A spectroscopic characterization of the structure of supported metal catalysts Citation for published version (APA): Martens, J. H. A. (1988). A spectroscopic characterization of the structure of supported metal catalysts. Technische Universiteit Eindhoven. https://doi.org/10.6100/IR282367 DOI: 10.6100/IR282367 Document status and date: Published: 01/01/1988 Document Version: Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication: • A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal. If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement: www.tue.nl/taverne Take down policy If you believe that this document breaches copyright please contact us at: [email protected]providing details and we will investigate your claim. Download date: 31. Dec. 2020
Transcript
A spectroscopic characterization of the structure of supportedmetal catalystsCitation for published version (APA):Martens, J. H. A. (1988). A spectroscopic characterization of the structure of supported metal catalysts.Technische Universiteit Eindhoven. https://doi.org/10.6100/IR282367
DOI:10.6100/IR282367
Document status and date:Published: 01/01/1988
Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication
General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne
Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.
Dit proefschrift is goedgekeurd door de pronwtoren:
prof. dr. R. Prins
en
prof. dr. ir. D. C. Koningsberger
The research reported in this thesis has been carried out at the Laboratory of Inorganic Chemistry and Catalysis at the Eindhoven University of Technology and has been supported by the Netherlands Foundation for Chemical Research (SON) with financial aid from the Netherlands Organization for the Advancement of Pure Research (ZWO).
- v -
aan mijn vader
en aan mijn moeder
aan Angeliene
- VI -
Contents
Chapter 1 Introduction 1 1.1 Catalysis 1 1.2 Heterogeneous Catalysis 2 1.3 Characterization of Supported Catalysts 7 1.4 Scope of this Thesis 8 1.5 References 8
Calculating Spectra 2.9.4 Data Analysis 38 2.9.5 Experimental Method 43
2.10 References 44
Chapter 3 The Preparation of y -Al20 3 supported Monometallic 47 Rh and Pt and Bimetallic Rh-Pt Catalysts 3.1 Introduction 47 3.2 Experimental 48
3.2.1 NMR and Laser Raman Experiments 48 3.2.2 Adsorption Experiments 49 3.2.3 TPR of Rh, Pt and Rh-Pt/ A1 20 3 50
- VII -
3.3 Results and Discussion 51 3.3.1 NMR and Laser Raman Experiments 51 3.3.2 Adsorption Experiments 56 3.3.3 TPR of Rh, Pt and Rh-Pt/ Al 20 3 67
3.4 Conclusions 71 3.5 References 72
Chapter 4Ferric Iron in Reduced Si0 2 Supported 73 Fe-Ru and Fe-Pt Catalysts Evidence from Mt>ssbauer Spectroscopy and Electron Spin Resonance 4.1 Introduction 73 4.2 Experimental 75 4.3 Results and Discussion 76 4.4 Conclusions 81 4.5 References 81
Chapter 5Controlled Oxygen Chemisorption on an 83 Alumina Supported Rhodium Catalyst The Formation of a Metal-Metal Oxide Interface Determined by EXAFS 5.1 Introduction 83 5.2 Experimental 84 5.3 Results 86 5.4 Discussion 90
5.4.1 Rh/Al 20 3 after Reduction and Evacuation 90 5.4.2 A Model for the Oxidation of Metal Particles 95 5.4.3 Rh/ Al 20 3 after Oxygen Admission at 100 K 101 5.4.4 Rh/ A1 20 3 after Warming up to 300 K 102 5.4.5 General Remarks 104
5.5 Conclusions 108 5.6 References 109
Chapter 6 The Structure of the Metal-Support Interface 111 in Rh/ A1 20 3 Determined with the AS ED-MO Method 6.1 Introduction 111 6.2 Theoretical Method 112 6.3 Description of the Model and Results 114
8.4 Discussion 184 8.4.1 Rh/Ti0 2 after Reduction at 494 K 184 8.4.2 Rh/Ti0 2 after Reduction at 623 K 187 8.4.3 Rh/Ti02 after Reduction at 773 K 190 8.4.4 General Remarks 191
8.5 Final Conclusions 194 8.6 References 195
EXAFS Evidence for Direct Rh0-Tan+ Bonding 197 and Coverage of the Metal Part icles in a Rh/Ta20 5 Catalyst in the SMSI State 9.1 Introduction 197 9.2 Experimental 199
9.3 Results 201 9.3.1 Reference Compounds 201 9.3.2 Analysis of the EXAFS Spectra 203
9.4 Discussion 207 9.4.1 Rh/Ta20 5 after Reduction at 523 K 207 9.4.2 Rh/Ta20 5 after Reduction at 858 K 210 9.4.3 Rh/Ta20 5 after Admission of 0 2 215
in the S MS I State 9.5 Final Conclusions 9.6 References
216 217
Chapter 10 Concluding Remarks 219
Summary 225
Samenvatting 231
Dankwoord 237
Curriculum Vitae 240
List of Publications 241
page 1 Chapter 1
Chapter 1
Introduction
1.1 Catalysis
Chemistry is an area of vital importance in todays society, and
within chemistry. catalysis is a mainstay. not only in many indus
trial applications. but also in numerous processes in the chemistry of life. Catalysis performs a key role in processes such as the
conversion of crude oil into a wide scale of useful products, in the preparation of many nutrients, in the conversion of coal via syn
thesis gas into products like alcohols, gasoline, and in the removal
of noxious components from exhaust gases.
Catalysis is the science of accelerating chemical reactions that under normal conditions proceed only slowly or not at all. The rate
of a chemical reaction can be controlled by a few parameters only:
temperature, pressure and composition. lr:i addition, the choice of a suitable catalyst may change the reaction pathway . As a conse
quence, the overall reaction rate may be increased and/or new path
ways, and therefore new products, may become feasible. A chemi
cal reaction is the result of a collision of two or more molecules or
atoms. The function of a catalyst is merely to capture the participants of the reaction, to bring them in close contact and thus to
guide them through some reaction pathway . The combination of
catalyst and reactant( s) dictates the pathway. Two characteristics
are important in describing a catalyst. First of all its activity. that
is the rate at which the products are generated. The higher the
activity, the better the catalyst. Secondly, there is the selectivity. In most cases, a catalyst produces a (wide) range of products, some
of which are useful and others are not . By selectivity in general we
mean the fraction of (useful) products, usually expressed as a
Introduction page 2
percentage . A high selectivity indicates that the catalyst produces
mainly the desired products.
Catalysts are known in may varieties, but in principle they can
be classified into two categories : homogeneous and heterogeneous catalysts. Homogeneous catalysts can be mixed perfectly with the
reactants, i.e., up to molecular scale . Homogeneous catalysis mainly occurs in the liquid phase. The reactants and the catalyst
are liquids or dissolved in the liquid phase. All the enzymes at work
in our body are homogeneous catalysts . In heterogeneous catalysis.
the catalyst. the reactants and the products are in separate phases and therefore the mixing is far from perfect. The catalyst is usually
a solid and the reactants are liquids or gases. The automotive catalyst is an example of a heterogeneous catalyst. The reactants,
hydrogen carbon monoxide and nitric oxides, and the products, water, carbon dioxide and nitrogen, are gaseous while the active
catalyst is a combination of several precious metals supported on
some (inert) support.
1.2 Heterogeneous Catalysis
In this thesis we will describe heterogeneous catalysts containing precious metals. We have already met one application of such
catalysts : the use of noble metals in automotive catalysis. Another
important use of metals is in the Fischer Tropsch synthesis. In this
process, carbon monoxide and hydrogen are combined to long-chain hydrocarbons and oxygen-containing organic molecules like alcohols
( 1-3). The way Fischer Tropsch catalysts and automotive catalysts work is very similar. The metals are capable of adsorbing the reac
tants and of splitting them into smaller fragments or atoms. On the surface of a suitable metal, sometimes with the help of addi
tives. these fragmented molecules rearrange, in the case of Fischer Tropsch synthesis to useful products and in the case of automotive
page 3 Chapter 1
catalysis to harmless products . These products desorb and leave
the surface of the catalyst as gases. Here we have encountered on of the most important aspects of heterogeneous catalysis : only the
surface of the metals is exposed to the (gaseous or liquid) reactants and products. Therefore, only the metal atoms in the surface are
active in the process . The atoms directly beneath the surface
atoms may still play a (minor) role , while all the other atoms in the bulk are 'useless' . To give an impression of this 'waste' : in a metal
sphere or ball of size 1 mm, only one in about 1.3 million atoms is a
surface atom ; in a metal particle of one micron , one in .approximately 1.3 thousand atoms is in the surface of the particle and
therefore potentially useful in a heterogeneous reaction. For inexpensive and commonly available metals, this is hardly an objection .
For precious noble metals which are not only very expensive, but also sometimes very rare, this is worth consideration. The general
aim therefore is to reduce the size of the metal particles and consequently to use the metal as efficiently as possible. This can be
achieved by 'dispersing' the metal on an inert support material . A commonly used support is aluminum oxide, Al 20 3, known as
alumina. It has a typical surface area of several hundreds of square meters per gram . The surface area in a few grams of alumina (a tea spoon full) would typically cover the area of a football field . Several techniques are available to implant large amounts of 'highly dispersed ' metal particles on this surface. The size of these particles is measured in nanometers (1 nm= 10-9 m) or in Angstroms
(1 A= 10-10 m). In most cases, the fraction of metal· atoms in the surface is close to unity and in the majority of cases above one half. Typically, the activity in Fischer Tropsch synthesis of one gram of alumina loaded with one percent rhodium exceeds the activity of
one gram pure rhodium powder by one or two orders of magnitude, while the price of the alumina supported catalyst is lower by about
two orders of magnitude.
Very small metal particles may not be metallic : their chemical
properties, i.e. , the properties of the surf ace atoms, may deviate from the properties of the surface atoms in larger metal particles
Introduction page 4
(4-6). Another important characteristic of small metal particles is,
that they are more easily influenced than larger metal particles . A
striking example is a phenomenon discovered in 1978 (7-9) . For
metal particles supported on oxides of transition metals. such as
titania (Ti0 2), vanadia (V 20 3) and tantalum oxide (T a20 5), two
different 'states' are accessible. There is a 'normal' state, in which
the properties of the metal particles are comparable to the proper
ties of the same metal particles supported on inert oxides like
alumina (A1 20 3) and silica (Si0 2) . The surface area of such parti
cles can be estimated by measuring the amount of gas that adsorbs
on the surface of the particles. The other state is known as the
Strong Metal-Support Interaction (SMSI) state. The properties
differ markedly from the properties of the 'normal' metal particles.
Most pronounced is the decrease of the amount of gas that the
metal particles can adsorb . In either state, however, the basic
structure of the particles is the same and the metal particles can be
brought from the normal into the SMSI state and vice versa. After
reduction at low temperatures the particles are in the 'normal' state
and a subsequent reduction at high temperatures induces the SMSI
state. After oxidation at mild temperatures (up to 500 K) and a
subsequent reduction at low temperature. the 'normal' state is
restored. Catalysts that can be brought into the SMSI state have
one characteristic in common : in the temperature regime up to
800 K the support can be reduced to a suboxide and this suboxide
can be re-oxidized to the original oxide. Thus, SMSI state and pres
ence of suboxides go together. The reduction of the support is in
general catalyzed by the metal particles and is limited to the direct
environment of the metal particles. These reduced suboxides have
in general different properties than the original oxide . They may
have semi-conducting or even meta II ic properties (7-9), or they may
have an enhanced mobility ( 13-20). These special properties are
thought to be responsible for the SMSI state. Several models have
been proposed to explain the SMSI state. In the first model, already
adopted by Tauster and his co-workers (7-9), the origin for this
SMSI state is thought to be an enhanced interaction between metal
particles and supporting oxide, due to the electronic properties of
page 5 Chapter 1
the suboxide . In one of the first papers on SMSI (JO) even a direct bonding between metal atoms and Tin+ ions in the support was
suggested. The formation of alloys is another model that may explain the incapability of adsorbing gases in the SMSI state ( 1-
3.11 >. In this model it is assumed that during the reduction of the
support (for example, Ti02) metallic titanium is formed which may diffuse into the metal particles and form alloy particles. It is known that the gas adsorption capacity of these alloys is very low ( 12). The third and last model that is important in explaining the SMSI
state is the coverage model. It is assumed that reduced support species have an increased mobility and may cover the metal parti
cles, thus blocking the surface of the metal particles and decreasing their adsorption capacity. In many publications, evidence for coverage has already been reported ( 13-20). In chapters 7, 8 and 9 of this thesis, we will discuss the SMSI phenomenon and its for origin
rhodium catalysts supported on Ti02 and Ta20 5.
In the above discussion, it was assumed that pure metals were used to guide reacting molecules along their reaction pathway. This pathway can be modified by introducing additives on, in and/or
beneath the surface of the metal particles . Like the support materials mentioned above, these additives influence the properties of (the
surface atoms of) the metal particles. We can discern two classes of additives : promoters and (other) metals. The difference between the two classes can be found in their activities towards the desired process in the absence of their host metal. A promoter is,
on its own, incapable of catalyzing the desired reaction. Combined with some active host metal, however, the activity of the host may
be increased significantly. Promoters are found among alkali (Li, Na, K) ( 21-26), rare earth and some transition metal or metal
oxides (V20 3, Mo0 3, Th02) (27,28) . The other class of additives is in fact the class of active metals itself. The intention here is to
'combine' the properties of two (or more) active metals. There are cases known where the combination of two metals is 'better' than the 'sum' of the two monometallic cases, better in terms of activity and/or selectivity. For example, for the Fischer-Tropsch synthesis,
Introduction page 6
an increase in methanol and ethanol selectivities for Fe-Rh/Si02 catalysts ( 29) and an increase in ethane and propene selectivities
for Co-Rh catalysts (JO) has been reported. The activity of bime
tallics has been the subject of many studies . The major problem is to prepare metal particles that contain both metals and to verify
that indeed alloy formation has taken place. In most studies, TP R
is used to investigate .the formation of alloy particles. In preparing
bimetallic catalysts, two metal precursors are used. These precur
sors may have a different reducibility. Thus, in Temperature Programmed Reduction (T PR), they will be reduced separately. How
ever, when they are co-impregnated, already during the impregna
tion and subsequent drying step, particles containing both precur
sors may have been formed and in general, these particles will be
reduced at the the reduction temperature of the component that is reduced most easily. This component, once reduced, can adsorb
and dissociate hydrogen and thus can catalyze the reduction of the
second component. In that case, TP R may indeed point to the for
mation of bimetallic particles. Using TPR, it was found that in Cu- Ni/Si02 alloy particles were formed (3/ ,32). Evidence has also
been found for an intimate contact between the two constituent
metals in Pt-Re/ Al20 3 catalyst <33,34). Oxidation at high tempera
tures caused segregation of platinum oxide and rhenium oxide. In
(35-37), the formation of bimetallic Co-Rh catalysts supported on
A1 20 3, Ti02 and Si0 2 has been described. It was found that rho
dium aided the reduction af cobalt and that bimetallic particles were
formed. During oxidation, segregation occurred, but this did not
lead to the formation of monometallic particles during a subsequent
reduction. For Co-Rh/Ti02 (36), it was found that CoRh 20 4 was formed and that this mixed oxide was covered with Co30 4. Such a
segregation may also occur in metallic particles : the component with the lowest sublimation enthalpy will be present preferably in
the surface of the metal particles. For Co-Rh catalysts, it has been
reported that the outer shell of the alloy particles was enriched in cobalt ( 37 ). Segregation may be enhanced by the gas atmosphere;
this is known as gas induced surface enrichment. Clearly, bimetallic
catalysis is a delicate subject; the structure and composition of the
page 7 Chapter 1
alloy particles can be affected in numerous ways . In chapters 3 and
4, some aspects of bimetallic Rh-Pt/ Al 20 3, Fe-Ru/Si0 2 and FePt/Si02 catalysts will be discussed.
1.3 Characterization of Supported Catalysts
In order to control and to steer the properties of supported metal particles , it is of prime importance to know and to under
stand their structure. Once their structure has been related to the catalytic action of the catalyst, one may be able to develop 'better'
catalyst systems in a scientific manner. In determining the structure of a heterogeneous catalyst, the support is the major obstacle.
The active metal is present on the 'internal' surface, 'inside' the porous support material. The number of metal particles visible with
an ordinary light microscopy is only a small fraction of the total amount of metal particles present in the specimen. This is a major disadvantage of characterization techniques that use radiation that cannot penetrate the support. In this thesis, we will describe the
use of radiation of high enough energy to penetrate the support and thus to reach the metal particles. EXAFS uses high energy X-ray
radiation, ESR and NMR use radio- and microwaves and IVlossbauer uses gamma radiation. However, the amount of support exceeds
the amount of metal by one or two orders of magnitude. As a result, the signal-to-noise ratio and the separation of signals may become a draw back. Another way 'around the support' is to use gases which penetrate in the pores, reach the metal particles and are adsorbed on the surface of the particles (hydrogen chemisorption) or which in some way react with the metal particles (tempera
ture programmed reactions such as reduction and oxidation). In chapter 2 these techniques will be discussed in more detail.
lntrod uction page 8
1.4 Scope of this Thesis
In this thesis we will focus on determining the structure of
supported noble metal catalysts . Several interesting catalyst sys
tems have been studied. Chapter 2 describes the techniques used
to prepare and study the catalysts . In chapter 3, preparational
aspects of alumina-supported monometallic rhodium and bimetallic
rhodium-platina catalysts will be discussed. Chapter 4 deals with
the intriguing presence of ferric iron (Fe3+) in Si02 supported bime
tallic Fe-Ru and Fe-Pt catalysts, the existence of which has not
been realized for a long time. In chapter 5, the structure of an
alumina-supported rhodium catalyst during an oxidation process is
described . Chapter 6 deals with a computational approach of sup
ported rhodium catalysts . In chapters 7, 8 and 9 EXAF S studies of
catalysts that suffer from metal-support interactions will be
highlighted : Rh/Ti0 2 and Rh/Ta20 5. In chapter 10, the results are
discussed in a wider context and where possible interrelated .
35. van ·t Blik, H. F. J .; Prins , R. J . Catal . 1986, 97, 188
36. Martens , J . H. A. ; van ·t Blik, H. F. J .; Prins , R. J . Catal . 1986, 97,
200
37. van 't Blik, H. F. J .; Koningsberger , D. C. ; Prins, R. J. Catal . 1986, 97, 210
page 11 Chapter 2
Chapter 2
Experimental Techniques
2.1 Catalyst Preparation
2.1.1 Introduction
For the properties such as activity, selectivity and stability of the eventual catalyst, the method of preparation is of crucial impor
tance . Prime consideration- always is to keep the size of the metal particles within acceptable ranges . The smaller the metal particles,
the more active sites per gram of metal. Several techniques are known to bring forth small metal particles . The choice of the
preparation method depends on the metal and the support to be used . We will describe three preparation methods which were used
to prepare the catalysts that will be discussed in this thesis . The active material is in our case always a metal. It is impossible to
directly dispers a metal homogeneously on a porous support. In most cases and in all cases discussed in this thesis, a precursor,
usually a metal salt, is dissolved in a solute. This. solution can penetrate into the pores of the support and enable the metal precur
sor to be deposited on the internal surface area of the support. This is the part of the preparation in which the three methods differ. In
paragraphs 2.1.2 , 2.1.3 and 2.1.4 we will discuss this deposition of metal precursor for the three methods separately.
After fixing the metal precursor onto the support, the solvent is removed by filtering and drying and the precursor remains in the pores of the support. Sometimes not only the metal salt but in
addition some residue originating from the solvent stays behind as well. To discard of this, a calcination step may be introduced, i.e.,
Experimental page 12
the precursor catalyst is heated in air to a few hundred degree cen
tigrade. After calcination or drying, the metal precursor is brought
into the active. metallic state by a reduction in hydrogen . After
reduction, the catalyst is highly active and exposing the catalyst without further precaution to air would result in a process known as
'run away oxidation'. Already during the start of the oxidation pro
cess enough heat would be evolved to allow the oxidation process
to continue in an uncontrollable way. The temperature of the parti
cles would reach a level at which the mobility of the particles is
high enough to start a sintering process, in which several smaller
metal particles join to form larger metal particles. This process is
of course to be avoided. Therefore, direct after reduction , the active
catalyst is flushed with nitrogen in order to remove all hydrogen.
Thereafter , oxygen is added carefully to the nitrogen feed . The
oxygen content is increased slowly up to a level of 20%. This pro
cess is known as passivation : a careful and controlled oxidation of
the metal particles. For noble metal particles, passivation is in gen
eral limited to the surface of the metal particles . The active metal is then covered by an oxide layer, is made 'passive' and can be stored
in air to await further study. A reduction in hydrogen at low tem
peratures is in general enough to restore the catalyst to its active
state.
In the discussion above, the first and most important part of
the preparation, the introduction of the metal precursor onto the internal surface area of the support, has been skipped. We will now
continue to describe this essential part for three methods . The methods are the pore volume impregnation method, the ion
exchange method and the urea method.
page 13 Chapter 2
2.1.2 Pore Volume Impregnation
The pore volume impregnation method is the method which is most frequently used in catalyst preparation because of the elegant simplicity of the method. The quantity of metal salt needed to prepare the catalyst is dissolved in the exact amount of solute needed to fill the pores of the support. We illustrate this by the following example, in which we prepare 5 g of an alumina supported catalyst loaded with 2 wt% rhodium. The alumina used has an internal surf ace area of 180 m2 g-1 and a pore volume of 0.6 ml g- 1
.
The precursor is RhCl3·3H 20 with a molecular weight of 263.3 g. 5 g of the eventual catalyst will contain 4.9 g A1 20 3 and 0.1 g Rh, which is equivalent to 0.2559 g RhCl 3·3H 20 and has to be dissolved in 4.9*0.6 = 2.94 ml water. This solution is added slowly to the alumina, which is stirred vigorously in order to distribute the dissolved precursor evenly over the support. Because of capillary forces, the solution is soaked up quickly in the pores of the support. It is important that the pores are just filled. When too little solvent is used, the precursor is spread only on part of the surface area of the support, and this may result in larger metal particles. When too much solvent is used, part of the precursor will end up on the relatively small external surface area of the support, which may result in a few but very large metal particles outside the pores. Because the pores have to be filled precisely, th is method is also known as the incipient wetness technique : when the pores are.just filled, on the verge of 'flowing over', the support starts to feel wet. After impregnating or incipiently wetting the support, the catalyst precursor is, as described above, dried carefully, calcined if necessary,. reduced in hydrogen and finally passivated.
Experimental page 14
2.1.3 Jon Exchange
A technique more sophisticated than the pore volume impregnation method is the ion exchange technique. In this method, cations are fixed to the internal surface area of the support. An example is the ammonia ion NH 4 + which absorbs readily on supports like titania, Ti02• An ammonium solution (NH 40H) is added to the support and allowed to adsorb. After the adsorption process, the support is filtered off. This support, saturated with ammonia, is added to a solution of a metal salt, e. g, an aqueous solution of Rh ( N 0 3h It is essential that the metal ions in the solution are present as cations . In the case of Rh(N0 3b, rhodium is present as IRh(OH)n(H 20)6.nJ( 3·n)+ complexes. These (positively charged) complexes exchange readily with the absorbed N H4 + ions on the support. In this way, an equal spread of the metal precursor over the support can be achieved. If RhCl3 were used as metal precursor, rhodium would be present as (H 30+ +) [RhCl3(0H)n(H 20) 3_nJncomplexes and these negatively charged complexes will not exchange with the N H4 + ions on the support.
In this thesis, we will encounter an example in which no specific counter ion has been adsorbed on the support to exchange with a metal complex. The 4 wt% Rh/Ti02 catalyst described in chapter 7 has been prepared by exchanging Rh(N03h with the protons present in the surface hydroxyl groups in Ti0 2; the Rh/Ti02 catalyst described in chapter 8 has been prepared by ion exchange
. . using ammonia.
After the metal precursor has been exchanged, the support is filtered off and dried. In case ammonia has been used, a calcination step is usually applied to remove the ammonia left behind on the support. The dried or calcined catalyst is then reduced in hydrogen and finally passivated.
page 15 Chapter 2
2.1.4 The Urea Method
The urea method is founded on a controlled raise of the pH of
a solution of the metal precursor and urea (CO(NH 2)i) in which the
support is suspended ( 1,2). The solution is stirred vigorously. An
adequate temperature is selected ( ± 365 K) at which the urea
decomposes slowly according to the following reaction :
As a result, the pH value of the solution slowly increases as the urea decomposes. At a certain pH value a metal hydroxide starts
to form and precipitate on the support. Because the urea decompo
sition rate and thus the liberation of OW groups can be controlled
by regulating the temperature, the precipitation of the metal precur
sor can be controlled elegantly. As a result, the method provides a
homogeneous spread of metal hydroxide over the support, which is
a good starting point for obtaining highly dispersed metal particles .
After drying, a calcination step is essential in order to remove the
urea left behind in the sample. Finally , calcination is followed by
reduction and passivation.
2.2 Temperature Programmed Reduction
Temperature programmed reactions are used frequently to
study the chemical behavior of supported metal catalysts. Most
important among them is Temperature Programmed Reduction
(TP R). Other examples are Temperature Programmed Desorption
(TPD), Oxidation (TPO) and Sulfiding (TPS). The principles are
the same for all temperature programmed reaction techniques. As
the temperature of the catalyst is increased, some reaction of the
Experimental page 16
active phase with the gas atmosphere is studied. In TPR. the redu
cibility of a sample , e. g, an oxidized metal catalyst or a catalyst
precursor, takes place. The catalyst is flushed with a mixture of
4% H2 in N2 and the temperature of the sample is increased at a
constant heating rate of 5 K min-1. The following reaction schemes
illustrate the reduction processes that may proceed in case of a pre
cursor catalyst which has been prepared using RhCl 3, or in case of
an oxidized rhodium catalyst :
2 RhCl 3 + 3/2 H2 +::± 2 Rh + 3 HCI
Rh20 3 + 3 H2 +::± 2 Rh + 3 H20
By monitoring the consumption of hydrogen, one can get informa
tion on the reduction process. The hydrogen uptake as a function
of temperature is usually denoted as the 'TPR profile ' of the sam
ple . The temperatures at which the reduction proceeds reveal what
substance is being reduced and can be used as a 'fingerprint' . The
amount of hydrogen consumed provides the reaction stoichiometry
and/or the degree of reduction at a certain temperature during the
process.
Just as the other temperature programmed techniques, TP R is
used most often as a fingerprint technique. For detailed descrip
tions, we refer to the review by Hurst (3) and a few of the earliest
papers on TPR (4-11 ). The apparatus we used has been described
extensively by Boer et al. ( 12).
page 17 Chapter 2
2.3 Hydrogen Chemisorption
In supported metal catalysis , the amount of metal atoms exposed to the gas atmosphere is one of the most important characteristics of the catalyst. The dispersion of a catalyst is used to quantify this and is defined as the fraction of metal atoms in contact with the gas atmosphere . Adsorption and desorption techniques can be used to estimate the dispersion. These techniques use a selected gas that adsorbs only on the metal particles and not on the support. Dependent on the specific technique used, the amount of gas that adsorbs or desorbs is measured. The hydrogen chemisorption technique used to measure the dispersion of the catalysts discussed in this thesis is extensively described in ( 13,14).
Briefly, the experimental procedure comprised the following steps. A catalyst sample, dried , calcined, passivated or oxidized, is reduced
in situ in 100% H2 at the desired temperature. After the reduction procedure. the sample is evacuated at some elevated temperature, in general 473 K. The chemisorption cell contains two sections, of which the exact volume is known. The first compartment is used as a reference chamber, the second contains the (now reduced and evacuated) catalyst sample. A stop cock separates the two sections . After reduction and evacuation , a known amount of hydrogen is admitted into the reference chamber. After the valve between the reference and catalyst section is openeq . hydrogen starts to adsorb on the catalyst . In most cases, adsorption is an activated and therefore slow process. To circumvent this, the catalyst is temporarily heated , usually to 473 K, to speed up the adsorption process. After cooling down and equilibrating, the amount of gaseous hydrogen can be computed by measuring the
pressure and the amount of adsorbed hydrogen can be calculated. The experiment is then continued in the desorption mode. The stop cock between the two chambers is closed and the hydrogen pressure in the reference chamber is lowered. After opening the stop cock, hydrogen desorbs from the catalyst because of the lower pressure, and the amount can again be computed by measuring the
Experimental page 18
equilibrium pressure. Thus. the ratio of the amount of adsorbed hydrogen and the amount of metal present. the H/M value. is monitored as a function of pressure. The linear part of this desorption isotherm is extrapolated to zero pressure in order to nullify small errors in the volume of the catalyst section and to eliminate adsorption of hydrogen on the support.
As indicated in ( /4), this H/M value is not a direct measure for the dispersion. but can be used to compare catalysts and is therefore very useful as a fingerprint. However. since in the same paper the hydrogen chemisorption method has been calibrated with an independent technique. EXAFS. we can estimate very accurately. dispersion and particle sizes from the H/M values .
2.4 Electron Spin Resonance Spectroscopy
In case a catalyst contains paramagnetic centers. Electron Spin Resonance is a useful technique to study these centers. For an isolated electron. a paramagnetic center. two states are accessible, one with spin +1/z (O'), the other with spin -1/z (y) . In the absence of a magnetic field these two states are degenerated : E( O') = E( y). A magnetic field removes this degeneracy. the two states differ in energy by
!iE = £(0') - E(y) - gel3H 12.4.1]
in which 13 is the Bohr rnagneton and H is the magnitude of the magnetic field . The electron g-factor, ge, for an isolated electron in a spin-only case. is equal to 2.0023. Transitions between the two levels O' and 13 can be generated by a suitable electromagnetic radiation. In practice. the frequency of the electromagnetic radiation is kept constant while the magnetic field is varied linearly with time.
page 19 Chapter 2
When the photon energy h ii equals the difference in energy of the
two states, transitions between ex and {3 may occur. This 1s
accompanied by absorption of the electromagnetic radiation.
In practice, electrons are never isolated and we have to expand
the theory to perturbed unpaired electrons. In case of a free elec
tron, the electron spin is solely responsible for the electrons mag
netic moment and the Hamiltonian H can be written as
H gf3(H·S) [2.4.2]
In relevant cases, however, the electron 'moves' in an orbit 'around' a nucleus and this gives rise to an orbital angular momentum cou
pling (f3 (H ·L )) and a spin-orbit coupling (.\ (L ·S)). Conse
quently, the Hamiltonian can be written as
H = g{3(H·S) + f3(H·L) + .\(L·S) [2.4.3]
The complication induced by the latter two phenomena can be cir
cumvented be defining an effective spin Hamiltonian Herr which
operates only on fictive spin states :
Herr = f3 (H ·g err"S:) [2.4.4]
The effects therefore manifest themselves in the value of the
effective g-tensor g err· Measuring the effective g-values therefore
provides information on the magnitude of the orbital angular
momentum and the spin-orbit coupling. Two important properties
of the g-tensor need to be mentioned. The electron spin is coupled
to its orbital momentum and therefore to the lattice in which the
atom or ion is situated. Hence, relaxation phenomena are related to the effective g-value. Enhanced relaxations may be accompanied by
large deviations of g err from 2.0023. The second property to
Experimental page 20
mention 1s the symmetry of the g-factor. A crystal field with spherical symmetry gives rise to an isotropic g-value. For an axial field along the z-axis. two g-values, gxx = gYY and g
11• may be
observed.
In Chapter 4, an ESR study of Fe3+ ions will be described. Fe3+ has a 3d5 configuration and in a high spin case in an octahedral crystal field. Fe3+ has five unpaired electrons. Whenever the site symmetry deviates slightly from perfect octahedral. which is usually the case, this gives rise to a very characteristic ESR signal centered at g = 4.2 ( 15-17).
2.5 Nuclear Magnetic Resonance Spectroscopy
For Nuclear Magnetic Resonance, the basic principles are the same as those for ESR. While in ESR transitions between electron spin states are observed, in NM R transitions between nuclear spin states are studied. For nuclei with I= +1/z in a magnetic field, again two states are accessible : one with the nucleus' magnetic moment parallel and the other anti-parallel to the external magnetic field. When irradiated with a suitable electromagnetic radiation, transitions between those states can be generated and an absorption of the electromagnetic radiation can be observed. As . for ESR, the frequency of the electromagnetic radiation is kept constant and the absorbance is monitored as a function of the external magnetic field. The magnetic field which the nucleus experiences is in general not equal to the applied magnetic field : the electrons surrounding the nucleus modify the external magnetic field. Thus, the shift in NMR spectra gives information on the chemical environment of the nucleus under study and is therefore called the chemical shift. For example, the chemical shift for Pt in H2PtCl6 is different from the chemical shift of Pt in Na2Pt(OH) 6 ( 18,19) and can therefore, although both salts are present in the same sample, be used to
page 21 Chapter 2
study these salts separately. In chapter 3 we will encounter an
example in which 195 Pt NM R is used to estimate the amount of Pt
present in H2PtCl6 crystallites in an impregnated and dried catalyst
sample.
2.6 Mossbauer Spectroscopy
In 1957, Rudolf L. Mossbauer demonstrated that nuclei can
resonantly absorb gamma rays which originate from similar nuclei
decaying from excited states ( 20-22). The basic principles are
explained in Figure 2.1a. In this example, 57Co is used as a source
and decays according to the scheme in Figure 2.1. The 14.4 keV
emission can be used to generate transitions in 57 Fe nuclei in the
sample between I = 1h and I = %. In practice, the energy of the
emitted gamma quanta differs slightly from the energy needed to
excite the 57 Fe nucleus under study. To circumvent this, the source
is give a velocity and because of the Doppler effect, the emitted
energy is modulated slightly. The absorption therefore, is measured
as a function of the Doppler velocity of the source and peak positions and shifts are reported in mm s-1
. This is ii lust rated In
Figure 2.1b. There are three hyperfine interactions which are respon
sible for the fact that the 57 Fe nuclei in the sample absorbs quanta
of a different energy than those emitted by the excited 57 Fe nuclei
in the source . The first is the isomer shift (J.S.) which is a meas
ure for the electron density around the nuclei under study . The iso
mer shift is caused by the Coulomb interaction between the posi
tively charged nucleus and the negatively charged s-electrons,
whose wave functions overlap with the nucleus. The isomer shift
gives information on the oxidation state of the iron . The second
hyperfine interaction is the quadrupole splitting. In its ground state,
the 57 Fe nucleus has a spherical charge distribution and therefore no quadrupole moment. In the excited state, however, the nucleus has
an ellipsoidally shaped charge distribution and therefore has a
Experimental page 22
Figure 2.1 Mossbauer Spectroscopy
(a) Basic Principles
(b) Experimental set up
(c) Three basic Mossbauer spectra : (1) no hyperfine interactions (the spectrum of stainless steel), (2) the influence of quadrupole splitting (the spectrum of sodium nirtoprusside) and (3) magnetic hyperfine splitting (the spectrum af a-Fe) .
57Co
~lectron b CJ \apture
.. source sample detector
9% 91%
137 keV 123 keV
- - -'.----- I = 3/ 2
--~-I=1/2 (14.4 keV---
____L _ _ __,__
Doppler velocity I I I I I I I I I I I I I I I I I I I I I
-10 -5 0 5 10 c
~ (1)
w (21
(3)
page 23 Chapter 2
positive electric quadrupole moment . In case the nucleus experiences an electric field gradient , two orientations are allowed for the quadrupole moment and a splitting of the excited level is observed (see Figure 2.1c) . The splitting between the two excited levels t::i..£ 0
is proportional to the electric field gradient at the nucleus ; The third interaction is a magnetic hyperfine splitting. The magnetic moment of the excited nucleus will react on any external magnetic field, including the magnetic field induced by the surrounding electrons . Two orientations are accessible for the ground state and four for the excited level. Thus, a splitting of the ground level and the excited level is observed (see Figure 2.1c) . Therefore , for a nucleus with a magnetic moment , eight transitions are possible , two of which are forbidden, leaving six transitions.
Thus , isomer shift , quadrupole splitting and magnetic hyperfine splitting are used to identify the environment of 57Fe nuclei in a sample . Mossbauer spectroscopy can of course be used for other elements as well. Examples are Ir and Ru. In chapter 4 we will discuss an example in which the state of iron in Si02 supported bimetallic Fe-Ru and Fe-Pt catalysts is studied using Mossbauer spectroscopy.
2. 7 Laser Raman Spectroscopy
In Raman spectroscopy , the em1ss1on spectra of excited molecules are studied. A laser is used as a light source. The electric field E of light will give rise to a redistribution of the charge in a
polarizable molecule and thus induce a dipole, the dipole moment 71 being equal to
[2.7 .1]
Experimental page 24
in which O' is the polarizability of the molecule. For light with frequency v 0• the dipole moment is equal to
µ - 0'£0sin(21Tv0 t) [2.7.2)
When during the absorption process the excited molecule decays to a state different from the ground state, a state with an internal vibration with frequency vi, the polarizabil ity O' will oscillate accordingly :
[2 .7.3)
and therefore
µ [O'o + 13 sin(217' vi t ) )Eosin(217' v0 t ) [2.7.4)
- O'oEosin(21Tvo t)
+ 1hf3E0 ,cos(21T(vo-vi)t) - cos(21T(v0+vi)t)]
The excited molecule will therefore emit radiation with frequencies equal to v0 (Rayleigh scattering), v0 - vi and v0 + vi, the Stokes and Anti-Stokes lines respectively. The frequency shifts in Raman therefore correspond directly to the frequency of the induced vibration. As in infrared (IR) spectroscopy, these frequencies are very specific for the molecule under study. In chapter 3 we will discuss the Raman spectra of the PtCl6
2- unit. The basic unit of this
molecule is an octahedron and three of its vibrations are Raman active. Figure 2.2 shows the PtCl6
2- unit. For centro-symmetric
systems, the vibrations that are inactive in Raman spectroscopy, are active in IR and vice versa. However , IR turned out to be incapable of_ detecting any vibrations in the PtCl6
2- unit of impregnated
and dried Pt/ A1 20 3 catalysts. Details of the Raman spectroscopy experiments wil I be given in chapter 3.
page 25 Chapter 2
Figure 2.2 The structure of H2PtCl 6
Cl Cl I / Cl I / Cl
Cl-Pt-Cl ········ ·· Cl--Pt-Cl /I /I Cl .Cl Cl Cl
I / er I / er c1 Cl-Pt-Cl ···········C l--Pt - -Cl
/ I. / I Cl Cl Cl Cl I / Cl I / Cl
- Pt-· · C 1--Pt---C 1 / I ,, I Cl Cl Cl ,,
I/Cl I/Cl Cl C 1----Pt---C 1 · ··· C 1- Pt-C l
c1/ I c1/ I Cl Cl
2.8 ASED-MO Computations
ASED is an acronym for atom superposition electron delocalization. The ASED-Molecular Orbital theory (23) has been applied in
numerous and diverse studies to predict structures, reaction mechanisms and vibrational and electronic properties . The theory
uses for input data the ionization potentials and valence state Slater orbital exponents for the constituent atoms (24-28) . These parame
ters are sometimes altered, particularly in treatments of ionic hetero nuclear molecules, to ensure reasonably accurate calculations. The
electronic charge density function of a molecule or solid can be partitioned into components in any number of arbitrary ways. The
ASED-MO theory is based on a partitioning of the density function
Experimental page 26
into free atomic components and the rest. The atomic components
are spherically symmetric in a field-free space and are centered on
the nuclei . They follow the nuclei 'perfectly' . The remainder
changes its shape depending on the geometry. With respect to the nuclei. it is 'non-perfectly following' . Figure 2.3a shows an example
of perfectly following atomic densities Pa and Pb for atoms a and b
in a diatomic molecule and a nonperfectly following charge density
Pnpf . The interaction energy is made up of a repulsive part which
can be computed from the perfectly following charge densities and an attractive part, which can be calculated from the nonperfectly following charge density. In Figure 2.3b both repulsive and attrac
tive energy terms and their sum, the interaction energy are shown schematically . The first assumption in the ASED method is, that the atoms are first instantaneously superimposed. Based on this, the repulsive energy term follows from
[2.8.1]
in which Z is the nuclear charge, p the atomic charge density function, R the coordinate of the nuclei and r the coordinate of the elec
trons. E R is repulsive because nuclear repulsion energy is greater
than the attractive energy between nucleus b and Pa. The nonperfectly following energy term is attractive because of the concen
tration of charge in the internuclear region due to bonding. The attractive energy term is given by
z Rfb r (. R . ) d J i - - b oo J Pnpf
1 • b dR; I R;- r I drdR; [2.8.2]
It is impossible to evaluate equation [2.8.2] . However , the nonperfectly following or attractive energy term is due to electron delo
calization and is roughly equal to the difference in atomic and
page 27 Chapter 2
Figure 2.3 ASED-MO terms for a diatomic molecule .
(a) The perfectly and non perfectly following charge densities for a diatomic molecule AB
(b) The attractive energy term. the repulsion term and the interaction energy as a function of interatomic distance.
a b
/
/
R a-b
Etot E
npf
ER
molecular orbital energies. Thus, E att is successfully approximated by
!).£mo
'° n Eab £., I I
[2.8.3)
which is a summation over the molecular orbitals i; n; is the orbital occupation number (0, 1 or 2), Eia and Eib are the atomic orbital
energies (in practice, VSIP) and Eiab the molecular orbital energies. The molecular orbital energies are the solutions of the (extended Hi.ickel) Hamiltonian with
in which S;'1/' is the overlap integral of a and b. R the internuclear
distance. Finally. the total energy is the sum of ER and E att :
Etot - L, ER(a .b) + Eau a > b
[2.8.7]
For a detailed discussion, we refer to ( 23 ,29 ). In chapter 6 we will describe the results of this kind of calculation for 10 atom rho
dium metal clusters supported on y-Al 20 3. From these calcula
tions, we will be able to derive information about the binding of
rhodium metal particles to the alumina support, the binding energy
and the structure of the metal- support interface.
2.9 Extended X-ray Absorption Fine Structure
2.9.1 Basic Principles
EXAFS is an acronym for Extended X-ray Absorption Fine
Structure. It refers to the 'wiggles', the fine structure that can be
observed in the X-ray absorption spectra of condensed phases at
the high energy side of absorption edges. Let us focus on a 1s elec
tron which is subjected to monochromatic X-ray radiation. When
the photon energy tiw is lower than the binding Eb energy of the 1s
electron, absorption will not take place. When the photon energy
equals the binding energy. the 1s electron may be excited and this
page 29 Chapter 2
gives rise to a sharp increase in the x-ray absorption. The probabil
ity of exciting the electron is given by
[2.9.1]
in which l/Ji and l/Ji are the electron initial and final state. When we
assume that the photon has its electric field polarization in the zdirection, a simple deduction ( 30) leads to the formula for the
attenuation coefficient µ. :
µ. - [2.9.2]
in which w is the frequency of the photon, c the speed of light, Na
Avogadro's number and p(E t) the density of final states.
At photon energies above the electron binding energy, the transition probability and therefore the attenuation coefficient slowly
decrease with increasing photon energy. A typical example is the case of a monoatomic gas such as krypton. Figure 2.4a gives an
ideal absorption spectrum of such an unperturbed transition .
In case the excited atom is surrounded by neighboring atoms, the absorption changes . The neighboring atoms have a pronounced influence on the final state and therefore on the attenuation
coefficient. Because of the neighboring atoms, some final states are favored and others are disfavored with respect to the unperturbed case where no nearest neighbors are present. The result of this is illustrated in Figure 2.4b : at the high energy side of the adsorption
edge of rhodium foil the attenuation coefficient µ. shows an oscillatory behavior which damps out slowly .
Experimental page 30
Figure · 2.4 The influence of neighboring atoms on the X-Ray absorption spectrum
(a) An unperturbed adsorption spectrum
(b) The X-Ray absorption spectrum of the K-edge of rhodium foil
1
E (e V) 24000
E (e V) 25000
With scattering theories, we can describe the final state in an understandable way and derive a mathematical expression for the fine structure in the absorption. In our model, the outgoing part of
the wave function l/Jf will scatter from the neighboring atoms. The scattered state will interfere with the outgoing state and depending on the relative phases of both states, the final state, being the sum of outgoing and scattered state, is amplified in case both states are in phase and will be attenuated in case both states are out of phase. The electron wavelength in the outgoing and scattered state is given by
h (2.9.3)
Both outgoing and scattered state are in phase when
page 31
R n A. 2
Chapter 2
[2 .9.4]
in which R is the interatomic distance between absorbing and
scattering atom. Thus, when the photon energy equals
E - hw [2.9.5)
both outgoing and scattered state will be in phase. the final state
will be augmented and absorption will increase relative to the unper
turbed state. When we convert the energy scale to a wave vector (k) scale using
k 2.J'i.TTm ,/tiw-£. h I
(2.9.6]
we will find maxima in the attenuation coefficient at
k n 1T
R (2.9.7}
The maxima, and therefore the minima and the nodes in µ will occur at constant k-intervals of fj,k = TT/ R. Therefore, EXAFS
spectra are always represented in k-space rather than in energy space.
We will not further explore the physics of the EXAFS phenomenon. In the above discussion the basic principles are
explained sufficiently to give a good insight into the spectra which
will be discussed in the next chapters. A rigorous derivation of a
more complete EXAFS formula includes phase shift functions, back
scattering amplitude, the lifetime of the final state, the mean free path of the electron, disorder and finally, corrections can be made
Experimental page 32
for approximations induced by the small atom method used and for
multiple scattering phenomena ( 30). Such a derivation is clearly
out of place in this thesis and we will therefore conclude this intro
duction by discussing a general formula for the EXAFS function
x (k) :
x(k) = L,Aj(k)sin(2kRj + ct>j(k)) J
(2.9.8]
The amplitude function Aj (k) is given by :
Clearly, X(k) is a sine function, modulated by an amplitude func
tion. In the argument of the sine function we find the interatomic distance R . between absorbing atom and scattering atom j and a
J phase shift function ct> j (k ). The amplitude function Aj (k) is
governed by the number of neighbors in shell j (Nj ), their distance
with respect to the absorbing atom (Rj ), a backscattering function
Fj (k), SJ which corrects for relaxations in the absorbing atom
and multi-electron excitations, and two exponential terms which account for disorder (the term with o 2) and for the main free path
of the electron (>-e ).
The backscattering amplitude F (k) indicates 'how well' the neighboring atom performs in scattering the excited electron. F (k ) is therefore element specific for the scattering atom. In general, heavier atoms scatter more and therefore have a higher backscatter
ing amplitude. The backscattering amplitude is also a function of the wave vector k and F (k ) for high-Z elements may have a very . characteristic k-dependence . In Figure 2.5a, F (k) for 0, Ti, Rh and Ta are given. The data have been taken form (]J ).
page 33 Chapter 2
Figure 2.5 The k-dependence of F (k ) and cf> (k )
(a) The backscattering amplitude for 0. Ti . Rh and Ta
(b) The backscattering phase shift function for 0. Ti. Rh and Ta
- Ta ··· ··· Rh ---- Ti -0
a b 12
. I' 9 ~ :, \
. . ... . .. . . ·.· . . 6
0.5 -----3
0 ------------------------ ---3
5 10 15 5 10 15
k r,J-11 k [ ,J -11
Just as F (k ) , the phase shift function cf> (k ) is element
specific. However, in contrast to F (k ) . cf> (k ) is a function of both absorbing and scattering atom. The contributions of absorbing and
scattering atom to the total phase shift are independent <31-33) :
[2.9.10)
cf>~ is the contribution of the absorbing atom A and cb~ the contri
bution of the scattering atom B to the total phase shift function :
The constant 8 is equal to unity for K and L1 edges and is zero for L11 and L111 edges. In Figure 2.5b, the phase shift functions for the
absorber-scatterer pairs Rh-0, Rh-Ti, Rh-Rh and Rh- Ta are given.
As we will see in the next sections, F (k) and c/>(k) perform a unique role in identifying the neighboring atom . And by determining the interatomic radii and the coordination numbers, EXAFS 1s
clearly a very powerful technique in analyzing local structures.
Experimental page 34
2.9.2 Fourier Transformations
In general, more than one shell of neighbors is present around
an atom and this makes the analysis of the EXAFS spectrum more
complicated . Therefore, Fourier transforms are used frequently to
separate the contributions of these neighboring shells of atoms . The mathematical formulation for a Fourier transformation is
[2.9.11]
A Fourier transform is no more than a mathematical tool
which 'sorts' the information in a frequency spectrum in a more
convenient way. The result of the Fourier transform given above is
a radial distribution function en (r), a function that has maxima at
radii that are related to the coordination distances R; in the original
function. In Figure 2.6a the EXAFS function of rhodium foil and its
Fourier transform are shown. In the Fourier transform, the first and
higher order shells are clearly visible. In case a peak in a Fourier
transform is clearly separated from neighboring contributions (like the first and main peak in the Fourier transform in Figure 2.6b), an
inverse Fourier transform can be used to extract a single shell EXAFS function. In case two or more peaks are included in the
back-transformation range, a two or more shell EXAFS function 1s
obtained. An inverse Fourier transform is given by
rmax
x· (k) - 1 J e (r )e - 2ikr dr kn .J2.7r . . n
r1nm
[2.9.12)
In Figure 2.6c, the result of an inverse Fourier transform over an ,._ range from 1.2 to 3.3 A, i.e. the first shell, is given.
page 35 Chapter 2
Figure 2.6 EXAFS functions and Fourier transforms
(a) The EXAFS function of rhodium foil
(b) k1-weighted Fourier transform of X (k ) of rhodium foil
(c) The inverse Fourier transform of the k1-weighted Fourier transform of X (k ) of rhodium foil
(d) k1-weighted Fourier transform of X (k) of rhodium foil, corrected for phase and amplitude of Rh.
Backscattering amplitude and phase shift functions can com
plicate Fourier transforms. In Figure 2.6b, a k 1-weighted Fourier
transform of the EXAFS function of rhodium foil is given. The first peak is clearly asymmetric, for the magnitude as well for the ima
ginary part of the Fourier transforms. This is caused by a k
dependence in both functions. Usually, the phase shift function has
Experimental page 36
a linear k-dependence like
<t> (k ) - <t>o + <J>k [2.9.13]
The argument of the sine in the function can then be written as
2kR + <i>(k) - 2kR + <i>o + <i>k
- 2k (R + ~ ) + <i>o
[2.9.14]
and consequently, the Fourier transform 'peaks' at a distance of
R + <t>/2. However, when the phase shift function is known, we
can correct for this by multiplying x (k) with e- i ¢(k I prior to the
Fourier transformation. The effect of the backscattering amplitude is only of importance when F (k) shows one or more extrema .
That is the case for high-Z scatterers . In that case, the EXAFS function is 'modulated' by a more or less periodic function, and the
corresponding peak in the Fourier tran~form will be accompanied by
one or more sidelobes. When F (k) is known , we can circumvent this by dividing x (k) by F (k) prior to the Fourier transformation . Thus, a phase and backscattering amplitude corrected Fourier
transform is given by the following equation :
e~(r) k "'°' -i¢(k)
_1_Jx(k)e kne2ikrdr J21T k . F (k)
mm
[2.9.15]
In Figure 2.6d, the result of a phase and backscattering amplitude corrected and k 1-weighted Fourier transform of the EXAFS function of rhodium foil is given. Clearly, the first peak is now centered at
the correct Rh-Rh distance (2.687 A) and has no sidelobe any more . Corrected Fourier transforms look less complicated than uncorrected
transforms and are therefore used whenever possible in this thesis .
page 37 Chapter 2
2.9.3 Reference Compounds and Calculating Spectra
The mainstay of analyzing experimental data is to calculate an EXAFS functions that resembles the measured function as accurate as possible. In addition to the parameters N, R, a 2, we need to know F (k ) and <t> (k ) to calculate an EXAF S spectrum. Both these function can be extracted from the EXAFS function of a suitable reference compound. For example, for a Rh-Rh EXAFS function, we extract FRh(k) and <t>Rh-Rh(k) from the EXAFS spectrum of rhodium foil and for a Rh-0 EXAFS function, we extract F 0 (k) and <t>Rh-o{k) from the EXAFS spectrum of Rh 20 3. The procedure is in fact very simple but very sensitive towards the choice of Fourier transform ranges and should therefore be carried out with great care. Assuming we have correctly extracted x (k ) from the raw data (see the next paragraph), we perform a Fourier transform over the largest possible range in k-space, selecting krnin and krnax in nodes of the EXAFS function to avoid cut-off effects. An inverse Fourier transform containing only but completely the desired shell yields the EXAFS function from which the amplitude function Aj (k) and the argument of the sine function (2kR + <t>(k )) can be extracted. Using crystallographic data (N and R) we can derive the phase shift function and a normalized amplitude function
F. (k) A (k) kR 2 N
[2.9.16]
This amplitude function has the same general behavior as the backscattering amplitude, but in addition contains the disorder term, the main free path term and SJ (k). We will assume that the latter two terms will be the same for the samples under study and the reference compound. We will account for disorder by introduc
ing the term e- 2k
2t:.<Y
2 in which tia 2 is the Debye-Waller factor, the
Experimental page 38
disorder in the sample relative to the disorder in the reference com
pound. Thus, an EXAFS function can be calculated using
x (k)
in which F • (k ) and <b (k ) are the experimentally determined backscattering amplitude and phase shift functions.
2.9.4 Data Analysis
Analyzing EXAFS spectra is a very delicate procedure. An excellent review has been published by Sayers in ( 34). The reliabil
ity of the resulting structural parameters depends up to a high
degree on the accuracy with which data reduction and data analysis have been carried out. There are many pitfalls in the data reduction
and data analysis procedure. For example, subtracting a background which has not been optimized thoroughly will result in incorrect coordination numbers and Debye Waller factors and possi
bly in wrong coordination distances. Even a small glitch or a jump
in a spectrum will induce effects in a Fourier transform and thus, may give rise to artificial peaks in the radial distribution function.
Before the actual procedure of analyzing the data, a data
reduction procedure is needed to extract the EXAFS function from
the experimentally determined X-ray absorption spectrum. Dependent on the software which controlled the experiments at the sta
tion where the measurements have been preformed, the monochromator position may have to be converted to an energy scale and the
read out of the two ion chambers may have to be converted to absorbance. In Figure 2.7, as an example, the X-ray absorption
spectrum is given for the 4 wt% Rh/Ti0 2 sample to be discussed in
page 39 Chapter 2
Figure 2.7 The Victoreen curve (dashed line) and the background (dot
ted line) used to extract X (k) from the experimental data
chapter 7. From this spectrum a Victoreen curve is subtracted
(the dashed line in Figure 2.7). After the Victoreen curve has been subtracted, the inflection point of the edge is chosen to represent the actual edge position. After a Victoreen curve has been sub
tracted, the gradient in the data is small. Therefore, at this stage any glitches or jumps have to be removed carefully from the spectrum. During the next step, a smooth background is constructed to fit loosely the data at the high energy side of the edge. This background is supposed to represent the unperturbed absorption spectrum, i. e., the spectrum in case the absorbing atom is not sur
rounded by neighboring atoms. The dotted line in Figure 2.7 represents the background. This background is subtracted from the
experimental data to give a preliminary EXAFS function. The background is constructed using a cubic spline routine. The parameters
are chosen such that the background does not contain any EXAFS
oscillations and that the resulting preliminary EXAFS function contains 'as little background' as possible (34). A Fourier transform of the resulting spectrum and the derivative of the background are
used to optimize the cubic spline parameters. Slow background oscillations present in the EXAFS function manifest themselves in a Fourier transform as peaks below 1 or 1.5 A. In case EXAFS oscillations are present in the background, the main peak(s) in the
Experimental page 40
Fourier transforms will have decreased in intensity and these oscil
lations can be observed in the derivate of the background . Clearly.
in the ideal case the contribution in the Fourier transform below 1 A. is negligible and the major peaks still have their maximum attainable intensity .
After the background is subtracted. the preliminary EXAFS
function is normalized with respect to the height of the edge and
has therefore a per-atom dimension . This allows a quantitative
analysis of the spectrum. The data reduction procedure is finally
concluded by converting the energy scale to a wave vector (k) scale
using equation (2.9.6) and taking Ei equal to the inflection point.
Since the inflection point is not a correct measure for the actual edge position but the only objective alternative at hand, a small
correction on the edge position will have to be included in calculat
ing EXAFS spectra. This correction 11£0 thus comprises the error
in the edge position of the reference compound used to calculate the
spectrum and the error in the edge position of the sample under
study .
The data analysis procedure consists of calculating EXAFS
spectra that resemble the measured spectra as accurately as possible . The mainstay of this process is the comparison of the calcu
lated and experimentally determined spectra. In this respect,
Fourier transforms are preferred above the data in k-space, especially when more shells . are present. In addition, corrected Fourier
transforms give a good insight in the effect of changing the parameters used to calculate the spectra. For example, a wrong coordina
tion distance will give a peak at the wrong position and an incorrect
value of 11£0 will result in a different symmetry in the imaginary
part of the Fourier transforms, provided the correct reference has
been used to correct the Fourier transform . When the coordination
number is too low or the Debye Waller factor (!io 2) is too high , the
resulting peak in the Fourier transform is lower than the
corresponding peak in the experimental data . Here we encounter a
first problem : N and !io 2 seem to be interrelated. at least in a
page 41 Chapter 2
Fourier transform . There is an infinite number of combinations of
N and ~a 2 that give the same peak height in the Fourier
transform. However, N and ~a 2 both have a different k
dependence : a higher coordination number will increase the ampli
tude of the EXAFS function over the whole k-range, a lower Deb ye
Waller factor will result in an increase in amplitude especially at
higher k-values, in other words, the damping of the EXAFS f unc
tion is less for a lower ~a 2. There is of course only one combina
tion of N and ~a 2 that will fit the experimental data best and this
combination can be found as follows. There are two restrictions.
One is that the contribution to be optimized should be separated
reasonably from the other contributions in the spectrum and the
second is that R and ~Eo are chosen correctly. We start by deter
mining N as a function of ~a 2 using a k 1-weighted Fourier
transform and assuring that the peak height of the calculated and
measured spectrum are the same over the whole ~a 2 range. This is
done easiest by taking a few fixed values for ~a 2 and determining
the corresponding coordination number. For a fixed value of ~a 2,
N is directly proportional to the peak height and an incorrect choice
of N can be corrected easily. (For a fixed value of N, ~a 2 is not
directly proportional to the magnitude of the peak in the Fourier
transform, see equations [2 .9 .9) and (2.9.17]) We repeat this pro
cedure, now using a k 3-weighted Fourier transform. Since in a k 1-
and k3-weighted Fourier transform the data are weighted differently
with respect to k and because N and ~a 2 have a different k
dependence, we will find two curves for N as a function of !J.a 2.
The intersection point of the two curves represents the combination
of N and ~a 2 that will result in the best fit over the whole range in k-space.
When there is one dominant contribution and one or more
smaller contributions in the spectrum, we can use the difference file
technique. We calculate and optimize a one-shell EXAFS spectrum
that matches the main peak in the Fourier transform of the experi
mental data best. We then subtract this calculated spectrum from the experimental data and analyze the difference spectrum, possibly
Experimental page 42
using once more the difference file technique. In this way , all con
tributions are analyzed. This technique has been introduced first by van Zon ezal. ( J S).
We can expand this technique to a recurrent optimization pro
cess. We subtract the calculated EXAFS function(s) that. matches
the difference spectrum (spectra) from the experimental data and
use this new difference file to optimize the major contribution in the
experimental data. We thus initiate a recurrent process that, when
carried out correctly , will converge to a set of parameters that yields
an EXAFS function that accurately resembles the measured spec
trum. However, care should be taken that the final set of parame
ters does not represent a ' local minimum '.
Using these procedures, the EXAFS spectra in the next
chapters have been analyzed. The details of the analysis procedure
depend highly on the system under study and thus , for details we
refer to the experimental part of the chapters 5, 7, 8 and 9.
Figure 2.8 The experimental set up at station 9 .2 at the SRS in Dares
bury .
Mo no-chr omat or
Wiggl er X-rays
~ ·~ ~ Samp l e
Stor age e
~ D D D ring
I C1 IC2
page 43 Chapter 2
2.9.5 Experimental Metfwd
A reliable analysis can only be done on high quality data and
high quality data can only be collected when the experimental set
up during the measurement has been carried out with utmost care.
The set-up as used on station 9.2 at the Synchrotron Radiation Source (SRS) in Daresbury, where most of the spectra discussed in
the next chapters have been measured , is shown schematically in Figure 2.8 . The synchrotron was operated at 1.8 or 2.0 GeV and
the ring current ranged from 100 to 300 mA. From the wiggler magnet (or from a bending magnet), a primary X-ray beam enters
the monochromator chamber . IP contains a whole spectrum of wavelengths. The beam leaving the monochromator, I 0, is highly
monochromatic : the wavelength obeys the Bragg-relation :
2d sin{ 8) (2.9 .18]
in which /... is the wavelength. d the d-spacing in the Crystals of the monochromator and e the angle between the incident beam and the crystals . The photon energy is in general given by
E = 12398.52 /...
[2.9.19)
Thus. taking into account only the first order reflection (n = 1). we find for the photon energy in I 0 :
E - 12398.52 2dsin(8)
[2 .9 .20]
{When d is in A. E is in eV). The crystals used at station 9.2 were two Si [220] crystals with a d-spacing of 1.916 A. A measure for the absorbance of the sample is given by ln(J 0/1). This is however
Experimental page 44
not an absolute measure for the attenuation coefficient µ.. but is
directly proportional to it . We will therefore use In (I 0/ I) as a
measure for µ. .
An optimum signal-to-noise ratio in µ. can only be achieved by
selecting optimum absorbances for the two ion chambers 11 and 12
and for the sample. The gas filling of the first ion chamber is such
that it absorbs 20% of the incoming beam I 0, the second ion
chamber absorbs 80% of the photons leaving the sample. The
thickness of the sample is chosen such that the absorbance (µ.) at
the edge is 2.5, i. e ., the sample absorbs about 8% of the photons
leaving the first ion chamber , 11. We refer to (36) for more details .
So far the description of the experimental set-up. The actual
data collection is now rather straightforward . The catalyst is
pressed into a thin , self supporting wafer with an absorbance of 2.5
at the edge and mounted in an in situ cell. In this · cell, in situ treatments can be carried out prior to the measurement. The pre
treatments of the different samples are described in the relevant
chapters. After treating the sample in the in situ cell. the cell is
placed in the beam between the two ion chamber and after a correct
alignment. the absorption spectrum can be measured . Usually . the
spectrum is recorded while the sample is cooled with liquid nitrogen
to about 100 K. The reference compounds are measured in exactly
the same way as the catalyst samples.
2 .10 References
1. Hermans, L. A. M._; Geus , J . W. "PreJX1.ration of Catalysts II" : Delmon, B.; Grange, P.; Jacobs , P. A., Eds .; Elsevier , Amsterdam 1983, p. 113
2. Geus, J . W "Prepa.ratiOn of Catalysts III" : Poncelet , G.; Grange , P.; Jacobs , P . A., Eds .; Elsevier, Amsterdam 1983, p. 1
page 45 Chapter 2
3. Hurst, N. W.; Gentry, S. J.; Jones, A. Catal. Rev., Sci. Eng. 1982, 24, 233
4. Robertson, S D.; McNicol, B. D.; de Baas, J. H.; Kloet, S. C.; Jen
kins, J. W J. Catal. 1975, 37, 424.
5. Jenkins, J. W.; McNicol, B. D.; Robertson, S. D. Chem. Tech 1975, 7, 316.
6. Wagstaff, N.; Prins, R. J. Catal. 1979, 59, 435.
7. Wagstaff, N.; Prins, R. J. Catal. 1979, 59, 445.
8. Vis, J. C.; van 't Blik, H. F. J.; Huizinga, T.; van Grondelle, J.; Prins, R. J. Catal. 1985, 95, 333
9. van "t Blik, H. F. J.; Prins, R. J. Catal. 1986, 97, 188
10. Martens, J. H. A.; van 't Blik, H. F. J.; Prins, R. J. Catal. 1986, 97,
200
11. van 't Blik, H. F. J.; Koningsberger, D. C.; Prins, R. J. Catal. 1986, 97, 210
12. Boer, H.; Boersma, W. J.; Wagstaff, N. Rev. Sci. Inst. 1982, 53, 349
13. Kip, B. J.; van Grondelle, J.; Martens, J. H. A.; Prins, R. Appl. Catal.
1986, 26, 353
14. Kip, B. J.; Duivenvoorden, F. B. M.; Koningsberger, D. C.; Prins, R'.
J. Catal. 1987, 105, 26.
15. Castner. T.. Newell. G. S .. Holton. W. C.. Slichter. C. P. J. Phys. Chem. 32. 668 (1960).
16. Wickman. H. H .. Klein. M. P .. Shirley. D. A. J. Phys. Chem.42. 2113 (1965).
17. Dowsing. R. D .. Gibson. J. F. J. Phys. Chem. 50. 294 (1969).
18. Mehring, M. "High Resolution NMR Spectroscopy in Solids";
Springer, 1976
19. Harris, R. K.; Mann, B. E. "NMR and the Periodic Table";
Academic Press, 1978
20. Mossbauer, R. L. Z. Physik 1958, 151, 124
21. Mossbauer, R. L. Naturwissenschaften 1958, 45, 538
22. Mossbauer, R. L. Z. Naturforsch. 1959, 149, 211
23. Anderson, A. B. J. Phys. Chem. 1975, 62, 1187
Experimental page 46
24. Richardson, J. W.; Nieuwpoort, W. C.; Powel, R. R.; Edgel, W. F. J. Phys . Chem. 1962, 36, 1057
25. Clementi, E. ; Raimondi, D. L. J. Phys. Chem. 1963, 38, 2686
28. Moore, C E. Atomic Energy Levels; NBS Circ. no. 467; National
Bureau of Standards ; U.S. Government Printing Office; Washington, DC, 1958
29. Anderson, A. B.; Grimes, R. W; Hong, S. Y. J. Phys . Chem. 1987,
91,4245
30. Stern , A . E, "X-Ray Absorption"; Koningsberger, D. C. ; Prins, R ..
Eds.; John Whiley & Sons, 1987; Chapter 1, p. 3
31. Teo, B. K.; Lee, P A. J. Am. Chem. Soc. 1979, 101, 2815.
32. Citrin, P. H.; Eisenberger, P.; Kincaid, B. M. Phys. Rev. Let . 1976,
36, 1346
33. Sinfelt, J. H.; Via, G. H.; Lytle, F. W.; Greegor, R. B. J . Chem. Phys. 1980, 72(9), 4832
34. Sayers, D. E. "X-Ray Absorption" ; Koningsberger, D. C.; Prins, R.,
Eds.; John Whiley & Sons, 1987; Chapter 6, p. 211
35. van Zon, J. B. A. D.; Koningsberger, D. C.; van 't Blik , H. F. J.; Sayers , D. E. J . Chem . Phys. 1985, 12, 5742.
36. Heald, S., "X-Ray Absorption"; Koningsberger, D. C.; Prins, R., Eds.; John Whiley & Sons , 1987; Chapter 3, p. 87
page 47 Chapter 3
Chapter 3
The preparation of y-Al20 3 supported Monometallic Rh and Pt and Bimetallic Rh-Pt Catalysts
3.1 Introduction
The introduction of the metal precursor onto the internal sur
face area of the support, the very first step in the preparation of a
supported metal catalyst, is of vital importance. This first step dic
tates the lower limit of the size of the eventual metal particles, and
in addition, it may be the prelude for the formation of bimetallic
particles. Here we will describe the pore volume impregnation
method and the adsorption method to prepare RhCl3 and H2PtCl6 catalyst precursors supported on y-Al 20 3. For the H2PtCl6 sam
ples, we used 195Pt NMR and Laser Raman spectroscopy to follow the growth of H2PtC16 crystallites during the pore volume impregna
tion method. From these experiments it became evident that above
a loading of 0.18 mmol H2PtC16 per gram of Al 20 3 crystals of
H2PtCl6 started to form. This confirmed the suggestion in the
literature that H2PtCl6 can adsorb as single entities on the Al20 3 support at low loadings ( 1 ,2). Furthermore, we studied the adsorp
tion of H2PtCl6 and RhCl3 on y -Al 20 3. Furthermore, we will
describe TPR profiles of RhCl3, H2PtCl6 · and (RhCl3 + H2PtCl6)
supported on Al20 3 in order to follow the formation of bimetallic
particles.
Preparation of Rh. Pt and Rh-Pt/ Al 20 3 page 48
3.2 Experimental
3.2.1 NMR and Laser Raman Experiments
In all experiments we used y-Al 20 3 from Ketjen (000-1.5E)
with a internal surface area of 180 m2 g-1 and a pore volume of
0.65 ml g- 1. The radius of the y-Al 20 3 particles was approximately
0.1 mm. H2PtC1 6/A1 20 3 samples were prepared containing 0.087,
0.169. 0.215. 0.256. 0.297 and 0.343 mmol H2PtCl 6 per g Al 20 3 using the pore volume impregnation method. The samples were
dried at room temperature for 24 h and at 393 K for another 24 h.
The l\IMR experiments were carried out on a Bruker CXP300
spectrometer under Magic Angle Spinning conditions in a Beams
Andrew rotor. For each spectrum, 4096 Bloch decays were accumu
lated at room temperature and Fourier transformed to give the
resulting spectrum.
For the Raman experiments, the dried samples were pressed
into circular tablets which were mounted in a holder and rotated at
an angle of 45° with respect to the laser beam, in order to avoid damage to the sample by heating and possibly burning. The mono
chromator was aligned with respect to the reflected beam. thus at
an angle of 45° to the sample. Using a dye liquid, the laser was
tuned to a wavelength of 5140 A and a power of 90 mW. The spec
tra were recorded at room temperature, at energies between 50 and
500 cm -l away from the energy of the incident light.
page 49 Chapter 3
3.2.2 Adsorption Experiments
For the adsorption experiments, two solutions containing
RhCl3 (15.5 and 31.2 mM) and two solutions containing H2PtCl6 (9.76 and 17.5 mM) were prepared. Various amounts of Al20 3,
ranging from 0.1 to 1.4 g, were added to a fixed amount of these
solutions (20 ml for the lower concentrated and 10 ml for the higher
concentrated solutions). After allowing the metal salts to adsorb
on the support for four days, the equilibrium concentrations of
RhCl3 and H2PtCl6 in the solution were measured. From these con
centrations, the amount of metal salt adsorbed on the support could
be calculated. For the set of lower concentrated solutions, the pH value of the solution in equilibrium was measured as well. The con
centrations of RhCl 3 and H2PtC16 were measured according to the
following procedure . 5 ml of a concentrated HCI solution and 5 ml
of a 1 M SnCl2 solution were added to 5 ml of the metal salt solu
tion. This mixture was heated on a water bath to the boiling point.
Highly colored [Rh2C14(SnC1 3) 4( and [PtC12(SnC13ht complexes were formed, and their extinction coefficients were measured at
A. = 471 nm for the rhodium complex and at A. = 401 nm for the platinum complex. For the rhodium complex, the extinction
coefficient is linearly dependent on the concentration only between 4 and 20 mM and for the platinum complex between 3 and 12 mM.
Therefore, in some cases the amount of 5 ml metal salt solution
mentioned above had to be diluted with distilled water in order to
assure that the conc~ntration of the final solution was in the desired range. For both complexes, the extinction coefficient was calibrated.
In order to establish the influence of the acidity of the solution on the adsorption process, the same procedure, using the lower con
centrated solutions, was followed again. but in this case, after the A120 3 had been added, and during the adsorption process, the pH value of the solution was maintained at 1.9 ± 0.2 using a 0.1 M HCI
solution.
Preparation of Rh, Pt and Rh-Pt/ Al 20 3 page 50
We also prepared 5 solutions of 20 ml for both the rhodium (15 .5 mM) and platinum (9.76 mM) salt and added about 0.1 g Al 20 3 to the solution. Directly after the addition of A1 20 3, the pH
value was brought to values ranging from 0.5 to 5.0. After allowing the adsorption process to come to equilibrium, the pH value and the amount of adsorbed metal salt were measured . We repeated these experiments, by bringing the pH of the solution to values ranging from 0.1 to 3.0 before the addition of the Al 20 3. After a week, the pH value and the amount of adsorbed metal salt were measured .
Finally, in order to establish whether there was any competition between RhCl 3 and H2PtC1 6 during the adsorption process , 0.6 g Al20 3 was added to a solution of 10 ml containing 31.2 mM RhCl 3 and 17.5 mM H2PtCl6. After a week, the rhodium and platinum content of the solution and the support were measured using the procedure described above.
3.2.3 TPR of Rh, Pt. and Rh-Pt / Al20 3
Six bimetallic Rh-Pt/y -Al 20 3 precursor catalysts were prepared using the pore volume impregnation method . For all the samples. the RhCl 3 and H2PtCl6 loadings were 0.100± 0.005 mmol g- 1 Al 20 3. Four of these samples were prepared using equimolar solutions of rhodium chloride (RhCl3·3H 20) and a platinum chloride. The acidity of the solution and the chloride content were increased by using respectively PtCl4·5H 20 and H2PtCl6·3H 20, dissolved in distilled water, and H2PtCl6 dissolved in 0.5 and 1.0 M HCI. 0.5 M HCI was equivalent with 3 HCI molecules per rhodium or platinum atom, 1.0 M HCI with 6 HCI molecules per rhodium or platinum atom. These samples will be denoted as RPA1, RPA2, RPA3 and RPA4, respectively.
page 51 Chapter 3
The other two samples were prepared using equimolar
amounts of RhCl3·3H 20 and H2PtCl6·3H 20 dissolved 1n either
CH 30H or C2H50H . The sample impregnated with CH 30H is
denoted as RPA5, the sample impregnated with C2H50H as RPA6.
After impregnating and carefully drying, TP R experiments were car
ried out as described in chapter 2, using a heating rate of 5 K min -1.
Drying of the samples impregnated with CH 30H and C2H50H gave
problems. Drying at elevated temperatures (up to 500 K for 10 h,
even in vacuum) was insufficient to remove the alcohol completely.
The only effective procedure we found was to dry the sample at
393 K for a few hours and then for at least two weaks at room tem
perature. Obviously, CH 30H and C2H50H adsorbed strongly on the
y-Al 20 3 and desorption was very slow.
3.3 Results and Discussion
3.3.1 NMR and Laser Raman Experiments
In Figure 3.1a, . the NMR spectra of four supported H2PtCl6
samples are shown. The chemical shift indicated that the peak
observed originated from Pt in IPtC16f octahedra ( 3,4). Apart from
this, no other peaks were observed. As a measure for the intensity
of the peaks, we used the peak height (in arbitrary units), multi
plied by the width of the peak at half height (in arbitrary units).
These intensities are tabulated in Table 3.1 and plotted in Figure
3 .1b as a function of H2PtCl6 loading . As indicated in Figure 3.1b.
a straight line can be constructed through these points and this line
intersects the abscissa at a loading of 0.16 ± 0.01 mmol H2PtCl6 per g Al 20 3. Obviously, H2PtCl6 crystals start to form only above
this limit. Below this limit, the platinum salt are adsorbed strongly on the support, as will be demonstrated by the adsorption experi
ments . We could not observe other peaks in the spectrum that
Preparation of Rh, Pt and Rh-Pt/ Al20 3
Figure 3.1 Results of the NMR experiments .
1
2
3
4
(a) NMR spectra of H2PtCldAl20 3 with loadings of :
1 : 0.215 mmol g- 1, 2 : 0.256 mmol g- 1
3 : 0.297 mmol g- 1, 4 : 0.343 mmol g- 1
(b) NMR intensities as a function of H2PtC1 6 loading
possibly could be attributed to such adsorbed complexes. A possi
ble reason might be the following. It is known that small deviations in the chemical environment of platinum result in enormous
chemical shifts and this alone makes it difficult to detect them.
page 53 Chapter 3
Therefore, platinum chloride complexes which are adsorbed on the
support and which most probably have exchanged one or more
chloride ions for an oxygen ion or a hydroxyl group of the support,
will have a different chemical shift than [PtCl6t complexes. Evidence for such an exchange was found by Lagarde (5), who
observed that after impregnating a y-Al20 3 support with H2PtCl6,
the platinum complexes had lost approximately one Cl- ligand
(NPt-CI = 5.0 ± 0.5) . The loading of that sample, prepared via pore volume impregnation, was low enough to assure that no cry
stalline H2PtCl6 had been formed (1.5 wt% Pt, which is equal to .
0.077 mmol per gram Al20 3; the surface area of the Al20 3 was
240 m2 g-1). In addition, we expect that the variation in the chemi
cal environment of these strongly adsorbed complexes may be (relatively) large, which will give rise to a large spread in chemical shifts and therefore to low and broad peaks, which are difficult to detect.
In Figure 3.2a, the Raman spectra of a H2PtCl6 solution and of
three supported H2PtCl6 samples are shown. In these spectra, the
Raman active frequencies of the [PtC16]2- units are clearly visible.
The line width of the peaks around 340 and 320 cm -l could not be
measured accurately. Therefore, the peak height was taken as a measure for the intensity. These 'intensities' are collected in Table
3.2 and depicted in Figure 3.2b. The general behavior of the intensities as a function of loading is the same for the three bands.
Above a loading of about 0.18 ± 0 .02 mmol H2PtCl6 g-1 Al20 3, the
intensity increases more with increasing H2PtCl6 loading than below
that loading. This can be explained as follows. Below the limit of 0.18 mmol H2PtCl6 g-1 Al20 3, H2PtCl6 adsorbed as single entities on
the support and these adsorbed platinum complexes were not per
fect [PtC16f octahedra, as could be concluded from the shift in the
frequencies in the spectra (cf. Table 3.2). The Raman frequencies
are related to the vibrational force constants. Since these frequencies changed only slightly, the force constants had not changed
drastically, which indicates that the [ PtC 16]2- octahedra were
slightly deformed or had exchanged one (or more) chloride ions for
-0 H- groups of the support. As discussed for the NM R
Preparation of Rh, Pt and Rh-Pt/ Al20 3 page 54
Figure 3.2 Results of the Raman experiments .
a
1
(a) Raman spectra of H2PtCICi and H2PtCl<i/Al20 3 (spectra are not to scale) :
1 : H2PtCl6 in aqueous solution,
2 : 0.169 mmol g- 1, 3 : 0.256 mmol g-1
, 4 : 0.343 mmol g-1
(b) Raman intensities as a function of H2PtCICi loading
340 cm-1 : solid line, 320 cm-1
: dashed line, 150 cm-1 : dotted
line
2
b
~ I
Q /
,•
~~ ,
/
>- ,i' (/) 0
J
c .O" (1) • _.,...--- ,, c .··
I
4 ,,
. ..0 ... ·
0 " 0 "
0 " 400 100 0.0 0.2 0.4
-Wavenumber (cm·1 ) Loading --
experiments, this will be accompanied by a broadening of the bands and therefore, the peak heights will be lower. Above the limit of 0.18 mmol H2PtCl6 g-1 Al20 3, the frequencies of the three bands corresponded precisely with the frequencies of pure H2PtCl6, which indicated that (perfect) H2PtCl6 crystals had started to form. The solid line in Figure 3.2b depicts the intensity of the bands of pure H2PtCl6 in the samples and the dotted line the estimated contribution of the adsorbed platinum complexes.
page 55 Chapter 3
Table 3.2 Raman data
H2PtCl6 Frequency Peak loading height
(mmol g- 1) (cm- 1) (a.u .)
0.087 338 0.42 324 0.30
0.169 343 1.18 325 0.96 154 0.35
0.256 343 1.53 319 1.34 159 0.48
0.343 343 3.36 316 3.04 157 1.09
The main conclusion from the NMR and Raman experiments is
that up to loadings of 0.16-0.18 mmol H2PtCl6 g- 1 Al20 3 platinum
complexes were adsorbed on this support, and were probably molec
ularly dispersed. These complexes differed from perfect [PtCl6)2-
octahedra and had probably exchanged one or more Cl- ligands for
oxygen ions or hydroxyl groups from the support. Above this limit,
H2PtCl6 crystals had started to form. This limit is far above the
amounts of H2PtClri normaliy used to prepare Pt/ Al 20 3 catalysts.
A metal loading of 2 wt% corresponds to a loading of 0.1 mmol H2PtCl6 per g Al 20 3. After reduction, these catalysts are
never atomically dispersed . Typical particles sizes for Pt/A1 20 3
catalysts are about 8 A and these particles contain about 10 plati
num atoms (6). Thus, the formation of these larger metal particles
in the final form of the catalyst must have taken place during the
reduction procedure. It should be noted, that these results were obtained for a y-Al 20 3 with a surface area of 180 m2 g-1. For other
aluminas, the results may be different.
Preparation of Rh, Pt and Rh-Pt/ Al20 3 page 56
3.3.2 Adsorptwn Experiments
Table 3.3 Adsorption isotherms for RhCl 3 and H2PtCl6 on Alp3
a [M] is the concentration ( mM) of the metal salt in the solution above
Alp3 in equilibrium
b [M/S] is the metal-salt loading (mmol g- 1) on the Alp3 support
In order to obtain information of the state of the complexes adsorbed on the support, we performed adsorption experiments to
establish adsorption isotherms for RhCl 3 and H2PtCl 6 on Al 20 3.
Table 3.3 summarizes the results. The maximum amount of RhCl3 that was adsorbed was 0.326 ± 0.009 mmol g- 1 for the more concentrated and 0.47± 0.1 mmol g-1 for the less concentrated RhCl3 solu
tion. For H2PtCl6, these amounts were 0.225 ± 0.003 and 0.24± 0.07 mmol g-1
, respectively. (l\Jote that for the experiments where the equilibrium concentration was lower, the uncertainty in the loading will be lower as well. The uncertainty in the concentra
tions ranged from 2 to 4%. For concentrations > 1 mM. the
page 57 Chapter 3
uncertainty is about 4%. for concentrations > 0.1 mM ± 3%. for
concentrations > 0 .01 mM ± 2%. Thus, the uncertainties in the
maximum attainable coverages are highest. This rjtiight suggest
that the uncertainty in the actual maximum attainabl.e coverages is
high . But the uncertainties in the second figures in li able 3.3a, b, c
and d are much lower . These uncertainties are about 0 .002-0.004) .
During the experiments with the lower concentrated solutions, the
pH value of the solution in equilibrium was measured. At the max
imum attainable coverage, the pH value of the solution was 3.0 for
the RhCl 3 solution and 2.6 for the H2PtCl 6 solution . At low metal
salt coverages, the pH value was about 3.6. A1 20 3 served as a
buff er agent and stabilized the pH value to 3.6 ± 0.1 over a wide
concentration range.
Table 3.4 Adsorption capacities for RhCl3 and H2PtCl6 at low pH values
a [ MJ is the concentration ( mM J of the metal salt in the solution above A1p3 in equilibrium
b [M/SJ is the metal-salt loading (mmol g- 1) on the A1p3 support
Table 3.5 summarizes the results of the adsorption experi
ments which were performed using different starting acidities.
From these results, it is clear that acidifying the solution had a
major impact on the adsorption capacity of the Al 20 3. It did not
matter whether the pH value of the solution was adjusted before or
after the addition of Al 20 3. The reason is, that the process of buffering the solution to a pH value of 3.6 was rather slow. Typi
cally, it took several hours for the A1 20 3 to adjust the pH value of
the solution to 3.6. To bring the solution (before or after the
page 59 Chapter 3
addition of the A120 3) with HCI to the values listed in Table 3.5 was a matter of a few minutes . During that short period of time,
· the Al20 3 had no time to buffer the pH value of the solution.
Therefore, the two sets of experiments can be regarded as the same. Note, that in case the starting acidity was lower than about
1.5, the Al20 3 did not manage to stabilize the pH value to
3.6 ± 0.1. The amount of Al 20 3 has been chosen low in order to
ensure that complete coverage would be obtained in all cases.
Therefore, the loadings in Table 3.5 reflect the maximum adsorption
capacity as a function of equilibrium pH value. The obvious con
clusion is, that the higher the starting pH value, the more RhCl3 or
H2PtCl6 can be adsorbed on the A120 3.
At this point, it is easy to understand why a lower amount of
metal salt was adsorbed when a more concentrated solution of
these. salts was brought in contact with Al20 3 (cf. Table 3.3). The
more concentrated solutions obviously has a lower pH value and
therefore less metal salt was adsorbed from these solutions.
A similar pH-dependence for the adsorption of H2PtCl6 on yAl203 has been reported by Heise et al. (7 ). They found a max
imum in the adsorption capacity at pH 3.5-4.0. In that study, the penetration depth was measured of the platinum complexes in yA203 pellets mounted above the solution and in contact with that
solution. The penetration depth they reported (0.2-0.5 mm) is larger than the radius of the Al20 3 particles we used (0.1 mm).
Thus, there is no reason to assume any depletion of the solvent in the pores during our experiments.
In order to explain the results of these adsorption experiments,
we must focus on the reactions which may take place during the
adsorption process. For example, for RhC13, the following reactions are of importance :
Preparation of Rh , Pt and Rh- Pt/ A120 3 page 60
RhCl3 + 3 H20 -+ RhC13(H 20)3 (3.1)
Rhc13(H2o h + H2o +:t !Rhc13(H 20)20Hr + H3o+ 13.21
We assume that in aqueous solutions the majority of the oxy
gen ions from the support which are exposed to the solvent are sur
face -0 H groups. In acid solutions , such as solutions of RhCl3 and
H2PtCl 6, these hydroxyl groups will be partially protonated :
l-OHJIH30+J
1-0Ht)
101 I [-OH2+] I og I-OH]
OH
13.3)
[3.4)
+pH
Cl
[3.5)
. I / Cl H20-/ Rti-OH2
Cl +H20
0
Thus, in the acid solutions of RhCl3 and H2PtCl6, the Al20 3 surface
is positively charged and the rhodium complexes are negatively
charged or are neutral. Adsorption of the complexes to the surface
of the support will therefore take place, due to Coulomb forces. A possible mechanism for the adsorption of rhodium complexes is
shown in equation [3.5]. The amount of protonated surface hydroxyl groups can be calculated from the pH value of the solution
and the pK0 value which is assumed to be equal to 3.6, the pH
page 61 Chapter 3
value of solutions stabilized by Al 20 3. In (7 ), Heise et al. found
that at pH= 8, the A1 20 3 surface was electrically neutral. Thus, at this pH, the fraction of protonated sites will be close to zero and
the fraction of unprotonated sites will be close to unity . Assuming that the amounts of protonated and non-protonated sites vary with
the pH of the solution as reported by Bowers (8), we can estimate
that the amount of protonated and unprotonated surface hydroxyl
groups will equal each other at approximately pH= 4-5, which is close to the value of 3.6 we assumed . Using equation (3 .4J, we can
estimate the amount of protonated and non-protonated sites on the Al 20 3 surface.
Table 3.6 Amount of protonated and non-protonated surface hydroxyl
The TPR profiles of the samples RPA1 to RPA6 and the TP R profiles of RhCl3/ A1 20 3. PtC14/ A1 20 3 and H2PtC16/ A1 20 3 are presented in Figure 3.3. In Table 3.7, the results of the TPR experiments are summarized. Note, that the metal loading was 0.10 mmol g-1 in all cases. This loading is lower than the limit above which crystalline H2PtCl6 started to form according to the NMR and Raman experiments (0.16-0.18 mmol g-1
). Although this value was obtained for H2PtCl6, we observed that the adsorption capacity for RhCl 3 was always higher than for H2PtCl6 and therefore we expect that the loading above which crystalline RhCl3 starts to form is at least equal or even higher than the value reported for
Preparation of Rh, Pt and Rh-Pt/ Al20 3 page 68
H2PtCl6. Hence, we assume that in the samples in which RhCl 3,
PtCl 4 and H2PtCl6 were dissolved in water (RPA1 and RPA2), the
salts were adsorbed as single , isolated entities on the support. In
the cases where HCI has been added to the solution (RPA3 and
RPA4), we may expect that the adsorption capacity of the support
has decreased and thus, that crystalline material may have been
Tmax is the temperature in the TPR profile of the maximum hydrogen uptake.
For the monometallic samples. RhC1 3/ Al 20 3 reduced at 415 K,
PtC1 4/ Al 20 3 at 510 K and H2PtC16/ A1 20 3 at 505 K. For the bimetallic samples, imprengated with H20 , the reduction took place in
one step at temperatures where monometallic RhCl 3 was reduced.
For the samples impregnated with HCI solutions , the hydrogen
uptake at higher temperatures increased (cf. Figure 3.3). Using the
results from the adsorption experiments, we can explain these
findings as follows. Going from RPA1 to RPA4 , the acidity of both the solut ion and , because of the increasing chloride content, of the
Al 20 3 increased. From the NMR and Raman experiments we may
page 69 Chapter 3
conclude that for RPA1 and RPA2 no crystalline RhCl 3 or crystal
line H2PtCl 6 was present. Therefore, on these samples. the rhodium
and platinum complexes were adsorbed as isolated complexes on
the support. After the reduction process, the H /M value deter
mined with hydrogen chemisorption for RPA2 was 0 .77 . Using the
calibration described in (6), we can estimate that the particles were
about 15 A in diameter and contained about 50 atoms . Obviously,
during the reduction process, the atomically dispersed rhodium and
platinum complexes have been reduced and have grown or sintered
to larger metal particles. In order to explain why RhCl 3 and H2 PtCl6
were reduced in one step, we propose the following model. During
the reduction process, the adsorbed RhCl 3 complexes were reduced
first. The metallic rhodium atoms that were formed had only very
little interaction with the support and therefore had a significant mobility. These mobile rhodium metal atoms can 'diffuse' over the
support and, in addition, adsorb and dissociate hydrogen. Therefore,
when they encounter an unreduced rhodium or platinum complex
during this 'random walk'. they can catalyze the reduction of that
complex. Thus, during the reduction process, the average particle
size gradually increases due to the accumulation of additional metal
atoms, until the metal particles are too large to have any significant
mobility. or until all the RhCl 3 and H2 PtCl 6 is reduced. For the more acid solutions, RPA3 and RPA4, we may expect that not only
monodispersed rhodium and platinum complexes have been formed,
but crystalline material as well. These crystallites may 'capture'
diffusing small metal clusters and become reduced. But once they
are reduced, these larger crystallites have no significant mobility and
thus smother the reduction process . This is most pronounced for R PA4 : after a sharp increase, due to the reduction of monodisperse
RhCl3 and the start of the reduction of larger crystallites, the hydro
gen uptake decreased immediately. The reduction of the remaining
crystallites was not catalyzed but depended on the reducibility of the particles and thus on their (surface) composition : crystallites
containing more RhCl3 were reduced more easily than crystallites containing more H2PtCl6. This explains the tailing behavior of the
reduction profile for RPA3 and RPA4. In the TPR profile of RPA4,
Preparation of Rh, Pt and Rh-Pt/ A1 20 3 page 70
some hydrogen uptake was still visible in the region were pure
H2PtCl6 is reduced. This indicates that some monometallic H2PtCl6 crystallites were present. The main conclusion is that the reduction
process was dominated by mobile metal clusters that once formed
catalyzed the reduction of the whole sample, if their mobility was
not reduced drastically. When crystalline material was present, the
reduction process was s.mothered.
The TP R profiles of the two alcohol-impregnated samples
were completely different. There was a peak in the hydrogen
uptake at very low temperatures, below the reduction temperature
of RhCl 3, and a peak exactly at the reduction temperature of
H2PtCl6. This indicates that RhCl3 and H2PtCl6 were separately
present on the support. In the experimental part we noted, that the
alcohols adsorbed strongly on the support. Therefore, we may
assume that there were no sites left for RhCl 3 and H2PtCl6 to
adsorb and thus, that the metal salts were not adsorbed as single
complexes. Hence, during the drying procedure, only crystalline
RhCl3 and H2PtCl6 is formed. However, the hydrogen uptake of the
two separate peaks did not agree with the amount of RhCl3 and
H2PtC16 present in the sample, although the total hydrogen uptake
agreed very well. The hydrogen uptake in the low temperature peak
was too high for the reduction of RhCl3 only. Apparently, some
H2PtCl6 had been incorporated in the RhCl3 crystallites. Why RhCl3 and H2PtCl6 crystallized separately is not yet clear. The RhCl3 crys
tallites were reduced below the temperature where RhC13/ Al20 3 is
reduced. We believe that this effect is due to the presence of
CH 30H or C2H50H in the crystallites replacing crystallization water.
Pure RhCl3 contains 3 molecules crystallization water per RhCl3.
Because these samples were prepared using CH 30H and C2H50H
and RhCl3.3H 20 (see experimental), during recrystallization part of
the crystallization water may have been replaced by CH 30H or
C2H50H. Because these molecules are substantially larger than
H20, they perturb the structure, and thus may decrease the reduci
bility of the particle. This effect is most pronounced for the CH 30H sample, because CH 30H is smaller than C2H50H and thus
page 71 Chapter 3
could replace crystallization water more easily. For the reduction
peak of the H2PtCl6 crystals we did not observe a lowering of the
reduction temperature. This can be explained by the fact that
H2PtCl 6 already has a very weak crystal structure and the sites in
the structure for the crystallization water are fairly large (see Figure
2.2 in chapter 2). Thus, replacing H20 for CH 30H or C2H50H will
have no major influence on the structure and on the reducibility of
the H2PtCl6 crystallites.
3.4 Conclusions
From the NMR and Raman experiments we conclude that dur
ing the pore volume impregnation method H2PtCl6 can adsorb as
single, isolated complexes on the Al20 3 support up to a loading of
about 0.16 mmol g-1. Above this limit, H2PtCl 6 crystals are formed.
The adsorption experiments indicated that for RhCl 3 the same
behavior could be expected. The maximum attainable mono
disperse coverage was even larger for RhCl 3 than for H2PtCl 6. This
maximum attainable coverage was dependent on the pH of the solution, with a decreasing adsorption capacity with decreasing pH. During the reduction process, the monodisperse rhodium complexes were reduced first and because of their mobility they could catalyze
the reduction of the rest of the metal complexes. The presence of larger crystallites smothered the reduction process . When an
alcohol was used as solute, no monodisperse adsorption had taken
place. During the drying process, RhCl 3 and H2PtC16 were
separated . Because there were no monodisperse rhodium complexes
and therefore no mobile rhodium atoms formed during the reduction
process, the RhCl 3 and H2PtCl6 crystallites were reduced separately.
Preparation of Rh. Pt and Rh-Pt/ Al20 3 page 72
3 .5 References
1. Santecesaria, E.; Carra, S .; Adima , I. Ind . Eng . Chem ., Prod . Res . Dev. 1977, 16(1), 41
3. Mehring, M. "High Resolution NMR Spectroscopy in Solids"; Springer, 1976
4. Harris, R. K.; Mann, B. E. "NMR and the Periodic Table"; Academic Press, 1978
5. Lagarde , P .; Murata, T.; Vlaic, G.; Freund, E.; Dexpert, H. J. Catal . 1983, 84, 333
6. Kip, B. J.; Duivenvoorden, F. B. M.; Koningsberger, D. C.; Prins, R. J. Catal. 1987, 105, 26.
7. Heise, M. S.; Schwarz, J. A. J . Col. Int. Sci. 1985, 107, 237
8. Bowers, A. R.; Huang, C. P .; J . Col. Int. Sci. 1985, 105 ( 1),197
9. Cox, L. E.; Peters, D. E.; Wehry, E. L. J. lnorg . Nucl . Chem. 1972,
34, 297
10. v.d . Berg, G.H.; Rijnten, H. T. "Preparation of Catalysts II"; Del
mon, B.; Grange, P .; Jacobs, P . A., Eds.; Elsevier, Amsterdam 1979,
p. 285
11. Botman , M. J. P., thesis, State University of Leiden (Netherlands),
1987.
12. Le Page, J. F. Catalyse de Contact, Technip, Paris, 1978, p. 589
13. Castro, A. A.; Scelza, 0 . A.; Benvenuto, E. R.; Baronetti, G. T.; de
Miguel, S . R.; Parera, J. M: "Prepa.ration of Catalysts Ill"; Pon
celet, G.; Grange, P.; Jacobs, P . A., Eds.; Elsevier, Amsterdam 1983,
p. 47
14. Heise, M. S .; Schwarz, J. A. J. Col. Int. Sci. 1985, 113, 55
page 73 Chapter 4
Chapter 4
Ferric Iron in Reduced Si02 Supported Fe-Ru and Fe-Pt Catalysts
Evidence from Mossbauer Spectroscopy and Electron Spin Resonance
4.1 Introduction
Supported bimetallic catalysts consisting of iron and one of the more noble group VIII metals M (M = Ru, Rh, Pd, Ir and Pt) have extensively been studied by Mossbauer spectroscopy ( 1-11 ). In general, the Mossbauer spectra of reduced Fe-M/Si02 catalysts contain two contributions, one due to an Fe-M alloy and the other to a doublet with an isomer shift (IS) of about 0.65 mm s- 1 relative to sodium nitroprusside and a quadrupole splitting (QS) in the range of 0.6 - 1.0 mm s-1
. These parameters are entirely characteristic of high-spin Fe3+, and several authors ( 5-11) have made th is assignment. Garten ( 1 ), Lam et al. ( 2), Vannice et al. ( J) and Garten and Sinfelt (4), on the other hand, favor the interpretation that the doublet in the Mossbauer spectra of reduced Fe-M /Si02 and FeM/ Al 20 3 catalysts corresponds to zero-valent iron atoms in the surface of the Fe-M alloy particles. The high isomer shift is explained by the assumption that the electron density for surface iron atoms is lower than for bulk iron atoms ( 2). The Mossbauer spectra of reduced Fe-Rh /Si02 and Fe-Ir /Si02 catalysts, measured in sizu at 4 K, however, do not support the interpretation in terms of zerovalent surface iron but are in agreement with the assignment of the
Fe3+ in Fe-Pt and Fe-Ru/Si02 page 74
doublet to Fe3+ (9,J J ).
From the viewpoint of Mossbauer spectroscopy the assign
ment of the doublet with the parameters as given above to zero
valent iron seems unlikely and interpretation in terms of Fe3+ would
be preferred. From a chemical point of view, however, it is not
readily apparent why substantial amounts of ferric iron should be
stabilized in the presence of a noble metal. which in general facili
tates the reduction of the less noble component, iron. In most
Fe/Si02 and Fe/ Al20 3 catalysts, iron can be reduced to at least the Fe2+ state ( J ,JS), although in some cases, such as the promoted
ammonia or Fischer-Tropsch synthesis catalysts small amounts of
Fe3+ are also observed (JO.JS). In conclusion, the presence of ferric
iron in reduced Fe-M/Si02 catalysts, as deduced from Mossbauer
spectroscopy, seems somewhat unexpected and confirmation by
another in situ technique would be highly desirable.
Electron Spin Resonance (ESR) is very sensitive in detecting
F 3+ . d b 1 · d . . F 3+ . h h e ions an can e app 1e zn sztu. e ions ave a very c arac-
teristic ESR signal centered at g = 4.2 whenever the site symmetry deviates slightly from the perfectly octahedral or tetrahedral sym
metry ( J2-J4). Trivalent iron has been the subject of many ESR
studies and the corresponding g = 4.2 ESR signal cannot be mis
taken for divalent or zerovalent iron.
In this note we report ESR and Mossbauer results of reduced
SiOrsupported Fe-Ru and Fe-Pt. These catalysts represent the combination of iron with the least noble and the most noble metal
in the Fe-M/Si02 series. The ESR experiments confirm that both
catalysts contain ferric iron, in amounts comparable to those deter
mined by Mossbauer spectroscopy.
page 75 Chapter 4
4.2 Experimental
Catalysts were prepared by impregnating the Si02 support
(Cab-0-Sil, EH-5, 310 m2 g-1) with aqueous solutions of
Fe(N03)J.9H 20 and RuCl 3.3H 20 or H2PtCl6.6H 20 under frequent stirring, until the incipient wetness point was reached. The Fe
Ru/Si02 catalyst contained 0.46 wt% iron and 4.15 wt% ruthenium: the Fe-Pt/Si02 catalyst 0.28 wt% iron and 4.72 wt% platinum. The iron was 10% enriched in the isotope 57 Fe. Catalysts
were dried in air at 295 K for a week , at 330 K for 24 h and at 400 K for 72 h. The catalysts were reduced at 400 K for 0.5 h and sub
sequently at 725 K for 6 h in the Mossbauer in situ reactor.
Mossbauer spectra were measured at room temperature with a
constant acceleration spectrometer . Doppler velocities are reported with respect to the isomer shift of sodium nitroprusside at 295 K. After measuring the Mossbauer spectra , the catalysts were passivated in air at 295 K and transferred to an in situ ESR sample
holder, described in ( 16). The samples were reduced in flowing hydrogen at 725 K. It was checked that Mossbauer spectra of the
catalysts after passivation and rereduction at 725 K are identical to those obtained after the first reduction treatment.
The X-band ESR spectra were recorded with a Varian E-15 spectrometer equipped with an Oxford Instruments ESR-9 continuous flow cryostat. In order to quantitatively determine Fe3+ concen
trations we measured ESR spectrum intensities of the two reduced catalysts and of a reference compound with a known Fe3+ concentration (A1 20 3 CK300, Ketjen : 0.03 wt% Fe3+) at different tem
peratures between 4 and 80 K.
Fe3+ in Fe-Pt and Fe-Ru/Si02 page 76
4.3 Results and Discussion
Mossbauer spectra of the reduced Fe-Ru/Si02 and Fe-Pt/Si02 catalysts are shown in Figure 4.1. The spectra have been analyzed by
computer to determine the l\llossbauer parameters of the iron compounds
present and their spectral contributions: see Table 4.1 for the results. The
spectrum of Fe-Ru/Si02 consists of two quadrupole doublets, one charac
teristic of iron in hep Fe-Ru (17 ,18) and the other of high-spin Fe3+. The
spectrum of Fe-Pt/Si02 has been fitted with two doublets as well. One is
identical to the doublet reported for iron in an ordered tetragonal Fe-Pt
alloy (19 ), the other doublet is characteristic for high-spin Fe3+. As Table
4.1 shows, the contribution of Fe3+ to the Mossbauer spectra at 295 K of
reduced Fe-Ru/Si02 and Fe-Pt/Si02 is in the order of 80%. This number
should be considered as a lower limit for the actual Fe3+ content, because a
previous study of the Fe-Rh/Si02 system has shown that the recoilless
fraction of Fe3+ is considerably smaller than that of zero-valent iron in the
Fe-Rh alloy ( 19) Hence, the actual Fe3+ content of the reduced Fe-Ru and
Fe-Pt catalysts may well exceed 80%.
Figure 4.2 shows the ESR spectra of reduced Fe-Ru/Si02 and Fe-Pt/Si0 2 and of the Fe3+-containing Al 20 3 reference compound. All spectra show the characteristic Fe3+ spectrum at g = 4.2. The
presence of ferric iron has thus been established. The amount of ferric iron in both catalysts has been obtained by measuring the
ESR intensity at different temperatures. The spectral intensity fol
lows from the formula
I - H (WPP )2 12.9.1]
in which I is the intensity, H the peak-to-peak height of the spec
trum (corrected for receiver gain) and W the peak-to-peak line width of the spectrum in Gauss. PP
page 77 Chapter 4
Figure 4.1 Mossbauer spectra of reduced Fe-Ru and Fe-Pt/Si02 ca
talysts, measured in situ under H2 at 295 K
Ci) I ..... .4 c:: :::J 0 (.)
(!)
0 .s >- 1.3
..... "ii) c:: (!) ..... c::
1 :5 Fe-Ru I Si02
-5
nFeRu nFe3+
0
1: 5 Fe-Pt/Si02
nFePt nFe3+
5 -5 0 Doppler Velocity (mm;s)
3.8
3.6
5
Table 4.1 Mossbauer parameters of Fe in Fe-Ru/Si02 and Fe-Pt/Si02
after reduction in H2 at 725 K
Mossbauer Parameters
IS QS % Assigned Catalyst -! mm s- 1 to mm s
Fe-Ru 0.27 0.19 16 Fe3+ in Fe-Ru
0.69 0.71 84 Fe3-'-
Fe-Pt 0.56 0.43 17 Fe0 in Fe-Pt
0.69 0.76 83 Fe3+
Figure 4.3 shows the calculated reciprocal intensity for the three samples as a function of temperature. At temperatures above 10 K the magnitude of 1 I I depends linearly on T. Above 60 K
Fe3+ in Fe-Pt and Fe-Ru/Si02 page 78
Figure 4.2 ESR spectra of the reduced Fe-Ru and Fe-Pt/Si02 catalysts
measured in situ at 4 K. Spectrum c corresponds to the 0.03 wt% Fe3+-in-Al20 3 reference. For clarity, the curves have
been shifted : the arrows correspond tog = 4 .2. The intensities are not to scale .
a
c
lOOG
a Fe-Ru/Si02 b Fe-Pt/Si02
c Al20 3
b
c
saturation occurs, giving deviation from the linear dependence . Since the slope of the linear part of the 11 I curve is inversely proportional to the Fe3+ concentration, the latter follows from the formula
c, - I ~: II i c, [2.9.2]
in which C is the concentration of Fe3+ in wt%, D the bulk density of the sample and S the slope of the linear part in the reciprocal-
page 79 Chapter 4
Figure 4.3 Reciprocal intensities of the g = 4.2 ESR lines versus tem
perature for the reduced Fe-Ru and Fe-Pt/Si02 catalysts and
the 0.03 wt% Fe3+-in-Al 20 3
:i ~ -.5 ...... ,...
1.0
a Fe-Ru/Si02
b Fe-Pt/SiC>i
c Al20 3
50 T(K)
100
intensity versus temperature plot . The indices s and r denote sample (catalyst) and reference compound (the 0.03 wt% Fe3+-in
Al203). Table 4.2 summarizes these results.
The ESR analyzes confirm that Fe-Ru/Si02 and Fe-Pt/Si02 catalysts contain substantial amounts of ferric iron which survives reduction in H2 at 725 K, notwithstanding the presence of a noble
metal. For the Fe-Ru/Si02 catalyst, both Mossbauer spectroscopy and ESR indicate that at least 80% of the iron is in the ferric state.
For Fe-Pt/Si0 2, on the other hand, the Fe3+ contents as determined by Mossbauer spectroscopy and ESR are 83 and 40%,
respectively. It should be noted that ESR detects Fe3+ provided that these ions are not antiferromagnetically ordered as in the common bulk iron(JJJ)oxides, Fe20 3 and FeOOH. Also, the intensity of the g = 4.2 signal may depend slightly on the deviation of the site
Fe3+ in Fe-Pt and Fe-Ru/Si02 page 80
Table 4.2 ESR results and comparison with Mossbauer results
(1.) : wt% Fe3+ for the catalyst as determined by ESR (2) : percentage of iron present as Fe3+ as determined by ESR (3) : Contribution of Fe3+ to the Mossbauer spectra at 295 K
symmetry from octahedral or tetrahedral. Therefore, the amounts of Fe3+ calculated from the ESR intensities should be considered to be semi quantitative. Nevertheless, they prove that substantial amounts of Fe3+ are present in reduced Fe-Ru and Fe-Pt/Si02 catalysts and that these amounts are of the same magnitude as those determined by Mossbauer spectroscopy.
The fact that relatively large amounts of Fe3+ can be detected
by ESR is in agreement with the conclusion based on Mossbauer spectra that the ferric iron is present in a highly dispersed state, in close contact with the Si02 support (8,9,JO). The reason why unreduced iron occurs predominantly as Fe3+ in bimetallic Fe-M/Si02 catalysts and as Fe2+ in most monometallic catalysts is probably due to differences in dispersion. According to Guczi ( S), the iron and the noble metal impede each others migration over the support during reduction and maintain each other in a state of high dispersion. In this view, the formation of ferrous iron in Fe/Si02 catalysts is accompanied by sintering of the iron to some extent. Experiments to test the validity of this explanation are in preparation.
page 81 Chapter 4
4.4 Conclusions
With Mossbauer spectroscopy, the presence of ferric iron in
reduced Fe-Ru/Si02 and Fe-Pt/Si0 2 has been observed. However, from the Mossbauer spectroscopy point of view, their might be
doubts about this assignment. In ESR, Fe3+ ions cannot be mistaken with other Fe ions. Therefore, the ESR experiments clearly
indicated that the assignment made by Mossbauer was indeed correct. Moreover, the amounts of ferric iron as determined by ESR
agreed very well with the amount as found with Mossbauer. Therefore, these observations make the model in which iron and the noble
metal maintain each other in a highly dispersed state, a state in which iron cannot be reduced to ferrous iron, very Ii kely.
4.5 References
1. Garten, R.L. 11 Mossbauer Effect Methodology 11 I. J. Gruverman, Ed.; Vol. 10, p. 69, Plenum, New York, 1976.
2. Lam, Y. L.; Garten, R. L. 11 Proceedings, the 6th Ibero-American Symposium on Catalysis " Rio de Janeiro, 1978
3. Vannice, M.A.; Lam., Y. L.; Garten, R. L. Advan. Chem. 1979, 178,
15.
4. Garten. R. L., Sinfelt, J. H.J. Cata!.. 1980, 62, 127.
6. Minai, Y.; Fukushima, T.; Ichikawa, M.; Tominaga, T. J. Radioanal.
Nucl. Chem., Lett. 1984, 87, 189.
7. Niemantsverdriet, J. W.; van der Kraan, A. M.; van Loef, J. J.; Delgass, W. N. J. Phys. Chem. 1983, 87, 1292.
8. Niemantsverdriet, J. W.; Aschenbeck, D. P.; Fortunato, F: A.; Delgass, W. N. J. Mol. Cata!.. 1984, 25, 285.
Fe3+ in Fe-Pt and Fe-Ru/Si02 page 82
9. Niemantsverdriet, J. W.; van der Kraan , A. M.; Delgass , W. N. J . Catal . 1984, 89, 138.
10. Niemantsverdriet, J. W.; van Kaam, J. A. C.; Flipse, C. F. J.; van der Kraan, A. M. J. Catal. 1985, 96, 58.
11. Niemantsverdriet, J. W.; van der Kraan, A. 1\/1 . Surf. Interface Anal.
1986,9,221.
12. Castner, T.; Newell, G. S.; Holton, W. C. ; Slichter , C. P. J . Phys. Chem. 1960, 32, 668.
13. Wickman, H. H.; Klein, M. P.; Shirley , D. A. J. Phys . Chem. 1965, 42, 2113.
14. Dowsing, R. D.; Gibson, J. F. J . Phys. Chem. 1969, 50, 294.
15. Tops~ , H.; Dumesic, J . A.; M~rup, S. "Applications of Mossbauer Spectroscopy" Cohen , R. L., Ed. , Vol. II , p. 55, Academic Press, New York, 1980.
16. Konings , A. J. A.; van Dooren, A. M.; Koningsberger, D. C.; de Beer, V. H. J .; Farragher , A. L.; Schuit, G. C. A. J. Catal. 1978, 54, 1.
17. Rush , J. D.; Johson, C. E.; Thomas, M. F. J. Phys. Chem. 1976, 6,
2017 .
18. Williams ; J . M.; Pearson, D. I. C. J . Phys. (Paris) 1979, C6, 401.
19. Barholomew, C. H.; Boudart , M. J. Catal. 1973, 29, 278.
The Fonnation of a Metal-Metal Oxide Interface Determined by EXAFS.
5.1 Introduction
Oxidation of bulk metals is a process that is in general well understood ( 1,2). In contrast, the oxidation of supported metal
catalysts is less well understood. In technological applications, heterogeneous catalysts need to be regenerated several times and
oxidation is an important step during the regeneration process of supported metal catalysts. Therefore, oxidation of small metal par
ticles is a process that needs to be understood better. One of the techniques which can be used to study the reduction and oxidation behavior of a catalyst is Temperature Programmed Reduction (TPR) and Oxidation (TPO) (3-10). Reduction of metal catalysts is in general a fast process and, hence, T PR is a very sensitive technique in describing the reducibility of supported metal catalysts ( 3-6). Oxidation is in general a slow process, limited by diffusion of oxygen or metal ions through the oxidic skin around the metal particles, after the oxidation process has started ( 1,2,7-10). Consequently. the usefulness of TPO is limited.
EXAFS is a technique that has proved to be very adequate in studying supported metal catalysts ( 11,12). Here, we will present the results of an EXAFS study in which we followed the oxidation of small rhodium metal particles (25 ± 5 .A) at temperatures of
0 2 Chemisorption on Rh/ A1 20 3 page 84
100 K up to 300 K. In this temperature regime, oxidation is not
complete. A careful oxidation at room temperature is in general
known as passivation (7-10.13-16). · As a result of the incomplete
ness of the oxidation, a cherry-like situation develops in which a
small metallic kernel is covered with metal oxide ( 16).
For highly dispersed catalysts, a careful analysis of the EXAFS spectra of the fully reduced sample usually indicates metal-oxygen
. distances to be present in the range of 2.6- 2.8 A ( 12,17-24).
These distances have been ascribed to 0 2- neighbors of Rh0 atoms
in the interface between metal particle and supporting oxide. Two
observations confirm this assignment . The first is that these dis
tances are always found when oxidic type supports are used (Al20 3 (12,17-21). Ti0 2 (22,23) and even X-zeolite (24)), and regardless of
the metal (Rh (12,17,18,22,23), Pt (19), Ir (20,21), Pd (24)), the
distance is always in the range of 2.6 - 2.8 It Secondly, the relative
number of 0 2- neighbors decreases with increasing rhodium particle
size ( 17 ), indicating that these Rh-0 distances are not related to
the bul.k of the metal particles but to the surface or the metal
support interface. In this study we will present additional evidence
that indeed metal-to-metal oxide interfaces can be detected with
EXAFS and that in these interfaces metal-to-oxygen distances are
present in the range of 2.6 - 2.8 A.
5.2 Experimental
A 1.9 wt% Rh/ Al 20 3 catalyst was prepared by incipient wet
ting of the Al 20 3 support (000-1.5E, Ketjen : surface area 180 m 2 g- 1• pore volume 0.65 ml g- 1
) with an aqueous solution of
RhClf3H 20 (Drijfhout) . After impregnation, the catalyst precursor
was carefully dried (24 h at room temperature, 12 h at 395 K, heating rate 5 K min-1
) . The dried catalyst was pressed into a thin self
supporting wafer with an absorbance (µ.x) of 2.5 at the rhodium
page 85 Chapter 5
K-edge. assuring an optimum signal-to-noise ratio in the rhodium
EXAFS spectra . The catalyst wafer was first reduced in an EXAFS in situ cell at 623 K (heating rate 5 K min-1
, 5 ml H2 min- 1) and
after reduction, the catalyst was evacuated at the same tempera
ture. After this, an EXAFS spectrum was recorded with the sample
at 100 K. At the same temperature, a small amount of oxygen was
admitted and after 10 min a second EXAFS spectrum was recorded .
Thereafter, the sample was allowed to warm to 300 K under oxygen
and after 10 min at 300 K, the catalyst was quickly cooled down to
100 K and a third EXAFS spectrum was recorded. The EXAFS
spectra were recorded at the synchrotron radiation source (SRS) in
Daresbury, U.K. The storage ring was operated at 2.0 GeV, the ring
current was in the range of 100-300 mA.
In order to analyze the EXAFS spectra of the catalyst, EXAFS
spectra of reference compounds are needed. From these, back-
scattering amplitude (F(k)) and phase shift ( <b(k)) functions have
to be extracted. With these functions , we can calculate EXAFS
spectra for the catalyst samples . In the calculations we have four
independent parameters, the coordination number N, the coordina
tion distance R, the Debye Waller factor /:J,.o 2 which accounts for
disorder and "£o, which allows for a correction on the position of the
adsorption edge. By changing these parameters, EXAFS spectra that fit the measured spectra best are calculated. Rhodium foil was
used as a reference for the Rh0-Rh0 contributions and Rh20 3 as a reference for Rh0-02
- and Rh 3+-0 2- contributions. The thickness of
the rhodium foil was 20 µm (corresponding to an absorbance
µ x = 1.4). For Rh20 3 a wafer with an absorbance of 2.5 was
prepared by mixing and crushing 70 mg of Rh 20 3 with 30 mg of
A1 20 3. The EXAFS spectra of the reference compounds were
recorded with the sample at room temperature.
page 86
5.3 Results
The EXAFS functions, x(k), were obtained from the X-ray absorption spectra by subtracting a Victoreen curve, followed by a cubic spline background removal ( 25). Normalization was performed by division to the height of the edge . In Figures 5.la , 5.lb and 5.lc , the raw EXAFS spectra of the catalyst after reduction and evacuation. after oxygen admission at 100 K and after warming to 300 K are shown .
For a complete description of the analysis procedure we refer to ( 17 ,23 .26 ,n ). Briefly , the analysis consisted of the following steps which led to a recurrent optimization process. For the reduced and evacuated sample and for the sample oxidized at 100 K, an RhRh EXAFS function was calculated using F( k) and <b< k) obtained from rhodium foil . The parameters N. R, D.a 2 and Eo were chosen to give the best agreement in r-space of the main peaks in a k 3
-
weighted Fourier transform. In Figures 5.ld and 5.le the k 1-
weighted Fourier transforms of the experimental data and the calculated Rh-Rh EXAFS functions are shown. Corrections were made in the Fourier transformation for the k-dependence in backscattering amplitude and phase shift using F(k) and <b< k) of the rhodium foil ( 17). From Figures 5.ld and 5.le it is obvious that apart from a Rh-Rh contribution , other contributions are present. Note that these differences are less obvious in a k3-weighted Fourier transform . The calculated Rh-Rh EXAFS spectrum was subtracted from the measured spectrum . The resulting difference spectrum contained two Rh-0 contribut ions . A two shell Rh-0 EXAFS function (using F( k)
and <f>( k) from Rh20 3) and its Fourier transform were calculated to give the best agreement with the difference spectrum and with the k 1-weighted Fourier transform of the difference spectrum (corrected for the k-dependence in the Rh-0 phase shift). In Figures 5.2a and 5.2b the k 1-weighted Rh-0 corrected Fourier transforms of the difference spectra and the calculated two-shell Rh-0 spectra are shown . Th is calculated two-shell Rh-0 EXA F S function was again
page 87 Chapter 5
subtracted from the experimental data , in order to obtain a
difference spectrum with a major Rh- Rh contribution and almost no
Rh-0 contributions. Again a best fitting Rh-Rh EXAFS function
was calculated, subtracted from the experimental data in order to
once more optimize the Rh-0 contributions. In Figure 5.2d and 5.2e
the k 1-weighted Rh-Rh corrected Fourier transforms of the experi
mental data and the Fourier transforms of the sum of the calculated
Rh- Rh and two shell Rh-0 EXAFS functions are shown . The results
of this analysis procedure are presented in Table 5.1.
For the sample that has been oxidized at 300 K, the procedure
was different. Because in this case the main contribution originated
from oxygen neighbors, a Rhn+ _0 2- EXAFS spectrum using F(k)
and cb(k) from Rh20 3 was calculated first to give the best agree
ment in the k 1-weighted Rh-0 corrected Fourier transform with the
experimental data. Figure 5.1f shows the k 1-weighted Rh-0
corrected Fourier transform of the experimental data and the calcu-
Figure 5.1 Raw EXAFS data of
(a) Rh/A120 3 after reduction and evacuation at 623 K
(b) Rh/A120 3 after oxygen exposure at 100 K
(c) Rh/Alp3 after oxygen admission at 300 K
Imaginary parts of the Fourier transforms of the original EX
AFS spectra in Figure 5.1a-5.1c (solid lines) and calculated
Rh-Rh EXAFS spectra (dashed lines). The Fourier transforms
are k 1-weighted and corrected for the Rh-Rh phase shift and
backscattering amplitude. The Fourier transform ranges in k
space are indicated in brackets.
(d) Rh/Al 20 3 after reduction and evacuation at 623 K (2 .97 -
14.58)
(e) Rh/Al20 3 after oxygen exposure at 100 K (3.26 - 12.03)
(f) Rh/Alp3 after oxygen admission at 300 K (2.55 - 12.45)
In Figure 5.1f the dashed line represents the calculated dom
inant Rh3+-0 2- contribution, this Fourier transform is k 1
!).a 2, the De bye Waller factor , is a measure for the disorder and E0
is a correction on the edge position: see ref ( 17) for more details .
lated Rh-0 2- EXAFS function. The remaining difference spectrum
could be fitted with a combination of two contributions, a Rh0-Rh0
and a Rh0-02- EXAFS function. The major contribution in the
difference spectrum originates from oxygen scatterers . Therefore, a k 1-weighted Rh-0 phase corrected Fourier transform rather than a
k 3-weighted Fourier transform (corrected for Rh-Rh phase and
backscattering amplitude) was used to optimize the different contri
butions in the EXAFS spectrum. Since in this Fourier transform an
incorrect phase shift function is used for the (small) Rh0-Rh0 contri
bution , the accompanying Rh- Rh peak is shifted and coincides with
the peak originating from the Rh0-0 2- bond of about 2.76 A. The
fact that the peak at the right hand side of the Rh3+-0 2- contribu
tion in Figure 5.1f is indeed the result of two contributions (Rh0-
0 2 Chemisorption on Rh/ Al20 3 page 90
Rh0 and Rh0-o2- ). becomes clear from a careful study of Figure
5.2c. In Figure 5.2c. the magnitude of the Fourier transform of the original data (solid line) and of the Fourier transforms of a calculated Rh3+-0 2
- + Rh0-Rh0 EXAFS spectrum (dotted line) and of a calculated Rh3+ -02
- + Rh0-02- EXAFS spectrum (dashed line) are
presented. In the Fourier transform of the former calculated EXAFS function. a strong destructive interference is visible in the region between both peaks, whereas in the latter a slightly constructive interference is observed. In the same region in the Fourier transform of the original data, a slightly destructive interference is present. Therefore, we conclude that the right-hand side peak in Figure 5.2c is the result of the sum of two contributions, namely Rh0-Rh0 and Rh0-0 2
- . In a k 1-weighted Rh-Rh phase and backscattering amplitude corrected Fourier transform, all contributions are separated, but , as mentioned above, this Fourier transform is not suitable to optimize the dominant low-Z scatterer contribution. Figure 5.2f shows once again the k1-weighted Rh-0 corrected Fourier transforms of the experimental data and the calculated {Rh-02
- + Rh0-02
- + Rh0- Rh0) EXAFS functions . The results of this analysis
are also presented in Table 5.1.
5.4 Discussion
5.4.1 Rh/ Al20 3 after Reduction and Evacuation
After reduction and evacuation of the catalyst at 623 K the major contribution in the EXAFS spectrum is from rhodium neighbors. In the difference spectrum (see Figure 5.2a) two contributions are present which both originate from oxygen neighbors.
Figure 5.2 Imaginary parts of the Fourier transforms of the difference spectra (original EXAFS spectra minus th e calculated Rh-Rh EXAFS function s. solid lines) and the Fourier transforms of the calculated two shell Rh-0 EXAFS functions (dashed lines) . The Fourier transforms are k1-weighted and corrected for Rh-0 phase shift. The Fourier transform ranges in k
space are indicated in brackets.
(a) Rh/A1p3 after reduction and evacuation at 623 K (3.46 -8.33),
(b) Rh/Alp3 after oxygen exposure at 100 K (2 .54 - 8.00)
(c) Magnitude of the Fourier transform of
solid line : Rh/ Alp3 after oxygen admission at 300 K (2.55 -12.45) ,
dotted line : Calculated Rh h- 0 2- + Rh 0-Rh 0 EXAFS function ,
dashed line Calculated Rh 3+-02- + Rh 0-02
- EXAFS func
tion . (See text for further details)
Fourier transform of the original EXAFS spectra (solid lines) and the Fourier transforms of the calculated best fitting EXAFS spectra (using the parameters presented in Table 5.1 . The Fourier transforms are k1-weighted and corrected for Rh-Rh phase s hift and backscattering amplitude The Fourier transform ranges ink-space are indicated in brackets .
(d) Rh/A1p3 after reduction and evacuation at 623 K (2 .97 -
14 .58)
(e) Rh/Alp3 after oxygen exposure at 100 K (3.26 - 12.03)
(f) Rh/Alp3 after oxygen admission at 300 K (2 .55 - 12.45)
The rhodium atoms and ions in the sample have on the aver
age 5.6 rhodium nearest neighbors. The coordination distance of
2.635 A points to zerovalent rhodium atoms having zerovalent rho
dium neighbors; the contribution originates from rhodium atoms in
metallic particles . From (28) we can estimate that the particles con
tain about 15 atoms on the average, and are roughly 10 A. in diame
ter. However. the results underestimate the size of the metal
page 93 Chapter 5
particles . For , in the EXAFS spectrum, an additional contribution from oxygen neighbors at 2.055 A is present and this contribution arises from oxidic type Rh-0 2
- bonds . From TPR studies , which are usually carried out in 5% H2 in some inert gas, it is evident that at 623 K reduction in 100% H2 should have been complete ( 17 ). During the evacuation procedure at high temperature however, the hydroxylated A120 3 surface looses water and at the high evacuation temperature, the metal particles may possibly be partly reoxidized by H20. Another possibility is that a small leakage of air into the cell has occurred during the evacuation procedure. Anyway , it is obvious that the metal particles were partly oxidized after the reduction treatment, during the evacuation procedure. In bulk Rh20 3 the Rh3+ -0 2
- distance is 2.05 A and each rhodium ion has 6 oxygen neighbors. We found a Rh-0 2
- coordination number of 0.6. The coordination numbers measured with EXAFS are averaged coordination numbers, i.e .. they are averaged over all rhodium atoms and ions present in the sample. The zerovalent rhodium atoms in the metal particles will not have oxygen neighbors at an oxidic type distance. Therefore, the EXAFS oscillations in the X-ray absorption spectrum before the normalization procedure, originating from these rhodium atoms will be proportional to the number of neighbors on metallic Rh-Rh distances The rhodium ions in an oxidic phase in the catalyst will have only oxygen neighbors in their first coordination shell at about 2.05 A. The EXAFS oscillations originating from these rhodium ions will be proportional to the number of oxygen neighbors at 2.05 A. In the normalization procedure, we divide these EXAFS oscillations by the height of the edge. However, both rhodium metal atoms (in the metal particles) and rhodium ions (in the oxidic phases) will contribute to the height of the edge jump. Therefore, the ampl itude of the oscillations in the normalized EXAFS function which originate from metallic rhodium atoms will be proportional to the number of neighbors at metallic distances divided by the total number of rhodium atoms and ions present in
the sample : A ~e~~l =A ,~e~11a1n metal/ (n meta1+n oxidic ) and therefore. N meas - N real l l I (n +11 ) ·1n wh"1ch A meas and N meas meta l - meta l meta l meta l oxidic • meta l meta l
are the measured amplitude of the EXAFS oscillations and the
0 2 Chemisorption on Rh/ Al 20 3 page 94
measured coordination number. based on these oscillations. A ,~~~11;i 1 and N r';~~ 1;i 1 are the real amplitude and the real coordination number in the EXAFS function . n nwlJI and n oxidi, are the number of rhodium atoms in the metal particles and rhodium ions in the oxidic phase that contribute the the adsorption. Thus , the coordination numbers measured by EXAFS are always fractional coordination numbers . Thus, when we want to determine the real or corrected coordination numbers in order to be able to estimate particle sizes, we will have to correct fot this . Therefore, we need to know the fraction
of rhodium atoms present in the metal particles or the fraction of rhodium ions present in the oxidic phase . Using the following procedure, we can estimate the fraction of rhodium ions in the oxidic phase. In bulk Rh 20 3 each rhodium ion has six oxygen neighbors at 2.05 A. That indicates, that the rhodium ions in the oxidic phase in the catalyst sample will have up to six oxygen neighbors . As a lower limit, we estimate that these rhodium ions will have at least four oxygen neighbors . Thus, we assume that N ~;~i( is 4-6. the
measured N ~~~iz is 0.6. thus. we find that fox= 0.6/6 - 0.6/4. This means that 15 to 10% of the rhodium atoms is oxidized (0.6/4 = 0.15 > fox > 0.6/6 = 0.10, f ox is the fraction of rhodium atoms that has been oxidized) . Because the Rh-Rh coordination number of 5.6 is also an average coordination number . the Rh-Rh coordination number of rhodium atoms ·present in metallic particles
t b t d f II . N real - N rneils / f . h" h mus e correc e as o ows . rnet;il - me.till metill in w IC
f metill = 1 - f ox = 0.85-0.90 As a consequence, the average RhRh coordination number is 6.4 ± 0.2 and the metal particles will consist approximately of 25 atoms, their diameter is about 15 A. Also, although the catalyst had been reduced at high temperature (623 K), after evacuation the sample was partly oxidized , as can be concluded from the presence of the Rh-0 2
- distance of 2.05 A.
One more contribution needs to be discussed , the 2.73 A Rh-0 contribution (cf. Table 5.1). This contribution arises from oxygen ions in the surface of the support, which are neighbors to zerovalent rhodium atoms in the metal-support interface. This type of rhodium-to-oxygen contribution has frequently been reported
page 95 Chapter 5
( 12,17-24). Like the Rh-Rh contribution, this Rh0-02- contribution
must be corrected for the presence of rhodium cations. The
corrected coordination number is 2.0/(1 - fox) = 2.3 ± 0.1. How
ever , this number does not yet represent the number of oxygen
anions in contact with rhodium atoms in the metal-support inzerface, because for three dimensional particles only p~rt of the rho
dium atoms will be present in this interface. To be able to deduce
the real number of neighboring oxygen ions for a rhodium atom in
the metal-support interface we need to know the size and the shape
of the metal particles. The size is obtained from the Rh-Rh coordi
nation number , but no information on the shape is available. Alter
natively, if we know the real number of oxygen neighbors for a
interfacial rhodium metal atom, we can determine the shape of the
metal particles . In the following we assumed that the rhodium metal
particles rest on a [111] face of the y-Al20 3 support. The size of
the rhodium atoms is about equal to the size of the oxygen ions
(the Rh-Rh distance is 2.635 A, the 0 2--02- distance is 2.80 .A), the
real Rh0-02- coordination number of a rhodium atom in the metal
support interface is 3.0. With this knowledge we can make a model
for the size and shape of the metal particles.
5.4.2 A Model for the Oxidation of Metal Particles
In th is paper we will try to describe the oxidation process of
small metal particles . At low temperatures, oxygen readily adsorbs
and even partially dissociates on rhodium metal (29). Considering
the fact that we study very small metal particles, oxidation may be
a relatively fast process. In the following, we will assume that the
metal atoms in the particle with the least negative binding energy
(i. e. atoms which are the most coordinative unsaturated) will be
oxidized first. We have calculated the cohesive energy of every atom in the metal particle by summation of the Lennard-Jones energy
contributions of all the other atoms in the metal particle. The
0 2 Chemisorption on Rh/ Al 20 3 page 96
Lennard-Jones binding energy is given by
E(R) = Al-1- - Ji_ R12 R6 (5.4.1]
where R is the interatomic rhodium-rhodium distance. The constant A in the expression is not important when one compares binding energies of the same type and has therefore been taken equal to unity. B determines the interatomic distance at which the minimum in the binding energy occurs and has been given the value of 0.005989 A.6. This value corresponds to Rmin = 2.635 A, which is equal to the Rh-Rh distance in the metal particles (see Table 5.1). Thus, the relative binding energy of atom i in the metal particle has been calculated by using
E= L i ;"= j
1 R.12
l j
5.989 10- 3
RiJ (5.4.2]
In most cases, the R i) 12 term is negligible compared to the
Ri)6 term . Therefore, a model in which only the number of nearest rhodium neighbors is taken as a measure for the binding energy of a given atom gives nearly the same result as the model described above, which takes into account the long range interactions as well. In Table 5.2 the calculated binding energies and the number of rhodium nearest neighbors is given during a hypothetical oxidation process in which the rhodium atoms, one by one, become 'oxidized ' to Rh . Once a rhodium atom is oxidized, it will preferentially be surrounded by 0 2
- ions. Hence, it will not contribute to the EXAFS Rh0-Rh0 coordination number and it will also not contribute to the binding energy of the remaining rhodium atoms in the metal particle. After an atom has been oxidized (i.e., removed from the actual cluster) , the new Rh0-Rh0 coordination number, the binding energy and the number of rhodium neighbors for each rhodium atom in the remaining particle were re-calculated.
page 97 Chapter 5
Table 5.2 Lennard-Jones binding energies (E , a u.) and number of rhodium neighbors (N) for each rhodium metal atom during a hypothetical oxidation process
Atom Number of atoms oxidized : nrs 0 1 2 3 4 5 6 7 8 9
In Figure 5.3, a 26 atom metal particle resting on a [111] yAl203 face is drawn. The averaged Rh- Rh coordination number is
6.38. The particle is assumed to be grown epitaxially on an alumina
(111] crystal face, which according to (JO) is the most stable crystal
face. In the 26 atom metal particle in Figure 5.3, atom number 5 and atom number 9 have the lowest binding energy. The atoms
with the next lowest binding energy are atoms 1 and 4. With
fox = 0.10 - 0.15, it follows that about 3 rhodium atoms in this
particle were in the oxidic phase. Removing (arbitrarily) atom 5
from the particle induces a new situation in which atom 1 has the
lowest binding energy in the 25 atom metal particle (see Table 5.2, column under '1'). After subsequently removing atom 1. atom 9 has
(again) the lowest binding energy. When atom 9 is removed as well,
the Rh- Rh coordination number averaged over all 23 atoms and 3 ions equals to 5.54 (i. e. , the coordination number as it would have
page 99 Chapter 5
been measured with EXAFS), in excellent agreem~nt with the experimentally observed value of 5.6. However , after removing atom 9, atom 4 obtains a very low binding energy (Table 5.2 . column under '3') . We therefore assume that when atom 9 is oxidized, atom number 4 will immediately be oxidized as well . As a consequence, the situation after reduction and evacuation will be represented by an average of two situations, one in which two atoms (atom 1 and 5 or 9 and 4) and one in which four atoms (atoms 1, 5, 9 and 4) are oxidized . The situation in which three atoms are oxidized is unlikely to occur. In Figure 5.4a and 5.4b, a bare 24 and 22 atom metal particle are shown. For these metal particles , the calculated Rh0-0 2
-
coordination number is on the average equal to 1.5 (1.38 and 1.62 respectively; note that these are already average coordination numbers), which is too low compared to the measured value of 2.0. However, up to now we have only considered 0 2
- anions of the support which are in contact with the remaining metallic rhodium atoms . But the rhodium oxide phase which has been formed may well be in contact with the remaining metal particle and this would explain the higher Rh0-0 2
- coordination number. In Figure 5.4c and 5.4d, a possibility is presented in which several oxygen ions take the place of the oxidized rhodium atoms. Using that configuration, the calculated and averaged Rh0-02
- coordination number is 2.0. We have assumed that the oxidation process leads to stoichiometric Rh 20 3 and thus, in Figure 5.4 (as well as in Figures 5.5 and 5.6) three 0 2
- ions are incorporated for every two rhodium atoms which are assumed to be oxidized to Rh3+. The Rh3+ ions are not shown in these Figures. They most probably reside in the octahedral or tetrahedral sites present between the oxygen ions from the support and the oxygen ions in the rhodium oxide phase.
By varying the assumptions, we found that is was impossible to reproduce the measured Rh0-02
- coordination number of 2.0 with situations which are principally different from the situation depicted in Figures 5.4c and 5.4d . In all cases, it was essential that .rhodium oxide was in contact with the metal particle. We therefore assume that Figures 5.4c and 5.4d give a fairly accurate description of the
0 2 Chemisorption on Rh/ Al 20 3 page 100
Figure 5.4 Model of the rhodium metal particles of y-A20 3 after evacua-tion .
(a) Bare 24 atom rhodium metal particle
(b) Bare 22 atom rhodium metal particle
(c) 24 Atom rhodium metal particle in contact with Rhp3
(d) 22 Atom rhodium metal particle in contact with Rhp3
a
\ ..... ,,X., ... > y
/' ;,,.,,. ........ l~: ...... _ f '---~._... ........ __,.........__/ ' '::-.......... ·· ' "····~. y
metal particles in the catalyst after reduction and evacuation . As a
consequence, the 26 atom metal particle in Figure 5.3 will give a good representation of the average metal particle in our Rh / A1 20 3 sample directly after reduction.
c
d
page 101 Chapter 5
5.4.3 Rh/ Al20 3 after Oxygen Admission at JOO K
After admission of oxygen to the catalyst at 100 K, the
EXAFS spectrum changed drastically (see Figure 5.1). The Rh0-Rh0
contribution drastically diminished and the contribution of the
oxidic type of Rh-0 2- bonds increased in magnitude (cf. Table 5.1).
Clearly, oxidation has occurred.
We continue to describe the oxidation process as indicated in
the foregoing section. Assuming that the rhodium ions in the oxide
phase have 4 to 6 oxygen neighbors, we find that
0.23 < fox < 0.35, indicating that 6 to 9 rhodium atoms have been
oxidized. Thus, the remaining metal particles contain 17 to 20 rho
dium atoms. Using the same procedure as described above, we find
that successively atoms (5 , 1, 9, 4.) 14. 15 and 16 will be oxidized
(i. e. 'removed' from the metal particle) . The calculated and aver
aged Rh0-Rh0 coordination number is 4.46, which is higher than the
measured value of 4.1 . In this situation , atoms 10 and 13 both have
a low binding energy . Removing only atom 10 or 13 gives a 19
atom metal particle with a Rh 0-Rh0 coordination number of 4.15, in
very good agreement with the measured value. However. because of
their low binding energy, it is more likely that both atoms 10 and 13
are oxidized at the same time . In that situation, the Rh0- Rh 0 coordi
nation number is 3.85. Again. we assume that reality is an average
of two or possibly three situations . In all cases, atoms 5, 1, 9, 4, 14, 15 and 16 are oxidized. In addition. (a) no atoms. (b) one atom (10 or 13) or (c) both atoms 10 and 13 may be oxidized . The average
Rh0-Rh0 coordination number will be somewhere between 3.85 and
4.46. With the two situations most likely to occur, situations (a)
and (c), in a one-to-one ratio , the average · Rh0-Rh0 coordination
number would be equal to 4.1, and would be in excellent agreement
with the experimental data.
0 2 Chemisorption on Rh/ Al20 3 page 102
The calculated Rh0-02- coordination numbers (averaged over
all 26 rhodium atoms and ions) of the 19. 18 and 17 bare atom
metal particles (Figures 5.5a. 5.5b and 5.5c) are 1.04. 0.93 and 0.81
respectively. They are too low compared to the measured value of 1.4 . Figures 5.5d, 5.5e and 5.5f give a representation of a possible
arrangement of rhodium oxide around a remaining 19. 18 and 17 rhodium metal particle after 7, 8 and 9 atoms have been oxidized.
The calculated Rh0-Rh0 coordination number (4 .15. the average over 4.46. 4.15 and 3.85 for the 19, 18 and 17 atom metal particle)
agrees very well with the measured value (4.1), and the calculated and measured Rh0-02
- coordination numbers (1.5 and 1.4 respec
tively) also agree nicely. Clearly. at least part of the rhodium oxide is in contact with the metal particle. Obviously, the interface
between rhodium metal particle and rhodium oxide is rather stable
under the experimental conditions.
5.4.4 Rh/ Al20 3 after warming up to 300 K under oxygen
After oxygen admission at 100 K, the sample was warmed to 300 K under oxygen atmosphere . The resulting spectrum (Figure
5.1c) was completely different form the two other spectra. The most dominant contribution is now from oxidic Rh-0 2
- bonds and
almost no Rh0-Rh0 contribution is left (see Table 5.1) .
From the measured Rh-0 2- coordination number, we calculate
that 0.60 <fox < 0.90. However, because oxidation has progressed that far, we expect that the number of oxygen neighbors in the rho
dium oxide phase will be close to 6 and therefore the fraction of oxidized rhodium atoms will be close to 0.6. That means that about 16
atoms will be oxidized and about 10 atoms will remain metallic. Our
model calculations indicate that during the oxidation as observed,
atoms number 17, 19. 20, 23, 24, 25 and 26 will be oxidized as well.
The calculated Rh0-Rh0 coordination number of the remaining .
Figure 5.5 Model of the rhodium metal particles of y -Al 20 3 after oxida-tion at 100 K
(a) Bare 19 atom rhodium metal particle
(b) Bare 18 atom rhodium metal particle
(c) Bare 17 atom rhodium metal particle
( d) 19 Atom rhodium metal particle in contact with Rhp 3
( e) 18 Atom rhodium metal particle in contact with Rhp 3
(f) 17 Atom rhodium metal particle in contact with Rhp 3
particle , averaged over all 26 rhodium atoms , is 1.85, is good agree
ment with the measured value of 1.9. (The Rh0-Rh0 coordination
number averaged only over the 10 atoms in the remaining metallic
kernel of the particle, the corrected coordination number, is 4.8.)
We once more found it imperative that at least part of the rhodium
oxide formed is in contact with the metal particle and compared to the situation in the preceding paragraph, the rhodium metal-to
rhodium oxide interface has even increased in magnitude. In Figure
5.6a, a bare 10 atom metallic kernel is shown . For this metal parti
cle, if not covered with rhodium oxide, the Rh0-0 2- coordination
number is 0 .81, which is about a factor two lower than the meas
ured value (1.7) . Figure 5 .6b represents a plausible arrangement of
rhodium oxide around the 10 atom metallic kernel.
5.4.5 General Remarks
In the discussion above, we used the Rh0-Rh0 coordination
numbers determined with EXAFS to estimate the size of the metal
particles and the Rh-02- coordination numbers to estimate the
amount of oxidized rhodium. The Rh0-02- coordination numbers
were used to estimate the extent of the interface between metal and
oxide. The accuracy of the latter coordination number, however, is lower than that of the others. The reason for this is shortly as fol
lows. We used a Rh0-Rh0 absorber-scatterer pair (R = 2.687 , the
page 105 Chapter 5
Figure 5.6 Model of the rhodium metal particles of y-A1p 3 after oxidation at 300 K.
(a) Bare 10 atom rhodium metal particle
(b) 10 Atom rhodium metal particle in contact with Rhp3
a
Q Rh atom
Rh-Rh distance in bulk rhodium) to extract F(k) and (/>(k), and
used these to calculate Rh0-Rh0 EXAFS spectra with R = 2.63-2.645 A, very close to the reference . The Rh-02
- EXAFS functions
(with R = 2.03-2.055 A) were calculated using the absorberscatterer pair Rh 3+ -0 2
- (R = 2.05 A). The coordination distances in
calculated EXAFS function and reference compound differ only
slightly and, therefore, also this calculation is reliable. However, to
use the same absorber-scatterer pair to calculate Rh0-0 2- EXAFS
spectra with R = 2.7 A is not correct. Mainly because of the longer
distance, the calculated coordination number underestimates the
real coordination numbers . The extent of underestimation may be
as much as 20% ( 17,3 0 . However , this does not affect our main
conclusion that during oxidation a rhodium-rhodium oxide interface is formed. Since the real Rh0-0 2- coordination numbers will be
higher than the values reported in Table 5.1, the metal-oxide interfaces will even be larger and the coverage by rhodium oxide on the
metal particles will be more complete.
b
0 2 Chemisorption on Rh/ A1 20 3 page 106
In the above discussion it was assumed that energetic con
siderations played the key role during the oxidation process. How
ever, as oxidation proceeds , the metal particles become covered with
rhodium oxide. As a consequence, diffusion will sooner or later
become the rate limiting step in the oxidation process. In the situa
tion described above, we assumed that rhodium atoms with the
lowest binding energy will be oxidized. The binding energy calcula
tions indicate that metal atoms which become oxidized have a (rela
tive) binding energy less than 4.1 (or less than 5 rhodium nearest
neighbors) . In the metal particles in Figure 5.3, 5.4 and 5.5, all rho
dium atoms which have this low binding energy, and which are to
be oxidized in the next step, are exposed to the gas atmosphere and
therefore, a kinetic limitation of the oxidation process is most
unlikely. In Figure 5.6, a 10 atom metal particle is shown. In this
metal particle, 6 rhodium atoms have a binding energy lower than
4.1 (and less than 5 rhodium neighbors). Therefore, these are can
didates to be oxidized in the next stage. They are, however, (partly)
covered with rhodium oxide and oxidation may therefore be hampered by diffusion. The three top rhodium atoms still have a bind
ing energy of 4.9, far above the limit (4.1) we used in order to
establish whether an atom will be oxidized. Thus, the situation in
Figure 5.6b may represent a stable situation in which oxidation will
proceed only after elevating the temperature . Obviously, up to tem
peratures of about 300 K. oxidation is not limited by diffusion under
the experimental conditions : the oxidic skin is too small to screen
the metallic kernel from the surrounding gas phase.
In most cases, during oxidation, the rhodium atom that was to
be oxidized in a next stage of oxidation had a binding energy in the range of 4 .0 . In some cases, removing one rhodium atom resulted in
a severe decrease in binding energy of one of its direct nearest rho
diu m neighbors. In those cases , we assumed that this weakly bound
rhodium atom would be oxidized immediately. We never encoun
tered situations in which more than two atoms would be oxidized at
the same time. Therefore , we conclude that oxidation has to be a rather smooth and straightforward process.
page 107 Chapter 5
Even after oxidation (passivation) at 300 K, some clean metal
surface is still exposed (see Figure 5.6b): the Rh0-0 2- coordination
number was too low to account for complete coverage. The binding
energy of the three rhodium atoms on top is 4 .9 and they are there
fore rather stable. This may very well explain why passivated metal
particles are very easily reduced : the metallic kernel adsorbs and
dissociates hydrogen even at low temperature and because of the
dissociated (activated) hydrogen, reduction of the neighboring oxide
phase is very fast and can proceed at temperatures far below the
reduction temperature of the bulk oxide.
The rhodium metal particles and rhodium oxide phase have
been grown epitaxially on the Al20 3 [111] crystal face, which is not unlikely for small metal particles. nor for oxide particles. It has been
shown that rhodium oxide grows epitaxially on rhodium [111). According to Castner and Somorjai ( 32) Rh20 3 [0001 J fits on Rh
1111) when the rhodium oxide unit cell expands by about 5%. The
strain induced by this mismatch has only a very small influence on
the cohesive energy calculated by using the Lennard-Jones poten
tials. Rh20 3 has the same corundum structure as A1 20 3. Epitaxial
growth of rhodium oxide on the alumina support is also not
unlikely. Altogether, the assumptions made in the model described
above, epitaxial growth of rhodium particles on the alumina support
and epitaxial growth of rhodium oxide on both rhodium metal parti
cles and alumina support, seem justified .
We used only the rhodium atoms in the metal particle in calcu
lating the Lennard-Jones binding energies of the rhodium atoms:
the influence of neighboring oxygen ions has been neglected. The
additional contribution of oxygen neighbors is expected to be small
and not to influence the results significantly. At the moment we are
studying the binding of rhodium metal particles on Al20 3 surfaces using the AS ED-MO method described by Anderson et al. (33) and
we expect to be able to quantify the contribution of oxygen ions to
the binding energy in the near future .
0 2 Chemisorption on Rh/ A1 20 3 page 108
In the discussion above we described the oxidation of a 26 atom rhodium metal particle. We applied the same procedure to metal particles ranging from 22 to 30 atoms per particle in order to find out whether the observed trends also hold for these particles. In all cases we found no major differences, and the results presented above and the conclusions to be presented below are representative for the cases we have studied.
5.5 Conclusions
Using Rh0-Rh0, Rh0-02- and Rh-02- coordination numbers determined with EXAFS. a model has been derived to describe the oxidation of small rhodium metal particles supported on A1 20 3.
After reduction, the metal particles were about 15 A in diameter and contained about 26 rhodium atoms; after evacuation, the particles were partly oxidized, either by water formed during evacuation or by leaking in of oxygen. and the metal particles contained about 22 to 24 atoms. After oxygen admission at 100 K, oxidation had proceeded and the metal particles contained about 17 to 19 rhodium atoms. After warming the catalyst under oxygen to 300 K, the metal particles contained only about 10 rhodium atoms . The Rh0
-
02- coordination numbers indicated that in all cases the rhodium oxide formed was in contact with the metal particles . The remaining metallic kernel was partly covered with rhodium oxide. We therefore conclude, that under the experimental conditions, the rhodium-rhodium oxide interface is more or less stable. Our main conclusion is that metal-support interfaces indeed can be detected with EXAFS and that the distance between interfacial zerovalent rhodium atoms and divalent oxygen atoms is about 2.7 A, in accordance with literature data.
page 109 Chapter 5
5 .6 References
1. Cabrera, N. "Semiconductor Surface Physics" ; Kingston , R.H ., ed .; J. Wiley & Sons, New York 1960.
2. Hauffe, K. "The Surface Chemistry of Metals and Semiconductors" ;
Gatos , H.C. , Ed .; J. Wiley & Sons , New York 1960.
3. Robertson , S. D.; McNicol , B. D.; de Baas , J. H.; Kloet, S. C. ; Jenkins, J. W J. Catal . 1975 , 37 , 424.
4. Jenkins , J. W.; McNicol, B. D.; Robertson , S. D. Chem. Tech 1975, 7, 316.
5. Wagstaff, N.; Prins , R. J. Catal . 1979, 59, 435.
7. Vis , J. C.: van 't Blik, H. F. J.; Huizinga, T.; van Grondelle, J.; Prins , R. J. Catal . 1985, 95 , 333
8. van 't Blik , H. F. J.; Prins, R. J . Catal . 1986, 97, 188
9. Martens . J. H. A. ; van 't Blik, H. F. J.; Prins , R. J. Catal . 1986, 97 , 200
10. van 't Blik , H. F. J. ; Koningsberger , D. C.; Prins , R. J. Catal. 1986, 97, 210
11. Sinfelt , J. H.; Via , G. H. ; Lytle , F. W. J. Chem. Phys. 1917, 67 , 3831 .
12. van Zon , J. B. A. D. ; Koningsberger , D. C. ; van 't Blik, H. F. J.; Prins ,
R.; Sayers , D. E. J . Chem. Phys. 1984 , 80 , 3914.
13. Uchijima, T. ; Herrmann , J. M .; Inoue, Y.; Burwell. R. L. Jr.; Butt , J. B.; Cohen, J. B. J. Catal . 1977, 50 , 464.
14. Koboyashi , M .; Inoue . Y ; Takahashi , N.; Burwell , R. ,L_ Jr .; Butt , J. B.; Cohen, J. B. J. Catal . 1980, 64, 74.
15. Nandi, R. K.; Georgopoulos , P ; Cohen , J. B.; Butt , J. B.; Burwell , R.
L. Jr. J . Catal . 1982, 77 , 421 .
16. Huizinga , T ; van Grondelle, J.; Prins , R. A ppl . Catal. 1984, 10, 199
17. van Zon , J. B. A. D.; Koningsberger , D. C. ; van 't Blik, H. F. J. ;
Sayers, D. E. J . Chem. Phys . 1985, 12, 5742.
0 2 Chemisorption on Rh/ Al 20 3 page 110
18. Koningsberger , D. C. ; van Zon , J . B. A. D.; van 't Blik , H. F. J .; Mansour, A N.; Visser , G. J .; Prins , R.; Sayers , D. E.; Short, D. R.;
Katzer , J. R. J. Chem. Phys. 1985, 89 . 4075
19. Koningsberger , D. C. ; Sayers , D. E. Solid St . Jon . 1985, 16, 23
20. van Zon. F. B. M .; Visser , G. J .; Koningsberger , D. C. 9th Interna
tional Congress on Catalysis , Calgary . 1988. to be published
21. Koningsberger, D. C.: Duivenvoorden , F. B. M.; Kip, B. J.; Gates , B. C. "EXAFS and Near Edge Structure"; Lagarde, P.; Raoux , D.; Petiau, J Eds.; Les Editions de Physique , 1986; vol. 1, p. C8-255.
22 . Koningsberger, D. C.; Martens .. J. H. A.; Prins, R.; Short, D. R.;
Sayers , D. E. J. Phys . Chem. 1986, 90 , 3047 .
23. Martens , J. H. A.; Prins , R.; Zandbergen, H.; Koningsberger , D. C.;
accepted for publication in J . Phys . Chem ..
24 . Moller , K .; Bein , T. "EXAFS and Near Edge Structure "; Lagarde, P.; Raoux, D.; Petiau , J. Eds .; Les Editions de Physique , 1986; vol. 1,
p. C8-231.
25. Cook, J. W.; Sayers, D. E. J. Appl. Phys. 1981, 52, 5024 .
26. Duivenvoorden, F. B. M. ; Koningsberger, D. C.; Uh, Y. S.; Gates , B. C. ]. Am. Chem . Soc . 1986, 108, 6254
27 . van 't Blik, H. F. J .; van Zon , J . B. A. D.; Huizinga, T. ; Vis , J. C.;
Koningsberger , D. C. ; Prins , R. J . Am. Chem. Soc. 1985, 107, 3139.
28. Kip, B. J. ; Duivenvoorden , F. B. M .; Koningsberger, D. C. ; Prins , R. J . Catal . 1987, 105 , 26.
31. Stern, E. A .; Bunker , B. A.; Heald, S. M . Phys. Rev . B 1980, 21 ,
5521.
32. Castner, D. G. ; Somorjai , G. A. Appl. Surf. Sci. 1980, 6 , 29.
33. Anderson, A. B.; Ravimohan, Ch.; Mehandru, S. P Surf. Sci. 1987,
183, 438.
page 111 Chapter 6
Chapter 6
The Structure of the Metal-Support Interface in Rh/ Al20 3 Determined with the ASED-MO Method
6.1 Introduction
In Extended X-ray Absorption Fine Structure ( EXAFS) spectra
of the metal edge of fully reduced supported metal catalysts.
metal-oxygen contributions are present . These contributions have
been ascribed to metal atoms in the metal -support interface having
divalent oxygen ions of the support as neighbors . This assignment
was based on the following observations . Firstly , this contribution
is always observed when oxidic type of supports are used (A1 20 3,
Ti02. X-zeolite) . and regardless of the metal , the corresponding
metal-to-oxygen bond distance is always in the range of 2.6-2 .8 A ( 1-7 ). Secondly , the relative number of oxygen neighbors decreases
with increasing metal particle size ( 2), indicating that these contri
butions are not related to the bulk of the metal particles. Hence, it
was concluded that the metal-oxygen contributions originated from
the metal-support interface. The metal-oxygen bond length (2.6-
2.8 A, depending on support and metal) was explained assuming
that the metal atoms and oxygen ions could be regarded as hard
spheres with radii equal to the radius of the metal atom and the
radius of divalent oxygen ions respectively , both in the range of
1.3-1 .4 A. Since EXAFS is the only technique that has been able to identify these contributions , confirmation by another, independent
technique seemed desirable. In this study we will present theoretical evidence for these metal-oxygen coordination in the metalsupport interface.
Rh/A120 3 characterized with AS ED-MO page 112
6.2 Theoretical Method
We used the atom superpos1t1on and electron delocalization
molecular orbital (ASED-MO) method to study the binding of rho
dium metal particles to y-Al 20 3. The theory has been outlined in
chapter 2 and has been described extensively in the literature (8,9).
Therefore, we will only give a brief summary of the basics of the
method. The theory is based on a partioning of the electronic charge density function (p) of a molecule, or in this case a cluster,
into rigid free atoms parts (Pi) which center on the nuclei and fol
low them perfectly, and a non-perfectly following or bond charge
density function (Pnpd :
atoms
L Pi + Pnpf !6.2.l]
The force on nucleus b that is due to Pa and nucleus a is repulsive
because the repulsion component of this force increases more
rapidly than the attractive force of Pa due to the penetration of the
Pb charge cloud. The force on nucleus b due to Pb is zero. Hence,
as the force on nucleus b that is caused by atom a is integrated, a
repulsive energy ER• is obtained. The Pnpf density will cause an
attractive force on nucleus b and thus, its integral is an attractive
energy, Enpf· The binding energy curve is then the sum of two
nonzero components :
E !6.2.2]
However, since Pnpf is not a know function, Enpf cannot be obtained
from an integral of the force it causes on a nucleus . Nevertheless,
the electron delocalization energy can be well approximated by a molecular orbital delocalization energy in most instances (8,9) .
page 113 Chapter 6
E ::::::: ER + ~mo [6.2.3]
were b..£ 1110 is the total one-electron molecular orbital 1energy minus the one-electron atomic orbital energies. The appropriate oneelectron Hamiltonian is a function of experimentally determined valence state ionization potentials (VSIP), valence Slater atomic orbital overlap integrals and internuclear distances . For aluminium and oxygen, these data have been taken from the Al20 3/Ni and Al 20 3/Pt studies by Anderson et al. ( 10,11 ). For rhodium, these data have been taken from the compilation of Lotz ( J 2), from the work of Basch ( 13) and the work of Moore ( 14). These parameters are listed in Table 6.1. The ionization potential for the rhodium 4d orbital had to be adjusted in order to obtain physically relevant results . The Fermi level for the bare y-Al20 3 cluster was -10.60 eV. For the bare rhodium duster, with the 4d VSIP equal to the value found in the literature, -9.56 eV, the fermi level was -9.24 eV. As a result, when the total energy of the metal cluster plus the Al 20 3 support was calculated, there was a net charge of approximately + 10 on the rhodium metal particle due to the difference in Fermi levels, independent of the distance between metal particle and support cluster. This is not only a physically irrelevant situation, but in addition it did not give rise to a situation in which bonding between metal particle and support cluster occurred. The rhodium 4d VSIP has therefore been taken equal to -10.95 eV. In that case, the metal particle far away from the support cluster had a zero net charge. For more negative values for the rhodium 4d VSIP , the metal cluster obtained a negative net charge of -10 and also in that case, no situations were found in which bonding occurred between rhodium metal particle and Al20 3 support clus ter.
Rh/ Al 20 3 characterized with AS ED-MO page 114
Table 6.1 Parameters used in the calculations principal quantum
Atom
Rh s p
Al s p
0 s
p
number n, Slater exponent ~ (a .u.) , ionization potential IP (eV) and the coefficients for the double~ d orbitals .
We used the AS ED-Mo method to calculate the total energy of
a y-Al 20 3 cluster of 54 oxygen atoms , 32 aluminium and 12 hydro
gen atoms. A 10 atom rhodium metal particle was places on this
cluster. The A1 20 3 cluster has been composed using the structural
data from Lippens (JS) and Knozinger ( 16). The rhodium metal
particle had an fee structure and it consisted of 7 atoms in the
metal support interface and 3 atoms on top. Both clusters are
shown in Figure 6.1a. The rhodium-rhodium distance in the metal
particles has been taken equal to the oxygen-oxygen distance in the
[111) Al 20 3 surface plane (2 .80 A) , which is almost equal to the
rhodium-rhodium distance in bulk rhodium (2 .687 A) . As a result , the metal particle could be fitted epitaxially on the support cluster .
From Figure 6 .1a it can be seen that the interfacial rhodium atoms
may rest on different sites. When we only take the oxygen ions
into account , we can discern one-fold (r on-top), two-fold and
thre~fold coordinated sites . These sites are illustrated in Figure
6.1b. Because of the symmetry, these sites are the same for all
seven interfacial metal atoms. The on-top positions 1 and 5 in Fig
ure 6.1b are in theory the same. In practice , they differ slightly
page 115 Chapter 6
because of the limited dimensions of the support cluster. In addi
tion, there is one two-fold and there are three-fold coo~dinated sites.
In one of these two three-fold sites, there is an oxygen ion of the
second layer directly beneath the site (position 4 in Figure 6 .1b) , for
the other three-fold site this is not the case (position 2 in Figure
6.1b).
Figure 6.1 Model for the A1p3 support cluster and the rhodium metal
particle
(a) the support cluster and the rhodium particle
(b) a schematic representation of the five sites · 1 = on-top , 2 = 3-fold, 3 = 2-fold , 4 = 3-fold and 5 =on-top.
Rh
1 2 3 4 5
b
Rh/ Al 20 3 characterized with AS ED-M 0 page 116
The total energy of the system has been computed as a f unction of the height of the cluster above the Al 20 3 support cluster . The range in height was chosen such that the Rh-0 bond lengths ranged from 2.0 to 3.0 A. Each curve contained 10 points. The value of the total energy of the system with the particle at 10 A above the cluster was taken as zero energy . In all cases, th is energy was equal to the sum of the total energy of the bare metal particle and the bare support cluster. The min ima in these Lennard-Jones type of curves represent energetically stable situations and the accompanying binding energy and Rh-0 bond length could be calculated from these minima. Such curves have been calculated for the metal particle above the five sites shown in Figure 6.1b.
In practice, the surface of the y-Al20 3 support is a hydroxylated surface ( 15.16). In fact , the chemical formula for y-Al 20 3 is Al20 3·nH 20 . Therefore, 12 protons have been incorporated in the Al20 3 support cluster. These protons were assumed to be located on top of oxygen ions in the first layer, thus resembling surface hydroxyl groups. The 0 -H bond length has been obtained by varying the 0-H bond length between 0.8 and 1.5 A for the bare support cluster. A sharp minimum was found at a bond length of 0.95 A. There are several possibilities to distribute these 12 protons over the 27 oxygen ions in the surface of the support cluster. We have studied three cases. In situation A, no protons were present under the rhodium metal particle. The protons were situated on the edge of the support cluster . In situation B, the protons were distributed evenly over the support, such that the rhodium atoms in the metalsupport interface had both bare oxygen ions and hydroxyl groups as direct neighbors . In situation C, the protons were located on the oxygen ions in the middle of the surface of the support cluster and the rhodium atoms in the metal-support interface had only hydroxyl groups as nearest neighbors.
page 117 Chapter 6
Thus, we have calculated total energy curves as a function of height of the rhodium particle above the support cluster for 15 cases, in the three situations A. B and C as mentioned above. and above the 5 sites depicted in Figure 6.1b. The results of these calculations are presented in Table 6.2 .
6.4 Discussion
From the results in Table 6.2, it is obvious that the most stable arrangement is the one in which the interfacial rhodium atoms rest in 3-fold sites of surface hydroxyl groups (sites 2 and 4 in situation C) . When we compare this to the metal particle in a 3-fold site in situation A, it is obvious that the interaction between metal particle and support cluster has decreased drastically : the total binding energy decreased from 146.3 to 85 .9 kcal mol-1 {for site 2) . In situation A, the fully dehydroxylated surface, on-top sites are favored. The binding per Rh-0 bond in an on-top site in situation A is very strong : 20.8 kcal moi-1
. This bond strength is almost equal to the sum of the three bond strengths in sites 2 and 4 in situation C, the fully hydroxylated surface : 20.9 and 20.5 kcal mol-1. Obviously, the protons have a pronounced influence on the binding of the metal particle to the support. When the interfacial rhodium atoms are situated on top of hydroxyl groups, the protons are situated in between the oxygen ions and the rhodium atoms . This situation is clearly very unfavorable : the binding energy of a rhodium atom on top of a bare oxygen ion is approximately 20.8 kcal mol- 1
• on top of a hydroxyl group the binding energy has decreased to 11.3 kcal mol 1
-. Above 3-fold sites, however, the situation is reversed : the binding per rhodium atom and per Rh-0 bond increased when a bare oxygen ion was replace by a hydroxyl group. The increase is such that 3-fold sites on a hydroxylated surface are even preferred above on-top sites of a dehydroxylated surface . Note, that the binding energy per Rh-0
Rh/ A1 20 3 characterized with AS ED-MO page 118
Table 6.2 Final results of the ASED-MO calculations on a 10 atom Rh cluster above a A120 3 support cluster with 54 0 , 32 Al and 12 H atoms .
site 1 site 2 site 3 site 4 site 5 on-top 3-fold 2-fold 3-fold on-top
H min the height of the metal particle above the support cluster in the minimum of the energy-height curve
E,01 The net binding energy in kcal mor1 in the minimum of the energy-height curve (£101 = E,0 ,(H =H min) - E,0 ,(H = 10A))
RRh-0 the rhodium-oxygen bond length
E Rh the binding energy in kcal per mol interfacial rhodium atoms
(=E101/7)
E Rh- 0 the binding energy in kcal per mol rhodium-oxygen bonds (=E 101/7n in which n= 1, 2 or 3, for 1-fold , 2-fold and 3-fold sites respectively)
A All 12 hydrogen atoms on the edge of the Al20 3 cluster : no -OH ions underneath the rhodium metal particle.
B All 12 hydrogen atoms evenly distributed over the Alp3 support cluster
C All 12 hydrogen atoms in the middle of the Al20 3 support cluster : only -OH ions underneath the rhodium metal particle
page 119 Chapter 6
bond in an on-top site is still larger th an the binding energy per
Rh-0 bond in a three-fold site. However, since in a three-fold site
three Rh-0 bonds per interfacial rhodium atom are present, the
three-fold sites are favored above the on-top sites.
For supported metal catalysts . metal-oxygen distances in the
metal-support interface ranging from 2.6-2.8 A have been reported.
For Rh/ Al 20 3 catalysts it has been found that the rhodium atoms in the metal-support interface have two to three oxygen neighbors
( 2) . Thus, the results from this AS ED-MO study agree nicely with
the results from the EXAFS studies. We also found that on a fully
dehydroxylated surface on-top sites were preferred and that the
accompanying Rh-0 bond length has decreased to 2.1 A. It is very difficult to study catalysts under fully dehydroxylated conditions
with EXAFS, since the H20 vapor pressure under high vacuum con
ditions in an EXAFS cell is high enough to sustain a fully hydroxy
lated y -Al20 3 surface. However, in the EXAFS study on lr/A120 3 ( 17), in which the catalyst had been evacuated at high tempera
tures, a decrease in lr-0 distance in the metal-support interface was
reported : after the evacuation procedure the lr-0 distance had con
tracted from 2.6 to 2.15 A. This is of course in very good agree
ment with the Rh-0 bond length calculated for the meta l particle on a fully dehydroxylated surface.
It is clear that protons play a key role in binding a rhodium
metal particle to the support. What the influence of the protons
exactly is , is a matter that has to be studied more closely . The
intention of this study merely was to show that indeed the Rh-0 distances that have been detected before with EXAFS are
trustworthy. The model which has been used up to now to explain
these Rh-0 distances , which assumed that the rhodium atoms and
oxygen ions can be regarded as hard spheres, however , is shown to be incorrect.
Rh/ Al 20 3 characterized with AS ED-MO page 120
Although the theoretical results agree nicely with the experi
mental data. their reliability still has to be ascertained. The results may be model-dependent. For instance. the position of the protons
is a matter that may be of importance. They may even be placed inside the Al20 3 support cluster . Or we could leave them out .
Secondly. the rigid structure of the metal particle and the cluster can be questioned. It might be better to optimize the position of
each rhodium atom separately and the position of each proton separately. Unfortunately, such an optimization would be virtually
impossible with this method. In addition to this model-dependence, the method · does not take electrostatic and van der Waals forces
into account . Although the support cluster and the metal particle were neutral and van der Waals forces may be relatively small,
these forces may have· an effect of the final results. Because of these uncertainties, and also because of the semi-empiric nature of
the ASED-MO method, the results presented above should only be used in a semi-qualitative, or better qualitative way . In this con
text, we think that not too much attention should be paid to small differences in energies and bond lengths. But the dramatic
influence of hydroxylation on the position of the rhodium atoms and on the rhodium-oxygen bond length is expected to be independent of the approximations in the theory.
6.5 Conclusions
Using the ASED-MO theory, we have shown that the binding
of a rhodium metal particle to an y-Al20 3 support cluster is governed by the hydroxylation state of the surface of the support. Energetically, three-fold sites on a fully hydroxylated surface are favored. The accompanying Rh-0 bond length is about 2.5-2.6 A. This is very close to the values reported with EXAFS for such metal-oxygen distances and with the fact that according to EXAFS
rhodium atoms in the metal-support interface have approximately 2
page 121 Chapter 6
to 3 oxygen neighbors in Rh/ Al20 3 catalysts. For a fully dehy
droxylated A120 3 surface. the rhodium atoms in the metal-support interface prefer on-top sites: the accompanying Rh-0 bond length
decreases to 2.1 A..
6 .6 References
1. van Zon, J. B. A . D.; Koningsberger , D. C. ; van 't Blik , H F. J .; Prins,
R.; Sayers , D. E. J. Chem. Phys . 1984, 80, 3914.
2. van Zon, J. B. A. D.; Koningsberger, D. C.; van 't Blik, H. F. J .; Sayers, D. E. J . Chem. Phys. 1985, 12, 5742.
3. Koningsberger, D. C. ; van Zon, J . B A. D. ; van 't Blik, H. F. J.; Mansour, A . N.; Visser , G. J.; Prins , R. ; Sayers, D. E.; Short , D. R.; Katzer , J. R. J. Chem. Phys . 1985, 89 , 4075
4. Koningsberger , D. C. ; Duivenvoorden , F. B. M .; Kip, B. J .; Gates , B. C. "EXAFS and Near Edge Structure"; Lagarde, P.; Raoux, D.; Petiau, J. Eds.; Les Editions de Physique, 1986; vol. 1, p. C8-255.
5. Koningsberger, D. C.; Martens, J. H. A.; Prins, R.; Short, D. R.;
Sayers , D. E. J. Phys. Chem. 1986, 90, 3047.
6. Martens, J. H. A. ; Prins, R.; Zandbergen, H.; Koningsberger, D. C.;
accepted for publication in J. Phys . Chem ..
7. Moller, K.; Bein, T. "EXAFS and Near Edge Structure "; Lagarde, P.; Raoux, D.; Petiau , J. Eds.; Les Editions de Physique, 1986; vol. 1, p. C8-231.
8. Anderson , A . B. J. Phys. Chem. 1975, 62, 1187
9. Anderson, A . B.; Grimes, R. W.; Hong, S. Y. J. Phys . Chem. 1987, 91 , 4245
10. Anderson , A. B.; Mehandru , 5 . P ; Smialek, J. L. J. Electrochem. Soc .
1985, 132, 1695
11. Anderson, A . B.; Ravimohan , Ch .; Mehandru , S. P Surf. Sci . 1987, 183,438
Rh/A1 20 3 characterized with ASED-MO page 122
12. Lotz . F. W. J. Opt . Soc. Am. 1970, 60 . 206
13. Basch , H. ; Gray , H. B. Th.ear . Chim. Acta 1966. 4 , 367
14. Moore , C. E. Atomic Energy Levels ; NBS Circ . no. 467 ; National
Bureau of Standards, U.S. Government Printing Office: Washington ,
DC, 1958
15. Lippens, B. C. Thesis , Delft, 1961
16. Knozinger, H.; Ratnasamy, P Cata!.. Rev .-Sci. Eng . 1978, 17(1), 31
17. Kampers , F. W. H.; Sayers, D. E.; Koningsberger, D. C. to be pub
lished
page 123
Chapter 7
The Structure of Rh/Ti02 in the Nonnal and the SMSI State
as Determined by EXAFS and H RTEM
7 .1 Introduction
Chapter 7
In heterogeneous metal catalysis the support is used to provide
a large surface area to facilitate the preparation of well dispersed
catalysts and to prevent sintering of the small supported metal par
ticles, in order to preserve their state of high dispersion. It is often found that support materials modify the chemical reactions of the
metal catalyst. Examples are shape selectivity induced by a zeolitic
support and bif unction al catalysis of metal particles dispersed on an
acidic support. where the metal component catalyzes the
hydrogenation/dehydrogenation reactions_ and the acidic support facilitates isomerization of olefinic compounds. In addition, the sup
port may have a more direct influence on the chemical properties of
supported metal particles, especially after reduction at high
(> 650 K) temperature. Thus, it is well known that for metals dispersed on certain transition metal oxides, the capacity to adsorb
hydrogen or carbon monoxide drastically diminishes when the
catalyst is reduced at high, rather than low temperatures ( < 650 and usual > 450 K) even though the particle size remains
unchanged ( J-3). Non-transition metal oxides like Al20 3 and Si02 do not influence the capacity to adsorb gasses; the decrease in
adsorption after reduction at high temperature of the metal particles dispersed on these supports can be accounted for by sintering.
Another interesting phenomenon is observed in chemical processes.
EXAFS and HRTEM of Rh/Ti0 2 page 124
When metal particles are dispersed on transition metal oxide sup
ports, their properties m chemical reactions such as (de)hydrogenation and Fischer-Tropsch synthesis differ markedly
from those traditional oxides, like A1 20 3 and Si02.
A clear distinction between the two classes of support materi
als can be based on their reducibility. Oxides like Al 20 3 and Si02
are hard to reduce, while transition metal oxides like Ti0 2 and
Ta 20 5 can be reduced to suboxides at moderate temperatures. These suboxides are thought to be responsible for the inability of
the metal particles to adsorb gasses after reduction at high temperature. Most of these suboxides are known to have semiconducting properties . In the first reports dealing with this phenomenon ( J-6), these semiconducting suboxides were thought to have a
strong (electronic) influence on the supported metal particles and the phenomenon was labeled SMSI, an acronym for strong metal
support interaction. Thus, SMSI refers to the state of inability of supported metal particles to adsorb hydrogen and carbon monoxide,
invoked by a reduction at temperatures where the support is known to be at least partially reduced.
Oxidation at mild temperatures (> 450 K) restores both the original oxide and the original properties (i.e. the properties when
reduced at mild temperatures) of the metal particles, demonstrating that suboxide formation and inability to adsorb hydrogen and carbon monoxide are related ( 1,6-8). Hence. the SMSI state can be
removed in under (slightly) oxidizing conditions at elevated temperatures.
Since the first discovery, many studies have been devoted to
SMSI. In addition to the model based on an electronic effect several other explanations for SMSI have been suggested. The most impor
tant of them is the coverage model (9-17 ). Reduced transition metal oxides can wet metal, in contrast to unreduced oxides . Thus, it has
been suggested that after reduction at high temperature, the suboxides cover the metal particles and consequently reduce their
page 125 Chapter 7
capacity to adsorb gasses.
SMSI has so far been studied mostly by using model systems
like metal films deposited on oxidized titanium (Ti02), or TiOx
deposited on a metal. The techniques frequently used in these stu
dies are 'surface sensitive', such as Auger and XPS, and in the
majority of cases these studies report coverage . These techniques ,
however, are not truly surface sensitive. In XPS even up to five
layers of the sample can contribute to the spectrum and the results
of sputtering can be questioned because of the destructive character
of the technique. Kelley et al. ( 18), using EELS and other surface
sensitive techniques, did not report complete coverage. Their results
pointed rather to 'electron structural changes' in both metal and support .
Recently, evidence for alloy formation under SMSl-like condi
tions has been reported by Beard and Ross ( 19). In the alloy forma
tion model it is assumed that part of the Ti0 2 supporting oxide is
reduced to metallic Ti and forms an alloy with the supported metal
particles. For each noble metal (M) at least three stable titanium
alloys are known : M3 Ti, MTi and MTi 3. It has been shown that
alloying can reduce the hydrogen and carbon monoxide adsorption
capacity as well (20) and alloying may therefore be another plausi
ble explanation for SMSI.
Up to now, no hard evidence in favor of any of the models to
explain SMSI in real catalysts has been presented in literature. The
model of covered metal particles, though, has been accepted most
widely. In addition, suggestions have been made that coverage alone
cannot explain the anomalous properties of the supported metal
particles under SMSI conditions, and it has been said that an elec
tronic influence of the covering oxide on neighboring (bare) metal
sites might also play an important role ( 10-13,16,17).
EXAFS and HRTEM of Rh/Ti0 2 page 126
EXAFS has proven to be an excellent tool to investigate the
local environment around metal atoms in a supported metal catalyst
(21,22) . Since only metal atoms in the metal-support interface will
be sensitive to changes in the supporting oxide and because only
surface metal atoms will be sensitive to (changes in) coverage, it is
evident that highly dispersed catalysts must be used . Furthermore,
since the contribution to EXAFS spectra of the low-Z atoms of the
support (0 2- and Ti4+) will be low, high quality data are a prere
quisite as well. In a preceding study (23), we presented the results
of an EXAFS study of the structure of the rhodium metal particles
in a Rh/Ti0 2 catalyst. This highly dispersed 2.85 wt% Rh/Ti0 2 catalyst was studied after reduction at low temperature and high
temperature, the latter leading to the SMSI state. After reduction at
low temperature, the Rh-Rh coordination number was 3.2, proving
that the metal particles were very small indeed. The rhodium atoms
in the metal-support interface had oxygen neighbors at 2.75 A. These oxygen neighbors originated from the support. From this it
was concluded that the metal particles rested on a (001] anatase
crystal face. When reduced at higher temperature, in the EXAFS
spectrum a 3.4 A Rh- Ti contribution could be detected, indicating
that the (001] anatase crystal face was reduced as well. Since the
Rh0-02- coordination number hardly changed upon reduction at
higher temperature and no other type of oxygen neighbors could be
detected, we concluded that with EXAFS no evidence for coverage
was found.
In the following, we will present the results of a consecutive EXAFS study of a highly dispersed 4 wt% Rh/Ti0 2 catalyst in the
'normal' and the 'SMSI' state . Because of the higher metal loading,
a better signal-to-noise ratio could be realized. High Resolution
Transmission Electron Microscopy ( H RT EM) has been used to ver
ify the average metal particle size determined by EXAFS. Since in
our earlier EXAFS study we found no evidence for coverage, we
have complemented these experiments by investigating the the
catalyst after reduction at higher temperature (723 K compared to
673 K in the previous study) and we have studied the influence of
page 127 Chapter 7
oxygen on the metal particles in the normal and the SMSI state.
After reduction at 723 K and evacuation at 623 K, the Rh/Ti02
catalyst was exposed to oxygen at liquid nitrogen temperature and
at room temperature and after each exposure an EXAFS spectrum was recorded. The same experiments were carried out on a
Rh/ Al 20 3 catalyst in order to study oxygen adsorption on 'normal' rhodium metal particles .
7 .2 Experimental
7 .2.1 Catalyst Preparation
In order to obtain highly dispersed catalysts, a high surface
area support is imperative. Since surface areas of commercially available Ti02 are low, in the range of 10 - 50 m 2g-1
, we have
prepared our own Ti02 support according to the following pro
cedure. A solution of 8 ml of Ti(OC3H7) 4 in 200 ml ethanol was
added slowly, dropwise, to 4 I of a 1:1 well stirred mixture of ice
and distilled water. The precipitated Ti(OH) 4 was filtered off,
washed with distilled water and dried for 24 h at room temperature, 1 h at 363 K (heating rate 1 K min-1
) and finally 12 h at 393
(heating rate 1 K min-1) . The sample was powdered and calcined
for 3 hat 923 K (heating rate 5 K min-1) . The uncai"cined sample
had a surface area of 700 m2g-1. After calcination the surface area
had decreased to 130 m2g-1 and the pore volume was 0.65 ml g-1.
Calcining for longer than 3 hours at 923 K did not affect surface
area, nor pore volume.
A 4 wt% Rh/Ti02 catalyst was prepared by adding 5 ml of an
aqueous solution of Rh(N03h6H 20 (90 mg mr1) to 3 g of Ti0 2.
After 48 hours , the Ti02 with adsorbed Rh3+ was filtered off,
washed, filtered off and dried as described above for the Ti(OH) 4
EXAFS and HRTEM of Rh/Ti0 2 page 128
prec1p1tate (heating rate 5 K min-1). The dried catalyst was cal
cined at 623 K for 3 h (heating rate 5 K min-1). This sample was
used as starting material for further experimentation .
The H/Rh value as determined by hydrogen chemisorption for
the calcined sample after reduction at 525 K. was found to be 1.2. Recently we published an empirical calibration of hydrogen chem
isorption with EXAFS results for several supported rhodium. plati
num and iridium catalysts (24). According to this calibration a
H/Rh value of 1.2 corresponds to a EXAFS Rh-Rh coordination
number in the range of 7 to 8. For half-spherical particles , this
coordination number corresponds to particles containing roughly 40 to 60 atoms and thus to rather large metal particles . However. as
will be discussed , our · EXAFS results and High Resolution
Transmission Electron Microscopy prove that the metal particles
are very small with five o.r six metal atoms per particle and a parti
cle size of about 7 A. Two explanations for this contradiction are
possible. In the first place, the rhodium metal particles in the
present sample are much smaller than the (rhodium) metal particles in the most highly dispersed catalyst in ( 24). Their chemistry ,
therefore, may deviate from the behavior of the larger metal parti
cles in hydrogen atmosphere as described in ( 24). Because of their
size, the small rhodium metal particles approach the quantum size limit and consequently may not be able to adsorb as much hydro
gen per surface metal atom as a larger (fully metallic) particle.
Another plausible explanation is that during reduction at 525 K the
formation of TiOx suboxides has already started and that the influence of SMSI is reflected in a low adsorption capacity of these
very small rhodium particles. After reduction at 723 K, the H/M value decreased to 0.2, indicating that the catalyst is in the SMSI
state.
In temperature programmed reduction, a sharp peak at 330 K
was observed for the calcined catalysts. After oxidation at 773 K, the temperature programmed reduction profile contains one single
and sharp peak at 350 K. According to Vis et al. (25), a TPR
page 129 Chapter 7
reduction peak at 330 - 350 K corresponds to highly dispersed
metal particles.
1 .2.2 EXAFS Measurements
EXAFS spectra were recorded at the synchrotron radiation
source (SRS) in Daresbury, United Kingdom. The storage ring was
operated at 1.8 or 2.0 GeV, the ring current was in the range of 100 - 300 mA. The samples were pressed into thin self supporting
wafers. The thickness of the wafers was chosen to give an absorbance (µ.x) of 2.5, assuring an optimum signal-to-noise ratio. The
pressed samples were mounted in an in situ EXAFS cell, enabling in situ treatments and measurements in different gas atmospheres.
The EXAFS spectra of the rhodium K-edge were recorded with the sample at approximately 100 K. EXAFS spectra of the reference
compounds were recorded at room temperature.
The experiments on the catalyst were carried out in two series. In all cases the heating rate was 5 K min -1
. After each pretreatment an EXAFS spectrum was recorded. The samples were cooled
using liquid nitrogen, the temperature of the samples was approximately 100 K. In the first series, the catalyst was reduced at 473 K for 0.5 h and subsequently at 723 K for 1 h. In the second series, a fresh (calcined) sample was reduced at 723 K for 1 h and subse
quently evacuated at 623 K for 2 h. After recording the EXAFS
spectrum, a small amount of 0 2 was admitted to the evacuated
sample at 100 K. After recording an EXAFS spectrum the sample was allowed to warm up to room temperature and after 10 min at
room temperature an EXAFS spectrum was recorded at liquid nitrogen temperature. The same experiments have been carried out with
a Rh/A1 20 3 catalyst after reduction at 623 K.
EXAFS and HRTEM of Rh/Ti0 2 page 130
Phase shifts and backscattering amplitudes from reference
compounds were used to calculate EXAFS spectra and to correct in
the Fourier transformations for the k-dependence in both phase
shifts and backscattering functions . Rhodium foil was used as a
reference for Rh-Rh contributions , Rh20 3 for Rh-0, Rh Ti alloy for
Rh-Ti contributions in the EXAFS spectra . The Rh Ti alloy was
prepared by arc melting equimolar amounts of rhodium and
titanium. The structure and homogeneity were checked by X-ray
diffraction and micro probe analysis. After careful powdering and
sieving, 43 mg of the alloy was mixed and crushed with 32 mg of
Al 20 3 and pr~ssed into a self supporting wafer with absorbance
µx =2.5. For Rh 20 3, a supporting wafer with an absorbance of 2.5
was prepared in the same way (70 mg Rh20 3 and 30 mg Al 20 3). A
rhodium foil was chosen with a thickness of 20 µm thick, µx =1.4.
7.2.3 HRTEM Experiments
Before the HRTEM experiments, the 4 wt% Rh/Ti0 2 catalyst
was reduced at 773 K in a 4% H2 in Ar mixture (heating rate
5 K min-1). After the reduction treatment the catalyst was care
fully passivated at room temperature in a 4% 0 2 in He mixture . In
order to study the passivated sample in the electron microscope. the
catalyst was suspended in methanol. A droplet of this suspension
was put on a carbon coated Formvar holey copper grid. For the
H RTEM recordings , an objective aperture of 7 nm -l was used . The
photographs were taken with a magnification of 5.105 and an exposure time of 1 sec.
An electron microscope image is formed from two contributions, the absorption contrast and the phase contrast. Absorption
contrast originates primarily from an intensity deficiency caused by
the exclusion of a number of electrons with scattering angles larger
than the aperture used. Secondly, inelastically scattered electrons
will have wave lengths different from the incident and elastically
page 131 Chapter 7
scattered electrons and will contribute only non-constructively to
the image. Phase contrast arises from constructive interference of diffracted and undiffracted electrons within the aperture used .
When low scattering and thin supports are used. metal particles can be detected best at focus values very close to zero. Since at
this focus phase contrast is almost zero, absorption contrast, due to the higher electron scattering amplitude of the metal particles, is
most pronounced . However, .near zero focus, the resolution of the absorption contrast is approximately twice the point resolution of
the microscope and the particle size determinations will have about the same uncertainty . Since the point resolution of the microscope
used is approximately 2.4 A, the uncertainty in the particle sizes is about 5 A, provided no image calculations are carried out. The best
resolution is obtained near Scherzer focus, where, for thin specimen, phase contrast dominates the absorption contrast. Here, the uncer
tainty is about the point resolution of the microscope. Compared to zero focus, where absorption contrast dominates, the visibility of
the metal particles is worse at Scherzer focus. We have taken a number of photographs at different focus. Such a through-focus
series can help to identify the presence of metal particles (at zero focus) and to determine their size (at Scherzer focus). Therefore, the uncertainty in the metal particle size determined in this way 1s about the point resolution of the microscope, which is 2.4 A.
In case of very small metal particles, it is difficult to distinguish them from artefacts. There is also the possibii"ity that very
small particles are overlooked. Partial in situ sintering of the metal particles with a very intense electron beam can help in this respect.
By comparing the total volumes of the metal particles before and after sintering, an estimation can be obtained of the metal particle
size before sintering. The in situ sintering experiment has been performed by taking out the condensor aperture and focusing the elec
tron beam on the agglomerate of Ti02 crystallites and rhodium particles. The degree of sintering can be regulated by the focus of the
condensor lens .
EXAFS and HRTEM of Rh/Ti0 2 page 132
7.3 Results
7.3.1 Analysis of the EXAFS spectra
The EXAFS functions (x(k)) were obtained from the X-ray
absorption spectra by subtracting a Victoreen curve, followed by a
cubic spline background removal (26). Normalization was performed by division to the height of the edge. In Figure 7.1, the raw EXAFS
functions of the Rh/Ti0 2 catalyst after reduction at 473 and at 723 K and the raw EXAFS functions of the Rh/ Al 20 3 and Rh/Ti02 catalysts after evacuation at 623 K and after 0 2 admission at 100 and 300 K are shown .
The spectra of the reference compounds, used to .obtain phase
shifts and backscattering functions, were processed in the same way as the catalyst samples. To obtain phase sh if ts and backscattering amplitudes, the EXAFS spectra of the reference com
pounds were Fourier transformed over the largest possible range in
k-space. To avoid cut-off effects, kmin and kmax were chosen in nodes of the EXAFS function . Table 7.1 presents the Fourier
transform ranges and the crystallographic data for the reference
compounds (27 ). In the Fourier transforms of the EXAFS functions of the rhodium foil and Rh 20 3, the Rh-Rh and Rh-0 peaks are clearly separated from higher coordination shells. An inverse transformation over a limited range in r-space gave the EXAFS
spectra for the single shell Rh-Rh and Rh 3+-02- absorber-scatterer
pairs . Since the Fourier transforms have not been corrected for the
k-dependence in phase shift and backscattering amplitude, the transforms contain side lobes. These side lobes were included in the
inverse transformation range. The required phase and backscattering amplitude have been derived from these spectra .
page 133 Chapter 7
* 10-2 * 10-2
3 5
a b f--j
::r:: -0 0 CJ
-3 -5 0 5 10 0 5 10 * 10-2 * 10-2
5 5
c d
~ ::r:: CJ
0 0
-5 -5 0 5 10 15 0 5 10
* 10"'"2 * 10-2
5 5
e f f--j
::r:: 0 0 CJ
-5 -5 0 5 10 15 0 5 10 * 10-2 * 10-2
5 5
g h f--j
::r:: 0 0 CJ
-5+-'-~~-+-"~~-t--'~~~ -5-+-'~~~,__._~~_.____.~~
0 5 0 10 15 0 5 0 10 k [A- 1] . k [A- 1]
EXAFS and HRTEM of Rh/Ti0 2 page 134
Figure 7 .1 Raw EXAFS data of
(a) Rh/Ti02 after reduction at 473 K
(b) Rh/Ti02 after reduction at 723 K
(c) Rh/A1p3 after reduction and evacuation at 623 K
(d) Rh/Ti02 after reduction at 723 K and evacuation at 623 K
(e) Rh/A1p3 after oxygen exposure at 100 K
(f) Rh/Ti02 after oxygen admission at 100 K
(g) Rh/A1p3 after oxygen admission at 300 K
(h) Rh/Ti02 after oxygen exposure at 300 K
Table 7 .1 Crystallographic data and Fourier transform ranges for the reference compounds
Rb Fourier transformation
a
b
(
d
Compound NNa
Rh foil Rh
Rhp3 0
RhTi Rh Ti
Nearest Neighbor
Coordination Distance (A)
Coordination Number
NC
2.687 12
2.05 6
2.949 4 2.676 8
Weighting factor in Fourier transformation
nd k-range
3 2.16 - 24.0
1 2.35 - 20.0
3 2.83 - 16.6 1 3.00 - 15.0
e On these data , no direct inverse transformation has been applied .
Crystallographic data derived from (27)
r-range
1.42 - 3.00
0.00 - 2.10 e
1.06 - 2.59
For the Rh T i alloy this procedure was more complex. In the Fourier transform, two contributions are present . The first contribution is from 8 titanium neighbors at 2.676 A and the second from 4
page 135 Chapter 7
rhodium neighbors at 2.949 A. (27 ). Since these contributions over
lap, they could not be separated by an inverse Fourier transform over a window in r-space . A Rh-Rh contribution was calculated
using the phase shift and backscattering amplitude obtained from the rhodium foil. The best agreement in r -space in the region of the
main Rh-Rh peak in the k 3-weighted Fourier transform of the measured and calculated C:XAFS function was obtained with the follow
ing Rh-Rh parameters N = 4.0 , R = 2.95. !:::.a 2 = 0.0042 and !:::.£ 0 = -1. (N is the coordination number, R the coordination dis
tance, !:::.a 2 the De bye Waller factor , a measure for the disorder and !:::.E 0 is a correction on the edge position; see ref ( 28) for more
details.) This calculated Rh-Rh EXAFS was subtracted from the
experimental data and the difference spectrum was used to obtain the Rh- Ti phase shift and backscattering amplitude. The Fourier transform of the difference spectrum showed one single peak at
2.16 A.. The Fourier transform has not been corrected for the k
dependence in phase shift and backscattering amplitude . Because of
the phase shift, the main peak in the Fourier transform has shifted
from the real coordination distance. Inverse transformation over the r-range 1.06-2.59 resulted in the Rh- Ti EXAFS function , from which the Rh-Ti phase shift and backscattering amplitude could be
obtained.
The spectra of the different catalyst samples as presented in Figure 7.1 may contain several contributions. Because of the higher
backscattering amplitude of the high-Z elements, the contribution from rhodium neighbors will be dominant. Other contributions are to be expected from oxygen and possibly from titanium neighbors.
The information of the low-Z elements such as oxygen and titanium is limited in k-space to about k = 7 or 8 A.- 1
. For high-Z elements such as rhodium, the information extends, depending on coordina
tion number, up to k = 11 or 13 ;...-1 and for high coordination numbers even up to k = 15 f...-: 1
. A k 3-weighted Fourier transform emphasizes the high k part of the EXAFS spectrum , therefore
strongly enhances the high-Z element information relative to the low-Z scatterer information and has thus been used to separate
EXAFS and HRTEM of Rh/Ti0 2 page 136
high- and low-Z scatterer contributions .
For a detailed description of the data analysis procedure, we
refer to (28-30). Briefly, the analysis consisted of the following
steps . An Rh-Rh EXAFS function was calculated. The parameters
N, R, /),.a 2 and f),.£0 were chosen to give the best agreement in rspace with the main peaks in the k 3-weighted Fourier transform of
the calculated and the measured EXAFS function. This calculated
Rh-Rh EXAFS spectrum was subtracted from the measured spec
trum. In most cases, the resulting difference spectrum contained up
to 3 or 4 different contributions. Since it was impossible to further
separate these contributions, a 2-, 3- or 4-shell EXAFS spectrum
was calculated and optimized to model the difference spectrum in
k-space as well as the k 1-weighted Fourier transform in r-space (the
Fourier transform was corrected for Rh-0 phase shift obtained from
Rh20 3). In order to calculate Rh-0 and Rh-Ti EXAFS spectra,
phase shifts and backscattering amplitudes obtained from Rh20 3
and Rh Ti were used. For each contribution, a separate spectrum
was calculated. These calculated spectra were added to give the
resulting 2-. 3- or 4- shell Rh-0.Ti EXAFS spectrum.
The resulting calculated Rh-0. Ti EXAF S function was sub
tracted from the original spectrum, in order to start an optimization
cycle. Since this new difference spectrum contained mostly Rh-Rh
information, both the k 1- and k 3-weighted Fourier transforms and
the data in k-space could be used to calculate the best fitting Rh
Rh EXAFS. This improved Rh-Rh EXAFS spectrum was subtracted
from the original data and the resulting difference spectrum was
analyzed to further optimize the Rh-0 and (if present) Rh-Ti
parameters . In this way a cyclic optimization process was started.
The procedure of subtracting a calculated spectrum to separate the
information of the low-Z scatterers from the high-Z scatterer was
followed until the parameters N, R, /),.a 2 and /),.£ 0 for each contri
bution became constant. Figure 7.2 shows the k 3-weighted Rh-Rh
corrected Fourier transform of the original data and the calculated Rh-Rh EXAFS for all samples are shown. The observed differences
Figure 7 .2 Imaginary parts of the Fourier transforms of the original EX
AFS spectra (solid lines) and calculated Rh-Rh EXAFS spec
tra (dashed lines). The Fourier transforms are k3-weighted
and corrected for the Rh-Rh phase shift and backscattering
amplitude. The Fourier transform ranges in k-space are indi
cated in brackets.
(a) Rh/Ti02 after reduction at 473 K (3.33 - 10.24 .&.-1)
(b) Rh/Ti02 after reduction at 723 K (2 .82 - 10.13 .&.-1)
(c) Rh/A1p3 after reduction and evacuation at 623 K (2.97 -14.58 A- 1
)
(d) Rh/Ti02 after reduction at 723 K and evacuation at 623 K
(i.85 - 10.06 ,&.-1)
(e) Rh/A1p3 after oxygen exposure at 100 K (3.26 - 12.03 .&.-1)
(f) Rh/Ti02 after .oxygen admission at 100 K (2.84 - 10.06 .&.-1)
(g) Rh/Alp3 after oxygen admission at 300 K (2.55 - 12.45 .&.-1)
In this case, the dashed line represents the dominant Rh3+-
02- contribution, the Fourier transform is k 1-weighted and
corrected for Rh-0 phase shift.
(h) Rh/Ti02 after oxygen exposure at 300 K (3.22 - 10.18 .&.-1)
are due to low-Z scatterer contributions present in the original data.
Figure 7.3 shows the k 1-weighted Rh-0 corrected Fourier transform of the final difference spectrum (original data minus calculated Rh
Rh EXAFS, which contains oxygen and. if present, titanium contri
butions) and the calculated Rh-0. Ti EXAFS functions for all sam
ples . Finally, Figure 7.4 shows the kt-weighted Rh-Rh corrected
Fourier transform of the original EXAFS function and the EXAFS
function obtained by adding the calculated Rh-Rh and Rh-0.Ti EXAFS functions. In a kt -weighted Fourier transform the low-Z
scatterer is much more pronounced than in a k3-weighted Fourier
transform (as in Figure 7.2). From Figures 7.3 and 7.4 it is evident
that the calculated spectra reproduce the measured spectra extremely well, which emphasizes the reliability of the results.
page 139 Chapter 7
* 10-2 *10-2
4 4
a b
I-ll_ 0 0
-4 -4 0 3 6 0 3 6
* 10-2 * 10-2
3 3
c d
I--:. -0 -0 ll_
-3 -3 0 3 6 0 3 6
* 10-2 * 10-2
5 3
e f
I-ll_ 0 0
-5 -3 0 3 6 0 3 6
* 10-2 * 10-2
7
10 g h . . I-
..... ,\
ll_ 0 5
-7-1------L..~-'--+~-'----'~~
3 0
R [A] 6 0 3 0
R [A] 6
EXAFS and HRTEM of Rh/Ti0 2 page 140
Figure 7.3 Imaginary parts of the Fourier transforms of the difference spectra (original EXAFS spectra minus the calculated Rh-Rh EXAFS functions , solid lines) and the Fourier transforms of the calculated Rh-0.Ti EXAFS functions (dashed lines) . The Fourier transforms are k1-weighted and corrected for Rh-0 phase shift. The Fourier transform ranges in k-space are indicated in brackets.
(a) Rh/Ti0 2 after reduction at 473 K (2.56 - 8.73 .A- 1)
(b) Rh/Ti02 after reduction at 723 K (2.95 - 8.39 .A-1)
(c) Rh/ A1p3 after reduction and evacuation at 623 K (3.46 -8.33 ,A-1)
(d) Rh/Ti02 after reduction at 723 K and evacuation at 623 K (2.94 - 8.48 A- 1
)
(e) Rh/Alp3 after.oxygen exposure at 100 K (2 .54 - 8.00 A- 1)
(f) Rh/Ti02 after oxygen admission at 100 K (2.97 - 8.53 A- 1)
(g) solid line : Rh/ Alp3 after oxygen admission at 300 K (2.55 -12.45 A- 1
)
and Fourier transforms of two calculated EXAFS functions :
dotted line : Rh 3+-02- + Rh0-Rh0
,
dashed line : Rh 3+ -02- + Rh0-02
-.
(See text for further details)
(h) Rh/Ti02 after oxygen exposure at 300 K (2 .87 - 8.92 A- 1)
The structural parameters obtained in this way are presented
in Table 7.2. The contributions of the Rh-0 and Rh-Ti absorber
scatterer pairs in the difference spectra are sometimes small and
could be due to artefacts induced by an incorrectly calculated Rh-Rh
EXAFS. In order to ensure that these contributions are indeed real,
and to ensure that the set of parameters we obtained is indeed the
solution that led to the best fit with the experimental data, for all
Rh/Ti0 2 catalyst samples Rh-Rh EXAFS spectra were calculated
for which N and 1::3.a 2, or R and 1::3.E 0, were varied over a large range.
1::3. a 2 was varied concurrently with N so as to give a constant magnitude in the k 3-weighted Fourier transform, while 1::3.£ 0 was varied
concurrently with R in order to prevent the main peak in the Fourier
Figure 7.4 Fourier transform of the original EXAFS spectra (solid lines) and the Fourier transforms of the calculated best fitting EXAFS spectra (using the parameters presented in Table 7.2. The Fourier transforms are k1-weighted and corrected for Rh-Rh phase shift and backscattering amplitude. The Fourier transform ranges in k-space are indicated in brackets.
(a) Rh/Ti02 after reduction at 473 K (3.33 - 10.24 .&.-1)
(b) Rh/Ti02 after reduction at 723 K (2 .82 - 10.13 .&.-1)
(c) Rh/ Alp3 after reduction and evacuation at 623 K (2.97 -14.58 .&.- 1
)
(d) R.h/Ti02 after reduction at 723 K and evacuation at 623 K (2 .85 - 10 06 .&.- 1
)
(e) Rh/A1p3 after oxygen exposure at 100 K (3 .26 - 12.03 A-1)
(f) Rh/Ti02 after oxygen admission at 100 K (2.84 - 10.06 .&.-1)
(g) Rh/ Al20 3 after oxygen admission at 300 K (2 .55 - 12.45 A-1)
(h) Rh/Ti02 after oxygen exposure at 300 K (3.22 - 10.18 .&.- 1)
transform of the calculated EXAFS from shifting with respect to
the main peak in the Fourier transform of the original data. In all
cases, the difference spectra contained the same features and the
same contributions; only the magnitude of the contributions varied
slightly with varying Rh- Rh parameters. In all cases, the data
analysis procedure as described above led to the parameters as
presented in Table 7.2. The experimental errors have been
estimated by slightly varying the parameters in Table 7.2. Small
deviations from the 'best fit ' in the Fourier transform were not
allowed. In this way. a good approximation for the experimental errors could be established. The overall errors which are presented
in T abel 2, are the sum of this experimental error and an estimation
for the systematic error. The systematic error is induced by the
procedure of analyzing the data and incorporates errors induced by phase and magnitude transferribility, electron main free path , etc.
page 143 Chapter 7
Coordination Distance acr 2 ll Eo D
Treatment NN number !Al (* 10-3 ft.-2) (eV) (a\ (al (a) (a}
0 = Admission of oxygen at the temperature indicated
(a) = Estimated overall (experimental+ systematic) error b
t:.a 2, the Debye Waller factor , is a measure for the disorder and 11Eo is a correction on the edge position : see ref ( 29) for more details .
Figure 7 .5 H RTEM micrograph of the Rh /Ti02 catalyst after reduction
at 773 K en subsequent passivation. The arrows indicate me
tal particles. The micrograph is taken near Scherzer focus .
Therefore, the vis ibility of the metal particles is not very
good , but the uncertainty in their size is best.
page 145 Chapter 7
7.3.2 Characterization with HRTEM
The metal particles were observed to have a very uniform and
narrow size distribution. In Figure 7.5, a micrograph near Scherzer
focus , the metal particles size was determined to be between 7 and
8 A, the uncertainty being about 2 to 3 A. in very good agreement
with the results from EXAFS, based on the Rh-Rh coordination
number of 3.2 - 3.4 (see Discussion). It can be seen that the metal
particles seem to favor positions at the edges of the surface planes
and not on flat surfaces. The number of metal particles observed on
the micrograph is roughly in agreement with the estimated occupa
tion based on the .size of the Ti02 crystallites, the 4 wt% metal load and a metal particle size of about 7 or 8 A.. The in situ sinter
ing experiments (see Figure 7.6) indicate that indeed the major part
of the metal particles is visible. During sintering by exposing the
catalyst to an intense electron beam, large rhodium metal particles
were formed. The d-spacings in these particles indicated that all
metal particles were fee rhodium; alloy particles in which Rh and Ti
are ordered (RhTi3, RhTi or Rh3Ti) have not been observed. The
same sintering experiments on Ti02 supported Ir catalysts resulted
in lr3 Ti alloy formation . Clearly, for rhodium catalysts, alloy forma
tion in the SMSI state is unlikely to occur.
During the HRTEM experiments, several Rh/Ti02 catalysts
have been studied extensively after reduction at high temperature.
In these experiments no sign of even the slightest coverage has
been found. Image calculations (JJ) show that a monolayer of TiOx
on top of a rhodium particle containing 5 rhodium atoms should be
visible. The visibility depends on the defocus of the microscope, the
atomic configuration of the monolayer and the thickness and orien
tation of the support. Before transferring the sample to the micro
scope, the sample will always be passivated. On larger rhodium par
ticles (other Rh/Ti0 2 catalysts have been investigated as well) islands of TiOx will be formed on top of the metal particles during
passivation. Since these islands will be several atomic layers thick,
EXAFS and HRTEM of Rh/ Ti0 2 page 146
Figure 7.6 HRTEM micrograph of the Rh/ Ti02 catalyst after reduction
at 773 K en subsequent passivation and in situ sintering.
they must be visible with HRTEM. Although investigations were performed to image these in particular , such islands were never
observed. In conclusion, we feel that coverage is unlikely to occur for Rh / Ti02 catalysts.
In Figure 7.7 it is shown that the crystallographic surface plane
exposed most is evidently the [101) crystal face. Other planes exposed are [001) and [103). In Figure 7.7 an example is given. The particles on these micrographs are found mainly on edges of the
Ti0 2 crystallites and on [101) crysta l faces ; only few particles were found on [001) crystal faces.
page 147 Chapter 7
Figure 7.7 HRTEM micrograph of the Rh / Ti02 catalyst after reduction
at 773 K en subsequent passivation . Several exposed crystal
faces are indicated .
7.4 Discussion
7 .4.1 Rh/ Ti02 after Reduction at 473 K
After reduction at 473 K and cooling down under H2, the aver
age Rh-Rh coordination number is 2.5 (Table 7.2), demonstrating
that the metal particles were highly dispersed. The Rh-Rh distance
was found to be equal to the Rh- Rh bulk distance : 2.687 A. In
addition to rhodium nearest neighbors , two different rhodium
oxygen contributions could be discerned. These contributions arise
from approximately 1.3 oxygens at 2.07 A and 1.0 oxygen at 2.78 A. The first Rh-O distance is only slightly longer than the Rh3+ -02
-
distance in bulk Rh 20 3 (2.05 A). This indicates that reduction was
incomplete and that part of the rhodium was still in the calcined,
oxidized state. This might seem to contradict our TPR results,
which indicated that reduction was complete at 400 K. There is ,
however , a slight difference between the two reduction treatments .
In TP R, the H2 mixture is forced to flow through the catalyst bed,
whereas in the in situ EXAFS cell H2 flows along and not through
EXAFS and HRTEM of Rh/Ti02 page 148
the self supporting sample wafer. Thus, in TPR the removal of
water out of the catalyst bed is by convection and consequently
much faster than during the in situ reduction in the EXAFS cell,
where water is removed only by diffusion. In the EXAFS cell, therefore, reduction may take longer to complete. In our earlier paper
(23), we found that the catalyst was completely reduced after
reduction at 473 K. This can readily be explained by the lower
heating rate (2.5 K min-1 vs. 5 K min-1) and the longer reduction
treatment (2 h vs. 0.5 h) used in our earlier study.
Reduction of supported noble metal oxide particles is known to
be a fast process, limited only by nucleation, not by particle size
( 25). As a consequence, once the reduction of an oxide particle has started (nucleation) the· reduction process is rapidly completed. We
expect the particles to be either fully metallic or fully oxidized.
Therefore, in the catalyst reduced at 473 K, metal particles as well
as oxide particles will exist. Since the information in EXAFS is aver
aged over all rhodium atoms and ions present in the sample, the
actual Rh- Rh coordination number for the rhodium atoms in the
metal particles is higher than the measured coordination number
(32). We will assume that the Rh-0 coordination number in the
oxide particles is 6, the same as in bulk Rh20 3 (32). This assump
tion ts reasonable, even for small oxide particles, since Rh3+ ions may have 0 2
- ions from the support as one or more of the six oxy
gen nearest neighbors . The fraction of rhodium present in the oxidic
form is then 1.3/6 = 0.22 and the fraction of rhodium atoms in the
metal particles is 0.78. This means that the actual Rh-Rh coordina
tion number for the rhodium atoms in the metal particles is equal to
2.5/0.78 = 3.2.
As in earlier papers ( 28,32), we ascribe the oxygen neighbors
at 2.78 J.. to oxygen ions from the supporting oxide. The rhodium atoms in the metal-support interface have oxygen ions from the
support as nearest neighbors . The radius of zerovalent rhodium is about 1.34 A, the radius of divalent oxygen ions is about 1.4 A.. One
may therefore expect a Rh 0-02- di stance of about 2. 7 4 J.., in good
page 149 Chapter 7
agreement with the reported value of 2.78 A. Since the oxygen ions
at 2.78 A are only nearest neighbors to rhodium atoms in the metal
lic particles, the Rh0-02- coordination number must be corrected in
the same way as we have corrected the Rh0-Rh 0 coordination
number: the corrected value is 1.0/0.78=1 .3. To find the real
number of oxygen neighbors for each interfacial rhodium metal
atom, this value has to be divided by the fraction of metal atoms in
the metal-support interface ( 28).
It is often found that background subtraction is difficult in
EXAFS spectra of catalysts in which the metal component is
(partly) oxidized. Using the cubic spline method. there is no unam
biguous criterion for a correct background subtraction. For metallic
catalysts, the choice of the smoothing parameter in the cubic spline
routine should be such that the magnitude of the Fourier transform
in the region below 1 A is low. For oxid ic samples as the present
one, this can be achieved, but it is always observed , that the first peak in the Fourier transform, in this case the Rh3+-0 2
- contribu
tion, has slightly decreased in intensity. The second peak, in general, has not decreased in intensity. Therefore, the Rh3+-02
- coordi
nation number will be slightly underestimated and · so will be the
corrected Rh0-Rh0 and Rh0-02- coordination numbers. In general,
the error will not exceed 20% .
In the analysis described above, we have used the phase shift
and backscattering amplitude obtained from Rh 20 3 to calculate the Rh0-0 2
- EXAFS functions. Since in Rh 20 3 the absorber-scatterer
pair is Rh 3+-02- with a coordination distance of 2.05 A, it is
incorrect to use data from Rh20 3 EXAFS functions to calculate 2.78 A Rh0-02
- EXAFS functions in which the coordination distance is
considerably larger and the valence state of one of the members of
the absorber-scatterer pair is different . It has been shown (28,33) that in such cases the calculated distances are reliable. The Rh0-0 2
-
coordination numbers however, are underestimated. There is no indication for the degree of underestimation. The actual average
Rh0-0 2- coordination number must therefore be higher than the
EXAFS and HRTEM of Rh/Ti0 2 page 150
above reported value of 1.3.
Figure 7 .8 Small metal particles
(a) four-atom metal particle
(b) five-atom metal particle
(c) seven-atom metal particle
(d) eight-atom metal particle
At this point, we are in a pos1t1on to estimate the size of the
metal particles. In Figure 7.8, examples of four, five, seven and
eight-atom metal particles are shown. In the four-atom metal particle, each metal atom has three direct rhodium nearest neighbors.
The average Rh- Rh coordination number for this particle therefore
is 3.0. In the five-atom metal particle, each interfacial atom has
three nearest neighbors, two interfacial atoms and the top atom.
The top atom has four direct nearest neighbors , the four interfacial
atoms. The average number of rhodium nearest neighbors in this
particle is therefore 3.2 and the average diameter is about 6.5 A.. The average coordination number for the seven and eight atom
page 151 Chapter 7
metal particles is 4.0 and their diameter is about 8.5 A. Since the
observed Rh-Rh coordination number is about 3.4 ± 0.3, we con
clude that the metal particles in the Rh/Ti0 2 catalyst contain
roughly 5 or 6 metal atoms. The results of th.e HRTEM characterization, a very uniform and narrow particles size distribution around
7 A, are in good agreement with this conclusion. Obviously, even for very small metal particles, the EXAFS coordination number very
. accurately determines the metal particle size. These results, how
ever, leave some room for a discussion about a particle size distribution. The EXAFS results give a Rh-Rh coordination number of 3.4 ± 0.3. This corresponds to inetal particles containing from 4 up to
6 or 7 or even 8 atoms. The particle diameter determined with
HRTEM is about 7.5 ± 1.5 A. This corresponds to metal particles
containing roughly 3 up to about 9 atoms per particle. Because the average particle size as determined by both EXAFS and HRTEM
points to particles containing about 5 atoms, we assume the metal
particles contain between 4 and 8 metal atoms. The average metal
particle will contain 5 atoms. Even explanations in t~rms of bimodal size distributions are possible. In that case, a minor part of the
Rh atoms is present in particles containing one or twp metal atoms and the major part in larger particles, which contain bn the average
7 or 8 atoms per particle. We believe, however, that the latter situa
tion is unlikely.
7 .4.2 Rh/Ti02 after Reduction at 723 K
After reduction at 723 K, we found no evidence for the pres
ence of rhodium oxide particles. Apparently, at this temperature the
reduction was complete. The Rh-Rh coordination number was 3.4, in good agreement with the corrected (.and possibly underestimated)
value of 3.2 reported above. This indicates that during the high temperature reduction no sintering had taken place. The Rh-Rh
coordination distance had decreased markedly . For non-SMSI
EXAFS and HRTEM of Rh/Ti0 2 page 152
catalysts such as Rh/ Al 20 3, such a decrease in the Rh-Rh distance
has been observed after reduction and evacuation at high tempera
ture (34) . After removal of adsorbed hydrogen, the remaining sur
face metal atoms contract in order to compensate for the loss of hydrogen nearest neighbors. Obviously, since the measurement was
performed in an H2 atmosphere at 100 K. the metal particles in the
Rh/Ti0 2 catalyst after reduction at 723 K did not chemisorb H2.
Clearly, the metal particles were in the SMSI state. In our earlier
Rh/Ti0 2 study we reported the same observations.
The presence of the long distance Rh-0 contribution is as
expected. However, the bond length between interfacial rhodium
atoms and supporting 0 2- ions has decreased from 2.78 to 2.60 A.
The decrease in Rh-Rh distance, 0.053 A, indicates a decrease in the
average rhodium atomic radius of 0.027 A. This decrease, however,
is too small to account for the observed decrease of 0.18 A in the
Rh0-0 2- bond. Although it is obvious that this decrease is the result
of a change in the metal-support interaction, it is not evident what
kind of interaction has caused this decrease. In our earlier study we reported a decrease in the Rh0-02
- distance of 0.04 A. That catalyst
was reduced at a temperature 50 degrees below the reduction tem
perature of the present catalyst (673 and 723 K, respectively) . The
metal-support interaction induced by the high temperature reduction is obviously stronger when the catalyst is reduced at higher
temperatures.
In the SMSI state, two additional contributions were detected,
titanium neighbors at 3.41 and 4.39 A. These contributions could not be detected after the low temperature reduction. In our previous
paper (23), we reported that a 3.4 A Rh-Ti contribution was present
after reduction at 673 K. The better signal-to-noise ratio of the
present experiments and the enhanced metal-support interaction obviously enable us to distinguish even more contributions. To
explain these contributions, we need to take a closer look at the supporting oxide, anatase Ti02. The H RTEM micrographs show
that the [101] anatase crystal face is exposed most. HRTEM also
page 153 Chapter 7
showed that the majority of the metal particles was present on the
edges of the Ti02 crystallites, the rest of the metal particles was present predominantly on [101] crystal faces. It is therefore not evi
dent what the structure of the (average) crystal face is on which the particles rest. The majority of the crystal faces are 1101] faces.
It is not unlikely, that the crystal face on the edge of the crystallites. between two [101] crystal faces, is indeed a [001] type crystal
face . Surface energy calculations have shown that the [001] crystal face is indeed a stable crystal face (JS). We therefore assume that the metal particles rest on [001] and [ 101] type crystal faces.
In Figure 7.9a and 7.9b, both a [001] and (101] anatase crystal
face are shown. Both crystal faces consist of a two dimensional rectangular array of oxygen ions, with as many octahedral sites as oxygen ions. Half of these octahedral sites is occupied by Ti4+ ions.
We assume that the five (or eight) atom metal particle (see Figure
7.8) rests on these crystal faces. Figure 7.9c and 7.9d show the most plausible arrangement for this metal particle on both faces;
Figure 7.9 Ti02 crystal faces
(a) 1101] anatase crystal face
(b) [001] anatase crystal face
( c) [101] anatase crystal face with a five atom metal particle
(d) 1001] anatase crystal face with a five atom metal particle
(e) [101] anatase crystal face with a five atom metal particle after
reduction
(f) 1001] anatase crystal face with a five atom metal particle after
reduction
Ri = 2.7 A R2 = 2.0 A R3 = 3.4 A R4 = 4.3 A
(Rh0-o2-)
(Rh0-Ti)
(Rh0-Ti)
(Rh0-Ti)
EXAFS and HRTEM of Rh/Ti0 2 page 154
a b
c d
e f
eTin+
page 155 Chapter 7
four metal atoms are interfacial atoms and each interfacial atom has four oxygen nearest neighbors . The average Rh0-0 2
- coordination
number in this arrangement is 3.2. Earlier we argued that our
method of calculating Rh0-0 2- coordination numbers underestimates
the real Rh0-02- coordination numbers. The apparent discrepancy
between the expected and the measured value of 1.9 is therefore acceptable.
In the models we have described , we assumed that in a first approximation the rhodium atoms of the metal particles are
situated on lattice-oxygen positions, that is, positions which would
have been occupied by oxygen ions, if another Ti0 2 layer had been
deposited on the crystal. Therefore, it does not matter whether the metal particles rest on [101] or [001] or any other crystal face.
Because the rhodium atoms occupy oxygen-equivalent positions, the Rh-0 2
- distances (and Rh-Ti distances, these will be discussed
later) will be the same, regardless of the crystal face on which the particles rest. The coordination distances are model-independent.
The coordination numbers , however, may vary , but only slightly . Another consequence of this assumption is, that the metal particles
grow epitaxially on the supporting oxide, which is not unlikely for very small metal particles . The metal particles in Figure 7.8a and
7.8c, which have been grown epitaxially on [111] type crystal faces, also fit on the [001] and [101] type crystal faces. Thus, the four, five, seven and eight atom metal particles in Figure 7.8 represent possible structures of the metal particles. For the sake of clarity,
however , we will focus only on the five atom metal particle on a
[001] or [101] type anatase crystal face . One more detail needs closer attention. Anatase is build of slightly distorted octahedra . As a result, in pure anatase, two 0 2
- -02- distances are encountered . If
the Rh atoms are situated on 0 positions, one should expect the same two Rh-0 2
- distances as well. We found however only one
Rh-0 2- distance. This can be accounted for as follows. In the [101] plane (see Figure 7.9a), the lower 0 2
- ions, the ions underneath Ti4+ ions , are shifted alternately slightly to the right and slightly to the left with respect to their position as shown in Figure 7 .9a. This
EXAFS and HRTEM of Rh/Ti0 2 page 156
has very little influence on the Rh-0 2- distances because the dis
placement is perpendicular to the distance. For metal particles
situated on [101.] type crystal faces we thus expect only one Rh-0 2-
distance with possibly a increased disorder around that distance. From Table 7.2, it is indeed clear that the disorder in the Rh0-0 2
-
distances is larger than any other disorder (note that the disorder is
expressed relative to the disorder in the reference compounds) . The
0 2- ions in a ideal (001] plane (see Figure 7.9b) are situated alter
nately a little above and a little underneath the horizontal plane.
But if we assume that the outermost [001] plane of 0 2- ions has
relaxed in such a way that all 0 2- ions are in the same plane, only
one Rh-0 2- distance is present with possibly an increased disorder.
A closer look at ou; model in Figure 7.9 reveals three Rh0-Ti4+
distances : R2 = 2.1 , R3 = 3.4 and R4 = 4.3 A. The average coordi
nation numbers for these contributions are respectively 0.4 , 1.6 and
3.2. (The latter coordination number is higher than one might derive from Figure 7.9. In Figure 7.9, however , only the Ti4+ ions in the
top layer are shown . The Ti4+ ions in the second layer will also con
tribute to the 4.3 A Rh0-Ti4+ contribution . Another 'hidden' 4.3 A Rh0-Ti 4+ contribution arises from the top rhodium atom in the five
metal atom particle and the Ti4+ ion directly underneath the top
atom). These coordination numbers are low and one would expect
their contributions , especially for the longer distance, not to be
detectable by EXAFS. The EXAFS amplitude is in a first approxi
mation inversely proportional to the square of the distance. The
values of N/R 2 for the three contributions are 0.09, 0.15 and 0.16,
respectively. Because of the Debye Waller factor, however , these
values will change . For comparison, 1\J/R2 for the Rh0-02- contribu
tion is 0.44. Since the backscattering amplitudes for the Rh-Ti and
Rh-0 contributions do not differ much in magnitude (the Ti backscattering amplitude is slightly larger) , it is evident that the Rh0-
Ti4+ contributions will be hard to detect , compared to the Rh0-0 2-
contribution . Because of the Debye Waller factor, the 3.4 A contri
bution is expected to be the most dominant of the three Rh0-Ti4+
contributions . In Figure 7.3a, the difference spectrum for the 473 K
page 157 Chapter 7
reduced catalyst, a small contribution around 3.4 A can indeed be
observed. But this contribution is too small to allow a reliable analysis.
In our previous paper , we already reported the presence of a 3.4 A Rh-Ti4+ coordination in the SMSI state catalyst . In the 723 K
reduced sample of the present study , both the 3.4 and 4.3 A RhTi4+ contributions are indeed observed, but the coordination
numbers are higher than anticipated . The explanation for the enhanced presence of these Rh-Ti4+ contributions in the SMSI state
can be found in the reducibility of the Ti0 2 supporting oxide. During the reduction process, oxygen is removed from the Ti0 2 surface
in the form of water . This process is centered around the metal particle, which provides reactive hydrogen atoms and thus catalyzes
this reduction process. After a few oxygen ions have been removed , the second [001] or [101] crystal plane becomes exposed. Bare Ti3+
ions from the first (outermost) crystal face remain on top of the 0 2
- ions of the second plane. This situation is energetically very
unstable and these Ti3+ ions will migrate to empty ocrtahedral sites in the second oxygen layer and possibly the third or even further
crystal planes . What we see here, is the formation of a TiOx suboxide and especially the formation of part of a shear plane . The five atom metal particle is positioned in the same way on the second crystal planes as it was on the outermost plane before reduction,
but now the interfacial rhodium atoms have more Ti 3+ and Ti4+ nearest neighbors at 3.4 and 4.3 A (see Figure 7.9e,f). This increase
in Ti nearest neighbors explains our EXAFS results. We thus conclude that after reduction at high temperature, the metal particles
rest on partially reduced Ti02.
Our first conclusion from this analysis is that the metal parti
cles are very small indeed. The average rhodium metal particle in the 4 wt% Rh/Ti0 2 catalyst contains about 5 rhodium atoms. The diameter of these particles is about 6.5 A. This was confirmed by HRTEM . This technique revealed even more information. After
exposing the sample to an intense electron beam, which resembles
EXAFS and HRTEM of Rh/Ti0 2 page 158
reduction at high temperature , the metal particles had sintered to
large fee rhodium metal particles; the presence of any ordered alloy
of Rh and Ti could be ruled out. The same experiments on Ir /Ti02 resulted in lr3 Ti alloy particles. Thus, it is unlikely that in the
Rh/Ti0 2 catalyst in the SMSI state alloy formation takes place.
This in in accordance with the results resported by Beard and Ross ( 19). They found that upon heating a platinum catalyst supported
on carbon and impregnated with TiCl4 Pt3 Ti alloy particles were formed . For iridium and platinum alloy formation may play a role in
the SMSI state, but for rhodium catalysts this is not the case.
Secondly, we found with EXAFS no evidence for coverage, and
HRTEM confirmed this. In the HRTEM study, this catalyst and
other Rh/Ti0 2 catalysts with larger metal particle sizes were stu
died extensively after reduction at high temperatures and subsequent passivation . In these studies no sign of even the slightest
coverage was found . It is known , that after passivation and subsequent reduction at low temperature the normal hydrogen adsorption
capacity of the metal particles is partly , but not completely restored (7 ). Therefore, if coverage occurs after reduction at high tempera
ture and suppresses the hydrogen chemisorption by decreasing the exposed metal surface area, passivation will not completely remove
the covering oxide . Since in our HRTEM experiments on Rh/Ti0 2 catalysts we never observed covering, we conclude that coverage is
unlikely to occur for Rh/Ti0 2.
In the following we will discuss oxygen adsorption experiments
which have been performed in order to investigate whether the metal surface of the rhodium particles in the SMSI state is blocked
by a covering TiOx (sub)oxide . We exposed the catalyst in the in situ cell to pure 0 2 at 100 K. Before oxygen admission the
catalyst had been evacuated at 623 K. Subsequently the temperature was raised to room temperature and another EXAFS spectrum
was recorded . We performed the same experiments on a Rh/ Al 20 3 catalyst (reduced at 623 K) in order to compare oxygen adsorption properties of rhodium metal particles in the 'SMSI state' and the
page 159 Chapter 7
'normal state' (A1 20 3 supported). For the sake of clarity, we will
discuss alternately the results of the experiments on the Rh/ Al20 3 and Rh/Ti02 catalysts.
7 .4.3 Evacuation at 623 K
After reduction and evacuation of the Rh/ A1 20 3 catalyst, the major contribution is from rhodium neighbors with a Rh-Rh coordination number of 5.6. This indicates that the metal particles contain about 15 atoms and are roughly 10 A in diameter (24). Compared to the bulk value, the Rh-Rh coordination distance has decreased by 0.052 A. As we have described in the foregoing this contraction is the result of the evacuation procedure. Because of the removal of adsorbed hydrogen the surface metal atoms contract in order to compensate for the loss of neighbors . A Rh0-02
- contribution is also present, originating from neighboring 0 2
- ions in the metal-support interface. The small contribution at 2.06 A has to be ascribed to the presence of rhodium oxide. The maximum tempera
ture during the reduction treatment (623 K) was high enough to obtain complete reduction of the Rh/A120 3 catalyst (28). However,
it is possible that partial oxidation has taken place due to formation of water during the evacuation procedure or because of a small inleakage during or after the evacuation treatment.
EXAFS and HRTEM of Rh/Ti0 2 page 160
7.4.3.2 Rh/Ti02
After reduction at 723 K of a fresh Rh/Ti0 2 sample followed
by evacuation at 623 K, the EXAFS spectrum (in k-space as well in
r-space after Fourier transformation) resembles very closely the
EXAFS spectrum measured directly after reduction. All differences are within the indicated experimental errors . The Rh-Rh distance is
still significantly smaller than the bulk value (2.687 A) and almost equal to the Rh-Rh distance in the evacuated Rh/ A1 20 3 catalyst.
Thus, we have confirmed that in the SMSI state the rhodium metal particles do not adsorb H2.
7 .4.4 Oxygen Admission at 100 K
After admission of 0 2 to the Rh/ Al 20 3 catalyst at 100 K, the
EXAFS spectrum changed drastically . The Rh-Rh contribution diminished, as is clearly reflected in the decreased amplitude of the EXAFS function at higher k-values (k > 7 A-1
• see Figure 7.1e). Obviously, after 0 2 admission at this temperature all the metal particles are partially oxidized. This situation is different from the par
tially reduced Rh/Ti0 2 catalyst, where some of the particles are already fully reduced and others were still oxidized . It is impossible to calculate the size of the remaining metal kernel because now this metallic kernel is covered with Rh20 3 or possibly another form of
rhodium oxide and the Rh 3+-0 2- coordination number in the cover
ing shell of oxide will be different from six . At lower k-values, the
influence of the pronounced presence of the two Rh-0 contributions
is obvious. These qualitative conclusions are confirmed by the quantitative analysis (cf. Table 7.2). The decrease in the Rh-Rh
page 161 Chapter 7
coordination number and the increase of the Rh3+-0 2- coordination
number indicate that the oxidation process has started. Note, that
the high k-value part of the EXAFS function is very sensitive for
changes in the Rh-Rh contribution (i.e. average metal particle size) and the low k-value part is sensitive for low-Z scatterer contribu
tions.
7 .4.4.2 Rh/Ti02
The results for the Rh/ Al 20 3 catalyst presented above are in
sheer contrast to the results for oxygen admission to the Rh/Ti0 2 catalyst. At higher k-values, the EXAFS spectrum of the Rh/Ti0 2 catalyst after 0 2 admission at 100 K resembles very closely the
spectra of the reduced and the evacuated samples (cf. Figures 7.1d
and 7.1f). No changes in the Rh 0-Rh0 parameters could be detected.
Obviously, for the Rh/Ti0 2 catalyst, the basic structure of the
metal particles remained intact and oxidation has not taken place.
At lower k-values there are small differences. This becomes clear
from the Fourier transform of the difference spectrum, Figure 7.3f.
Apart from the Rh-0 and Rh- Ti contributions discussed before, a Rh-0 contribution at 2.09 A. is clearly visible with a coordination
number of 1.0. This Rh-0 contribution must be different, however,
from that of the sample reduced at 423 K. In the EXAFS spectrum
of that sample, the 2.06 A. Rh-0 contribution was assigned to Rh 3+-0 2- absorber-scatterer pairs, present in Rh 20 3 particles. But in
the spectrum taken after oxygen admission (cf. Figure 7.1f) no
differences with the spectrum taken after evacuation (Figure 7.1d)
can be observed at k > 6 A.-1. In this region, EXAFS is very sensi
tive to changes in the Rh-Rh contribution, therefore neither the
Rh-Rh coordination number nor the metal particle size have
changed. The results of the detailed data analysis confirm this (see
Table 7 .2). Clearly, oxidation of the rhodium particles on the Ti02 support has not taken place. The only possibility left is to ascribe
the 2.09 A. Rh-0 contribution to oxygen adsorbed on the surface of
EXAFS and HRTEM of Rh/Ti0 2 page 162
the metal particles. The atomic radius of zerovalent rhodium in the
metal particles is 2.64/2 = 1.32 and the radius of covalent oxygen
is about 0.73. The expected Rh0-o0 distance (2.05 A) is in good
agreement with the calculated distance of 2.09 A.
A very important result of this analysis is that the metal parti
cles in the SMSI state are capable of adsorbing oxygen. In order to
be able to adsorb oxygen, the metal particles must be bare,
uncovered, or at least not fully covered with a TiOx suboxide. Even
though oxygen was able to adsorb on the metal particles, oxidation
did not take place. in contrast to the Rh/ Al20 3 catalyst. This
suppressed oxidation can only be the result of an electronic
influence from the TiOx .suboxide.
7 .4.5 Oxygen Admissi.on at 300K
After oxygen admission at 100 K, the Rh/ Al20 3 catalyst was
warmed up to room temperature. The EXAFS spectrum (Figure 7.1g) differed completely from the spectra of the Rh/ A120 3 catalyst
after evacuation and oxygen admission at 100 K (Figures 7.1c and
7.1e). At high k-values, almost no high-Z scatterer EXAFS is visible, while at low k-values. the low-Z scatterer contribution differs markedly from the low k- value part of the two preceding spectra.
The main reason for this is a decrease in the Rh-Rh coordination number and an increase in the oxidic Rh3+-02
- contribution.
Because of the high Rh3+-0 2- contribution, the analysis of this ·
spectrum was different from the procedure as described above. The
major contribution to the spectrum originated from oxygen scatter
ers and therefore a k 1-weighted Rh-0 phase corrected Fourier
transform rather than a k 3-weighted Rh- Rh phase and
page 163 Chapter 7
backscattering corrected Fourier transform was used to calculate
the different contributions in the EXAFS spectrum. Since in this
Fourier transform an incorrect phase shift function has been used
for the (small) Rh0-Rh 0 contribution, the accompanying Rh-Rh peak
is shifted and coincides with the peak originating from the Rh0-0 2-
bond of about 2.76 A. The fact that the peak at the right hand side of the Rh3+-0 2
- contribution is indeed the result of two (Rh0-Rh0
and Rh0-0 2-) contributions is indicated if Figure 7.3g, in which the
magnitude of the Fourier transform of the original data and of the
Fourier transforms of the calculated EXAFS spectra of Rh3+-02- +
Rh0-Rh0 and of Rh3+-02- + Rh0-02
- are presented. In the Fourier
transform of the former calculated EXAFS function, a strong des
tructive interference is visible in the region between both peaks, in
the latter Fourier transform there is a slightly constructive interfer
ence. In the same region in the Fourier transform of the original
data, there is a slightly destructive interference. Therefore, we
must conclude that the right-hand side peak in Figure 7 .3g is the
result of the sum of two contributions, namely Rh0-Rh0 and Rh0-
02-. In a k 1-weighted Rh-Rh phase and backscattering amplitude
corrected Fourier transform, all contributions are separated, but, as mentioned above, this Fourier transform is not suitable to optimize
the dominant low-Z scatter contribution. Because of this, the errors in the parameters of the Rh-Rh and Rh0-0 2
- contributions used to
calculate the best fitting spectrum, are larger than the errors in the
same parameters of the other EXAFS spectra of the Rh/ A1 20 3
catalyst.
As argued before (see 7.4.1), the Rh 3+-02- coordination
number is not very accurate. The value of 3.6 m:.ist be considered
as a lower limit. The Rh 0-Rh0 and the Rh0-0 2- coordination
numbers on the other hand are more reliable. The Rh0-Rh0 contribution had diminished markedly, while the contribution of the long
distance Rh0-0 2- coordination number increased relative to the Rh
Rh coordination number. From this we conclude that the remaining
metal kernels of the rhodium particles are covered with rhodium
oxide. The presence of rhodium oxide on top of the metal particle
EXAFS and HRTEM of Rh/Ti02 page 164
creates an extra interface in which zerovalent rhodium is in contact
with an oxide. In this interface, new Rh0-02- bonds will be present.
This proves that EXAFS is capable of detecting coverage. It impli
cates, that coverage of titania supported metal particles in the SMSI state is even more unlikely than we have suggested up to now .
7 .4.5.2 Rh/ Ti02
When the Rh/Ti0 2 catalyst after oxygen admission at 100 K
was warmed up to room temperature, the EXAFS spectrum at
higher k-values still re~embled quite closely the spectrum of the
sample after evacuation. This indicates that the basic structure of
the metal particle size had not changed and that the formation of
rhodium oxide had not yet taken place. Detailed analysis confirmed
that the Rh-Rh coordination number remained constant. Compared
to the spectra after evacuation and after oxygen admission at 100 K, the differences in the EXAFS spectrum at lower k-values
were much more pronounced. The main reason for this is an
enhanced influence from the short distance Rh0-o0 contribution.
Obviously, since oxidation had still not taken place, the metal particles had adsorbed more oxygen . This will be explained in the fol
lowing paragraph.
The contributions from the support ions (Ti and 0 2-) all
relaxed to longer distances . For the Rh0-Ti contributions, this may seem to be in contradiction with the indicated error . However, the
errors indicated in Table 7.2 are overall errors . They are the sum of
the experimental error and the systematic error . The experimental
error is the result of (in)accuracies during analyzing the data. The
systematic error is the result of the procedure of analyzing the data
and may be regarded as being equal for comparable contributions.
For example, the quoted error of 0.03 in the Rh0-Ti distance of 3.43
A in the sample after oxygen admission at 100 K indicates that the
real Rh0-Ti distance has a 95% probability of being in the range of
page 165 Chapter 7
3.40 to 3.46 A. When comparing two distances, however, the sys
tematic error can be ruled out and only the experimental error
should be taken in account. To a good approximation, the experi
mental error in the Rh-Ti distances is about 0.01 A. Thus, the Rh0-
Ti distance in the sample after oxygen admission at 300 K is in
fact significantly larger than the distance found after oxygen admission at 100 K. The Rh0-02- contribution relaxed to 2.75 A, which is
very close to the Rh0-02- distance for the catalyst in the normal
state. The Rh0-Ti contributions relaxed to 3.48 and 4.36 A, indicat
ing that the metal-support interaction has weakened. Both Rh0-Ti
coordination numbers have decreased . The decrease in the coordi
nation number of the 3.48 A contribution is almost within the
experimental error, but the decrease in the 4.36 A Rh0-Ti contribution is more pronounced. Clearly, Ti ions in the vicinity of the metal
particles have disappeared. This indicates that the TiOx suboxide
around the metal particles has started to re-oxidize. This is, of
course, in agreement with the fact that the metal-support interac
tion has decreased in strength, the SMSI state has been removed partly and it may explain the fact that the metal particles have
adsorbed more oxygen.
7.4.6 Different Rh0-Ti Contributions
In the preceding paragraphs, we discussed two Rh-Ti contribu
tions present in the EXAFS spectra. Based on these findings, on
literature data and on HRTEM, we concluded that the rhodium
metal particles in Rh/Ti0 2 probably rest on anatase (001] and (101] crystal faces. In a model in which a five atom metal particle rests
on a [001] or [101] anatase crystal face, the distances found with EXAFS are indeed present. It is evident that EXAFS will not be able
to detect the distances longer than 4.4 A. However, the short 2.1 A Rh-Ti distance should be detectable by EXAFS but we have not
been able to straightforwardly detect this Rh- Ti contribution. One
EXAFS and HRTEM of Rh/Ti0 2 page 166
of the most important reasons is that especially in the normal state, the expected coordination number for this 2.1 A bond is low. Consider for example the particle on the [001] plane : two rhodium atoms have each one Ti neighbor; the coordination number therefore is 2/5=0.4. We may expect that this coordination number will increase for the metal particle in the SMSI state by a factor of two (or less). Thus, the maximum coordination number one may expect for this contribution is 0.8 and therefore still too small to be detectable by EXAFS. Compare for example Figures 7.3b and 7.3d. The Rh 0-02
- coordination number is 1.9. This contribution is well above the noise level (about 1 *10- 2
) but decreasing it by a factor 2 brings it already very close to the noise level. Therefore, a Rh0-02
- coordination number of 1 would be difficult to detect and even more difficult to analyze. Consequently, a Rh0-Ti contribution with a coordination number less than about 1 is almost impossible to detect or analyze. However, in the imaginary parts of the Fourier transforms of the difference spectra of the Rh/Ti02 catalyst in the SMSI state (Figures 7.3b and 7.3d) small deviations between theory and experiment are visible around 2 A. These deviations could not be completely eliminated, but adding a 2.1 A Rh-Ti contribution improved the agreement between calculated and experimental spectra . Because of the low coordination number, there remains some uncertainty about the scatterer. It was not possible to unambiguously identify the scatterer as oxygen or titanium , although titanium as neighbor resulted in a better fit. It is evident that at the low-R side in the Fourier transform of the EXAFS spectrum, a small contribution is present and it is likely that this contribution is due to a Rh-Ti coordination.
page 167 Chapter 7
Figure 7.10
(a) Magnitude of the k1 -weighted Fourier transforms of
solid line : EXAFS spectrum of the Rh Ti alloy
dotted line : Calculated Rh-Ti EXAFS function
dashed line : Calculated Rh-Rh EXAFS function
(b) Reference spectra of
solid line : Calculated Rh-Ti EXAFS function
dotted line Rh-Ti EXAFS function from inverse Fourier
transform 0. 15 -,----,------r--~--~
0.10
0.05
-RhTi ··········Ti -----·Rh
..-
'i /~', f\
t\ ' '
! \
,/ .. " ,, ~
~· ··· .
0. 00 +---'-----+----'--__;_-"! 0 2
R [A] 4
0.10
0.05
0 . 00
-0 . 05
-0 . 10
-0 .15 2
7 .4. 7 Comparison with Literature Data
-Calcul . ···········Inv. FT
14
Recently, similar EXAFS spectra of a Rh/Ti0 2 catalyst in the SMSI state were reported by Sakelson et al. (36). They ascribed the
major peak in the magnitude of the Fourier transform at the low rside of the dominant Rh-Rh contribution to a Rh-Ti contribution, while our results point to a Rh0-02
- contribution for the same peak
in the Fourier transform. We call the assignment of this peak to Rh-Ti into question for the following reasons :
( i) The above mentioned contribution, referred to as Rh-X, resembles closely the peak in the Fourier transform of the EXAFS functions of Rh/ A1 20 3 catalysts, which has unambiguously been ascribed to a Rh-0 contribution (28,32).
EXAFS and HRTEM of Rh/Ti02 page 168
Furthermore, we observed the same Rh-X contribution in the
'normal' as well as in the SMSI state of the catalyst. Since in the normal state there is no reason to assume Rh-Ti coordina
tion, the assignment of the Rh-X peak to oxygen neighbors in the metal-support interface is much more likely.
(ii) In the fitting procedure described in (36), only titanium as a
neighbor was taken into account; oxygen was not tried as a possible neighbor. The fitting procedures on our spectra yield,
in terms of sum-of-least-squares and variance, better results with oxygen than with titanium as scatterer in the Rh-X contributio·n.
(iii) The Rh-Rh and Rh-Ti contributions in the RhTi reference
compound overlap both in the k 1- and in the k 3-weighted
Fourier transforms. A rough approximation indicates that the overlap is at least 25%. Also, the well known side lobe of the
Rh-Rh peak (due to the k-dependence in phase and backscattering functions) is hidden under the Rh-Ti peak. In Figure
7.10a, the k 1-weighted Fourier transforms of the calculated Rh-Ti (N=8, R=2.676) and Rh-Rh (N= 4, R = 2.949) EXAFS spectra are shown, demonstrating their overlap. Thus, inverse transformation as performed in (36), using a window to
separate both contributions, definitely results in incorrect phase and backscattering functions for the Rh-Ti absorberscatterer pair. In Figure 7.10b the Rh-Ti EXAFS function is shown as we have calculated it and the same contribution
which has been achieved by an inverse Fourier transform over the window 0.00 - 2.38 A.. It is obvious that especially at lower k-values the differences are substantial.
(iv) Sakelson et al. stated that after a k 1 -weighted Fourier
transform, the Rh-Rh and Rh-X contribution can be separated
much better (36). However, in a k 1-weighted Fourier transform , the low-Z scatterer information is much more pronounced than in a k 3-weighted Fourier transform. As a result, both Rh- Rh and Rh-X peaks still overlap, and the overlap is even larger . Because of the larger overlap, there is a strong destructive interference in the region between the two peaks.
page 169 Chapter 7
This can best be seen in the imaginary part of the Fourier
transform . Because of the anti- phase behavior of the imaginary part of the separate contributions. there is a strong
destructive interference and the magnitude of the Fourier transform shows a sharp minimum between both peaks. In a
k 1-weighted Fourier transform, the Rh-Rh and Rh-Ti contributions are therefore separated even less clearly than in a k 3
-
weighted Fourier transform. This can also be seen in Figure
7.10, where the magnitude of the k 1-weighted Fourier
transform of the Rh Ti alloy is compared to the separate Rh-Ti
and Rh-Rh contributions. Because of the destructive interfer
ence, the sum of the two separate contributions (solid line) is lower in magnitude than the respective contributions at their
maximum. A careful inspection also reveals that the peak maxima are shifted slightly. Also for this reason , it is incorrect
to use a window in a k 1-weighted Fourier transform to
separate the Rh-Rh and Rh-X contributions.
(v) It is insufficient to use a k\(k) plot to show the result of the
best fit (36), since the k 3 term strongly enhances the high-Z Rh-Rh contribution which makes the plot less sensitive to the
low-Z Rh-X contribution. A plot of x<k) is necessary to com
pare experimental and calculated data.
Rh-Ti phase and backscattering functions have in a first approximation the same general k-dependence as the Rh-0 phase
and backscattering functions . This is as expected, because of the
low-Z character of both oxygen and titanium (37 ). As a result, it is
always possible to fit a Rh-0 EXAFS function with Rh-Ti parameters , and vice versa. A difference in backscattering amplitude can
be compensated with a particular (i.e. wrong) choice of coordination
number and Debye-Waller factor. Because of a different phase func
tion, however , such a fit will result in incorrect coordination distances. Clearly , to discriminate between Rh- Ti and Rh-0 contri bu
tions, a very careful data analy s is procedure is required. The procedure which we have used in our analysis procedure is superior to
the procedure described in (36). We believe that the assignment of
EXAFS and HRTEM of Rh/Ti0 2 page 170
the Rh-X contribution to oxygen neighbors is correct and that the
conclusion drawn by Sakelson et al. about RhTi alloy formation in
the SMSI state, is not justified.
7 .5 Final Conclusions
From tbe observed neighboring Ti ions, we conclude that in
the SMSI state the metal particles rest on a TiOx suboxide. This has already been suggested in the first reports on SMSI ( 1,3). In the
SMSI state, the metal p<,1rticles are incapable of adsorbing H2 and as
a consequence the Rh-Rh distance has contracted by about 0.05 A. The Rh0-02- distance in the metal-support interface has contracted
as well, by about 0.18 A. Th is is the result of an interaction between metal and support. In the SMSI state, the rhodium metal
particles are capable of adsorbing 0 2 at 100 and 300 K. Therefore, and because we found no evidence for a covering suboxide with
EXAFS nor with HRTEM , we conclude that coverage is not complete or is even absent in Rh/Ti0 2. Oxidation of the metal particles
in the SMSI state, however, is suppressed, even at 300 K. This
again must be the result of an electronic metal-support interaction
and cannot be explained by coverage.
Another interesting phenomenon observed has been observed
with EXAFS. When warming the oxygen exposed Rh/Ti0 2 catalyst to room temperature, the TiOx suboxide in the vicinity of the metal
particles starts to reoxidize. At the same time, the metal-support interaction weakens . This proves that indeed the suboxide underneath and around the metal particles plays a key role in the anomalous properties of the supported rhodium metal particles in ·
the SMSI state. At this point , however, it is not clear what the role of the suboxide is . A clue for further studies may possibly be found
in the contribution present around 2.1 A in the EXAFS spectra of the Rh/Ti02 catalyst in the SMSI state.
page 171 Chapter 7
During the past years, the majority of the literature dealing
with SMSI reported coverage (9-17 ). Since the results of this study
do not support complete coverage, we will focus on covering in the
following. Covering of metal particles by an oxide can take place in
two ways. The covering oxide may have a chemical interaction, or
may have no interaction at all with the metal particles. When there
is an interaction between covering oxide and metal particle, the
oxide will 'wet' the metal surface. Most authors see this kind of
covering as the origin for the decrease in adsorption capacity of
metal particles in the SMSI state. Such a covering would greatly
enhance the number of 0 2- neighbors for the metal atoms in the
surface of the metal particles . Since in such a case there is an
attractive interaction between oxide and metal particle (and because
the state of coverage has been formed at high temperatures), the
disorder in these Rh0-02- bonds that are formed in the SMSI state
is expected to be small. Therefore, EXAFS should be able to detect
such bonds. For instance , in the 5 atom metal particle in Figure 7.9,
each rhodium atom is a surface atom. In the case of complete cov
erage, the interfacial rhodium atoms should have seven (three more
than in the 'normal' state) 0 2- neighbors and the top rhodium atom
should have eight new 0 2- nearest neighbors. Therefore, the Rh0-
02- coordination number should increase at least by a factor of 2 or
3. For the Rh/ Al 20 3 catalyst reduced at 623 K and oxidized at
room temperature, the Rh0-Rh0 coordination number decreased by a
factor 3, the Rh0-02- coordination number by a factor 1.2 (cf. Table 7.2). Relative to the Rh0-Rh0 coordination number, the Rh0
-
02- coordination number increased by a factor of 2.5. From this we
concluded that the metallic kernel of the oxidized particles was
covered with rhodium oxide and it proves that EXAFS can detect a
covering oxide. For the Rh/Ti02 catalyst in the SMSI state we cer
tainly did not find any evidence for such an increase in Rh0-02-
coordination number . In this study, after reduction at 723 K. the
Rh0-02- coordination number increased from 1.3 ± 0.3 to 1.9 ± 0.3. In our previous study, we found a small decrease in the Rh0-0 2-
coordination number . The analysis of the EXAFS spectrum of the
Rh/Ti02 catalyst reduced at 473 K was hampered by the presence
EXAFS and HRTEM of Rh/Ti0 2 page 172
of Rh20 3. Therefore. the Rh0-02- coordination number for this
catalyst is rather inaccurate (after correction. this coordination
number is 1.3) and. in our view, too low. Therefore. the apparent
increase in the Rh 0..,o2- coordination number when increasing the
reduction temperature is insignificant and certainly too small to jus
tify the conclusion that the metal particles are tightly covered with
TiOx . Based on these Rh0-02- coordination numbers. one could
state that perhaps a small (partial) covering of the rhodium metal
particles has taken place. It is also possible that a very loose cover
ing occurs: this induces only a small number of additional 0 2-
neighbors. In the following paragraph we will discuss the effects of
such a loose covering.
When there is no interaction between metal and support. wet
ting and tight packing of the oxide around the metal particles will
not take place. Coverage then simply means the physical presence
of oxide crystallites over the metal particles. The spread in the
Rh0-0 2- distances will then be large, which makes it impossible for
EXAFS to detect the oxide (38). Based on the EXAFS results . we
therefore cannot exclude the presence of a loosely bound TiOx
suboxide on top of the metal particles. Such a coverage does not
have to be complete. but may still be able to suppress adsorption of
hydrogen and carbon monoxide .
From the oxygen adsorption experiments we learned that sur
face metal atoms are exposed . Our results therefore exclude com
plete coverage for rhodium metal particles supported on Ti02 in the
SMSI state. The metal particles may. however, be partially covered
with TiOx if this covering oxide does not tightly adhere to the metal
surface. But such a loosely packed covering oxide. which does not
interact with the metal particles. cannot be responsible for the
observed behavior of the SMSI state catalyst during oxygen expo
sure.
page 173 Chapter T
We feel that a loose coverage may occur in the SMSI state and
may also explain the difference in catalyst performance between
metals supported on the traditional supports like A1 20 3 and on
reducible supports . (see for example ref (39 )) . It is our conviction,
however. that coverage alone cannot explain the large decrease in
adsorption capacity (usually one order of magnitude) of the rho
dium metal particles in the SMSI state in the discussed Rh/Ti02 catalyst. From this and many other studies in this field (7 ,10-13,16-20) it is clear that SMSI is a complex phenomenon and may
vary from system to system. It may well prove impossible to con
dense this complexity into one model which cari explain all the
results presented up to now in the literature on SMSI. Rather, the
explanation for SMSI may vary from system to system. The
present study indicates that coverage is a phenomenon unlikely to
occur in Rh/Ti02. Although , based on the EXAFS results , coverage
cannot completely be excluded, it is evident that if a covering oxide
is present, the interaction between this covering oxide and the sur
face metal atoms is weak. But such a loose packed weakly interact
ing oxide cannot account for the fact that oxidation was greatly
suppressed when the surface metal atoms were exposed to oxygen.
Therefore, an electronic perturbation is the most likely explanation
for the anomalous properties of these small rhodium metal particles
supported on Ti02 in the SMSI state.
7 .5 References.
1. Tauster , S. J .; Fung, S. C. ; Garten , R.L. J. A m . Chem. Soc. 1978, 100, 170.
2. Tauster , S J.; Fung, S. C. J. Catal. 1978, 55, 29.
3. Tauster , S. J .; Fung, S. C. ; Baker , R. T. K .; Horsley , J . A. Science (Washington , D .C.) 1981 , 211, 1121.
4. Horsley , J. A. J . Am. Chem. Soc . 1979, 101 , 2870.
EXAFS and HRTEM of Rh/Ti0 2 page 174
5. Baker, R. T. K.; Prestridge, E. B.; Garten, R. L. J. Catal. 1979, 56,
390.
6. Baker, R. T. K.; Prestridge, E. B.; Garten, R. L. J. Catal. 1979, 59,
293.
7. van 't Blik, H. F. J.; Vriens, P. H. A.; Prins, R. ACS Symposium
1985, 298, 60:
8. Meriaudeau, P.; Ellestad, 0. H.; Dufaux, M.; Naccache C. J. Catal.
1982, 75, 243.
9. Meriaudeau, P.; Dutel, J. F.; Dufaux, M.; Naccache C. 11
Studies of Surface S~ience and Catalysis 11 1982, 11.
10. Belton. D. N.; Sun, Y. -M.; White, J. M. J. Phys. Chem. 1984, 88,
1690.
11. Belton, D. N.; Sun, Y.. -M.; White, J. M. J. Phys. Chem. 1984, 88,
5172.
12. Simoens, A. J.; Baker, R. T. K.; Dwyer, D. J.; Lund, C. R. F.; Madon, R. J. J. Catal. 1984, 86, 359.
13. Chung, Y. M.; Xiong, G.; Kao, C. C. J. Catal. 1984, 85, 237.
14. Sadeghi, H. R.; Henrich, V. E. J. Catal. 1984, 87, 279.
15. Sun, Y. -M.; Belton, D. N.; White, J. M. J. Phys. Chem. 1986, 90, 5178.
16. Ko, G. S.; Gorte, R. J. J. Catal. 1984, 90, 59.
17. Raub, G. B.; Dumesic, J. A. J. Phys. Chem. 1984, 88, 660.
18 Kelley, M. J.; Short, D. R.; Swartzfager, D. E. J. Catal. 1983, 20, 235.
19. Beard, B. C.; Ross, P. N. J. Phys. Chem. 1984, 90, 6811.
20. Brewer, L. 11 Phase Stability in Metals and Alloys 11; Rudman, P.;
Stringer, J.; Jaffee, R., ed.; MacGraw-Hill : New York, 1967; pp 39-61.
21. Sinfelt, J. H.; Via, G. H.; Lytle, F. W. J. Chem. Phys. 1977, 67, 3831.
22. van 't Blik, H. F. J.; van Zon, J. B. A. D.; Huizinga, T.; Koningsberger, D. C.; Prins, R. J. Phys. Chem. 1983, 87, 2264.
23. Koningsberger, D. C.; Martens, J. H. A.; Prins, R.; Short, D. R.; Sayers, D. E. J. Phys. Chem. 1986, 90, 3047.
24. Kip, B. J.; Duivenvoorden, F. B. M.; Koningsberger, D. C.; Prins, R. J. Catal. 1984, 105, 26.
page 175 Chapter 7
25. Vis, J. C.; van 't Blik, H. F. J.; Huizinga, T.; van Grondelle , J .: Prins, R. J. Mal. Catal. 1984, 25, 367.
26. Cook, J W.; Sayers, D. E. J. Appl. Phys . 1981 , 52 , 5024 .
27 Crystal structures :
Rh metal Wycklwff Crystal Structures 1963, 1, 13.
Rh 20 3 Structure Reports 1974, 40a, 301.
RhTi Structure Reports 1964, 29 , 13.
28. van Zon, J. B. A. D.; Koningsberger, D. C. ; van 't Blik, H. F. J.;
Sayers, D. E. J. Chem. Phys . 1985, 12, 5742.
29. Duivenvoorden, F. B. M.; Koningsberger , D. C.; Uh, Y S.; Gates, B. C. J . Am. Chem. Soc. 1986 , 108, 6254.
30. van 't Blik, H. F. J. ; van Zon, J. B. A. D.; Huizinga, T.; Vis, J. C.;
Koningsberger, D. C.; Prins, R. J. Amer . Chem . Soc. 1985 , 107, 3139.
31. Zandbergen, H. W.; Martens, J. H. A . (to be published)
32. Koningsberger, D. C.; van Zon, J. B. A . D.; van 't Blik, H. F. J.; Prins,
R.; Mansour, A . N.; Sayers, D. E.; Short, D. R.; Katzer, J. R. J. Phys . Chem. 1985, 89, 4075 .
33. Stern, E. A .; Bunker , B. A.; Heald , S. M . Phys . Rev . B 1980, 21,
5521.
34. van 't Blik, H. F. J.; van Zon, J . B. A. D.; ~oningsberger, D. C.; Prins,
R. J . Mal . Catal . 1984, 25, 379.
35. Woning, J.; Santen , R. A. Chem . Phys . Leu. 1983, 101 (6), 541.
36. Sakelson, S.; McMillan, M.; Haller, G. L. J . Phys . Chem . 1985 , 90 ,
1733.
37. Teo, B. K .; Lee , P. A. J. Am. Chem. Soc. 1979, 101, 2815.
38. Eisenberger , P.; Brown, G. S. Solid State Comm. 1979, 29, 481.
39. Levin, M. E.; Salmeron, M.; Bell, A. T.; Somorjai, G. A. Far. Symp. Chem. Soc. 1986, 21, paper 10.
EXAFS and HRTEM of Rh/Ti0 2 page 176
page 177 Chapter 8
Chapter 8
Strong Metal-Support Interactions in Rh/Ti02 Prepared with Ion Exchange
8.1 Introduction
The special properties of small metal particles supported on reducible transition-metal oxides have attracted considerable atten
tion during the last decade. As Tauster et al. already pointed out in their first papers ( 1-3), the properties of these catalysts after reduction at low temperature resemble the properties of such metal parti
cles supported on inert, unreducible support materials like Al 20 3 and Si02. After reduction at high temperature, however, the pro
perties change drastically. Most pronounced is the decrease in hydrogen and carbon monoxide adsorption capacity, suggesting that
the exposed metal surface has decreased due to sintering. Electron Microscopy revealed, however, that the metal particle size remained
virtually unaffected during the high temperature reduction. Several models have been proposed to explain this discrepancy. The first model, adopted by Tauster et al. ( 1-3) and later by many other authors, assumes a strong electronic interaction between metal par
ticle and the reduced support. Another model assumes covering of the metal particles by reduced support oxide species to be responsible for the SMSI state of the catalyst (4-7 ). Alloy formation is
another model proposed to explain the incapability of adsorbing
hydrogen and carbon monoxide by metal particles in the SMSI state ( 1-3,8). In a very recent EXAFS study (9). Sakelson et al. reported
alloy formation for a Rh/Ti02 catalyst in the SMSI state.
SMSI in Rh/Ti0 2 ex ion exchange page 178
In two recent EXAFS studies ( 10,11) and in chapter 7 of this
thesis, we reported on the structure of a Rh/Ti02 catalyst in the
normal and in the SMSI state. In these studies, we reported that
metal particles in the SMSI state rested on reduced Ti02 and also, that there was no evidence for alloy formation or for coverage of the
metal particles by reduced TiOx species. Consequently, we con
cluded that the decrease in hydrogen adsorption capacity and the
suppression of oxidation after reduction at 723 K was the result of
an electronic interaction between metal and support. On the other hand, Sakelson in the group of professor Haller (Yale University)
(9) reported alloy formation in the SMSI state, based on EXAFS
results for a similar Rh/Ti02 catalyst in the SMSI state. This
catalysts had been prepared with ion exchange, measured at CHESS
and analyzed with their own data analysis procedure. Thus. there was a serious problem. The results of Sakelson, alloy formation,
contradicted the results of our study, which clearly indicated no
alloy formation. The difference between their and our Rh/ Ti02 catalyst was the preparation method. The Rh/Ti0 2 sample in (9)
was prepared with ion exchange using ammonia, reduced and finally calcined, while the sample in our studies ( 10,l J) was prepared with
ion exchange without ammonia and was directly calcined. However,
also the EXAFS analysis procedure used by Sakelson differed from
the analysis procedure we used. Hence, the difference between the
two studies may be either the result of a different preparation
method or of a different analysis procedure. We reached an agree
ment with professor Haller that we should try to solve this contrad
iction. Professor Haller supplied us the raw EXAFS data of the
Rh/Ti02 sample of the Yale University which were used for the study in (9), in order to subject them to our data analysis pro
cedure. In this chapter we will describe the results of the analysis
of these EXAFS data from Yale University using the Eindhoven
data analysis procedure.
page 179 Chapter 8
8.2 Experimental
8.2.1 Catalyst Preparation
The catalyst was prepared at the Yale University using ion
exchange. The Ti0 2 was first immersed in a NH 40H solution at
pH= 11 for 15 h: this solution was stirred continuously. The Ti02 was then washed with distilled water until a pH of 7.5 was
obtained, filtered and dried for 24 h at 373 K. The ammonia treated
Ti0 2 was placed in a round bottom flask with water in a ratio of
60 cm3 g-1 Ti0 2. The temperature was raised to about 323 K and,
while vigorously stirring, 1 cm3 of a Rh(N03h solution was added
per gram Ti0 2. The solution contained 6.6 mg Rh cm-3 and was
added over a period of 5 h. Subsequently, the solution was allowed
to stir overnight at room temperature. The resulting catalyst was
centrifuged and washed several times with hot distilled water,
allowed to dry at room temperature for several days and then dried
for 5 h at 383 K. After reduction in flowing hydrogen at 773 k for
2 h and oxidation at 623 k for 2 h, the catalyst was stored in a
desiccator. The metal loading measured by HCI extraction and
atomic absorption was 0.47% and the H/Rh determined at Yale University with hydrogen chemisorption was 1.00± 0.05.
8.2.2 EXAFS Measurements
The X-ray absorption spectra of the rhodium K-edge of the
samples were measured by the Yale group at the Cornell high
energy synchrotron source (CHESS). All spectra were measured
with the sample cooled to liquid nitrogen temperature . The samples were pressed into thin self supporting wafers of approximately
20 mg cm-2. Two of these wafers were placed into a cell which
SMSI in Rh/Ti0 2 ex ion exchange page 180
allows in situ reduction. The design of the cell is such that in case
of inleakage, the influx of gases into the catalyst section consisted
of pure helium. The hydrogen flow was high enough to ensure no significant dilution with helium. X-Ray absorption spectra of the
reference compound were measured as well. The EXAFS spectra of
the reference compounds, rhodium foil, Rh02 and RhTi were also
measured by the Yale group at CHESS. These EXAFS spectra were
used to extract backscattering amplitude and phase shift functions
for the Rh-Rh, Rh-0 and Rh- Ti absorber-scatterer pairs; using the Eindhoven data analysis procedure.
8.3 Results
The extraction of backscattering amplitudes and phase shift
functions for the absorber-scatterer pairs Rh- Rh, Rh-0 and Rh-Ti
has been described extensively in chapter 7. The crystallographic
data for the reference compounds were taken from ( 13) and are
summarized, together with Fourier transform ranges in Table 8.1. Briefly, the analysis procedure was as follows. The EXAFS func
tion of the reference compound was Fourier transformed over the indicated range in k-space. An inverse Fourier transform over a lim
ited range in r-space yielded the EXAFS function for the mentioned
absorber-scatterer pairs. From these functions, backscattering
amplitude and phase shift functions could be derived. For the Rh Ti reference compound, first a best fitting Rh- Rh contribution was cal
culated and this EXAFS function was subtracted from the experi
mental data . Froms this difference spectrum, the desired Rh-Ti
backscattering amplitude and phase shift function were acquired as
described above .
page 181 Chapter 8
Table 8.1 Crystallographic data and Fourier transform ranges for the reference compounds
Rb Fourier transformation
a
b
(
d
Compound NN;i
Rh foil Rh 2.687
Rh02 0 1.963
Rh Ti Rh 2.949 Ti 2.676
Nearest Neighbor
Coordination Distance (A)
Coordination Number
Ne d n
12 3
6 1
4 3 8 1
Weighting factor in Fourier transformation
Crystallographic data were obtained from ( 12)
k-range r-range
2.81 - 14.50 1.52 - 3.02
2.49 - 16.98 0.00 - 1.84
2. 73 - 15.40 2.86 - 13.96 1.10 - 2.56
The procedure of analyzing the Rh/Ti0 2 EXAFS data has been described in chapter 7 and in ( 10, 11 ). Briefly, the procedure comprised the following steps. For all spectra , the Rh-Rh contribution was dominant. A k 3-weighted Fourie~ transform was used to
obtain a first estimate for the Rh- Rh parameters (N , R , !:!:.a 2 and t:J.£ 0) . Using these parameters , a Rh- Rh EXAFS function was cal
culated and subtracted from the experimental data. The resulting difference file was used to analyze the residual Rh-0 and Rh- Ti con
tributions . The calculated best fitting Rh-( O+ Ti) EXAFS function
was subtracted from the experimental data and the resulting
difference spectrum, now containing mostly Rh-Rh contribution, was used to optimize the Rh-Rh parameters . In this way, a recurrent optimization process was started that converged to the final set of parameters as presented in Table 8.2. In Figure 8.1 are shown the raw EXAFS spectra , their k 1-weighted Fourier transforms and the Fourier transforms of the best fitting Rh-Rh
a Estimated overall (experimental+ systematic) error b
fj,a 2, the Debye Waller factor , is a measure for the disorder and E0 is a
correction on the edge position .
contributions . In Figure 8.2, the k 1-weighted Fourier transforms of the difference spectra and the best fitting Rh- (O+ Ti) EXAFS functions are shown as well as the k1-weighted Fourier transforms of
the raw data and the best fitting Rh-( Rh+O+ T i) EXAFS functions . Clearly, even in the difference files (Figures 8.2a-c) , the calculated spectra are in excellent agreement with the measured spectra, demonstrating the reliability of the results .
Figure 8.2 Imaginary parts of the k1-weighted Rh-Rh corrected Fourier transforms of the difference files (raw EXAFS minus calculated Rh-Rh contribution , solid lines) and the best fitting calculated Rh-(0+ Ti) contributions (dotted lines) for (Fourier transform ranges are indicated in brackets) :
(a) Rh/Ti02 after reduction at 494 K (4.00- 9.00 .&.- 1)
(b) Rh/Ti02 after reduction at 628 K (3.33 - 8.75 .&.-1)
(c) Rh/Ti02 after reduction at 773 K (3.38 - 9.00 .&.-1)
Imaginary parts of the k1-weighted Rh-Rh corrected Fourier transforms of the raw EXAFS functions (solid lines) and the best fitting calculated Rh-(Rh+O+ Ti) contributions (dotted lines) for (Fourier transform ranges are indicated in brackets)
(d) Rh/Ti02 after reduction at 494 K (3.15 - 9.95 .&.-1)
(e) Rh/Ti02 after reduction at 628 K (3.38 - 10.74 .&.-1)
(f) Rh/Ti02 after reduction at 773 K (3.38 - 10.83 .&.- 1)
Apart from the Rh- 0 contribution at 2.0 A, a second Rh-0
contribution at 2.59 A was observed. This contribution can be ascribed to rhodium metal atoms having oxygen anions as neigh
bors . This kind of metal-oxygen contribution has been reported
before (see chapter 5 and ref 14-17). The oxygen ions may ori
ginate from the support and, in this case, also from unreduced rho
dium oxide which is in contact with the rhodium metal particles
(see chapter 5 and ref 15) . This can be concluded from the following observations . First of all. the accompanying Rh0-02
- coordina
tion number is relatively high . The corrected Rh0-0 2- coordination
number is 2.2/0.8 = 2.8. For similar metal particles we reported a Rh0-0 2
- coordination number of 1.9 ( 1 J). Secondly. the Rh-Rh dis
tance (2.630 A) was contracted significantly with respect to the dis
tance in bulk rhodium metal (2.687 A) and the Rh-Rh distance of the sample reduced at 628 K (2.668 A) . Such a decrease in Rh-Rh
distance has been ascribed to the absence of adsorbed hydrogen on the surface of the metal particles ( 18). This may be induced by the
SMSI state or may be the result of the presence of oxide species on
page 187 Chapter 8
top of the metal particles (see chapter 5 and ref ( 15)) . Although
some neighboring Ti ions could already be observed at 3.54 A, we
assume that the catalyst was not in the SMSI state because the
number of neighboring Ti ions was relatively low (in chapter 7 and in ref ( 11 ), coordination numbers ranging from 2.5 to 3.3 were
reported for the sample in the SMSI state). In addition, the reduc
tion temperature was rather low (SMSI behavior is observed usually
after reduction at temperatures above 600 K) . As reported in
chapter 7 and in ref ( 11 ), the Ti neighbors at 3.5 A are Ti ions in
the support underneath the metal particles. The coordination
number of 1.3 found in this study corresponds nicely with the value
based on the model described in chapter 7 and in ref ( 11) (for a 5
atom metal particle resting on unreduced Ti02, a Rh0-Ti coordina
tion number of about 1.5 may be expected). Since the measured and expected Rh0-Ti coordination numbers in the normal state
agree very well. we conclude that the particles fit well on the sup
port according to the model described in chapter 7 and in ref ( 11)
(see Figure 8.3) . Since the catalyst is not in the SMSI state, we conclude that the absence of hydrogen on the surface of the metal
particles, leading to a decrease in the Rh-Rh distance, is the result
of rhodium oxide species on top of or in contact with the metal par
ticles. In conclusion, after reduction at 494 K the situation was as
follows : about 80% of the rhodium was reduced . The metal parti
cles contained about 5 atoms and rested on unreduced Ti0 2.
Finally, the particles were covered or at least in contact with unre
duced rhodium oxide.
8.4.2 Rh/ Ti02 after Reduction at 628 K
After reduction at 628 K, the Rh-Rh coordination number has increased and the Rh3+-02
- contribution at 2 A. has decreased. The
estimated amount of unreduced rhodium oxide is about 10%. Therefore, the corrected Rh-Rh coordination number 1s
SMSI in Rh/Ti0 2 ex ion exchange page 188
Figure 8.3 A 13 atom metal particle with N=4.9 , resting on a anatase
1001] face .
ORh 0Ti4
+
( .Ti 4+ under Rh
4.6/0.9 = 5.1. corresponding to metal particles containing about 15 rhodium atoms (I 2). The amount of rhodium oxide still present
after reduction at 494 K was about 20%, while the metal particles contained approximately 4 to 7 rhodium atoms . This means that
for every metal particle, or for every 4-7 rhodium atoms in the metallic state , there will be one or two rhodium ions in the oxidic
state. After complete reduction of this rhodium oxide phase, the metal particles would, on the average, contain about 5 to 9 atoms,
which is clearly not the case : the corrected coordination number of 5.1 points to particles with 15 metal atoms . Obviously, some
sintering has occurred during further reduction. Since the catalyst had been pre-reduced at 773 K in the preparation, one might con
clude that during subsequent reduction experiments, no sintering should occur at reduction temperatures lower than 773 K. Neverthe
less, we observed sintering even at 628 K. This can be explained as described by Wang and Schmidt for Ir /Si02 (/9 ) . After pre
reduction at 773 K, metal particles are formed. During subsequent calcination, rhodium oxide is formed . This oxide phase may break
up into several smaller oxide particles. These smaller oxide particles are close together, but not necessarily in intimate contact. Upon reduction at lower temperatures, these smaller oxide particles
page 189 Chapter 8
will be reduced to small metallic particles . Since these metal parti
cles are not far apart, sintering will take place below the temperature of 773 K. After the sintering process has stopped, the metal
particle size wil I be equal or even larger than the metal particle size after the first reduction treatment at 773 K.
The Rh-Rh distance in the metal particles after reduction at 628 K (2 .668 A) has relaxed towards the Rh-Rh bulk distance
(2.687 A), indicating that there was no more coverage of rhodium oxide and that the metal particles had adsorbed hydrogen. Hence,
also this catalyst was not in the SMSI state. Because of the larger metal particles, the Rh0-02
- and Rh0-Ti coordination numbers can
not be compared directly to the corresponding coordination numbers for the smaller metal particles . Therefore, we take as a model a 13
atom rhodium metal particle which has 9 atoms in the metalsupport interface and 4 atoms on the top. In Figure 8.3, this metal
particle on a anatase [001) crystal face is shown. The 9 interfacial metal atoms constitute a 3*3 square and can be fitted adequately
on the Ti02 support (see the models in Figure 7.9 in chapter 7). The calculated coordination number for this particle is 4.9. When the Ti0 2 is not reduced, we may expect a Rh0-02
- coordination number of 2.5 and a Rh0-Ti coordination number of 1.5. Both coor
dination numbers measured are significantly lower. Hence, we conclude that the metal particles did not fit 'perfectly' on the support .
This may be the result of the sintering process : from the coordination numbers we may estimate that two to three rhodium particles
of about five atoms each, have sintered to one particle of about 15 atoms. Because of the low temperature, the particles may not have
reorganized completely and therefore. the structure of these particles may be imperfect and will not fit adequately on the support.
Confirmation for this can be found in the Debye-Waller factor , the disorder for the Rh-Rh contribution (see Table 8.2) . For the sample
reduced at 628 K, the Debye-Waller factor is higher by about a factor of two compared to the samples reduced at 494 and 773 K. Concluding, after reduction at 628 K, the catalyst was not in the SMSI state. The particles had sintered and were covered with
SMSI in Rh/Ti0 2 ex ion exchange page 190
adsorbed hydrogen . The reduction process was not complete .
8.4.3 Rh/Ti02 after Reduction at 773 K
After reduction at 773 K, there was no evidence for unreduced
rhodium oxide. Clearly, reduction was complete. The Rh-Rh coor
dination number was 4.9, in excellent agreement with the corrected
coordination number measured after reduction at 628 K and indicat
ing that no additional sintering had occurred during reduction between 628 and 773 .K. The Rh 0-0 2
- and Rh0-Ti coordination
numbers both increased significantly. In addition, the Rh0-0 2- coor
dination distance decreased markedly. We ascribed this decrease to
the interaction between metal and support (see chapter 7 and ref ( 11 ). Because of this enhanced interaction and because of the
higher reduction temperature, the metal particles had reorganized
(the Debye-Waller factor had decreased) and fitted better on the
Ti0 2 support. The Rh0-Ti coordination number is higher than for a
metal particle on unreduced Ti0 2 (from the model described above,
this was estimated to be about 1 .5). Obviously, as reported in
chapter 7 and in ref ( 11 ), the particles rested on reduced Ti0 2. In
addition, the Rh-Rh distance has again decreased . Hence, the sample was in the SMSI state and the metal particles had not adsorbed
hydrogen. We found no titanium neighbors at metallic distances.
Thus, alloy formation can be ruled out as an explanation for the
SMSI state. Also, the measured Rh0-0 2- coordination number (2.0)
was lower than the expected coordination number for the 13 atom
metal particle ( ± 2.8), indicating that the particles were not covered
with reduced TiOx. The fact that the measured coordination
number was lower than the expected coordination number has been
explained in ( 11) as being the result of the fact that the reference
Rh-0 distance (1.963 A) was shorter than the calculated Rh-0 distance (2.51-2.77 A) , giving rise to an error in the mean free path
term in the EXAFS equation ( 16). Hence, we cannot but conclude
page 191 Chapter 8
that an electronic interaction is responsible for the SMSI state.
8.4.4 General Remarks
The results presented above agree nicely with the results of
our earlier paper ( 11 ). However, there are differences. First of all,
even after reduction at 628 K, the sample was still not completely
reduced. The reason for this may be found in the preparation method. The support was immersed in water and stirred vigorously
several times. Consequently, the support has been powdered very
finely. As a result , after exchanging ammonium groups with rho
diu m nitrate, the support had to be centrifuged in order to separate
it from the solution . When such a finely powdered support is
pressed into a wafer , the wafer may be very compact and gases will
diffuse only slowly through the sample. Consequently, hydrogen will
diffuse only slowly into the catalyst wafer and the water formed
during the reduction process will diffuse only very slowly out of the
catalyst sample. Hence , the reduction process will continue very
slowly and unreduced rhodium oxide may be found in the sample
even at relatively high reduction temperatures.
Another difference can be found in th~ Rh0-Ti contributions. In
the present study we found only one contribution clearly present :
the 3.5 A contribution. This distance is within the experimental
error equal to the distances reported in (10,JJ). In (JJ) , however ,
an additional contribution at 4.3 A was found . This contribution
has not been observed clearly in the present sample. In the difference spectra in Figure 8.2a-c , however , small differences at
about 4.3 A are still visible and these could be fitted with Rh0-Ti
contributions, but the accompanying coordination numbers were
Very small , N < 0.4, in contrast to 1 .4 < N < 2.8 in ( 11 ). Although the particles were larger, we would still expect this contri
bution to be present . However, as argued above, the reduction process may have been delayed . As a consequence. the reduction of
SMSI in Rh/Ti0 2 ex ion exchange page 192
the support may have been limited to the direct environment of the
metal particle , whereas in ( 11 ), the support was reduced to a larger
extent . This is the result of the more open structure of the support
used in ( 11) (the Ti02 in ( 11) has not been stirred using a mag
netic bar) . Hence, the reduction process of both metal particle and
support could proceed more easily and thus. the SMSI state could
be invoked at lower temperatures . Consequently. the binding of the
metal particles to the (reduced) support was strong already at low temperatures. This explains why we did not observe any sintering
after reduction at 773 K in chapter 7 and in ( 11) (for that sample,
no pre-reduction has been applied), whereas in this study sintering
was already observed at 628 K, notwithstanding the higher metal
loading (4 wt% in ( J J) and 0.47 wt% in this study).
Now we come to the most important part of this study.
Clearly, our data analysis procedure produces results completely
different from the results obtained by the procedure followed by
Sakelson. The explanation for this discrepancy is in fact very sim
ple. Sakelson used a k 3-weighted Fourier transform to study the
data. The transform range was in all cases choosen from k=3.0 to
15.0 A_-l and this is the major reason for the differences in the
results. Above k=11 A-1 the noise level increases drastically. For
the samples reduced at 494 and 628 K even a few glitches and jumps had to be removed from the raw data in order to obtain reli
able EXAFS functions. Anyhow, above k=11 A.- 1, the data are
unreliable . This can be seen in Figure 8.4, in which we have plotted
the EXAFS function of the sample reduced at 773 K, weighted by
k 3 and the k 3-weighted EXAFS function which fitted these data
between k = 3.38 and 10.83 A-1 and was calculated using the param
eters from Table 8.2. Note. that these functions are in fact the
functions that are transformed in a k 3-weighted Fourier transform.
The differences are clearly visible and these differences are the result of the noise in the spectrum above k=11 A_-l. As a result , a Fourier transforms up to k=15 A_-l will result in a strongly distorted
radial distribution function and will inevitable result in incorrect parameters.
page 193 Chapter 8
Figure 8.4 k3*CHI for the experimental data of the sample reduced at 773 K (solid line) and the calculated best fitting EXAFS func
tion (dotted line)
13
9
-----..._ 5 ~ '---
f-j 1
J: lJ * -3
('T)
~ -7
-11
-15 0 5 15
k
In general, next to the potential errors indeuced by choosing
wrong Fourier transform ranges, many more pittfals are present in
the data analysis procedure. One can think of removing glitches
and jumps and background subtraction. Another very important
aspect is imposed by the reference compounds. Even more care
should be taken during the data analysis procedure, for backscatter
ing amplitudes and phase shift functions obtained from reference
compounds are especially sensitive towards incorrect background
subtraction and Fourier transform ranges. And unreliable reference
spectra will only add to the unreliability of the analysis of the spectra of the catalyst samples. Obviously, analyzing EXAFS spectra
should be carried out with the utmost caution and only a data
analysis performed with such utmost care will result in reliable
parameters.
SMSI in Rh/Ti0 2 ex ion exchange page 194
8.5 Final Conclusions
After reduction at 494 K, reduction was not complete and very
small metal particles were formed, containing about 5 atoms. Because of their small dimensions, these particles fitted easily on
the support. Unreduced rhodium oxide covered the metal particles.
After reduction at 628 K, still 10% of the rhodium oxide was not
reduced: the particles were not in the SMSI state. The binding of the metal particles to the support was relatively weak. Conse
quently, the ·metal particles had sintered and the larger metal parti
cles did not fit perfectly on the support. After reduction at 773 K.
the reduction process was complete and the catalyst was in the SMSI state: the supp.ort underneath the metal particles was
reduced . Because of the stronger binding of the metal particles to the support in the SMSI state, the metal particles had rearranged
with respect to those after reduction at 628 K and fitted more perfectly on the support . Because of the deferred reduction process,
the SMSI state was invoked only at relatively high temperature. In the SMSI state no evidence for alloy formation nor for coverage was
found . Thus, we conclude that an electronic interaction between metal particle and reduced support is responsible for the SMSI state
with particles of the size as studied in this chapter .
The differences between the results of the data analysis followed by Sakelson in (9) and the data analysis procedure used at the Eindhoven University could be ascribed to the increased noise
level at higher k-values and an incorrect choice of Fourier transform ranges .
page 195 Chapter 8
8 .6 References.
1 Tauster . S J; Fung, S C.; Garten, R.L. J. Am Chem . Soc . 1978. JOO. 170
2 Tauster . S J.; Fung. S. C ] . Catal. 1978 . 55, 29
3 Tauster . S. J; Fung, S C. ; Baker , R T. K.; Horsley, J . A Science
{Washington, D.C.) 1981. 211, 1121 .
4. Meriaudeau, P.; Dutel , J F ; Dufaux. M.; Naccache C "Studies of Surface Science and Catalysis" 1982. 11 .
5 Belton . D. N : Sun . Y. -M; White , J . M J . Phy.1. Chem . 1984. 88,
1690.
6 Simoens. A. J.; Baker. R. T. K: Dwyer , D J; Lund , C RF ; Madon . R. J. J. Catal . 1984 . 86. 359.
7. Sadeghi. H. R.; Henrich , V E. J . Catal. 1984. 87, 279.
8 Beard , B. C. ; Ross . P. N. J. Phys. Chem . 1984. 90 . 6811 .
9. Sakelson , S .; McMillan , M.; Haller, G. L. J . Phys . Chem . 1985, 90, 1733
10. Koningsberger, D. C.; Martens , J. H. A.; Prins, R. ; Short, D R : Sayers, D. E. J . Phys . Chem. 1986, 90 , 3047.
11. Martens, J . H. A ; Prins, R. ; Koningsberger, D. C. J . Phys . Chem.,
accepted for publication .
(chapter 7 of this thesis)
12. Kip, B. J.; Duivenvoorden, F. B. M.; Koningsberger , D. C.; Prins , R J. Catal. 1987, 105 , 26.
13. Crystal structures •
Rh metal Wyckhoff Crystal Structures 1963. 1 . 10.
Rh0 2 Structure Report s for 1968 1975 , 33a , 271 .
RhTi Structure Reports f or 1964 1972 , 29 , 13
14 Koningsberger, D. C.; van Zon , J . B. A D; van 't Blik, H.F . J .; Mansour. A N ; Visser, G. J .; Prins , R; Sayers , D. E.; Short, D. R.; Katzer , J . R. J . Chem . Phys . 1985 , 89, 4075
15. Martens , J H. A ; Prins , R ; Koningsberger , D. C. J . Phys . Chem.,
submitted for publication
SMSl _in Rh/Ti02 ex ion exchange page 196
(chapter 5 of this thesis)
16. van Zon, J. B. A. D.; Koningsberger, D. C.: van ·t Blik, H. F. J.; Sayers, D. E. J. Chem. Phys. 1985, 12, 5742.
17. Koningsberger, D. C.; Duivenvoorden, F. B. M.; Kip, B. J .; Gates, B. C. "EXAFS and Near Edge Structure"; Lagarde, P.; Raoux, D.; Petiau, J. Eds.; Les Editions de Physique, 1986; vol. 1, p. C8-255 .
18. van 't Blik, H. F. J.; van Zon, J. B. A. D.; Koningsberger, D. C.; Prins, R. J. Mol. Catal . 1984, 25, 379.
19. Wang, T.; Schmidt, L. D. J. Catal . 1980, 66 , 301
page 197 Chapter 9
Chapter 9
EXAFS Evidence for Direct Rh0 -Ta"+ Bonding and Coverage of the Metal Particles
in a Rh/Ta20 5 Catalyst in the SMSI State
9.1 Introduction
Since the discovery of the phenomenon of Strong MetalSupport interaction (SMSI) in the late seventies ( J-3), a lot of research effort has been devoted to the behavior of various metal catalyst systems in this SMSI state and to the search for the explanation of the anomalous properties of such catalysts. Shortly, the SMSI state can be defined as follows : after reduction at lower temperatures, the properties of metal particles supported on selected transition metal oxide supports can be classified as 'normal'. The
capacity to adsorb hydrogen and carbon monoxide is one of those properties. After reduction of the catalyst at higher temperatures, the capacity to adsorb hydrogen and carbon monoxide diminishes drastically. This reduced adsorption capacity is always observed for such catalyst systems and is therefore generally used to define the state of a catalyst.
In two recent studies ( 4,5) and in chapters 7 and 8, we reported on the structure of a titania supported rhodium catalyst in the normal and the SMSI state. From these studies it became evident that in the SMSI state alloy formation (which is one of the proposed explanations for SMSI) had not taken place. Although it was found that coverage of the metal particles by reduced support species (coverage or decoration, another explanation for SMSI that has been proposed in the literature) was unlikely to occur, coverage
page 198
could not be excluded completely. Oxygen adsorption experiments
indicated that oxidation was suppressed, but that the metal particles were exposed to the gas atmosphere and were covered with
absorbed gaseous oxygen ( 5). A Rh/ A120 3 catalysts become oxidized under these conditions (see chapter 5). Thus, if the metal par
ticles in the Rh/Ti02 sample in chapter 7 were covered with a TiOx suboxide, this coverage was not complete and could not explain the
behavior of the catalyst in the SMSI state. In the EXAFS spectra
of the catalyst in the SMSI state, contributions from neighboring
titanium ions at 3.4 and 4.3 A were present. It was concluded that
the metal particles rested on reduced titania. Since this was the
only change in the EXAFS spectra after reduction at high tempera
ture, it was suggested that there was an interaction between the
supporting oxide underneath the metal particles and the metal particles themselves. In a model of the structure of small rhodium metal
particles on (001J and (101J anatase crystal faces after reduction at high temperature, Rh0-Ti"+ coordination distances were observed
which were equal to those present in the EXAFS spectra ( 5 ). However, in this model titanium ions at about 2 A are present as well.
The number of titanium neighbors at 2 A is lower than the number of titanium neighbors at 3.4 and 4.3 A by more than a factor 2.
Because of this , and because of the fact that titanium has a relatively low backscattering amplitude, the contribution of these
titanium ions is expected to be small. In practice, in the Fourier
transform of the EXAFS spectra of the rhodium catalyst in the
SMSI state, a small contribution was present at the low r-side of
the main contributions. Although this contribution could be fitted
with a Rh0-Ti"+ contribution at about 2 A, the contribution was too low to allow a reliable analysis.
In the present study we have tried to observe these support
cations at short distance by modifying the Rh/Ti0 2 system. The
coordination number is a quantity which we do not have in hand. However, by choosing another cation than that of titanium, we can
manipulate the backscattering amplitude and use it to our advan
tage. In this study we will describe the results of an EXAFS study
page 199 Chapter 9
of a Ta 20 5 supported rhodium catalyst. We choose Ta 20 5 because
it is known to be an SMSI support ( 2) and, more importantly, because tantalum has a backscattering amplitude which is higher by
a factor 2 to 4 than that of titanium at higher k-values . In Figure 2.5 (chapter 2), the backscattering amplitudes for Rh, Ta and Ti
according to Teo and Lee (6) are shown . Clearly, at higher k-values, the contribution of tantalum is much more pronounced than that of
titanium. If tantalum ions are present at short distances in the SMSI state, we may expect that EXAFS will be capable of detecting such contributions.
9.2 Experimental
9.2.1 Catalyst Preparation
A high surf ace area T a20 5 support was prepared according to
the following procedure. 20 g T aCl 5 was dissolved in a 100 ml concentrated HCI solution. The resulting solution was added carefully
to a mixture of 4 I distilled water and ice, which was acidified with HCI to pH= 0.0. Approximately 300 ml of a NH 40H solution was
added dropwise to the TaCl 5 solution during a period of 100 min while vigorously stirring the solution . After all the ammonia had
been added to the TaCl5 solution, the pH had increased to 6.0. After stirring for another 100 min, the precipitated Ta(OH)s was filtered off and washEd several times with distilled water. thereafter it was carefully dried at 393 K for 24 h (heating rate 2 K min-1),
cooled down to room temperature. powdered, dried again as
described above, and finally calcined for 1 hr at 873 K (heating rate 2 K min -l). The resulting 7 g T a20 5 had a surface area of approxi-
2 -1 mately 100 m g .
page 200
A 3 wt % Rh/T a20 5 catalyst was prepared using the urea
method (7,8). 350 ml distilled water was stirred and heated to
365 K. 3 g of the high surface area Ta20 5 was added and the solu
tion was acidified with 8 N HCI to pH = 2.5. Then 0.53 g of urea (a tenfold excess based on the amount of RhC1 3) was added and
finally 0.2254 g of RhCl3. At 365 K the urea slowly decomposed
and the pH of the solution increased very slowly. At a pH value of
approximately 4, Rh( OH h started to precipitate. After 10 h the
catalyst precursor was filtered off and dried as described above for
the T a20 5 support. In order to remove the remainings of the urea,
the sample was calcined at 923 K, pre-reduced in hydrogen at 773 K
and finally oxidized at 573 K. This sample was stored for further
use. Temperature programmed reduction experiments indicated that
reduction was complete· at 470 K when using 4% H2 in N2. Hydro
gen chemisorption measurements after reduction at 523, 773 and
873 K gave H/Rh values- of 0.93, 0.14 and 0.06, respectively.
9.2.2 EXAFS Measurements
The dried, calcined, reduced and subsequently oxidized catalyst
was pressed into a thin self supporting wafer. The thickness of the
wafer was such as to give an absorbance (µx) of 2.5 at the rho
dium K-edge (23219.8 eV). assuring an optimum signal-to-noise
ratio in the rhodium EXAFS spectra. The wafer was mounted in an
EXAFS cell which enabled in situ pretreatments in different gas
atmospheres at temperatures ranging from 100 to 873 K. The sam
ple, once mounted in the cell, was reduced in 100 % H2 at 523 K for
1 h (heating rate 5 K min-1). After cooling with liquid nitrogen to
100 K an EXAFS spectrum was recorded with the sample still under
H2. Thereafter, the sample was reduced in H2 at 858 K for 15 min (heating rate 5 K min-1
), cooled down to 100 K and a second
EXAFS spectrum was recorded. Finally, the sample was evacuated
at 523 K for 1 h, oxygen was admitted to the sample at 100 K and
page 201 Chapter 9
another EXAFS spectrum was recorded, with the sample still under
oxygen atmosphere. The EXAFS spectra of the reference com
pounds, rhodium foil , Rh 20 3• RhCl3• Rh3 Ta alloy. Ta powder and
TaCl5, were recorded at 100 K as well. The absorption spectra were recorded at the synchrotron radiation source (S RS) in Dares bury.
U.K. The ring was operated at 2.0 GeV and with ring currents from 100 to 300 mA.
9.3 Results
9.3.1 Reference Compoonds
The backscattering amplitude F (k ) and the phase shift func
tion </> (k) which are necessary for analyzing the EXAFS data have been obtained from reference compounds. Following the same pro
cedure as described in chapters 2, 7 and 8, we extracted F (k ) and
</> (k) for Rh-0 contributions using Rh20 3, for Rh-Cl contributions using RhCl3, for Ta-Ta contributions using Ta powder and for Ta
Cl using TaCl5. In Table 9.1 all the relevant information concerning the references is given. The crystallographic data were obtained from (9).
Extracting F (k ) and </> (k ) for Rh-Ta contributions was not as straightforward as described above. In the most obvious refer
ence compound, the Rh3Ta alloy, rhodium has rhodium neighbors · and tantalum neighbors . . In the Fourier transform of the EXAFS
spectrum of the alloy, however, the Rh-Rh and Rh-Ta peaks over
lapped almost completely and ex fr acting F (k ) and </> (k ) for Rh-Ta was impossible. We therefore choose to compose F (k ) and </> (k) using other reference compounds. The basic principle is the following. The backscattering amplitude is only function of the scattering
atom (6). Therefore, F (k) is the same for the absorber-scatterer
page 202
Table 9.1 Crystallographic data and Fourier transform ·ranges for the reference compounds
Compound Edge NNa
Rh foil Rh . K Rh
Rh70 1 Rh. K 0
RhCI ~ Rh. K Cl
Ta powder Ta . L 111 Ta
TaCI~ Ta . L111 Cl
a : Nearest Neighbor b : Coordination Distance (A) t : Coordination Number ·
pairs A- B and B- B. Thus, we used F (k) from the Ta-Ta contribu
tion to represent F(k) for Rh- Ta. F(k) for Ta-Ta has been
extracted from the L111 EXAFS spectrum of tantalum powder. The
phase shift function <f>A-B (k) can be written as (6)
<t>A-B(k) = <t>t_(k) + <t>~(k) - 07T
In this equation, <f>J.(k) is the contribution of the absorbing atom A
and <f>s(k) of the scattering atom B to the phase shift in the EXAFS function. For K and L1 edges, withs-symmetry, the factor S is equal to unity , for L11 and L111 edges, with p-symmetry, S is zero.
We measured the Rh K-edge EXAFS spectrum of RhCl3 and the Ta
L111 EXAFS spectra of Ta powder and of TaCl5• From these EXAFS
functions we extracted ¢Rh-c1(k ), <l>Ta-Ta(k) and ¢Ta-c1(k ). A linear
page 203 Chapter 9
combination of these three functions yielded the desired phase shift
function for Rh-Ta contributions (6,10,J J ), as is shown below. Note, that the factor TT remains in the expression, so the resulting
phase shift function can indeed be used for Rh K-edge EXAFS spectra (even when for Ta powder and T aCl 5 the Ta K or L1 edges
Our procedure of analyzing the EXAFS spectra has already been described extensively in the literature (5,12-14) and in
chapters 2, 7 and 8. Briefly, the analysis comprised the following steps. Using the cubic spline routine (JS), a smooth background
was subtracted from the experimental data. The resulting EXAFS function was then normalized by division to the height of the edge.
Then, using the backscattering amplitude F (k) and the phase shift function </>(k) of suitable reference compounds, EXAFS spectra
containing one or more shells were calculated. In calculating EXAFS spectra, for each shell of neighbors, four parameters can be varied :
the coordination number N, the coordination distance R, the Debye
Waller factor !::.a 2 to account for any disorder, and E-o, which allows a small correction on the edge position . By varying these parameters one tries to make the calculated EXAFS spectra resemble the
page 204
measured spectra as accurately as possible. In two previous studies
(4,5) we described a recurrent optimization process in order to
separately analyze the contributions from high-Z and low-Z scatter
ing neighbors. Since in this study the contribution of the low-Z
scatterer (oxygen) turned out to be very small and the remaining
contributions originated from high-Z scatterers (rhodium and tantalum), this procedure could not be used. Moreover, the Rh-Rh and
Rh-Ta contributions overlapped in the Fourier transform (see Figure 9.2f), making it impossible to use the difference file technique. As a
result, the analysis we used was a single step multiple shell
approach wi.th four parameters (N, R. !J,.a 2 and Eo) for each shell to
optimize : we calculated a Rh-Rh EXAFS function and two Rh-Ta
EXAFS functions, added them and Fourier transformed the result- .
ing spectrum in order to compare with the (Fourier transform of the) measured data. Because of the high-Z character of the main
contributions (Rh and Ta), the use of k3-weighted Fourier transforms was essential. Differences between the two spectra were minimized by varying the Rh-Rh and Rh-Ta parameters. The
results of this analysis procedure are presented in Table 9.2.
In Figure 9.1a, the raw EXAFS data for the sample reduced at 523 K and the calculated best fitting Rh-Rh EXAFS function are
shown. The imaginary part is shown in Figure 9.1b and the magni
tude of the Fourier transforms of these EXAFS functions in Figure
9 .1c. From Figure 9.1, it is obvious that apart from the Rh-Rh contribution, no other contribution is present in the spectrum. In Fig
ures 9.2a and 9.3a the raw EXAFS spectra and the calculated best fitting Rh-Rh EXAFS functions for the samples after reduction at 858 K and oxygen admission are shown . In Figures 9.2b and 9.3b
are shown the imaginary parts of the k 3-weighted Fourier
transforms of the measured data and calculated Rh-Rh EXAFS functions . Figures 9.2c and 9.3c show the magnitude of these
Fourier tranforms. The differences in Figures 9.2b, 9.2c, 9.3b and 9.3c at the left hand side of the main Rh- Rh peak are due to the neighboring tantalum ions. We tried to fit these differences with rhodium and oxygen neighbors as well, but the fits resulted in
a b r:w 2, the Debye Waller factor , is a measure for the disorder and E0 is
a correction on the edge position.
physically irrelevant parameters : for Rh-Rh , coordination distances
of about 1 A and for oxygen coordination numbers higher that 10. In addition, the resulting fit was worse than the fit with tantalum
neighbors. In Figures 9.2d, 9.2e and 9.2f are shown the raw data
and the calculated best fitting Rh-(Rh+ Ta) EXAFS f1,mctions, the
imaginary parts of their k 3-weighted Fourier transforms and the
magnitude of the Fourier transform of the raw data and of the three
separate contributions. Clearly, the agreement at the left hand side
of the main Rh-Rh peak in the Fourier transform is better . The
Fourier transforms of the EXAFS spectra were complicated by the
k-dependence in F (k) and <!>(k ). Therefore. the transforms were
corrected for F (k) and c/>(k) from rhodium foil, the reference for the Rh-Rh contribution, which was the major contribution in all
spectra. As a result, in the Fourier transforms, the Rh-Rh contribu
tions ' peaked ' at the correct Rh-Rh distance and the imaginary
page 206
Figure 9.1 Rh/Ta20 5 after reduction at 523 K •
(a) Raw data (solid line) and calculated Rh-Rh EXAFS funct ion (dotted line)
(b) Imaginary parts of the k3-weighted Fourier transform of the raw data (solid line) and calculated Rh-Rh EXAFS function (dotted line)
(c) Magnitude of the k3-weighted Fourier transform of the raw data (solid line) and calculated Rh-Rh EXAFS function (dotted line)
* 10-2 30 b 5 ~~~~~~..--;-.,~~~~~
a
-30
40 c
20
-5 0 5 10 15 20 25
k [A-1 J 0
0 2 4 6
R [AJ parts of the Fourier transforms were more or less symmetric. The
peaks corresponding to other minor contributions can shift to seemingly longer or shorter distances and can be asymmetric. However, since the same correction has been applied to measured and calcu
lated data, the calculated coordination radii and coordination numbers represented those in the sample as accurately as possible.
page 207 Chapter 9
9.4 Discussion
9.4.1 Rh/ Ta20 5 after Reduction at 523 K
According to the TPR experiments, reduction of the sample
was complete at 523 K. A careful analysis of the EXAFS spectrum
confirmed this . From Figures 9.1a, 9.1b and 9.1c it is obvious that
apart from a Rh-Rh contribution, no other contribution was present.
The deviations in Figure 9.'.lb and 9 .1c are very small: the deviation
at the right hand side of the main peak could not be fitted with a
Rh- Rh. a Rh-0 or a Rh-Ta contribution . From Table 9.2 it can be concluded that on the average each rhodium atom had approxi
mately 7 .9 rhodium neighbors at 2.658 A, which is slightly shorter
Figure 9 .2 Rh/Ta20 5 after reduction at 858 K •
(a) Raw data (solid line) and calculated Rh-Rh EXAFS function
(dotted line)
(b) Imaginary parts of the k 3-weighted Fourier transform of the
raw data (solid line) and calculated Rh-Rh EXAFS function
(dotted line)
( c) Magnitude of the k 3-weighted Fourier transform of the raw
data (solid line) and calculated Rh- Rh EXAFS function (dot
ted line)
(d) Raw data (solid line) and calculated Rh-(Rh+ T a) EXAFS
function (dotted line)
(e) Imaginary parts of the k 3-weighted Fourier transform of the
raw data (solid line) and calculated Rh- (Rh+ Ta) EXAFS
function (dotted line)
(f) Magnitude of the k 3-weighted Fourier transform of the raw
data (solid line) , calculated Rh-Rh EXAFS function (dotted
line) , the calculated Rh-Ta EXAFS functions (dashed lines)
and the sum of the calculated Rh-Rh and Rh-Ta contributions
than the Rh-Rh bulk distance (the Rh-Rh distance in rhodium foil is 2.687 A) . Using the calibration procedure as described in ( 16), we
could estimate from the Rh-Rh coordination number that the parti
cles were approximately 17 A in diameter and contained about 73 ± 5 rhodium atoms. Acco~ding to ( 16), the H/Rh value deter
mined with hydrogen chemisorption should be about 0.95 according to both the experimental calibration method and the computer
model calculations. This agreed excellently with the measured value, H/Rh = 0.93, and therefore we conclude that the hydrogen
chemisorption capacity was 'normal' and thus, that the metal particles were in the normal state.
In the EXAFS spectra of fully reduced supported metal catalysts, a metal-oxygen contribution from the metal atoms in the metal-support interface having oxygen neighbors has frequently been reported (4,5,12,17). In the Fourier transform, such a contri
bution is situated at the left hand side of the main Rh-Rh peak. Since the relative extent of the metal-support interface decreases
with increasing particle size, the contribution from neighboring oxygen ions decreases with increasing particle size as well ( 12 .17). For
a 73 atom metal particle, about 30% of the rhodium atoms are situated in the metal support interface. When these rhodium atoms each have 2 to 3 oxygen neighbors, the average Rh0-02
- coordination number is about 0.6 to 0.9 and. because of the low coordina
tion number and the low backscattering amplitude of oxygen, the corresponding contribution in the Fourier transform should be
smaller by more than an order of magnitude compared to the Rh-Rh contribution. Accordingly, in the EXAFS spectrum of the sample
reduced at 523 K, no contribution from neighboring oxygen ions could be detected.
page 210
Figure 9.3 Rh/Ta20 5 after reduction at 858 K and oxygen admission at 100 K:
(a)
(b)
(c)
Raw data (solid line) and calculated Rh-Rh EXAFS function (dotted line)
Imaginary parts of the k3-weighted Fourier transform of the raw data (solid line) and calculated Rh-Rh EXAFS function (dotted line)
Magnitude of the k3-weighted Fourier transform of the raw data (solid line) and calculated Rh-Rh EXAFS function (dotted line)
* 10-2 30 b 6 ,,--.-.-,,....,-.-,--,-,...T"T""T"""T""">,,....,~"T"""T"""~~
After reduction at 773 K, the H/Rh value determined with hydrogen chemisorption decreased to 0.14, and after reduction at
873 K to 0.06. Clearly, after reduction at 858 K the metal particles
were in the SMSI state. The Rh- Rh coordination number remained
unchanged. Obviously, the basic structure of the metal particles
page 211 Chapter 9
remained intact. In the earliest studies on SMSI. it was suggested
that in the SMSI state the metal particles might spread over the support and that ' pillboxes' would be formed ( 18.19). Such a spread
of the metal particles should be accompanied by a significant decrease in the Rh-Rh coordination number. We did not observe
such a decrease and therefore we conclude that in the case of
Rh/T a20 5 spreading of the rhodium particles did not occur. Another explanation was the formation of alloys ( 20). We did not observe contributions from neighboring tantalum atoms at dis
tances in the range of 2.7-2.8 A (in the Rh3Ta bulk alloy, the Rh-Ta distance is 2.729 A (9)). Therefore, alloy formation can be ruled
out. The Rh-Rh coordination distance decreased by 0.008 A which is low for a metal particle which is not covered with adsorbed hydrogen. In ( 5) we reported a decrease of 0.05 A for very small rhodium metal particles. This contraction was due to the absence
of hydrogen on the surface of the metal particles in the SMSI state. The difference between the decrease observed for the Rh /T a20 5 catalyst and for the Rh/Ti02 catalyst in (5) can be explained as fol lows. In vacuum or under inert · gases, the first interatomic layer
distance decreases (21-23). Thus, for very small metal particles a considerable decrease in the average Rh-Rh distance is to be
expected and has indeed been reported (5,12,17). In general , however , the second inter atomic layer distance expands, the third contracts, and so forth. The deeper the layers, the smaller the decrease or increase. As a result, for larger metal · particles the decrease in
the average Rh-Rh distance is less pronounced, as is the case for
the 17 A metal particles in the Rh/T a20 5 catalyst in this study.
In the spectrum, two more contributions were present which both originated from tantalum neighbors at notably short distances .
In the Fourier transforms, the peaks from both Rh-Ta contributions overlapped to a large extent with the major Rh-Rh contribution . As a result, th.e uncertainty in the Rh-Ta coordination number is larger
than the uncertainty in the Rh-Rh coordination number . Because the Rh-Rh contribution is much larger than the Rh-Ta contribu
tions, the uncertainty in the Rh-Rh contribution is not affected. The
page 212
estimated uncertainty in the 2.0 A Rh-Ta coordination number is
about 0 .2 and in the distance about 0 .1 A. The uncertainty in the 1. 7 A Rh-Ta contribution, however, is larger. An irregularity around
k = 9 A-1 was visible in the EXAFS spectra of the Rh/Ta20 5 sam
ples reduced at 858 K (see Figures 9.2a and 9.3a) . This irregularity
gave rise to a peak in the Fourier transform centered at r = 1.6 A. In the Fourier transforms of the sample reduced at 523 K (see Fig
ures 9.1b and 9.1c) , the contribution from this artefact was very
low and was clearly separated from the Rh-Rh peak and caused only
small deviations in the Fourier transform. The 1.7 A Rh-Ta contribution had a. main peak at 2.3-2 .4 A in the Fourier transform and
two sidelobes at 1.8 and 1.4 A (see Figure 9.2f). The main peak is shifted from the real Rh-Ta distance because the Fourier transform
was corrected for Rh-Rh phase shift and backscattering amplitude. The two sidelobes interfered with the artefact at 1.6 A and thus ,
only the main peak of this contribution could be used to determine the accompanying parameters. This caused an additional uncer
tainty in the parameters for the 1.7 A Rh-Ta contribution : about 0.2 in the Rh-Ta coordination number and approximately 0.1 A in
the Rh-Ta coordination distance (additional to the uncertainty in the 2.0A Rh-Ta contribution). In the Rh-Ta distances, another
uncertainty has to be considered . The phase shift function for this absorber-scatterer pair has been composed from three phase shift
functions. Hence, the resulting uncertainty in the final phase shift function was the sum of the uncertainties in the three individual
phase shift functions . We estimated that based on this, the uncertainty in the resulting calculated Rh-Ta distances was about 0.1 A. In Table 9.2, the final uncertainties of all parameters are summarized .
Because the irregularity around k = 9 A-1
created problems, we tried to eliminate this from the spectra by means of the following
procedure. We subtracted the best fitting calculated spectra from
the measured data . The difference spectra contained mostly this artefact. Using a Hanning window between k= 8 and k= 9.5 A._-1, we
isolated this artefact from the difference file and subtracted this
page 213 Chapter 9
from the measured data. The Fourier transform of these spectra
were indeed better than the spectra of the raw data : the artificial
peak around r = 1.6 A was almost completely removed. This is illustrated in Figure 9.4. However, this procedure induced new smaller artefacts around k= 8 and k= 9.5 A.- 1
. It was found impos
sible to completely remove the artefact without introducing new artefacts. Nevertheless, our conclusion is, that the peak in the
Fourier transform around r=1.6A was indeed induced by the artefact around k= 8-9 A.- 1
.
Figure 9.4 The EXAFS function for the Rh/Tap5 sample reduced at 858 K (solid lines) and the spectrum for this sample. in which
the artefact has been 'removed· artificially (dotted lines)
(a) The EXAFS functions
(b) The k3-weighted Fourier transforms
(Dashed line : the calculated Rh-(Rh+ Ta) contribution)
Altogether, the overall uncertainty in the Rh- Ta parameters
was quite large and made detailed interpretation meaningless. Nevertheless, there was no doubt that both contributions were present. Taking the uncertainties in the Rh-Ta distances into account, these results show that tantalum ions are present at
page 214
distances ranging from 1.5 to 2.3 A. These are very short coordina
tion distances and can arise only from Ta ions in almost direct con
tact with rhodium atoms in the metal particles which are in contact
with tantalum (sub)oxide. They may therefore be located directly
underneath the rhodium metal particles. This indicates that indeed
the Ta20 5 support under the metal particles had been reduced.
However, the coordination numbers of the Rh- Ta contributions are
relatively high . When only the rhodium atoms in the metal-support
interface have Ta neighbors, and the support does not expose a
large amount of bare Ta ions, these coordination numbers cannot
exceed 0.3 (in the 73 atom metal particle, about 30% of the metal
atoms is in the metal-support interface and we assumed that each
interfacial rhodium atoms could have up to one tantalum neighbor
at such short distance) .. Therefore. we must conclude that also the
surface rhodium metal atoms, which are not in the metal-support
interface, must be in direct contact with Ta ions. We thus conclude that the metal particles were partly covered with reduced T a20 5.
The direct contact between rhodium atoms and tantalum ions after
reduction at high temperature could result in a strong interaction
between metal particle and support:
In (4,5) and in chapters 7 and 8, we reported Rh0-Ti distances
of 3.4 and 4.3 A. Such longer distances have not been observed for
the Rh/Ta20 5 catalyst in the SMSI state. The reason is the follow
ing. Ti02 has a very simple and very regular structure. The 3.4 and 4 .3 A distances are therefore very well defined. For supports with a
more complicated crystal structure like Al 20 3 and Ta20 5 this is
(unfortunately) not the case. At short distances, the Rh-Ta (or
Rh-Al 3+) distances are well defined, but between about 3 and 5 A, many Rh-Ta distances can occur, each with a very low coordination
number. This makes makes it almost impossible to observe the
longer Rh-Ta distances and explains also why Rh-Al 3+ distances
have never been observed.
page 215 Chapter 9
9.4.3 Rh/Ta20 5 after Admission of 0 2 in the SMSI State
After evacuating the sample at 523 K and admitting oxygen at
100 K, the major Rh-Rh contribution remained unaffected. Obviously, just as for the Rh/Ti0 2 sample in the SMSI state (5), oxida
tion had not taken place. Only a small but detectable change in the Rh-Ta parameters occurred. The distance in the most reliable Rh
T a contribution increased by 0.1 A and the coordination number increased slightly, but the increase was less than the overall uncer
tainty . The coordination number of the 1.7 A Rh-Ta contribution decreased. Regarding the uncertainties in these parameters. the only
conclusion we may draw is that no structural change has taken place and that possibly the reduced suboxide underneath the metal
particle might have started to reoxidize. The covering T a20 5 might be the reason why the oxidation was suppressed, but an electronic
effect might be the reason as well and may even be more likely. because we cannot estimate the extent of coverage (only a complete
coverage could suppress oxidation).
In chapter 7 and in ( 5), we did the same kind of oxygen
adsorption experiments on a Rh/Ti0 2 catalyst in the SMSI state. For the Rh/Ti02 sample, we found that oxygen could adsorb on the metal particles. Because the rhodium metai particles on the Ta20 5 support are much larger than those in the sample in chapter 7 and
in ( 5), and because the particles in Rh/Ta20 5 are partly covered with TaOX, the relative extent of the surface of the metal particles
exposed to the gas atrnosphere is very small, Therefore, for the Rh/T a20 5 sample it is impossible to detect oxygen absorbed on the
metal particles.
page 216
9.5 Final Conclusions
With the Rh/Ti0 2 samples in chapters 7 and 8 we did not
observe any covering, but we could not exclude partial covering.
The fact that the rhodium metal particles in Rh/Ta20 5 are covered
to a possibly larger extent than the metal particles in Rh/Ti0 2 can
be explained by several reasons. First of all, the metal particles in
Rh/T a20 5 are much larger than the metal particles in the Rh/Ti02 samples in chapters 7 and 8 and coverage has up to now only been
reported in literature for larger metal particles. Another reason
might be the fact that the Rh/T a20 5 sample was reduced at much
higher temperatures (858 K) than the Rh/Ti0 2 samples (723 and 773 K). A third reason may be found in the preparation method :
the Rh/Ti02 samples in chapter 7 and 8 were prepared by exchang
ing with a solution of Rh(N03h which had a relatively high pH.
The Rh/T a20 5 sample was prepared using the urea method and
therefore, the starting pH was low. Thus. during the preparation of
the Rh/Ta20 5 sample. some Ta20 5 might have dissolved and later
precipitated on top of the rhodium metal particles. This kind of
coverage has already been reported for Rh/V 20J/Si02 by Kip et al.
(24) and for Rh/V 20 3 by van der Lee et al. (25) and by Bastein
et al. . After reduction at high temperature, this Ta20 5 on top of
the metal particles will become reduced and may have an intimate
contact with the metal particle, giving rise to Rh-Ta bonding.
In summary, apart from the Rh-Rh contribution, no other con
tributions could be detected In the EXAFS spectra of the Rh/Ta20 5 catalyst reduced at 523 K. From the Rh- Rh coordination number it
was concluded that the particles were about 17 A in diameter and
contained about 70 to 80 rhodium atoms . After reduction at 858 K
the sample was in the SMSI state. The Rh-Rh coordination number
remained unchanged. Thus, the basic structure of the metal parti
cles remained intact. Any spread of the metal particles over the · support like a pillbox formation could be excluded . Alloy formation
had not taken place either. In the SMSI state the rhodium atoms in
page 217 Chapter 9
the metal-support interface had tantalum ions as neighbors at very
short distances. ranging from 1.5 to 2.2 A. This indicated that the
Ta 20 5 oxide directly underneath the metal particles was reduced to
a lower oxide of Ta 20 5. In addition. the metal particles were at
least partly covered with reduced T a20 5. Since the only change in
the EXAFS spectra was from neighboring Ta ions. we conclude that
these neighboring Ta ions must be the origin for the SMSI state.
9.6 References.
1. Tauster, S J.; Fung, S. C.; Garten, R.L. J. Am. Chem. Soc. 1978. 100.170.
2. Tauster, S. J.; Fung, S. C. J. Catal. 1978, 55, 29.
3. Tauster, S. J.; Fung, S. C.; Baker, R. T. K.; Horsley, J A Science
(Washington, D.C.)1981, 211, 1121.
4. Koningsberger, D. C.; Martens, J. H. A.; Prins, R.; Short, D. R.;
Sayers, D. E. J. Phys. Chem. 1986. 90, 3047.
5. Martens, J. H. A.; Prins, R; Koningsberger, D. C. J. Phys. Chem.,
accepted for publication . (see chapter 7 of this thesis)
8. Geus, J. W. "Preparation of Catal_ysts 111 ": Poncelet. G.; Grange,
P.; Jacobs, P.A., Eds.; Elsevier. Amsterdam 1983. p. 1
9. Crystal structures :
Rh metal Wyckhojf Crystal Strnctures 1963, 1. 10.
Rhp3 Structure Reports for 1974 1976. 40a, 301.
RhCl 3 Structure Reports for 1964 1972, 29, 275.
Structure Reports for 1964 1972. 29, 130.
Ta metal Wyckhojf Crystal Structures 1963, 1, 16.
TaCl 5 Structure Reports for 1958 1968, 22, 237.
page 218
10. Citrin, P. H.; Eisenberger, P.; Kincaid, B. M. Phys. Rev. Let. 1976, 36, 1346
11. Sinfelt, J. H.; Via, G. H.; Lytle, F. W.; Greegor, R. B. J. Chem. Phys.
1980, 72(9 ), 4832
12. van Zon, J. B. A. D.; Koningsberger, D. C.; van 't Blik, H. F. J.;
Sayers, D. E. J. Chem. Phys. 1985, 12, 5742.
13. Duivenvoorden, F. B. M.; Koningsberger, D. C.; Uh, Y. S.; Gates, B.
C. J. Am. Chem. Soc. 1986 , 108, 6254.
14. van 't Blik, H. F. J.; van Zon, J. B. A. D.; Huizinga, T.; Vis, J. C.; Koningsberger, D. C.; Prins, R. J. Amer. Chem. Soc. 1985, 107, 3139.
15 Cook, J. W.; Sayers, D. E. J. Appl. Phys. 1981, 52, 5024.
16. Kip, B J.; Duivenvoorden, F. B. M.; Koningsberger, D. C.; Prins, R.
J. Catal. 1984, 105, 26.
17 Koningsberger, D. C.; van Zon, J. B. A. D.; van 't Blik, H. F. J.;
Visser, G J; Prins, R.; Mansour, A. N.; Sayers, D. E.; Short, D. R.;
Katzer, J. R. J. Phys. Chem. 1985, 89, 4075.
18. Baker, R. T. K.; Prestridge, E. B.; Garten, R. L. J. Catal. 1979, 56,
390.
19 Baker, R. T. K.; Prestridge, E. B.; Garten, R. L. J. Catal. 1979, 59,
293.
20. Beard, B. C.; Ross, P. N. J. Phys. Chem. 1984, 90, 6811.
21. Somorjai, G. A. "Chemistry in Two Dimensions", Cornell University
press, 1981
22. "The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis", King, A. D.; Woodruff, D. P., Eds, Elseviers, Amsterdam
1981
23. Jona, F.; Marcus, P. M. Proceedings JCSOS II, Amsterdam, 1987, in
press
24. Kip, B. J.; Smeets, P. A. T.; van Grondelle, J.; Prins, R. Appl. Catal.
1987, 33, 181
25 van der Lee, G.; Schuller, B.; Post, H.; Favre, T. L. F.; Ponec, V. J. Catal. 1986, 98, 522
25 Bastein, A. G. T. M.; van der Boogert, W. J.; van der Lee, G.; Luo, H.; Schuller, B.; Ponec, V. Appl. Catal. 1987, 29, 243
page 219 Chapter 10
Chapter 10
Concluding Remarks
In this thesis, we discussed several interesting aspects of supported metal catalyst. Using the results from l\JMR, Raman and
adsorption experiments, we described a model that explained the reduction process of bimetallic Rh-Pt catalysts and the formation of
bimetallic particles. It appeared that adsorbed RhCl 3 complexes were reduced most easily and that the resulting rhodium atoms had
no significant interaction with the support. Therefore, these mobile rhodium atoms could migrate over the support and catalyze the
reduction of other adsorbed complexes. Since adsorbed RhCl 3 as well as adsorbed H2PtCl6 complexes were present, metal clusters
were formed that contained both platinum and rhodium. The composition of these particles was therefore statistically determined. In
chapter 4, however , a completely different situation was encountered : one of the two metals was iron, the other was a more noble
metal (Ru or Pt). It appeared that also in that case the iron was reduced to metallic iron and incorporated in bimetallic particles.
But not all of the iron was reduced, the other part remained in the Fe3+ state. In monometallic iron catalysts, however , iron can be
reduced to Fe2+ and further. Thus , noble metals seem to be able to either catalyze the reduction of a second component, or to suppress
the reduction of that second component . This laher phenomenon is up to now still not solved completely and additional research effort
has to be directed towards this problem. A solution may be found in the interaction between Fe3+ ions and the support. We could assume that Fe3+ in direct contact with the support or even in the surface of the support will not be reduced to Fe2+. Therefore , for
larger, three-dimensional iron oxide or iron hydroxide crystals , Fe3+ will always be present in the particle-support interface, even after
reduction , while the rest of the particle may become reduced. Thus, it may be that in bimetallic Fe-M catalyst systems the noble
Concluding remarks page 220
component plays a role in keeping the Fe3+ phase in a highly
dispersed form in which it cannot be reduced. And it has indeed
been shown by Mossbauer that the Fe3+ phase is a highly dispersed
phase. We could use the results from chapter 3 to put this model in
another light. Fe(N0 3)J is always used As precursor for iron
catalysts . It is common knowledge that these nitrates hydrolyze and
form large clusters in solution. This process can be suppressed in
acid media . Hence, when a chloride such as H2PtCl6 is used as a
precursor for the noble metal component, the acidity of the solution
will increase , hydrolyzation will be suppressed and iron complexes
may adsorb in a highly dispersed way on the support . This highly
dispersed iron is then ' immune' for reduction. This once more
confirms the importance of the first two steps in catalyst prepara
tion, the choice of the metal precursors and the impregnation step,
and their impact on the final state of the catalyst.
Regeneration is an important aspect of a supported metal
catalyst : during processes such as Fischer T ropsch synthesis , the
activity of the catalyst slowly decreases . Therefore, after a period
of time, the catalyst has to be regenerated and oxidation is an
important step in the regeneration of a supported metal catalyst.
Oxidation is also important for storing a catalyst. Once a catalyst
has been reduced in hydrogen and is highly active, it cannot be
stored in air without further precautions. Passivation is a process
in which the small metal particles are oxidized in a slow and con
trolled way in order to prevent sintering. Oxidation of small metal
particles is a process that has not been studied carefully up to now.
In chapter 5 we have described a model for the oxidation of small
rhodium metal particles based on data obtained with EXAFS.
EXAFS indicated that during the oxidation of the metal particles a
new metal-to-oxide interface was created : a rhodium-to-rhodium oxide interface. It was shown that during the oxidation process
rhodium oxide covered the metal particles. After careful oxidation
at room tempera~ure , a small kernel of the particles remained metallic and this metallic kernel was only partly covered with rhodium
oxide. Thus, a part of the metal phase was still exposed to the gas
page 221 Chapter 10
atmosphere. This explains why passivated noble metal catalysts
can be reduced easily : the metallic kernel can adsorb and dissociate
hydrogen especially at low temperature and thus can catalyze the
reduction of the rhodium oxide. Another important aspect has been shown in chapter 5 : EXAFS proved to be capable of detecting
oxide on top of the metal particles, or covering metal particles .
This will be of interest in chapters 7 , 8 and 9, in which the question
is addressed whether metal particles supported on transition metal
oxide supports are covered with support material in the SMSI state.
The results of the ASED-MO computations in chapter 6 show
that this kind of modelling can indeed be of importance in determin
ing structures in metal catalysts. It is, however, merely a star.t that indicates that a lot of useful information can be obtained in this
way . In order to ascertain that the results are indeed reliable, addi
tional research will have to be done. The main question is, how
model-dependent the calculations are. For example, what would
happen if we did not incorporate any protons in the Al20 3 cluster ? We even could place the protons inside the Al20 3 support cluster.
What happens if we do not assume that the metal particle has a
rigid structure and we optimize the coordinates of each rhodium
atom in the metal particle separately ? One of the main problems
that has not been discussed in chapter 6 is of more practical nature : the computer and the (computer) time involved. For each calcula
tion a memory capacity (storage capacity in I BIVI terminology) of 10
megabyte was needed . Since this is not at hand on commonly avail
able mini computers, we had to install the software on the IBM 4381 main frame. Even on such a main frame, one CPU hour was
needed to calculate one point in the energy-distance curves. A com
plete curve would typically take about 12 hours CPU time. On a
time-sharing base, this will take about a day or even more in real time. Hence, apart from the theoretical problems, a lot of practical
problems still have to be solved. But, once these are solved, a com
plete new horizon of applications becomes available. For example,
we could perform these calculations for m~al particles supported on Ti02 and suboxides of Ti02. One could check whether coverage
Concluding remarks page 222
may occur, whether electronic influences are likely, what exactly
they are and how they influence the adsorption capacity of the metal particle, or we could find out if indeed alloy formation can
occur or whether the metal particles will spread on the reduced sup
port.
In chapters 7, 8 and 9, the special properties of metal catalysts supported on transition metal oxides have been addressed. For
these metal catalyst , two states are attainable : a normal state and an SMSI state . In the normal state, they behave like metal particles
supported on inert oxides like Al 20 3, in the SMSI state their properties change. Most pronounced is the decrease in hydrogen and car
bon monoxide adsorption capacity. Several models have already been proposed to explain this phenomenon . The most important
models are ( i) the model of coverage, (ii) the electronic influence from the reduced support oxide on metal particle, (iii) a model of
'pill box ' formation , which assumes that the metal particles spread over the support, and (iv), the model of alloy formation. With two
Rh/Ti0 2 catalysts and a Rh/Ta 20 5 catalyst we found that in the SMSI state the metal particles rested on reduced support oxide.
With the Rh/Ti02 samples, we did not observe any covering oxide. We could, however , not exclude any loose coverage of metal parti
cles in the Rh/Ti0 2 catalyst samples . But, with the Rh/Ta 20 5 we found that the metal particles were at least partly covered with a
suboxide of Ta 20 5. Several explanations are possible. First of all, the metal particles in the Rh/T a20 5 are much larger than the parti
cles in the Rh/Ti0 2 catalysts and coverage has up to now mostly been reported for such larger metal particles. Why small metal par-'
ticles may not be covered epitaxially, however, is not clear . A possible explanation is that larger metal particles expose more defined
crystal faces on which epitaxial growth of an oxide phase is likely. A second explanation may be found in the preparation method : the pH of the impregnating solution was much lower for the Rh/Ta 20 3 sample. Hence, some Ta20 5 may have been dissolved and during the drying procedure, this dissolved Ta 20 5 may have precipitated on top of the rhodium precursor particles. A third explanation can be
page 223 Chapter 10
found in the reduction temperature : the Rh/Ta20 5 sample was
reduced at a much higher temperature than both Rh/Ti0 2 samples.
Thus, we could assume that the metal particles in the Rh/Ti0 2 samples in the SMSI state were indeed partly covered with a TiOx suboxide, but that this coverage very loose. it was not an epitaxially
grown oxide. This covering oxide will rearrange and cover the metal particles more perfectly only after reduction at higher temperatures
( > 773 K), as we have found for the Rh/T a20 5 catalyst. Note, that the covering oxide on top of the metal particles in the Rh/ Al 20 3 catalyst in chapter 5 was rhodium oxide which can grow easily in an epitaxial form on rhodium metal.
The metal particles in Rh/Ti02 and in Rh/Ta 20 5 in the SMSI state did not oxidize upon exposure to oxygen at 100 K. Under
these circumstances, a ' normal ' catalyst, Rh/ Al 20 3 in chapter 5, started to oxidize. For the Rh/Ti0 2 catalyst, we could show that
the metal particles were (partly) covered with adsorbed oxygen . For the Rh/T a20 5 catalyst , we did not observe adsorbed oxygen .
Form the hydrogen chemisorption experiments , however , we may conclude that still some surface area is exposed to the gas atmo
sphere. Therefore, we may conclude that in the Rh/Ta 20 5 sample in the SMSI state and under oxygen , part of the surface of the
metal particles is in contact with oxygen . Thus , if there is no further influence from the support, oxidation should have started.
Since we did not observe this, we must conclude that for Rh/Ti0 2 and also for Rh /T a20 5, an electronic influence from the support is responsible for the suppression of the oxidation process. This electronic influence may also alter any other property of the metal parti
cles. However, research still has to be done in order to establish the nature of this electronic influence and its impact on the properties
of the metal particles. As indicated in the preceding paragraph, the ASED-MO method is very promising and may give valuable infor
mation in this respect.
Concluding remarks page 224
page 225 Summary
Summary
In this thesis we focussed on a few aspects which were impor
tant for the structures in supported metal catalysts : the preparation of supported metal catalysts and the reduction process in
which the inactive metal salts are reduced to active metal particles. We found that, when two metal salts were used to prepare a bime
tallic catalyst, the specific combination of the two metals had a pronounced influence on the final state of the catalyst . When the -salts
of two noble metals were used, the noble metal that was reduced first aided the reduction of the second metal. However, when iron
was one of the two metals, the noble metal kept part of the iron in its highest valence state, Fe3+. The other part was reduced to
metallic iron which formed alloy particles with the noble metal. We also studied the process of passivating a catalyst. That process is
of importance for storing reduced (i.e., active) metal catalysts. In their active state, metal catalysts cannot be stored in air without
risking run-away-oxidation and consequently sintering of the metal particles. We used the ASED-MO theoretical method to ascertain
that the information from EXAFS about the metal-support interface was indeed reliable. The metal-support interface has received much
attention in this thesis, especially for catalyst systems that suffer from (strong) metal-support interactions (SMSI). In these systems,
the metal-support interface plays a key role and the answer for the anomalous properties of the metal particles m the SMSI state can
be found in the metal-support interface.
In chapter 3, the preparation of y-Al 20 3 supported Rh, Pt and
Rh-Pt catalysts has been investigated. From NMR and Raman spectroscopy experiments we concluded that during the pore
volume impregnation method H2PtCl6 could adsorb as separate monometallic complexes on the Al 20 3 support up to a loading of
about 0.18 mmol g- 1. Above this limit, H2PtCl6 crystals were
formed. Adsorption experiments indicated that for RhCl 3 the same
behavior could be expected. The maximum attainable coverage of
Summary page 226
monometallic complexes was even larger for RhCl 3 than for
H2PtCl 6. This maximum attainable coverage depended on the
amount of non protonated surface hydroxyl groups present on the
support during the adsorption process. Therefore, the adsorption
capacity decreased with decreasing pH value of the solution. Dur
ing the reduction process, the monodispersed rhodium complexes
were reduced first and because of their mobility could catalyze the
reduction of the complexes and crystal I ites which are harder to
reduce. Thus. during the reduction process, metal clusters could
'wander' over the support and on their way, catalyze the reduction
of unreduced material. Consequently, as the reduction proceeded,
these metal clusters increased in size. When they encountered a
large crystal of metal salt, this crystal became reduced but, because
of the size of the cluster after the reduction process, the metal clus
ter had no significant mobility. Hence, these larger crystallites
merely 'captured' the mobile metal clusters and smothered the
reduction process. When an alcohol was used as solute, mono
disperse adsorption did not take place. During the drying process,
RhCl 3 and H2PtCl6 crystallized separately. Because there were no
monodisperse rhodium complexes and therefore no mobile rhodium
atoms formed during the reduction process, the RhCl3 and H2PtCl6
crystal! ites were reduced separately.
Chapter 4 describes ESR investigations of the presence of ferric iron (Fe3+) in reduced SiOrsupported Fe-M bimetallic catalysts.
With Mossbauer spectrocopy, the presence of ferric iron in reduced
Fe-Ru/Si02 and Fe-Pt/Si0 2 has been observed. However. from the
Mossbauer spectroscopy point of view, there might be doubts about
this assignment. In ESR, Fe3+ ions cannot be mistaken with other
Fen+ ions. Therefore, the ESR experiments clearly indicated that
the assignment made by Mossbauer spectroscopy was indeed
correct. Moreover, the amounts of ferric iron as determined by ESR
agreed very well with the amounts as found with Mossbauer spec
troscopy. Therefore, these observations make the model in which
iron and the noble metal maintain each other in a highly dispersed
page 227 Summary
state, a state in which iron cannot be reduced to ferrous iron, very
likely. Note the difference with the previous results for Rh-Pt
catalysts : in that case the more easily reduced component aided
the reduction of the other component.
In chapter 5, an alumina-supported rhodium catalyst has been
studied with EXAFS . After reduction and evacuation, oxygen was
admitted at 100 K and at 300 K. EXAFS spectra of the catalyst
after oxygen admission at 100 K indicated the beginning of oxida
tion. At 300 K only a small part of the rhodium particles remained
metallic and this metallic 'kernel' was partly covered with rhodium
oxide. In the rhodium metal to rhodium oxide interface the same 2.7 A Rh0-0 2
- distances are present as in the metal-support interface. Using Rh0-Rh0. Rh0-02- and Rh"+ -02- coordination numbers deter
mined with EXAFS. a model has been derived which describes the
oxidation of small rhodium metal particles supported on A120 3.
The interactions between a ten atom rhodium metal cluster and a 98 atom y-Al20 3 cluster serving as support have been stu
died in chapter 6, using the ASED Molecular Orbital method. The
bonding of the metal particle was strongest when only surface
hydroxyl groups and no bare support oxygen ions were present
underneath the metal particle. The rhodium atoms in the metalsupport interface rested preferable on three-fold coordinated sites.
The accompanying Rh-02- bond length was approximately 2.54-
2.57 A, which is in good agreement with the values reported with
EXAFS (2.6-2.8 A). When only bare oxygen ions · were present
underneath the metal particle. the interfacial rhodium atoms preferred on-top sites and the Rh-0 bond length decreased to about
2.09-2.10 A. Thus, it appeared that the protons in the surface
hydroxyl groups played a key role in the bonding of the metal parti
cle to the support. Protons disfavor on top coordination of the metal atoms and favor binding on three-fold coordination sites.
Summary page 228
In chapter 7 and 8, EXAFS and HRTEM have been used to
study the structure of two Rh/Ti0 2 catalysts . One of these sam
ples was prepared in Eindhoven and measured in Daresbury, the
other sample was prepared at Yale university and was measured in Cornell (CHESS) . The results of both studies are very similar. For
the Eindhoven sample it was found that after reduction in H2 at
473 K (when the catalyst is in the normal state) the metal particles
contain on the average five rhodium atoms and are situated prefer
ably on edges of the Ti02 crystallites, but also on [101] and to a
lesser extent on [001.] anatase crystal faces . Reduction in H2 at
723 K leads to the SMSI state. Besides oxygen neighbors from the
support, the rhodium metal atoms in the metal-support interface
have Tin+ neighbors at 3.4 and 4.3 A. These distances and their
coordination numbers fit well with a model in which the metal particles rest on a Ti02 suboxide. This indicates that the supporting
oxide near the metal particle has been reduced to a suboxide of
Ti02. In the SMSI state no indication for coverage has been found
with either EXAFS or HRTEM. On the contrary, exposing the catalyst in the SMSI state to oxygen at 100 K resulted in changes
in the EXAFS spectrum due to physisorption of oxygen . Conse
quently, in the SMSI state the particles are either not covered or are
incompletely covered with TiOx. Since a Rh/ Al20 3 catalyst under the same conditions became partly oxidized, it is evident that for
the Rh/Ti0 2 catalyst oxidation has been suppressed. This is most
probably the result of an electronic influence from the reduced sup
porting oxide. Even after oxygen admission at room temperature,
the rhodium particles on the TiOx support remain in the metallic
state. The TiO" suboxide in the vicinity of the metal particles starts to re-oxidize and the metal-support interaction becomes weaker.
For the Rh/Ti0 2 catalyst prepared at Yale University , the
reduction process proceeded only very slowly because of the very
compact structure of the sample once it was pressed into a wafer. Even after reduction at 628 K rhodium oxide was present in the
sample. Because of the delayed reduction process, the SMSI state was only invoked at relatively high temperature. Consequently, at
page 229 Summary
lower temperatures, the binding of the metal particles to the sup
port was weak and sintering occurred. Once the SMSI state was established, sintering process stopped. As for the Rh/Ti0 2 sample
in chapter 7, it was found that in the 'normal' state the metal particles rested on unreduced Ti0 2 and in the SMSI state on reduced
Ti0 2. Again, there was no evidence for coverage nor for alloy formation in the SMSI state.
In chapter 9, a Rh/T a20 5 catalyst has been studied with EXAFS. After reduction at 523 K in H2, the rhodium particles were fully reduced and in the 'normal' state. The metal particles con
tained about 73 rhodium metal atoms and were approximately 17 A in diameter. Apart from a contribution from rhodium neighbors, no
other contribution was detected in the EXAFS spectrum. After reduction at 858 K the catalyst was in the SMSI state. In addition
to a contribution from rhodium nearest neighbors, two contributions from neighboring tantalum ions could be detected. The tantalum
ions were located in the reduced supporting oxide directly underneath the rhodium metal particles and in tantalum oxide covering
the rhodium metal particles. Their were no indications for alloy formation, nor for the formation of pillbox or raftlike structures. After
admitting oxygen at 100 K to the catalyst, the metal particles remained unaffected and indications were found that the supporting
tantalum suboxide underneath the metal particles and covering the metal particles had started to reoxidize.
Concluding, for two different supported rhodium catalysts (Rh/Ti02 in chapter 7 and 8 and Rh/Ta20 5 in chapter 9) different
results were obtained : for Rh/Ti0 2 no evidence for coverage has been found, but some coverage could not be excluded while for
Rh/T a20 5 indications for coverage have indeed been found. However, it is impossible to indicate to what extent coverage in the
Rh/T a20 5 catalyst had taken place. Note, that the Rh/Ta 20 5 sample was reduced at a much higher temperature. Thus, a possible
explanation is that, because the Rh/Ti0 2 samples were reduced at
Summary page 230
lower temperatures, coverage was only loose and could not be
detected by EXAFS. Upon reduction at higher temperatures, how
ever, this loose covering oxide may re-organize into epitaxially
grown (sub )oxide.
page 231 Samenvatting
Samenvatting
In dit proef schrif t hebben we de aandacht gevestigd op struc
turen in gedragen metaalkatalysatoren. We hebben een aantal
aspecten de revu laten passereri : de bereiding van katalysatoren,
het proces waarin de niet aktieve metaalzouten op het dragerma
teriaal worden aangebracht en daarna worden gereduceerd tot
aktieve metalen. Wanneer twee metaalzouten aangebracht worden
op het dragermateriaal, dan blijkt dat de combinatie van die twee
metalen bepalend is voor de eindtoestand van de katalysator. In
sommige gevallen helpt het ene {meestal het meest edele metaal) bij
het reductieproces van het andere metaal, in andere gevallen onder
drukt het meest edele metaal de reductie van het andere metaal.
We hebben ook het proces van passiveren bestudeerd. Oat proces
is van belang om katalysatorsystemen te kunnen bewaren. In hun
aktieve toestand is het niet mogelijk gedragen metaalkatalysatoren
in lucht te bewaren zonder het gevaar te lopen dat de metaaldeeltjes
groter worden en daardoor hun aktiviteit verliezen. We hebben
modelberekeningen gebruikt om aan te tonen dat de informatie die
ons de EXAFS-techniek verschafte over het grensvlak tussen het
metaaldeeltje en de drager betrouwbaar is. Oat metaal-drager
grensvlak is een aspect dat in het bijzonder is uitgediept in dit
proefschrift, en met name voor dragermaterialen die een bijzondere
invloed uit kunnen oefenen op de metaaldeeltjes die zich op dat
dragermateriaal bevinden. We hebben kunnen aantonen dat het
dragermateriaal de electronische eigenschappen van die metaal
deeltjes kan beinvloeden en soms de metaaldeeltjes gedeeltelijk kan
bedekken.
In hoofdstuk 3 hebben we de bereiding van monometallische
Rh, Pt en bimetallische Rh-Pt katalysatoren gedragen op y-Al 20 3
besproken. Op grond van NMR en Raman experimenten hebben we
kunnen vaststellen dat tijdens porie-volume impregnatie het zout H2PtCl6 in de vorm van afzonderlijke molekulen aan het dragerma
teriaal kan adsorberen tot aan een grens van ongeveer 0.18 mmol
Samenvatting page 232
per gram A1 20 3. Boven deze grens warden tijdens het draogproces
aan dat we voor RhCl3 hetzelfde gedrag mogen verwachten . De
maximale 'monodisperse' bedekkingsgraad voor RhCl3 is zelfs grater
dan voor H2PtCl6. Voor beide zouten wordt die maximale mono
disperse bedekkingsgraad beinvloed door de zuurgraad van het
oplosmiddel waarmee de zouten op het dragermateriaal warden
aangebracht : bij een lagere pH kan er minder zout aan het drager
materiaal adsorberen. Tijdens het reductiepraces warden de mono
disperse rhodium-complexen als eerste gereduceerd en er ontstaan
rhodium-metaalatomen. Die atomen hebben in tegenstelling tot de
geadsorbeerde complexen nauwelijks een interactie met het drager
materiaal en kunnen zich over het oppervlak van de drager bewegen.
Zodoende kunnen zij het reductiepraces van andere, moeilijker redu
ceerbare platina complexen versnellen . Zo ontstaan er kleine
metaalclustertjes die zich over het oppervlak van de drager bewegen,
niet gereduceerde complexen reduceren en dus steeds grater warden.
Wanneer er grotere zoutkristallen op het dragermateriaal aanwezig
zijn, dan kunnen die de 'randzwervende' metaalatomen en kleine
metaalclusters invangen en zodoende het reductieproces afremmen.
Het is gebleken dat ook het oplosmiddel een belangrijke in vloed op
het bereidingspraces heeft. Gebruiken we een alkohol in plaats van
water (zoals in het hiervoorgaande het geval was) dan vindt er geen
adsorptie plaats en kristalliseren de zouten tijdens het draogpraces
afzonderlijk uit . Er ontstaan RhCl3 and H2PtCl6 kristallen die ook
afzonderlijk warden gereduceerd tijdens het reductieproces.
In hoof dstuk 4 hebben we met behulp van ESR-experimenten
kunnen aantonen dat Fe in bimetallische Fe- Pt en Fe-Ru katalysa
toren voor een graot gedeelte in de vorm van Fe3+ aanwezig is . Er
waren al aanwijzingen vanuit de techniek van Mossbauer spectros
copie dat een groot gedeelte van het ijzer driewaardig was, maar de
interpretatie van de spectra was niet eenduidig. ESR heeft nu
onomstotelijk kunnen aantonen dat dat wel het geval is , omdat we
met ESR Fe3+ ionen niet kunnen verwarren met andere Fe"+ ionen.
In katalysator-systemen waar ijzer aanwezig is zonder een ander
page 233 Samenv att ing
3+ 2+ {edel)metaal, wordt Fe gereduceerd tot Fe en ook verder. Dat
betekent dat in dit geval het edelmetaal een heel aparte rol vervult :
het houdt het ijzer in een hoge valentietoestand , of het verhindert
dat het ijzer gereduceerd wordt. Dit geheel in tegenstelling tot de bevindingen in hoofdstuk 3, waar rhodium de reductie van platina
aanzienlijk kon versnellen. Het model voor de bimetallische Fe-M
katalysatoren is nu als volgt : beide metalen houden elkaar in een
hoog-disperse toestand en in die toestand kan ijzer niet meer gereduceerd worden.
In hoof dstuk 5 hebben we het passiveren van een Rh/ Al 20 3 katalysator besproken. Passiveren is voorzichtig oxideren, zodat we
de katalysator zonder problemen aan de lucht kunnen blootstellen . Nadat de katalysator was gereduceerd en geevacueerd hebben we
zuurstof toegelaten bij 100 K en bij 300 K. EXAFS spectra toonden
aan dat bij 100 K oxidatie begon en dat bij 300 K nog slechts een
kleine kern van de oorspronkelijke deeltjes metallisch is. De rest
van het rhodium is geoxideerd tot rhodium-oxide en bedekt voor een
gedeelte het metallische kerntje. De rhodiumatomen die in contact
zijn met het drager materiaal en de rhodiumatomen die in contact
zijn met het bedekkende rhodiumoxide hebben zuurstofburen op een
afstand van ongeveer 2.7 A. Op grond van Rh0-o2--coordinatie
getallen hebben we een eenvoudig model opgesteld waarmee we het beginnende oxidatieproces kunnen beschrijven.
Met behulp van een computerprogramma hebben we in
hoof dstuk 6 de interacties tussen een rhodium-metaaldeeltje en het
dragermateriaal y-Al 20 3 bestudeerd. Het programma was
gebaseerd op de ASED-MO methode. Wanneer het y -Al 20 3
oppervlak onder het metaaldeeltje volledig uit -OH groepen bestond,
een volledig gehydrateerd oppervlak, was de binding tussen metaal
deeltje en drager energetisch het meest gunstig. De rhodiumatomen in het metaal-drager grensvlak prefereren dan sites waarin ze drie
-0 H groepen als naaste buren hebben . De Rh-0 af stand in het
metaal-drager grensvlak is ongeveer 2.5-2.6 A. Wanneer het
oppervlak onder het metaaldeeltje volledig uit -02- ionen bestaat ,
Samenvatting page 234
een volledig gedehydrateerd oppervlak, is de binding tussen metaaldeeltje en drager iets zwakker. Toch ziet de binding er totaal anders uit : de rhodiumatomen in het metaal-drager grensvlak prefereren nu posities boven op de zuurstof atomen en de Rh-0 afstand is nog maar 2.1 A. Het blijkt dus, dat de protonen in de hydroxylgroepen een belangrijke rol spelen in de binding tussen het metaaldeeltje en de drager. De protonen beinvloeden de on-top binding ongunstig en de binding in drievoudige sites daarentegen gustig. Een belangrijks resultaat in dit hoofdstuk was dat de metaal-zuurstof afstanden in het metaal-drager grensvlak zoals die gevonden zijn met EXAFS heel aardig kloppen met de afstanden die we op deze manier, op grond van theoretische berekeningen gevonden hebben .
In hoofdstukken 7 en 8 hebben we EXAFS en HRTEM gebruikt om de struktuur te bepalen van Rh/Ti0 2 katalysatoren. Deze katalysatoren kunnen in de SMSI toestand worden gebracht, een toestand waarin de eigenschappen van de metaaldeeltjes drastisch zijn veranderd. Een van de monsters was gemaakt in Eindhoven en de EXAFS spectra zijn gemeten in Dares bury, het andere sample was gemaakt door de groep van prof. Haller in Yale University (U.S.A.) en de EXAFS spectra van die groep zijn gemeten in Cornell. De resultaten van beide studies zijn vergelijkbaar . Voor het monster van Eindhoven vonden we dat na een reductie in H2 bij 473 K (de katalysator is dan in de 'normale ' toestand) de metaaldeeltjes gemiddeld ongeveer vijf rhodiumatomen bevatten en zich bevinden op hoeken van de TiOrkristallen en op 11011 en 10011 vlakken van de drager. Na een reductie in H2 bij 723 K (de katalysator is dan in de SMSl-toestand) zijn er nogal wat veranderingen in het EXAFS spectrum. De rhodium atomen in het metaal-drager grensvlak hebben nu niet alleen zuurstofburen van de drager, maar ook nog Ti"+ ionen van de drager op afstanden van 3.4 en 4.3 A. De coordinatie-getallen en -afstanden kloppen heel ardig met een model waarin de metaaldeeltjes rusten op gereduceerd Ti0 2. Hieruit blijkt dat inderdaad tijdens het reductieproces bij hogere temperatuur de drager in de buurt van het metaaldeeltje mee kan reduceren. We hebben geen aanwuzmgen gevonden voor bedekking van het
page 235 Samenvatting
metaaldeeltje met dragermateriaal. In tegendeel, wanneer we
zuurstof toelaten bij de katalysator in de SMSI toestand kunnen we
in het EXAFS spectrum zien dat er zuurstof aan de metaaldeeltjes is
geadsorbeerd. Dus moet op zijn minst een gedeelte van het
oppervlak van de deeltjes onbedekt zijn. Ook werden de metaal
deeltjes die bedekt waren met zuurstof niet geoxideerd, terwijl een
Rh/ A1 20 3 katalysator onder dezelfde omstandigheden wel wordt
geoxideerd. Hieruit blijkt , dat in de SMSl-toestand, als er al
bedekking optreedt, deze bedekking niet verantwoordelijk kan zijn
voor het abnormale gedrag van de metaaldeeltjes . We concluderen
dan ook dat een electronische interactie tussen drager en metaal
deeltje verantwoordelijk moet zijn voor de verandering in de eigen
schappen van de metaaldeeltjes in de SMSl-toestand .
De Rh/Ti0 2 katalysator die bereid was in, en waarvan de spec
tra gemeten waren door, de groep van prof. Haller van Yale Univer-
. sity kon in de EXAFS-cel slechts moeilijk gereduceerd worden.
Zelfs na een reductie bij 628 K was er nog rhodiumoxide aanwezig.
Door die vertraging in het reductieproces kwam de katalysator ook pas bij een hogere temperatuur in de SMSl-toestand. Dat kan er de
oorzaak van zijn dat de metaaldeeltjes zijn gaan sinteren. Voordat sintering optrad, waren de deeltjes erg klein en ongeveer even groot
als de metaaldeeltjes in de Rh/Ti0 2 katalysator van Eindhoven : ongeveer 5 atomen per metaaldeeltje. Na sintering waren de
deeltjes aanzienlijk groter : ze bevatten elk ongeveer 15 rhodiumato
men . Eenmaal in de SMSl-toestand is er een sterke interactie, een
sterke binding tussen drager en metaaldeeltje die ervoor kan zorgen
dat er geen sintering meer optreedt . Zoals voor de Rh/Ti0 2 kataly
sator van Eindhoven, bleek dat ook voor deze katalysator in de 'nor
male ' toestand de metaaldeeltjes op niet gereduceerd Ti0 2 en in de
SMSI toestand op gereduceerd Ti0 2 liggen . Wederom waren er geen
aanwijzingen voor bedekking. Ook nu moeten we concluderen dat
een interactie tussen metaaldeeltje en drager de oorzaak is van de
eigenschappen van de metaaldeeltjes in de SMSl-toestand.
I I .
I
Samenvatting page 236
Tenslotte hebben we in hoofdstuk 9 een vergelijkbaar katalysatorsysteem bestudeerd : Rh/Ta20 5. Na redutie in H2 bij 523 K waren de rhodiumdeeltjes volledig gereduceerd, er was geen rhodiumoxide meer aanwezig. Na reductie bij 858 K was de katalysator in de SMSl-toestand. In de SMSl-toestand hadden de rhodium-atomen in het metaaldrager · grensvlak en de rhodiumatomen in het oppervlak van de metaaldeeltjes Tan+ ion en als buren op heel korte afstanden : 1.5-2.3 A. Dus waren deze metaaldeeltjes in de SMSltoestand wel (gedeeltelijk) bedekt met gereduceerd dragermateriaal. We hebben verder geen aanwijzingen gevonden voor de vorming van een RhTa-legering. Ook hebben we kunnen uitsluiten dat in de SMSl-toestand de metaaldeeltjes zich spreiden over het drageroppervlak.
Concluderend, voor twee verschillende systemen (Rh/Ti02 en Rh /T a20 5) hebben we verschillende resultaten gezien wanneer de systemen in de SMSl-toestand gebracht werden . Voor de Rh/Ti0 2
katalysatoren vonden we geen aanwijzingen voor bedekking, voor de Rh/Ta20 5 katalysator wel. We konden echter niet uitsluiten dat de rhodium-metaaldeeltjes in de Rh/Ti0 2 katalysatoren in de SMSltoestand gedeeltelijk, maar dan wel losjes bedekt waren met TiOx. De Rh/T a20 5 katalysator was echter bij veel hogere temperatuur gereduceerd dan de beide Rh/Ti02 katalysatoren (723 en 773 K). Het is heel goed mogelijk die losse bedekking na reductie bij hogere temperatuur overgaat in epitaxiale bedekking, en die bedekking kunnen we met EXAFS wel aantonen.
page 237 Dankwoord
Dankwoord
Een promotie-onderzoek is iets dat een promovendus zeker niet
in zijn eentje kan volbrengen. Aan het proef sch rift dat U nu in Uw handen heeft hebben dan ook heel veel mensen een steentje
bijgedragen. ledereen die ook maar enigszins heeft bijgedragen aan het voltooien van het mijn onderzoek en het tot stand komen van
dit. proefschrif t dank ik van ha rte .
Om te beginnen natuurlijk Roel Prins. Roel, je was een fantas
tische begeleider. Zeker op wetenschappelijk gebied denk ik dat ik me geen betere eerste promotor had kunnen wensen . Zonder alle
verhelderende discussies, zonder jouw vele kritische kanttekeningen en zonder jouw snelle correcties op de vele proef versies van dit werk
zou dit boekje nooit tot stand zijn gekomen. Op een eerlijke tweede plaats, Diek Koningsberger. Beste Diek, jij bent degene die me de
techniek van EXAFS in alle geuren en kleuren hebt bijgebracht. Jij was degene die de aandacht wist te vestigen op kleine bijdragen in
de spectra, kleine verschillen die vaak grote gevolgen hadden. Jij bent dan ook degene die 'aan de wieg' heeft gestaan van een groot
aantal onderwerpen in dit proefschrif t.
Dan is er natuurlijk de rest van de vakgroep Anorganische
chemie, en in die vakgroep neemt de koffieclub van de groep metaalkatalyse de belangrijkste plaats in. De leden van die. koffieclub ben
in natuurlijk alle heel dankbaar voor de dagelijkse rustpunten. lk denk dat ik het zonder de gezellige koffiepauzes en de enervende
spelletjes bridge minder rustig zou hebben afgebracht . Maar natuurlijk is hun bijdrage veel meer geweest. In de eerste plaats ben ik
veel dank verschuldigd aan het technisch vernuft van de groep metaalkatalyse, Joop van Grondelle, die konstant in de weer was om
alle apparatuur draaiende te houden. Dan natuurlijk mijn dagelijkse collega's, de mede-lotgenoten/promovendi. Op de eerste plaats Bert Kip, die samen met Joop van Grondelle al die jaren met mij een kamer heeft gedeeld . Frans Kampers, Fanny van Zon en Joop van
Dankwoord page 238
Grondelle hebben het allereerste begin van veel van de EXAFS studies meegemaakt. Zonder hun hulp in Daresbury hadden we nooit zoveel metingen kunnen verrichten en was dit boekje aanzienlijk dunner geweest . Ad de Koster, niet in het minst omdat hij me op weg heeft geholpen met het ASED-MO programma. Rutger van Santen wil graag danken voor zijn inspanningen en wetenschappelijke bijdrage aan de ASED-MO berekeningen. Dick van Langeveld voor zijn altijd verhelderende discussies en kritische kanttekeningen . En zeker niet in de laatste plaats Hans Niemantsverdriet voor de prettige samenwerking tijdens de ESR experimenten.
Ook de hulp van collega's van de Rijks Universiteit in Leiden is onmisbaar geweest voor de metingen in Daresbury : met name Hans den Hartog. Marjan Botman en Frans Mijlhof ben ik hiervoor erkentelijk.
In het kader van hun praktikum of afstudeeronderzoek hebben een aantal studenten bijgedragen aan verschillende hoof dstukken van dit proefschrift : John Jansen, Theo van Dijk, Peter Wijnen, Mark Savelsberg, Ton Janssens, Arthur de Jong, Tom Janssens, Toine Ketelaars, Frank van Doormalen, Mark Brouwer, Frits de Koning en Jan van Casteren . Hen wil ik bedanken voor hun inzet, doorzettingsvermogen en enthousiasme.
T enslotte zijn er nog een aantal mensen in de vakgroep die ik zeker niet mag vergeten in dit dankwoord : Wout van Herpen voor zijn technische ondersteuning, Adelheid Elemans-Mehring voor de vele analyses, Frans Sanders voor leveren van al het glaswerk en andere magazijn artikelen en Henk van Lieshout voor alle administratieve beslommeringen die hij steeds snel en vakkundig wist af te handelen .
Dan ben ik, TGTAJM, de medewerkers in het rekencentrum veel dank verschuldigd voor hun niet aflatende inspanningen om alle computers optimaal te laten draaien . In de eerste plaats Gert Jan Visser, die alle software voor de EXAFS data analyse op touw heeft
page 239 Dankwoord
gezet en voortdurend aan onze steeds veranderende eisen aanpast.
Dan, op de tweede plaats, Peter Bregman. de VAX/VMS systeem manager, die de VAXen ondanks de vele EXAFS data-analyses in
optima forma weet te houden en vele problemen wist op te lossen. Dan natuurlijk Frans Galle, die de zware taak heeft de VAX/UNIX
ondanks de vele en zware troff-processen draaiende te houden. Ook Tony van Langeveld voor zijn welwillende hulp bij het opstarten van
de ASED programma's op het IBM mainframe. En tenslotte alle medewerkers van de balie die niet alleen snel alle probelemen met het computerpark oplossen, maar ook dagelijks de vaak enorme stapels output van de printers sorteren en ervoor zorgen dat ze bij de
goede gebruikers terecht komen. Zonder jullie plezierige medewerking zoud veel van het rekenwerk van TGTAJM niet zo succesvol zijn afgerond !
Many of the catalysts discussed in this thesis have been stu
died with EXAFS. The EXAFS measurements have been performed at the Synchrotron Radiation Source in Daresbury (U. K.). I would like to express my gratitude to the staff of .the SRS Laboratory for their skilful! and ever ready assistance. Especially dr. G. Daikun,
station master of the EXAFS station 9.2, for his assistance during the set-up of the station to our needs , dr. N. Greeves, coordinator
of line 9 and Alf Neild, the beamline technician.
T enslotte, mijn vrouw Angeliene, wie ik meest dank verschul
digd ben. Omdat ze ondanks de gezelligheid die ze , vooral gedurende het laatste jaar, 's avonds heeft moeten ontberen, al het geduld dat ze heeft moeten opbrengen, me altijd ter zijde staat. En ook mijn ouders, die me voortdurend hebben gestimuleerd tijdens mijn studie
en promotie, wil ik op deze plaats bedanken.
Th is dissertation has been supported by the Netherlands foun
dation for Chemical Research (SON) with financial aid from the Organization for the Advancement of Pure Research (ZWO).
Curriculum Vitae page 240
Curriculum Vitae
Johannus Hubertus Anna Martens werd geboren in Elsloo (L) op 18 maart 1958. Na het be.halen van het eindexamen Gymnasium f3 aan de Scholengemeenschap Sint Michiel te Geleen in 1976 begon hij de studie Chemische T echnologie aan de T echnische Universiteit in Eindhoven . Na zijn afstudeer onderzoek, dat gericht was op de struktuur van kobalt en kobalt-rhodium katalysatoren gedragen op titaandioxide, onder leiding van prof. dr. R. Prins in de vakgroep Anorganische Chemie, studeerde hij af in augustus 1983. Op 1 september 1983 trad hij in dienst van de Nederlandse Organisatie voor Zuiver Wetenschappelijk Onderzoek en begon een promotieonderzoek dat gericht was op het bestuderen van gedragen metaalkatalysatoren. Het onderzoek werd uitgevoerd onder leiding van prof. dr. R. Prins en prof. dr. ir. D. C. Koningsberger. De resultaten van dit onderzoek zijn in dit proefschrift beschreven . Op 21 juni 1986 trad hij in het huwelijk met Angelina Theresia Carolina van den Hof.
page 241 Lijst van Publicaties
Lijst van Publicaties
1. Characterization of Supported Cobalt and Cobalt-Rhodium Catalysts : II. TPR and TPO of Co/Ti0 2 and Co-Rh/Ti02 Martens. J. H. A.: van 't Blik H. F. J.: Prins, R. }. Catal. 1986, 97, 200
2. Ferric Iron in Reduced Si02 Supported Fe-Ru and Fe-Pt Catalysts : Evidence from Mossbauer Spectrtoscopy and Electron Spin Resonance Martens, J. H. A.: Prins, R. Niemantverdriet, J. W. }. Cata!. 1987. 108, 259
3. Preparation and Characterization of Very Highly Dispersed Iridium on Al20 3 and Si02 Catalysts Kip, B. J.: van Grondelle, J .: Martens. J. H. A.: Prins, R. Appl. Catal. 1986, 26, 353
4. The Stucture of a Rh/Ti02 Catalyst in the Strong Metal Support Interaction State Determined by EXAFS Koningsberger, D. C.; Martens, J. H. A.; Prins, R.; Short, D. R.; Sayers. D. E. }. Phys. Chem. 1986, 90, 3047
5. The Structure of Rh/Ti02 in the Normal and the SMSI State as Determined by EXAFS and HRTEM Martens. J. H. A.; Prins. R.; Zandbergen, H.; Koningsberger, D. C. J. Phys. Chem., 1n press
6. Controlled Oxygen Chemisorption on an Alumina Supported Rhodium Catalyst : The Formation of a new Metal-Metal Oxide Interface Determined with EXAFS Martens, J. H. A.: Prins, R. Koningsberger, D .. C. J. Phys. Chem., to be published
Lijst van Publicaties page 242
7. Preparation of Monometallic Rh and Pt and Bimetallic Rh-Pt Catalysts Supported on y-Al 20 3 Martens. J. H. A.: Prins, R. to be published
8. The Structure of the Metal-Support Interface m Rh/ A1 20 3 Determined with ASED Martens, J. H. A.: van Santen, R. A.: Prins , R. to be published
Stellingen
Behorende bij het proefschrift
A Spectroscopic Characterization of the
Structure of Supported Metal Catalysts
van
J. H. A. Martens
- II -
Stellingen
1. lnterne silanolgroepen in ZSM-5 met een hoog siliciumgehalte
zijn niet afkomstig van verbroken en gehydrateerde silicium
zuurstof-siliciumbindingen in vierringen. zoals verondersteld door Nagy et al. en Boxhoorn et al. Men kan namelijk aantonen dat
vier van deze groepen vacatures omringen van T-atomen (T =Al,
Si), die willekeurig verspreid zijn door het rooster .
Nagy , J. B.; Gabelica , Z.; Derouane, E. G.; Jacobs . P A. Chem. Lett. 1982, 2003
Boxhoorn , G.; Kortbeek, A.G. T. G. ; Hays. G. R.; Alma . N. C. M. Zeolites 1984 4, 15
Kraushaar, B.; van de Ven, L. J . M.; de Haan , J . W.; van Hooff, J . H. C. •Studies in Surface Science and Catalysis". Elseviers Science Publishers, Amsterdam, 1988 (in press)
2. In tegenstelling tot hetgeen Rezek er al. veronderstellen. zijn de
door hen gepubliceerde data over het verband tussen laagdikte
en groeitijd van lnGaPAs kristallen, wel degelijk in overeenstem
ming met normale diffusiebeperkte kristalgroei. als de transport
snelheid maar in de berekeningen wordt meegenomen.
Rezek, A. E. ; Vojak , B. A.; Chin , R.; Holonyak , Jr . J . Electronic Mater.1981 , JO (1), 255
3. De door Williams en Nelson gevonden temperatuuraf
hankelijkheid van de oppervlaktesegregatie in PtRh-legeringen is
waarschijnlijk geen intrinsieke eigenschap van het systeem.
Gezien de door de auteurs gebruikte experimentele condities kan
de invloed van verontreinigingen niet uitgesloten worden . Daar
naast kan bij lage temperaturen door diff usielimitering het
preparaat zich in een niet-evenwichtstoestand bevinden.
Williams , F. L.; Nelson , G. C. Appl. Surf. Sci. 1979, 3, 409
van Delft , F. C. M. J . M. ; van Langeveld , A. D. ; Nieuwenhuys , B. E. Surf. Sci 1987. 1891190. 1129
- Ill -
4. Het verdient aanbeveling om in Raman-spectroscopische studies
naar de rek van hoge-modulus-vezels meer aandacht te besteden
aan de polarisatierichting van de invallende laserbundel ten
opzichte van de vezel.
Bool, R. P.; BretzlafL R. S.: Boyd, R. H. J. Pol. Sci., par1 B :
Polymer Physics 1986, 24, 1039
Robinson. J. M.; Yeung. P. H. J.; Galiotis, C.: Young. R. J.:
Batcheler. D N. J. Mal. Sci. 1986, 21, 3440
5. Bij het bestuderen van selectiviteiten naar zuurstofhoudende
produkten in de inloopperiode in de Fischer-T ropsch-synthese
wordt te weinig rekening gehouden met het feit dat dragerma
terialen de gevorrnde verbindingen kunnen adsorberen. Met
name y-Al 20 3 kan alcoholen sterk adsorberen. De bevindingen
van Kip et al .. waarin veranderingen in de oxo-selectiviteiten
gerelateerd worden aan chloorgehaltes zijn waarschijnlijk geen
intrinsieke eigenschappen van het katalysatorsysteem. maar kun
nen verklaard worden aan de hand van het adsorberend vermo
gen van het dragermateriaal.
Kip, B. J.; Dirne, F. W. A.; van Grondelle. J.; Prins, R. Am.
Chem. Soc., Div. Pelr. Chem., Mechanisms of Fischer Trospch
Chemistry, 1986, 31, 43
Kip, B. J.; Smeets, P. A. T.; van Grondelle, J: Prins. R. Appl. Catal. 1987, 33, 181
6. De gevoeligheid en de kwaliteit van verschillende EXAFS
opstellingen kan beter vergeleken worden aan de hand van nog
meetbare gewichtsfracties dan aan de hand van nog meetbare
concentraties van het adsorberend element.
7. Het feit dat EXAFS-metingen aan kobalt-molybdeensulfide
hydrotreatingkatalysatoren tot nu toe geen directe aanwijzingen
hebben gegeven voor een interactie tussen kobalt en molybdeen
kan teruggevoerd worden op hetzij een te lage signaal/ruis
verhoud ing. hetzij een ontoereikende data-anal yseprocedure.
