria
lattice with different cations.[20, 21] Doping with elements such as Cr, Cu, and Bi
yields active and selective catalysts, where doping with elements such as Pd, Pt and
Fe yields catalysts with a low selectivity.[21, 22] The dopant acts as active site for the
selective hydrogen combustion, but also affects physical properties of the ceria,
such as the crystallite size.[21] Temperature programmed reduction (TPR) studies
showed that this also affects the redox properties of the catalysts.[23-25] The above
experiments were performed using a simple model system, with hydrogen or
propene as reducing agent. In the actual selective hydrogen combustion, however, a
mixture of propane, propene and hydrogen is present. Here, we investigate the
effect of the crystallite size, dopant type, and the dopant concentration on the
Chapter 3.4 Characterisation
203
catalytic properties in the selective hydrogen combustion reaction conditions. We
compare the activity, selectivity, and stability of three catalyst types: plain CeO2,
Cr-doped ceria and Cu-doped ceria, and discuss the factors that govern catalyst
performance.
Results and Discussion Catalyst preparation and characterisation. The doped ceria catalysts 1–
18 were prepared by co–melting mixtures of the metal nitrate hydrate precursors.[26,
27] After the precursor has liquefied, the pressure was lowered and a solid mixed
metal nitrate formed. This was converted into the mixed oxide by calcining in static
air at 700 °C for 5 h. Figure 1 shows pictures of three catalysts before (left, mixed
nitrates), and after calcination (right, mixed oxides). Whereas pure cerium nitrate is
white, and CeO2 is pale yellow, the mixed oxides show a variety of colours. After
the vacuum drying, the copper-containing catalyst precursors are still blue, as is the
copper nitrate starting material (Figure 1, bottom left). After calcination the
catalyst is black (Figure 1, bottom right). This indicates that CuO is present instead
of Cu2O (the latter is yellow/red[28]).
Each catalyst was characterised using powder X-ray diffraction, to confirm
it consists of a uniform phase. That is, only the diffraction lines of the cerias
fluorite structure are observed, and no separate oxides of the dopants are present.
Importantly, the catalysts were not prepared by impregnating CeO2 supports. The
co-melting of the cerium nitrate with the nitrates of the appropriate metals yields a
well mixed liquid catalyst precursor. This allows for incorporation of the dopants
into the ceria fluorite structure after calcination.
In total, we synthesised eighteen catalysts in three sets, all varying in
crystallite size: first, a set of undoped ceria catalysts as reference (1–6), second a
set of 5 mol% Cr-doped catalysts (7–12), and third a set of 0.1–10 mol% Cu–doped
ceria (13–18).
Chapter 3.4 Characterisation
204
3
16
8
3
16
8
Figure 1. Photos of three catalysts (3 CeO2, 8 Ce0.95Cr0.05O2 and 16 Ce0.93Cu0.07O2)
after heating under reduced pressure (left, mixed nitrates) and calcining (right, mixed
oxides).
Crystallite size effects on activity and selectivity. First, we studied the
relationship between crystallite size and selectivity in the selective hydrogen
combustion, by synthesising a set of undoped ceria catalyst with increasing
crystallite size. Table 1 shows the properties of the undoped catalysts 1–6. The
crystallite size was adjusted by varying the calcination temperature (catalyst 1 and
2), or by exposing the catalyst to hydrogen at 800 °C over extended periods of time
(catalysts 6a and 6b, which were prepared by treating 6 in 66% hydrogen at 800 °C
for 4 and 16 h, respectively). Catalysts 3–6 reflect the sample-to-sample variation
in crystallite size. In a typical selective hydrogen combustion reaction, a stream
simulating the effluent from industrial propane dehydrogenation (4:1:1% v/v
Chapter 3.4 Characterisation
205
C3H8:C3H6:H2 in Ar at 50 mL/min total flow rate) was fed to the reactor containing
~250 mg of doped ceria at 550 ºC. After this 10 min reduction step, the reactor was
purged with Ar, followed by an oxidation step (1% v/v O2 in Ar, 18 min) and a
second purge step with Ar to complete the cycle. The reaction was monitored using
mass spectrometry (MS, reoxidation step) and online gas chromatography (GC,
reduction step). These are used to assess the amount of hydrocarbon combustion,
coking, and hydrogen combustion (Scheme 2). The selectivity is defined
as 10083632
2 HCHCH
H
conversion
conversion . The amount of hydrocarbon combustion is determined
by measuring the amount of CO2 formed in the reduction step (no CO observed).
The coke is quantified by measuring the amount of CO and CO2, originating from
the combustion of the coke, in the oxidation step. The activity of the catalysts is
given by two parameters. The first is the so-called ‘oxygen demand’, which is the
amount of oxygen used in the oxidation step for refilling the reduced lattice
vacancies as well as for combusting the coke. Thus, the oxygen demand represents
both selective and unselective processes. The second is the ‘hydrogen activity’,
representing the percentage of the hydrogen feed combusted by each catalyst.
Cha
pter
3.4
Cha
ract
eris
atio
n
206
T
able
1. P
hysi
cal a
nd c
atal
ytic
pro
pert
ies
of p
lain
cer
ia c
atal
ysts
.
Cat
alys
t C
alci
natio
n
tem
pera
ture
(°C
)
Red
uctio
n tim
e at
800
°C /
66%
H2
(h)
crys
tall
ite
size
(nm
)
Sur
face
are
a
(m2 /g
)
Lat
tice
spa
cing
(Å)
Oxy
gen
dem
and
(mol
O /
kg)[a
]
1 55
0 0
9.8
84
5.40
98
0.40
2 62
5 0
12.3
75
5.
4102
0.
36
3 70
0 0
18.3
54
5.
4091
0.
28
4 70
0 0
22.0
39
5.
4106
0.
26
5 70
0 0
26.6
43
5.
4089
0.
38
6 70
0 0
29.9
22
5.
4090
0.
36
6a[b
] 70
0 4
188
0[c]
5.40
99
0.18
6b[d
] 70
0 16
24
5 0[c
] 5.
4116
0.
13
[a] T
he o
xyge
n de
man
d is
def
ined
as
the
amou
nt o
f ox
ygen
whi
ch i
s co
nsum
ed b
y th
e ca
taly
st i
n th
e re
oxid
atio
n st
ep.
It i
s th
e am
ount
of
oxyg
en n
eede
d fo
r re
fill
ing
the
latt
ice
and
com
bust
ing
the
coke
pre
sent
on
the
cata
lyst
sur
face
. [b
] Sam
e as
cat
alys
t 6,
but
tre
ated
for
4 h
at
800
ºC u
nder
flo
win
g hy
drog
en. [c
] The
sur
face
are
a of
thes
e ca
taly
sts
is to
o lo
w to
be
accu
rate
ly d
eter
min
ed. [d
] A s
econ
d ba
tch
of c
atal
yst 6
,
trea
ted
for
16 h
at 8
00 º
C u
nder
flo
win
g hy
drog
en.
Chapter 3.4 Characterisation
207
H2
SOR-O
(De)hydrogenation
C CC
SOR-O
Coking
COx
SOR-
Hydrocarboncombustion
H2O
SOR-
Hydrogencombustion
H2H2 CH3
Cracking
SOR-O
H2
SOR-O
(De)hydrogenation
C CC
SOR-O
Coking
COx
SOR-
Hydrocarboncombustion
H2O
SOR-
Hydrogencombustion
H2COx
SOR-SOR-
Hydrocarboncombustion
H2O
SOR-
Hydrogencombustion
H2H2 CH3
Cracking
SOR-O
CH3
Cracking
SOR-O
Scheme 2. Cartoon showing possible interactions between the dehydrogenation gas
mixture and the SOR catalyst. The so-called oxygen demand is the total amount of catalyst
oxygen used by the processes.
The undoped ceria is unselective, all catalysts shown in Table 1 give a
negligible conversion of hydrogen but do convert the hydrocarbons, mainly at the
beginning of the reduction cycle. Both hydrocarbon combustion and coking occur.
Figure 2 shows the level of coking, combustion and oxygen demand of the
undoped ceria catalysts against crystallite size. The data show that coking and
combustion are related to crystallite size. The small crystals primarily coke the
hydrocarbons (Figure 2, top), while large ones primarily combust them (Figure 2,
middle). Note that the trends remain when the data is normalised for surface area.
The oxygen demand (Figure 2, bottom), is the sum of the combustion and coking
processes. Indeed, the curve is u-shaped, stemming mainly from coking for the
small crystallites, and mainly from combustion for the larger crystallites. Figure 2
also shows that there is an optimal size for the ceria crystallites of around 20 nm,
where both of these unwanted processes are minimised. This can be taken into
account when using ceria in reactions with hydrocarbons. Note that 6a and 6b,
containing very large crystallites, show even lower amounts of coking and
combustion than the ones presented in Figure 2. However, their low surface area
renders them unsuited as catalyst supports.
Chapter 3.4 Characterisation
208
0.15
0.25
0.35
0.45
5 10 15 20 25 30 35Crystallite size (nm)
Oxy
gen
dem
and
(mol
O /
kg) Oxygen demand
0.5
1.5
2.5
3.5
4.5
5 10 15 20 25 30 35
Crystallite size (nm)
CO
2fo
rmed
(%
v/v
)
Combustion
Coking
Crystallite size (nm)
0.0
0.2
0.4
0.6
0.8
5 10 15 20 25 30 35C
arbo
n (m
g)
0.15
0.25
0.35
0.45
5 10 15 20 25 30 35Crystallite size (nm)
Oxy
gen
dem
and
(mol
O /
kg) Oxygen demand
0.5
1.5
2.5
3.5
4.5
5 10 15 20 25 30 35
Crystallite size (nm)
CO
2fo
rmed
(%
v/v
)
Combustion
Coking
Crystallite size (nm)
0.0
0.2
0.4
0.6
0.8
5 10 15 20 25 30 35C
arbo
n (m
g)
Figure 2. Level of hydrocarbon coking (top), combustion (middle), and the
oxygen demand (bottom) of the plain ceria catalysts vs. crystallite size. Determined at the
selective hydrogen combustion reaction conditions (alternating feeds of 4:1:1% v/v
C3H8:C3H6:H2 in Ar (10 min), and 1% v/v O2 in Ar (18 min), at 550 ºC).
Chapter 3.4 Characterisation
209
Figure 2 (middle) shows that small ceria crystallites do not combust
hydrocarbons at all in the selective hydrogen combustion, whilst larger ones do. In
H2-TPR, however, this phenomena is reversed: the smaller ceria crystallites do
show the highest level of hydrogen combustion (vide infra). Clearly, the
combustion behaviour of the catalysts depends on the gas feed. When
hydrocarbons are present, the smaller crystallites preferably coke them, even
though there is plenty of oxygen available. Possibly, the high adsorption affinity of
the lower coordinated surface atoms results in a higher level of coking, as
compared to combustion.
Chromium-doped ceria: varying crystallite size at a constant doping
level. To investigate the effect of the crystallite size on the catalytic properties of a
doped catalyst, we have prepared a set of catalysts containing 5 mol% of chromium
dopant, with the ceria crystallite size ranging from 7 to 50 nm (Table 2). The
crystallites size was adjusted by varying the calcination temperature from 450 to
800 °C. The presented crystallite sizes are determined from the broadening of the
Ce(111) XRD peak.
Cha
pter
3.4
Cha
ract
eris
atio
n
210
T
able
2. P
hysi
cal a
nd c
atal
ytic
pro
pert
ies
of c
eria
cat
alys
ts d
oped
wit
h 5
mol
% C
r.
Cat
alys
t C
alci
natio
n
tem
pera
ture
(°C
)
Cer
ia c
ryst
alli
te
size
(nm
)[a]
Lat
tice
spa
cing
(Å)
Sel
ectiv
ity
(%)[b
]
Oxy
gen
dem
and
(mol
O /
kg)[c
]
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
7 45
0 7.
1 5.
4102
92
0.
62
5
8 55
0 11
.5
5.40
90
90
0.61
6
9 62
5 15
.8
5.40
99
95
0.55
6
10
700
31.0
5.
4105
89
0.
46
7
11
750
37.4
5.
4110
98
0.
40
9
12[d
] 80
0 49
.2
5.41
10
91
0.45
15
[a
] D
eriv
ed f
rom
the
peak
bro
aden
ing
of th
e C
e(11
1) X
RD
pea
k us
ing
the
Sch
erre
r eq
uatio
n. [b
] The
initi
al u
nsel
ectiv
e co
mbu
stio
n is
not
take
n
into
acc
ount
whe
n ca
lcul
atin
g th
e se
lect
ivity
. [c] T
his
is th
e am
ount
of
oxyg
en w
hich
is c
onsu
med
by
the
cata
lyst
in th
e re
oxid
atio
n st
ep. T
hat
is,
it i
s a
com
bina
tion
of
the
oxyg
en n
eede
d to
ref
ill
latt
ice
and
com
bust
the
cok
e pr
esen
t on
the
cat
alys
t su
rfac
e. [d
] Tra
ces
of C
r 2O
3 w
ere
obse
rved
by
XR
D, a
nd th
e co
lour
of
the
sam
ple
is s
ilver
inst
ead
of li
ght b
row
n.
Chapter 3.4 Characterisation
211
Table 2 shows that varying the calcination temperature is an effective tool
for adjusting the cerias crystallite size. Note, however, that at high calcination
temperatures the dopant atoms can segregate to the surface and form a separate
phase.[29] Indeed, traces of Cr2O3 are observed for catalyst 12, calcined at 800 °C.
To assess if the surface concentration chromium is related to the calcination
temperature, we analysed catalysts 8, 11 and 12 with X-ray Photoelectron
Spectroscopy (XPS, see Table 3). The data shows that the surface concentration Cr
does not increase with calcination temperature. The values of about 2–4 mol% are
slightly lower than the expected bulk value of 5 mol%.
Cha
pter
3.4
Cha
ract
eris
atio
n
212
T
able
3. S
urfa
ce c
once
ntra
tion
s of
Ce,
Cr
and
O a
s de
term
ined
by
XPS
.
Cat
alys
t
Cal
cina
tion
tem
pera
ture
(°C
)
Cer
ia c
ryst
alli
te
size
(nm
)[a]
Con
cent
ratio
n C
e
(mol
%)
Con
cent
ratio
n C
r
(mol
%)
Con
cent
ratio
n O
(mol
%)[b
]
Rat
io
Cr/
Ce
8 55
0 11
.5
29.7
1.
7 57
.0
0.06
11
750
37.4
27
.4
4.2
55.8
0.
15
12[c
] 80
0 49
.2
26.9
2.
1 58
.7
0.08
[a
] D
eriv
ed f
rom
the
pea
k br
oade
ning
of
the
Ce(
111)
XR
D p
eak
usin
g th
e S
cher
rer
equa
tion.
[b] T
he s
ampl
es a
lso
cont
ain
carb
on a
nd s
odiu
m,
ther
efor
e, th
e ad
ditio
n of
the
conc
entr
atio
ns o
f C
r, C
r an
d O
doe
s no
t yie
ld 1
00%
. [c] T
race
s of
Cr 2
O3
wer
e ob
serv
ed b
y X
RD
.
Chapter 3.4 Characterisation
213
0.5
1.0
1.5
2.0
5 15 25 35 45 55
Crystallite size (nm)
CO
2fo
rmed
(%
v/v
)
Combustion
Coking
0.0
0.1
0.2
0.3
0.4
0.5
5 15 25 35 45 55
Crystallite size (nm)
Car
bon
(mg)
0.5
1.0
1.5
2.0
5 15 25 35 45 55
Crystallite size (nm)
CO
2fo
rmed
(%
v/v
)
Combustion
Coking
0.0
0.1
0.2
0.3
0.4
0.5
5 15 25 35 45 55
Crystallite size (nm)
Car
bon
(mg)
Figure 3. Hydrocarbon coking (top) and combustion (bottom) obtained using the
chromium doped ceria catalysts vs. crystallite size. Determined at the selective hydrogen
combustion reaction conditions (alternating feeds of 4:1:1% v/v C3H8:C3H6:H2 in Ar (10
min), and 1% v/v O2 in Ar (18 min), at 550 ºC).
Figure 3 shows the amount of coking and combustion for the Cr-doped
catalysts in the selective hydrogen combustion. As with undoped ceria, the
catalysts with a larger crystallite size show a lower level of hydrocarbon coking
(Figure 3, top). Contrary to undoped ceria, however, combustion does occur at
small crystallite sizes (Figure 3, bottom). Apparently, oxygen from the added
chromium oxide is used for hydrocarbon combustion. Note that the combustion
Chapter 3.4 Characterisation
214
occurs mainly during the first 25 s of the 10 min reduction cycle (vide infra), and is
probably caused by an unselective reaction with adsorbed oxygen. Indeed, the
combustion levels are roughly the same for all catalysts, as is the doping level (5
mol% Cr).
Importantly, adding Cr increases both the hydrogen activity and the
selectivity. This is seen in Figure 4, which shows the conversions of hydrogen,
propene and propane during a reduction cycle of undoped ceria 2, and the Cr-doped
catalysts 8 and 11. All three catalysts combust some of the hydrocarbons at the
start of the reduction cycle (Figure 4, first data point). In the remainder of the
reduction cycle, however, the Cr-doped catalysts selectively combust the hydrogen.
Moreover, the undoped ceria 2 shows formation of hydrogen gas, due to coking
(the hydrogen conversion has a negative value, that is, hydrogen is formed).
Chapter 3.4 Characterisation
215
2, CeO2 12 nm
Con
vers
ion
(%)
-40
0
40
80
200 400 600
Time (s)
HydrogenPropenePropane
8, Cr-CeO2 12 nm
-40
0
40
80
200 400 600
Time (s)
Hydrogen
Propene
Propane
11, Cr-CeO2 38 nm
-40
0
40
80
200 400 600
Time (s)
Hydrogen
Propene
Propane
Con
vers
ion
(%)
Con
vers
ion
(%)
2, CeO2 12 nm
Con
vers
ion
(%)
-40
0
40
80
200 400 600
Time (s)
HydrogenPropenePropane
8, Cr-CeO2 12 nm
-40
0
40
80
200 400 600
Time (s)
Hydrogen
Propene
Propane
11, Cr-CeO2 38 nm
-40
0
40
80
200 400 600
Time (s)
Hydrogen
Propene
Propane
Con
vers
ion
(%)
Con
vers
ion
(%)
Figure 4. Time resolved conversion profiles showing the H2 (▲), C3H6 (◊) and
C3H8 (○) conversion during a reduction cycle. Catalysts: pure CeO2 (2, 12 nm crystallites,
top); Ce0.95Cr0.05O2 (8, 12 nm crystallites, middle); and Ce0.95Cr0.05O2 (11, 38 nm
crystallites, bottom). The negative conversions indicate the production of H2 via coking.
Reaction conditions: 550 °C, 10 min reduction cycles with a gas feed of 4:1:1% v/v
C3H8:C3H6:H2 in Ar.
Chapter 3.4 Characterisation
216
Figure 4 shows that the hydrocarbons are mainly converted at the start of
the reduction cycle. It follows that the coking also occurs here. Indeed, plotting the
hydrocarbon conversion, occurring at 25 s into the reduction cycle, against
crystallite size (Figure 5), yields a curve with the same trend as the amount of
coking against crystallite size shown in Figure 3. Note, for example, that both the
hydrocarbon conversion and the coking level remain constant for the three smallest
catalyst, and then drop. Interestingly, the hydrogen activity of the catalysts follows
the same trend, but in reverse: when the level of coking drops, the specific activity
increases (Figure 5), and vice versa. Because no oxygen is consumed in the coking
process, the lower specific activity must originate from shielding of the catalysts
surface by the coke. This is in agreement with the data shown in Figure 4, where
the hydrocarbon conversion (coking) occurs at the start of the reduction cycle, and
the selective hydrogen combustion in the remainder of it.
Our data show that the larger Cr-doped crystallites show less hydrocarbon
conversion and an increased specific activity. Therefore, in case of the 5 mol% Cr-
doping, the larger crystals are the best suited for the selective hydrogen
combustion.
Hyd
roge
n ac
tivity
(% H
2co
mbu
sted
)
Propane
PropeneActivity
0
25
50
75
100
5 15 25 35 45 55
Crystallite size (nm)
Con
vers
ion
at 2
5 s
(%)
0
4
8
12
16
Hyd
roge
n ac
tivity
(% H
2co
mbu
sted
)
Propane
PropeneActivity
0
25
50
75
100
5 15 25 35 45 55
Crystallite size (nm)
Con
vers
ion
at 2
5 s
(%)
0
4
8
12
16
Figure 5. Initial hydrocarbon conversions and hydrogen activity of the chromium
doped catalysts against crystallite size. Reaction conditions: 550 °C, 10 min reduction
cycles with a gas feed of 4:1:1% v/v C3H8:C3H6:H2 in Ar.
Chapter 3.4 Characterisation
217
The time scale of the catalysts hydrogen activity. Figure 4 shows that
the time in which the catalysts are active increases with increasing particle size.
Catalyst 11, consisting of larger crystallites, is active over a longer time period
compared to 8. Note that catalysts 10–12, with the largest crystallites, are active
during the entire 10 min reduction cycle. We therefore cannot asses if these
catalysts coke some of the hydrocarbons at the end of the run, as is the case for 7–
9, consisting of smaller crystallites. To check if the larger crystallites indeed show
less coking, we subjected catalyst 11 to a 20 min reduction cycle, that is, 10 min
longer than its active period. The data show no indication of ‘end of run’ coking (a
net hydrogen production). Indeed, some dehydrogenation occurs: a small amount
of propene and hydrogen are formed, together with some propane conversion (not
shown). Note that Cr is used as a commercial dehydrogenation catalyst.[7]
Furthermore, an equal amount of coke is observed for either a 10 or 20 min
reduction cycle. Clearly, the catalyst is stable in the reductive feed: when 11 is
subjected to the dehydrogenation gas feed for an extra 10 min after the selective
hydrogen combustion has stopped, no extra coke is formed. This robustness is
important for when the catalysts are applied in the actual dehydrogenation process.
Copper-doped ceria: the effect of dopant concentration on the physical
and catalytic properties. Table 4 shows the physical properties and catalytic data
of the set of copper doped ceria catalysts varying in concentration from 0.1 to
10 mol%. The data show that the crystallite size and the copper concentration are
not varied independently. The crystallite size decreases with increasing copper
concentration (see also Figure 6). The addition of copper does increase the
selectivity of the ceria, as was the case for Cr-doping (Table 4).
Chapter 3.4 Characterisation
218
10
15
20
25
0 2 4 6 8 10
Concentration Cu (mol%)
Cry
stal
lite
size
(nm
)10
15
20
25
0 2 4 6 8 10
Concentration Cu (mol%)
Cry
stal
lite
size
(nm
)
Figure 6. The relationship between ceria crystallite size and dopant concentration
for a set of copper doped ceria catalysts.
Cha
pter
3.4
Cha
ract
eris
atio
n
219
T
able
4. P
hysi
cal a
nd c
atal
ytic
pro
pert
ies
of C
u do
ped
ceri
a ca
taly
sts.
Cat
alys
t C
once
ntra
tion
Cu
(mol
%)
Cer
ia c
ryst
alli
te
size
(nm
)[a]
Lat
tice
spa
cing
(Å)
Sel
ectiv
ity
(%)[b
]
Oxy
gen
dem
and
(mol
O /
kg)[c
]
Hyd
roge
n ac
tivity
(% H
2 co
mbu
sted
)
13
0.1
23.4
5.
4096
0
0.33
0
14
1 19
.2
5.40
87
86
0.36
6
15
3 16
.2
5.41
05
94
0.50
8
16
7 16
.7
5.41
00
95
0.71
8
17
8 13
.7
5.41
14
90
0.84
6
18
10
14.9
5.
4102
89
0.
88
7 [a
] Der
ived
fro
m t
he p
eak
broa
deni
ng o
f th
e C
e(11
1) X
RD
pea
k us
ing
the
Sch
erre
r eq
uatio
n. [b
] The
ini
tial
unse
lect
ive
com
bust
ion
is n
ot
take
n in
to a
ccou
nt w
hen
calc
ulat
ing
the
sele
ctiv
ity.
[c] T
his
is t
he a
mou
nt o
f ox
ygen
whi
ch i
s co
nsum
ed b
y th
e ca
taly
st i
n th
e re
oxid
atio
n
step
. Tha
t is,
it is
a c
ombi
nati
on o
f th
e ox
ygen
nee
ded
to r
efil
l lat
tice
and
com
bust
the
coke
pre
sent
on
the
cata
lyst
sur
face
.
Chapter 3.4 Characterisation
220
Figure 7 shows the amount of coking and combustion for the copper doped
catalysts vs. crystallite size. The coking decreases for larger crystallites. The coking
level is indeed correlated to the crystallite size, and not with dopant type or
concentration: the same correlation is observed for undoped ceria, ceria with a
constant doping level (5 mol% Cr), and ceria with varying doping level (0.1−
10 mol% Cu).
Hydrocarbon combustion is related to the addition of a dopant. The small
Cu-doped crystallites combust part of the hydrocarbon feed, similar to the Cr-
doped ones, and contrary to the undoped ceria (compare Figure 2, middle, with
Figure 7, bottom). Moreover, in case of the Cr-doping, the dopant concentration
and hydrocarbon combustion were roughly equal. In case of the Cu-doping, the
amount of combustion increases with the dopant concentration. This is shown in
Figure 8, where the level of coking and combustion are plotted against copper
concentration, instead of crystallite size. Figure 8 (bottom), shows the high
correlation between the level of combustion and the copper concentration. Indeed,
things are complicated in case of the copper doping, since the dopant concentration
affects the crystallite size. However, the correlation between the hydrocarbon
combustion and dopant concentration is stronger than with the crystallite size
(compare Figures 7, bottom, and 8, bottom). Higher amounts of copper result in a
higher level of combustion. Note again that this combustion occurs mainly at the
initial part of the reduction cycle, and probably reflects unselective reaction with
adsorbed oxygen.
Chapter 3.4 Characterisation
221
0
2
4
6
10 15 20 25
Crystallite size (nm)
CO
2fo
rmed
(%
v/v
)
0.0
0.1
0.2
0.3
0.4
0.5
10 15 20 25
Crystallite size (nm)C
arbo
n (m
g)
Combustion
Coking
13
0
2
4
6
10 15 20 25
Crystallite size (nm)
CO
2fo
rmed
(%
v/v
)
0.0
0.1
0.2
0.3
0.4
0.5
10 15 20 25
Crystallite size (nm)C
arbo
n (m
g)
Combustion
Coking
13
Figure 7. Level of hydrocarbon coking (top) and combustion (bottom) of the
copper doped ceria catalysts in the selective hydrogen combustion at 550 ºC against
crystallite size. In case of the hydrocarbon combustion, catalyst 13 is an outlier. Its doping
level of 0.1 mol % is probably too low to affect the properties of the ceria. Indeed, its
crystallite size, selectivity, and hydrocarbon combustion are comparable to that of undoped
ceria.
Chapter 3.4 Characterisation
222
2
4
6
0 2 4 6 8 10
CO
2fo
rmed
(%
v/v
)
Cu concentration (mol %)
Combustion
Coking
0.0
0.1
0.2
0.3
0.4
0.5
0 2 4 6 8 10
Cu concentration (mol %)
Ca
rbo
n (
mg)
2
4
6
0 2 4 6 8 10
CO
2fo
rmed
(%
v/v
)
Cu concentration (mol %)
Combustion
Coking
0.0
0.1
0.2
0.3
0.4
0.5
0 2 4 6 8 10
Cu concentration (mol %)
Ca
rbo
n (
mg)
Figure 8. Level of hydrocarbon coking (top) and combustion (bottom) of the
copper doped ceria catalysts in the selective hydrogen combustion at 550 ºC against copper
concentration.
Activity of the copper-doped cerias. As with chromium, adding copper
increases selectivity (see Table 4 and Figure 9).[30] The hydrogen activity, however,
remains constant from 1 to 10 mol% Cu (see Figure 9). The increased level of
hydrocarbon combustion and coking at high copper loadings (Figure 8) counters
the beneficial effect of adding extra copper-oxide for the hydrogen activity. Note
that doping ceria with copper results in catalysts with a rather low activity. In case
of more active catalysts, the hydrogen activity does increase with increasing dopant
concentration.
Chapter 3.4 Characterisation
223
Analysis of the level of propane and propene conversion of the copper
doped catalysts shows that the propane conversion increases linearly with copper
concentration, whilst the propene conversion remains unaffected by the addition of
the copper (see Figure 10). Possibly, the propene is mainly converted via coking,
and the propane is mainly converted via combustion.
25
50
75
100
0 2 4 6 8 10 12
Cu concentration (mol%)
Sel
ectiv
ity (
%)
2
4
6
8
Hyd
roge
n ac
tivity
(% H
2co
mbu
sted
)
SelectivityActivity
25
50
75
100
0 2 4 6 8 10 12
Cu concentration (mol%)
Sel
ectiv
ity (
%)
2
4
6
8
Hyd
roge
n ac
tivity
(% H
2co
mbu
sted
)
SelectivityActivity
Figure 9. Selectivity and hydrogen activity of the copper doped cerias.
25
50
75
100
0 2 4 6 8 10
Cu concentration (mol%)
Con
vers
ion
at 2
5 s
(%)
Propane
Propene
25
50
75
100
0 2 4 6 8 10
Cu concentration (mol%)
Con
vers
ion
at 2
5 s
(%)
Propane
Propene
Figure 10. Hydrocarbon conversions during the selective hydrogen combustion of
the copper doped cerias. Data is taken at 25 s, since here the main hydrocarbon conversion
occurs.
Chapter 3.4 Characterisation
224
The effect of crystallite size and doping on catalyst sintering. Sintering
can lead to loss of activity via loss of surface area. We have assessed the effect of
doping and crystallite size on sintering by subjecting the catalysts to redox cycling
at high temperature (800 °C), using TPR and TEM to evaluate the amount of
sintering. Figure 11 (top) shows the TPR profiles of the undoped ceria catalysts 1,
6, 6a and 6b, which increase in crystallite size from 10 nm to 245 nm. The TPR
profiles contain the two typical ceria peaks at about 470 °C and 700 °C (peaks A
and B, respectively). These peaks are ascribed to the reduction of surface oxygen
and bulk oxygen, respectively.[31] They are also explained by the reduction of small
and large crystallites, since small particles are ‘mostly surface’, and large particles
are ‘mostly bulk’.[23] Since our reaction is performed at 550 °C, we will only
concern ourselves with the TPR features below 600 °C (peak A).
Chapter 3.4 Characterisation
225
400
800
1200
0 200 400 600 800
Temperature (°C)
TC
D s
igna
l (a
u)
1, 10 nm
6, 30 nm
6a, 188 nm6b, 245 nm
A B
100
300
500
700
0 200 400 600 800
Temperature (°C)
TC
D s
igna
l (a
u)
freshspent
400
800
1200
0 200 400 600 800
Temperature (°C)
TC
D s
igna
l (a
u)
1, 10 nm
6, 30 nm
6a, 188 nm6b, 245 nm
A B
100
300
500
700
0 200 400 600 800
Temperature (°C)
TC
D s
igna
l (a
u)
freshspent
Figure 11. TPR profiles of catalysts 1, 6, 6a and 6b (top), and of fresh and spent
catalyst 1 (bottom). The solid curve denotes fresh catalyst 1, the dashed curve spent catalyst
1 (after heating to 800 ºC in 66% hydrogen). In between the measurements, the catalyst
sample has been reoxidised at 300 ºC for 30 min in 5% O2/Ar.
Chapter 3.4 Characterisation
226
Table 5. Quantitative TPR data of catalysts 1, 6, 6a and 6b.
Catalyst Calcination
temperature (°C)
crystallite size
(nm)
Surface area
(m2/g)
Size of TPR peak A
(mol O / kg)[a]
1 550 9.8 84 0.14
6 700 29.9 22 0.09
6a[b] 700 188 0[c] 0.04
6b[d] 700 245 0[c] 0.03 [a] Data obtained by calibrating the TCD detector using a CuO standard. The peak area of
this standard is integrated and the area is correlated to the amount of oxygen present in the
CuO. [b] Same as catalyst 6, but treated for 4 h at 800 ºC under flowing hydrogen. [c] The
surface area of these catalysts is too low to be accurately determined. [d] A second batch of
catalyst 6, treated for 16 h at 800 ºC under flowing hydrogen.
Figure 11 shows that the smaller ceria crystallites have a lower reduction
onset, and a wide reduction range (broad peak A).[32] The broadness of peak A has
been ascribed to the a broad size distribution of smaller crystallites, and the
simultaneous reduction and sintering of the small crystallites.[24] Indeed, peak A of
catalyst 6 (30 nm), shows a broad reduction feature between 300 and 400 °C. After
this catalyst is sintered into large particles by high temperature reduction (6a, 4h at
800 °C), this broad feature disappears, and peak A has become more symmetrical
(compare 6 and 6a in Figure 11). The size of peak A decreases further upon
sintering catalyst 6 into even larger particles (6b, 6 h at 800 °C), but it does not
disappear completely (for quantitative data of this process see Table 5). Peak A
does disappear, however, when smaller crystallites are sintered by high temperature
reduction. This is shown in Figure 11 (bottom) for catalyst 1. Peak A completely
disappears after catalyst 1 has been subjected to 30 min at 800 °C in hydrogen,
where it is still present for catalyst 6 after 6h at 800 °C in hydrogen. That is,
starting from the smaller crystallites (1 – 3, 10 – 18 nm), sintering occurs sooner,
compared to starting from the larger crystallites (4 & 6, 22 nm and 30 nm).[33] This
was confirmed by TEM experiments (Figure 12). Fresh catalyst 1 sinters into much
larger crystals as compared to fresh catalyst 6 (starting from 10 and 30 nm crystals,
respectively). Note the different scaling of the images of the unsintered (A, B) and
sintered (C, D) catalysts.
Chapter 3.4 Characterisation
227
Importantly, we have assessed the sintering at conditions which are severe
compared to the catalytic selective hydrogen combustion–tests, namely heating to
800 °C in 66% H2/Ar for the sintering, as compared to 10 minute reduction cycles
at 550 °C in 4:1:1 vol% of C3H8:C3H6:H2 for the catalytic tests. Indeed, catalyst 1
shows no indication of sintering during the catalytic tests (the activity, amount of
coking and conversions of the feed during the reduction cycles are stable).
Figure 12. TEM images of catalyst 1 and 6 fresh (left hand side) and sintered (1a
and 6a, right hand side). The spent catalyst 1 (1a) was subjected to milder sintering
conditions compared to spent catalyst 6a (0.5 h instead of 4 h at 800 °C in 66% H2/Ar).
Still, it sinters into larger crystals (right). Note the different scaling of images A and B
compared to C and D.
Chapter 3.4 Characterisation
228
Effect of doping on the sintering behaviour. It is known that ceria
zirconia mixed oxides are more sinter stable than plain ceria.[34] Our data show that
doping with Cr or Cu also increases the sinter stability of the ceria (see Figure 13).
In case of the ‘small’ undoped ceria crystallites (10–18 nm), peak A completely
disappears upon sintering (Figure 13, top). This is not the case for similarly sized
Cr and Cu-doped catalysts (Figure 13, middle and bottom). Note that peak A of the
copper doped catalyst 18 is far less reduced in size as compared to the chromium
doped catalyst 8 (quantitative data given in Table 6). The latter also showed a grey
band at the reactor exit after the sintering experiment, indicating that part of the
chromium has evaporated. Note again that the sintering conditions are severe
compared to the selective hydrogen combustion experiments. Indeed, we
previously showed that a Cr and Zr doped catalyst (Ce0.90Cr0.05Zr0.05O2) was highly
selective and active over 250 selective hydrogen combustion-redox cycles (a total
of 148 hours on stream), with no phase segregation or change in particle size (see
Chapter 2.1).[21]
Chapter 3.4 Characterisation
229
CeO2 (1), 10 nm
0
200
400
600
800
0 200 400 600 800
Temperature (°C)T
CD
sig
nal (
au)
Ce0.95Cr0.05O2 (8), 12 nm
0
200
400
600
800
0 200 400 600 800Temperature (°C)
TC
D s
igna
l (au
)
0
200
400
600
800
0 200 400 600 800Temperature (°C)
TC
D s
igna
l (au
)
Ce0.90Cu0.10O2 (18), 14 nm
fresh
spentCeO2 (1), 10 nm
0
200
400
600
800
0 200 400 600 800
Temperature (°C)T
CD
sig
nal (
au)
Ce0.95Cr0.05O2 (8), 12 nm
0
200
400
600
800
0 200 400 600 800Temperature (°C)
TC
D s
igna
l (au
)
0
200
400
600
800
0 200 400 600 800Temperature (°C)
TC
D s
igna
l (au
)
Ce0.90Cu0.10O2 (18), 14 nm
0
200
400
600
800
0 200 400 600 800Temperature (°C)
TC
D s
igna
l (au
)
Ce0.90Cu0.10O2 (18), 14 nm
fresh
spent
Figure 13. TPR of fresh (solid curve) and spent (dashed curve) catalysts 1, 8 and
18. Spent: after heating to 800 ºC in 66% hydrogen. Note that the doping level and
calcination temperature of these catalysts differ. The trends are the same, however, for
equal doping levels and calcination temperatures.
Chapter 3.4 Characterisation
230
Table 6. Quantitative TPR data of fresh and spent catalysts 1, 8 and 18.
Size of TPR peak C
(mol O / kg)[a, b] Catalyst Composition
First TPR
(‘fresh’)
Second TPR
(‘spent’)
1 CeO2 0.14 0.00
8 Ce0.95Cr0.05O2 0.77 0.12
18 Ce0.90Cu0.10O2 0.75 0.55 [a] Data obtained by calibrating the TCD detector using a CuO standard. The peak area of
this standard is integrated and the area is correlated to the amount of oxygen present in the
CuO. [b] Peak A in case of catalyst 1.
Chapter 3.4 Characterisation
231
Conclusions The activity, selectivity and stability of ceria in the selective hydrogen combustion
from a mixture with propane and propene are increased by either chromium or
copper doping. The doping often alters the crystallite size as well. In case of
undoped ceria, small crystallites (< 20 nm) mainly coke the hydrocarbons, and
larger ones (>20 nm) combust them. This results in an optimal crystallite size of
about 20 nm, where the least of the unwanted hydrocarbon coking and combustion
occurs. This effect should be taken into account when using ceria based catalyst in
reactions involving hydrocarbons. Interestingly, the smaller ceria crystallites show
the highest level of hydrogen combustion in H2-TPR. In the selective hydrogen
combustion however, they do not combust hydrocarbons at all, instead, they show
high levels of hydrocarbon coking.
Doping with Cr or Cu increases both the selectivity and activity of the ceria.
Contrary to undoped ceria, however, the small doped crystallites do combust some
of the hydrocarbons. Increasing the copper concentration from 0.1 mol% to
10 mol% does not affect the propene conversion, but the propane combustion
increases linearly. Since this also uses up oxygen, the hydrogen activity (amount of
hydrogen combusted), does not increase with increasing amount of dopant. The
sintering stability of the ceria, however, is increased by the copper doping. The
combustion does not vary with crystallite size in case of the Cr-CeO2 catalysts, all
doped with 5 mol% Cr. As coking levels do drop with increasing crystallite size,
best results are obtained with the larger crystallites (up to 50 nm), showing the least
amount of coking, and the highest activity. Importantly, the Cr-doped catalysts are
stable in the reductive gas feed. When the reduction time is doubled, no extra coke
is formed. This robustness is important for when the catalysts are applied in the
actual dehydrogenation process.
Chapter 3.4 Characterisation
232
Experimental Section Materials and instrumentation. Chemicals were purchased from Sigma-
Aldrich or Merck and used as received. Gasses were purchased from Praxair and
had a purity of 99.5% or higher. The O2, He, Ar and N2 streams were purified
further over molsieves and/or BTS columns. Powder X-ray diffraction
measurements were performed using a Philips PW-series X-ray diffractometer with
a Cu tube radiation source (λ = 1.54 Å), a vertical axis goniometer and a
proportional detector. The 2θ detection measurement range was 10 ° – 93 ° with a
0.02 ° step size and a 5 second dwell time. Lattice constants and crystallite sizes
were obtained after Rietveld refinement (structure fit) using PANalytical's X'pert
software package. GC analysis was performed on an Interscience Compact GC
equipped with TCD detectors, separating water, CO2 and C2 and C3 hydrocarbons
on a Porabond Q column (He carrier gas) and H2, CO, CH4, O2 and N2 on a 5 Å
molsieve column (Ar carrier gas). MS analysis was performed using a Pfeiffer
QMS 200 mass spectrometer (m/z range 0–200). XPS measurements were
performed on a KRATOS AXIS Ultra DLD spectrometer. TEM was performed on
an JEOL JEM2100 Transmission electron microscope.
Procedure for catalyst synthesis. The procedure for catalyst preparation
was described in detail previously.[21, 26] The metal nitrates are weighed in a
porcelain crucible and heated to about 100 °C, so that the cerium nitrate melts. The
mixture is stirred until all components are dissolved or have melted. The crucible is
placed in a vacuum oven set at 140 °C and the pressure is carefully lowered to < 10
mbar (in 10–15 min), making sure no vigorous boiling occurs. After 4 h, the
samples are placed in a furnace and calcined under static air at 700 °C (ramp rate
300 °C/h, 5 h hold). The resulting solid is pulverized, ground and sieved in
fractions of 125–212 µm (selectivity assessment) and < 125 µm (XPS, TPR, TEM
and XRD measurements).
Procedure for testing catalytic activity. The activity and selectivity were
determined using an automated cyclic redox reactor system built in-house,
described in detail elsewhere.[26] In a typical experiment, about 250 mg of sample
(125–212 μm) was placed on a quartz wool plug in a 4 mm id quartz reactor. The
reactor was placed in a water cooled oven and heated to 550 °C at 1200 °C/h, under
oxygen flow. At this temperature, redox cycling was started. The selectivity was
Chapter 3.4 Characterisation
233
determined by GC during the 10 min reduction in 4:1:1% v/v C3H8:C3H6:H2 in Ar
(total flow 50 mL/min), with 5 mL/min of N2 added as internal standard. The gas
hourly space velocity (GHSV) is 13200 / h (at the typical bed volume of 0.25 cm3
and the reduction cycle's total flow of 55 mL/min). The weight hourly space
velocity (WHSV) is 1.2 / h, and is calculated from the weight of C3H8 + C3H6 + H2
per h per the weight of the catalyst. The 4:1:1 ratio of reductive gases is chosen
since this is the equilibrium mixture of a conventional dehydrogenation catalyst.[11]
After a 4 min purge step (pure Ar), the sample was reoxidised for 18 minutes in
1% v/v O2 in Ar (50 mL/min total flow). The redox cycle is completed by another
purge step in pure Ar. The selectivity is determined as the ratio H2 conversion:total
conversion. Activity is determined as the percentage hydrogen combusted during
the reduction step (hydrogen activity), and the amount of oxygen needed in the
oxidation step (oxygen demand). Both selectivity and activity are averaged over
eight redox cycles. The amount of coking was assessed by the amount of CO and
CO2 formed in the reoxidation step, determined by MS. All signals were
normalised using a small amount of helium added to the gas feed (1 vol %), and
were integrated using CasaXPS v2.1.18 software.
Procedure for TPR experiments. In a typical measurement, 100 mg of
sample is placed on a quartz wool plug in a 4 mm i.d. quartz reactor. The sample is
calcined in situ to 300 °C (ramp rate 10 °C/min, 30 min hold time) in 5% v/v
oxygen in argon (50 mL/min total flow). After cooling to room temperature and
purging with pure argon, the system is allowed to equilibrate in 67 % hydrogen in
argon (20 mL/min total flow) for about 1 h. For the actual TPR measurement, the
sample is heated with a 5 °C/min heating rate to 800 °C (no hold). When the final
temperature is reached, the sample is allowed to cool to room temperature. When
subsequent measurements are performed, the sample is reoxidised in 5% v/v
oxygen in argon (300 °C, 30 min hold time).
Chapter 3.4 Characterisation
234
Acknowledgements We thank T. Franssen-Verheijen of Wageningen University for the TEM
measurements, L. Massin of IRCELYON-CNRS (Lyon, France) for performing the
XPS measurements, Dr. M.C. Mittelmeijer–Hazeleger for the BET surface area
measurements, A.C. Moleman and W.F. Moolhuijzen for help with the XRD
measurements, and NWO–ASPECT for financial support and feedback.
Chapter 3.4 Characterisation
235
References [1] E. Burridge, Chem. Bus., 2006, 1, 45. [2] J. Plotkin and E. Glatzer, Eur. Chem. News, 2005, 82, 20. [3] N. Alperowicz, Chem. Week, 2006, 168, 17. [4] N. Alperowicz, Chem. Week, 2007, 169, 27. [5] G. Parkinson, Chem. Eng. Prog., 2004, 100, 8. [6] A. L. Waddans, Chemicals from Petroleum, 4th edn., John Murray Ltd., London,
1978. [7] T. A. Nijhuis, S. J. Tinnemans, T. Visser and B. M. Weckhuysen, Chem. Eng. Sci.,
2004, 59, 5487. [8] M. P. Lobera, C. Téllez, J. Herguido and M. Menéndez, Appl. Catal. A: Gen.,
2008, 349, 156. [9] M. M. Bhasin, J. H. McCain, B. V. Vora, T. Imai and P. R. Pujado, Appl. Catal. A:
Gen., 2001, 221, 397. [10] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Abstr. Pap. Am. Chem. S.,
1999, 217, U687. [11] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 1. [12] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 9. [13] J. G. Tsikoyiannis, D. L. Stern and R. K. Grasselli, J. Catal., 1999, 184, 77. [14] C. H. Lin, K. C. Lee and B. Z. Wan, Appl. Catal. A: Gen., 1997, 164, 59. [15] L. Låte, J. I. Rundereim and E. A. Blekkan, Appl. Catal. A: Gen., 2004, 262, 53. [16] L. Låte, W. Thelin and E. A. Blekkan, Appl. Catal. A: Gen., 2004, 262, 63. [17] J. Beckers, R. Drost, I. van Zandvoort, P. F. Collignon and G. Rothenberg,
ChemPhysChem, 2008, 9, 1062. [18] L. M. van der Zande, E. A. de Graaf and G. Rothenberg, Adv. Synth. Catal., 2002,
344, 884. [19] A. Trovarelli, C. de Leitenburg, M. Boaro and G. Dolcetti, Catal. Today, 1999, 50,
353. [20] G. Rothenberg, E. A. de Graaf and A. Bliek, Angew. Chem., Int. Ed., 2003, 42,
3366. [21] J. Beckers, F. Clerc, J. H. Blank and G. Rothenberg, Adv. Synth. Catal., 2008, 350,
2237. [22] J. Beckers and G. Rothenberg, Dalton Trans., 2008, 6573. [23] F. Giordano, A. Trovarelli, C. de Leitenburg and M. Giona, J. Catal., 2000, 193,
273. [24] E. Aneggi, M. Boaro, C. de Leitenburg, G. Dolcetti and A. Trovarelli, J. Alloy
Compd., 2006, 408, 1096.
Chapter 3.4 Characterisation
236
[25] J. H. Hwang and T. O. Mason, Z. Phys. Chem., 1998, 207, 21. [26] J. H. Blank, J. Beckers, P. F. Collignon, F. Clerc and G. Rothenberg, Chem. Eur.
J., 2007, 13, 5121. [27] J. H. Blank, J. Beckers, P. F. Collignon and G. Rothenberg, ChemPhysChem,
2007, 8, 2490. [28] N. N. Greenwood and A. Earnshaw, Chemistry of the Elements, Pergamon Press,
Oxford, 1989. [29] X. C. Zheng, X. L. Zhang, X. Y. Wang, S. R. Wang and S. H. Wu, Appl. Catal. A:
Gen., 2005, 295, 142. [30] Catalyst 13, containing 0.1 mol% of copper, is and exception. The doping level is
too low. It is not selective, and has a conversion profile similar to that of undoped ceria (2 in Figure 4).
[31] H. C. Yao and Y. F. Yao, J. Catal., 1984, 86, 254. [32] Note that the TPR profiles of catalysts 2, 3, and 4 have the same shape as that of 1,
and 5 has a similar shape as 6. [33] W. Huang, P. Shuk and M. Greenblatt, Chem. Mater., 1997, 9, 2240. [34] R. T. Baker, S. Bernal, G. Blanco, A. M. Cordón, J. M. Pintado, J. M. Rodríguez-
Izquierdo, F. Fally and V. Perrichon, Chem. Commun., 1999, 149.
Summary
237
Summary
The first chapter gives a general background of ceria-based materials and
their use in selective oxidations. It is divided in three main parts: an introduction
about ceria based materials as catalysts; some specific properties of these relevant
to catalysis; and the use of ceria based catalysts in various selective oxidations. The
focus of this thesis is on the application of doped cerias as solid ‘oxygen reservoirs’
(SORs), for selective hydrogen combustion. This reaction is part of a novel redox
process for propane oxidative dehydrogenation. The lattice oxygen of the SOR
selectively burns hydrogen from the dehydrogenation mixture at 550 °C. This gives
three key advantages: it shifts the dehydrogenation equilibrium to the desired
products side, it can generate heat in situ, which aids the endothermic
dehydrogenation, and simplifies product separation.
The second chapter deals with the discovery of selective,
active and stable SORs. In section 2.1, we apply a genetic algorithm (GA) to
screen doped cerias for their performance in the selective hydrogen combustion.
GAs mimic evolutionary biology in silico. Several generations of catalysts are
synthesised and tested, and the new generations are selected based on the
performance of the previous ones. We used 26 different dopant metals to
synthesise and test five generations of 97 catalysts in total. We identified six
dopant atoms which lead to selective hydrogen combustion catalysts, namely Bi,
Cr, Cu, K, Mn, Pb and Sn (‘good’ dopants). The other dopants either result in
unselective catalysts (e.g. Ru, Pd, Pt) or inactive catalysts (e.g. Yb, Nd, Sr). There
is little synergy, and the behaviour of bi-doped catalysts can be predicted from the
behaviour of the singly doped ones. Analysis of the effect of electronegativity,
ionic radius and dopant concentration shows that most elements yielding a high
fitness have electro negativities ranging from 1.5–1.9. The average fitness (a
measure of activity and selectivity) increases up to generation 3, and then
stabilises. Importantly, the doped cerias show a high stability in the redox cycling,
much higher than that of supported oxides. A Cr and Zr doped catalyst
(Ce0.90Cr0.05Zr0.05O2) was highly selective and active over 250 redox cycles (a total
of 148 hours on stream), with no phase segregation or change in particle size.
Summary
238
In section 2.2, the possibilities of applying perovskite type oxides as SORs
are explored. The screening of fourteen perovskites shows that the catalytic
properties depend strongly on the composition. Changing the B atom in a series of
LaBO3 perovskites shows that Mn and Co give a higher selectivity than Fe and Cr.
Moreover, replacing part of the La-atoms with Sr or Ca also affects the catalytic
properties. The best results are achieved with Sr-doped LaMnO3. La0.9Sr0.1MnO3 is
active and selective, and shows excellent stability, even after 125 redox cycles at
550 °C (70 h on stream). Notably, the activity per unit surface area of the
perovskite catalysts is higher than that of doped cerias, the current benchmark of
solid oxygen reservoirs.
Section 2.3 focuses on lead containing SORs. Lead oxide supported on
alumina or silica, or lead doped ceria yields highly active and selective SORs, but
the lead(oxide) is not stable under the redox cycling. Good results are obtained
with lead chromate. This catalyst is more active and selective than any other SOR
tested. Prolonged testing (125 redox cycles at 550 °C) shows a drop in activity of
25 percent of the initial value after 40 cycles. The resulting activity is still higher
than that of other SORs, and the tests were carried out on ‘as received’ PbCrO4, of
which the stability can possibly be increased.
The third chapter deals with the characterisation of the SORs. In section
3.1, the redox properties of six monodoped cerias are investigated using TPR and
TGA. We show that the doped cerias generally release more oxygen compared to
plain ceria. Secondly, the temperature where the oxygen is released is generally
lower for the doped cerias as well, and varies from 110 °C (Cu-CeO2) to 550 °C
(Ca-CeO2, determined by H2−TPR). This enables catalytic applications over a wide
temperature range. The H2-reduction rate at 550 °C is correlated to the reduction
onset of the catalyst. Catalysts with a relatively low reduction temperature, such as
Cu-, Mn-, Bi- and Pb-CeO2, show a high reduction rate at 550 °C. Conversely,
catalysts with a high reduction temperature, such as Fe-CeO2 and plain ceria,
reduce slower.
Section 3.2 explores the redox properties of doped and supported copper-
ceria catalysts. Using TPR and XRD, we show that reduction occurs at ~110 °C,
~150 °C, or ~190 °C, depending on the catalyst type. The reduction at 110 °C is
ascribed to highly dispersed copper species doped in the ceria lattice (doped ceria),
Summary
239
and that at 190 °C to CuO crystallites supported on ceria. Remarkably, both types
converge to the 150 °C feature after redox cycling. The reduction temperature of
the doped catalyst increases after redox cycling, indicating that stable Cu clusters
form at the surface. Conversely, the reduction temperature of the ‘supported’
catalyst decreases after redox cycling, and the CuO crystallites disappear. With this
knowledge, a copper–doped ceria catalyst is analysed after application in selective
hydrogen combustion (16 consecutive redox cycles at 550 °C). No CuO crystallites
are observed, and the sample reduces at ~110 °C. This suggests that copper-doped
ceria is the active oxygen exchange phase in selective hydrogen combustion.
Furthermore, calorimetric measurements show that the hydrogen combustion by
doped cerias can indeed be a net exothermic process.
In section 3.3, we show four ways to increase the selectivity of bismuth
doped ceria. Bismuth is a promising dopant, but its selectivity can be improved.
We found that this can be achieved by increasing the hydrogen content of the feed,
by co-doping with Pt, resulting in a Pt-Bi alloy which is more selective than the
separate Pt or Bi, by co-doping with Sn (which prevents coking), or by adjusting
the reaction temperature (optimal performance at 400 °C). We rationalise this
optimal reaction temperature from hydrogen and propene TPR: 400 °C lies above
the reduction maximum of hydrogen, yet below that of propene. That is, this
temperature is sufficiently high to facilitate rapid hydrogen combustion, but low
enough to prevent hydrocarbon conversion. Indeed, in case of the unselective plain
ceria, the reduction maximum of propene lies below that of hydrogen, and for the
selective Bi-doped or Pb-doped ceria, the reduction maximum of propene lies
above that of hydrogen.
In section 3.4, we investigate the relationship between the catalytic
properties of the ceria based SORs and the crystallite size and dopant
concentration. We show that the level of hydrocarbon coking is related to the
crystallite size (smaller crystallites coke more), and that the level of both
hydrocarbon and hydrogen combustion are increased upon dopant addition. Doping
with Cr or Cu increases the selectivity, activity and stability of the ceria. The
propane combustion, however, also increases linearly with the Cu-concentration.
The best results are obtained with Cr-doped ceria, with selectivities up to 98%, and
combusting up to 15% of the hydrogen feed. The larger Cr-doped crystallites (up to
Summary
240
50 nm), show the least amount of coking, and the highest activity. Importantly, the
Cr-doped catalysts are stable in the reductive gas feed. No extra coke is formed
when the catalyst is subjected to an extra 10 min in the dehydrogenation mixture,
after the hydrogen combustion reaction has stopped.
Finally, a list of all the doped ceria catalyst which were made is given in
Appendix I, showing the relationship between the dopant type and its
concentration and the success of doping. In Appendix II, the activities of the SORs
are expressed in various units, and grouped by the type of method used to
determine them. The relationship between the SORs' activity and the amount of
SOR needed in the proposed combined dehydrogenation and selective hydrogen
combustion process is also given here.
Samenvatting
241
Samenvatting
Het eerste hoofdstuk is een algemene inleiding op de toepassing van ceria
en op ceria gebaseerde materialen als katalysatoren voor selectieve oxidaties. Het
hoofdstuk is in drieën opgedeeld: ten eerste een inleiding op ceria en op ceria
gebaseerde materialen als katalysator; dan volgt een bespreking van enkele
specifieke eigenschappen van ceria en op ceria gebaseerde materialen die van
belang zijn voor de katalyse; ten slotte het gebruik van ceria en op ceria gebaseerde
materialen als katalysatoren voor selectieve oxidaties. De kern van dit proefschrift
is het gebruik van gedoopte ceria's als vaste zuurstofreservoirs (Eng. solid oxygen
reservoir, SOR) voor de selectieve oxidatie van waterstof. Dopen betekent hier
‘toevoegen aan’ of ‘vervangen’: een deel van de ceriumatomen in het kristalrooster
wordt vervangen door die van een ander element. De selectieve waterstofoxidatie is
onderdeel van een nieuw oxidatief propaan-dehydrogenatieproces. Hierbij wordt de
roosterzuurstof van de SOR gebruikt om het waterstof selectief weg te oxideren uit
het reactiemengsel. Dit gebeurt bij een temperatuur van 550 °C. Dit proces heeft
drie belangrijke voordelen: het verschuift het dehydrogenatie-evenwicht naar de
productkant, het kan ter plaatse hitte genereren - wat voordelig is in verband met de
endotherme dehydrogenatiereactie - en het vereenvoudigd de productscheiding.
Het tweede hoofdstuk gaat over de zoektocht naar actieve, selectieve en
stabiele SOR's. Deel 2.1 beschrijft hoe een genetisch algoritme (GA) wordt
gebruikt om gedoopte ceria's met de gewenste katalytische eigenschappen te
vinden. Dit type algoritmes bootst de evolutionaire biologie na. Een generatie
katalysatoren wordt gemaakt en getest, en de volgende generatie wordt
geselecteerd aan de hand van de prestaties van de voorgaande generaties. Selectie
vindt plaats op basis van de zogenaamde fitness (geschiktheid) van de
katalysatoren, die bepaald wordt door de activiteit en de selectiviteit. We hebben
zesentwintig verschillende elementen gebruikt om vijf generaties gedoopte
ceriakatalysatoren te maken, met als resultaat in totaal zevenennegentig
katalysatoren. We hebben zes elementen gevonden waarmee selectieve
katalysatoren gemaakt kunnen worden, namelijk: bismut, chroom, koper, kalium,
mangaan, lood en tin (de zogenaamde ‘goede’ elementen). Het dopen met andere
metalen levert inactieve katalysatoren op (bijvoorbeeld bij dopen met ytterbium,
Samenvatting
242
neodymium of strontium) of niet selectieve katalysatoren (bijvoorbeeld bij dopen
met ruthenium, palladium of platina, de zogenaamde ‘slechte’ elementen). Er is
weinig synergie en het gedrag van katalysatoren gedoopt met twee elementen kan
afgeleid worden uit het gedrag van katalysatoren gedoopt met de afzonderlijke
elementen. De analyse van de elektronegativiteit, de ionradius en de concentratie
toont aan dat de elektronegativiteit van de meeste ‘goede’ elementen tussen de 1,5
en 1,9 ligt. De gemiddelde fitness van de generaties neemt toe tot en met de derde
generatie en blijft dan gelijk. Een belangrijke eigenschap van de gedoopte ceria's is
dat ze veel stabieler zijn dan gedragen metaaloxides. Een met chroom en zirkonium
gedoopte katalysator (Ce0.90Cr0.05Zr0.05O2) vertoonde een hoge selectiviteit en
activiteit gedurende tweehonderdvijftig cycli met een totale tijd van 148 uur, bij
550 °C. Er vond geen fasescheiding plaats tijdens de reactie en de deeltjesgrootte
bleef gelijk.
In deel 2.2 wordt onderzocht of perovskieten gebruikt kunnen worden als
SOR's. Uit de veertien geteste perovskieten blijkt dat de katalytische
eigenschappen in grote mate van de samenstelling van de katalysator afhangen. Het
veranderen van het B atoom in LaBO3 toont dat mangaan en kobalt een hogere
selectiviteit geven dan ijzer en chroom. Daarnaast kunnen de katalytische
eigenschappen ook beïnvloed worden door een deel van de La-atomen te
vervangen door strontium of calcium. De beste resultaten worden verkregen met
strontium-gedoopt LaMnO3. La0.9Sr0.1MnO3 is actief, selectief en stabiel, zelfs na
125 redox cycli bij 550 °C (totale reactietijd 70 uur). Een belangrijke ontdekking is
dat de perovskieten een hogere activiteit per vierkante meter hebben dan de
gedoopte ceria's.
Deel 2.3 gaat over met lood gedoopte ceria's. Loodoxide op alumina of
silica en met lood gedoopt ceria zijn zeer actief en selectief, maar het lood(oxide) is
niet stabiel gedurende de reactie. Met loodchromaat werden wel goede resultaten
verkregen. Dit materiaal is het meest actieve en selectieve dat we tot nu toe getest
hebben. Na 125 cycli zakt de activiteit weliswaar naar 25% van de beginwaarde,
maar deze activiteit is nog steeds hoger dan die van de gedoopte ceria's en de
perovskieten. Bovendien is voor de metingen puur loodchromaat gebruikt. De
stabiliteit hiervan kan mogelijk verder verhoogd worden.
Samenvatting
243
Het derde hoofdstuk gaat over de karakterisering van de katalysatoren. In
deel 3.1 worden de reductie- en oxidatie-eigenschappen (Eng. ‘redox’) van zes
gedoopte ceria's onderzocht met behulp van temperatuurgeprogrammeerde reductie
(TPR) en thermogravimetrische analyse (TGA). De gedoopte ceria's geven over het
algemeen meer zuurstof af dan pure ceria. Verder is de temperatuur waarbij
zuurstofafgifte plaatsvindt meestal ook lager, variërend van 110 °C (Cu-CeO2) tot
550 °C (Ca-CeO2, bepaald met H2−TPR). Dit betekend dat de katalysatoren over
een groot temperatuurgebied gebruikt kunnen worden. De reductiesnelheid bij
550 °C is gerelateerd aan de TPR-reductietemperatuur. Katalysatoren met een
relatief lage TPR-reductietemperatuur, zoals met koper, mangaan, bismut of met
lood gedoopt ceria, vertonen een hoge reductiesnelheid bij 550 °C. Omgekeerd
hebben katalysatoren met een hoge TPR-reductietemperatuur, zoals Fe-CeO2 en
pure ceria, een lage reductiesnelheid bij 550 °C.
In deel 3.2 worden de reductie- en oxydatie-eigenschappen van met koper
gedoopt ceria en koperoxide op ceria onderzocht. Met behulp van TPR en
Röntgendiffractie (XRD) tonen we aan dat de reductie plaatsvindt bij ~110 °C,
~150 °C, of ~190 °C, afhankelijk van het type katalysator. We schrijven de
reductie bij ~110 °C toe aan in grote mate gedispergeerd koper, gedoopt in het
ceria, en die bij ~190 °C aan koperoxide op ceria. Het is opvallend dat beide types
na een redoxcyclus een reductietemperatuur van 150 °C hebben. De
reductietemperatuur van het gedoopte ceria neemt toe na de redoxcyclus,
waarschijnlijk door de vorming van koperclusters aan het oppervlak. Omgekeerd
neemt de reductietemperatuur van het koper op ceria af doordat de
koperoxidedeeltjes zich over het ceria-oppervlak verspreiden. Met deze kennis is
een met koper gedoopte katalysator geanalyseerd na 16 cycli in de selectieve
waterstofoxidatiereactie (bij 550 °C). Na de reactie werd geen koperoxide
aangetroffen en de reductietemperatuur was ongeveer 110 °C. Dit toont aan dat het
met koper gedoopt ceria de actieve fase is voor de selectieve waterstofoxidatie.
Calorimetrische metingen tonen bovendien aan dat de waterstofoxidatie door
gedoopte ceria's inderdaad exotherm kan zijn.
In deel 3.3 beschrijven we vier manieren om met bismut gedoopte
ceriakatalysatoren selectiever te maken. Het met bismut gedoopte ceria is een
veelbelovende katalysator, maar de selectiviteit ervan zou verbeterd moeten
Samenvatting
244
worden. We hebben ontdekt dat de selectiviteit hoger wordt door verhoging van de
waterstofconcentratie, toevoeging van platina (waardoor een platina-bismut
legering wordt gevormd die selectiever is dan de aparte metalen), toevoeging van
tin (dit verhindert het zogenaamde coken, dat wil zeggen de vorming van vast
koolstof op het katalysatoroppervlak) en door aanpassing van de
reactietemperatuur. De beste resultaten worden verkregen bij een
reactietemperatuur van 400 °C. Met behulp van waterstof- en propeen-TPR kunnen
we beredeneren waarom dit zo is: 400 °C is hoger dan het reductiemaximum van
waterstof maar lager dan het reductiemaximum van propeen. Anders gezegd, deze
temperatuur is hoog genoeg om de waterstof te verbranden maar te laag voor een
reactie met de koolwaterstoffen. Bij het niet selectieve pure ceria zijn de
reductiemaxima omgekeerd: het reductiemaximum van waterstof ligt hier boven
dat van propeen.
In deel 3.4 onderzoeken we de relatie tussen de katalytische eigenschappen
van gedoopte ceria's en de kristalgrootte, en de hoeveelheid toegevoegd element.
We tonen aan dat de mate van het coken van de koolwaterstoffen gerelateerd is aan
de kristalgrootte (kleine kristallen coken meer) en dat de mate van
koolwaterstofoxidatie toeneemt door het dopen. Het toevoegen van chroom of
koper verhoogt zowel de selectiviteit als de activiteit en de stabiliteit van het ceria.
De propaanoxidatie neemt echter lineair toe met de hoeveelheid koper. Het beste
resultaat verkregen we met de met chroom-gedoopte ceria's, die een maximale
selectiviteit van 98% gaven, en een hoeveelheid waterstofoxidatie van maximaal
15% van de toegevoerde waterstof. De grote met chroom gedoopte ceriakristallen
(tot aan 50 nm groot) vertonen de minste coking en de hoogste activiteit. Ook zijn
deze katalysatoren stabiel in het reducerende gas: zelfs als de katalysator, nadat de
waterstofoxidatie gestopt, is gedurende tien minuten in de dehydrogenatie-
gasstroom blijft, wordt geen extra coke gevormd.
Bijlage I bevat een tabel met alle katalysatoren die voor dit onderzoek
gemaakt zijn en laat de relatie zien tussen het succes van de synthese en het type en
de concentratie van het element waarmee gedoopt wordt. Bijlage II bevat een
overzicht van de activiteit van de SOR's, gegroepeerd naar analysemethode. Hierbij
wordt ook aangegeven wat de relatie is tussen de hoeveelheid SOR die nodig is in
het voorgestelde oxidatieve dehydrogenatieproces en de activiteit van de SOR.
List of publications
245
List of publications
[21] 'Lead-containing solid oxygen reservoirs for selective hydrogen
combustion', Jurriaan Beckers and Gadi Rothenberg, Green Chem. 2009, DOI:
10.1039/b913994j *
[20] 'Ce0.95Cr0.05O2 and Ce0.97Cu0.03O2: Active, selective and stable catalysts for
selective hydrogen combustion', Jurriaan Beckers and Gadi Rothenberg, Dalton
Trans. 2009, 5673. *
[19] 'Bismuth-doped ceria, Ce0.90Bi0.10O2: A selective and stable catalyst for
clean hydrogen combustion', Jurriaan Beckers, Adam F. Lee and Gadi Rothenberg,
Adv. Synth. Catal. 2009, 351, 1557.*
[18] 'Marrying gas power and hydrogen energy: A catalytic system for
combining methane conversion and hydrogen generation', Jurriaan Beckers, Cyril
Gaudillère, David Farrusseng and Gadi Rothenberg, Green Chem., 2009, 11, 921.*
[17] 'Selective hydrogen oxidation catalysts via Genetic Algorithms', Jurriaan
Beckers, Frédéric Clerc, Jan Hendrik Blank and Gadi Rothenberg, Adv. Synth.
Catal. 2008, 350, 2237.*
[16] 'Redox properties of doped and supported copper-ceria catalysts', Jurriaan
Beckers and Gadi Rothenberg, Dalton Trans. 2008, 6573.*
[15] 'Selective hydrogen oxidation in presence of C3 hydrocarbons using
perovskite oxygen reservoirs', Jurriaan Beckers, Ruben Drost, Ilona van Zandvoort,
Paul F. Collignon and Gadi Rothenberg, ChemPhysChem 2008, 9, 1062.*
* These articles originated from the PhD-project (2005–2009).
List of publications
246
[14] 'Redox kinetics of ceria-based mixed oxides in selective hydrogen
combustion', Jan Hendrik Blank, Jurriaan Beckers, Paul F. Collignon and Gadi
Rothenberg, ChemPhysChem 2007, 8, 2490.*
[13] 'A “green route” to propene through selective hydrogen oxidation', Jan
Hendrik Blank, Jurriaan Beckers, Paul F. Collignon, Frédéric Clerc and Gadi
Rothenberg, Chem. Eur. J. 2007, 13, 5121.*
[12] 'Clean diesel power via microwave susceptible oxidation catalysts',
Jurriaan Beckers, Lars M. van der Zande and Gadi Rothenberg, ChemPhysChem,
2006, 7, 747.
[11] '‘Hot spot’ hydrocarbon oxidation catalysed by doped perovskites –
towards cleaner diesel power', Jurriaan Beckers and Gadi Rothenberg,
ChemPhysChem, 2005, 6, 223.
[10] 'Nanocluster-based cross-coupling catalysts: A high-throughput approach',
Mehul B. Thathagar, Jurriaan Beckers and Gadi Rothenberg, Catal. Org. React.,
2005, 104, 211.
[9] 'Design and parallel synthesis of new oxidative dehydrogenation catalysts',
Gadi Rothenberg, Bart E.A. de Graaf, Jurriaan Beckers and Alfred Bliek, Catal.
Org. React., 2005, 104, 201.
[8] 'The effect of the reduction temperature on the structure of Cu/ZnO/SiO2
catalysts for methanol synthesis', Erdni D. Batyrev, Johannes C. van der Heuvel,
Jurriaan Beckers, Wim P.A. Jansen and Hessel L. Castricum, J. Catal., 2005, 229,
136.
* These articles originated from the PhD-project (2005–2009).
List of publications
247
[7] 'Palladium-free and ligand-free Sonogashira cross-coupling', Mehul B.
Thathagar, Jurriaan Beckers and Gadi Rothenberg, Green Chem., 2004, 6, 215.
[6] 'Dielectric heating effects on the activity and SO2 resistance of
La0.8Ce0.2MnO3 perovskite for methane oxidation', Ye Zhang-Steenwinkel, Hessel
L. Castricum, Jurriaan Beckers, Erica Eiser and Alfred Bliek, J. Catal. 2004, 221,
523.
[5] 'Using La1-xCexMnO3 Perovskites as the active components for soot filter
regeneration by dielectric fields', Ye Zhang-Steenwinkel, Jurriaan Beckers, Hessel
L. Castricum and Alfred Bliek, Proceedings of the Third World Congress on
Microwave and Radio Frequency Applications, 2003, Sydney, Australia.
[4] 'Combinatorial design of copper-based mixed nanoclusters: new catalysts
for Suzuki cross-coupling', Mehul B. Thathagar, Jurriaan Beckers and Gadi
Rothenberg, Adv. Synth. Catal. 2003, 345, 979.
[3] 'Copper-catalyzed Suzuki cross-coupling using mixed nanocluster
catalysts', Mehul B. Thathagar, Jurriaan Beckers and Gadi Rothenberg, J. Am.
Chem. Soc. 2002, 124, 11858.
[2] 'Dynamic behavior of the surface structure of Cu/ZnO/SiO2 catalysis', Wim
P.A. Jansen, Jurriaan Beckers, Johannes C. van der Heuvel, Denier A.W. van der
Gon, Alfred Bliek and Hidde H. Brongersma, J. Catal. 2002, 210, 229.
[1] 'Surface Properties and Catalytic Performance in CO oxidation of cerium
substituted Lanthanum-Manganese oxides', Ye Zhang-Steenwinkel, Jurriaan
Beckers and Alfred Bliek, Appl. Catal. A: Gen. 2002, 235, 79.
248
List of abbreviations
249
List of abbreviations
BET Brunauer Emmett Teller
DSC Differential scanning calorimetry
EPR Electron paramagnetic resonance
EXAFS Extended X-ray absorption fine structure
FCC Fluid catalytic cracking
GA Genetic algorithm
GC Gas chromatography
GHSV Gas hourly space velocity
ICP Inductive coupled plasma
LEIS Low-energy ion scattering spectroscopy
MS Mass spectrometry
MTBE Methyl tertiary butyl ether
ODH Oxidative dehydrogenation
POM Partial oxidation of methane
PP Polypropene
PROX Preferential oxidation
SHC Selective hydrogen combustion
SMART Styrene monomer advanced reheat technology
SMSI Strong metal-support interaction
SOR Solid oxygen reservoir
STAR Steam active reforming
TCD Thermal conductivity detector
TEM Transmission electron microscopy
TGA Thermo gravimetric analysis
TPR Temperature programmed reduction
TWC Three-way catalyst, or three-way converter
WHSV Weight hourly space velocity
XPS X-ray photoelectron spectroscopy
XRD X-ray diffraction
250
Dankwoord
251
Dankwoord
Dit zou dan het moeilijkste stukje moeten zijn. Nou, ik vind het toch
makkelijker dan een artikel schrijven:
Bedankt!
Maar goed: wie dan, en waarvoor? Nu volgt wel iets moeilijks, omdat een
hele hoop mensen bovenaan moeten. Let dus alstublieft niet teveel op de volgorde!
De eerste plaats is natuurlijk wel duidelijk: Gadi, man achter en voor de schermen.
Het is mij op congressen vaak overkomen dat er een glimlach op het gezicht van
een voor mij onbekend persoon verscheen als ik vertelde voor wie ik werkte: “Ah,
Gadi, ja!” Dat zegt op zich genoeg, maar ik ga toch uitwijden. Slechts twee zinnen
voor je promotor kan natuurlijk niet! Gadi, het was erg plezierig om met je samen
te werken. Door de hoge graad van faciliteren (van alles regelen) kon ik veel tijd
aan onderzoek besteden en was er veel mogelijk. Het was ook leuk dat, als ik dacht
dat ik helemaal vast zat, ik iedere keer na een gesprek met jou weer opgelucht en
enthousiast over mijn werk was en dat er ook altijd wel tijd voor zo'n gesprek was.
Daar heb je (onder meer) voor gezorgd, bedankt! En ook bedankt voor de vele
leestips, waaronder je eigen boek ;-).
Goed, dat is één. Dan de tweede persoon zonder wie deze promotie niet
mogelijk was geweest. Kees, ik weet dat we over onderzoek weinig gesproken
hebben (anorganische chemie is toch nog wel erg organisch!) maar zonder
promotor had ik natuurlijk nooit kunnen beginnen. Dat is volgens mij te danken
aan je vruchtbare instelling dat dingen in principe mogelijk zijn, mits verantwoord,
in plaats van Heel Erg Moeilijk.
Freek en Jorrit, bedankt voor de samenwerking en de kritische blik.
Jan Hendrik, het jaar dat we samengewerkt hebben heeft een enorme push
aan mijn onderzoek gegeven. Dit heeft geresulteerd in drie publicaties. Mijn
briljante idee om de eerste generatie monsters op de six-flow te screenen terwijl ik
mijn opstelling afbouwde, bleek al snel niet te werken. Toen zijn we samen flink
aan het bouwen gegaan, met goed resultaat. Verder heb je waanzinnige Excel
macro's opgesteld, die ik tot het eind toe gebruikt heb (“Kan dat ook met Excel?”
Dankwoord
252
“Ja, joh, staat gewoon in de helpfunctie!”), heb je je bezig gehouden met Karl
Fischer metingen, TGA, de synthese van katalysatoren, XRD enzovoorts. Niet
zeuren, gewoon doorwerken en altijd oog voor fundamentele aspecten en het
begrijpen van de zaak. Mooi hoor! Gelukkig ben je ook gaan promoveren.
Marjo, toen Lars en ik net analist waren bij de UvA, was een dagelijkse
spreuk “even aan Marjo vragen”. Eigenlijk ben ik daar tijdens mijn promotie
gewoon mee door gegaan (maar gelukkig niet elke dag meer!). Dat je ook eens
kwam vragen hoe het ging toen ik het even wat minder zag zitten, heb ik ook heel
erg gewaardeerd, bedankt!
Paul, zonder jou had ik veel minder data kunnen vergaren. Het deels
bouwen en onderhouden van de opstelling en zeker het automatiseren ervan zijn
goud waard (ik kon 's nachts gewoon slapen terwijl de meting draaide). Niet
onbelangrijk: het was ook heel leuk om met je samen te werken.
Bart de Graaf, je hebt de basis gelegd voor mijn promotie. Ik heb nog vaak
in je proefschrift gebladerd en kon ook nog altijd bellen met vragen. Dankjewel!
Bert en Willem, de XRD analyse van mijn (vele) katalysatoren was
essentieel voor mijn onderzoek. Ik stel het zeer op prijs dat ik zoveel bij jullie heb
kunnen meten en dat er, als er heel af en toe iets met de apparatuur was, het zo
goed als direct weer gerepareerd was. Hartelijk bedankt!
Leo en Ton, ook bedankt voor al het ICP doe- en denkwerk. Het was leuk
en nuttig en ik heb met plezier op jullie lab gewerkt.
Ruben en Ilona, mooi dat er uit jullie project ook een publicatie is gerold!
Ik beloof dat we ‘een volgende keer’ wel katalysatoren zullen maken.
Bart van der Linden, nu hebben we in Amsterdam en in Delft
samengewerkt! Bij mijn volgende werkgever weer?
Jan en Tiny, het was voor mij, met mijn voorliefde voor Wageningen, een
groot plezier om bij jullie op bezoek te komen. Het was gezellig.
Natuurlijk ook dank aan alle ASPECT leden en de ASPECT-AiO's en Post
doc's, in het bijzonder aan Arlette en Dorine voor al het geregel. Ik vond het zeer
waardevol om regelmatig presentaties te geven voor een groot publiek bestaande
uit mensen van de industrie en andere universiteiten!
Gooitzen, we hebben maar weinig samen gewerkt maar het was wel leuk.
Dankwoord
253
Lars, we hebben na 2004 niet meer samen gewerkt, maar daarvoor wel heel
erg prettig! Jammer dat het niet langer kon, maar wie weet, misschien in de
toekomst weer?
Han, ook wij hebben tijdens mijn promotie niet samengewerkt, maar
daarvoor wel, en dat was heel leuk, bedankt.
Peter Verschuren, we hebben eigenlijk nooit samengewerkt, maar de
lunches waren altijd gezellig.
Fédéric Clerc, c’était un réel plaisir de travailler ainsi que de publier deux
articles ensemble. Merci pour ton aide, même lorsque tu avais fini ton PhD et que
tu étais très occupé par ton nouveau travail pour une entreprise commerciale.
Par la même occasion, afin de rester en France: Claude, David, Cyril,
Vesna, Cécile, Laurence, Aline: Merci pour tous les bons moments passés à Lyon.
J’ai pris beaucoup de plaisir à vivre et travailler dans votre pays et, depuis lors, je
me demande, parfois, pourquoi je vie dans ce pays si froid et si gris qu’est le mien!
Bien entendu, j’ai vraiment apprécié de travailler avec vous tous d’un point de vue
personnel.
Adam and Karen, thanks for the XPS measurements which are a great
addition to my thesis! Thanks for being so helpful and kind.
Luckily, I was not alone in my office. Thanks to Tony and Nina for your
company during my PhD research. Erik Jan, het was heel handig om in mijn laatste
jaar twee keer per week een vraagbaak tegenover me te hebben!
Ron en Johan, ik vond het een leuke ervaring om eens praktijkles te geven.
De proeven waren leuk en de sfeer was goed, ook dankzij de AiO's in de
‘lesgeefgroep’ en de studenten natuurlijk!
Ook nog dank aan de mensen die in mijn promotiecommissie plaats
hebben willen nemen, ook al hebben ze het heel druk.
Arjen, wij zitten op precies hetzelfde opleidingspad. Een soort twee-mans
lotgenotengroep. Twee is veel beter dan één.
Klaas, bedankt voor het nauwkeurig doorlezen van mijn concept-
proefschrift, het is er beslist beter van geworden!
Sebastiaan, bedankt voor het prachtige kaftontwerp.
Goed, nu gaan en we een versnelling hoger, want het wordt te gek. Alle
(ex) AiO's en Post doc's en medewerkers van Gadi en op de 6e: Jia, Hessel, Laura,
Dankwoord
254
Anil, Erdni, Ye, Mehul, Susana, Enrico, Zbig, Nina, Mikeal, Tony, Dion (ok, niet
van de 6e!), Ana, Irene, Zea (thanks for the French translations & good luck with
your PhD!) etc etc: bedankt voor alle gezelligheid, overal.
Iedereen van de mechanische werkplaats (Wietze, Theo, Daan, Kees,
Henk, Tjerk, en alle anderen) hartelijk bedankt voor alle hulp. Hetzelfde voor de
mensen van de glasblazerij (o.a. Gerry, Bertus) de bieb (Marijke, Judith e.v.a.) het
centraal magazijn (Mike en Martin e.a.) en de inkoop (o.a. Wim, Wim, Boudewijn,
we hebben altijd heel goed samengewerkt, jammer dat jullie verhuisd zijn!). Ook
Maureen, Renate, Marianne Braaksma, Peter Scholts en Petra bedankt. Jetske, Ida
en Helene bedankt voor de extra snelle afhandeling van het ‘Verzoek tot
samenstelling van de promotiecommissie’!
Mensen van de 7e (Peter, Stefan, Luc, Jan Meine, Dorette, Hans Werner,
David Dominguez, Erica, Alexandre, Jeroen, Michael, Fred, etc. etc.): ik vond het
heel leuk om eens per dag koffie te drinken met een andere groep (even wat
anders!) en het was ook leuk tijdens group meetings en congressen.
Mensen van andere verdiepingen, zoals Bert Sandee (allebei uit Zeeland en
dan ook nog dezelfde mensen kennen, en op elkaars bruiloft geweest!), Petra, en de
sloot mensen van de 9e natuurlijk.
Zo, dat is eruit, nu kunnen we weer wat langzamer. Hans, Peter en
Constant, onze ‘vaste prik’ op zondagochtend was erg belangrijk en plezierig voor
me. Even iets dat niets met scheikunde of promoveren van doen had. Het laatste
(half) jaar is het er niet meer zo van gekomen, maar ik hoop dat we na de promotie
de draad weer oppikken, onder het motto: alles is al gedaan, maar daar moet je je
geen zak van aantrekken, en gewoon doorgaan met mooie dingen maken!
Dan nog dank aan alle niet scheikundigen waaronder mama, Mathijs, Jos
(semi-scheikundig), Marijke, Jan en Digna voor het vragen hoe het ging en het
vertrouwen dat het wel zou lukken (gebaseerd op...?).
Dankwoord
255
En dan als laatste natuurlijk, (of toch beter als eerste? Nee, dit is een goede
plek): Sanne. Jij verenigt het ‘niets met scheikunde te maken hebben’ met het
‘mogelijk maken van de promotie’, wat op zich al bijzonder is. Dat begon al bij het
maken van de keuze om er überhaupt aan te beginnen en ging door tijdens de
promotie. Het heeft mij zo vaak opgelucht om aan het eind van de dag even te
praten en mijn gedachten te ordenen. Ik vraag me af hoe ik er alleen doorgekomen
zou zijn!
256
Appendix I Success of doping
257
Appendix I: Success of doping
The success of doping is determined by XRD and evaluation of the colour
of the sample. Doping is considered unsuccessful when a separate dopant-metal or
dopant oxide phase is detected by XRD, or when the sample's colour is not
homogeneous (spots appear during calcination).
The samples are grouped on the type of dopant, and ordered for increasing
dopant concentration (mol%). The type and concentration of the second dopant is
added, when applicable. Note that bi-doped samples appear twice in the table, once
for each dopant type. Also, doping can be successful for dopant 1, but not for
dopant 2 of the same sample. All catalyst were prepared via the standard
preparation-method (see below), and calcined at 700 °C, unless stated otherwise.
General procedure for catalyst synthesis. The metal nitrate precursors
(or chlorides or ammonium metallates, where nitrates were not available) were
weighed into a crucible and placed on a heater. When liquefied, they were mixed
with a spatula. If necessary, 2–6 drops of water were added to aid the solution of
the precursors. After about 5 minutes, the crucible was placed in a 140 °C vacuum
oven. Pressure was reduced to < 10 mbar in about 10 minutes. The latter was
performed carefully to prevent vigorous boiling. After 4h, the crucible was placed
in a muffle oven and calcined for 5h at 700 °C in static air (ramp rate: 300 °C/h).
The resulting solid was pulverized, ground and sieved in fractions of 125–212 µm
(selectivity assessment) and < 125 µm (XRD and BET measurements). The final
metal concentration was calculated from the amount of precursor weighed in,
corrected for the water content as determined on catalysts G1–01 to G1–18 by ICP.
Appendix I Success of doping
258
Table AI. The success of the synthesis of various doped cerias.
Successful doping Unsuccessful doping
Dopant
type Conc.
(mol%)
Second
dopant, conc.
(mol%)
Conc.
(mol%)
Second
dopant, conc.
(mol%)
Type of extra
phase[d]
Comments [a,b,c]
Ag 8 Sr, 5 Ag metal
Au 10 Au metal
Al 2
Al 2 Cu, 2
Al 2 Cr, 2
Al 2 Pt, 2
Al 2 La, 5
Al 2 Yb, 5
Al 2 Cu, 5
Al 2 Bi, 8
Al 5
Al 8 Ta, 5
Al 10
Bi 2
Bi 2 Gd, 2
Bi 2 Mn, 2
Bi 2 La, 2
Bi 2.5 Cr, 2.5
Bi 5 Pt, 5
Bi 5 Cr, 5
Bi 5 K, 5
Bi 5 Cr, 8
Bi 8
Bi 8 Cu, 2
Bi 8 Al, 2
Bi 8 Sn, 5
Bi 10
Ca 2
Ca 2 Pt, 2
Ca 2 Sr, 2
Ca 5 Pb, 5
Ca 10
Ca 10 Cu, 5
Cu 0.1
Appendix I Success of doping
259
Table AI, continued.
Successful doping Unsuccessful doping
Dopant
type Conc.
(mol%)
Second
dopant, conc.
(mol%)
Conc.
(mol%)
Second
dopant, conc.
(mol%)
Type of extra
phase[d]
Comments [a,b,c]
Cu 1
Cu 2 Al, 2
Cu 2 Zr, 2
Cu 2 K, 2
Cu 2 Mn, 2
Cu 2 Ru, 2
Cu 2 Bi, 8
Cu 2 W, 8
Cu 2 Mn, 10
Cu 3
Cu 4 Pb, 2.5
Cu 4 Zr, 4
Cu 5 Al, 2
Cu 5 Mn, 2
Cu 5 Sn, 2
Cu 5 Gd, 8
Cu 5 Ca, 10
Cu 7
Cu 8
Cu 8 Mg, 5
Cu 8 Zr, 8
Cu 10
Cu 10 Mn, 2
Cu 3 CuO C800
Cu 7 CuO C800
Cu 8 Mn, 8 Cu-Mn-O
Cu 8 Sn, 5 CuO a.m.
Cu 10 Ru, 2 CuO
Cu 10 Pr, 8 CuO
Cu 10 Cr, 2 CuO
Cu 15 CuO
Cr 2
Cr 2 Al, 2
Cr 2 Ta, 5
Appendix I Success of doping
260
Table AI, continued.
Successful doping Unsuccessful doping
Dopant
type Conc.
(mol%)
Second
dopant, conc.
(mol%)
Conc.
(mol%)
Second
dopant, conc.
(mol%)
Type of extra
phase[d]
Comments [a,b,c]
Cr 2 W, 10
Cr 2 Cu, 10
Cr 2 Fe, 10
Cr 2.5 Bi, 2.5
Cr 5 a.m., C450
Cr 5 a.m., C550
Cr 5 a.m., C625
Cr 5 a.m., C750
Cr 5 a.m.
Cr 5 Pt, 2
Cr 5 Zr, 5
Cr 5 Bi, 5
Cr 8
Cr 8 a.m.
Cr 8 Bi, 5
Cr 8 Sn, 5 a.m.
Cr 8 Fe, 8
Cr 5 spotted
Cr 5 Cr2O3 a.m., C800
Cr 8 Ti, 2 Cr2O3
Cr 8 Ru, 5 Cr2O3
Cr 10 Cr2O3
Fe 2 Zr, 2
Fe 2 Nd, 8
Fe 5 Y, 5
Fe 5 Ru, 5
Fe 8 Sr, 2
Fe 8 Mn, 2
Fe 10
Fe 10 Cr, 2
Fe 8 Ti, 2 Fe2O3
Fe 8 Cr, 8 Fe2O3
Gd 2
Gd 2 Bi, 2
Appendix I Success of doping
261
Table AI, continued.
Successful doping Unsuccessful doping
Dopant
type Conc.
(mol%)
Second
dopant, conc.
(mol%)
Conc.
(mol%)
Second
dopant, conc.
(mol%)
Type of extra
phase[d]
Comments [a,b,c]
Gd 2 Pr, 2
Gd 2 Mn, 5
Gd 2 Yb, 5
Gd 2 Yb, 8
Gd 5
Gd 8 Cu, 5
In 5 In2O3
In 10 In2O3
K 2
K 2 Cu, 2
K 2 Yb, 5
K 5 Bi, 5
K 10
Li 10 Li-oxide
La 2
La 2 Bi, 2
La 5 Al, 2
La 8 Sn, 5
La 10
Mg 5 Cu, 8
Mg 8 Zr, 2
Mg 10 MgO
Mn 2
Mn 2 Bi, 2
Mn 2 Cu, 2
Mn 2 Cu, 5
Mn 2 Fe, 8
Mn 2 Cu, 10
Mn 5
Mn 5 Gd, 2
Mn 5 Sr, 2
Mn 5 Sr, 5
Mn 10 Mn2O3
Mn 10 Cu, 2 Mn-oxide
Appendix I Success of doping
262
Table AI, continued.
Successful doping Unsuccessful doping
Dopant
type Conc.
(mol%)
Second
dopant, conc.
(mol%)
Conc.
(mol%)
Second
dopant, conc.
(mol%)
Type of extra
phase[d]
Comments [a,b,c]
Mn 8 Cu, 8 Cu-Mn-O
Mo 10 Mo-oxide
Ni 10 Sm, 2 Ni-oxide
Nd 2
Nd 2 Sn, 2
Nd 8 Fe, 2
Nd 10
Pb 2 a.m.
Pb 2.5 Cu, 4
Pb 8
Pb 8 a.m.
Pb 8 a.m.
Pb 2 spotted
Pb 2.5 Sr, 4 PbO
Pb 5 PbO
Pb 5 Zr, 2 PbO
Pb 5 Ca, 5 spotted
Pb 5 Sr, 8 PbO
Pb 10 Pb-oxide
Pd 2 Sn, 2
Pd 5
Pd 10
Pr[e] 2
Pr 2 Gd, 2
Pr 2 W, 2
Pr 2 Zr, 5
Pr 8 Cu, 10
Pt 2
Pt 2 Al, 2
Pt 2 Ca, 2
Pt 2 Mn, 10
Pt 2 Sn, 2 Pt metal
Pt 2 Cr, 5 Pt metal
Pt 2 W, 8 Pt metal
Appendix I Success of doping
263
Table AI, continued. Successful doping Unsuccessful doping
Dopant
type Conc.
(mol%)
Second
dopant, conc.
(mol%)
Conc.
(mol%)
Second
dopant, conc.
(mol%)
Type of extra
phase[d]
Comments [a,b,c]
Pt 5 Bi, 5 Pt metal
Ru 2
Ru 2 Cu, 2
Ru 2 Cu, 10
Ru 5
Ru 5 Cr, 8
Ru 5 Sm, 5
Ru 5 Fe, 5 RuO2
Ru 8 RuO2
Sm 2 Ni, 10
Sm 5 Ru, 5
Sn 2
Sn 2 Nd, 2
Sn 2 W, 2
Sn 2 Pd, 2
Sn 2 Pt, 2
Sn 5 Bi, 8
Sn 5 La, 8
Sn 8 a.m.
Sn 2 Cu, 5 SnO2
Sn 5 Cr, 8 SnO2 a.m.
Sn 5 Cu, 8 SnO2 a.m.
Sn 10 Sn oxide
Sr 2
Sr 2 Mn, 5
Sr 2 Ca, 2
Sr 4 Pb, 2.5
Sr 5
Sr 5 Mn, 5
Sr 5 Y, 5
Sr 5 Ag, 8
Sr 8 Pb, 5
Sr 2 Zr, 5 Sr(CeO3)
Sr 10 Ce-Sr-O
Appendix I Success of doping
264
Table AI, continued.
Successful doping Unsuccessful doping
Dopant
type Conc.
(mol%)
Second
dopant, conc.
(mol%)
Conc.
(mol%)
Second
dopant, conc.
(mol%)
Type of extra
phase[d]
Comments [a,b,c]
Ta 2 Ta2O5
Ta 5 Ti, 5 Ta2O5
Ta 5 Cr, 2 Ta2O5
Ta 5 Ti, 8 Ta2O5
Ti 2
Ti 2 Fe, 8
Ti 2 Cr, 8
Ti 5 Ta, 5
Ti 8
Ti 8 Ta, 5
V 8 Mo, 8 V-oxide
W 2
W 2 Sn, 2
W 2 Pr, 2
W 10 Cr, 2
W 2 W-oxide
W 8 Cu, 2 W-oxide
W 8 Pt, 2 W-oxide
W 10 W-oxide
Y 2
Y 5
Y 5 Fe, 5
Y 5 Sr, 5
Yb 2
Yb 5 K, 2
Yb 5 Al, 2
Yb 5 Gd, 2
Yb 8 Gd, 2
Yb 10
Zr 2
Zr 2 Fe, 2
Zr 2 Cu, 2
Zr 2 Pb, 5
Zr 2 Fe, 8
Appendix I Success of doping
265
Table AI, continued.
Successful doping Unsuccessful doping
Dopant
type Conc.
(mol%)
Second
dopant, conc.
(mol%)
Conc.
(mol%)
Second
dopant, conc.
(mol%)
Type of extra
phase[d]
Comments [a,b,c]
Zr 2 Mg, 8
Zr 4 Cu, 4
Zr 5 Pr, 2
Zr 5 Sr, 2
Zr 5 Cr, 5
Zr 8
Zr 8 Cu, 8
Zr 10 [a] a.m. : the sample is made with the adjusted preparation method described in chapter 2.3. The
difference with the standard preparation method is that the dopant precursors are placed in a crucible
without adding the cerium nitrate, and dissolved in as little water as possible. Then, the cerium nitrate
is added and mixed with the dissolved dopant precursor(s) into a slurry. This is heated on a heating
plate until the cerium nitrate melts, and then placed in a vacuum oven. In the standard preparation
method, the dopant precursors and the cerium nitrate are both placed in a crucible, mixed, and heated
on a heating plate. If one of the precursors does not melt or dissolve, a few drops of water are added,
and the mixture is placed in a vacuum oven. [b] C[value] means that the sample is not calcined at
700 °C but at the value given. [c] ‘Spotted’ means that no XRD was performed since the colour of the
sample was not homogeneous after calcination, indicating an extra phase has formed. [d] Determined
by XRD. If the precise phase cannot be detected, the extra phase is labelled ‘dopant-oxide’. [e] The
XRD peaks of ceria and Pr-oxide are similar, so it is hard to determine of a separate dopant phase has
formed.
266
Appendix II Catalyst activity
267
Appendix II: Catalyst activity
II.a Catalyst activity expressed in various units
II.b Catalyst activity in the mixed
dehydrogenation and selective hydrogen
combustion process
Appendix II Catalyst activity
268
Appendix II Catalyst activity
269
II.a Catalyst activity expressed in various units
The amounts of oxygen released by the SOR catalysts is determined via the
reduction in hydrogen (TPR and TGA analysis), and by the amount of hydrogen
combusted from the dehydrogenation gas mixture (4:1:1% v/v C3H8:C3H6:H2 in Ar
at 550 °C). This ‘oxygen storage capacity’ can be expressed in several ways, such
as mass over mass, mass over volume and so on. In this section, the activity data,
mainly from the catalysts of Chapter 3.1, is given in several units.
Tables A.1 and A.2 show the oxygen release of several SORs determined
by H2-TPR. Generally, the ‘mol O / kg catalyst’ allows for an easy comparison.
The data of catalysts 1–7 is taken from Chapter 3.1. Catalyst 8 is plain ceria
calcined at low temperature (550 °C instead of 700 °C), resulting in a larger surface
area. The ‘Ref.’ data is the theoretical maximum amount of oxygen which can be
released from ceria, when full (surface and bulk) reduction to Ce2O3 occurs. That
is, a maximum 25% of the available oxygen can be released (note that we did not
heat any sample above 800 °C).
Table AII.1 shows the oxygen release of the catalysts calculated from the
size of TPR peak C (surface reduction, see Scheme 2 in Chapter 3.1). Catalysts 7
and 8 show that the amount of oxygen released (by weight or volume catalyst),
nearly doubles when doubling the catalyst surface area. This shows that optimising
the catalyst synthesis method is a promising route for increasing catalyst activity.
Where our synthesis method yields surface areas of about 20–50 m2/g, alternative
synthesis methods can give surface areas of 125–160 m2/g for ceria or ceria
zirconia mixed oxides.[1, 2] The highest surface area reported is 230 m2/g (when
calcining at 450 °C).
Table AII.2 shows the ‘total’ oxygen release of the catalysts, calculated
from both TPR peak C and peak B (surface and bulk reduction, see Scheme 2 in
Chapter 3.1). Since bulk reduction occurs above 600 °C, this oxygen will not be
available in the dehydrogenation reaction (generally performed at 550–600 °C).
App
endi
x II
Cat
alys
t act
ivit
y
270
T
able
AII
.1. Q
uant
itat
ive
TP
R d
ata
part
I: s
urfa
ce r
educ
tion
(TP
R p
eak
C).
[a]
Siz
e T
PR
pea
k C
(su
rfac
e re
duct
ion)
[a]
Cat
alys
t/ C
ompo
sitio
n S
urfa
ce a
rea
(m2 /g
)
Cry
stal
lite
siz
e
(nm
) m
g O
/ 10
0 m
g
sam
ple[b
]
mg
O /
m2
surf
ace
area
kg O
/ m
3
sam
ple[c
]
mol
O /
kg
sam
ple
kmol
O /
m3
sam
ple[c
]
1 C
e 0.9
1Mn 0
.09O
2 56
11
0.
88
0.16
67
0.
57
4.2
2 C
e 0.9
0Bi 0
.10O
2 33
18
1.
24
0.38
95
0.
80
5.9
3 C
e 0.9
0Cu 0
.10O
2 47
15
1.
22
0.26
93
0.
75
5.8
4 C
e 0.9
0Fe 0
.10O
2 50
14
1.
15
0.23
88
0.
75
5.5
5 C
e 0.9
2Pb 0
.08O
2 56
13
1.
24
0.22
95
0.
80
5.9
6 C
e 0.9
1Ca 0
.09O
2 22
28
0.
47
0.22
36
0.
29
2.2
7 C
eO2
38
26
0.38
0.
10
29
0.23
1.
8
8 C
eO2
‘C55
0’ [
d]
84
10
0.71
0.
08
54
0.43
3.
4
Ref
. CeO
2 →
Ce 2
O3[e
] -
- 4.
65
- 35
6 2.
91
22.2
[a
] Pea
k A
in
case
of
cata
lyst
7 a
nd 8
(ce
ria)
. [b
] Dat
a ob
tain
ed b
y ca
libr
atin
g th
e T
CD
det
ecto
r us
ing
a C
uO s
tand
ard.
The
pea
k ar
ea o
f th
is
stan
dard
is
inte
grat
ed a
nd t
he a
rea
is c
orre
late
d to
the
am
ount
of
oxyg
en p
rese
nt i
n th
e C
uO.
[c] N
ote
that
the
val
ues
per
volu
me
sam
ple
are
give
n as
‘kg
’ an
d ‘k
mol
’ in
stea
d of
‘m
g’ a
nd ‘
mol
’. [d
] Thi
s pl
ain
ceri
a sa
mpl
e w
as c
alci
ned
at 5
50 º
C in
stea
d of
700
ºC
(1–
7), r
esul
ting
in a
high
er s
urfa
ce a
rea
and
smal
ler
crys
tall
ite
size
. [e
] Add
ed a
s re
fere
nce,
thi
s is
the
max
imum
of
avai
labl
e ox
ygen
fro
m c
eria
, fr
om t
he f
ull
(sur
face
and
bul
k) r
educ
tion
of
CeO
2 to
Ce 2
O3
(a q
uart
er o
f th
e ox
ygen
can
be
rele
ased
).
App
endi
x II
Cat
alys
t act
ivit
y
271
T
able
AII
.2. Q
uant
itativ
e T
PR
dat
a pa
rt I
I: s
urfa
ce a
nd b
ulk
redu
ctio
n (T
PR
pea
k C
+B
).[a
]
Siz
e T
PR
pea
k C
+B
(su
rfac
e an
d bu
lk r
educ
tion
)[a]
Cat
alys
t/ C
ompo
sitio
n S
urfa
ce a
rea
(m2 /g
)
Cry
stal
lite
siz
e
(nm
) m
g O
/ 10
0 m
g
sam
ple[b
]
mg
O /
m2
surf
ace
area
kg O
/ m
3
sam
ple[c
]
mol
O /
kg
sam
ple
kmol
O /
m3
sam
ple[c
]
1 C
e 0.9
1Mn 0
.09O
2 56
11
1.
18
0.21
90
0.
77
5.6
2 C
e 0.9
0Bi 0
.10O
2 33
18
1.
96
0.60
15
0 1.
28
9.4
3 C
e 0.9
0Cu 0
.10O
2 47
15
1.
68
0.36
12
9 1.
04
8.0
4 C
e 0.9
0Fe 0
.10O
2 50
14
1.
93
0.39
14
8 1.
26
9.2
5 C
e 0.9
2Pb 0
.08O
2 56
13
1.
56
0.28
11
9 1.
01
7.5
6 C
e 0.9
1Ca 0
.09O
2 22
28
1.
72
0.78
13
2 1.
04
8.2
7 C
eO2
38
26
1.21
0.
32
93
0.75
5.
8
8 C
eO2
‘C55
0’ [
d]
84
10
1.03
0.
12
79
0.63
4.
9
Ref
. CeO
2 →
Ce 2
O3[e
] -
- 4.
65
- 35
6 2.
91
22.2
[a
] Pea
k A
+B
in c
ase
of c
atal
yst 7
and
8 (
ceri
a). [b
] Dat
a ob
tain
ed b
y ca
libra
ting
the
TC
D d
etec
tor
usin
g a
CuO
sta
ndar
d. T
he p
eak
area
of
this
stan
dard
is
inte
grat
ed a
nd t
he a
rea
is c
orre
late
d to
the
am
ount
of
oxyg
en p
rese
nt i
n th
e C
uO.
[c] N
ote
that
the
val
ues
per
volu
me
sam
ple
are
give
n as
‘kg
’ an
d ‘k
mol
’ in
stea
d of
‘m
g’ a
nd ‘
mol
’. [d
] Thi
s pl
ain
ceri
a sa
mpl
e w
as c
alci
ned
at 5
50 º
C in
stea
d of
700
ºC
(1–
7), r
esul
ting
in a
high
er s
urfa
ce a
rea
and
smal
ler
crys
tall
ite
size
. [e
] Add
ed a
s re
fere
nce,
thi
s is
the
max
imum
of
avai
labl
e ox
ygen
fro
m c
eria
, fr
om t
he f
ull
(sur
face
and
bul
k) r
educ
tion
of
CeO
2 to
Ce 2
O3
(a q
uart
er o
f th
e ox
ygen
can
be
rele
ased
).
App
endi
x II
Cat
alys
t act
ivit
y
272
T
able
AII
.3 a
nd A
II.4
sho
w t
he o
xyge
n re
leas
e of
the
cat
alys
ts w
hen
redu
ced
in h
ydro
gen
at 5
50 °
C i
n a
TG
A s
et-u
p
(see
Cha
pter
3.1
). T
able
AII
.3 s
how
s th
e ox
ygen
rel
ease
d du
ring
the
fas
t pa
rt o
f th
e re
duct
ion
curv
e (s
urfa
ce r
educ
tion
), T
able
AII
.4 s
how
s th
e ox
ygen
rel
ease
d in
bot
h th
e fa
st a
nd s
low
par
t (s
ee S
chem
e 2
in C
hapt
er 3
.1).
Not
e th
at t
he e
xper
imen
t w
as
stop
ped
afte
r 15
min
. R
educ
ing
long
er w
ill
incr
ease
the
am
ount
of
oxyg
en r
elea
sed,
but
not
to
a gr
eat
exte
nt (
see
Fig
ure
3 in
Cha
pter
3.1
).
T
able
AII
.3. Q
uant
itat
ive
TG
A d
ata
part
I: ‘
fast
red
ucti
on’
part
(55
0 °C
).
Siz
e ‘f
ast r
educ
tion’
par
t
Cat
alys
t/ C
ompo
sitio
n S
urfa
ce a
rea
(m2 /g
)
Cry
stal
lite
siz
e
(nm
) m
g O
/ 10
0 m
g
sam
ple
mg
O /
m2
surf
ace
area
kg O
/ m
3
sam
ple[a
]
mol
O /
kg
sam
ple
kmol
O /
m3
sam
ple[a
]
1 C
e 0.9
1Mn 0
.09O
2 56
11
0.
56
0.10
43
0.
49
2.7
2 C
e 0.9
0Bi 0
.10O
2 33
18
1.
23
0.37
94
0.
92
5.9
3 C
e 0.9
0Cu 0
.10O
2 47
15
1.
00
0.21
77
0.
77
4.8
4 C
e 0.9
0Fe 0
.10O
2 50
14
1.
30
0.06
99
0.
83
6.2
5 C
e 0.9
2Pb 0
.08O
2 56
13
0.
84
0.15
64
0.
63
4.0
6 C
e 0.9
1Ca 0
.09O
2 22
28
0.
30
0.03
23
0.
23
1.4
7 C
eO2
38
26
0.33
0.
08
25
0.27
1.
6
8 C
eO2
‘C55
0’ [
b]
84
10
n.d.
[c]
n.d.
n.
d.
n.d.
n.
d.
Ref
. CeO
2 →
Ce 2
O3[d
] -
- 4.
65
- 35
6 2.
91
22.2
[a
] Not
e th
at th
e va
lues
per
vol
ume
sam
ple
are
give
n as
‘kg
’ an
d ‘k
mol
’ in
stea
d of
‘m
g’ a
nd ‘
mol
’. [b
] Thi
s pl
ain
ceri
a sa
mpl
e w
as c
alci
ned
at
550
ºC in
stea
d of
700
ºC
(1–
7), r
esul
ting
in a
hig
her
surf
ace
area
and
sm
alle
r cr
ysta
llit
e si
ze. [c
] Not
det
erm
ined
. [d] A
dded
as
refe
renc
e, th
is is
the
max
imum
of
avai
labl
e ox
ygen
fro
m c
eria
, fr
om t
he f
ull
(sur
face
and
bul
k) r
educ
tion
of
CeO
2 to
Ce 2
O3
(a q
uart
er o
f th
e ox
ygen
can
be
rele
ased
).
App
endi
x II
Cat
alys
t act
ivit
y
273
T
able
AII
.4. Q
uant
itativ
e T
GA
dat
a pa
rt I
I: T
otal
oxy
gen
rele
ase
(15
min
mea
sure
men
t, 55
0 °C
).
Tot
al o
xyge
n re
leas
e (1
5 m
in m
easu
rem
ent)
Cat
alys
t/ C
ompo
sitio
n S
urfa
ce a
rea
(m2 /g
)
Cry
stal
lite
siz
e
(nm
) m
g O
/ 10
0 m
g
sam
ple
mg
O /
m2
surf
ace
area
kg O
/ m
3
sam
ple[a
]
mol
O /
kg
sam
ple
kmol
O /
m3
sam
ple[a
]
1 C
e 0.9
1Mn 0
.09O
2 56
11
0.
78
0.14
60
0.
49
3.7
2 C
e 0.9
0Bi 0
.10O
2 33
18
1.
54
0.47
11
8 0.
92
7.4
3 C
e 0.9
0Cu 0
.10O
2 47
15
1.
22
0.26
93
0.
77
5.8
4 C
e 0.9
0Fe 0
.10O
2 50
14
1.
32
0.27
10
1 0.
83
6.3
5 C
e 0.9
2Pb 0
.08O
2 56
13
1.
07
0.19
82
0.
63
5.1
6 C
e 0.9
1Ca 0
.09O
2 22
28
0.
32
0.14
24
0.
23
1.5
7 C
eO2
38
26
0.43
0.
10
33
0.27
2.
0
8 C
eO2
‘C55
0’ [
b]
84
10
n.d.
[c]
n.d.
n.
d.
n.d.
n.
d.
Ref
. CeO
2 →
Ce 2
O3[d
] -
- 4.
65
- 35
6 2.
91
22.2
[a
] Not
e th
at th
e va
lues
per
vol
ume
sam
ple
are
give
n as
‘kg
’ an
d ‘k
mol
’ in
stea
d of
‘m
g’ a
nd ‘
mol
’. [b
] Thi
s pl
ain
ceri
a sa
mpl
e w
as c
alci
ned
at
550
ºC in
stea
d of
700
ºC
(1–
7), r
esul
ting
in a
hig
her
surf
ace
area
and
sm
alle
r cr
ysta
llit
e si
ze. [c
] Not
det
erm
ined
. [d] A
dded
as
refe
renc
e, th
is is
the
max
imum
of
avai
labl
e ox
ygen
fro
m c
eria
, fr
om t
he f
ull
(sur
face
and
bul
k) r
educ
tion
of
CeO
2 to
Ce 2
O3
(a q
uart
er o
f th
e ox
ygen
can
be
rele
ased
).
App
endi
x II
Cat
alys
t act
ivit
y
274
T
able
AII
.5 a
nd A
II.6
sho
w t
he c
atal
yst
acti
vity
in
the
sele
ctiv
e hy
drog
en c
ombu
stio
n fr
om a
mix
ture
with
pro
pane
and
pro
pene
.
The
rea
ctio
n co
nditi
ons
are:
10
min
cyc
les
of 4
:1:1
% v
/v C
3H8:
C3H
6:H
2 in
Ar
(tot
al f
low
50
mL
/min
), w
ith 5
mL
/min
of
N2
adde
d as
inte
rnal
stan
dard
, 550
°C
, 250
mg
of c
atal
yst (
abou
t 0.2
5 cm
3 ), G
HS
V 1
3200
/ h
and
WH
SV 1
.2 /
h (c
alcu
late
d fr
om th
e w
eigh
t of
C3H
8 +
C3H
6 +
H2
per
h pe
r th
e w
eigh
t of
the
cat
alys
t).
The
act
ivity
is
expr
esse
d in
tw
o w
ays,
Tab
le A
II.5
sho
ws
the
cata
lyst
s ‘o
xyge
n de
man
d’.
Thi
s is
the
tota
l am
ount
of
oxyg
en u
sed
by th
e ca
taly
sts
in t
he r
eoxi
dati
on s
tep,
to r
efil
l the
cer
ia la
ttic
e an
d to
com
bust
the
cok
e. I
t th
eref
ore
repr
esen
ts
both
sel
ecti
ve a
nd u
nsel
ecti
ve p
roce
sses
.
T
able
AII
.5. A
ctiv
ity in
the
sele
ctiv
e hy
drog
en c
ombu
stio
n at
550
°C
par
t I: o
xyge
n de
man
d.[a
]
Oxy
gen
dem
and
Cat
alys
t/ C
ompo
sitio
n S
urfa
ce a
rea
(m2 /g
)
Cry
stal
lite
siz
e
(nm
) m
g O
/ 10
0 m
g
sam
ple
mg
O /
m2
surf
ace
area
kg O
/ m
3
sam
ple[b
]
mol
O /
kg
sam
ple
kmol
O /
m3
sam
ple[b
]
1 C
e 0.9
1Mn 0
.09O
2 56
11
1.
90
0.34
14
5 1.
19
9.1
2 C
e 0.9
0Bi 0
.10O
2 33
18
1.
70
0.52
13
0 1.
06
8.1
3 C
e 0.9
0Cu 0
.10O
2 47
15
1.
40
0.30
10
7 0.
88
6.7
4 C
e 0.9
0Fe 0
.10O
2 50
14
3.
00
0.60
23
0 1.
88
14.3
5 C
e 0.9
2Pb 0
.08O
2 56
13
1.
10
0.20
84
0.
69
5.3
6 C
e 0.9
1Ca 0
.09O
2 22
28
0.
20
0.09
15
0.
13
1.0
7 C
eO2
38
26
0.60
0.
16
46
0.38
2.
9
8 C
eO2
‘C55
0’ [
c]
84
10
0.40
0.
05
31
0.25
1.
9
Ref
. CeO
2 →
Ce 2
O3[d
] -
- 4.
65
- 35
6 2.
91
22.2
[a
] Con
diti
ons:
10
min
cyc
les
of 4
:1:1
% v
/v C
3H8:
C3H
6:H
2 in
Ar
at 5
0 m
L/m
in t
otal
flo
w, 5
50 °
C, 2
50 m
g of
cat
alys
t (a
bout
0.2
5 cm
3 ), G
HS
V 1
200
/h. T
he
oxyg
en d
eman
d is
the
oxy
gen
cons
umed
by
the
cata
lyst
dur
ing
reox
idat
ion,
bot
h to
ref
ill t
he la
ttice
oxy
gen
and
to c
ombu
st t
he c
oke.
[b] N
ote
that
the
val
ues
per
volu
me
sam
ple
are
give
n as
‘kg
’ an
d ‘k
mol
’ in
stea
d of
‘m
g’ a
nd ‘
mol
’. [
c] T
his
plai
n ce
ria
sam
ple
was
cal
cine
d at
550
ºC
ins
tead
of
700
ºC (
1–7)
, re
sult
ing
in a
hig
her
surf
ace
area
and
sm
alle
r cr
ysta
llite
siz
e. [d
] Add
ed a
s re
fere
nce,
thi
s is
the
max
imum
of
avai
labl
e ox
ygen
fro
m c
eria
, fr
om t
he f
ull
(sur
face
and
bul
k) r
educ
tion
of
CeO
2 to
Ce 2
O3
(a q
uart
er o
f th
e ox
ygen
can
be
rele
ased
).
Appendix II Catalyst activity
275
Table AII.6 shows the ‘hydrogen activity’ of the catalysts. This is the
amount of the hydrogen feed which is combusted by the catalysts, and therefore
represents the selective reaction only. Besides catalysts 1–8 and the Ref. data, the
data of the most active doped cerias (9–12), the most active perovskites (13, 14,
and 15), and the most active and selective material discovered up to date, PbCrO4
(16), are given. Note that the surface areas of the perovskites and the PbCrO4 are
very low (0–6 m2/g), leaving room for improvement. Interestingly, the ‘real’
activity of PbCrO4 (16), i.e. the hydrogen combustion in simulated reaction
conditions (in the presence of the hydrocarbons and at 550 °C), is comparable to
the theoretical maximum activity of the doped cerias (Ref., 2.8 mol O /kg and
2.9 mol O / kg, respectively). Note that we have only done preliminary tests using
pure PbCrO4 powder, which activity drops to about 25% of the initial value during
prolonged redox cycling (125 cycles, 73 h on stream, see Chapter 2.3). This
catalyst has yet to be optimised concerning stability.
App
endi
x II
Cat
alys
t act
ivit
y
276
T
able
AII
.6. A
ctiv
ity in
the
sele
ctiv
e hy
drog
en c
ombu
stio
n at
550
°C
par
t II:
Hyd
roge
n ac
tivity
.[a]
Hyd
roge
n ac
tivity
Cat
alys
t/ C
ompo
sitio
n
Sur
face
are
a
(m2 /g
) /
Sel
ecti
vity
(%
)[b]
% H
2
com
bust
ed[c
]
mg
O /
100
mg
sam
ple
mg
O /
m2
s.a.
kg O
/ m
3
sam
ple
mol
O /
kg
sam
ple
kmol
O /
m3
sam
ple
1 C
e 0.9
1Mn 0
.09O
2 56
/ 93
4.
6 0.
07
0.01
6
0.05
0.
3
2 C
e 0.9
0Bi 0
.10O
2 33
/ 77
32
.9
0.46
0.
14
36
0.29
2.
2
3 C
e 0.9
0Cu 0
.10O
2 47
/ 89
7.
4 0.
12
0.03
9
0.07
0.
6
4 C
e 0.9
0Fe 0
.10O
2 50
/ 0
-[d]
- -
- -
-
5 C
e 0.9
2Pb 0
.08O
2 56
/ 92
46
.1
0.70
0.
13
54
0.44
3.
4
6 C
e 0.9
1Ca 0
.09O
2 22
/ 0
- -
- -
- -
7 C
eO2
38 /
0 -
- -
- -
-
8 C
eO2
‘C55
0’ [
e]
84 /
0 -
- -
- -
-
9 C
e 0.9
0Bi 0
.10O
2 ‘4
00 º
C’[f
] 33
/ 98
88
.9
1.28
0.
39
98
0.80
6.
1
10 C
e 0.9
5Cr 0
.05O
2 ‘C
800’
[g]
n.
d.[h
] / 91
14
.8
0.24
n.
d.
18
0.15
1.
1
11C
e 0.8
7Cr 0
.08S
n 0.0
5O2/
SnO
2[i]
n.d.
/ 94
33
.1
0.52
n.
d.
40
0.32
2.
5
12 C
e 0.8
7Bi 0
.08S
n 0.0
5O2
55 /
84
44.6
0.
74
0.13
56
0.
46
3.5
13 (
La 0
.7Sr
0.3)
0.98
MnO
3 6
/ 85
71.4
1.
12
1.93
n.
d.
0.70
n.
d.
14 L
a 0.8S
r 0.2M
nO3
5 / 9
2 44
.1
0.69
1.
35
n.d.
0.
43
n.d.
15 L
a 0.9S
r 0.1M
nO3
3 / 9
2 31
.3
0.49
1.
58
n.d.
0.
31
n.d.
16 P
bCrO
4 <
1[j] /
100
48.6
[k]
4.47
-
282
2.79
[l]
17.6
Ref
. CeO
2 →
Ce 2
O3[m
] -
/ -
4.
65
- 35
6 2.
91
22.2
[a
] Con
ditio
ns: 1
0 m
in c
ycle
s of
4:1
:1%
v/v
C3H
8:C
3H6:
H2
in A
r at
50
mL
/min
tota
l flo
w, 5
50 °
C, 2
50 m
g of
cat
alys
t (ab
out 0
.25
cm3 ),
GH
SV
1200
/h.
Hyd
roge
n ac
tivity
is
the
amou
nt o
f th
e hy
drog
en f
eed
whi
ch i
s co
mbu
sted
dur
ing
a re
duct
ion
cycl
e. [
b] S
elec
tivity
for
hyd
roge
n
App
endi
x II
Cat
alys
t act
ivit
y
277
com
bust
ion
from
a m
ixtu
re w
ith p
rope
ne a
nd p
ropa
ne (
1:1:
4, r
espe
ctiv
ely)
at
550
°C,
expr
esse
d as
hyd
roge
n co
nver
sion
/tota
l co
nver
sion
.
The
uns
elec
tive
firs
t da
ta p
oint
is
not
take
n in
to a
ccou
nt. [
c] T
his
is t
he p
erce
ntag
e of
the
hyd
roge
n fe
ed w
hich
is
com
bust
ed b
y th
e ca
taly
st
duri
ng t
he 1
0 m
in r
educ
tion
cyc
le. [
d] T
he u
nsel
ectiv
e ca
taly
sts
show
a n
et p
rodu
ctio
n of
hyd
roge
n vi
a co
king
, th
eref
ore,
the
ir l
evel
of
hydr
ogen
com
bust
ion
and
sele
ctiv
ity c
anno
t be
dete
rmin
ed. [
e] T
his
plai
n ce
ria
sam
ple
was
cal
cine
d at
550
ºC
inst
ead
of 7
00 º
C, r
esul
ting
in a
high
er s
urfa
ce a
rea
and
smal
ler
crys
tall
ite
size
. [f] T
his
is c
atal
yst
2, b
ut t
he c
atal
ytic
tes
ting
was
per
form
ed a
t 40
0 °C
ins
tead
of
550
°C,
whi
ch r
esul
ts i
n a
high
er s
elec
tivity
and
act
ivity
, se
e C
hapt
er 3
.3. [
g] T
his
cata
lyst
was
cal
cine
d at
800
°C
ins
tead
of
700
°C,
resu
lting
in
a
larg
er c
ryst
alli
te s
ize.
Tra
ces
of C
r 2O
3 w
ere
obse
rved
by
XR
D, s
ee C
hapt
er 3
.4. [
h] N
ot d
eter
min
ed. [i
] XR
D a
naly
sis
show
ed t
hat
part
of
the
tin i
s pr
esen
t as
a s
epar
ate
SnO
2 ph
ase.
[j] T
he s
urfa
ce a
rea
is t
oo s
mal
l to
be
dete
rmin
ed. [
k] N
ote
that
for
thi
s ca
taly
st,
~40
mg
inst
ead
of
250
mg
sam
ple
was
use
d. T
here
fore
the
per
cent
age
hydr
ogen
com
bust
ed s
houl
d no
t be
use
d to
com
pare
the
cat
alys
ts. [
l] T
his
is t
he i
nitia
l
activ
ity,
afte
r pr
olon
ged
redo
x cy
clin
g, t
he a
ctiv
ity s
tabi
lises
at
abou
t 25
% o
f th
is v
alue
. [m
] Add
ed a
s re
fere
nce,
thi
s is
the
max
imum
of
avai
labl
e ox
ygen
fro
m c
eria
, fro
m th
e fu
ll (
surf
ace
and
bulk
) re
duct
ion
of C
eO2
to C
e 2O
3 (a
qua
rter
of
the
oxyg
en c
an b
e re
leas
ed).
Appendix II Catalyst activity
278
Appendix II Catalyst activity
279
II.b Catalyst activity in the mixed
dehydrogenation and selective hydrogen
combustion process
In the combined dehydrogenation and selective hydrogen combustion
process, pure propane is fed over a reactor bed containing a dehydrogenation
catalyst and the SOR. Table AII.7 shows how much of the SOR must be added in
the reactor to combust 90% of the hydrogen formed in the dehydrogenation
reaction, and in case of 10 min reaction cycles (which is similar to the Catofin
dehydrogenation process).[3] A hydrogen conversion of 90% is chosen, since at too
low hydrogen levels, hydrocarbon coking will occur. Note that a model study of
this process type has been published by De Graaf et al.[4]
We have used the propane dehydrogenation conditions from data obtained
by Grasselli et. al, namely a 0.7 wt % Pt-Sn-ZSM-5 dehydrogenation catalyst, pure
propane feed at 2/h WHSV (17 cc propane / g cat /min), and about 25% propane
conversion (540 °C).[5] The selectivity of both catalysts is set to 100%. The activity
of the SOR catalyst is based on our measurements in a simulated propane
dehydrogenation gas mixture, unless stated otherwise (4:1:1% v/v C3H8:C3H6:H2 in
Ar at 50 mL/min total flow, 550 °C). Note again that increasing the catalysts
surface area can increase the activity, and that the initial activity of the PbCrO4
catalyst (16) is close to the theoretical maximum activity of (doped) ceria.
Appendix II Catalyst activity
280
Table AII.7. The amount of SOR catalyst needed in the combined
dehydrogenation and selective hydrogen combustion process.[a]
Weight based Volume based
Catalyst/ Composition
Amount of
SOR needed
per kg DH
catalyst (kg)
SOR
needed
(wt %)
Amount of
SOR needed
per m3 DH
catalyst (m3)
SOR
needed
(vol %)
5 Ce0.92Pb0.08O2 3.9 80 1.3 56
9 Ce0.90Bi0.10O2 ‘400 ºC’[b] 2.1 68 0.6 40
15 La0.9Sr0.1MnO3 5.5 85 8.7 90
16 PbCrO4[c] 0.6 38 0.2 20
Ref. CeO2 → Ce2O3[d] 0.6 37 0.2 16
[a] Conditions: 10 min propane dehydrogenation cycles at 540 °C (25% propane
conversion), selectivity of both catalyst is 100%, and the SOR combusts 90% of the
hydrogen produced. Dehydrogenation catalyst: 0.7 wt % Pt-Sn-ZSM-5, pure propane feed
at 2/h WHSV (17 cc propane / g cat /min). The activity of the SOR catalyst is based on our
measurements in a simulated propane dehydrogenation gas mixture, unless stated otherwise
(4:1:1% v/v C3H8:C3H6:H2 in Ar at 50 mL/min total flow, 550 °C). [b] This is catalyst 2, but
the catalytic testing was performed at 400 °C instead of 550 °C, see Chapter 3.3. [c] These
values are based on the initial activity of the catalyst. After prolonged redox cycling, the
activity stabilises at about 25% of this value. [d] Added as reference, this is the maximum of
available oxygen from ceria, from the full (surface and bulk) reduction of CeO2 to Ce2O3 (a
quarter of the oxygen can be released).
Appendix II Catalyst activity
281
Table AII.8 shows the amount of SOR catalyst needed to combust the
hydrogen produced by the full dehydrogenation of 1 kg propane. Note that in a real
reaction at 550 °C, the propane conversion is about 25%. This means that in this
case, one quarter of the amounts of SOR given in Table AII.8 are needed.
Table AII.8. The amount of SOR catalyst needed to combust the hydrogen
produced from the dehydrogenation of 1 kg propane.[a]
Catalyst/ Composition Weight based
(kg)
Volume based
(m3.10-3)
5 Ce0.92Pb0.08O2 51.7 6.8
9 Ce0.90Bi0.10O2 ‘400 ºC’[b] 28.4 3.7
15 La0.9Sr0.1MnO3 73.3 46.1
16 PbCrO4[c] 8.1 1.3
Ref. CeO2 → Ce2O3[d] 7.8 1.0
[a] The activity of the SOR catalyst is based on our measurements in a simulated propane
dehydrogenation gas mixture, unless stated otherwise (4:1:1% v/v C3H8:C3H6:H2 in Ar at
50 mL/min total flow, 550 °C). [b] This is catalyst 2, but the catalytic testing was performed
at 400 °C instead of 550 °C, see Chapter 3.3. [c] These values are based on the initial
activity of the catalyst. After prolonged redox cycling, the activity stabilises at about 25%
of this value. [d] Added as reference, this is the maximum of available oxygen from ceria,
from the full (surface and bulk) reduction of CeO2 to Ce2O3 (a quarter of the oxygen can be
released).
Appendix II Catalyst activity
282
References [1] D. Terribile, A. Trovarelli, J. Llorca, C. de Leitenburg and G. Dolcetti, Catal.
Today, 1998, 43, 79. [2] D. Terribile, A. Trovarelli, J. Llorca, C. de Leitenburg and G. Dolcetti, J. Catal.,
1998, 178, 299. [3] T. A. Nijhuis, S. J. Tinnemans, T. Visser and B. M. Weckhuysen, Chem. Eng. Sci.,
2004, 59, 5487. [4] E. A. de Graaf, G. Zwanenburg, G. Rothenberg and A. Bliek, Org. Process. Res.
Dev., 2005, 9, 397. [5] R. K. Grasselli, D. L. Stern and J. G. Tsikoyiannis, Appl. Catal. A: Gen., 1999,
189, 9.