Transition Metal Oxides

Studies in Surface Science and Catalysis 45

TRANSITION METAL OXIDES: Surface Chemistry and Catalysis

This Page Intentionally Left Blank

Studies in Surface Science and Catalysis Advisory Editors: B. Delmon and J.T. Yates

Vol. 45

TRANSITION METAL OXIDES: Surface Chemistry and Catalysis

Harold H. Kung Chemical Engineering Department, The Technological Institute, Northwestern University, 2 145 Sheridan Road, Evanston, IL 60208, U.S.A .

ELSEVI ER

Amsterdam - Oxford - New York - Tokyo

1989

ELSEVIER SCIENCE PUBLISHERS B.V. Sara Burgerhartstraat25 P.O. Box 2 1 1, 1000 AE Amsterdam, The Netherlands

Distributors for the United States and Canada: ELSEVIER SCIENCE PUBLISHING COMPANY INC. 655, Avenue of the Americas New York, NY 10010, U.S.A.

First edition 1989 Second impression 1991

ISBN 0-444-87394-5 0 Elsevier Science Publishers B.V., 1989 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior written permission of the publisher, Elsevier Science Publishers B.V./ Academic Publishing Division, P.O. Box 330, 1000 AH Amsterdam, The Netherlands. Special regulationsfor readers in the USA - This publication has been registered with the Copyright Clearance Center Inc. (CCC), Salem, Massachusetts. Information can be obtained from the CCC about conditions under which photocopies of parts of this publication may be made in the USA. All other copyright questions, including photocopying outside of the USA, should be referred to the publisher.

No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Although all advertising material is expected to conform to ethical (medical)standards, inclusion in this publication does not constitute a guarantee or endorsement of the quality or value of such product or of the claims made of it by its manufacturer. This book is printed on acid-free paper. Printed in The Netherlands

V

To my mentor

Professor Robert L. Burwell, Jr and my wife Mayfair

This Page Intentionally Left Blank

vii

PREFACE

In the past twenty years there have been rapid advances in the understanding of surface phenomena as a result of the availability of many new and powerful techniques to study solids and their surfaces. Clever experimentation coupled with techniques such as laser Raman spectroscopy, solid state NMR, extended X-ray absorption fine structure spectroscopy, as well as a host of surface sensitive electron spectroscopies have provided information on the atomic structure and composition of surfaces and near surface regions. To date, the bonding geometries, the positions of adatoms. the orientations of adsorbed molecules, the structures of overlayers, the extent of relaxation of surface atoms, and even the atomic positions of surface defects have become accessible. Indeed, many of these data have been obtained for metallic surfaces, especially transition metal surfaces where the attention of most surface scientists has been focused. These data have resulted in significant advances in the understanding of the surface chemistry and catalytic properties of these materials. It is anticipated that a similar rapid increase in our level of understanding of surface phenomena will be forthcoming for transition metal oxides. Indeed, a growing number of investigators have turned their attention to oxidic materials. Simultaneously, new catalytic properties of transition metal oxides are being discovered. New reactions are being reported, such as metathesis, selective oxidation of butane to maleic anhydride and various photo-enhanced processes, and new catalytic materials are being synthesized, especially various mixed oxides. It appears that the level of understanding in the area of surface chemistry and catalysis of transition metal oxidcs is poised for a quantum leap. In this book I have attempted to summarize existing information on the structure, electronic properties, chcrnistry, and catalytic properties of transition metal oxide surfaces so that it can serve as a uscrul source of information for investigators in this field and as a comprchensivc overview of the subject for graduate students. The book is inlended for surface physicists, chcmists, and catalytic engineers. By presenting physical, chemical and catalytic properties in one volume, it is hopcd that the interrclationship among them will become more apparent.

...

VIU

The subjects in this book can be divided into three sections. The first section (chapters 1 to 3) deals with the structural, physical, magnetic, and electronic properties of transition metal oxides. Although the emphasis is on surface properties, relevant bulk properties are also discussed. The second section (chapters 4 to 7) covers surface chemical properties. It includes topics that describe the importance of surface coordinative unsaturation in adsorption, the formation of surface acidity and the role of acidity in determining surface chemical properties, the nature and reactivities of adsorbed oxygen, and the surface chemistry in the reduction of oxides. The third section is on the catalytic properties (chapters 8 to 14). Various catalytic reactions including decomposition, hydrogenation, isomerintion, metathesis. selective oxidation, and reactions involving carbon oxides are discussed. Emphasis is placed more heavily on reaction mechanisms and the role of catalysts than on kinetics and processes. A chapter on the preparation of oxide catalysts and one on photo-assisted processes are also included. Whenever appropriate, relationships among various topics are indicated. It would have been impossible to complete this book without the encouragement and help of a number of people. The most significant of them is Professor Robert L. Burwell, Jr. who provided numerous suggestions and comments on its content. The constant encouragement and helpful discussions with Dr. Mayfair C. Kung are also greatly appreciated. The enthusiasm and dedication of my students have made the study of this subject enjoyable. Much of our research in this area has been supported by the U. S. Department of Energy without which I would not have the necessary background to undertake the task. Finally, permission by various publishers to reproduce the figures and tables in the text is also gratefully acknowledged.

Harold H. Kung 1988

ix

CONTENTS

PREFACE

vii

Chapter 1. INTRODUCTION References 5

1

Chapter 2. BULK AND SURFACE STRUCTURE OF TRANSITION METAL OXIDE 2.1 Bulk Structure 6 2.2 Shear Structure in Intermediate Oxides of Ti, V, and Mo 12 2.3 Structure, Stability and Reconstruction of Oxide Surfaces 15 2.4 Structure of Supported Oxides 20 References 25

6

Chapter 3. PHYSICAL AND ELECTRONIC PROPERTIES 3.1 Surface Composition 27 3.2 Ionicity of Oxides 28 3.3 Magnetic Properties of Small Oxide Particles 36 3.4 Quadrupole Splittings of Surface Ions 43 3.5 Surface Electronic Structure 46 3.6 Surface Vibration 48 References 50

27

Chapter 4. SURFACE COORDINATIVE UNSATURATION 4.1 Formation of Surface Coordinative Unsaturation 53 4.2 Chemical Properties of Surface Coordinatively Unsaturated Sites 57 4.3 Adsorption of Small Molecules 61 References 69

53

X

Chapter 5. SURFACE ACIDITY 5.1 Surface Acid Sites 72 5.2 Formation of Surface Acid Sites 74 5.3 Determination of Acidity 80 5.4 Role of Acid Sites in Catalytic Reactions 88 References 89

72

Chapter 6. REDUCTION OF OXIDES 6.1 Introduction 91 6.2 Thermodynamics of Reduction 92 6.3 Kinetic Models 93 6.4 Mechanism of Reduction 96 6.5 Effect of Support 98 6.6 Effect of Other Components 100 6.7 Reactivity of Reduced Surfaces 100 6.8 Influence of Reduced Oxides on the Properties of Transition Metals 102 References 107

91

Chapter 7. OXYGEN ON OXIDES 7.1 Nature of Adsorbed Oxygen 110 7.2 Detection of Adsorbed Oxygen 111 7.3 Reactivity of Adsorbed Oxygen 116 References 119

110

Chapter 8. PREPARATION OF OXIDES 8.1 General Considerations 121 8.2 Preparation of Unsupported Single Component Oxides 123 8.3 Preparation of Supported Oxides 129 8.4 Preparation of Multicomponent Oxides 131 References 134

121

Chapter 9. METATHESIS AND ISOMERIZATION 9.1 Metathesis 136 9.2 Isomenzation of Alkenes 140 References 144

136

Chapter 10. DECOMPOSITION, HYDROGENATION AND 146 RELATED REACTIONS 10.1 Decomposition of Alcohols 146 10.2 Decomposition of Nitrous Oxide 153 10.3 Hydrogenation, H-D Exchange of Hydrocarbons, and H2-D2 Scrambling 155 10.4 Reduction of NO 162 References 166

xi

Chapter 11. SELECTIVE OXIDATION REACTIONS I 11.1 Introduction 169 11.2 Types of Selective Oxidation Reactions 170 11.3 Features of Catalytic Selective Oxidation 170 11.4 Chemical Factors Affecting Selectivity 175 11.5 Oxidation of Propene to Acrolein and Ammoxidation to Acrylonitrile 181 11.6 Effect of Water on Propene Oxidation 194 References 195

169

Chapter 12. SELECTIVE OXIDATION REACTIONS II 12.1 Selective Oxidation of Butenes 200 12.2 Selective Oxidation of Butane 210 12.3 Selective Oxidation of Methane 212 12.4 Selective Oxidation of Methanol 217 References 223

200

Chapter 13. CATALTYIC REACTION BETWEEN HYDROGEN AND CARBON OXIDES 13.1 Introduction 227 13.2 Alcohol Synthesis on Copper-Zinc Oxide and Other Cu-based Catalysts 228 13.3 Group VIII Metalmetal Oxide Catalysts 236 13.4 Isosynthesis Reaction 240 13.5 Water-Gas Shift Reaction 244 References 248

227

Chapter 14. PHOTO-ASSISTED SURFACE PROCESSES 14.1 Introduction 252 14.2 Photo-Assisted Adsorption and Desorption 256 14.3 Photocatalysis: Gas Phase Reactions 258 14.4 Photocatalysis: Liquid Phase Reactions 266 14.5 Photocatalysis by Metalmetal Oxide and Oxide/Oxide Composites 271 References 273

252

INDEX

277

This Page Intentionally Left Blank

Chapter 1 INTRODUCTION

Transition metal oxides are technologically important materials that have found many applications. For example, in the chemical industry, these oxides are the functional components in the catalysts used in a large number of processes to convert hydrocarbons to other chemicals. They are also used as electrode materials in electrochemical processes. In the electronics industry, they are used to make conductors in films. The recently discovered high temperature superconductors are multicomponent transitional metal oxides. Among these applications, perhaps the use of transition metal oxides as catalysts is the most technologically advanced and economically important. It is also an area in which much progress has been made in recent years in terms of the understanding of the fundamental processes that occur, primarily because advances in instrumentation and experimental techniques have made it possible to study the chemistry of the interface between the transition metal oxide and the fluid phase in greater detail than ever before. In particular, developments in surface science techniques have provided very detailed pictures about the surface structures, chemical compositions, and electronic properties of the surfaces. Some of the chemical processes that make use of transition metal oxides are listed in Table 1-1. As can be seen from the table, many of the processes require high selectivity for a particular product, and many involve oxidation of the reactant molecules. In fact, selective oxidation, ammoxidation, and selective dehydrogenation probably constitute the most important catalytic uses of transition metal oxides. The different oxidation states available in these oxides make it possible to control the selectivity in oxidation with the properties of the oxides. Some transition metal oxides can also catalyze selective hydrogenation and are used in some commercial processes. As the demand for specialty chemicals (that is, specific chemicals for spccific processes) increases in the future, demand for high selectivity will increase in a wide variety of reactions including amination, alkylation, aldol condensation, and carbonylation, in addition to those in Table 1 - 1 . It is quite possible that transition metal oxides will occupy an increasingly 1

2

Table 1-1 Examples of Chemical Processes in which Transition Metal Oxides are Catalysts Process

Example

Oxidation

Production of SO3 from SO2 CO oxidation in emission control Production of styrene from ethylbenzene

Dehydrogenation (nonoxidative) Dehydrogenation (oxidative) Selective oxidation

Selective ammoxidation Selective reduction Metathesis Water-gas shift

Production of formaldehyde from methanol and butadiene from butenes Production of acrolein from propene, and maleic anhydride from benzene or butane Production of acrylonitrile from propene Reduction of NO, selective hydrogenation of unsaturated ketones Production of long chain alkenes Production of hydrogen

important position as catalysts in these chemical processes. Not listed in the table is the production of methanol by the hydrogenation of CO or C 0 2 . The earlier generation of catalysts for this process are based on zinc-chromium oxide. However, there is controversy over the current copper-zinc oxide catalyst as to whether the active component is the oxide or the copper metal. In addition to being used as catalysts, transition metal oxides are also precursors for other important catalysts. The cobalt-molybdenum sulfide catalyst for hydrodesulfurization is an example. This catalyst is prepared by sulfiding cobalt-molybdenum oxide (often supported on alumina). Another example is the chromium-based catalyst for ethylene polymerization. The catalyst can be made from supported chromium oxide as a precursor. Finally, many noble mctal catalysts are prepared by reduction of the corresponding oxides. In addition, these metal catalysts are often stored in air and are converted to their oxides during storage. It is quite conceivable that in these cases, the detailed structures, morphologies, or other properties of the transition metal oxide precursors could affect the properties of the final catalysts. Understanding catalysis requires an understanding of surface chemistry, which deals with the bonding and reaction of an adsorbate with the surface and the influence of the surface on the bonding and reaction between adsorbates. It is apparent that an important part of any effort toward obtaining such an undcrstanding is the ability to characterize thc physical and chemical properties of a surface. In recent years, much progress has been made in the understanding of

INTRODUCTION

3

Table 1-2 Properties that are Important in the Surface Chemistry of Transition Metal Oxides

Presence of cations and anions in stoichiometric ratios and in well-defined spatial (structural) relationships Possibility of covalent and ionic bonding between cations and anions Presence of a strong electric field normal to the surface due to the coulombic nature of the ionic lattice Presence of charged adsorbed species Presence of surface acidity and basicity Presence of cationic and anionic vacancies Ability of cations to undergo oxidation and reduction High mobility of lattice oxygen and the possibility that the lattice oxygen are reactants in a reaction Interaction of the solid with incident photons that leads to photo-assisted surface chemical processes

metallic surfaces [l-41. Progress has also been made, but at a slower pace, for the transition metal oxides because of the higher level of complexity in the experimental techniques involved. There are significant differences between the chemistry of transition metal oxides and the corresponding metals. Table 1-2 provides a list of properties that are important in the surface chemistry of transition metal oxides. Many of them either do not apply or apply only to a limited extent to the metals. Common to many of the properties listed in Table 1-2 is the fact that transition metal oxides are made up of metallic cations and oxygen anions. The ionicity of the lattice, which is often less than that predicted by the formal oxidation states, results in the presence of charged adsorbate species, and the common heterolytic dissociative adsorption of molecules (that is, a molecule AB is adsorbed as A+ and B3. Surface exposed cations and anions form acidic and basic sites as well as acid-base pair sites. The fact that the cations often have a number of commonly obtainable oxidation states has resulted in the ability of the oxides to undergo oxidation and reduction and the possibility of the presence of rather high densities of cationic and anionic vacancies. As can be seen throughout the discussions in this book, these properties have determining effects on the interaction of molecules with the oxide surfaces. This book starts with discussions of the structural (Chapter 2) and physical and electronic properties of transition metal oxides (Chapter 3). Knowledge of surface structure is as important as knowledge of molecular structure in understanding surface chemistry. Physical properties of these oxides, especially those for small crystallites, are often used for identification and characterization purposes. Sometimes they are used as means to monitor chemical interactions.

4

Surface electronic properties determine the mode of bonding of the adsorbates, and both surface and bulk electronic properties determine the photo-assisted surface chemical processes. Discussions of the surface chemical properties then follow. The importance of surface coordinative unsaturation in governing the adsorptive properties of many oxides will be discussed in Chapter 4. The discussion of surface acidity will be found in Chapter 5 , and the nature and reactivities of adsorbed oxygen will be discussed in Chapter 7. Reduction of transition metal oxides, which is initiated at oxide surfaces, will be discussed in Chapter 6. After surface chemistry, the catalytic properties of transition metal oxides will be discussed. The discussions will begin with the methods of preparation of oxides and the dependence of the final properties of the oxides on the preparation methods (Chapter 8). Then various catalytic reactions will be discussed. These are metathesis and isomerization (Chapter 9), decomposition and hydrogenation (Chapter lo), selective oxidation (Chapters 11 and 12), reactions of carbon oxides (Chapter 13), and finally photo-assisted surface processes (Chapter 14). In the last chapter, photo-assisted surface chemical reactions will be described together with photo-assisted catalytic reactions. The emphases of the discussions will be on transition metal oxides. However, whenever appropriate, nontransition metal oxides will also be discussed, especially when they are used to illustrate certain concepts or for comparisons. In particular, properties of ZnO will be discussed rather extensively because ZnO is among the best understood oxides whose surface chemistry, catalytic, electronic and structural properties have been studied extensively, and because its behavior is in many ways similar to many transition metal oxides. Throughout this book, reference to information obtained from various experimental techniques will be made. The readers are referred to some excellent treatises that describe these techniques 11.4-91. The following is a list of standard acronyms that are often used and their meanings:

AES ELS EPR EXAFS FTIR HREELS IR ISS LEED NMR

SEM STEM

TEM TPD TPR UPS

Auger electron spectroscopy Electron energy loss spectroscopy Electron paramagnetic resonance spectroscopy Extended X-ray absorption fine structure Fourier transform infrared spectroscopy High resolution electron energy loss spectroscopy Infrared spectroscopy Ion scattering spectroscopy Low energy electron diffraction Nuclear magnetic resonance spectroscopy Scanning electron microscopy Scanning transmission electron microscopy Transmission electron microscopy Temperature programmed desorplion or decomposition Temperature programmed reduction or reaction Ultraviolet photoelectron spectroscopy

INTRODUCTION

5

UV-vis Ultraviolet-visible absorption spectroscopy XANES X-ray absorption near edge spectroscopy XPS X-ray photoelectron spectroscopy Other techniques will also be mentioned. They include: Cyclic voltammetry Electric conductivity measurement Magnetization measurement Mossbauer spectroscopy Raman or laser Raman spectroscopy It will be greatly beneficial for the readers to have some general knowledge about all of these techniques, especially regarding the types of information that are obtainable, so as to better understand the discussions of the data presented.

REFERENCES 1. G. A. Somorjai, "Chemistry in Two Dimensions: Surfaces," Cornell University Press, Ithaca, NY, 1981. 2. 'The Nature of Surface Chemical Bonds." T. N. Rhodin, and G. Ertl, ed., NorthHolland Publ. Co., NY, 1979. 3. S. R. Morrison, 'The Chemical Physics of Surfaces." Plenum Press. NY. 1977. 4. G. Ertl, and I. Kuppers, "Low Energy Electrons and Surface Chemistry," Weinheim, Germany, 1985. 5 . "Experimental Methods in Catalytic Research," Vol. 1-3, ed. R. B. Anderson, et. al., Academic Press, NY, 1968-76. 6. W. N. Delgass, G. Haller, R. Kellerman. and I. Lunsford, "Spectroscopy in Heterogeneous Catalysis," Academic Press, N.Y., 1984. 7. D. P. Woodruff, and T. A. Delchar. "Modem Techniques of Surface Science," Cambridge University Press, Cambridge, 1986. 8. "Electron Spectroscopy for Surface Analysis," Topics in Current Physics, vol. 4, ed. H. Ebach, Springer, Berlin, 1977. 9. 'The Chemical Physics of Solid Surfaces and Heterogencous Catalysis, D. A. King and D. P. Woodruff ed., Elsevier Scientific Publ., Amsterdam, 1?8l, 1983.

Chapter 2

BULK AND SURFACE STRUCTURE OF TRANSITION METAL OXIDE

2.1 BULK STRUCTURE With the exception of some complex oxides of unusual stoichiometries, multicomponent compounds, and oxides that are only stable at high tcmpcrature and high pressure, the bulk structures of most transition metal oxides are known. An examination of the known structures shows that transition metal oxides exist in many different crystallographic forms. There does not appear to be a simple generalization that relates the structure to the stoichiometry or the position in the periodic table. In fact, it is not uncommon to find a certain oxide in more than one crystal structure at ordinary temperatures as a result of the high activation energy in the process of transforming from a less thermodynamically stable to a more stable structure. There is perhaps one generalization, which is the fact that the ionic radii of transition metals are smaller than that of 02-.Thus the oxygen ions are usually close-packed with the smaller metal ions situated in the octahedral and tetrahedral holes among the oxygen ions. There are many excellent treatise on the subject of crystal structures, such as the one by Wells [l]. In this section, structures of some of the oxides that are commonly used in studies of surface chemistry and catalysis are described. The structures of the other oxides are necessarily left out. Oxides commonly studied as catalytic materials belong to the structural classes of corundum, rocksalt, wurtzite, spinel, perovskite, rutile, and layer structure. Table 2-1 lists the stable structures of some binary oxides. These arc the structures often reported for the oxides prepared by common methods under mild conditions. In some cases, other structures exist. Furthermore, the structures indicated represent the general type. The positions of the ions may not be at the ideal positions of the highest symmetry. For example, distortions are found for FeO, NiO, MnO, and COO from the cubic lattice, and V 0 2 , Nb02, Mo02, W02 from the perfect rutile structure. The rocksalt structure is made up of a three-dimensional array of alternating 6

U

Table 2-1 Crystal Structural Classes of Some Common Transition Metal Oxides

R

v)

Sc2Qcs

Ti0 r* TizQ cr

VO

VzO, cr

C r 2 Q cr Cr02 t

MnO r Mn304sp*

Ti02

VOz

C I Q or

MnzQcs* Fe@4 sp MnOz t*,and

t,a,b

r

t* vZo5 a

FeO r F e 2 Q n,sp

COO r Co304 sp

NiO

r

CUO

cuzo

s c

zno

w

C

w

M

others

Y 2 4 cs

LazQmt

ZrO r Z r O 2 m,tet

NbOz t*

Mooz m,(t*) TcOz m.(t) MOO, l,(or) T C Za~

RuOz t

NbZO5 mt

H Q m

Ta02 t

WOz m,(t*) Re02 m,(t)

OsO2 t

Ta205 or

WO,

m

Re@

R h z Q cr*

PdO

s

AgzO c

CdO

r

HgO

or,and others

Rho2 t

cub OsO, (perovskite)

cl 5cr C c1 cl

m

m

t

Pt304 cub PtOz

t

Re207 a

r =rocksalt s =PtSscructure, w=wurtzite, t =rutile, a = anatase, c = interpenetrating cristobalite, f = fluoride, cr = corundum, sp = spinel, or = orthorhombic, tet = tetragonal, m = monoclinic. cub = cubic, 1 = layer, mt = multiple modifications, * = distorted or defective

b = brookite, cs = C structure,

8

cations and anions (Fig. 2-la). Each ion is in the center of an octahedron whose vertices are ions of the opposite type. The structure can be viewed as being made up of comer-sharing octahedra (Fig. 2-lb). A wurtzite structure is made up of a three-dimensional net of corner-sharing tetrahedra (Fig. 2-lc,d). Each ion is in the center of a tetrahedron in which the opposite ions are at the vertices. The corundum, the rutile, and the spinel structures are made up of layers of close-packed oxygen ions. If the oxygen ions are modeled by hard spheres, (neglecting the cations for the moment,) each ion in a close-packed layer is in contact with six others (see Fig. 2-2). When one such layer is stacked on top of another such that an ion in this layer is in contact with the maximum possible number of ions in the other layer, this ion will be sitting above a triangular hole of the other layer (point B or C), and it will be in contact with three ions in the other layer. Now consider a case of two close-packed layers stacked in this manner as in Fig. 2-2. When a third layer is put on top of these two, its ions can take the positions vertically on top of the ions in the first layer such as above point B, or the positions above point C. In the former case, the spatial positions of the layers follow the sequence ababab... The resulting structure is called hexagonal closepacking (h.c.p.), and it forms the basis for the corundum and the rutile structure. In the latter case, the spatial sequence of the layers is abcabcabc... The resulling s.tructure is called cubic close-packing (c.c.P.),and it forms the basis for the spinel structure. Between adjacent layers of oxygen ions in both h.c.p. and c.c.P., the interstices (holes) are bound by either four or six oxygen ions (Fig. 2-2). They are commonly referred to as tetrahedral and octahedral holes, respectively. There are as many octahedral holes as the number of oxygen ions, and half as many tetrahedral holes as octahedral holes. In the ideal rutile structure, half of the octahedral holes are filled with cations, while the tetrahedral holes are empty. Thus the compound has a formula M02 (e.g. TiOz). In the corundum structure, two-thuds of the octahedral holes are filled. The tetrahedral holes are empty, and the compound has a formula M203 (e.g. a-FezO3). An ideal spinel structure would have one-half of the tetrahedral holes and one-half of the octahedral holes filled, and the formula of the compound is M304 (e.g. Fe304). It is easily seen that for charge neutrality, the cations must be of two different oxidation states. The most common oxidation states are +2 and +3, and the formula can be rewritten as MIMJII04. Both M r I and MI11 can occupy the tetrahedral or octahedral holes. The equilibrium distribution depends on the nature of the cation and the temperature. For entropic reasons, increasing the temperature tends to randomize the distribution. A normal spinel is one in which all MI ions are in the tetrahedral holes, and all MI1 ions in the octahedral holes. An example is ZnFez04. An inverse spinel has all MI ions in the octahedral holes. The MI1 ions are distributed equally between the octahedral and the tetrahedral holes. Examples are Fe304 and MgFe200. Mixed spinels have intermediate distributions, The fact that some of the octahedral or tetrahedral holes arc unoccupied makes it possible that other ions may occupy these holes when thcse are exposed on the surface.

OXIDE STRUCTURE

9

Figure 2-1 a. A rocksalt structure; b. Two comer-sharing octahedra, one centered around ion A. Octahedra centered around ions A and B in FIg. 2-la are edge-sharing; c. A wurtzite structure; d. Two comer-sharing tetrahedra, one centered around ion C.

Figure 2-2 Close-packed layers of oxide ions. If ions in the third layer are above B, it is a hexagonal close-packed structure. If these ions are above C, it is a cubic closepacked stucture. Ions that define octahedral and tetrahedral holes are also shown.

y-Fe2O3 is also a spinel. As suggested by the chemical formula, the compound has fewer cations than needed to complete an ideal spinel structure. Indeed, the formula is sometimes written as Fe3(Fe5@)OI2to represent the fact that for every twelve oxygen ions (which provide twelve octahedral holes and six tetrahedral holes), there are three FeIII ions in the tetrahedral holes, and five FeIII ions in the octahedral holes. Compared to the idcal spinel structurc, the occupancy of the tetrahedral holes is the same, but the occupancy of the octahedral holes is 1/6 less. This sixth position is denoted by @ to represent a cation vacancy. Thus there is one cation vacancy for every three spinel units. When this vacancy is ordered, the repeating unit is a trispinel. There is another way to view the spinel and the corundum structures. Instead of constructing the solid with close-packed layers of oxygen ions, the same structure can be formed using the octahedral units as building blocks. At the comers of the octahedral units are the oxygen ions, and at the centers are the cations. Of course, these cation positions do not have to be completely filled. The corundum structure is then made up of a three dimensional net of such octahedra in which some octahedra share comers, edges, or faces, whereas the spinel structure is made up of octahedra that share comers and edges. Although not apparent in these cases, it is sometimes more convenient to discuss structures with networks of octahedra. The rutile structure is one such case. In this structure, the sheets of closepacked oxygen ions are rather distorted. The cations are in the center of octahedra of oxygen ions, as shown in Fig. 2-3a. Along the c-direction (the vertical direction), the octahedra are linked by sharing edges (Fig. 2-3b) to form a chain. Adjacent chains are connected by sharing vertices (Fig. 2-3c). For Ti02, the octahedra are distorted such that four metal-oxygen distances are of one value, and the other two are of a different value. For some others like the dioxides of V, Nb, Mo, and W, the metal-metal distances along the octahedra chain are not regular, but alternate between a longer and a shorter distance. Molybdenum trioxide M a 3 has a unique layer structure made up of chains of octahedra that share comers. Two such chains are connected by sharing two edges of the octahedra to form a double chain (Fig. 2-4a). These double chains are then connected together in the third dimension (perpendicular to the plane of the double chain) by sharing comers to form a sheet-like structure. Thus for each octahedron, three 0 atoms are shared by three octahedra of the same double chain, two are shared by two octahedra of adjacent double chains, and one is unshared. This unshared unit is commonly referred to as a Mo=O unit. Finally, these twodimensional sheets are stacked on top of each other with rather weak interaction between layers (Fig. 2-4b). Vanadium pentoxide V 2 0 5 also has a layer structure. The basic units are chains of tetrahedra linked through two comers. Two chains are then linked by placing a fifth oxygen ion from one chain to each V ion in the other chain to form a double chain and the basic structure of metavanadates. In this manner, each vanadium ion is regarded to have five oxygen neighbors. There are two V - 0 bonds that are shorter where the oxygen is not shared. When one of these oxygen is shared bctween double chains so that they are linked, the layer structure of V 2 0 5 is formed. The oxygen ions in this structure are

OXIDE STRUCTURE

0 0

11

oTi

Figure 2-3 a. A unit cell of rutile TiO,; b. Two edge-sharing octahedra of Ti06 units from adjacent unit cells; c. A network of octahedra that makes up TiOz.

Figure 2-4 Moo3 structure. a. A double-chain unit; b. Cross-section of the layer structure of sheets of double-chains. The Octahedra are shown as squares with diagonals.

12 considered to be attached to one, two, or three vanadium ions. The one attached to one V is the shortest, and is referred to as V=O. The vanadium ion is considered to be in a distorted octahedron in which the sixth oxygen is from another layer, and is very far from the vanadium ion. The last simple structure to be discussed is the perovskite structure. Compounds of this structure are usually of the formula M1MIIO3. An ideal perovskite structure is made up of a cubic net of corner-sharing octahedra (Fig. 2-5). The smaller and more highly charged cation, MII sits in the center of an octahedron, and the larger cation MI sits in the center of the cavity defined by a cube of eight octahedra. Thus this latter cation is coordinated to twelve oxygen ions. Usually, the MI1 ions are the transition metal ions, and the MI ions are the alkali, alkali earth or lanthanide ions. Some examples of perovskites are KTa03, SrTi03, and LaCoO3. Some compounds have ideal perovskite structures, but many others, especially those that have large cations, are often distorted. In addition to those described, there exist many other structures, such as scheelite, pyrochlore, and wolframite. The variety increases with the number of components in the compound. The readers are referred to the text by Wells [l] for discussions of these structures.

2.2 SHEAR STRUCTURE IN INTERMEDIATE OXIDES OF Ti, V, AND Mo Many transition metal ions possess multiple stable oxidation states. This is evident from the compounds listed in Table 2-1 where a number of oxides of different stoichiometries are found for many elements. Among these elements, four have unusually large number of stoichiometries. They are Ti, V, Mo and W. Table 2-2 lists the known oxides of Ti, V, and Mo that have well-defined threedimensional crystal structures. It is evident that many of these oxides differ only slightly from each other. In fact, their structures are very similar and can be constructed from the same building blocks. This contributes to the easy conversion of one oxide to another of adjacent stoichiometry by oxidation or reduction. The easy oxidation and reduction, and the existence of cations of different oxidation states in the intermediate oxides have been thought to be important factors for these oxides to possess desirable properties in selective oxidation catalysis. As an example, the structures of M o and~ M 0~~ 0 ~ ~~ h athe v~ ecommon building blocks of a three-dimensional cubic network of comer sharing octahedra (the Re03 structure, which is also the building block for perovskite). The structures of these compounds are formed when the positions of some octahedra are shifted that pairs of octahedra share edges instead of comers (Fig. 2-6). The shift in the positions are regular along certain directions called shear planes. The farther is the stoichiometry from Moo3, the higher is the density of shear planes.

OXIDE STRUCTURE

Table 2-2 Intermediate Oxides of Ti, V, and Mo Ti30 Ti20 Ti0 Ti203 Ti305 Ti407 Ti509 Ti02

Figure 2-5 A unit cell of perovskite MIMIIOJ. The MII ion is in the center of a comersharing octahedron.

13

14

Figure 2-6 Shear structure in molybdenum oxides. a. Re03 structure; b. MogOZ6;c. MoSOZ3.Each Moo6 octahedron is shown as a half-filled square block.

OXIDE STRUCTURE

15

2.3 STRUCTURE, STABILITY AND RECONSTRUCTION OF

OXIDE SURFACES The positions of the surface ions may or may not be the same as those defined by simple extension of the bulk structure, depending on whether the free surface reconstructs or not. The driving force for reconstruction is to lower the surface Gibbs energy per unit surface area to attain a thermodynamically more stable system. However, metastable surface structure can exist if the energy barrier for reconstruction is too high. At present, there are few reliable experimental values for surface Gibbs energies of oxides. Reliable calculated surface energies are also difficult to find. The uncertainties in the true ionic charge, the degree of covalency, and other surface properties such as compressibilities have made it difficult to perform accurate calculations. Therefore, information on the thermal stability of a certain surface plane comes entirely from experimental data. Table 2-3 lists the surfaces of transition metal oxides that have been studied. A surface is regarded stable (in some cases metastable), if it has been reported to yield a 1x1 LEED pattern at low temperature. A thermodynamically stable surface structure will not reconstruct on heating, provided that the temperature is not too high, while a metastable surface would usually undergo reconstruction. In the absence of knowledge of surface Gibbs energy, it is useful to develop qualitative guidelines as a first approximation to compare the stabilities of different surface structures. Correlations with two effects appear reasonable. One is the polarity of the surface, and the other is the degree of coordinative unsaturation of a surface cation. When a crystal of a binary oxide is cleaved to generate two new surfaces, the ions in the cleavage plane are partitioned into the two separating solids in such a manner that charge neutrality is maintained in each solid. The structure of the two newly created surfaces, however, may or may not be identical. If they are identical, or if the surface plane contains a stoichiomeuic ratio of cations and anions, the surface will be dipoleless and it is called a nonpolar surface. If they are not, the surface will probably (but not necessarily) possess a strong dipole and the surface is a polar surface. A schematic representation of the two types of surfaces is shown in Scheme 1. Examples of nonpolar surfaces are rocksalt (100) surface, rutile (110), (loo), and (001) surfaces, pervoskite (100) surface, and corundum (047) surface. Examples of polar surfaces are wurtzite (0001) and (0001) and rocksalt (111) surfaces. With other factors being identical, a polar surface is less stable than a nonpolar surface. The presence of a dipole moment increases the surface Gibbs energy. Comparing a metal-polar and an oxygenpolar surface, the latter is usually more stable because oxygen ions are more polarizable than metal ions. Polarization lowers the surface electric field and the surface energy. The degree of coordinative unsaturation of a surface cation measures the number of bonds involving the cation that have to be broken to form a surface.

16

Table 2-3 Stability of Oxide Surfaces

Surface

Stabil- Cau'on coordination itya bulk surface

+ NiO (100) c o o (100) + MnO (100) + TiO, (110) + (100) +/(001) +/SrTiO, (100) + (111) +/T$O,(047) + v p , (047) + v,o, (010) + a-Fe,O, (001) BaTiO (001) ZnO (13oio) + (1120) + (4041,5031) + (Oool) +/-

6 6 6 6 6 6 6 6 6 6 6 6 6 4 4 4 4

+I-

4

+

6 6

(mi)

MOO, (010) wo, (100)

+

5 5

5 6 and 5 5 4 5 3 5 5 6 and 5 3 5 3 3 3 3 4 6 6

Polar or nonpolar

n n n n

n n n P

P n n n n P P n n

Ref. Comment

1-3 3 3 4-6 5 73 22 9 10 10 11 12 13 14-17 16 17,18 18,19 19 20 21

Footnotes: a ) -,+/-, and + represent structures of increasing stability to thermal treatment. b)reconstruct to (2x2) and (6x6) structures at high temp. (ref. 23). C)reconstruct to (1x3). (1x5), and (1x7) structures on heating. d, facet to (011) and then to (114) which is stepped (001) on heating. e, reconstruct to (2x2). (6x &)R30" and splitting of spots. reconstruct to (6 x g , 6-fold "1x1" at low temp., spot-splitting at 300-400°C. (2x2). (5x1), and (6x a R 3 0 " above 700°C. h, 6-fold "1x1". ' ) distorted "1x1". f ,

6.

References: 1. F.P. Netzer and M. Prutton, J. Phys. C., 8, 240 (1975). 2. M.R. Welton-Cook, and M. Prutton, J. Phys. C., 13, 3993 (1980). 3. M. Prutton, J.A. Walker, M.R. Welton-Cook, R.C. Felton, and T.A. Ramsey, Surface Sci., 89, 95 (1979).

17

OXIDE STRUCTURE Table 2-3continued 4. R.H. Tait and R.V. Kasowski, Phys. Rev. B. 20,5178 (1979). 5. Y.W. Chung, W.J. Lo, and G.A. Somorjai, Surface Sci., 64,588 (1977). 6. V.E. Henrich, G. Dresselhaus, and H.J. Zeiger, Phys. Rev. Lett., 36, 1335 (1976). 7. V.E. Henrich. and R.L. Kurtz, Phys. Rev. B, 23, 6280 (1981). 8. L.E. Firment, Surface Sci., 116. 205 (1982). 9. W.J. Lo, and G.A. Sornorjai, Phys. Rev. B, 17,4942(1978). 10. R.L. Kurtz, and V.E. Henrich, Phys. Rev. B. 25. 3563 (1982);26,6682 (1982). 11. L. Fiermans, and J. Vennik. Surface Sci., 18,317 (1969); 9, 187 (1968). 12. R.L. Kurtz, and V.E. Henrich, Surface Sci., 129,345 (1983). 13. A. Aberdam, and C. Gaub, Surface Sci.. 27,571 (1971);27, 559 (1971). 14. M.F. Chung, and H.E. Famsworth, Surface Sci., 22. 93 (1970). 15. J.D. Levine, A. Willis. W.R. Bottoms, and P. Mark, Surface Sci., 29,944 (1972). 16. S.C. Chang. and P. Mark, Surface Sci.. 45,721 (1976). 17. W.H. Cheng, and H.H. Kung, Surface Sci., 122,21 (1982). 18. W.H. Cheng. and H.H. Kung, Surface Sci., 102. L21 (1981). 19. S.C. Chang. and P. Mark, Surface Sci.. 46,293 (1974). 20. L.E. Firment, and A. Ferretti. Surface Sci., 129. 155 (1983). 21. M.A. Langell, and S.L. Bernasek, J. Vac. Sci. Technol., 17, 1287, 1296 (1980). 22. V.E. Henrich, G.Dresselhaus, and H.J. Zeiger, Phys. Rev. B, 17,4908(1978). 23. M. Langell. N. Cameron, Surface Sci., 185, 105 (1987).

I

-M-O

1

l

l

-

"

l

l

M - 0 -M

l

-

I

l

f nonpolar surfaces

I

\o/ M\o' M\o'

-0-M-O-M-Ot cleavage

L

+

t 0-polar surface M-polar surface

1

Nonpolar surfaces

Mewl-polar (lower) and Ox ygen-polar (upper) Surfaces

Scheme 1

18

The degree of coordinative unsaturation of a surface anion can be similarly defined. Qualitatively the smaller the number of bonds broken, the more stable is the surface. Thus with all other factors being identical, a surface is more stable if its surface cations and anions are more coordinatively saturated. This factor, together with the surface dipolar factor, qualitatively explains the trends observed in Table 2-3. When a metastable surface acquires enough thermal energy to overcome the barriers for atomic migration, reconstruction by rearrangement of surface atoms takes place to lower the areal surface Gibbs energy. The structure of a reconstructed surface is no longer a simple extension of the bulk. At present, only a few structures of reconstructed surfaces have been analyzed. One example is the TiOz (001) surface. Upon heating, this surface facets. Analysis of the LEED pattern shows that the faceted surface is a stepped surface which is composed of (011) facets at low temperatures, and (114) faccts at high temperatures [2]. Other examples are the ZnO polar surfaces. A LEED pattern showing a "1x1" pattern of six-fold symmetry is often obtained for the ZnO (0001) and (0001) surfaces, although the truncated bulk structure possesses only a three-fold symmetry. Two possible explanations for the six-fold symmetry have been proposed. One explanation comes from the LEED intensity analysis of the Zn-polar surface [3]. It has been found that the outermost Zn layer relaxes inward by about 0.02 nm (see Table 2-4). This brings the layer of Zn closer to the layer of 0 atoms, and the surface behaves like a hexagonal close-packed layer of atoms in the LEED analysis. The other explanation is that double layer steps are present on these surfaces. Since adjacent double layers are rotated by 60°, a random occurrence of these steps would result in a six-fold LEED pattern [4]. It should be emphasized that the stability of these surfaces has been studied under ultra high vacuum conditions such that the surfaces are presumably clean. The environment during catalytic reactions or during oxide powder preparation is either that of an aqueous solution or of gases of atmospheric pressure or higher. Adsorption of molccules can substantially change the stability of a certain surface structure. Adsorption often increases the stability of an otherwise unstable surface. In vacuo, some surfaces lose oxygen atoms to lower the Gibbs energy of the system. This is accompanied by reduction of the cations. If the oxygen atoms are lost such that the anion vacancies form a periodic structure, a superstructure is observed in the LEED pattern, This phenomenon has been used to explain the data for a TiOZ (100) surface [5]. Upon heating this surface to increasingly higher temperature, (1x3), (1x5) and (1x7) patterns arc formed consecutively. Simultaneous AES data indicate increasing loss of surface oxygen. In addition to reconstruction, small relaxations of the surface atoms from their truncated bulk positions also occur. The bonds of covalent compounds arc not very compressible, and it would be expected that the anion-cation bond lengths would remain approximately constant during relaxation [6,7], as is observed for GaAs (110) [8]. The extent of relaxation also depends on the presence of electron

OXIDE STRUCTURE

19

Table 2-4 Relaxation of Surface Atoms of Oxides Deduced from LEED Data

Surface

Relaxationa 60

NiO (100)

coo (100) ZnO (lOi0) ZnO (0001) coo (111)

%s,dJ

6M

0 to -3% 0 to -3% small 0.051

Rumpling s,-60

Ref.

-5to5% Oto3%

a b b

4 . 3 f 0.11 4 . 1 4.21 0 0.1 to 0.21 -15% +_

C

d e f

Footnote: a) negative relaxation is towards the solid. References: a) C.G. Kinniburgh, and J.A. Walker, Surface Sci., 63,274 (1977). b) M. Prutton, J.A. Walker, M.R. Welton-Cook, R.C. Felfon, J.A. Ramsey. Surface Sci., 89, 95 (1979). c) C.B. Duke, A.R. Lubinsky, S.C. Chang, B.W. Lee, and P. Mark, Phys. Rev. B, 15, 4865 (1977). d) C.B. Duke, A.R. Lubinsky, Surface Sci.. 50, 605 (1975). e) R.E. Watson. M.L. Perlman, and J.W. Davenport, Surface Sci.. 115, 117 (1982). f ) A. Ignatiev, B. Lee, and M. Van Hove, Proc. 7th Intern. Vacuum Congr.. Vienna. R. Dobrozemsky, et al. ed., 1977, p. 1733.

lone pairs on the surface [9,10]. For example, the oxygen atoms at the (OOOT) oxygen-polar surface of ZnO have two electrons in the lone-pair orbital pointing outward from the surface, while the Zn atoms in the (0001) Zn polar surface do not have such filled orbitals. Surface atom relaxation is expected to be different for the oxygen and for the Zn atoms, assuming that ZnO is a covalent compound. For ionic compounds, relaxation of a surface ion is determined by three factors [ l l ] : changes in h e ionic size because of reduced surface charge transfer resulting from a reduced Madelung potential at the surface, imbalance of the ionic forces because of the termination of the lattice, and the residual influence of covalent bonding. The extent and the direction of relaxation depend on the relative contributions of these factors. Since the restriction on bond directions and bond lengths in an ionic compound is much less severe than in a covalent coompound, it is possible that both the metal ion and the oxygen ion relax inward. This difference between covalent and ionic

20

surface

bulk ionic

covalent

SIDE VIEW

Figure 2-7 Side view of a solid showing relaxation of surface atoms in an ionic and a covalent compound. The dotted circles represent truncated surface lattice positions.

relaxation is schematically illustrated in Fig. 2-7 for the surface atoms of the zinc oxide (1010) plane [l 11. Available data on the relaxation of surface atoms of transition metal oxides are summarized in Table 2-4. For these oxides, the relaxation is of the ionic type, and both the metal and the oxgyen atoms relax inward. These results are not too surprising as these oxides are of the first transition period. They are relatively ionic. Since the ionicity decreases on going to the lower right hand comer of the periodic table, one would expect the mode of relaxation of the surface atoms of oxides of ,t'F Pd, Cu, etc. to be different. It should be emphasized that substantial amounts of defects may exist in surfaces of oxide crystallite powders prepared by conventional techniques. It has been reported that steps of different heights exist on surfaces of Moo3 crystallites. These surfaces also reconstruct readily upon thermal heating or irradiation by electron beams [12]. In fact, single crystal surfaces may also have well-ordered domains coexisting with disordered regions [ 131.

2.4 STRUCTURE OF SUPPORTED OXIDES Supported oxides that do not interact strongly with the support form threedimensional crystallites whose properties are similar to large bulk crystals. However, some oxides interact strongly with oxidic supports such that a monolayer of an oxide of properties different from the bulk oxide is formed. Growth of threedimensional crystallites occurs only after a substantial fraction of the surface is covered by the monolayer. This is often the case for oxides of Cr, Mo, W, V, and

OXIDE STRUCTURE

21

Re supported on alumina, and for vanadium oxide supported on anatase titania. The absence of three-dimensional crystallites and the presence of monolayer structure are readily observed by Raman spectroscopy. Figure 2-8 shows the Raman spectra of vanadia. The spectrum for V 2 0 5 (curve a) shows intense sharp peaks at around 996, 703, 530,483, and 406 cm-' [14]. The 996 cm-' band is assigned to terminal V=O stretch, and the 703 and 530 cm-' bands are assigned to bridging VC&V symmetric and antisymmetric stretch, respectively. When 2.1 wt.% of vanadia is supported on y-AI2O3 (curve b), the spectrum does not show any of the prominent V2O5 peaks. Instead, two featureless broad peaks appear; one centers around 970 cm-' accompanied by a shoulder around 995-1000 cm-', and the other in the range between 800-830 cm-' [15]. When a larger amount of vanadia is supported on the alumina (curve c), the intense peaks characteristic of crystalline V2O5 appear. Interestingly, these characteristic peaks begin to appear at lower vanadia loadings on a silica support than on alumina [16]. This suggests that vanadia interacts more strongly with alumina than with silica so that it forms a monolayer structure with alumina but not with silica. Vanadia also interacts strongly with anatase Ti02 to form a layer structure [17]. Vanadia supported on anatase Ti02 is a much studied system. It is generally believed that the vanadia layer structure exposes the (010) plane and the V=O groups preferentially [ 18-20] because of the excellent match of the vanadate unit and the anatase structure. It has been proposed that on the (001) plane of anatase, vanadium exists as monovanadate (V04)"- groups bonded to the surface as [21]:

\

l I O\ l / O\ l /O\ l / Ti Ti Ti Ti

In this form, the V ion occupies roughly a position which another Ti ion would occupy if the bulk structure continues. The V-0 bridging bond length of 0.190 nrn is very close to the Ti-0 bond length of 0.193 nm. Indeed, EXAFS studies show that on anatase, the basic structural unit of vanadium contains two terminal V=O bonds of a bond length of 0.165 nm, and two bridging V - 0 bonds of 0.190 nm [221. The Dicture is also consistent with the observation that an average of one oxygen ion per vanadium can be readily removed on reduction [21]. However, this model has not been supported by IK data which do not show the expcctcd bandsplitting due to coupling of the two V=O groups [23]. On other crystallographic planes of anatase, however, there are no sites of such a good match with the monovanadate groups, and other structures may be present. On y-AI2O3, EXAFS and XANES studies show that vanadium ions are present in a more regular tetrahedral coordination than on anatase. Each vanadium ion is associated with two terminal V=O bonds of 0.167 nm, and two bridging bonds of 0.182 nrn bond length [21,22]. Upon reduction, each vanadium

22

1500

1000 Wavenumber

5bO cm-l

Figure 2-8 Raman spectra of supported and unsupported vanadia. a. V205; b. 2.1 wt.% V/y-A1203; c. 4.0 wt.% Vfi- A1203; d. 2.3 wt.% V/SiO2. Curves a-c are from Z i t . Phys. Chern. Neue Folge, 111, 215 (1978), curve d is from J. Phys. Chem., 84, 2783 (1980).

ion loses 0.6 oxygen atoms readily on the average. A picture that is consistent with these observations is that the vanadium ions exist as dimcric units on y-AI2O3, such as pyrovanadate units [21]:

Like vanadia supported on titania or A1203, molybdena on thcse supports also cxists as well-disperscd units or monolaycr until high loadings. This is illustratcd by Raman spectroscopy of MoO3/AI203 [24] and M003/ri02 1251. As shown in Fig. 2-9, the intense and sharp characteristic bands of crystalline Moo3 at 998

23

OXIDE STRUCTURE

I

I

800

I

1000

Wavenumber

I

1

1200 cm-’

Figure 2-9 Raman spectra of supported molybdena. a. 12 wt.% MoO3/q-AI2O3; b. 4 wt.% Mo03/A1203;c. 13.5 wt.% Mo03/Ti02;d. 1.8 wt.% MoO-,/riOZ. Curves a and b are from J. Phys. Chern., 82, 2002 (1978), and curves c and d are from J. Catal., 94, 108 (1985).

and 821 cm-’ are not present in samples of low loadings of Moo3. They appcar only in samples of high loadings. A monolayer structure is formed when the Mo loading is below about 5 Mo atoms per nm2. Below this value, the ratio of the intensity of the Mo signal to the A1 signal in XPS and ISS increases linearly with the Mo loading, consistent with the picture of a well-dispersed phase [26,27]. At higher loadings, this ratio reaches a plateau. Electron microscopy also shows a highly dispersed phase of molybdena at low loadings [28,29]. Interestingly, crystallites of MoS2 on alumina can be redispersed by reoxidation to form a dispersed phase of molybdena. One picture that has been proposed to explain these and other evidence is that the molybdenum ions exist as monomeric MOO^^- units at loadings up to about 1 Mo per nm2. Betwecn 1 to about 4.5 Mo per nm2, heptameric Mq0246p units and octahedrally coordinated polymeric surface species are also present [30-341. The heptamer is proposed to be adsorbed as a bilayer with four Mo ions lying next to the AI2O3and the remaining three on top [32]. At still higher loadings, crystalline Moo3 appears.

24

0

500

Wavenumber

1000 cm

-1

Figure 2-10 Raman spectra of tungsten oxides. a. WO,; b. A12(W04),; c. 10 wt.% W03/A1203. (From J. Catal.. 90. 150 (1984). copyright Academic Press.)

The existence of these well-dispersed or monolayer phases is actually quite common among supported transition metal oxides. Fig. 2-10 shows another example of WO3 supported on alumina [35,361. Commonly found in these welldispersed phases are the terminal metal-oxygen double bonds. These M=O bonds usually have characteristic Raman stretching frequencies in the 950-1000 cm-' region. These peaks are often at somewhat different frequencies and broader than the corresponding peaks for the crystalline oxides [371. Indeed such M=O groups have been detected for W6+, Mo6+, V5+, Re7+, and Cr6+ oxides on y-AI2O3 [37,38], and Mo6+ and V5+ on titania [39,401. The frequencies of these terminal double bonds vary somewhat depending on the surface coverage of the dispersed phase and on whether the double bond is interacting with adsorbed water or surface hydroxyl groups [41 and references thercin]. Nickel oxide also forms a well-dispersed phase on y-A1203 instead of crystallites of NiO. This well-dispersed two-dimensional phax may actually be Ni occupying the octahedral and/or the tetrahedral holes at the surface of the alumina to form a layer of spinel-like NiA1204 [26,27,42]. When Ni is supported together with Mo on y-AI2O3, there is interaction between Ni and Mo which affects their locations on the support. It has been

OXIDE STRUCTURE

25

suggested that at low calcination temperatures, the Ni ions interact strongly with the molybdenum species to partially shield the oxomolybdenum ion. At high calcination temperatures, the Ni ion migrates to the tetrahedral or octahedral holes of the alumina, exposing the molybdenum ions [27,43]. Similar incorporation into the alumina has also been observed in Ni-W/AI2O3 [44]and for Co in the CoMo/A1203system [451.

REFERENCES 1. A. F. Wells, "Structural Inorganic Chemistry", 4th ed., Clarendon Press, London, 1975. 2. L. E. Firment, Surface Sci., 116, 205 (1982). 3. C. B. Duke, and A. R. Lubinsky, Surface Sci., 50, 605 (1975). 4. V. E. Henrich, H.J.Ziegler. E. I. Solomon, and R. R. Ray, Surface Sci.. 74, 682 (1978). 5. Y.W. Chung. W. Lo, and G. A. Somorjai. Surface Sci., 65, 419 (1977). 6. J. D. Levine, and S. Freeman, Phys. Rev. B., 2, 3255 (1970). 7. J. E. Rowe, S . B. Chrisman, and G. Margaritondo, Phys. Rev. k f t . 35, 1471 (1975). 8. A. R. Lubinsky, C. B. Duke, B. W. Lee, and P. Mark, Phys. Rev. Lett. 36. 1058 (1976). 9. H. C. Gatos, and M. C. Levine, J . Electrochem. Soc.,107, 427 (1960). 10. H. C. Gatos, J . Appl. Phys., 32, 1232 (1961). 11. C. B. Duke, A. R. Lubinsky, B. W. Lee, and P. Mark, J. Vac. Sci. Technol., 13, 761 (1976). 12. J. M. Dominguez-Esquivel, 0. Guzman-Mandujano, and A. Garcia-Borquez, J . Catal., 103, 200 (1987). 13. L. E. Firment, and A. Ferretti. Surface Sci., 129, 155 (1983). 14. I. Beattie, and T. Gibson, J . Chem. SOC. A, 2322 (1969). 15. F. Roozebaum, J. Medema, and P. Gellings, &it. Physik. Chem. Neue Folge, 111, 215 (1978). 16. F. Roozebaum, M. Mittelmeijer-Hazeleger, J. Moulijn, J. Medema, V. de Beer, and P. Gellings, J . Phys. Chem., 84, 2783 (1980). 17. I. Wachs, R. Saleh, S . Chan, and C. Chersich, Appl. Catal., 15, 339 (1985). 18. A. Vijux, and P. Courtine, J . Solid Stare Chem., 23, 93 (1978). 19. G. C. Bond and P. Konig, J . Catal., 77, 309 (1982). 20. Y.Murakami, M. Inomata, K. Mori, T. Ui, K. Suzuki. A. Miyamoto, and T. Hattori, in Preparation of Catalysts III. G. Poncelet, P. Grange, and P. Jacobs ed.,Elsevier Science Publ., Amsterdam, 1983, p. 531. 21. J. Haber, A. Kozlowska, and R. Kozlowski, J. Catal., 102, 52 (1986). 22. R. Kozlowski, R. Pettifer. and J. M. Thomas, J . Phys. Chem., 87, 5176 (1983). 23. G. Busca, G. Centi, L. Marchetti, and F. Trifio, Langmuir. 2. 568 (1986). 24. H. Knozinger. and H. Jeziorowski, J . Phys. Chem., 82, 2002 (1978). 25. Y. Liu, G. Griffin, S . Chan, and I. Wachs, J. Card., 94, 108 (1985). 26. P. Dufresne, E. Payen, J. Grimblot, and J. P. Bonnelle, J. Phys. Chem., 85, 2344 (1981).

26 S. Kasztelan, J. Grimblot, and J. P. Bonnelle, J . Phys. Chem., 91, 1503 (1987). T. Hayden, and J. Dumesic, J . Catal.. 103. 366 (1987). T. Hayden, J. Dumesic. R. Sherwood, and R. Baker, J . Catal., 105, 299 (1987). S. Kasztelan, J. Grimblot. J. P. Bonnelle, E. Payen, H. Toulhoat, and Y.Jacquin, Appl. Cafal.,7 , 91 (1983). 31. L. Wang, and W. K. Hall, J . Cafal..77. 232 (1982). 32. W. K. Hall, Proc. 4th Conf. Chemistry and Uses of Mo, H. F. Barry, and P. C. H. Mitchell, ed., 1982, p.224. 33. H. Weigold, J. Catal., 83, 85 (1983). 34. N. Giordano, J. C. J. Bart, A. Castellan, and G. Martinotti, J . Cafal.,36. 81 (1975). 35. S. Chan, I. Wachs. and L. Murrell, J . Catal., 90, 150 (1984). 36. S. Chan, I. Wachs. L. Murrell. L. Wang, and W. Hall, J . Catal., 88, 5831 (1984). 37. I. E. Wachs. F. D. Hardcastle, and S . S . Chan, Spectroscopy, 1. 30 (1986). 38. S. S. Chan, and I. E. Wachs, J . Catal., 103, 224 (1987). 39. C. P. Cheng and G. L. Schrader, J . Cafal.,60, 276 (1979). 40. K. Y. S. Ng, and E. Gulari, J . Catal., 92, 340 (1985). 41. R. Quincy. M. Houalla, and D. Hercules, J . Catal., 106, 85 (1987). 42. S . Kasztelan, J. Grimblot, and J. P. Bonnelle. J . Chim. Phys., 80, 793 (1983). 43. H. Jeziorowski, H. Knozinger, E. Taglauer, and C. Vogdt, J . Cafal.,80, 286 (1983). 44. B. Horrell. D.L. Cocke, G. Sparrow, and J. Murray, J . Cafal.,95, 309 (1985). 45. F. Delannay, E. Haeussler. and B. Delmon, J . Cafal.,66, 469 (1980). 27. 28. 29. 30.

Chapter 3 PHYSICAL AND ELECTRONIC PROPERTIES

3.1 SURFACE COMPOSITION The surface composition of a single component oxide is determined by the surface anion to cation ratio, which, for an ideal surface, depends on the stoichiometry of the oxide and the orientation of the exposed crystal plane. It is often important to determine whether or not a surface is stoichiometric. Nonstoichiometry often arises from preferential removal of surface oxygen leading to a slight reduction of the surface. The extent of nonstoichiometry depends on the pretreatment of the sample. Stoichiometric surfaces can often be obtained for surfaces that have low surface areal Gibbs energies (the stable surfaces) by low temperature annealing. For example, stoichiometric SrTi03 (loo), TiOz (1 lo), TiOz (100) [l-31, ZnO (1070) [4], and Moo3 (010) [5] surfaces have been prepared. High temperature annealing in vacuo or ion-sputtering preferentially removes oxygen, and the cations near the surface are reduced to lower oxidation states. For oxides that have empty d bands, reduction of the surface to generate nonstoichiornetry rcsults in the appearance of band-gap states which are readily detectable by ultraviolet photoelectron spectroscopy and electron energy loss spectroscopy. For multicomponent oxides, in addition to the surface anion to cation ratio, the ratio of the different cations is also of interest. At least two questions are asked. If the two component oxides form a solid solution, are the surface and the bulk cation ratios the same? If the component oxides form a bulk compound, does the surface have the same chemical stoichiometry as the bulk? Whether the surfacc cation ratio is the same as the bulk ratio depends on a number of factors. Thesc factors include: (1) the surface tension (or surface Gibbs energy) of the component oxides. The lower energy component tends to be segregated to the surface; (2) thc bulk strain of the solid solution due to mismatch of the ionic sizes or the coordination symmetry. The larger ion tends to be segregated to the surface, as docs the ion whose preferred coordination symmetry 27

28

differs from that provided by the matrix; (3) the nature of the adsorbate. Chemisorption lowers the surface energy of the solid. Thus chemisorption tends to induce surface segregation of the component that binds more strongly with the adsorbate; (4) formation of a surface compound. Even if the bulk oxide is a true solid solution, a surface compound of a certain stoichiometry may be formed. The surface composition is then determined by the surface compound. One driving force for the formation of a surface compound is the presence of surface adsorbates that causes the surface cations to have an oxidation state different from the bulk oxidation state. For example, in the presence of oxygen, the surface chromium ions in Cr203have an oxidation state of +6, whereas the bulk ions are +3. In this case, the adsorbed oxygen are being incorporated as surface lattice oxygen ions. There are only a few experimental results on solid solutions. In the dilute solution of ZnO in MgO, the surface is enriched with Zn [6]. This can be explained by point (2) in that Zn prefers tetrahedral coordination, while the coordination symmetry in MgO is octahedral. In the solid solutions containing Cr, which includes CoO-Cr203, NiO-Cr203 [7], and Fq03-Cr203 [8], Cr is enriched on the surface in the region of low Cr concentrations. The surface cation ratios seem to remain constant at intermediate Cr concentrations. This phenomenon has been interpreted as due to the formation of a compound that is segregated onto the surface, such as CoCrZO4 spinel [7]. The surface enrichment of Sb in the Sn02Sb02 solution [9] can be explained by the larger size of Sb than Sn [lo]. In an oxidic multicomponent compound, the surface composition is constrained by the periodicity of the compound, unless the driving force for surface segregation is sufficiently large to disrupt the periodicity. In the limited number of cases reported, the surface has the same cation ratio as the bulk. For example, the surface Bi/Mo ratios for Bi2Mo06,Bi2M0209and Bi2Mo-,OI2 are within 10% of the bulk values [ll].

3.2 IONICITY OF OXIDES The formal oxidation states of transition metal oxides range from one in Cu20 to eight in Os04, for example. However, there is little doubt that the true ionic charges in many transition metal oxides are less than those predicted from formal oxidation states. Since ionicity affects oxide properties including the surface electric field gradient, the surface electrostatic potential, and the mode of surface relaxation, knowledge of its magnitude is important. Recently, experimental measurements of the degree of charge transfer between the cation and the anion become available from X-ray diffraction data [ 12151, and X-ray photoelectron spectroscopy data [ 16-18]. Some theoretical calculations are also available that provide charge lransfcr information. These data are summarized in Table 3-1. In the X-ray diffraction technique, electron dcnsity distribctions are deduced directly from X-ray data. From the distribution maps, boundaries of ions are assumed to correspond to the point of minimum (local minimum) electron density. The number of electrons associated with each ion, and thus the true cation charge,

PHYSICAL AND ELECTRONIC PROPERTIES

29

q, is then obtained by summing the electron density within the boundary. Reliabliliy of the values of q obtained depends on the validity of this method of partitioning of electrons to the various ions. When properly analyzed, the method appears to be adequate for electrostatic potential calculations using a point charge model or a shell model which disregards the nature and spatial orientation of the electron orbitals. The method using X P S binding energy shift data was originally proposed by Siegbhan et al. [17]. The values of q reported here are obtained using the method of Bagus and Braughton [16]. In this method, it is assumed that the experimental binding energy of an ion in a solid differs from the binding energy of the ion in the gas phase by the lattice self-potential at the ion site and the final state extra-atomic relaxation effect. (The final state extra-atomic relaxation effect is an effect that affects the kinetic energy of the escaping photoelectron due to the relaxation of the electrons in the lattice surrounding the ion in response to the removal of the photoelectron from the ion). Assuming that the final state effects are the same for the cation and the oxygen ion, the following relationship is obtained [16,18]: BE(0,XPS) - BE(M,XPS) = BE(0,gas ion) - BE(M,gas ion) + O(0) - O(M)

(3-1)

The left hand side is the difference in the values of experimental binding energy of the observed electrons from the oxygen and the metal ion. The first two terms on the right hand side represent the difference in the binding energy of the same electrons from the gas phase oxygen and metal ions. The last two terms represent the difference in lattice self-potential of the oxygen and the cation in the solid. Since the electron binding energies of the gas phase ions and the term O(0)- @(M) are unique functions of the true ionic charge, it is in principle possible to calculate q using this equation. In calculating the values of q listed in Table 3-1, the electron binding energy calculated for the gas phase ions [19] have been used. The lattice self-potentials used are listed in Table 3-2. Assuming that the experimental values of binding energy are accurate, the accuracy of the ion charge thus obtained still depends on the accuracy of the calculated binding encrgy for the gas phase ions, the validity of the assumption that the final state extra-atomic relaxation effects of the oxygen ion and the metal ion cancel out, and the error introduced by using lattice self-potentials of bulk ions in calculations involving surface and near surface ions. With these limitations, the ionic charges thus obtained are at best qualitative and should be used only for qualitative comparison of ionicity among similar solids. The seriousness in neglecting the final state effect can best be illustrated using Si02 and CdO for which the extra-atomic relaxation effect has been included in the calculation. The charge for Si in Si02 using equation (3-1) is +4. If a relaxation of about 5 eV is included [20], the charge becomes +2. Similarly for CdO, a Cd charge of 0.4-0.8 would be obtained without relaxation compared to a charge of 1.0 with relaxation [21]. A variation of the above method has been used for a number of chromium compounds [22]. In this study, the authors make use of the fact that because of

30

Table 3-1 True Cation Charges of Oxides (From J. Solid State Chem., 52, 191 (1984), copyright Academic Press with update).

Compound

XPS

b

Method a X-ray

1.9 (a), 1.85 (b), 1.5 (c) 1.8 (c) 2 (c) 2 (c) 1.52 (o), 1.35" (0) 0-1 (d) 0.66 (1) 1.4 (a), 1.51 (b) 1.24 (o), 1.03" (0) 1.29 (o), 1.03" (0) 1.2 (a), 1.4 (b) 1.16 (o), 0.78" (0) 0.7 (a), 0.9 (b) 1.2-1.8 (m), 1.9 (n), 1.00 (o), 0.74"(0)

MgO CaO SrO BaO Ti0

1.25-1.75

MnO FeO coo NiO

1.5

CUO ZnO CdO Ago Ag(1) Ag(II1) a-Fe203 a-Cr2O3 a-A1203 Ti02 S i02 cro2 zrO2 crO3 Moo3

2 1.2-1.7, 1.11" (h) 1.05" (h) 0.53" (h) 1.87' (h) 2.7 0.15d (g), 2.6 2-2.5 (j) 2.8 4, 2" (i) 1.0 (k) 0.38d (g) 1.2 0.46d (g) 1.5-2 1.5 0.48" (h)

vo

wo3

&2O

1-1.5 1-1.5

(MI) 1.84 (b) (M2) 1.79 (b) LiA1Si206(Li ion) 0.7 (b) 2.4 (b) (A1 ion) CaMgSi206 (Mg ion) 1.44 (b) (Ca ion) 1.39 (b) Mn2Si04 (Mn ion) (Ml) 1.21 (b) (M2) 1.49 (b) Mg2Si206

(Mg iron)

Calculation

2.19" (p)

PHYSICAL AND ELECTRONIC PROPERTIES

31

Table 3-1 continued Footnotes: a) Letters in the brackets denote references. b, Unless noted, the charges are calculated using equation 3-1 in text. The XPS data are from references e and f except when indicated. ') The final state extra-atomic relaxation of the cation has been included in calculating these values. See text for detail. d, These are taken directly from reference g using a modified form of equation 3-1. See text for detail. ') These are for clusters that resemble a surface. References: a) S. Sasaki, K. Fujino. and Y. Takeuchi, Proc. Japan Acad., 55, Ser R, 43 (1979). b) S. Sasaki, K. Fujino, Y. Takeuchi, and R. Sadanaga, Acta Cryst. A36, 904 (1980). c) G. Vidal-Valef J.P. Vidal, and K. Kurki-Suonio, Acta Cryst., A34, 594 (1978). d) M. Morinaga, and J.B. Cohen, Acta Cryst., A32, 387 (1976). e) C.N.R. Rao, D.D. Sarma, S . Vasudevan, and M.S. Hegde, Proc. Royal SOC. London, A367, 239 (1979). f) D.D. Sarma, and C.N.R. Rao, J. Electron. Spectra. Related Phenom., 20, 25 (1980). g) T. Dickinson, A.F. Povey, and P.M.A. Sherwood, J. Chem. SOC.Faraday Trans. I, 72, 686 (1976). h) S.W. Gaarenstroom, and N. Winograd, J. Chem. Phys., 67, 3500 (1977). i ) F. Bechstedt, Phys. Stat. Sol. B, 91, 167 (1979). j ) V.I. Nefedov, D. Gati, B.F. Dzhurinskii, W.P. Sergushin. and Ya. V. Salyn, Russ. I. Inorg. Chem., 20, 1279 (1975). k) R.F. Steward, M.A. Whitehead, and G . Donnay, h e r . Mineralogist, 65, 324 (1980). 1) V.A. Gubanov, B.G. Kasimov, and E.Z. Kuxmaev, J. Phys. Chem. Solid, 36, 861 (1975). m)A.B. Anderson, Chern. Phys. Lett., 72, 514 (1980). n) P.S. Bagus, and U. Wahlgren, Mol. Phys., 33. 641 (1977). o) C. Satoko, and M. Tsukada, IMS Ann. Review, 1979, p. 33; M. Tsukada, H. Adachi, and C. Satoko, Prog. Surface Sci., 14, 113 (1982). p) M. Tsukada, C. Satoko. and H. Adachi, J. Phys. SOC.Japan, 48. 200 (1980).

reduction by X-ray, XPS data could be recorded for Cr ion and reduced Cr atom from the same sample. Presumably, since both the Cr ion and the Cr atom are embedded in the same solid matrix, their extra-atomic relaxation should be the same. Then using an equation similar to equation (3-1) except that the difference between two Cr peaks are used instead of between oxygen and chromium, the Cr ion charge is calculated. It can be scen (Table 3-1) that the ionic charge thus obtained is smaller than that obtained using equation (3-1). Finally, the ionic charge provided by some theoretical calculations are included. These calculations are cluster calculations and the ionic charges are calculated from the Mulliken populations. It is apparent that different methods yield different ionic charges. The

32 Table 3-2 Lattice Self-potential a t the Cation Sites of Oxides a

Compound Potential, eV

MO

MgO CaO SrO BaO MnO FeO coo NiO

-23.9 -20.9 -19.5 -18.2 -22.7 -23.3 -23.6 -24.2

CdO CUO ZnO PbO

-21.4 -24.3 -24.0 -20.5

M02 Ti02 Sn02 Si02 ca2

zrO2

Compound

Potential, eV

-34.9 -34.8 -36.6 v2°3 -33.4 Ti203 -33.6 Ga203 -35.0 Rh203 -34.1 Pb203 Pb(1) -32.8 Pb(2) -28.1 -29.0 Nd203 -29.7

M2O3 a-Cr2O3 a-Fe203 a-A1203

44.7 M03 42.9 4 6 . 4 to -50.6b -46.2 42.3

Cr03 Ma3 wo3

-58.5 -58.5 -64.5

Footnotes: a) Taken from J. Solid State Chem., 52, 191 (1984); original data from J.Q. Broughtori and P.S. Bagus, J. Electron Spectros. Relat. Phenom., 20, 261 (1980). and W. Van Cool, and A.G. Piken. J. Mater. Sci., 4. 95 (1969). b, Value depends on the crystal structure.

differences, in addition to experimental variations, may be due to the fact that different methads measure ionic charges that are defined differently. Only in the limit that eleclrons associated with each ion are clearly defined and there is no overlap of electron density from different ions that the different methods could yicld the same charge. In spite of this, however, it is worth noting that if the data are analyzed as exactly as possible with minimal simplification, h e ionic charge does not differ significantly among the different methods. The values shown in Table 3-1 also confirm the fact that, in general, the trend of ionicity parallels that derived from electronegativity arguments. Thus the alkali metal oxides are rather ionic with ionicity increasing with increasing atomic masses as one goes down a given group. Transition metal oxides near the middle of the pcriod (groups V1 to IIB) have true ionic charges that arc about half of the formal

PHYSICAL AND ELECTRONIC PROPERTIES

33

charges. Finally, Si02 which is normally thought of as covalent does possess a cation charge of one. When the true ionic charge is not available from experimental or theoretical results, relative ionicity can still be estimated using the electronegativity scale of Pauling [23], Sanderson [24] or Phillips [251. The basic assumption in Pauling's ionicity is that if the heat of formation of a A-B bond, DAB exceeds the arithmetic average of the heats of formation of the homopolar A-A and B-B bonds, DAA and DBB, the extra energy is due to the transfer of electron from the less electronegative to the more electronegative atom in the bond. It is, therefore, ionic in origin. A scale of elemental electronegativities XA and XB can be defined from the relations:

The proportionality factor has the dimension of energy. XA and XD are dimensionless and increase by 0.5 with valence charge of unity for the first row of the periodic table. Fractional ionic character f,(A,B) is defined by:

to satisfy the conditions that 0 5 f, 5 1, and f,(A,B) = f,(B,A). Phillips [25] has proposed that the energy of a bond contains a homopolar, or covalent, and a heteropolar, or ionic part, the values of which can be defined accurately in terms of spectroscopic transition energy between bonding and antibonding states. This proposal stems from the observation that the bonding states have lower energy, are centered predominantly on the more electronegative atom, and point towards the nearest-neighbor atoms. The antibonding states are centered predominantly on the more electropositive atom, and point away from the nearest neighbors. The homopolar energy Eh is assumed to depend only on the bond length or nearest neighbor distance r, and the position in terms of the rows in the Period Table under concern. The average energy gap between bonding and antibonding state, E,, is then the geometric sum of Eh and the average ionic energy gap C: Eg2 = Eh2

+ c'

(3-4)

Phillip's ionicity fi is defined as:

The values of E, can be evaluated from the optical properties of the solids such as the valence electron plasma frequencies or the low frequency electronic dielectric constants, and Eh can be extrapolated from values of E, of elemental compounds such as diamond, silicon, germanium and tin using an empirical relation that Eh K r 0 , where r is the interatomic distance, and S an empirical constant with a value of 2.48.

34 Table 3-3 Phillips' Ionicity, f,

MnO 0.887 FeO 0.873 COO 0.858 NiO 0.841 C U ~ O 0.56 Ge20 0.730 Ti02 0.686 A1203 0.796 Cr2O3 0.777 F%03 0.677 LiNb03 0.825

Be0 ZnO MgO CdO CaO SrO 0.65.67 0.65-.66 BaO 0.53 0.53 Ge02 0.49 0.51 Sn02 Li20 si02

0.620 0.653 0.839 0.778 0.916 0.928 0.93 1 0.730 0.784 0.57 0.766 0.570 0.57-.59

0.57 0.57-.59

(Nb-0)

LiTa03 0.850 (Ta-0)

0.511

0.53

0.54

A1203 0.796 LiGa03 (Li-0) 0.815 (Ga-0) 0.653 TeOz

0.67

0.63

Ge02

Footnotes: a ) From B.F. Levine, Phys. Rev. B. 7. 2591 (1973). b , From B.F. Levine, J. Chem. Phys., 59, 1463 (1973). ') From F. Gervais. Solid State Commun., 18, 191 (1976);I I and to the polarization directions.

L

refer

The ionicities defined in this method are obtained from measuremcnts of the interaction of electrons in the solid with incident photons. Thus they are based on an effective dynamic charge e that differs from the static charge q by [26]:

e = q + R dq/dr

(3-6)

whcre R is the equilibrium interatomic distance, and dq/dr is the sensitivity of the static charge transfer to small shifts in the atomic positions. The difference bctwecn e and q arises from the fact that in the optical measurement of E,, the solid interacts with an electromagnetic force which draws atoms from their equilibrium positions. From the viewpoint of electrostatic interaction, the effect of this movement is equivalent to that of placing on an unperturbed lattice an electric dipole of magnitude c and a lenglh equaling the change in the interatomic distance.

PHYSICAL AND ELECTRONIC PROPERTIES

35

Since atoms or ions in a solid strongly interact with each other, their movements are strongly coordinated, giving rise to group vibrations (phonons). Therefore, it is possible to have different values of dynamic charges (Ze) for a given solid depending on the type, direction and frequency of the vibration. The relationship between this dynamic charge and the dielectric constants is given by Gervais [271:

WLo and WTO are frequencies for the longitudinal and transverse mode, p is the reduced mass of the ions, q is the dielectric constant of vacuum, and V is the molar volume. It can be shown that Ze is the actual transverse charge of the ions within the rigid ion model description. Phillip's original concept has been developed for simple AB crystals. It is later extended by Levine to compounds of other structures and stoichiometries, including oxides, ternary compounds, and compounds containing transition metal cations [28,29]. The ionicity reported for a number of oxides are listed in Table 3-3. By examining a large number of compounds including halides, halogenides (sulfides, tellurides, etc.), binary oxides and ternary oxides, it is found that values of Z/Zo, where Zo is the formal charge, is roughly linearly proportional to exp(f,) [271:

Z/Zo = 0.24 (efi - e)

(3-8)

An examination of the static and dynamic ionic charges in Tables 3-1 and 3-3 suggests that the ionic charges of transition metal oxides seldom equal their formal charges. In fact, in most cases, the charge is about half the formal charge. Furthcrmore, the static and dynamic charges do not differ significantly. What are the consequences to surface chemistry and catalysis that transition metal oxides are ionic? Qualitatively, the ionic character results in the presence of a strong electric field that points outward from the oxide surface. The separation of charges into cations and anions results in a strongly modulated electronic potential on the oxide surface. These effects lead to the common phenomenon of heterolytic dissociative adsorption of molecules. An example is shown for HZ:

(3-9)

Contrary to H adsorbed on metals, the two hydrogen atoms that are dissociatively adsorbed are not equivalent and carry opposite charges. It is likely that they have different reactivities. It should be noted that the charges shown in the equation do not necessarily represent the real charges which are usually unknown.

The ionic character may also increase the sticking probability of polar molecules, such as ammonia, water, alcohols, acids, ethers and amines. When these molecules approach the surface, their dipole moments interact with the electric field at the surface to orient the molecule, thus enhancing the probability of an attractive bonding interaction. Since coulombic interaction (that is, charge-charge interaction) is a longrange interaction, the surface chemistry of an oxide depends not only on the nature of the cation or anion at the immediate vicinity of a surface adsorbate, but also on the ionicity of the rest of the oxide matrix. The consequence of this effect in dilute solid solutions has been explored. It is concluded that the ionicity of the matrix oxide in a solid solution affects the activation energy of catalytic reactions in which the rate limiting step involves charge transfer between the oxide and the surface intermediate [301. This effect is also important in determining whether new acid sites are formed in oxide solid solutions [31]. This will be discussed in Chapter 5.

3.3 MAGNETIC PROPERTIES OF SMALL OXIDE PARTICLES In some supported transition metal oxide catalysts, the oxide is present as submicron-size crystallites. They may be so small that detection by X-ray diffraction is difficult. Recent studies on these particles indicate that some of their properties differ from those of the bulk oxides. One such property is the magnetic property which has been used to identify the presence of small crystallites of oxides and to determine their sizes. The magnetic properties of an ion or atom are determined by the orientation and the number of its electron spins. For transition metal oxides, the individual electrons are so strongly correlated in their motion that the spin of an ion is better characterized by one total spin (atomic spin) than the individual electron spins. The atomic spins of neighboring ions may also be strongly correlated with each other to form a spin sublattice. Depending on the magnitude, the orientation, and the number of spin sublattices, transition metal oxides possess different internal magnetic fields, as well as different responses to applied magnetic field. There are different types of magnetism. The readers are referred to textbooks on the subject, such as that of Cullity [32] or Morrish [33], for a more extensive discussion. Here we shall give a very brief introduction to magnetism, and then proceed to discuss magnetic properties of small oxide particles. Fig. 3-1 shows the types of magnetism commonly found in transition metal oxides, and the associated spin orientations. The temperature dependence of the magnetic susceptibility x, which is the magnetization per unit applied field, is shown in Fig. 3-2. Diamagnetism is related to changes in the orbital motion of electrons that occur when atomic systems are placed in a magnetic field. This induced motion of electrons (or currents) is set up in such a direction as to oppose the change in the magnetic flux, and persists as long as the magnetic field is present. The magnetic field produced by the induced current is opposite to the applied field, and the

PHYSICAL AND ELECTRONIC PROPERTIES

0 0 0 0 0 0 Diamagnetism

37

p”

Q-

“s

ST

Paramagnetism

0

P Antiferromagnetism

Ferrimagnetism

Figure 3-1 Types of magnetism in oxides. direction.

Direction of arrows indicates the spin

Paramagnetic

Diamagnetic

Anti ferromagnetic

Ferrimagnetic

a

Figure 3-2 Temperature dependence of magnetic susceptibility for different types of , magnetism, For antifcrromagnetism. the Nee1 temperature is shown. For femmagnetisrn, saturation magnetization 0,and the Curie temperature T, are shown.

38

Table 3-4 Room Temperature Magnetic Properties of Some Common Transition Metal Oxides a Paramagnetic Ti203

vo vo2

Nd02

Antiferromagnetic

Ferrimagnetic

Diamagnetic

cr203

MnO Mn203

Mn02 FeO a-Fe203 COO (3304

NiO v203

Footnote: a) Omitted are the ferromagnetic oxides, such as CrOz, in which there is only one spin sublattice where the spins are all oriented in the same direction.

magnetic moment associated with the current is a diamagnetic moment. Paramagnetism is related to the tendency of a permanent magnetic dipole to align itself parallel to a magnetic field. In transition metal ions, the permanent magnetic moment is associated with coupled electron spin and orbital motion of partially filled shells. In a paramagnetic material, these permanent moments are normally random and uncorrelated, but are aligned in an applied field. The magnetic susceptibility decreases with increasing temperature because thermal motion tends to randomize any spins that are aligned by the external field. In a ferromagnetic material, the interaction among spins is so strong that the magnetic moments are aligned parallel to each other. Such an internal interaction is called exchange field. However, few oxides are ferromagnetic. In a simple antiferromagnetic material, there are two sets of strongly correlated spins that form two spin sublattices of equal magnitude but opposite orientations. These two sublattices cancel each other, resulting in a solid that is like a paramagnet and has weak magnetization. Below the ordering or Ned temperature, the behavior of an antiferromagnetic material differs from that of paramagnetic material because at such low temperatures, the spin sublattice is so rigid that the effect of small imperfect cancellation of spins becomes evident. In a ferrimagnetic material, the magnitudes of the two opposing spin sublattices are different, which result in a net sizable spin. In an applied magnetic field, these spins align with the external field to yield a saturation magnetization M,, which can be attained at a relatively low field of about 100 Oe. On increasing

PHYSICAL AND ELECTRONIC PROPERTIES

39

100

-8 --- 8

(0

0 F

X

-.-.-

nm nm

9 nm

...........12

-

-0.--

nm

1 4 nm

22 nm

-x-x-400 I

1

1

100

200

300

Temperature

nm

K

Figure 3-3 Temperature dependence of magnetic susceptibility of a-Fe203 of different crystallite sizes. (From J. Phys. SOC.Japan, 17, supplement B-1, 690 (1962). copyright Physical Society of Japan).

temperature, thermal fluctuation randomizes the spin until the Curie temperature is reached when the spins are totally random that the response of the material is like that of a paramagnet. Table 3-4 lists the magnetic behavior of some common transition metal oxides. The behavior of small oxide crystallites may differ from the bulk behavior described above because of the large contribution from ions at the surface. The different magnetic and electric fields experienced by ions in the surface may result in canting and pinning of surface spins. The surface ions may have different atomic spins than the bulk ions if they are in a different oxidation state. Adsorbates may also affect the magnetic behavior. The magnetic susceptibilities of a number of antiferromagnetic oxides, 01Fe2O3, NiO and COO have been measured at different temperatures for different crystallite sizes [34-391. Fig. 3-3 shows the data for a-Fe203. The large crystallites (> 14 nm) behave like typical antiferromagnetic materials. The magnitude of the magnetic susceptibility is very small, and it decreases slowly with increasing temperature up to about 260K. For the small crystallites, the magnetic Susceptibility first increases and then decreases with increasing temperature. This is a typical behavior for superparamagnetic particles. Furthermore, the magnitude of thc low temperature susceptibility increases with decreasing crystallite size. This is also demonstrated in Fig. 3-4 which shows the magnetization as a function of applied field for small crystallites of a-F@03 supported on Si02 [401. The

40

o” 15 N Y

c W

5a 5 I

0 10

20

MAGNETIC

30

FIELD

40

KCe

50

Figure 3-4Magnetization at 1.7K as a function of applied magnetic field strength: a. 2.5 nm a-FezO,/SiOz; b. 7.5 m a-FezQ/SiOz; c. 9.5 nm a-Fe,Q/Si02; d. 14.5 m aFq@/SiOz; e. 25 m a-Fe2Q. (From J. Phys. Chem., 88, 2525 (1984). copyright American Chemical Society).

magnetization increases with decreasing crystallite size. The shape of the magnetization curves follows the behavior for superparamagnetic particles, and can be fitted satisfactorily with the classical Langevin equation:

(3-10)

where M is the instantaneous magnetization, which is the value at the applied field, Ms is the saturation magnetization, which is the value at an infinitely large applied field, p is the magnetic moment per particle, H is the applied field, k is the Boltzmann constant, and T is the temperature. A quantitative theory to explain the increase in magnetic susceptibility with decreasing particle size is not available. It has been suggested by Nekl [41] that spins which lie near the surface tend to orient parallel to the surface. For small particles, the number of surface spins is a large fraction of the total number of spins. Since the surface spins have different orientations from those of the bulk spins, cancellation of the antiparallel spins is no longer perfect. As the particle size decreases, the contribution of the surface relative to the bulk increases. Thus the magnetic susceptibility increases. Let us return to the temperature variation of the magnetic susceptibility for large particles of a-Fe203 in Fig. 3-3. These particles behave like typical antiferromagnetic material until a temperature of about 250 K when the susceptibility suddenly increases before it decreases again on further increase in temperature. The temperature at which this phenomenon occurs is called the

PHYSICAL AND ELECTRONIC PROPERTIES

41

Table 3-5 Morin Transition Temperature and Lattice Dilation of a-Fe203 (From J. Phys. SOC. Japan, 24, 23 (1968), copyright Physical Society of Japan).

cr-Fe203 Particle Size (nm) 80 26 20 17 14

Transition Temp TM (K) 245 227 120

Lattice Constants a (1) c (1) 5.0345 5.0365fD.0007 5.0370fD.0007 5.03739.0007 5.04049.0 1

13.749 13.7901tO.OOO7 13.79020.0007 13.805+0.0007 13.83920.01

Morin transition temperature. The increase in susceptibility for a-Fe203just above this transition temperature is due to the flipping of the magnetically ordered spin from an orientation parallel to the c-axis to parallel to the c-plane (basal plane) of the crystal. The crystal then changes from antiferromagnetic to weakly ferrimagnetic. For small a-Fe203 crystallites, however, this transition appears to be absent. This effect has been confirmed by magnetization measurements [35,36,391 such as those of Fig. 3-3, and by Mossbauer spectroscopy [42,43] which makes use of the fact that on going from the weakly ferrimagnetic to the antiferromagnetic state, there is a decrease in the magnetic dipolar field and a decrease in the magnitude of the magnetic hyperfine splitting in the Mossbauer spectrum. From these measurements, it has been found that the Morin transition temperature decreases with decreasing crystallite size until it disappears for very small sizes (see Table 3-5). The critical size for the disappearance of the transition temperature has not been firmly established owing to the difficulties in accurate size determination. For a-F%03, it is estimated to be about 20 nm. One explanation for this behavior is that the magnetically coupled spins in small crystallites fluctuate in the basal plane strongly enough that they are no longer oriented to the c-axis even at low temperatures. Another explanation is that the spins are pinned at the surface. For small crystallites in which the surface spins dominate, the magnetization vector does not change direction when the temperature is changed. Concurrent with the lowering and disappearance of the Morin transition temperature, the lattice of small a-Fe203 crystallites is found to dilate [39,43]. The extent of dilation increases with decreasing crystallite size as is shown in Table 3-5. The lattice dilation is probably homogeneous [39,43-451, and results in clearly observable shifts in the x-ray diffraction peaks. This dilation along the a-axis may contribute significantly to the lowering of the Morin transition temperature through the change in the dipolar magnetic field. It should be noted,

42

Table 3-6 Crystallite Size Dependence of Magnetic Hyperfine Field Compound Temp (K)

y-Fe203a

77

a-Fe2O3b

296 83 296 80

Crystallite size (nm) 9.3 sphere 17.5 sphere 30.0 sphere 800 length acicular 6/1 shape ratio

bulk bulk 18 18

Hyperfine field (KOe) A site B site 483 489 497 508

508 513 517 525

518 542 503 527

Footnotes: a) From K. Haneda. and A. Momsh, Phys. Lett.. 64A, 259 (1977). b, From W. Kiindig, H. Bommel, G. Constabaris. and R. Lindquist, Phys. Rev., 142, 327 (1966).

however, that this phenomenon of lattice dilation may not be general, and is not observed for y-Fe203 over the crystallite size range from 9.3 nm to over a few hundred nm [46]. Another effect observed in small magnetic crystallites is the decrease of magnetic hyperfine field with decreasing crystallite size. The magnetic hyperfine field is the magnetic field experienced by the nuclei in an oxide particle. The nuclear magnetic moment interacts with this field, and the interaction can be detected by various spectroscopic techniques, especially Mossbauer spectroscopy. Table 3-6 lists some illustrative data for a- and y-Fe203. There have been three proposals as to the origin of this decrease. The first one, based on a decrease in the Curie or Ned temperature with decreasing crystallite size can now be discounted. Magnetization measurements on small crystallites do not detect changes in these temperatures. The second interpretation assumes that surface ions have a smaller hyperfine field than the ions in the bulk. This effect has now been discounted also by Haneda and Monish [471, at least for y-FezO3. Haneda and Momsh have prepared y-Fe03 samples with a surface coating enriched in s7Fe. The magnetic hyperfine field of this enriched sample is found to compare well with the samples without enrichment. This suggests that the magnetic hyperfine field of the ions at the surface is the same as the bulk ions. The third interpretation is that of Mdrup and Topsde [481 who suggest that at temperatures below the superparamagnetic blocking temperature, thermally excited oscillations of the magnetization about an energy minimum reduce the

PHYSICAL AND ELECTRONIC PROPERTIES

43

average magnetization and thus the magnetic hyperfine field. The amplitude of the thermal oscillation depends on the magnitude of the energy barrier for flipping of the magnetization vector from one easy direction to another, which is expressed as a product of the anisotropy constant K and the volume of the crystallite. For small oscillation amplitudes, the hyperline field Hhr(V,T) for a crystallite of volume V at temperature T is:

where k is the Boltzmann constant. The theory predicts that Hhf decreases linearly with increasing temperature, and the rate of decrcase depends inversely on the crystallite size. Furthermore, Hhf is independent of the crystallite size at 0 K. These predictions have been separately confirmed for Fe304 [48] and 'y-Fe203 crystallites [47]. Although the magnetic hyperfine field of the surface ions is the same as the bulk ions, there is evidence that under an applied magnetic field, the canting angle of the surface spins is substantially different from that of the bulk ions. This is the origin of the much enhanced 2,5 peak in the Mossbauer spectrum of y-Fez03 [49]. Together with the variation in magnetic hyperfine field, the anisotropy constant also depends on the crystallite size. Table 3-7 shows that the anisotropy constant K decreases sharply from bulk material to 10 nm size crystallites. As yet the origin of this decrease is not known. The effect of such a decrease is that the rate of decrease of the energy barrier for the flipping of the magnetization vector in a particle and the temperature at which the particle becomes superparamagnetic (the superparamagnetic blocking temperature) decrease faster than the volume V of the crystallite, since both of these are proportional to KV.

3.4 QUADRUPOLE SPLITTINGS OF SURFACE IONS It has been observed in Mossbauer spectra of small iron oxide crystallites that the quadrupole splittings of surface ions are different from the bulk ions. This is not surprising because the quadrupole splittings are determined by the electric field gradient at the nucleus, and it is expected that such a gradient at the surface is different from in the bulk. Fig. 3-5 shows the behavior of small iron oxide crystallites supported on SiOz [50]. In the fully oxidized form, the spectrum shows a superparamagnetic doublet due to Fe3+ ions (curve a). After reduction (curve b), the spectrum consists of two sets of doublets due to Fe2+ ions. The inner doublet shows a quadrupole splitting of 0.91 mm s-l, and the outer doublet, a quadrupole splitting of 1.74 mm spl. It is common to assign the outer doublet to Fez+ ions of higher coordination such as those octahedrally coordinated by oxygen ions in the bulk, and the inner doublet to Fez+ of lower coordination, such as those in the surface [50521. The assignment is supported by the fact that the inner doublet can be converted to the outer doublet by adsorption of NH3, CH30H [51], pyridine

44

Table 3-7 Dependence of Anisotropy Constant K on Crystallite Size

Compound

Fe30

4

?I-Fe203

K erg/cm3

Crystallite size (nm)

6 10 12 large

1.2-1.4 x lo6 0.85-0.95 x lo6 0.85-0.90 x lo6 1.1 x 103

6 12 95.5 single domain powder epitaxially grown

cx-F%03

Ref.

10 12 13-18'

8 x lo5 3 x 105 1.2 x 106 3 104 4.6 x 104 3.5 x 105 5-6 x 105 4-7 x 104

f g h.i

* These values may not be accurate, see ref. f. References: a) L. Bickford, J. Brownlow, and R. Penoyer. Proc. Inst. Elect. Eng.. B104, 238 (1957). b) M. Boudart, A. Delbouille, J. Dumesic, S. Khammouma, and H. Topsde, J. Catal., 37,486 (1975). c) J. Coey, and D. Khalafalla, Phys. Stat. Sol. (A), 11, 229 (1972). d) A. Morrish, and E. Valtyn, I. Phys. SOC.Jpn.. 17, suppl. B1, 392. (1962). e) H. Takei, and S , Chiba. J. Phys. SOC.Jpn. 21, 1255 (1966). f) J. Amelse, K. Arcuri, J. Butt, R. Matyi, L. Schwartz, A. Shaxpiro, J. Phys. Chem., 85, 708 (1981). g) S . Mdrup, and H. Tops&, Appl. Phys., 11, 63 (1976). h) W. Kiindig, H. Bommel, G. Constabaris, and R. Lindquist, Phys. Rev., 142,327 (1966). i ) W. Kiindig, K. Ando, R. Lindquist, and G. Constabaris, Czech. J. Phys. B17, 467 (1967).

[50],NO [52], CO [53], and H 2 0 [54]. It is apparent from these examples that detecting the changes in quadrupole splittings of surface ions on adsorption of molecules is one method to investigate the nature of the adsorption sites.

45

PHYSICAL AND ELECTRONIC PROPERTIES

-4

-2

0 2 VELOCITY ( r n r n l s )

4

Figure 3-5 Room temperature Mossbauer spectra for 0.5 wt.% Fe/SiOz. a. After reduction in H, at 673K followed by oxidation in 0 2 at 423K; b. After reduction in H2 at 498K; c. After reduction in Hz at 673K. (From J. Catal., 101, 103 (1986), copyright Academic Press).

It might be expected that the electric field gradient due to the iorricity of the oxide matrix is larger at the surface than in the bulk, and would lead to a larger quadrupole splitting for the surface Fez+ ions than the bulk ions, which is contrary to what has been observed. This is due to the fact that the couloumbic field from the ions in the oxide matrix is only one contribution to the electric field at the nucleus. The smaller quadrupole splitting for surface Fez+ ions than bulk Fez+ ions has been interpreted as due to the opposing effect of larger crystal field gradicnt and smaller d electron field gradient at the surface than thc bulk [52,56,57]. Inner and outer doublcts of Fez+ ions have been observed on many samples of silica-supported small crystallites of reduced iron oxide, although thc magnitude of the quadrupole splitting differs somewhat from sample to sample. The prcsencc of inner and outer doublcts has also been observed using MgO, yA1203, and Ti02 as supports [%I. Clearly distinguished inncr and outer doublets of Fe3+ ions have not bccn rcportcd. However, it is well established that the average quadrupole splitting of

46

Table 3-8 Crystallite Size Dependence of Quadrupole Splitting in a-Fe203at 298K. (From Phys. Rev., 142, 327 (1966), copyright American Physical Society).

Size (nm) < 10 13.5 15.0 18.0

Q.S. mm s-1 0.98 0.57 0.55 0.44

Fe3+ in a crystallite increases with decreasing crystallite size, as is shown in Table 3-8. This increase has been explained by the larger asymmetry in the environment of surface ions than bulk ions. Some researchers have attempted to fit their Mossbauer spectra of small crystallites containing Fe3+ ions using two sets of peaks of different quadrupole splittings. The set with a larger quadrupole splitting is assigned to the surface ions [56,58], which is supported by the observed decrease in the quadrupole splitting on adsorption of water, methanol, and ammonia [59].

3.5 SURFACE ELECTRONIC STRUCTURE Knowledge of the detailed surface electronic structure is critical in the understanding of the chemisorptive and catalytic properties of oxides. The bulk electronic structures have been studied for a number of oxides [60]. Studies on surfaces have been concentrated on the detection of electronic states not found in the bulk that are generated by the presence of the surface. These states are called surface states. The energy of the surface states of interest in catalysis commonly lies near the Fermi energy of the solid (or valence states). These surface states are either partially filled with electrons so that they can both donate and accept electrons from the molecules interacting with the surface, or close enough to the filled valence band (or empty conduction band) such that together, they provide a pair of states to accept and donate electrons to the interacting molecules simultaneously. Consideration of electron bands in the solid is also important in photo-enhanced adsorption, desorption, and catalysis. These photo-assisted processes as well as a short description of electronic bands of oxides will be discussed in Chapter 14. The readers are referred to standard textbooks in solid state physics for more in-depth discussions. The nature and energy of any surface state depend on, among other factors, the ionicity of the oxides and the position of the ions. The major contributions to these factors are from ions in the surface region. Therefore, it is expected that surface structural rearrangements could lead to substantial changes in the surfxe electronic structure [61,62].

PHYSICAL AND ELECTRONIC PROPERTIES

47

Another effect expected for ionic surfaces is the reduction of the ionic charges of the surface ions. This results from the downward shift in energy of the surface orbitals compared to the corresponding bulk orbitals and the enhanced covalency of the cation-anion bond, as well as the polarization of the surface orbitals by the surface electric field. The surface-induced changes in the electronic properties are strongly localized in the outer few layers of the surface regions. For oxides that are predominantly covalent, the surface coordinative unsaturation results in the formation of valence orbitals projecting from the surface that are from the outermost ions (sometimes being referred to as dangling bonds) [63]. There are two sets of such orbitals. One set is empty and is located mostly at the cation, and the other is filled and is located at the anion. The separation in energy of these two sets increases with increasing ionicity of the solid. The electronic structures of some oxides have been studied by electron spectroscopies (UPS and ELS). For surfaces of oxides with empty d bands, such as stoichiometric Ti02 (loo), (001) and (110), SrTiO3 (100) [l-31 and ZnO [@I, no intrinsic surface states are identified within the bandgap. Similar results have been observed for W 0 3 [65] and Moo3 (010) [66] surfaces. The surface states of these oxides are close to the conduction or valence bands. For example, the surface states of ZnO are at -2 eV and -5 eV below the top of the valence band 185,861. Loss of surface oxygen ions in these transition metal oxides with empty dbands causes development of filled electronic states in the band gap. These states are detected by U P S as emission above the valence band emission (see Fig. 3-6) [67,68]. It is interesting to note that the intensity of this bandgap emission decreases when the oxide is annealed in vacuum. The decrease is due to reoxidation of the surface by oxygen diffusion from the bulk which is taking place faster than reduction by thermal desorption of oxygen. In a similar manner, this band gap state can be depopulated by adsorption of O2 [65,67,69,70]. For oxides with empty d bands, the bandgap surface state has been attributed to the cations at a lower oxidation state than the bulk, such as Ti3+ for Ti02 and SrTi03. Oxides with partially filled d-bands, such as Ti2O3, V2O3. a-Fe203 [70], T i 0 [711, MnO [72], FeO [72,73], COO [72,74], NiO [72,75], and CuO [76] have been investigated. The d states of the cations overlap significantly with the 0 2p bands. This makes it difficult to detect the presence of any surface states. Deconvolution of the experimental spectra shows that the bands are derived primarily from the metal 3d and oxygen 2p levels. The density of states of the 3d band shows two maxima near the Fermi level, and becomes more pronounced as the number of d electrons is increased. The absence of large differences between the U P S spectra of the surfaces of these compounds and the bulk band structures implies that their surface electronic structures do not differ significantly from the bulk. The effect of surface roughness has been investigated on Ti02 [77,78]. The surface states on the atomically rough surface are found to lie deeper in the bandgap than on the smooth surface. This could result in enhanced reactivity of the rough surface as compared to the smooth surface. Most of the studies todate concentrate on studying the surface electronic structures of the oxides. Only a few have been reported to study the interaction of molecules with these surfaces. Therefore, there is very limited knowledge of the

48

-1 2

-8

-4

Ef=O

Electron Binding Energy

eV

Figure 3-6 UPS spectra of the clean ordered SrTiQ (1 11) surfaces taken at: a. 27; b. 300; c. 6000C. (From Phys. Rev. B.. 17. 4942 (1978). copyright American Physical Society).

electronic structure of the surface-molecule bonds. The formation of such bonds is essential in surface chemistry and catalysis, and it requires that the molecule can physically approach the adsorption site (steric requirement), and surface orbitals of the appropriate energies and symmetry are available. Since information on the surface orbitals are not readily available, researchers in this field have made significant use of the steric requirement. This has led to the concept of coordinative unsaturation of surface ions, which will be discussed in the next chapter.

3.6 SURFACE VIBRATION The lattice vibration of surface atoms which is called the surface optical phonon vibration modes of some transition metal oxides have been studied using high resolution electron energy loss spectroscopy (HREELS). These oxides include NiO, TiOz,ZnO, and SrTi03,all of which exhibit phonon modes that strongly couple with the incident electron beam. The strong coupling makes it readily possible to detect not only the fundamental phonon modes but also the overtones. Fig. 3-7 shows a HREEL spectrum of a ZnO (10iO)surface which possesses a fundamental phonon mode at 68.8 meV [79]. Fig. 3-8 shows that for a NiO (001) surface whose fundamental phonon mode is at 69.5 meV [80]. The overtones in both spectra arc clearly visible. For the large ordered surfaces studied, the surface phonon vibrations generally lie in the range of 40-100 meV (about 320-800 cm-’), and the strong

PHYSICAL AND ELECTRONIC PROPERTIES

49

127 K

286 K

h Energy Loss

meV

Figure 3-7 Energy-loss sepctrum of 7.5-eV electlons after specular reflection from the (1070) surface of ZnO. (From Phys. Rev. Lett., 24, 1416 (1970), copyright American Physical Society).

overtones are in the region where vibrations involving surface adsorbates appear. This makes it difficult to identify adsorbates. The frequencies of the surface phonon modes can be described by classical analyses that apply well to cases where the vibrational wavelength is much larger than lattice parameters @I]. It has been shown that the surface phonons that are excited most strongly in HREELS have small wavevectors parallel to the surface and long characteristic penetration depths [82]. Thus the energy loss spectra are rather insensitive to details of the surface structure. Small differences are observed in the phonon frequencies of different ZnO surfaces. The observed values are 68.8 meV for the (lOi0) surface, and 67.3 meV for Lhe (0001) and (000i)surfaces [79]. These values are higher in energy than the infrared-active optical bulk modes for which the displacement of ions is pcrpcndicular to the surface. They agree well with the frequencies calculated from the classical treatment of ionic vibrations in the continuum approximation. The fundamental modcs of the NiO (100) and (111) surfaces are at about 69.5 mcV [80]. For the SrTi03 (100) surface, surface optical phonon modes at 57 meV (460 cm-') and 92 meV (740 cm-') have been observed, which are just below the bulk longitudinal modcs at 474 and 788 cm-' [831. Recently, a Fourier transform technique has been developed to remove the combination and overtone structures in HREEL spectra and retain only the loss slructure of the fundamental modes [84]. Using this technique, the energy loss spcclrum of H 2 0 adsorbed on a SrTi03 (100) surface has been observed.

50

-200

0

200

Energy Loss

400 meV

Figure 3-8 Specular electron energy loss spectrum of a NiO (001) surface. (From Surf. Sci.. 152/153, 784 (1985). copyright Elsevier Scientific Publ.).

REFERENCES 1. R. A. Powell, and W. E. Spicer, Phys. Rev. B. 13, 2601 (1976). 2. V. E. Henrich. G . Dresselhaus, and H. J. Ziegler, J . Vuc. Sci. Technof.. 15, 534 (1978); Phys. Rev. B. 17, 4908 (1978). 3. V. E. Hcnrich, and R. L. Kurtz, J . Vac. Sci. Technof., 18, 416 (1981); Phys. Rev. B, 23, 6280 (1981). 4. W. Gopel. J . Vuc. Sci. Technof., 15, 1298 (1978). 5. L. E. Firment, and A. Ferretti, Surface Sci.. 129 155 (1983). 6. C. Angeletti, A. Chino. V. Indovina. F. Pepe, and M. Schiavcllo, &it. Phys. Chern., N e w Fofge. 122. 237 (1980). 7. J. Haber, in Surface Properties and Catalysis by Nonmetals”. edited by J. P. Brunelle, B. Delmon. and E. Deronane, D. Reidel Publishing Company, 1983, p. 1. 8. M. C. Kung and H. H. Kung, Surface Sci., 104, 253 (1981). 9. Y. M. Gross, and D. R. Pyke, J . Catalysis, 58, 61 (1979). 10. S. Y.Liu, and H. H. Kung. Surfuce Sci., 110, 504 (1981). 11. D. van Oeffclen, J. van Hoeff, and 0 . Schuit, J . Cutuf., 95,84 (1985); I. Matsurra, R. Schut. and K. Hirakawa, J . Catul., 63, 152 (1980). 12. S. Sasaki, K. Fujino, and Y. Takeuchi. Proc. Japun Acad., 55 ser R, 43 (1979). 13. S. Sasaki, K. Fujino, Y. Takeuchi, and R. Sadanaga, Actu Cryst., A36, 904 (1980). 14. G. Vidal-Valata, J. P. Vidal, and K. Kurki-Suonio, Actu Crysf., A34, 594 (1978).

PHYSICAL AND ELECTRONIC PROPERTIES 15. M. Morinaga, and J. B. Cohen. Acta Cryst. A32. 387 (1976). 16. J. Q. Broughton, and P. S . Bagus, J . Electron Spectro. Related Phenom., 20, 261 (1980). 17. K. Sieghban, C. Nordling, A. Fahlman, R. Nordberg, K. Hamrin, J. Hedman. G. Johannson, T. Bergmark, S. E. Karlsson, I. Lundgren, and B. Lingberg, "ESCA: Atomic, Molecular, and Solid State Structure Studied by Means of Electron Spectroscopy". Nova Acta Regiae SOC.Sci., Upsalieusis Ser. IV, 1967. 18. P. H. Citrin, and T. D. Thomas, J. Chem. Phys., 57, 4446 (1970). 19. J. Q. Broughton, and P. S. Bagus. J. Electron Spectro. Related Phenom. 20, 127 (1980). 20. F. Bechstedt, Phys. Stat. Sol. B , 91, 167 (1979). 21. S. W. Gaarenstroorn, and N. Winograd, J. Chem. Phys. 67, 3500 (1977). 22. T. Dickinson, A. F. Povey, and P. M. A. Sherwood. J . Chem. SOC. Faraday Trans. I, 72. 686 (1976). 23. L. Pauling, "The Nature of Chemical Bond", Cornell University Press, Ithaca, N.Y., 1962. 24. R. T. Sanderson, J. Amer. Chem. SOC., 74, 272 (1952); Science, 114. 670 (1951). 25. J. C. Phillips, Phys. Today, 23, 23 (1970). 26. J. P. Suchet, J. Electrochem. SOC., 124, 30C (1977). 27. F. Gervais, Solid State Commun., 18, 191 (1976). 28. B. F. Levine, Phys. Rev. B , 7, 2591 (1973). 29. B. F. Levine. J. Chem. Phys., 59, 1463 (1973). 30. H. H. Kung, J. Catal., 37, 387 (1982). 31. H. H. Kung, J. Solid State Chem.. 52, 191 (1984). 32. B. Cullity, "Introduction to Magnetic Materials", Addison-Wesley, 1972. 33. A. Monish, "The Physical Principles of Magnetism", Wiley, New York, 1965. 34. T. Takada, and N. Kawai, J. Phys. SOC.Japan, 17 supplement B-1, 691 (1962). 35. K. M. Crecr, J. Phys. SOC. Japan, 17 supplemcnt R - I , 690 (1962). 36. Y. Bando. M. Kiyama, N. Yamamoto, T. Takada, T. Shinjo, and H. Takaki, J. Phys. SOC. Japan, 20, 2086 (1965). 37. G. J. Muench, S. Arajs, and E. Matijevic, J. Appl. Phys., 52, 2493 (1981). 38. W. J. Schuclc, and V. D. Deetsneek, J. Appl. Phys., 33, 1136 (1962). 39. N. Yamamoto, J. Phys. SOC. Japan, 24, 23 (1968). 40. F. Hong, B. Yang, L. Schwartz, and H. H. Kung, J. Phys. Chem., 88, 2525 (1984). 41. L. Neil, Compt. Rend. 252, 4075; 253, 9; 253, 203 (1961). 42. W. Kundig, H. Bommel, G. Constabaris, and R. H. Lindquist, Phys. Rev., 142, 327 (1966). 43. R. C. Niningcr, Jr., and D. Schroeber. J. Phys. Chem. Solidr. 39, 137 (1978). 44. D. Schroeber, and R. Niningcr, Jr., Phys. Rev. Lett., 19, 632 (1967); D. Schrocber, Phys. Leff., 27A. 507 (1968). 45. S. W. Marshall. and R. M. Wilenikk, Phys. Rev. Left, 16. 219 (1966). 46. K. Haneda, and A. H. Momsh, Solid Sfate Commun., 22, 779 (1977). 47. K. Haneda, and A. H. Momsh, Phys. Lett., 64.4, 259 (1977). 48. S. M q h p , and H. Topsde, Appl. Phys., 11, 63 (1976). 49. A. H. Momsh, K. Hancda, and 1'. J. Schurer. J. Phys. Collngue C6, supplement 12. C6-301 (1976).

51

50. G. Connell. and J. Dumesic, J. Catal., 101, 103 (1986). 51. M. Hobson. Jr., and H. Gager, J. Colloid Interface Sci., 34, 357 (1970). 52. S. Yuen. Y. Chen, J. E. Kubsh, J. A. Dumesic, N. Topsde, and H. Tops+, J . Phys. Chem.. 86, 3022 (1982). 53. B. Claussen, S. M$rup, and H. Topsde. Surf Sci., 106. 438 (1981). 54. H. Gager. J. Lefelhocz, and M. Hobson, Jr.. Chem. Phys. Lett., 23, 386 (1973). 55. G. Connell, and J. Dumesic. J. Catal.. 102, 216 (1986). 56. G. B. Raupp, and W. N. Delgass, J. Cafal., 58. 337 (1979). 57. R. Ingalls, Phys. Rev., 133, 787 (1964). 58. W. Kiindig, K. J. Ando, R. H. Lindquist, and G. Constabaris. Czech. J. Phys., B17,467 (1967). 59. H. M. Gager. M. C. Hobson, and J. F. Lefelhocz. Chem. Phys. Lett., 15, 124 (1972). 60. D. W. Bullett, in "Surface Properties and Catalysis by Nonmetals", edited by J. P. Bonnelle, B. Delmon and E. Derouane, D. Reidel Publishing Company, 1983, p. 47. 61. M. Tsukada, H. Adachi, and C. Satoko, Prog. Surface Sci., 14. 113 (1983). 62. I. Ivanov, and J. Pollmann, Phys. Rev. E, 24, 7275 (1981). 63. M. Tsukada. and T. Hoshino, J. Phys. SOC. Japan, 51. 2562 (1982). 64. W. Gopel, J. Pollmann, I. Ivonav, and B. Riehl, Phys. Rev. E, 26, 3144 (1982). 65. R. D. Bringans. H. Hochst. and H. R. Shank, Phys. Rev. E , 24. 3481 (1981). 66. L. E. Firmenf and A. Ferretti, Surface Sci., 129, 155 (1983). 67. V. E. Henrich, h o g . Surface Sci., 9, 143 (1979). 68. V. E. Henrich, Prog. Surface Sci., 14, 175 (1983). 69. M. A. Langell, and S. L. Bernasek, Prog. Surface Sci., 9, 165 (1979); Phys. Rev. E , 23, 1584 (1981). 70. S. Ferrer, and G . A. Somorjai, Surface Sci., 94.41 (1980); J . Appl. Phys., 52, 4792 (1981). 71. V. Henrich. H. Zeiger, andT. Reed. Phys. Rev. E , 19. 4121 (1978). 72. D. Eastman, and J. Freeouf, Phys. Rev. Lett.. 34, 395 (1975). 73. C. Brundle, Surface Sci., 66, 581 (1977). 74. Y. Jugnet. and T. Duc, J. Phys. Chem. Solid, 40,29 (1979). 75. G. Wertheim, and S. Hufner, Phys. Rev. Lett., 28, 1028 (1972). 76. C. Benndorf. H. Caw. B. Egert, H. Seidel. and F. Thieme, J. Electron Spect. Rel. Phenom., 19, 77 (1980). 77. R. B. Kasowski, and R. H. Tait, Phys. Rev. E . 20. 5168 (1979). 78. M. Tsukada, C. Satoko. and H. Adachi, J . Phys. SOC.Japan, 47, 1610 (1979). 79. H. Ibach, Phys. Rev. Lett., 24, 1416 (1970). 80. P. Cox, and A. Williams, Surface Sci., 1521153, 791 (1985). 81. R. Fuchs, and K. Klicwer. Phys. Rev., 6, A2076 (1965). 82. H. Liith, Festk&perprobleme. 21, 117 (1981). 83. A. Baden, P. Cox, R. Egdell, A. Orchard, and R. Willmcr. J. Phys. C , 14, L1081 (1981). 84. P. Cox, W. Flavell, A. Williams, and R. Egdell. Surface Sci., 1521153, 784 (1985). 85. R. DOm, H. Luth. and M. Ruchel, Phys. Rev., B16, 4675 (1977). 86. W. Gopcl. J. Pollmann, I. Ivanov, and €3. Reihl, Phys. Rev., B26, 3144 (1982).

Chapter 4 SURFACE COORDINATIVE UNSATURATION

4.1 FORMATION OF SURFACE COORDINATIVE UNSATURATION It is generally accepted that surface coordinative unsaturation is important in surface chemistry. This concept is analogous to that in coordination chemistry and arises from the fact that because of steric and electronic reasons, only a limited number of ligands or nearest neighbors can be within bonding distance of a metal atom or ion. In most transition metal oxides, the oxygen anions in the bulk form close-packed laycrs and the metal cations occupy holes among the anions as described in Chapter 2. In this picture, since the bulk oxide ions are as densely packed as possible, the oxide ion ligands around a metal cation are thought to have saturated the coordination sphere of the bulk cation, that is, the bulk cation is coordinatively saturated. We have seen in Chapter 2 that in the formation of a surface by cleaving an oxide crystal, metal-oxygen bonds have to be broken. Therefore the surface anions and cations have fewer numbcrs of nearest neighbors than the corresponding ions in the bulk. These surface anions and cations are coordinatively unsaturated (cus). In most instances, coordinative unsaturation results in ions that are active in bonding with adsorbates. However, as will be discussed later, not all surface coordinativcly unsaturated ions are necessarily active and have a high tendency to form chemical bonds with adsorbates. There are two approachcs to picture the formation of surface coordinative unsaturation depending on the way the surface is prepared. The fist approach applies to surfaces of microcrystalline samples prepared from aqueous solutions. The second applies to surfaces formcd from cleaving large single crystal samples. An oxide or hydrous oxidc sample prepared by precipitation from an aqueous solution is formed by condensation and polymerization of hydroxylated mctal ions. For example, a M(II1) metal ion existing as a monomeric unit possesses a saturated coordination of three hydroxyl and three watcr ligands. Condensation of two

53

54

hydroxyl groups from two different monomeric units links the monomers with a bridging oxygen to form a dimer. Condensation of more than one pair of hydroxyl is possible and the dimeric unit has more than one bridging oxygen ion. Bridging by hydroxyl ions is also possible. These units are schematically shown below:

Monomer

Dimers

Coordinatively unsaturated sites

Hydroxylated Surface

Partially Dehydroxylated Surface

When condensation occurs between many diffcrcnt monomeric units, which

is orten promoted by drying at elevated temperatures, a three-dimensional network is formed. Depending on the metal and the drying condition, the network may be A possible configuration of a amorphous, semicrystalline, or crystalline. hydroxylatcd surface of a network is shown above. In this schematic drawing, thc number of ligands around a metal ion is set to remain at six, but the number of ligands for the oxygen ions increases to four in the bulk and three in the surface. Other situations are possible, and the coordination number of the metal ions may change as the solid is formed. Dchydroxylation of this surface may take place between two adjacent hydroxyl groups, and the surface mctal ions involved would be coordinated to only five oxygen ions. They bccome coordinatively unsaturated. The surface oxide ions generated are also coordinativcly unsaturatcd and have lower coordination numbers than bulk oxide ions or oxide ions of the

COORDINATIVE UNSATURATION

55

surface hydroxyls. Often the M"+(cus) site bchavcs like a Lewis acid, and the 02-(cus) ion is more basic than the bulk ions. Such an acid-base pair site participates in hcterolytic dissociative adsorption. More extensive coordinative unsaturation is possible if more dchydroxylation takes place, or if surface lattice oxygen is rcmoved by reduction. Such a process of dchydroxylation has been used by Burwell, et al. [2] to describe the formation of a partially dchydroxylated chromia surface which has surface ions that are coordinatively unsaturated by one or two ligands. If dehydroxylation occurs randomly bctwccn pairs of hydroxyl groups, it would be difficult to achieve complete removal of surface hydroxyl groups. Evcn after high temperature evacuations, isolatcd surface hydroxyl groups are often present that show a sharp IR absorption band at high frequencies (?3700 cm-I). As an example, IR bands of surface hydroxyl groups have bcen identified on anatase at 3636, 3654, 3672, and 3707 cm-'. The intcnsitics of thcse bands decrease with increasing evacuation tcrnperature. The rate of decrease is the fastcst for the lowest frcquency band, whilc thc highest frcqucncy band is almost unchanged [ 11. The second approach to picture the formation of surface coordinative unsaturation is by clcaving a single crystal. Commonly one s w t s with an clcctrically ncutral singlc crystal whose faces are bound by nonpolar surfaces. If thc crystal is bound by polar faces, thcsc faces nccd to be populated with ion vacancies, hydroxyl groups, or other species so that the crystal does not posscss any pcriodic dipole momcnt. Once the crystal is dcfined, it is cleaved (conccptually) into two halves along the dcsircd direction to expose the planes of intcrcst. The common guidclincs for the partition of ions along the clcavage planc arc: (1) it rcsults in two halfcrystals that are electrically ncutral; (2) it minimizes the total amount of coordinativc unsaturation as much as possiblc; (3) it avoids the formation of ions with an unusually large dcgrce of coordinative unsaturation; and (4) it involves the brcaking of as small a numbcr of bonds as possible. Figurc 4-1 shows modcls of three ideal and ordcrcd Ti02 and Ti203 surfaces. The Ti02 surfaces would be fully oxidizcd surfaces cxccpt for the anion vacancics shown in Fig. 4-la and b. If thc anion vacancics arc fillcd by oxidc ions, thcsc surfaces can be generated theoretically by fracturing a Ti02 singlc crystal along thc dcsircd dircction, and partitioning the atoms along thc clcavagc planc equally bctwcen thc two parting faces. It should be cmphasizcd that thc modcls rcprcscnt highly ordcrcd surfaccs, and such ideal surfaccs arc not cxpcctcd to cxist ovcr large macroscopic surfaces in rcality. For cxamplc, the Ti02 (110) surfacc is pictured as one in which pcrfcct rows of oxygcn ions sit on altcrnatc rows of Ti ions in thc surface plane. Although such a configuration minimizes the total numbcr of coordinative unsaturation and thc number of bonds broken, it has low entropy. On a rcal surface, thc desire to maximize cntropy rcsults in lcss wcllordcrcd arrangcmcnts. When a crystal is fractured in a rcal cxpcriinent, it is oftcn observed that clcclrons arc emitted from thc surface for somc time. Thcsc arc callcd cxoelcctrons. In addition, oxygcn ions arc obscrvcd to be evolvcd. It is possiblc

56

Y Pe 1

-LJ Figure 4-la Model of a Ti02 (110) surface. Solid circles are Ti ions, open circles are 0 ions. One type 1 surface lattice oxygen is removed to show an anion vacancy.

Figure 4-lb Model of a Ti02 (100) surface. Shaded circles are surface lattice oxygen ions under the plane of Ti ions. Other symbols are the same as in Fig. 4-la. One type 1 oxide ion is removed to show an anion vacancy.

Figure 4-lc Model of a Ti203 (047) surface. Symbols arc the same as in Fig. 4-lb.

COORDINATIVE UNSATURATION

57

that neutral atoms are also evolved without being detected. This further points out the fact that an ideal perfect surface of macroscopic dimensions may be hard to obtain. An inspection of these models shows that different exposed surfaces possess ions of different degrees of coordinative unsaturation. On the Ti02 (1 10) surface, two types of Ti ions are present. Half of the Ti ions have six oxygen ion nearest neighbors and are coordinatively saturated. The other half have five oxygen ion neighbors. Similarly, there are two different types of lattice oxygen ions. Type one is bonded to two Ti cations that are six-fold coordinated, sits above the plane of the other ions, and, on Sn02 (1 10) [73], has been shown to be more easily removed than other anions. The other type is bonded to three Ti ions of six- and five-fold coordination. They are in the surface plane. The Ti02 (100) face has five-fold coordinated Ti ions, and the (047) face has four-fold coordinated Ti ions. In addition to these, ions of other types of coordinative unsaturation can be created by introducing defects on the surface. An example is shown in the Ti02 (100) surface in Fig. 4-la. An oxygen ion in the row sitting above the surface is removed, perhaps by reduction. This action exposes two Ti ions of four-fold coordination. In general, the presence of anion vacancies results in cations of a lower coordination and in a lower oxidation state. The resulting surface is commonly chemically more active.

4.2 CHEMICAL PROPERTIES OF SURFACE COORDINATIVELY

UNSATURATED SITES Coordinatively unsaturated sites are responsible for chemisorption and binding of molecules to a surface in most instances. Indeed this explains very well a variety of phenomena including poisoning of a surface, competitive adsorption, and the common requirement of heating of an oxide to activate it for chemisorption and catalysis. Activation of an oxide by heating is needed because of the strong adsorption of water on oxides. Thus an oxide surface becomes fully covered with adsorbcd water and hydroxyl groups once it is exposed to the moisture in the atmosphere. This process is shown by eq. (4-1) and (4-2):

In eq. (4-l), water is adsorbed molecularly to satisfy the coordinative unsaturation of the metal ion. In eq. (4-2), the coordinatively unsaturated cation and anion pair adsorbs a water molecule dissociatively as OH- and H+. These

58

a,

c.'

2"

0

N

r

Evacuation Temperature, C Figure 4-2 Surface hydroxyl content of a ZnO sample evacuated at various temperatures. (From J. Phys Chem., 84, 2054 (1984), copyright American Chemical Society).

surface ions become coordinatively saturated and unable to adsorb other molecules. Heating causes removal of water and formation of surface coordinative unsaturation. The extent of dehydroxylation depends on the temperature, as can be seen in Fig. 4-2 for ZnO [3], and Fig. 4-3a for Cr203 [2]. These figures are representative of most oxides. Referring to Fig. 4-2, it can be seen that water is lost continuously from ZnO when the temperature is raised. The water that is lost at low temperature are weakly adsorbed water. It is probably nondissociatively adsorbed water that is lost by a process similar to the reverse of eq. (4-l), as well as water held to the surface by hydrogen bonding. A rapid loss of water occurs around 200 to 350°C which corresponds to dehydroxylation of the surface by the reverse of eq. (4-2). Further dehydroxylation of the remaining hydroxyl groups bcyond 400°C is slow. These hydroxyl groups are probably isolated and difficult to be removed. Removal of the weakly adsorbed molecular water docs not necessarily activate the surface, especially if the water molecules are held to the surface by hydrogen-bonding. This is illustratcd in Fig. 4-3b. The amorphous Cr203 surface only becomes active in chemisorption of CO or O2 and in hydrogenation of 1-hexcne after the oxide is activated above 200"C, although some adsorbed water is already removed by this temperature (Fig. 4-3a) 121. The chemisorptivc capacity and the catalytic activity increase rapidly as the degree of dchydroxylation increases. The manner in which these quantities incrcasc dcpcnds on the molcculc and the reaction bccause different requirements of surface sites may bc involved. For example, the dependence of NH3 adsorption on Cr203 on the activation temperature is much less than for CO or 0 2 .

59

COORDINATIVE UNSATURATION

300 v,

5 -h

P,

200

D

; P,

.

0 N I

100

2

30

200

rn

40(

300

Activation Temp.

3 . !

C

Figure 4-3a Amount of water lost per C P f upon heating from 25OC and the surface area of amorphous Cr2O3 as a function of temperature of activation with Hz. Data from ref. 2.

1.2

-

300

-

/

0 0)

e0

.8

-

v,N

O

E

Q : c

cN2

0 2 J L

u.4

-

-

0 0 ,

0 0 o E

0 Activation Temp

C

Figure 4-3b Capacity for 0 2 or CO adsorption and catalytic activity for 1-hexene hydrogenation as a function of the tcmpcrature of activation of amorphous CrzO3 in H2. Data from ref. 2.

60

Surface normal

T

Figure 4-4 Adsorption of CO on a: ZnO (0001); b: ZnO (1070) surface.

There is evidence that on a surface where the coordinativc unsaturation is along a certain direction, (that is, directional valcnce orbital), a molecule covalently bonded to the surface lies roughly in the direction of thc valence orbital. The evidence is provided by CO adsorption on ZnO surfaces shown in Fig. 4-4. Using angle-resolved U P S to monitor the orientation of the adsorbate, it has been found that CO is adsorbed linearly along the direction of the surface normal of a ZnO (0001) surface and colinear with the surface valence orbital [4]. On the ZnO (1010) surface, the axis of the adsorbed CO molcculc is along the direction near the expected direction of a tetrahedral bond of a surface Zn ion [4,5]. On these two surfaces, CO are bonded to the surface Zn ions which have thrce oxygcn ion nearest neighbors instead of four as in the bulk. The situation for the 0-polar (0001) surface is less clear. The Zn ions in this surfacc have four oxygcn ion nearest neighbors. They arc coordinatively saturatcd. In one report, it is said that adsorption of CO on this surface takcs place only on cations in the surface stcp defects where coordinatively unsaturated Zn ions are exposed [4,6]. In another report, evidence is presented that CO may adsorb to form surface carbonate [721. It is further mentioned that the amount of CO adsorbcd on this surface is rcduccd if the surface is disordercd by ion-bombardment, although one might cxpcct that such treatment would gencrate more surfacc coordinativcly unsaturated cations. Furthermore, it is reported that this bchavior is oppositc to that of C 0 2 [72]. It follows that surface ions that are coordinatively unsaturatcd by more than

COORDINATIVE UNSATURATION

61

one ligand may adsorb more than one molecule. This has been observed on chromia [7,8]. When a sample of reduced chromia is exposed to a mixture I2CO and 13C0, three IR bands corresponding to Cr(12C0)2, Cr(12CO13CO) and Cr(13C0)2are observed. If each Cr ion adsorbs only one CO, only two bands for Cr12C0 and Cr13C0 are expected. Direct observation of the interaction between an adsorbate and a surface cation has been made. For example, UV-vis spcctroscopy shows that on a reduced Cr2O3/AI2O3sample, Cr ions of square pyramidal coordination are converted to a distorted octahedral coordination upon adsorption of ammonia or methanol [9]. When 13C0 is adsorbed on a Coo-MgO solid solution, the EPR signal of Co2+ shows hyperfine interaction with 13C [lo]. The Mossbauer spectrum of reduced supported iron oxide shows an inner doublet and an outer doublet of Fe2+ ions. Upon adsorption of H 2 0 [ll], pyridine [12], NO [13] or other mblecules, the inner doublet is converted to the outer doublet. Only in a few cases where surface coordinatively unsaturited oxide ions are presesnt without any associated coordinatively unsaturated metal ions. The 0polar surface of ZnO is one such case. It has been found that without accompanying M"+(cus) ions, the 02-(cus) ions are not very reactive, particularly for species that prefer acid-base pair sites. For example, binding of water on the 0polar surface is weak. Temperature programmed desorption of water shows a peak at 190 K from this surface, and a peak at 340 K from the Zn-polar surface [14]. The 0-polar surface is also less active than the Zn-polar surface in the catalytic decomposition of 2-propanol [15].

4.3 ADSORPTION OF SMALL MOLECULES Substantial information is available on the interaction of small molecules with transition metal oxides and other oxides. At present, most of the information on the molecular nature of the adsorbate has been obtained by infrared spectroscopy. Because of the strong infrared absorption by oxides in the region below about 1000 cm-', there is little information on the surface-adsorbate bond. Such data may become available as new techniques such as EXAFS and HREELS become more sophisticated, especially the latter when deconvolution techniques to rcmove surface phonon spectra are developed. In general, molecules may interact with an oxide surface in different ways: (1) Molecular (nondissociative) adsorption in which the interaction is mainly

by o-donation and/or x-bonding interaction. This can take place on a single surface coordinatively unsaturated ion. (2) Dissociative adsorption in which a molecule dissociates upon adsorption. Dissociation of H 2 0 into H+ and OH- upon adsorption is an example of hctcrolytic dissociative adsorption in which the molccule is dissociated into charged species. This type of adsorption usually requires an anion-cation coordinativcly unsaturatcd pair site. Homolytic dissociative adsorption in which ncutral species are formed may also occur but less

62

frequently. (3) Abstractive adsorption in which the adsorbate abstracts a species from the surface (often a proton). This commonly occurs on acidic oxides. If a proton is abstracted from the surface and the adsorbed species becomes cationic, the adsorbate could be held to the surface by electrostatic forces, and coordinatively unsaturated ions might not be involved. (4) Reductive adsorption in which an adsorbed molecule is oxidized while the surface is reduced. It may also be abslractive as in the case when a hydrocarbon molecule is oxidized on adsorption to a carboxylate utilizing the lattice oxygen while reducing the cation. In addition to adsorption, a surface may catalyze reactions of the adsorbate. Examples of various catalytic reactions are discussed in later chapters. Deprotonation and protonation of an adsorbate are examples of lssociative and abstractive adsorption, respectively. These are BrQnsted acid-base reactions between an adsorbate and a surface, and commonly occurs with molecules such as NH3, pyridine, and alcohols. They will be discussed in greater detail in Chapter 5. Reductive adsorption is commonly observed on transitional metal oxidcs which have cations that have readily accessible multiple oxidation statcs. This type of adsorption must be considcrcd whcn interpreting adsorption data. For example, adsorption of nitrogen-containing compounds may result in their oxidation to nitrates or nitrites, and oxidation of adsorbed hydrocarbons to carbonates or carboxylates also occurs frequently in the absence of gaseous oxygen. lat ice RCH3 l a t t i c e,0

RN02(ad)

+

2 OH(ad)

(4-3)

RC02(ad)

+

3 OH(ad)

(4-4)

Busca and Lorenzelli have summarized the infrared band frcquencies of carbonate ions, carboxylates, bicarbonates and formates of various compounds [16]. Because of the many different possiblc ways that these species can bond to a surface, there are no characteristic bands unique to each species to help thcir identification. Free carbonate ions possess two dcgcnerale IR-active asymmelric CO stretching (v3) bands at 1415 cm-’. The splilting between the two bands on adsorption may be used to identify the type of coordination of the carbonatc. If bonding is to a metal ion that has a high polarizability, the classificauon proposed by Nakamoto et a]. shown on the ncxt pagc may be uscd [171. The assignment of v3 splitting largcr than 300 cm-’ is not certain. On oxidcs of metals having low polarizing powcr, the splitting of v3 is smaller, and the distinction between a monodcntate and a bidcntatc coordination bccomcs less clear. Then consideration of thermal stability can help. A moncdcntatc structure should normally be less stable on a surface than a bidcntatc spccics [16]. Carboxylates show absorption in thc rcgion of 1550-1760 and 1150-1200 cm-’. Sometimes the observation of two bands in thcsc rcgions is confused with the v3 splitting of a carbonate species. Adsorbcd formate ions show four bands

63

COORDINATIVE UNSATURATION

Coordintation of Carbonates v3 splitting (cm-') zero 100 300

symmetric

monodentate

Type of coordination symmetric monodentate bidentate or bridged

bidentate

bridged

under optimal conditions: C-H stretching at about 2900 cm-', asymmetric C 4 stretching at 155CL1600 cm-', C-H bending at about 1400 cm-', and symmetric CO stretching at 1345-1385 cm-'. It is commonly observed that the form of adsorption of a molecule depends upon the pretreatment of the oxide. This results from different modes of adsorption requiring different surface sites. Protonation requires surface hydroxyl groups and thus depends on the extent of surface dehydroxylation. Abstractive reductive adsorption removes surface oxide ions and thus depends on the oxidation state of the surface. The presence of adsorbed oxygen could also oxidize adsorbates. As illustrated earlier, whether a molecule adsorbs nondissociatively (eq. 4-1) or dissociatively (eq. 4-2) depends on whether cation-anion coordinatively unsaturated pair sites are available. In what follows, illustrative examples of the adsorption of hydrogen, hydrocarbons, alcohols, CO and NO are presented to show the different types of adsorption. H2

Hydrogen is adsorbed on ZnO both molecularly and dissociatively. Molecular adsorption has been observed to occur at -195"C, and the adsorbed Hz shows an IR band at 4019 cm-'. This band is shifted to 3507 and 2887 cm-' for HD and DZ,respectively [18]. Dissociative adsorption is most likely heterolytic although there is no direct confirmation of it, and occurs at room temperature:

For Hz,a band assigned to Zn-H at 1712 cm-' and C t H at 3490 cm-' are observed [19]. It is interesting that the H atoms bound to Zn-H and 0-H exchange with D2 present in the gas phase to form HD in the gas phase, but not

64

1700

1650

1600 Wavenumber

cm

-1

Figure 4-5 Effect of increasing CO pressure on the Zn-H strctching band frcquencics (Pll2 = 100 Ton). (From J. Catal., 51, 160 (1978), copyright Academic I’ress).

with Zn-D or O-D [20]. This IR-active dissociative form of adsorbed H2 is involved in the catalytic hydrogenation of alkenes on ZnO. The frequency of the Zn-H band is affected by coadsorbates. When CO is adsorbed onto a hydrogen-coverd surface, two new Zn-H bands appcar whosc intensities vary syslematically with the CO coverage. This is illuslratcd in Fig. 4-5 [21]. The two new Zn-H bands represent Zn-H cornplcxes wilh one or both of its neighboring Zn ions coordinated to a CO molcculc.

Alkenes Alkenes are adsorbed both nondissociativcly and dissociatively on ZnO. Nondissociative adsorption is in the form a x-complex. This has k e n observed for C2H4 [22], propcnc [ 2 3 ] ,and butenes [24]. Thesc x-complcxes exhibit a C=C stretching band that is shifted to a lower frequency from the corrcsponding band in the gas phase by less than 50 cm-’ . These spccics are only weakly adsorbed. Propcne is also adsorbed on ZnO dissociatively to form x-allyl. The dissociation is probably heterolytic, and a x-ally1 anion is formcd [23,25]:

C ~ H + -Zn+ ~

4

CH~-CH-CH; -1-Zn

H+ I

(4-6)

COORDINATIVE UNSATURATION

65

This process is similar to H2 adsorption on ZnO. Upon adsorption of propene in this form, an 0-H infrared band is obscrvcd but no Zn-H band. Thus the x-ally1 species is adsorbed on the Zn ion. Another band at 1545 cm-' is also obscrvcd which is assigned to the antisymmctric stretch of thc C-C-C skeleton. This assignment is substantiated by observing shifts in this band position upon deutcrating the propene molecule at various positions and noting that the magnitudes of the shifts agree with those prcdictcd by group frequency analyses. It is also supportcd by thc fact that the hydrogen of the G H group formed arises from thc dlylic H of the propene molecule. This x-ally1 species can be reversibly removed from the surface on evacuation at slightly elevated temperatures. This fact strongly suggests that it is not an oxidized surface species. The relative amount of propene adsorbcd as x-ally1 and as n-complex depends on thc extcnt of dchydration of the surface [26]. Conclusivc evidence for the formation of n-ally1 from butene when it is cxposcd to ZnO has not been established. Although an IR band at about 15701580 cm-l has been observed, it cannot be readily removed by evacuation [24]. This, together with the fact that the kinetics of its formation does not appcar to correlate with the kinetics of butene isomerization, raises the possibility that this band might be due to a surface oxidized species. By applying the criterion that x-complex and x-ally1 can be readily dcsorbcd from the surface, it has also bcen claimcd that a 1600 cm-' band is due to the C=C strctch of a x-complcx of propcne on Ga-Mo oxide [27], and a 1580 cm-l band to that on cuprous oxide [28]. Bands at 1400 and 1440 cm-I have bcen assigncd to the n-ally1 species on thcsc two oxides, respectively. However, definitive assignmcnt of IR bands cannot be made readily without supplementary information from other observations, espccially in the absence of close analogs to the adsorbed spccics. Indccd, thc investigators caution that the 1440 cm-' band mentioned above may also be assigncd to deformation vibrations of the CH3 group in physically adsorbcd propene or a x-complcx. It is also intcrcsting to note that adsorption of allyl bromide on ZnO rcsults in a surface allyl species with a band at 1470 ern--' instead of 1545 cm-' which is observed on adsorption of propcnc [29]. Only x-complexes have bccn idcntificd on adsorption of ethene and propene on NiMo-MgO, Co-Mo-MgO 1301, and Ti02 [31]. They havc also been identified on adsorption of butcncs on a V-P-0 [32], where it is observed that the red-shift in the C=C stretch on adsorption depends on whether the oxide is oxidized or reduced. Thc shift is 33 cm-' on an oxidized sample, and 23 cm-' on a reduced sample. Like alkcncs, terminal alkynes may bc adsorbcd eithcr molecularly or dissociativcly [33,34]. It has bcen obscrvcd that acetylene and phcnylacctylcne are adsorbcd at room tcmpcraturc dissociativcly as acetylide (HCS-) and phcnylacctylidc, rcspectivcly, on ZnO [34]. Propyne is adsorbcd on ZnO to form both mcthylacetylide and propagyl spc5cs [33,35-391:

[ CH3X - C ]

mclhylacctylidc

ProPagYl

66

The propagyl species is the intermediate in the isomerization of propyne to allene. The formation of propagyl species on ZnO has been confirmed using XPS which shows only one carbon peak for the adsorbed species, instead of two peaks for the gas phase propyne molecule or the methylacetylide species on a silver surface [34]. Some of the dissociatively adsorbed species is further oxidized upon heating, resulting eventually in carbon oxides and water. The extent of dissociative adsorption depends on the crystallographic orientation of the surface. On ZnO, dissociative adsorption occurs on the Zn-polar surface, but only molecular adsorption occurs on the 0-polar surface [341.

Alcohob and Acetone With few exceptions, alcohols are adsorbed by heterolytic dissociation at room temperature on a dehydroxylatcd surface with a proton going to a surface lattice oxygen and an alkoxide to a surface cation. This is again similar to the other heterolytic dissociative adsorption shown in eq. (4-2), (4-3, and (4-6). The process on ZnO is shown below:

ROH

+

Z n 4 -+

RO H I I Zn- 0

(4-7)

For example, methoxide has been detected when methanol is adsorbed on ZnO [40] and Moo3 [41,42]. Similarly, ethoxide is formed from ethanol on Moo3 [43] and ZnO [44], 1-propoxide is formed from 1-propanol [44] and 2propoxide is formed from 2-propanol on ZnO [45]. Methanol has been shown to be adsorbed dissociatively at as low as 105 K [46]. Oxidation of surface alkoxide commonly occurs. The oxidation of methoxide to a surface formate is well documented on ZnO [47,48]. The zinc ion is reduced to zinc metal which can be desorbed from the surface. It also occurs on Cr2O3 [49]. 2-propoxide on ZnO is oxidized to a surface enolate which eventually is dcsorbed as acetone [45]. Adsorbed acetone dissociates heterolytically into a surface enolate species with simultaneous formation of a surface OH group. This reaction:

is observed to occur on both ZnO [45] and NiO [50]. The reversibility of this reaction has been demonstrated. Complete exchange of the hydrogen atoms of the methyl groups with deuterium has been observed when a sample of ZnO with adsorbed acetone is exposed to deuterium gas [451. Thus deuterium is adsorbed dissociatively on the oxide, and the OD group participates in the reverse reaction.

COORDINATIVE UNSATURATION

67

co Carbon monoxide is a rather common probe molecule in the study of the surface chemistry of transition metal oxides. Its behavior on these oxides is quite different from its behavior on the metals. Adsorption of CO on noble metals is much stronger than on oxides. CO is adsorbed dissociatively on transition metals at the left of the Periodic Table. Dissociative adsorption of CO on oxides has not been reported. Instead, oxidation of CO to C02 or carbonate occurs. Nondissociative adsorption of CO usually occurs on a surface coordinatively unsaturated cation. A characteristic feature of this CO is that the species has a higher infrared absorption frequency than the gas phase frequency of 2143 cm-'. A value between 2170 amd 2200 cm-' is observed [51]. In constrast, a shift to lower frequencies is observed on metals. The phenomenon is first noticed on NiO [52]. Explanations of this shift to higher frequencies on oxides have been based on the assumption that the adsorption involves an interaction between a surface cation and the carbon end of the CO molecule. This assumption has now been substantiated by single crystal studies on ZnO [4-61. In this picture, the C=O stretching frequency depends on the relative extent of a-donation from CO to the cation, which would increase the C - 0 bond strength and its frequency, versus back xbonding from the cation to the CO molecule, which has the opposite effect. It is argued that for cations in a high oxidation state, the cation size is small and its electron affinity is high. Thus the bonding with CO involves mainly a-donation and little x-bonding, which results in an increase in the CO stretching frequency. The importance of x-bonding increases when the cation is in a lower oxidation state. It has been proposed that the frequency for W'CO is above 2170 cm-', M'CO in the region of 2120-2160 cm-', and M"C0 below 2100 cm-' [53]. In addition to the molecular orbital picture of the bonding, the electric field near an oxide surface could also affect the CO stretching frequency. This effect has bccn analyzed [54,55]. It is agreed that the interaction of the surface electric field with the dipole moment of the CO molecule changes the interatomic distance in the molecule, and thus its force constant and the stretching frequency. However, quantitative predictions are not yet available. The mode of adsorption of CO depends on the pretreatment of an oxidc. On a fully oxidized surface, CO may adsorb reductively to form carbonates [57,58]. CO may also react with surface hydroxyl groups to form formate [59]. The frequency of the CO band can also be affected by coadsorbates. When CO and COz are coadsorbed on ZnO [56], the CO frequency is shifted from 2183 to 2212 cm-'. The stretching frequency also depends on the pretreatment. ZnO evacuated at 400°C followed by an oxygen treatment yields a CO band at 2212 cm-*. If the sample is evacuated at 400°C, the band is shifted to 2187 cm-' [60]. The reasons for these shifts are not well understood.

NO NO is another probe molecule commonly used to study transition metal oxides. Except for one additional unpaired electron, it has an electronic structure similar to CO. Thus in many ways its adsorption is similar to that of CO, and occurs on surface M"'(cus) ions, although its adsorption is often stronger.

68 Table 4-1 Infrared Absorption Peak Positions of Dinitrosyl or Dimeric Species of Adsorbed NO

Cation

Peak Positions, cm-'

W+,Cr3+ reduced MOO, Fe2+ co2+ reduced WO,

1745-1775, 1865-1895 1695-1713, 1800-1817 1810, 1910 1765-1795, 1840-1875 1685, 1795

Peak Separation, cm-1

120 - 130

100 100 80 - 90

110

NO may be adsorbed in one of the four different forms shown below [51]:

Adsorption as neutral species either as a single molecule or as a pair is the predominant mode observed. In a few cases, charged adsorbed species are formed. The infrared stretching frequency of adsorbed NO- is lower than the others, being in the region below 1735 cm-'. The frequency for the othcrs are mostly in the region 1740-1950cm-'. In addition to the above species, oxidation of NO occurs occasionally to form surface nitrate or nitrite species [61-631. Adsorption in pairs on one M"+(cus) ion is characteristic of NO adsorption, and it shows in infrared spectroscopy as a pair of absorption bands due to the symmetric and antisymmetric stretches of the adsorbed NO pair. The absorption frcqucncies of the pair dcpend on the oxide as well as on the particular sample of a given oxide. However, the separation between the two peaks is constant for a given cation and varies less from sample to sample. These separations are listed in Table 4-1 which is derived from the data summarized in ref. 51. The high tendcncy for NO to be adsorbed in pairs is due to the stabilization derived from the mutual interaction of the unpaired electron on each NO [64].Whether the pair exists in the form of a dimcr or dinitrosyl is still unclear.

Dinitrosyl

Dimcr

COORDINATIVE UNSATURATION

69

The arguments favoring a dinitrosyl species include the absence of absorption in the 1250-1350cm? region, which excludes the presence of a hyponitrite species the fact that the adsorbed NO pair is stable to quite a high (N202-) [&I], temperature (on Mo, W, and Cr cations), and the two nitrosyl bands have approximately the same intensity [65]. Other evidence supports a dimeric form for the adsorbed NO pair. It has been observed that the ratio of the symmetric and asymmetric stretching from a N202 dimer is very similar to the coupled nitrosyl bands on a silica-supported chromia [66]. When a reduced molybdena is first half-saturated with 15N0 and then exposed to 14N0, no infrared bands attributable to a mixed l 5 N W 4 N O complex is observed [67,68]. Since isotopic mixing is expected for dinitrosyl complexes, this observation supports a dimeric species. However, it may also be explained (though less likely) with a dinitrosyl complex if the adsorption sites differ greatly in their binding energy for NO or in their accessibility. It should be noted that the ratio of the intensities of the asymmetric and symmetric stretch of adsorbed NO pairs is related to the angle between the two M-0 bonds [69]. It has been suggested that linear M-NO groups possess infrared absorption frequencies higher than 1850 cmpl [70]. Whether a linear group is formed depends on the degree of coordinative unsaturation of the metal cation. It appears that a F e 2 + 4 0 complex is more linear if the Fe2+ ion is less coordinatively unsaturated than more unsaturated [71].

REFERENCES 1. N. D. Parkyns, in "Chemisorption and Catalysis," ed. by P. Hepple, Elsevier Publ. Co., Amsterdam, Netherland, 1971, p. 150. 2. R. L. Burwell, Jr., G . L. Haller, K. C. Taylor, and J. F. Read, Adv. Cafaf..29, 1 (1969). 3. M. Nagao. and T. Morimoto, J. Phys. Chem., 84, 2054 (1984). 4. R. R. Gay, M. H. Nadine, V. E. Henrich, H. J. Zeiger. and E. I. Solomon, J . Amer. Chem. SOC., 102. 6752 (1980). 5 . K. L. DAmico, F. R. McFeely, and E. I. Solomon, J. Amer. Chem. Soc., 105, 6380 (1983). 6. K. L. DAmico, M. Trenary, N. D. Shim, E. I. Solomon, and F. R. McFeely. J. Amer. Chem. SOC., 82, 1504 (1982). 7. A. Zecchina, E. Garrone, and E. Gulielminotti, Cdalysis (London), 6 , 90 (1983). 8. B. Rebenstorf' and R. Larsson, Z. Anorg. Allg. Chem.. 453, 127 (1979). 9. V.A. Shvets, Russ. Chem. Rev., 55, 200 (1986). 10. V. Indovina, D. Cordischi, and M. Occhiuzzi, J. Chem. SOC. Faraday Trans. I , 77,811 (1981). 11. H. M. Gager, J. F. Lefelhocz, and M. C. Hobson, Jr., Chem. Phys. Left.. 23, 386 (1973). 12. G.Connell, and J. A. Dumesic, J. Cafal.,101, 103 (1986). 13. S. Yuen. Y. Chen, J. E. Kubsh. 1. A. Dumcsic. N. Tops&. and H. Tops&. J . Phys. Chem., 86, 3022 (1982).

14. 15. 16. 17.

G . Zwicker, and K. Jacobi, Surface Sci., 131, 179 (1983). P. Berlowitz, and H. H. Kung, J. Amer. Chem. Soc.,108, 3532 (1986). G . Busca, and V. Lorenzelli, Mater. Chem., 7, 89 (1982). K. Nakamoto, J. Fugita. S. Tanaka. and M. Kobayashi, J. Amer. Chem. Soc., 79, 4904 (1957). 18. C. C. Chang, and R. J. Kokes, J. A m r . Chem. Soc., 93, 7107 (1971). 19. R. P. Eischens, W. A. Plisken, and M. J. D. Low, J. Catal., 1. 180 (1962). 20. S. Naito, H. Shimizu, E. Hagiwara, T. Onishi. and K. Tamaru,Trans. Faraday Soc., 67, 1519 (1971). 21. F. Boccuzzi, E. Garrone, A. Zecchina, A. Bossi. and M. Camia. J. Catal., 51, 160 (1978). 22. A. L. Dent, and R. J. Kokes, J. Phys. Chem., 73, 3772, 3781 (1969). 23. A. L. Dent, and R. J. Kokes, J. Amer. Chem. Soc., 92. 6709 (1970). 24. C. C. Chang, W. C. Connor, and R. J. Kokes. J. Phys. Chem., 77, 1957 (1973). 25. T.T.Nguyen, and N. Sheppard, J. Chem. Soc., Chem. Commun., 868 (1978). 26. A. A. Efremov, and A. A. Davydov, Kinet. Catal., 21. 383 (1980). 27. A. A. Davydov, J. Tichy, and A. A. Efremov. React. Kinet. Catal. Lett., 5 , 353 (1976). 28. V. G. Mikhal’chenko, V. D. Sokolovskii. A. A. Filippova, and A. A. Davydov, Kinet.Catal., 14, 1099 (1973). 29. A. A. Davydov, A. A. Efremov, V. G . Mikhal’chenko. V. D. Sokolovskii. J. Catal., 58, 1 (1979). 30. R. Grabowski, A. A. Efremov, A. A. Davydov, and E. Haber, Kinet. Catal., 22, 794 (1981). 31. A. A. Efremov, and A. A. Davydov, React. Kinet. Catal. Lett., 15, 327 (1980). 32. E. V. Rozhkova, S. V. Gorej, and Ya. B. Gorokhovatskii, Kinet. Katal., 15, 694 (1974). 33. R. J. Kokes. Intra-Science Chem. Rep., 6, 77 (1972). 34. J. M. Vohs, and M. A. Barteau, J. Phys. Chem., 91, 4766 (1987). 35. C. C. Chang, and R. J. Kokes. J. Catal., 28, 92 (1973). 36. C. C. Chang, and R. J. Kokes, J. A m r . Chem. Soc., 92, 7517 (1970). 37. J. Saussey, and J. C. Lavalley, J. Chim. Phys., 75, 506 (1978). 38. T. T.Nguyen. J. C. Lavalley, J. Saussey, and N. Sheppard, J. Catal., 61, 503 (1980). 39. T. T. Nguyen, J. Catal.. 61. 515 (1980). 40. A. Ueno, T. Onishi, and K. Tamaru,Trans. Faraday Soc., 67. 3585 (1971). 41. R. P. Groff, J. Catal., 86, 215 (1984). 42. M. Ito, Vib. Surf. (Proc. Intern. Cod.), 2nd, 1980 (published 1982), p. 71. 43. K. Aika. and J. H. Lunsford, J . Phys. Chem., 81, 1393 (1977). 44. M. Nagao, and T. Morimoto. J. Phys. Chem., 84. 2054 (1984). 45. 0. Koga, T. Onishi, and K. Tamaru,J. Chem. Soc. Faraday Trans. I , 76, 19 (1980). 46. W. Hirschwald. and D. Hoffmann, Surface Sci., 140, 415 (1984). 47. S. Akhter, W.H. Cheng, K. Lui, and H. H. Kung, J. Catal.. 85, 437 (1984); S. Akhter. K. Lui, and H. H. Kung, J . Phys. Chem., 89, 1958 (1985). 48. M. Bowker, H. Houghton. and K. C. Waugh. J. Chem. SOC. Faraday Trans. I, 77, 3023 (1981). 49. K. Yamashita, S. Naito, and K. Tamaru, J. Catal., 94, 353 (1985).

COORDINATIVE UNSATURATION 50. H. N. Rufov, A. A. Kadushin, and S. Z. Roginsky. Proc. 41h Intern. Cong. Catal.. vol. 3, 1968. 51. M. C. Kung, and H. H. Kung, Catal. Rev., 27,425 (1985). 52. R. P. Eischens, and W. A. Pliskin, Adv. Cafal.,9, 662 (1957). 53. Yu. A. Lokhov, and A. A. Davydov, Kinel. Katal., 21, 1093 (1980). 54. N. S. Hush, and M. L. Williams, J . Molec. Specfrosc..50, 349 (1974). 55. R. Larsson, R. Lykvist. and B. Rebenstorf, Z . Phys. Chem. (Leipzig). 263, 1089 (1982). 56. J. C. Lavalley, J. Saussey, and T. Rais, J . Molec. Catal., 17, 289 (1982). 57. P. G. Harrison, and E. W. White, J . Chem. SOC. Faraday Tram. I , 74, 2703 (1978). 58. E. Guglielminotti, L. Cermti, and E. Borello, Gazz. Chim. Ital., 107, 503 (1977). 59. M . He, and J. G.Eckerdt, J . Cafal.,72, 303 (1981). 60. C. H. Amberg, and D. A. Seanor, Proc. 3rd Intern. Cong. Catal., Amsterdam, 1965, p. 450. 61. P. G. Harrison, and E. W. White, J . Chem. SOC. Faraday Tram I , 74, 2703 (1978). 62. G. Busca, and V. Lorenzelli, J . Cafal.,72, 303 (1981). 63. J. W. London. and A. T. Bell, J . Catal., 31, 32 (1973). 64. A. Zecchina, E. Garrone, C. Morterra, and S. Coluccia, J . Phys. Chem., 79, 978 (1975). 65. A. Kazusaka, and R.F. Howe, J . Cafal.,63, 447 (1980). 66. E. L. Kugler. R. J. Kokes, and J. W. Gryder. J. Catal., 36, 142 (1975). 67. W. S. Millman, and W. K. Hall, J . Phys. Chem., 83, 427, (1979). 68. J. B. Peri, J . Phys. Chem., 86. 1615 (1982). 69. F. A. Cotton, and G.Wilkinson, "Advanced Inorganic Chemistry". 3rd ed., Wiley-Interscience, N.Y. 1972, p. 697. 70. J. H. Enemark. and R. D. Feltham, Coord. Chem. Rev.. 138. 339 (1974). 71. S. Yuen, Y. Chen, J. E. Kubsh. J. A. Dumesic, N. Topsde, and H. Topsde, J . Phys. Chem., 86. 3022 (1982). 72. C. Au, W. Hirsch, and W. Hirschwald, Surface Sci.. 197, 391 (1988). 73. D. F. Cox, T. B. Fryberger, and S. Semancik, Phys. Rev. B, 38. 2072 (1988).

71

Chapter 5 SURFACE ACIDITY

5.1 SURFACE ACID SITES A partially hydroxylated oxide surface has hydroxyl groups and coordinatively unsaturated metal cations and oxygen anions. Each of these species can participate in an acid-base reaction. Exposed coordinatively unsaturated cations may act as acceptors for free electron pairs of adsorbed molecules.

M"+(cus)

+

:B(g)

+ M"+:B

(5-1)

Such cations are Lewis acid sites. The strength of these acid sites depends on the charge and size of the cations, both of which may vary with the oxidation number of the cation. In general, according to the concept of hard and soft acids [ l ] , cations of a higher oxidation state are harder. For cations in the same group in the Periodic Table and of the same oxidation state, those in a later period are softer. Harder cations are smaller and less polarizable. They will adsorb or bind hard bases stronger than soft or polarizable bases. Complicating these considerations, however, is the fact that cations in oxides are usually surrounded by larger and more polarizable oxygen anions. The harder, smaller and less polarizable cations are sometimes partially shielded by the oxygen anions so that binding of molecules to the cations is sterically hindered. As a result, binding is weaker than expcctcd. Surface hydroxyl groups may act as BrBnsted acid sites. They may dissociate to protonate adsorbed bases:

The resulting conjugate acids and bases are stabilized on the surface by electrostatic interaction with each other and with the oxide. As will be discussed later, the oxide exerts an effect similar to that of an aqueous solution in slabilizing charges on the surface. 72

SURFACE ACIDITY

73

Table 5-1 Relative Acidities on Oxide Surfaces a

relative acidity order on ZnO MgO

1. 2. 3. 4.

C~HSSH HCOOH CH3COOH 1. CH3COOH HCN 5. C ~ H S O H 2. C6H50H 6. CH3SH 3. CH3SH 4. CH30H 7.fH30H C ~ H S O H 5. C ~ H S O H 9. C~HSCCH6. C~HSCCH 7. HCCH

M

acidb

PKil

kcal/mol

3.7 4.8 9.3 9.9 12.0 15.5 17.0 18.5 26

33 1.8 345.2 348.5 353.1 351.4 359.0 379.2 376.1 370.3 375.4

D@(B-H)' kcal/mol

83.3 106 105.8 123.8 86.5 90.7 104.4 104.2 132

Footnotes: a) From R. Spitz, J. Barton, M. Barteau, R. Staley, and A. sleight, J. Phys. Chem. 90, 4067 (1986), copyright American Chemical Society. b, Homolytic band dissociation energy. From J.E. Bartmess. R.T. McIver, Jr., in "Gas Phase Ion Chemistry", M.T. Bowers, ed., Academic Press, NY 1979, p. 87. ') Hcat of heterolytic dissociation. From D.F. McMillen, D.M. Golden, Ann. Rev. Phys. Chem. 33, 493 (1982).

An exposed coordinatively unsaturated oxygen ion participates in an acidbasc reaction as a conjugate Brbnsted base. It is one element of the pair sites for hctcrolytic dissociative adsorption (see eq. 4-2). The heterolytic dissociative adsorption of water, hydrogen, alcohols and alkynes described in Chapter 4 are cxamplcs where the surface oxygen ions act as conjugate bases. The dissociative adsorption of bcnzaldchydc and chloroform are other examples. In general, transition metal oxides do not have strongly basic sites. When considering the BrBnstcd acidity of surface hydroxyl groups or basicity of surface oxygen ions, it is important to know whether it is more appropriate to use the aqueous solution acidity scale or the gas phase acidity scale. The charged species are primarily stabilized by the dipoles of water molecules in an aqueous solution, and by other polarizable spccies in other solutions. In the gas phase, they arc stabilized by induction within the molecule. Thus molecules may show different strengths of acidity and basicity in different media. While it has long bccn conjectured that the aqueous solution scale is more appropriate for oxidc surfaces, this has only been demonstrated recently [2]. In this study, the adsorption of a series of organic acids that include thiols, alcohols, acids and alkyncs was studicd (see Table 5-1). The relative strength of adsorption

74

of these molecules was monitored by displacement/ titration experiments. In these experiments, one molecule was first adsorbed. After evacuation, the oxide was exposed to a second molecule. The displacement of the first molecule by the second molecule was followed with FTIR and the extent of displacement was evaluated from the IR peak intensities. Sometimes deuterated molecules were used to avoid overlapping peaks. Although this method is not an equilibrium method and quantitative values of relative acidity could not be obtained, qualitative ordering could be derived. It was shown that the titration of acids of similar strengths is completely reversible, but is irreversible for acids of very different strengths. The results of this study are shown in Table 5-1. The orders of the strength of adsorption on both ZnO and MgO are the same and follows the same order as aqueous pK, values. The order does not follow the acid strengths in the gas phase as measured by the enthalpies of heterolytic dissociation of the probe molecules, or the bond dissociation energies of these molecules. It may be concluded that the oxide must function llke an aqueous solution to stabilize the electric charge on the dissociated molecules. Although Table 5-1 shows the results of an experiment where the oxide ion of the solid serves as the conjugate base, the same phenomenon probably applies to the situation in which the surface hydroxyl groups act as Brdnsted acids, since the same charge stabilization mechanism is required. Therefore, it is probable that the order of acid strengths of surface hydroxyl groups on different oxides determined either by protonation of adsorbed molecules in a nonaqueous or gaseous environment or by the isoelectric points in aqueous solutions will be the same. However, quantitative comparisons between the two methods will probably be impossible until accurate calculations to evaluate charge stabiliiation become available. Complicating the matter is the fact that the acid strength of a hydroxyl group may depend on its environment including the presence of ion vacancies, impurities, and surface dislocation defects. The importance of surface impurities has been shown to be a critical factor in the variation of the isoclectric point of a particular oxide reported in the literature.

5.2 FORMATION OF ACID SITES Exposed coordinatively unsaturated metal cations and oxygen anions on the surface are Lewis acid and Brdnsted conjugate base sites, as described above. Brdnsted acid sites are present only when hydroxyl groups are present. Therefore the number of such sites depends on the extent of hydroxylation of the surface. As mentioned in chapter 4, hydroxylation and dehydroxylation involve the interaction between a water molecule and surface coordinatively unsaturated metal cations and oxygen anions:

SURFACE ACIDITY

75

It may seem appropriate to relate the appearance of Lewis acidity with dehydroxylation and disappearance of BrQnsted acidity, and vice versa. In some cases this can be readily demonstrated. On a ZnO sample that is fully hydroxylated, significant amounts of BrQnsted acid sites are present, as indicated by the formation of I W, ions + on adsorption of ammonia. But when the hydroxyl density is less than half of the fully hydroxylated surfaces, hardly any is observed [3]. On alumina-supported molybdena, chromia, rhenium oxide, or tungsten oxide, adsorption of pyridine results in the formation of pyridinium ions the amounts of which increases after exposing the oxides to water [4]. A similar observation has been made using the adsorption of ammonia [ 5 ] . In some other cases this interconversion may not be demonstrated so readily. This may be due to the fact that the surface oxygen anion is very strongly basic such that its protonated form is only a very weak BrQnsted acid - so weak that it protonates only strongly basic molecules. It is also possible that in the dehydroxylated form, the surface cation behaves like a very weak Lewis acid center because of its nature or because of steric shielding by the neighboring oxygen ions. Then dehydroxylation would not result in the development of readily detectable Lewis acidity. For example, anatase TiOz prepared from titanium tetraisopropoxide possesses Lewis acid sites which are not converted by the addition of water to Brdnsted acid sites strong enough to protonate pyridine. However, Lewis acid sites on anatase prepared by the hydrolysis of titanium oxide sulfate can be converted to Brdnsted acid sites on exposure to water vapor [ 6 ] . Since surface coordinatively unsaturated ions are important, it follows that the presence of cation or anion vacancies and other defects that results in greater exposure of the ions also affects acidity. However, very little understanding of this matter is available. One way to generate new and perhaps stronger acid sites on an oxide is by incorporation of a second oxide. The mixed oxide of silica-alumina is a classic example where a high density of new strong acid sites is generated. This situation is found in many other mixed oxides, as is shown in Table 5-2. Also shown in the table is the fact that not all mixed oxides develop new acid sites. Various models have been proposed to predict the formation of new acid sites in mixed oxides. Tanabe's model applies to dilute mixed oxides where a small amount of a second oxide is incorporated into the first oxide by cation substitution [7-91. This model assumes that the generation of new acid sites is caused by an excess of negative or positive charge in a model structure of a binary oxide. The model structure is constructed as follows: i) The coordination number of a cation in the component oxide is maintained in the binary oxide. ii ) The coordination number of the oxygen ion in the binary oxide is the same as in the major component oxide. As an example, the model structures of a dilute mixed oxide of TiOz and Si02 are shown in Figure 5-1. In TiOz, each Ti ion is coordinated to six oxygen ions and each oxygen ion to three Ti ions. In Si02, each Si ion is four-coordinated and each oxygen ion is two-coordinated. In the model structure, the coordination numbers of Ti and Si are maintained, and the coordination number of oxygen is that of the

w+

76 Table 5-2 Formation of New Acid Sites in Some Mixed Metal Oxides. (From J. Solid State Chem., 52, 191 (1984), copyright Academic Press).

New Acid Sitesb Matrix Oxide

Substituting Oxide

Experimenta

+

+

-

-

+ +

za2

CdO

wo3

Tanabe's Model

+ +

-

+

+ +

Kung's Model

+ ? ?

+ ?

Footnotes: a) Experimental data extracted from ref. 9 except for those involving WO, which arc taken from Yamaguchi, et al., J. Catal., 65, 442 (1980). b, + means affirmative, - means negative, ? means dependent on the conditions.

SURFACE ACIDITY

77

17-1

I

-0-0

/

'I

I

T-

-0

(a)

Si in T i 0 2

I

I

I

I

O 0 '

0

(b)

I/

Ti in Si02

Figure 5-1 Structures used in Tanabe's model for the prediction of formation of new acid sites in mixed oxides. a. Si in a matrix of Ti02; b. Ti in a matrix of SiOz. (From Bull. Chem. SOC.Jpn., 47, 1064 (1974). copyright Chemical Society of Japan).

major component. The formal charge of each ion is then assumed to be evenly distributed over the coordinating bonds. In figure 5-la, the +4 charge of the substituting Si ion is distributed over four bonds, while the -2 charge of an oxygen ion is distributed over three bonds. Thus each of the bonds surrounding Si bears a net charge of 4/4 2/3 (=+1/3). The excess charge at Si is then 4 x 1/3 =+4/3. In this case, a Lewis acid site is assumed to appear because of the presence of an excess positive charge. In Fig 5-lb, a similar calculation results in an excess charge of -2 at the substituting Ti ion site. In this case, a Brbnsted acid site is assumed to appear, because two protons are assumed to be associated with the site to maintain charge neutrality. If there is no excess charge at the site of the substituting ion by such a calculation, as in the case of AI2O3-Bi2O3,there will be no new acid sites. This model has been applied to many binary oxides, some of which are shown in Table 5-2. It has been found that the prediction is accurate over 90% of the time. The high success rate makes the model very useful, although it is limited by thc assumptions uscd. One limitation is the need to have a unique coordination number, which may be difficult to decide in systems of low symmetry. Another limitation is the use of formal oxidation states which may be quite different from thc real chargc (see Chapter 3). Since electron deficiency at a site refers to real charge at a site, the use of formal oxidation states may not be accurate. Finally the ~

78

-8

-

0 .

0

r,

0

r

-6-

5

p E

z

-4

0.0

.)a

-2-

.-U

2

0-

w

tl c

2-

=

4-

.-P,

0

0

0 . .

.

0

0

60 I

I

I

I

I

Averaged Electronegativi t y Figure 5-2 Highest acid strengths and average electronegativities of metal ions of binary oxides (molar ratio = 1). (From Bull. Chem. SOC. Jpn., 46, 2985 (1973). copyright Chemical Society of Japan).

model cannot predict acid strength. However, there appars to be a rough correlation between electronegativity of the cations and the strength of acid sites in many mixed oxides, as is shown in Fig. 5-2 [lo]. Another model proposed by Kung takes a rather different approach. In Tanabe's model, charge compensation at the substituting ion site by neighboring oxygen ions is important. In Kung's model, changes in electrostatic potential experienced by the substituting cation due to all the ions in the matrix oxide is important. Thus Tanabe's model is a localized model, and Kung's model i s a delocalized model. In Kung's model [ l 11, the difference AV between the electrostatic potentials experienced by a cation A in a matrix BO, and in AO, is given by:

(5-3) where qi is the charge of the ion at a distance ri from the A cation. The subscripts BO and A 0 denote the matrices BO, and AO,, respectively. When AV is negative, cation A in matrix BO, experiences a more negative potential than in

SURFACE ACIDITY

79

AO,. It will be electrostatically more stable. Therefore the electron energy levels of cation A arc lower in energy in matrix BO, and the cation can accept electrons inorc rcadily. It will act as a new Lewis acid site. When AV is positive, A is less rcadily i n accepting electrons in matrix BO, than in matrix AO,. No new Lewis acid site is generated at the substituting A cation. In addition to the changes in the electrostatic potential at site A, when the oxidation states of cations A and B are different, the overall charge neutrality of the solid will be maintaincd by a change in the matrix. Two possibilities exist: a substituting cation A is of a lower formal oxidation state than the matrix cation B, y < z, and the reverse case of A being of a higher formal oxidation state than B, y > L.

When y < z, a simple substitution of B ion by A ion would result in a solid with excess oxygen. This excess can be balanced by: (1) development of anion vacancies; (2) adsorption of protons on the surface; or (3) development of intcrstitial cation defects. (1) and (2) are intimately related if the solid is prepared by aqueous precipitation such that the surface is hydroxylated. If the surface stays hydroxylated, the protons present to balance the excess oxygen will act as new Brbnstcd acid sites not present in the component oxides. On heating, some of these protons may be removed as water which is formed with the concurrent formation of an anion vacancy. The anion vacancy site is a Lewis acid site, and could act as a BrQnstedacid site after it had been hydrated. The effect of (3) is less easily prcdictcd. It could bc important for solids that have open structures. In most common binary oxides, however, the concentration of interstitial defects is limited. Whether these defects lead to the formation of acid sites will likely depend on the nature of the cation. When y > z, a simple substitution of B ion by A ion would result in a deficiency of oxygen. The deficiency can be removed by: (1) adsorption of ncgativcly charged oxygen species onto the A ion; (2) adsorption of OH- onto the A ion; or (3) formation of cation vacancies. When (1) or (2) operates, the consequence of electrostatic potential change at the A ion is removed because the coordinative unsaturation of A is removed by the adsorbed oxygen or OH-. Since it is not likely that adsorbed oxygen acts as an acid site, new acidity is predicted not to dcvclop. Adsorbed OH- could provide Brdnsted acidity, but the acidity would be weak. When (3) operates, new Lewis acid sites could appear because cation vacancies are clcctron deficient. A comparison of experimental data and the predicted formation of new acid sitcs i n binary oxides by Kung's model is summarized in Table 5-2. It is interesting to note that although Tanabe's and Kung's models employ very diffcrent approaches, either method gives the same prediction in many cases. This is bccause Tanabc's model can be mathematically expressed as examining the diffcrence between the qF/c value of the substituting ion and the surrounding oxygcn ion, where qp is thc formal oxidation state, and c is the coordination nunibcr. Although the stoichiometry of an oxide and the qF/c ratio do not have a fixcd rclationship, this ratio is the same for most oxides of the same stoichiometry. Sincc the latticc potentials ( i s . , C q,/r,) are about the same for oxides of the same stoichiomctry and arc diffcrcnt from other stoichiomctries, there is a correlation

80

between the difference in the qF/c values and the value of AV of eq. (5-3). Thus the two models yield similar conclusions in most cases [111. Although they may yield similar results, they involve important differences. Tanabe's model is a localized model such that any new acid sites formed are at the substituting cation site. Kung's model is a delocalized model. New acid sites can be formed on the matrix surface far away from the substitution site as well as at the site. The two models described above apply to dilute oxide solid solutions. That is, isolated "impurity" cations occupy sites of the host oxide substitutionally. Cations deposited on an oxide surface but not incorporated substitutionally may also form new acid sites. A rather detailed study of this latter system has been reported for Fez+ and Fe3+ ions deposited on MgO, TiO2. A1203 and SO2, and Zn2+, Ga3+, Fe3+, M3+,Fez+, Sc3+, and Mg2+ deposited on Si02. Mossbauer spectroscopy and pyridine adsorption have been used to characterize these samples [12-141. The dependence of the amounts of acid sites in these systems on the pretreatment temperature indicates that BrQnsted acid sites may be removed by high temperature evacuation of the oxide, that is, dehydroxylation of the oxide. Lewis acid sites are present on samples pretreated by evacuation at low temperatures, and they persist to much higher temperatures. A model has been proposed by Connell and Dumesic [13,14] to explain the formation of Lewis acid sites for ions deposited on Si02. The model assumes that the formal charge on the deposited cation is balanced by the coordinating surface lattice oxygen ions, and that a cation is coordinatively saturated if' it has a coordination number four as Si ions have in SO2. If the number of coordinating oxygen ions is less than four, the deposited cation may be coordinatively unsaturated and act as a Lewis acid site. It is further assumed that since oxygen ions in the Si02 surface have a coordination number of two, the formal charge of oxygen along each cation to anion bond is negative one. A +2 cation on the Si02 surface requires two coordinating surface oxygen for charge neutrality. For a +3 cation, this number is three. In both cases, the number of coordinating surface oxygen is less than four, and Lewis acidity is expected to develop at these cation sites. This model successfully explains the data presented. In addition to the requirement of coordinative unsaturation of the deposited cation, the formation of a new Lewis acid site also requires that the matrix oxide is not basic. Finally, the strength of the acid site is determined by the electronegativity of the cation: more electronegative cations result in stronger acid sites.

5.3 DETERMINATION OF ACIDITY Since surface acidity is an important property that often determines the surface chemistry, various methods have been developed to mcasure its presence and strength. Three methods will be described. The most commonly used method involves spectroscopic investigation of adsorbed probc molecules. This method is perhaps experimentally the most convenient. With the much improved spectroscopic techniques available today, rapid determination of the presence of surface acid sites is possible. The sccond method employs the dctcrmination of

81

SURFACE ACIDITY isoclcctric point and the third method, titration with indicators.

Adsorption of Ammonia and Pyridine Infrared spectroscopic investigation of adsorbed probe molecules has been employed routinely to characterize various oxides [15]. The well developed technique of spectral subtraction, made convenient with computers and Fourier transform infrared spectroscopy, has made possible the investigation of colored samples including iron oxide, and samples of low surface areas. Because this is a dry tcchnique, there is a wide latitude in the pretreatment of the sample, and thus it is convenient to study the effect of sample pretreatment on the amount and strength of surface acidic sites. The most common probe molecules are pyridine and NH3, although other amines are sometimes used. NH3 and pyridine are bonded to the surface in three different modes. In the first mode, the molecules are adsorbed abstractively: it is protonated by a proton from a surface hydroxyl group. Therefore, it probes surface Brdnsted acid sites.

C ~ H ~+N H M

-+

C ~ H ~ N H ++

0-4

(5-5)

In the second mode, the electron lone pair of the nitrogen atom adds to the (cus) cation of the oxide which acts as a Lewis acid. Such bonding involves o-donation and requires an exposed surface coordinatively unsaturated cation. Hydrogen bonding is the third mode. It is the weakest mode of interaction. and has not been uscd to measure the acidity of surface hydroxyl groups, although it can be done in principle. The presence of BrBnsted acidity is revealed by the formation of protonated ammonium or pyridinium ions. These ions possess characteristic vibrational bands. Likewise, ammonia or pyridine molecules bonded to Lewis acid sitcs or hydrogen-bonded to the surface also possess characteristic bands. The characteristic band for a N&+ ion is located at 1400-1480 cm-' [16]. As a rcfercnce, a detailed table prepared by Tsyganeko [ 171 of NH3 infrared absorption bands in solutions of various proton-accepting and proton-donating solvents can be uscd to help identify the bands. NH3 bonded via the electron lone pair at the nitrogcn atom to a Lewis acid site possesses an asymmetric and a symmetric stretching frequency of v,, = 3330-3380 cm-' (compared with 3444 cm-' for a molccule in the gas phase) and v, = 3260-3280 cm-' (3336 cm-' in the gas phase) [ 31. Additional hydrogcn-bonding intcraction with neighboring anions would lower these frcqucnces [ 181. Whcn hydrogen-bonding to a surface hydroxyl group is the only interaction, the NH3 infrared bands appear near 3400 and 3200 cm-' [3]. For pyridinc, according to Knozinger [ 191, the ring-vibration modes (1 9b and 8a modes) arc most affected by the nature of intermolecular interaction via the nitrogcn atom. These two modes are observed at 1440-1447 and 1580-1600 cm-', rcspectively, for hydrogcn-bonded pyridine, at 1535-1550 and about 1640 cm-' for a pyridinium ion, and at 1447-1464 and 1600-1634 cm-' for pyridine

82

coordinatively bonded to Lewis acid sites. Although a general classification of the mode of bonding is readily possible, it has nor becn possible using this method to obtain quantitative information on the strengths of the acid sites as measured by the proton affinities of the conjugate bases formcd or the pK,'s of the acid sites. Attempts to determine the strength of adsorption by temperature programmed desorption of adsorbed NH3 or pyridine often fail because of oxidation of these molecules by the transition metal oxide. It has not been possible either to correlate the vibrational band frcqucncics with the nature of the metal ion or its coordination [20,21]. Athough ammonia and pyridine have been used to probe similar surface propcrties, they are not identical. One difference is their sizes. The molccular cross-sectional area of NH3 is 0.127 nm2, and of pyridine is 0.313 nm2. Thus it may be possible that more ammonia is adsorbed than pyridine. The larger size of pyridine may also result in weaker bonding to a Lewis acid cation if steric crowding is a factor, especially if the cation Lewis site is recessed into the surfacc. This latter effect has been used to explain the smaller amount of pyridine adsorbed than NH3 on Cr2O3 [ 181. Another difference betwecn the two molecules is their relative basic strcngth. In an aqueous solution, ammonia is a stronger base with a pKb of around 9 compared to a pKb of about 5 for pyridine [221. However, the basicity of pyridine in the gas phase is significantly higher [23,24]. It has been pointed out in section 5.1 that the polarizable surface oxygen ions make molecules on an oxide surfacc behave as if they were in an aqueous solution. Thus the rclativc basicity in aqueous solutions should be more appropriate to describe the relative Brdnstcd basicity of these two molecules. Indeed, on the oxidized and the reduced molybdena/alumina, N h + ions are detected. On the other hand, pyridinium ions are dctected only on the oxidized form [25]. The relative Lewis acidity is different. For example, pyridine is desorbed more slowly than ammonia from thc Lewis acid sites of F%03, which suggests that it is adsorbed morc strongly than NH3 [26]. Most of the first row transition metal oxides and a fcw of the othcrs have bccn studicd with these two moleculcs. As shown in Table 5-3, Lewis acid sitcs arc found on practically all of the oxides studied that have been prctreatcd by hcating to 400°C or higher in vacuo. On the other hand, Brbnstcd acid sitcs are found only on V205, Nb205,Moo3, W 0 3 , Re207, and Cr2O3/AI2O3[ l S ] . The density of acid sites dcpends on the sample prctreatmcnt. Figurc 5-3 shows an example of this effect on Nb205 as dctcrmined by pyridinc adsorption [27]. The sample dried at 100°C possesses a rather large amount of Brdnstcd and Lewis acid sitcs. Hcating Ihe sample to 300°C in vacuo results in a decrcasc in the numbcr of BrBnsted acid sites but an increasc in Lewis acid sitcs. Thus thcrc is an intcrconversion upon dehydration. Further heating to 500°C in vacuo rcsults in sharp dccreascs in both typcs of acid sitcs, which are due mostly to a rapid loss in surface arca. The cffcct of evacuation aftcr pyridinc adsorption is also shown in the figurc. As expcctcd, a highcr evacuation temperature rcsults in smallcr amounts of pyridinc adsorbed. I t is interesting to note that thc dccrcasc in the amount of pyridinc adsorbcd is qiiitc gradual with the incrcasc in evacuation

SURFACE ACIDITY

83

Table 5-3 Acid Sites on Various Transition Metal Oxides Determined by Ammonia or Pyridine Adsorption

Oxidea

H-bonding

Mode of Bonding Lewis Acid

+ +

+ +

+ + +

Brbnsted Acid

+ +

+ +

+ + +

+ +

+ +

+

Footnote: a) Pretreated by evacuation at 400°C or higher.

temperature. This suggests that there is a broad range of strength of interaction of pyridine with the surface, which implies the presence of sites of a broad range of acid strength.

Isoelectric Point A second method to determine the presence of acid sites is by determining the isoelecuic point of an oxide. This method makes use of the phenomenon that solid oxide particles in aqueous suspensions are often electrically charged. Charged particles are formed when there is an imbalance between the densities of adsorbed H+,OH-, and ionized surface OH groups. They may also be formed by adsorption of metal hydroxo complexes derived from the hydrolysis products of material dissolved from the solid, i.e., [Mz+(OH),,]zA species. The presence of a net charge on the oxide particle can be observed in electrophoresis experiments. The magnitude of the surface charge depends on the pH of the solution, and there exists a pH at which there is no net surface charge. This pH is called the isoelectric point, It is equivalent to h e point of zero charge, and is the pH at which suspended particles in H 2 0 do not move in an electric field. Since the oxide particles arc in an aqueous solution during the determination of isoelccuic point, much of the variation of surface acidity due to different extents

84

0

200

400

Pretreatment Temp.

600 C

Figure 5-3 Relative amounts of a: Lewis acid; b: Brdnsted acid sites of a sample of Nb2OS.nHzO as a function of pretreatment temperature as determined by the IR absorption peak intensity of adsorbed pyridine after evacuation at: 1, room temperature; 2. 100°C; 3, 200°C; and 4, 300°C. The surface areas of the sample after pretreatments at various temperatures are : lOO"C, 164 m2/g; 300"C, 126 m2/g; and 500°C. 42 m2/g. (From Bull. Chem. SOC.Jpn., 56, 2927 (1983), copyright Chemical Society of Japan).

of hydroxylation cannot be studied. However, the isoelectric point is still a rather sensitive function of the history of the sample. This is because the density of charges on any oxide surface is quite low. Thus the surface charge could be significantly altered by the presence of impurity cations or anions on the surface, as well as surface imperfection such as vacancies and dislocations. Its value could also depend on the nature of the ions in the aqueous phase that are used to vary the pH due to the possibility of specific adsorption of the ions. These factors have led to wide variations in the reported values of the isoelecuic points of oxides. Isoelecuic points can be determined in an elcctrophoresis apparatus where the zeta-potential of a solid is measured as a function of the pH of the solution. Since zeta-potential is the potential between a charged surface and the electrolyte solution, its value is zero when the net charge of the surface is zero. Therefore, the pH at which the zeta-potential is zero is the isoelecuic point, assuming that there are no adsorption of charged species other than protons and hydroxide ions on the

SURFACE ACIDITY

85

Figure 5-4 Zeta-potential at 22.5'C as a function of pH. Curve 1: NiAl204 and CoA1204; curve 2: C0304; curve 3: A1203; and curve 4: NiO. (From J. Catal.. 83, 225 (1983). copyright Academic Press).

surface. Figure 5-4 shows some sample data for a number of oxides [28]. The isoelectric points for a large number of oxides have been determined. Table 5-4 is an updated summary of the comprehensive data collected by Park [29,30]. From these data, Park proposes a generalization presented below based on the stoichiometry of the oxide:

M20

MO M2°3

MO2 M205, M03

Isoelectric point (IEP) IEP > pH 11.5 8.5 and directions, respectively. The rate of oxygen removal is first order in surface oxide concentration, and first order in the hydrogen pressure [4] The contracting sphere model is an extreme case of the nucleation model in that it is assumed that the number of reduced oxide grains formed on the surface of the sphere is so large that the boundaries of the grains overlap when the diameters of the grains are still small versus the radius of the sphere. This situation can be accurately modeled as a rapid formation of an uniform layer of reduced oxide or metal on the sphere (Figure 6-3). The thickness of this reduced layer grows uniformly, resulting in a spherical core of oxide that shrinks with time. Since the rcduction of the core oxide is accomplished by diffusion of ions and reductant across the oxide-reduced oxide interface whose area decreases with time, and the distance of this interface from the surface of the sphere increases with time, the rate of reduction decreases with time. This results in a characteristic curve for the extent of reduction shown in Figure 6-2. The time dependence of the degree of reduction for isothermal reduction by the contracting sphere model is described by eq. (6-4) [2]: k rodo (‘O

P

-

c,,>t= [l

-

-a 3- ( 1 -

where k, is the rate constant of reduction per unit area at the oxide-reduced oxide interface, kd is the diffusivity of reductant through the reduced layer, ro is the radius of the entire sphere, and C,, C,, are the actual concentration of reductant at the outer surface of the sphere and the equilibrium concentration. The expression assumes that the chemical rcaction at the interface is first order with respect to Cl-C,,, where C , is the reductant concentration at the interface. Another situation is possible that is in between the nucleation and the contracting sphere models. This situation is as follows. Small grains of reduced oxide are first formed on the surface of an oxide particle as in the nucleation model. However, these grains of reduced oxide are less active in activating the reductant molecules than the fully oxidized oxide. Therefore, their presence on the oxide surface inhibits the rate of reduction bccause it reduces the area of exposed fully

96 oxidized oxide. These grains grow as reduction continues, eventually cover up the oxide particle. Whence the oxide is reduced by the contracting sphere model. In this situation, the rate of reduction decreases continuously with increasing extent of reduction, and the fraction of reduction-versus-time curve should be qualitatively the same as the one for the contracting sphere model. The descriptions so far assume that the cations and anions in an oxide have homogeneous properties. However, the surface chemistry and catalytic properties depend heavily on the surface condition. In section 4.1, it is shown that a surface may possess more than one type of surface lattice oxygen. A Ti02 (1 10) surface is one such example (Fig. 4.1) where there are two types of surface lattice oxygen ions. Type one is bonded to two six-coordinated Ti cations, and sits above the plane of the other ions. The other type is bonded to one six- and two fivecoordinated Ti cations and is in the surface plane. Their different positions and bondings should make them removable with different degrees of difficulty. Removal of the first type would involve breaking two Ti-0 bonds. Removal of the second type would involve breaking three Ti-0 bonds, and should be more difficult than the first type. Indeed, this has been confirmed on the Sn02 (110) surface [67]. Even within each type, it is expected that there will be a gradual increase in the difficulty in removing the oxygen ions as the reduction increases. This consideration suggests that unlike the kinetic treatment above which applies well to bulk reduction, reduction of the surface is likely to proceed with a rate constant that depends on the orientation of the surface plane, the extent of reduction, and the density and nature of surface defects. Unfortunately no quantitative experimental data are available.

6.4 MECHANISM OF REDUCTION The most heavily studied system is perhaps the reduction of well-dispersed molybdenum oxide on a support. In the reduction of molybdcnurn oxide supported on alumina, it has been found that one molecule of water is relcased for every two hydrogen molecules consumed at low extents of reduction. Thus some hydrogen is retained by the partially reduced oxide [51. The Mo(V) specics has been detected by EPR, but its concentration is low compared with thc extent of reduction [6]. A model of the reduction process that is consistent with these results is shown in Figure 6-4 [7]. In this model, the well-dispersed supported molybdenum oxidc is present as an oxide chain. The initial step of reduction is the heterolytic adsorption of hydrogen molecules followed by reduction of Mo cations and migration of H from a cation to an oxide ion to form a p-hydroxl group. Thus two p-hydroxyl groups are formed for one hydrogen molecule adsorbed. The two p-hydroxyl groups may bridge the same Mo-Mo pair or diffcrcnt pairs as shown in structure 111, and these two modes probably occur with equal likelihood. On the other hand, dehydration probably occurs more readily if the p-hydroxyl groups span the same Mo-Mo pair. This would explain the observation of one molecule of water released for every two molecules of hydrogen consumed at low extents of reduction. This i s

97

REDUCTION

I\ / I o-o-o-

-o-o-o

-1-

H2°

H 0

(VI)

-o-o-o

o-o-o-

Figure 6-4 A possible mechanism for the reduction of supported MOO,

98

illustrated by the conversion of III -t IV. Reduction continues by further adsorption of hydrogen and dehydration (IV + V + VI). If the molybdenum species is part of a crystallite instead of the monolayer shown, diffusion of lattice oxygen from the bulk to the surface occurs. This is equivalent to the diffusion of anion vacancies from the surface into the bulk, and surface molybdenum oxide can then undergo another cycle of reduction. This process results in bulk reduction. The mechanism and the extent of reduction determine whether the reduced oxide can be rapidly reoxidized. In general, if the reduction preserves the essential structural feature of the oxidized form, rapid oxidation may be expected. A rapidly reversible system has been demonstrated for CaMn03 and CaMnO2.5, and for CazMn04and Ca2Mn03.s [8]. The reduction of M a 3 to the lower oxides by the formation of shear planes is also expected to be rapidly reversible. When there are a number of intermediate oxides between the fully oxidized state and the metallic state, then during reduction, reduced oxides are normally formed in the sequence of increasing degree of reduction. For example, in the reduction of large crystallites of CuO at 250°C in 2% H2/Nz, the oxide is almost totally reduced to CuzO before metallic Cu appears [9]. However, depending on the system and the reduction conditions, some intermediate oxides are further reduced immediately upon formation. For example, one might expect the reduction of F@03 to proceed via Fe304 and then FeO en route to metallic Fe. However, when Fez03 is reduced in Hz, F%04 is lirst formed, and Fe304 is then reduced directly to Fe metal. Little or no FeO is detected [lo]. Even with a given oxide, the reduction sequence can depend on the crystallite size, the crystallographic form, and the nature of the support. In the example of CuO mentioned above, much less Cu20 is detected during reduction of small crystallites, and a majority of CuO appears to be directly reduced to Cu metal [9]. The dependence on crystallographic form is illustrated by N b z 0 5 . The reduction of p-Nb205 proceeds readily in H2 at 750°C to form Nb02 [ l l ] . However, aNb205 is reduced through a series of complex intermediates, but no Nb02 is formed [12]. The effect of the support is quite complicated. It is separately discussed in the next section.

6.5 EFFECT OF SUPPORT The interaction of an oxide with a support could affect both the thermodynamics and the kinetics of its reduction. The effect arises partly because a support can act as a dispersing agent, and small crystallites may be reduced differently than large crystallites, such as the example of CuO mentioned above. A support can also interact chemically with the oxide. Often, but not always, such a chemical interaction increases the resistance of the oxide to reduction. There are quite a number of examples of increased resistance to reduction when an oxide is supported. W 0 3 supported on A1203 is more difficult to reduce than unsupported W 0 3 [13,14]. Fe2O3 supported on y-AI2O3 [2], and Pt oxide supported on y-AI2O3 [15] are similarly less reducible than their unsupported

REDUCTION

99

oxides. It has been reported that a F't+ signal is detected by EPR on a I"fl-Al203 sample even after reduction at 623 K [16]. There are disagreements for some other systems. One report mentions that V205 supported on Ti02 is less reducible than unsupported Vz05 [17]. Another lists the ease of reduction as decreasing in the sequence Vh-Al203, VDiO2, V2O5 [18]. In this case, the influence of impurities present in the support may be important. The rhodium oxide system is interesting. It has been reported that in temperature programmed reduction, carefully oxidized Rh crystallites on A1203or Ti02 that contain only a surface layer of Rh oxide are reduced more readily than the bulk oxide [19]. In another report, it is mentioned that RhIy-Al203 that is oxidized at a temperature above 873 K is reduced less readily than a similarly pretreated bulk oxide [20]. It has been proposed that a high temperature of oxidation is necessary to induce the strong interaction with the support that leads to lower reducibility. It is interesting to mention that temperature programmed reduction of a Rh/Ti02 sample fully oxidized at low temperature shows two distinct reduction peaks. The lower temperature peak is due to well-dispersed Rh2O3 on the support, and the higher temperature peak is due to large Rh2O3 crystallites [19]. It appears that in the supported Rh203 system, different pretreatments lead to different effects of dispersion and chemical interaction on the rates of reduction. In some cases, the apparent support effect may be actually due to the reaction between the oxide and the support to form a new compound. In this case, the increased resistance to reduction is due to the different properties of the new compound. This is the case for NiO/A1203.The reduction of this oxide often shows portions of different rates. It is now understood that the different rates correspond to the reduction of crystalline NiO, Ni2+ bound to alumina, and nickel aluminate [21-241. The increased resistance to reduction of V205/Mg0 is likewise due to the formation of Mg3(V04)2[25]. The reduction of a metallic ion in a zeolite depends on the location of the ion. It has been reported that a Ni2+ ion in a hexagonal prism of faujasite is more difficult to be reduced than a Ni2+ ion in the sodalite cage or supercage [26]. Presumably, the lattice oxide ions of the hexagonal prism stabilize the Ni2+ ion rather effectively. It should be noted that while a support affects the reducibility of an oxide, the presence of an oxide may also cause changes in a support during reduction. For example, during temperature programmed reduction of a V205/Ti02 (anatase) sample, reduction of V2O5 and transformation of anatase to rutile are simultaneously observed [27-291. When an oxide strongly interacts with a support, the support could determine the structure of the oxide and its reduction behavior. For example, reduction of vanadia well-dispersed on anatase Ti02 removes an average of 0.85-0.90 oxygen atoms per vanadium ion [30]. Reduction of vanadia well-dispersed on A1203 removes an average of 0.6 oxygen atoms per vanadium. Reduction of large crystallites of V205 removes 0.55 oxygen atoms per vanadium.

100

6.6 EFFECT OF OTHER COMPONENTS The presence of other components may enhance the rate of reduction of an oxide by promoting the rate of activation of the reductant. For example, the presence of a noble metal may enhance the dissociation of hydrogen. The activated reductant then migrates to the reduction site and reduction is enhanced. In some cases, migration is by the spillover phenomenon in which the activated reductant migrates over the support. This mechanism has been used to explain the results from a physical mixture of F ~ 0 3 & - A l 2 0 3and ptly-AlzO3. Reduction of FezO3/y-AIzO3commences at about 430 K in this mixture, much lower than 640 K withoutt'F [2]. The reduction of Re-Pt/Alz03 is another well studied example. The addition of Pt considerably lowers the temperature required to reduce Rez&. This has been explained by the surface migration of R%07 to Pt [31,32]. Such migration results in physical contact between the two components, and hydrogen atoms dissociated on Pt can migrate readily to R%07. A similar enhanced reduction by the addition of Pd to R%07/A1203is also observed [33]. The reduction of Pt is also affected by the physical contact between Pt and RezO7. In a temperature programmed reduction experiment, the reduction peak for Pt is moved to a higher temperature in the presence of Rez& than in its absence

WI. The formation of hydrogen bronze is much facilitated when the oxide is in contact with a metal that dissociates hydrogen molecules. It has been reported that HxW03, H,M0O3, and HXV2O5can be readily prepared by exposing the oxide deposited with platinum to hydrogen gas [64-661. It is interesting that only part of the hydrogen incorporated into the bronzes can be removed easily, such as by evacuation or by consumption in the hydrogenation of ethene [65,66].

6.7 REACTIVITY OF REDUCED SURFACES Reduced surfaces often possess chemisorptive and catalytic properties different from stoichiometric surfaces. It has been shown that NO is adsorbed strongly on Fez+ but only weakly on Fe3+ [35-371. and strongly on partially reduced tungsten oxide but not on fully oxidized W6+[38]. CO has been shown to adsorb on Cu+ strongly, but on Cu2+ weakly [39]. The reason for the difference between the oxidized and the reduced ions is probably the different electronic configurationsof the ions, and the lower degree of coordinative unsaturation in the fully oxidized oxides. Together with the presence of surface cations with a high degree of coordinative unsaturation, a reduced surface also has cations of a lower oxidation state and anion vacancies. Some of these factors lead to the enhanced activity of the surface to dissociatively adsorb molecules that contain oxygen atoms. The dissociative adsorption often results in reoxidation of the surface. The data in Table 6-1 illustrate this effect. In these experiments, the fully oxidized (i.e., stoichiometric) surfaces are prepared either by extensive annealing of the surface

101

REDUCTION Table 6-1 Chemical Properties of Oxidized and Reduced Single Crystal Surfaces Oxidized" or Reduced

Surface

Adsorption

Ref.

Ads. O2 readily as 02-; Ads. H 2 0 molecularly.

a b

Reduced

Ads. O2 readily as 02-; Ads. H 2 0 dissociatively.

a b

Oxidized

Ads. O2readily as 02-; Ads. H 2 0 dissociatively. Ads. O2 readily; Ads. H 2 0 dissociatively.

C

SrTi03(1 1) Reduced

Ads. H 2 0 reoxidizes the surface.

d

SrTi03(100) Oxidized

Ads. Ads. Ads. Ads.

v203(W'

1

Reduced

Reduced

Ti02(110)

Oxidized Reduced

H 2 0 molecularly; small amount of 02. O2 reoxidizes the surface; H 2 0 dissociatively.

Does not ads. CO or H2. Ads. O2dissociatively and reoxidizes the surface; Ads. H20 dissociatively.

C

j e e f g gh

f

Ti02(100) (1x3)

Oxidized Reduced

Ads. H 2 0 molecularly. Ads. H 2 0 dissociatively.

1

Ti02(100) (1x7)

Reduced

Ads. H 2 0 dissociatively.

1

Footnote: a) Fully oxidized sufaces are stoichiometric surfaces. References: a) R.L. Kurtz, and V.E. Henrich. Phys. Rev. B. 25. 3563 (1982). b) R.L. Kurtz, and V.E. Henrich. ibid. 26, 6682 (1982). c) R.L. Kurtz, and V.E. Henrich, ibid, 28, 6699 (1983).

I

102 Table 6-1 continued d) S. Ferrer. and G.A. Somorjai, Surface Sci.. 97. L304 (1980). e) V.E. Henrich, et al., J. Vac. Sci. Techno]., 15, 534 (1978). f) V.E. Henrich. et al., Solid State Commun., 24. 623 (1977). g) W. Gopel, et al., Phys. Rev. B, 28, 3427 (1983). h) V.E. Henrich, et al., Phys. Rev. Lett., 36, 1335 (1976). i) W. Lo. et al., Surface Sci., 71, 199 (1978). j) by HREELS. R.G. Egdell and P.D.Naylor, Chem. Phys. Lett., 91, 200 (1982).

or by cleavage of a single crystal. Reduction of the surface is achieved by ionsputtering or H2 reduction. Adsorption and dissociation of the molecules are monitored with UPS. The results in Table 6-1 are supported by other measurements such as temperature programmed desorption. When adsorbed CI8O2 is desorbed by heating from a partially reduced TiOz surface, the amount of C 1 * 0 desorbed increases with the extent of reduction 1401. Reduction of a surface does not always lead u, stronger interaction with molecules. The peak temperature in the temperature programmed desorption of CO from Ti02 does not change with increasing degree of reduction, although the amount increases [40]. On ZnO, the differential heat of adsorption of CO is much lower on a reduced (7 kJ/mole) than on an oxidized surface (44 kJ/mole) [41]. In contrast to NO which is adsorbed strongly on Fez+ but weakly on samples containing only Fe3+ [35-371, pyridine is adsorbed more strongly on samples containing Fe3+ than Fe2+ [42]. It is interesting to note that in this study, Mossbauer spectroscopy detects direct interaction between pyridine and Fe2+ of low coordination, but not that between pyridine and Fe3+. However, pyridine is shown by infrared spectroscopy to adsorb on Lewis acid sites. Since Fe3+ is a harder and smaller cation than Fez+, it is a stronger Lewis acid.

6.8 INFLUENCE OF REDUCED OXIDES ON THE PROPERTIES OF TRANSITION METALS It was first reported about ten years ago that certain transition metal oxides can have a dramatic effect on the chemisorptive and catalytic properties of metals when they are in close contact with the metal. This effect was originally tcrmed Strong Metal-Support Interaction, but is currently referred to as decoration effcct. The first report of this effect was on the suppression of hydrogen and carbon monoxide adsorption capacity of noble metals. It has been found that when a noble metal is supported on Ti02, the material chemisorbs H2 or CO at a stoichiometry of roughly one H atom or CO molecule per surface exposed noble metal atom, if the material is reduced at 200°C. If the material is reduced at 50O0C, its ability to chemisorb H or CO is essentially completely suppressed [43,44]. This is illustratcd with typical data in Table 6-2. It was later found that

103

REDUCTION

I

I 100

300

700

500

TA

C

Figure 6-5 Hydrogen chemisorption on iridium supported on various oxides as a function of activation in hydrogen for 1 h at each of various temperatures. T, is the activation tcmperature. and H/M is the atomic ratio of hydrogen adsorbed to iridium in the catalyst. (From Science, 211, 1121 (1981), copyright American Association for the Advancement of Science).

Table 6-2 Suppression of H2Adsorptive Capacity due to Decoration Effect (From Scicnce, 211, 1121 (1981), copyright American Association for the Advancement of Science).

2% Metal on Ti02 Support

Ru Rh

Pd 0s Ir Pt

H atom adsorbed/rotal metal atoms Reduction at 200°C Reduction at 500°C 0.23 0.7 1 0.93 0.2 1 1.60 0.88

0.06 0.01

0.05 0.1 1 0.00 0.00

this effect is not unique to Ti02 as the support. Fig. 6-5 shows the extent of suppression of the hydrogen chemisorptive capacity as a function of reduction temperature for various oxide supports. In general, oxides that are readily reducible at intermediate temperatures all show this effect. The effect is reversible. That is. reoxidation with oxygen followed by a low temperature reduction at 200°C restores most of the adsorptive capacity. The influence of the decoration effect on the catalytic properties of the metal depends on the reaction. Some typical examples are shown in Table 6-3. It can be seen that the decoration effect suppresses the activity of R for benzene hydrogenation and cyclohexane dehydrogenation,but enhances the activity in CO hydrogenation. The activity of the Fe catalyst in ammonia synthesis is slightly decreased, but the activation energy is greatly increased. In the case of butane hydrogenolysis, the selectivity and the activity are both altered. In addition to these examples, it has been shown that CO hydrogenation on Ni catalysts is also enhanced by the decoration effect, although the extent of enhancement may vary from very little [48] to rather substantial [49]. In ethane hydrogenolysis on titania-supported Rh, it has been found that the activity decreases rapidly with increasing reduction temperature (Fig. 6-6), whereas the cyclohexane dehydrogenation activity over the same catalyst remains almost unchanged [501. It is now rather well established that this effect is not due to the formation of alloys (e.g. R-Ti alloy), the encapsulation of the metal by the support, sintering of the metal, poisoning of metal by impurities in the support, or simply electron transfer between the bulk of the support and the bulk of the metal crystallite. These conclusions follow from a variety of experimental observations. For example, transmission electron microscopic studies as well as X-ray diffraction show no evidence of sintering [45,51], and Mossbauer spectroscopy shows that the bulk of the iron crystallites is the same whether or not the sample is exhibiting the decoration effect [47], even though the titania support near the metal crystallite is reduced from Ti02 to TbO, [51,52]. The current picture is that the origin of the effect is the migration of small particles of reduced titania (or other reduced support) onto the metal crystallites to "decorate" the metal surfaces. These decorating reduced oxide particles may partially block the metal surfaces from gas molecules, affect the electronic structure of the neighboring metal atoms, or provide an oxide-metal interface for interaction with molecules. Depending on the reaction, one or more of these effects may participate to affect the observed characteristics of the reaction. That decoration is the physical picture was first suggested by Dumesic [471 on the Fe/Ti02 system. Using Mossbauer spcctroscopy, it has been observed that the bulk properties of Fe crystallites are the same whether they are in the decorated state or not. Thus the effect must be a surface phenomenon. It is then proposed that titania species cover the iron crystallites. Such a decoration model would suggest that the extent of the effect should depend on the time allowcd for the reduced oxide particles to migrate onto the metal crystallites and thc interface between the metal and the oxide. These have been confirmed. It is observcd that using the rate of ethane hydrogenolysis as a measure, the extent of the decoration

105

REDUCTION Table 6-3 Effect of Decoration on the Catalytic Properties of Noble Metals

Reaction

Catalyst

Reaction Rate

C6H6 hydrogenation at 288K

4.8% W i O 2 523 K reduced 773 K reduced

40 mmole/h-g cat. 3.5 mmole/h-g cat.

C6HI2dehydroge- 2.7% IrKi02 nation at 523 K 523 K reduced 773 K reduced CO hydrogenation 1.9% Pt/Ti02 at 524 K 473 K reduced 773 K reduced

NH3 synthesis at 673 K

C4Hlo hydrogenolysis at 623 K

45

1400 mmole/h-g cat. 304 mmole/h-g cat. 46

0.01 11 molecules/s-Pt, 0.076 molecules/s-% (0.0195 if assumed the same dispersion as lowtemperature reduced sample) 0.03 1 ks-l

798 K reduced

0.01 1 ks-'

773 K reduced

Ref

45

1.14 % Fe/Ti02 713 K reduced

4.8% Pfli02 623 K reduced

Other Effects

47 Eact = 100 kJ mole-' Eact = 220 kJ mole-' 45

Relative product formation rate: c1=35, C2=47, C,=25, i-C4=88 C, =0.65, C2=1.1, C3=0.7, i-C4=0

effcct depends linearly on the square root of the reduction time [501, which is characteristic of diffusion processes. For a given reduction time, the hydrogenolysis activity decreases as the inverse of the particle diameter. Finally, the increasing suppression of the hydrogenolysis activity as the reduction time increases is found to parallel a similar suppression by the addition of copper to a nickcl catalyst, which is interpreted by the braking-up of nickel ensembles on the surface by copper atoms. Thus the data are consistent with the model that the surface rnctal ensembles are broken up by reduced titania particles decorating the surface.

106

Reduction Temp-

K

Figure 6-6 Ethane hydrogenolysis and cyclohexane dehydrogenation on Rh/Ti02 catalyst as a function of catalyst reduction temperature. (From J. Catal., 82, 279 (1983). copyright Academic Press).

Recently, it has been further proposed that the metal-oxide interaction occurs through the interaction of metal atoms with oxygen ion lattice vacancies in the reduced oxide. At temperatures sufficiently high to induce the decoration effect, the oxide support is reduced so that it has a high concentration of anion vacancies. The high concentration of anion vacancies enhances the diffusion of metal atoms into the near-surface region of the bulk, and results in the formation of a raft-like metallic cluster covered by a thin (atomic) layer of the support. When the support is reoxidized, the anion vacancies are filled, driving the metal atoms back to the surface [53]. When titania particles are deposited on model catalysts of a Ni (1 11) single crystal surface 1541 or a F’t foil [55,56], a similar suppression of the H2 or CO chemisorption capacity is observed which is similar to that resulting from high temperature reduction of supported metals. Enhanced catalytic activity in CO hydrogenation has also been observed on these low surface arca catalysts [54,56]. In the case of Pt foil, the activation energy is reduced from 126 to 80 W/mole. Furthermore, a small amount of deposited TiO, particles is found to enhance the methanation activity of Ni. At an optimum coverage of 8%, (recently there is doubt about this number because of questions about the calibration method employed), the activity is four times that of a clean Ni surface. On the other hand, complete suppression of the chemisorptive capacity rcquires complete coverage of the metal surface by the oxide particles 156,571.

REDUCTION

107

The extent of election transfer between the metal and the decorating particles is not established. For example, a study of Pt crystallites supported on a Ti02 single crystal surface by XPS and AES suggests electron transfer from Ti02 tot'F [%I, while a XANES study suggests electron transfer from Pt to Ti02 [591. The influence of the decorating effect on the heat of adsorption depends on the system. On Pt supported on titania, the effect results in a decrease in the initial heat of adsorption of H2 from 92 to 82 kJ/mole, but no change for CO adsorption [60]. The integral heats of adsorption of both CO and H2 are substantially reduced [61]. However, the integral heat of adsorption of CO or H2 on Pd is not affected [62,63].

REFERENCES 1. a) M. Langell, Surface Sci.. 186, 323 (1987); b) W. Uena. Y. Moro-oka, and T. Ikawa, J. Chem. SOC. Faraday Trans. I , 78.495 (1982). 2. N.W. Hurst. S. J. Gentry, A. Jones, and B. D. McNicol, Catal. Rev., 24, 233 (1982). 3. J. Haber, J. Less-Common Met., 54, 243(1977); B. Delmon. "Introduction a la Cinetique Heterogene", Edition Technip, Paris, 1969. 4. R. P. Furstenau, G . McDougall, and M. A. Langell, Surface Sci., 150, 55 (1985). 5. W. K. Hall, and F. E. Massoth, J. Catal., 34. 41 (1974). 6 . S. Abdo, R. B. Clarkson and W. K. Hall, J. Phys. Chem., 80, 2431 (1976). 7. H. Weigold, J. Caul., 83, 85 (1983). 8 . K. R . Poeppelrneier, M. E. h n o w i c z . and J. M. Longo, J . Solid State Chem.. 44, 89 (1982). 9. P. B. Himelfarb, F. E. Wawner, Jr.. A. Bieser, Jr.. and S. N. Vines, J. Catal.. 83, 469 (1983). 10. E. Unmuth, L. H. Schwartz, J. B. Butt, J. Catal., 61, 242 (1980). 1 1 . K. M. Nimmo, and J. S . Anderson, J. Chem. Soc.,Dalton Trans., 2328 (1972). 12. S. K. E. Forghany, and J. S. Anderson, J. Chem. Sw.Dalton Trans., 225 (1981). 13. J. Salvati, J. Phys. Chem., 85, 3700 (1981). 14. I. E. Wachs, C. C. Chersich, and J. H. Hardenbergh, Appl. Cafal..13, 335 (1985). 15. R. D. McNicol, H. Charcosset, M. T. Chenebaux, and M. F'rimet, "Scientific Bases for the Preparation of Heterogeneous Catal., 11". 1978, paper no. B8. 16. T. Huizinga, and R. Prins, J. Phys. Chem.. 87, 173 (1983). 17. I. E. Wachs, S. C. Chan, and R. Y. Saleh, J. Catal., 91. 366 (1985). 18. A. J. Van Hengsbom, J. G. van O m e n . H. Bosch, P. J. Gellings, Proc. 8 f h Infern. Cong. Cafal.,IV, 297 (1984). 19. J. C. Vis. H. F. J.van' T Blik. T. Huizinga. J. Van Grondelle, and R. Prim. J . Cafal.,95, 333 (1985). 20. H. C. Yao, S. Japan, and M. Shclef, J. Catal., 50, 407 (1977). 21. J. Bachelier, J. C. Duchet, and D. Comet, Bull. SOC. Chim. Fr.. 3, 112 (1978). 22. M . Wu, and D. M. Hercules, J. Phys. Chem., 83, 2003 (1979). 23. H. E. Swift, F. E. Lutinski, and H. H. Tobin, J. Catal., 5, 285 (1966).

24. P. Dufresne, E. Payen, J. GrimbloS and J. P. Bonnelle, J. Phys. Chem., 85, 2344 (1981). 25. M. Iwamoto, T. Takenaka, K. Matsukami, J. Hirata, S. Kagawa. and J. Izumi, Appf. Cataf.,16, 153 (1985). 26. H. J.-Jiang, PhD thesis, Northwestern University, 1988. 27. G. C. Bond, A. J. Sarkany, and G . D. Parfitt, J. Cataf.,57,476 (1979). 28. A. Vejux, and P. Courtine. J. Solid State Chem., 23, 93 (1978). 29. D. J. Cole, C. F. Cullis, and D. J. Hucknell, J. Chem. SOC.. Farad. Trans. I. 72, 2185 (1976). 30. J. Haber, A. Kozlowska, and R. Kozlowska, J. Cafaf.,102, 52 (1986). 31. N. Wagstaff, R. Ens. J. Cataf..59. 434 (1979) 32. B. H. Isaacs, E. E. Petersen, J . Cataf.,77, 43 (1982). 33. S. B. Ziemecki, G. A. Jones, and J. B. Michel. J. Catal.. 99. 207 (1986). 34. M. S.Nacheff, PhD thesis. Northwestem University, 1988. 35. S. Yuen, Y. Chen, J. E. Kubsh, J. A. Dumesic, N. Topsbe, and H. Topsbe, J . Phys. Chem., 86, 3022 (1982). 36. N. S. Gill, R. H. Nuttall, D. E. Scaife, and D. W. A. Sharp, J. Imrg. Nucl. Chem.. 18, 79 (1961). 37. G. W. Poling, and R. P. Eischens, J . Efectrochem. Soc.. 113, 218 (1966). 38. K. Segawa, and W. K. Hall. J. Catal., 77, 221 (1982). 39. Yu A. Lokhov, Z. M u d , and A. A. Davydov, Kinet. Kataf.. 20, 207 (1979). 40. G. B. Raupp, and J. A. Dumesic. J. Phys. Chem., 89, 5240 (1985). 41. E. Giamello and B. Fubini, J. Chem. SOC. Faraday Trans. I , 79. 1995 (1983). 42. G. Connell. and J. A. Dumesic. J. Catal., 101, 103 (1986). 43. S. J. Tauster, S. C. Fung, and R. L. Carten. J . Amer. Chem. Soc.,100, 170 (1978). 44. S. J. Tauster, S. C. Fung, R. T. K. Baker and J. A. Horsley, Science. 211, 1121 (1981). 45. P. Meriandeau, 0. H. Ellestad, M. Dufaux, and C. Naccache. J. Catal., 75, 243 (1982). 46. A. Vannice. and C. C. Twu, J. Cataf.,82. 213 (1983). 47. J. Santos, J. Phillips, J. A. Dumesic. J . Catal., 81. 147 (1983). 48. C. H. Bartholomew, R. B. Pannell, and J. L. Butler, J. Catal.. 65, 335 (1980). 49. M. A. Vannice, and R. L. Garten. J. Catal., 56, 236 (1979). 50. D. E. Resasco, and G . L. Haller, J. Catal., 82. 279 (1983). 51. R. T. K. Baker, E. . Prestridge, and R. L. Garten, J. Catal., 56, 390 (1979). 52. R. T. K. Baker, E. B. Prestridge, and R. L. Garten, J . Cataf.,59, 293 (1979). 53. M. G. Sanchez, and J. L. Gazquez, J . Catal., 104, 120 (1987). 54. Y. W. Chung, G. Xiong, and C. C. Kao, J . Cataf.,85, 237 (1984). 55. C. S. KO and R. J. Gorte. J . Cataf.,90, 59 (1984). 56. R. A. Demmin, C. S. KO. and R. J. Gorte, J . Phys. Chem. 89, 1151 (1985). 57. D. J. Dwyer, S. D. Camero. and J. Gland. Surface Sci.. 159, 430 (1985). 58. M. K. Bahl. S. C. Tsai, and Y. W. Chung, Phys. Rev. E . 21, 1344 (1980). 59. D. R. Short, A. N. Mansour, J. W. Cook, Jr., D. E. Sayers, and J. R. Katzer, J . Catal., 82, 299 (1983). 60. J. M. Hermann, M.Gravell-Rumeau-Maillot, and P. C. Gravelle, J . Cafal., 104, 136 (1987).

REDUCTION 61. M. A. Vannice, L. C. Hasselbring. and B. Sen, J. Caul., 97, 66 (1986); 85, 2972 (1985). 62. P. Chow, and M. A. Vannice, J . Cafal., 104. 1 (1987). 63. P. Chow, and M. A. Vannice, J . Cafal.,104, 17 (1987). 64. S. Koobiar, J . Phys. Chern., 68, 441 (1964). 65. J. Marcq, X. Wispenninckx. G . Poncelet, D. Keravis, and J. Fripiat, J. Cafal., 73, 309 (1982). 66. J. Marcq, G . Poncelet, and J. Fripiat, J . Cafal., 87, 339 (1984). 67. D. F. Cox, T. B. Fryberger, and S. Semancik. Phys. Rev. E, 38, 2072 (1988).

109

Chapter 7 OXYGEN ON OXIDES

7.1 NATURE OF ADSORBED OXYGEN Depending on the sample history, it is possible that there are oxygen atoms, either neutral or charged, on an oxide surface that are in positions different from the positions of surface lattice oxygen ions. These oxygen atoms and ions may have different charges than the lattice oxygen, have different energics of binding (or adsorption), and be desorbed at different temperatures. Adsorbed oxygen on a stoichiometric surface of a fully oxidized oxide is readily identifiable. It is ususally desorbed at temperatures lower than the sublimation temperature of surface lattice oxygen. On a partially reduced surface, adsorbed oxygen may result in reoxidation of the surface cations to different degrees, depending on the extent of charge transfer between the reduced center and the adsorbed oxygen. If the charge transfer is such that an adsorbed oxygen atom acquires the same elcctron density as a surface lattice oxygen ion, and it occupies a lattice site, the surface is reoxidized and the adsorbed oxygen becomes a lattice oxygen. If the charge transfer is less extensive and/or the oxygen species occupies a site diffcrcnt from a surface lattice site, it is an adsorbed oxygen. Adsorbed oxygen may be present as atomic or molecular species with various charges. On transition metal oxides, the most common species are O,O-, 0 2 and , 0 2 , On some basic alkali and alkaline earth oxides, 022has been reported [l]. On UV-irradiated TiOz, 03- species has been observed with EPR [2]. Various review articlcs discussing these and other species such as 03-havc appeared in recent years [2-41. In general, atomic oxygen spccics are adsorbed more strongly than molccular species. This is because strong surface-atomic oxygen bonds are needcd to compensate for the energy required to break the double bond of thc oxygen molecule. The rate of dissociativc adsorption of oxygen to atomic species is expcctcd to be lower than that of molccular adsorption. This is because a pair of neighboring surface sites must be available for the former process. Otherwise

110

SURFACE OXYGEN

111

lattice or surface diffusion of mononuclear oxygen ions (or lattice anion vacancies) is needed before dissociation of the oxygen molecule can be achieved. This is consistent with the fact that a saturation coverage of adsorbed atomic oxygen corresponds to a few atoms per nm2 of surface (see Table 7-1). On the other hand, molecular adsorption of oxygen can take place on an isolated surface site. This requirement of the availability of a pair of surface sites can be overcome by using species such as N20 instead of 02,the decomposition of which would leave an atomic oxygen on the surface. Indeed, decomposition of N 2 0 at low temperatures resulting in an adsorbed atomic oxygen (usually detect& as 0-)has been observed on many oxides [2-41.

7.2 DETECTION OF ADSORBED OXYGEN One method to detect the presence of adsorbed oxygen is by temperature programmed desorption after exposing the oxide to 02.Presumably, adsorbed oxygen is bonded differently on the surface than lattice oxygen, and would be desorbed at temperatures different from the vaporization temperature. The temperature programmed desorption profiles of a number of oxides have been reported [5-81. The general feature of these profiles is schematically shown in Fig. 7-1. Upon heating of an oxide with adsorbed oxygen from room temperature, three types of desorption peaks are observed. Type I occurs at a relatively low temperature and it represents the most weakly adsorbed oxygen. Normally this is assigned to adsorbed molecular oxygen which usually desorbs below 300°C. Type I1 which occurs at an intermediate temperature usually results from adsorbed atomic oxygen, and desorbs below 600°C. The amounts of oxygen desorbed in type I and I1 are small, of the order of a few percent of a close-packed monolayer. Type 111occurs at high temperatures and at a rate which may increase continuously with increasing temperature. The amount of type 111 may be much larger than those of the other two types. This type results from vaporization of lattice oxygen, and desorption of metal atoms occurs simultaneously, that is, the metal oxide is subliming at these temperatures. Table 7-1 summarizes the desorption temperatures of oxygen from a number of oxides [51. Except for those noted, oxygen is adsorbed by cooling the oxide in oxygen from 600°C to 10°C. In this manner, adsorption is achieved even for species whose adsorption is activated. It is clear from the table that the schematic desorption profile shown in Fig. 7-1 is oversimplified. For many oxides, there are more than two desorption peaks below sublimation temperature. For example, there are three desorption peaks for iron oxide. From their temperatures, one is probably adsorbed molecular oxygen and two are adsorbed atomic oxygen. Titanium dioxide appears to have three forms of adsorbed oxygen. The different forms of adsorbed oxygen reflect interaction with different environment of the surface. At present, there is little detailed understanding of the differences. Attempts to gain further understanding of the adsorbed oxygen species is complicated by the fact that the extent of interaction of the adsorbed species with the surface (and perhaps also with each other) may depend on the adsorption

112

Table 7-1 Desorption Temperatures and Total Amounts of Oxygen Desorbed from Metal Oxides a

Oxides

Desorption Temp. C"

Volume Desorbed ml(STP)/m2

450 50,270, 360,540 55, 350,486 30, 165, 380 35,335,425,550 125, 390 65 100 125,1190, 250 190, 320 80, 150

0 0 0 0 0 2.13 x 6.54 x 4.05 x 3.30 x 1.12 x 1.42 x 2.05 x 2.99 x 5.52 x 2.45 x 2.11 x

v205

M003 Biz03 wo3

Biz03.2Mo03 Cr203 &02

Fez03 (3304

NiO CUO A120zb SiOz Ti0 (anatase)b ZnO?b Sn02

lop2 lop2 10-3 lop2 10-2 10-1 104 10-5 10-5 104 10-3

Fo o motes: a) From Iwamoto. et al., J. Phys. Chem., 82, 2564 (1978),coyright American Chemical Society. Oxygen is adsorbed by cooling the sample in oxygen from 600 to 10°C. b, Oxygen is adsorbed at room temperature. ') Obtained with a heating rate of 20°C/min.

conditions. For example, the temperature programmed desorption profile of oxygen adsorbed on a-Fe203 has been shown to depend on the adsorption temperature. This is shown in Fig. 7-2. The profile in this figure can be explained as follows. There are three types of adsorbed oxygen on a-Fe203.The fist type produces a desorption peak at about 490°C. The second type produces a peak at about 370°C. The peak temperatures of these two types do not vary with the adsorption temperature. However, adsorption into these forms is activated and temperatures close to the desorption temperature are needed for adsorption to occur. The desorption of the 370°C species is second order in surface coverage, suggesting that it may be an atomic species [8]. The third type produces a peak below 200°C whose exact temperature depcnds on the adsorption temperature. The desorption temperature suggests that this peak is due to adsorbed molecular

113

SURFACE OXYGEN

TFigure 7-1 Schematic temperature programmed desorption profile of oxygen born an oxide. I: from adsorbed molecular oxygen; II: from adsorbed atomic oxygen; III: from sublimation of lattice oxygen.

0)

w

m

a

C

0 .-c,

n

L

0 VJ Q)

a I

100

I

I

I

I

300 Temperature,

500

C

Figure 7-2 Effect of adsorption tcmperature on the TPD profile of adsorbed oxygen on aFe203. Adsorption tcmperatures: a. Sample cooled from 600 to 10°C in 0,; b. 10°C; c. 110°C; d. 250°C; e. 400°C. (From Bull. Chem. SOC. Jpn., 51. 2765 (1978), copyright Chemical Society of Japan).

114

Figure 7-3Room temperature EPR spectrum of adsorbed 0 2 - and Zn' on ZnO. (From J. Phys. Chem., 82. 2564 (1978),copyright American Chemical Society).

oxygen (type I). The variable temperature of the desorption peak suggests that there is a broad distribution of adsorption sites. Similar dependence of the temperature desorption profiles on the adsorption conditions has been observed also on NiO [91 and Mn02 [51. The assignment of adsorbed atomic or molecular oxygen species to the various desorption peaks has been confirmed by EPR measurements on some oxides. On ZnO, adsorption of oxygen generates an anisotropic signal with g,=2.052, g2=2.009, and g3=2.003 (Fig. 7-3). This signal is assigned to 02-.The signal disappears upon evacuation at 20O0C, which coincides with a desorption peak at 190°C. It appears that the EPR signal and the desorption peak are from the same species [ 5 ] . It is interesting that this 02-signal and the Zn' signal are complementary with each other, showing an electron transfer from Zn+ to oxygen:

On anatase TiOz, oxygen adsorption results in three EPR signals that have been assigned to 0 2 - on three different sites. These three signals disappear upon heating to temperatures that are coincident with the three temperature programmed desorption peaks [ 5 ] . Thus these three desorption peaks are from adsorbed molccular oxygen. Direct spectroscopic observation of adsorbed oxygen has been made using IR and EPR. Because of the lack of a dipole moment, the oxygen molccule in the gas phase is IR-inactive. However, it is Raman active with a vibrational stretching frequency of 1552 cm-'. Adsorbed molecular oxygen will have a lower symmetry than in the gas phase because of perturbation by the surface, and some weak IR bands have been assigncd to adsorbed 02.Some of these assignments are summarized in Table 7-2. The assignment of the charge on the adsorbed oxygen molecules is generally made by comparison with coordination complexes. Examples of adsorbed superoxide ion (023detected by EPR have been given above for ZnO and Ti02. Adsorbed superoxide ion has been detected on many

115

SURFACE OXYGEN

Table 7-2 IR Detection of Adsorbed Oxygen Molecules on Some Transition Metal Oxides

Oxide

gas phase

Adsorbed Species

0 2

02022-

NiO

singlet O2 02-

Ti02

020 2 h

cr203

0 2 022-

singlet O2 ~-FQO~

02022-

Stretching Frequency, cm-l

1552 1140 850 1500 1140,1060 1180-1060 16GO-1580 1600-1700 985 1460 1350,1325,1300,1270 1100-900

Ref.

a-c

d d e f e g d i i

References: a) J. Shamir. et al. J. Amer. Chem. SOC.,90. 6223 (1968). b) K. Nakamoto, "IR and Raman Spectra of Inorganic and Coordination Compounds," Wiley-Interscience, NY, 3rd edition, 1978. c) N. Sheppard, in "Vibrational Properties of Adsorbates." R. F. Willis, ed.. Springer-Verlag, Berlin, 1980. d) A. Tsyganenko, et al., Spect~os.Lett. 13, 583 (1980). e) A. Davydov, et al., Kinet. Catal., 14, 1342 (1973). f) A. A. Davydov, et al., Symposium on "Adsorbirovanny Kislorod v Katalize," 1972, Institute of Catalysis, Acad. of Science, USSR, Novosibirsk. Preprint No. 19. g) A. Davydov, et al., Kinet. Catal., 13, 980 (1972). h) F. Al-Mashta. et al., J. Chem. SOC.Faraday Trans. I, 78, 979 (1982).

oxides including ZnO [5,10-121, Ti02 (5,131, supported V205 [14], supported Moo3 1151, mixed oxides of COO-MgO [161, and nontransition metal oxides [3,41. This species is characterized by an anisotropic signal of g1=2.015 to 2.077, g2=2.002 to 2.012, g3=2.001 to 2.011 [3]. There is a correlation between the magnitude of g1 and the formal oxidation state of the metal ion: the higher is the oxidation statc, [he lower is the value of gl because of larger crystal field inlcractions [3,17]. Interaction of the superoxide ion with the nuclear magnetic moment of the metal ion may result in hyperfine splitting of the EPR signal [14].

116 Because of the electronegativity of oxygen, electron transfer from the oxide to adsorbed oxygen commonly occurs, although it may not be a necessary condition for chemisorption. Therefore, adsorption of oxygen becomes much more facile if the oxide is reduced. It has been shown that chromium oxide reduced at 500°C and evacuated (to dehydroxylate the surface and make some surface chromium ions coordinatively unsaturated) can adsorb oxygen strongly at as low as -195°C [MI. Similarly, molybdena reduced at 500°C and evacuated chemisorbs oxygen at -78°C and below [19]. In these two cases, adsorption is believed to be on Cr203 and M a 2 . The facile adsorption has made it possible to use oxygen chemisorption to determine the surface area of supported chromia and molybdena [20,21]. However, the stoichiometry of adsorbed oxygen and surface cation is not unity, and an assumption has to be made that this stoichiomeuy is the same for the supported and the unsupported sample whose surface area can be independently determined. The nature of the adsorption site is still a subject of investigation [22]. There is indication that the adsorption site on molybdenum oxide is a surface Mo2+ center based on the competitive behavior between O2 and NO adsorption ~231.

7.3 REACTIVITY OF ADSORBED OXYGEN Adsorbed oxygen species vary greatly in their reactivities depending on their nature and the nature of the oxide. Among species that have been identified with EPR, 0- has been shown to be very reactive. It abstracts an H atom from an alkane molecule at as low as 77 K. 0 2 - is less reactive, and it forms complexes with adsorbed alkene molecules. The reactivities of those species not detected by EPR vary. On iron oxide, for example, the strongly adsorbed oxygen appears to be unreactive, in contrast to the weakly adsorbed oxygen which degrades adsorbed alkenes rapidly to combustion products [24]. Most of the studies on the reaction of adsorbed oxygen concerns the EPRactive 0- species. The common method to generate this species is either by irradiation of an oxide with y-ray or UV light, or by decomposition of N 2 0 on the oxide. There does not appear to be any significant difference in the species produced by either method. Adsorbed 0- is a very reactive species. It reacts readily with alkanes at low temperatures on a number of oxides including ZnO, vanadia, and molybdena. Cleavage of a C-H bond and formation of OH and an alkyl radical is a common first step of the reaction. In somes cases, partial oxidation products of alkanes are formed. Examples of these reactions are described below. Figure 7-4 shows such a reaction on ZnO. Adsorbed 0- is generated by irradiating ZnO with UV light at 90°K. The presence of this species is detected by EPR (curve b). On exposure of the sample at 90 K to methane, a new EPR signal of .CH3 is detected (curve c). Thus 0- abstracts an H atom from methane according to the equation: O-(ad) + CH4

-+

.CH3(ad)

+

OH-(ad)

(7-2)

117

SURFACE OXYGEN

A

___)

H

Figure 7-4 EPR spectra taken at 90 K showing reaction of C& with adsorbed 0- on ZnO. a. ZnO pretreated in 0,; b. Adsorbed 0- on ZnO generated by UV radiation; c. Spectrum of C H 3 after exposure to C&. (From React. Kinet. C a d . Lett., 18, 243 (1981). copyright Elsevier Scientific F'ubl.)

The same reaction has been detected on silica-supported vanadia, molybdena, and tungsta [26,27]. When the reaction is conducted at room temperature on V/Si02, a small amount of C2H, formed by coupling of methyl radicals is detected in the gas phase. Upon heating the oxide, large quantities of C2H, and CO are desorbed together with small quantities of CH, and C02. Similar observations have been made on UV-irradiated Ti02. It has been proposed that the methyl species is adsorbed as a surface methoxide, at least above ambient temperature [27]:

or

CH3

+

02- + CH30-

.CH3

+

0- + CH30-

+

e-

(7-3)

The formation of ethane, however, should result from the coupling of two methyl radicals. When the adsorbed 0- on V/Si02 is exposed to a mixture of methane and oxygen at room temperature, the 0- EPR signal is immediately replaced by an 03signal. Formaldehyde becomes the major reaction product instead of ethane. On TiOz, CO and C 0 2 are the major products. It is interesting that if ' * 0 2 is present in the gas phase, the oxygen in the formaldehyde formed is not labeled [27]. Adsorbed 0-reacts readily with ethane also. On Mo03/Si02, the reaction is instantaneous and proceeds to almost completion at room temperature [28]. Upon heating, C2H, is desorbed. However, the amount of ethene observed is more than the amount of 0- detected by EPR. This may indicate the presence of undetected

118

0- which is EPR inactive because of strong magnetic dipolar interaction with the solid. 0-adsorbed on V/SiO2 also reacts readily with C2H.j. Upon heating, CI&, C2&, and carbon oxides are the major desorbed species. No coupling product, butane, is observed [271. Adsorbed 0- on Co-MgO also reacts readily with ethane [29]. Upon heating, ethene and methane are the major products. Interestingly, the total amount of ethene and methane formed is about five times the amount of 0detected, as in the case of Mo03/Si02 mentioned above. The reaction mechanism has been proposed to be: O-(ad)

+ C2&

+

C2HS(ad)+ OH-(ad)

(7-4)

C2Hs(ad) + 02-(lattice) + C2HSO-(ad) + e-(s) C2HsO-(ad)

+ 2 02-(lattice)

CH,COO-(ad)

+ OH-(ad)

+

CH3COO-(ad)

+ C&

+ C032-(ad)

(7-5)

+ H20 + 4 e-(s)

(7-6) (7-7)

In addition to alkanes, adsorbed 0- also reacts with HZ,D2, and alkenes. Reactions of adsorbed 0- with H2 or D2 are also rapid, as is that with various alkenes. At 10°C the reactivity of 0-with various molecules shows the order D2 < H2 < C21& < CO < CH, [26]. The reactivity also depends on the nature of the oxide. For H2, D2, and CI&, it decreases in the sequence V205/Si02 > Mo03/Si02> W03/Si02 [26,31]. The reaction with H2 results in the formation of adsorbed OH-. The rate shows a normal deuterium kinetic isotope effect of about four at -lOO°C. and of about two at 10°C. The reaction with CO results in the formation of C02- [26,32]. A similar associative reaction of 0-with 0 2 to form adsorbed 0 3 - has been reported [27.29,30]. On warming and evacuation, the reaction is reversible and 0-and 02(g) are formed. The reaction of adsorbed 0- with Cz& is more complex and the detail depends on the oxide. On Mo03/Si02,exposure of adsorbed 0-to C2H4 at 110 K results in a new EPR signal that has been assigned to a linear CH2CH20- species. On warming, the signal is replaced by one that has been assigned to a CHCH2 radical [33]. The assignment to a linear .CH2CH20- species is supported by deuterium and 13C labelling studies [34]. On W03/Si02, the same reaction at low temperatures leads to the appearance of a EPR signal that has been assigned to a cyclic (-CH2CH03epoxide-like species that has not been clearly identified. On warming to 90 K, the -CHCH2species is also formed 1351. When a sample of Mo03/Si02 is irradiated with W light in the presence of C2&, 0- is presumably formed by the reaction: Mo6+a2- +Mo5+4-. An EPR signal assigned to an ethylene oxide species has been observed at 77 K [361 that is formed by: C2&

+

0-(ad)

-

(H2C-

H2)\/= 0

(7-8)

SURFACE OXYGEN

119

Under this condition, a small amount of propene is also detected in the gas phase. On warming, more propene as well as some 1- and 2-butene and formaldehyde are detected. It is proposed that the adsorbed ethylene oxide decomposes on warming to form Mo-methylene species, which participate in metathesis at the elevated temperatures to produce the higher hydrocarbons. Other surface species have been observed in the reaction of ethene with adsorbed 0-.On Co-MgO, the reaction at 25°C results in a EPR signal assigned to .CCHp--OH- [25]. On ZnO at 90 K, a polymeric radical species of R4O0-45O0C) p and a low temperature a form [105]. Excess M a 3 is required for high selectivity for acrolein in the Co-Mo-0 system. It has been reported that excess Moo3 stabilizes the a modification [ 1061 whereas excess Co304stabilizes the p form [107]. Although the matter has not been discussed, it is also possible that excess Moo3 is the active site for the reaction, as is observed in the Ni-Mo oxide system [108]. It has also been reported that with Ni-Mo oxides, only samples containing more Moo3 than needed for NiMo04 produce acrylic acid in propene oxidation [110]. When excess Moo3 is present in Co-Mo-0, acrolein can be produced from propene in the absence of gaseous oxygen [lo91 until an equivalent of one monolayer of lattice oxygen is consumed, The acrylic acid yield, however, decreases rapidly with increasing degrees of catalyst reduction, and the rate of reaction is slower in the absence than in the presence of gaseous oxygen. The dependence of the activity and selectivity on the surface crystallographic orientation of Moo3 has been studied recently. Volta, et al. have studied the reaction over a series of Mo03/graphite catalysts prepared with different calcination times and temperatures [ 111,1121. These catalysts have different

SELECTIVE OXIDATION I

193

distributions of exposed crystal faces. A linear correlation is observed between the yield ratio of acroleiqK02 and the ratio of (100)/(010) surface areas. It is concluded that acrolein is produced on the (100) face of Moo3, and C02 is produced on the (010) face. Other explanations of these data have been advanced [ 113,1141. In one report, it was concluded that acrolein was produced on the (010) face, and combustion, on the (100) and the (101) faces. This latter assignment is supported by the observation that the yield of acrolein in the oxidation of ally1 halides is proportional to the surface area of the (010) face of Moo3 [114]. Another example of crystal-face specificity has been reported for the multicomponent Mnl -x4xV2-2xM02x0,jcatalyst [ 1151.

Effect of Modifiers to Molybdates Various modifiers have been used to further improve the activity and selectivity of bismuth molybdates. These have led to the development of the multicomponent catalysts. However, detailed understanding of the chemical effects of the additives is lacking. In many cases, only observations of the effect are reported. When Ce is substituted for some Bi in Bi2(M00~)~, high catalytic activities in propene ammoxidation are observed at Bi/Ce ratios that correspond to the maximum solubility of Bi in Ce(Mo04) and of Ce in Bi2(Mo04),. A high activity is also observed at a Ce concentration that correspnods to the point of equal solubility of Ce in Bi molybdate and Bi in Ce molybdate [116,117]. Substitution of Bi by La is less effective than by Ce [95]. This is interpreted as due to the fact that Ce undergoes redox (e- + Ce4+ +Ce3+) more readily than La or Bi, and more effectively facilitates the reoxidation and reconstruction of the Bi-Mo-0 phase. Thcre are reports that addition of Fe203, CrzO3, CuO, Te02 or Se02 facilitates the formation of the a and p phases of bismuth molybdates during the preparation of the catalysts [118]. Addition of BiP04, F%MqO12,or Cr2M03012 to Bi2M03012also increases the formation of the a phase [119]. The current industrial multicomponent catalyst, Me,F%B&M%PiK,O, where Me = Ni, Co, or Mg is essentially a Me-Bi-Fe molybdate promoted by P and K. In some recent publications, it is suggested that the primary components of this catalyst are femc molybdate and Bi3(Fe04)(M00~)~ [ 120,1211. Alkali metals are promoters in the supported Moo3 system. At low alkali metal concentrations,the promoting effect decreases as Cs> Rb > K > Na > Li [92]. This trend is inversely related to the electronegativity of the alkali metal. Addition of V [I221 or Sn [123] to Moo3 does not produce a selective catalyst for acrolein in propene oxidation. On the other hand, Te02-Mo03 is very active and selective both for propene oxidation to acrolein [124,125] and for ammoxidation [126]. Ce-Mo-Te oxide is also an active and selective ammoxidation catalyst. A catalyst prepared by coprecipitation of all the components consists of essentially a ternary (Ce,Mo,Te) oxide, a-Ce2Mo4OI5 and/or pCe2M03013[127]. Ni-Mo-0 and Cu-Mo-0 have also been reported to have reasonable selectivities [128,129,130]. In the Co-Mo-0 system, a small amount of Te has been found to be an

194

effective promoter in the oxidation of both propene [128] and butene [131]. Te is found to be enriched at the surface [105]. Addition of Fe, Bi and V also increases the activity and selectivity in acrolein production [105]. In general the Co-Mo-0 system produces more acrylic acid than the Bi-Mo-0 system.

11.6 EFFECT OF WATER ON PROPENE OXIDATION When water is added to a feed of propene and oxygen, 2-propanol is often formed either as the major product or as the initial product which is then dehydrogenated to acetone. On a Mo03/Al203 catalyst [132], acetone is produced with 83% selectivity at 300°C. If H21S0 is used, the l 8 0 atom is incorporated into acetone. This contrasts sharply the situation for acrolein production where the proximate source of the oxygen atom in acrolein is a lattice oxide ion. If D20 is used, the deuterium atoms rapidly exchange with the hydrogen atoms of the end carbons of the reactant propene to form (DIH)3C-CH=C(D,H)2. The hydrogen at the center carbon does not undergo exchange. When deuterated propene is used, no kinetic isotope effect is observed if the deuterium atoms are at the end carbons, as would be expected if the exchange of these atoms with hydrogen atoms in water is rapid. A kH/kD ratio of 2.2 is observed if the deuterium atom is at the center carbon. This behavior of the kinetic isotope effect is opposite that observed in acrolein production. It has been proposed that the reaction proceeds via a carbenium ion intermediate [132]:

+

MoS+ H+(ad)

H20 -+ MoS+OH-

+

C3H,

+

H+(ad)

+ C3H7+(ad)

C3H7+(ad) + Mo5+OH- + Mo5+ C3H70H

+

(11-10)

1/202 + (CH3)ZCO

+

C3H70H

+ H2O

In this mechanism, a surface coordinatively unsaturated MoS+is assumed to be the active site. However, there is no direct evidence for the presence of MoS+ in the catalyst. Whether desorbed 2-propanol is a significant product is questionable at temperatures higher than about 30O0C when the equilibrium concentration of 2propanol is small. Under these conditions, it is possible that acetone is formed from a surface alkoxide [ 1341: H3qHp-h C&

+

bI

H 0 J

F===

?*

-H -+

(CH&CO

+

H 0 J

(11-11)

SELECTIVE OXIDATION I

195

The mechanism in eq. (1 1-10) explains the data over a Sn-Mo oxide catalyst at 150°C [133]. On this catalyst, 2-propanol has been shown to be the only product at low conversions. Acetone begins to appear as the conversion increases. The reaction rate at high conversions is suppressed by methyl ethyl ketone. In the presence of water, propene can also be oxidized to acetone on a VzOs catalyst at 200°C [135]. As the temperature increases, the acetone yield decreases, whereas the yields for acetic acid and C02 increase. Over a V-P oxide (VP about unity), selective production of acrylic acid is observed. Addition of Te suppressess the selectivity for combustion such that at 80% conversion, a combined selectivity to acrylic acid, acetic acid, and acrolein in excess of 60% is obtained. Decreasing the water partial pressure in the feed decreases the selectivity for acrylic acid and acetic acid, but increases that for acrolein [135].

REFERENCES 1. P. Mars and D. van Krevelen. Chem. Eng. Sci. Suppl.. 3,41 (1954). 2. B. Grzybowska, and A. Mazurkiewicz, Bull. Acad. Polon, Sci. Ser. Sci. Chim. 27, 149 (1979). 3. G. W. KeuLks, and L. D. Krenzke, Proc. 6th Intern. Cong. Catal., London, Paper B-20. 1976; J . Catal., 61, 316 (1980). 4. Y. Moro-oko. Y. Takita, and A. Ozaki, J . Catal., 27, 177 (1972). 5. W. Ueda, Y.Moro-oka, and T. Ikawa, J . Chem. SOC. Faraday Trans. I, 78. 495 (1982). 6. R. K. Grasselli, Appl. Cafal.. 15, 127 (1985). 7. R. K. Grasselli, J. F. Brazdil, and J. Burrington, Proc. 8th Intern. Cong. Catal.. 5, 369 (1984). 8. R. S. Liu, M. Iwamoto. and J. H. Lunsford, J . Chem. SOC. Chem. Commun., 78 (1982). 9. H. F. Lin, R. S . Liu, K. Y.Liew, R. E. Johnson, and J. H. Lunsford, J. Amer. Chem. SOC., 106, 4117 (1984). 10. L. Mendelovici. and J. H. Lunsford, J . Catal., 94. 37 (1985). 11. M. B. Ward, M. J. Lin. and J. H. Lunsford, J . Catal.. 50. 306 (1977). 12. Y.Takita. Y. Moro-Oka. and A. Ozaki, J . Catal., 52, 95 (1978). 13.T. Ono, T. Ikehata, and Y. Kubokawa, Bull. Chem. SOC.Jpn., 56, 1284 (1983). 14. W. E. Slinkard and P. B. DeGroot, J . Catal., 68, 423 (1981). 15. A. W. Sleight, and W. J. Linn, Ann. New York Acad. Sci., 272 22 (1976). 16. J. F. Brazdil, L. C. Kartisek. R. K. Grasselli. J. Catal.. 81, 142 (1983). 17. W. M. H. Sachtler. G. J. H. Dorgels, V. Farenforf and R. J. H. Voorhoeve. Proc. 4th Intern. Cong. Catal., 1, 355 (1970). 18. I. Ziolkowski, J . Catal.. 80, 263 (1983). 19. I. Haber, J. Jenas. M. Schiavello, and R. J. Tilley, J . Catal., 82, 395 (1983). 20. R. K. Grasselli, and T. D. Bumngton, Adv. Catal., 30, 133 (1980). 21.M. Ai, and T. Ikawa, J . Catal.. 40, 203 (1975).

196 22. R. A. Zhdan. A. P. Shepelin, Z. G . Osipova, and V. D. Sokolovski, J . Catal., 58, 8 (1979). 23. J. M. Tatibouet, J. E. Germain, and J. C. Volta, J. Catal., 82, 240 (1983). 24. J. M. Tatibouet, and J. E. Germain. J. Catal., 72. 375 (1981). 25. F. Tdiro, Chim. Ind. (Milan), 56, 835 (1974). 26. P. C. H. Mitchell and F. Trifiro, J. Chem. SOC. (A), 3183 (1970). 27. K. Mori, M. Inomata, A. Miyamoto. and Y. Murakami, J. Phys. Chem.. 87, 4560 (1983). 28. M. Iwamoto,Y. Yoda, N.Tamazoe, and T. Seiyama, J. Phys. Chem., 82, 2564 (1978). 29. B. L. Yang, M. C. Kung. and H. H. Kung, J. Catal., 89, 172 (1984). 30. G. Keulko, and Z. Yu, Proc. 8th Intern. Cong. Catal., Vol. III, p. 289, 1984. 31.M. C. Kung, W. H. Chang, and H. H. Kung. J . Phys. Chem., 83, 1737 (1979). 32. R. K. Grasselli. and J. D. Burrington, Ind. Eng. Chem. Prod. Res. Devel., 23,393 (1984). 33. J. Haber, in "Solid State Chemistry in Catalysis", ed. by R. K. Grasselli and J. F. Brazdil, American Chemical Society Symp. Series no. 279. 1985. p. 1. 34. R. S. Liu, M. Iwamoto,and J. H. Lunsford, J. Chem. Soc..Chem. Commun., 78, (1982). 35. M. M. Khan, and G. A. Somorjai, J. Catal., 91, 263 (1985). 36. H. F. Lin, R.-S.Liu. K. Y. Liew. R. E. Johnson, and J. H. Lunsford, J . Amer. Chem. Soc.,106,4117 (1984). 37.K. J. Zhen, M. M. Khan, C. H. Mak,K. B. Lewis. and G. A. Somorjai, J . Catal.. 94, 501 (1985). 38. M. Iwamoto,Japanese patent JP5892630. 39. H. H. Kung, Ind. Eng. Chem. Prod. Res. Develop., 25. 171 (1986). 40. J. L. Callahan, and R. K. Grasselli. AZChE J., 9. 755 (1963). 41.F. J. Berry, Adv. Catal., 30, 97 (1981). 42. J. D. Burrington, C. T. Kartisch. and R. K. Grasselli. J . Catal., 81,489 (1983). 43. J. D. Burrington, C. T. Hartisch, and R. K. Grasselli, J. Catal., 87, 363 (1984). 44.C.R. Adams, and T. J. Jennings, J. Catal., 2, 63 (1963). 45. C. R. Adams, and T. J. Jennings, J. Catal., 3, 549 (1964). 46. C. C. McCain, G. Cough, and G. W. Godin, Nature (London). 198, 989 (1963). 47.W. M. H. Sachtler. Rec. Trav. Chim. Pays-Bas., 82, 243 (1963). 48. W. M. H. Sachtler, and N. K. DeBoer, Proc. 3rd. Int. Cong. Catal., I. (1965) p.252. 49. H. H. Voge, C. D. Wagner, and D. P. Stevensen. J. Catal., 2, 58 (1963). 50. R. K. Grasselli, and D. D. Suresh, J . Catal., 25. 273 (1972). 51. G. W. Godin. C. C. McCain, and E. A. Porter, Proc. 4th Int. Cong. Catal., I, (1971) p.271. 52.N. W. Cant, and W. K. Hall, J . Cdal.. 22. 310 (1971). 53.N. W. Cant, and W. K. Hall, J. Phys. Chem., 75. 2914 (1971). 54. W. Martir, and J. H. Lunsford, J . Amer. Chem. Soc., 103, 3728 (1981). 55.D. Driscoll, and J. H. Lunsford. J. Phys. Chem., 89, 4415 (1985). 56. J. D. Burrington. C. T. Kartisch, and R. K. Grasselli, J. Org. Chem., 46,1877 (1981). 57. C. R. Adams, Proc. 3rd Intern. Cong. Catal. I , 1965, p. 240.

SELECTIVE OXIDATION I

197

58.J. M. Peacock, A. J. Parker, P. G. A s h o r e . and J. A. Hockey. J . Cafal., 15, 398 (1969). 59. R. K. Grasselli, and J. D. Burrington. Ind. Eng. Chem., Prod. Res. Devel., 23. 393 (1984). 60. B. Grzybowska, J. Haber, and J. Janus, J. Calal., 49, 150 (1977). 61.M. Imachi, R. L. Kuczkowski, J. T. Groves, and N. W. Cant, J. Cotal., 82. 355 (1983). 62. H A . Choi, J.-T. Lin, and R. L. Kuczkowski, J. Cafal.,99, 72 (1986). 63. J. D. Burrington, C. T. Kartisch, and R. K. Grasselli, J . Cafal.,63, 235 (1980). 64. C. Daniel, and G. W. Keulks, J . Catal., 24, 529 (1972). 65. C. W. Keulks, M. P. Rosynek, and C. Daniel, Ind. Eng. Chem. Prod. Res. Dev., 10, 138 (1971). 66. J. D. Krenze. and G. W. Keulks, J. Cafal.,61. 316 (1980). 67. G. W. Keulks, and M. P. Rosynek, Amer. Chem. Soc.Div. Petrol. Chem. Prepr., 14, C55 (1969). 68. A. P. Gershkov, J. K. Kolchin, I. M. Gribov, and L. Ya Margolis, Kinef. Katal., 9, 1068 (1968). 69. J. D. D~i.scoll,and J. H. Lunsford, J. Phys. Chem., 87. 301 (1983). 70. J. L. Callahan, R. K. Grasselli, E. C. Milberger, H. A. Strecker, I d . Eng. Chem. Prod. Res. Dev., 6 , 134 (1970). 71.M.V.C. Sastri, B. Viswanathan, V.C. Bhuvana, Proc. 7th Inter. Cong. Catal., Tokyo, (1980). B-32. 72. Y. Moro-oka, S. Tan, and A. Ozaki, J. Catal., 12, 291 (1968). 73.0. M. Vinogradov, G. F. Vitnov, I. V. Luiksaar, and 0. V. Altshuler, Kinet. Katal., 16, 671 (1975). 74.H. E. Swift, J. E. Bozik. and J. A. Ondrey. J. Catal., 21. 212 (1971). 75. K. Bruckmann. J. Haber. and J. Janus. Bull. Acad. Pol. Sci., Ser. Sci. Chim., 20, 813 (1972). 76.T. Osubo, H. Mirura, Y.Morikawa, T. Shirasaki, J . Cafal..36. 240 (1975). 77.K. Aykan, D. Halvorson. A. W. Sleight, D. B. Rogers, J. Catal., 35, 401. (1975). 78. H. Miura, Y. Arai. K. Sugiyama, and T. Matsuda. J . Cafal.,68, 264 (1981). 79. G. W. Keulks, J. Catal., 19. 232 (1970). 80.R. D. Wragg, P. G . A s h o r e , and J. A. Hockey, J. Cafal.. 22. 49 (1971). 81.K. M. Sancier, P. R. Wentrcek. and H. Wise, J. Catal., 39, 141 (1975). 82. J. R. Christie, D. Taylor, and C. C. McCain, J. Chem. SOC.Faraduy Trans. I , 72, 334 (1976). 83. P. Pendleton, and D. Taylor, J. Chem. Soc. Faraduy Trans I , 72, 1114 (1976). 84. Y. Moro-oka, Y. Takifa, and A. Ozaki, J. Cafal., 27. 177 (1972). 85. M. Akimoto, M. Akiyama, and E. Echigoya, Bull. Chem. Soc.Jpn., 49. 3367 (1976). 86. B. L. Kugler, and R. J. Kokes, J. Cafal..32, 170 (1974). 87. B. L. Kugler, and J. W. Gryder, J . Catal., 44, 126 (1976). 88. J. M. Peacock, M. J. Sharp, A. J. Parker, P. G. A s h o r e , and J. A. Hockey, J. Catal., 15, 379 (1969). 89. W. Linn. and A. W. Sleight, J. Cafal.,41, 134 (1976).

90. D.A. G . van Ocffclcn. J. H. C. Van Hoolf, and G. C. A. Schuit. J . Carol., 95. 84 (1985). 91. T. Osuba. H.Miura. Y. Morikswa, T. Shirosaki. J. C dal., 36. 240 (1975). 92. M. Akimoto. and E. Echigoya. J. Coral., 35, 278 (1974). 93. E. V. Hocfs. J. R . Monnicr. and G . W. Keulks, J. C d a h , 57, 331 (1979). 94. R. K. Grassclli, Appl. Carol., 15, 127 (1985). 95. R. K. Grasselli. J. F. Rraxdil, J. D.Burrington. Infern. Cong. C u d . , 8rh. V (1984) p. 369. 96. F. Trifiro. P. Cent014 I. Pasquon. P. Jiru. Prcx. 41h Irucrn. Cong. Carol., 1. 310 (1969). 97.P. C. H. Mitchell, and F. Trifiro. 1. Uhem. Soc. A , 3183 (1970). 98. L. Burlamncchi, G. Martini, and F. Trifiro, 1.C u d . . 30. 393 (1973). 99. F. Trifiro. C. Capuio, and P.L. Villa, J . Less Common Meroh, 36. 305 (1974). 100. Dh. A. Hatist. C. J. Kapteyus. B. C. Lippcns. and C.C. A. Schuit, J. CafuL, 7, 33 (1967). 101. G.Rlassc. 1. Inorgvi. Nuclear Chem., 28, 1124 (1966). 102. D.B. Dadyburjor, S. S. Jewur. and E. Ruckenstein, Catal. Rev., 19. 293 (1979). 103. R. K. Grassclli, J. B. Burrington, and J. E. Brardil. 1.Chem. Soc.,Fmoday Disc., 72, 203 (1982). 104. I. Marsurra R. Schul K. Hirakawa. J. Cdd., 63, 152 (1980). 105. B. Crzybowska. A. Miv.urkicwicc and J. Sl~~zynski, Appl. Carat.. 13, 223 (1985). 106. T. C. Alkhazav, K. Yu A&hamov, and N. Kh. Allakhvcrdovc, Kinel. Karol., 15, 1492 (1974). 107. P. Boutny, J.C. Daumas, R. Montarnal. P. C o d n c , and C. Pannclicr, Bull. Soc. C h h . F r m e (1 968) 481 1 . 108. U . Ozkan, and G . L. Schreder, 1. Cuhl., 95, 120 (1985). 109. B. Crzybwska, and A. MazurkiewicL Bull. Acad. Polon. Sci. Ser. Sci. Chbn.. 27. 149 (1979). 110. C . Mwzocchia. F. DiKenzo, P. Centola, R. Del Rosso. in “Proc. 4th Inrrernniional Conierence in rhe Chmisrry and Uses oj Molybdenum“, Climax Molybdenum Co., Coldcn. Colo. 1982. 1 1 1 . J. C. Volla. M. Forissicr, F. Thmbald. and T.P. Pham, 1. C k m . SOC., Farad. Disc.. 72, 225 (1981). 112. J.C. Volta. and 8 . Moraweck, J. Chem. SOE.Chcrn. Commw.. (1980) 338. 113. J. Ziokowski, 1. Card, 80, 263 (1983). 114. K. 8riickmw K . Grabowski. 1. Hah. A. Ma~urkiewicr, J. Sloczynoki, and T. Wiltowski. 1. Curd.. 104. 71 (1987). 115, J. Xiotkowski. and J. Janass. 1. Caful., 81, 298 (1983). 116. J . F. Brazdil, and R. K . Grussclli. 1. Coral.. 79, 104 (1983). 117. 1. F. Rwdil, L.C. Glaesa. R. G. Tcllcr and R. K. Grassclli, Reprints ACS Division of Perroleurn Chemisrry, 28, 1285 (1983). 118. M . Ai. and S. Sumki. 1. Cord., 30. 362 (1973). 119. Ph. A. Batist. C. G.M. van der Moesdijk, I. Matsurra, and G.C . A. Schuit, 1.Carol., 20, 40 (1971). 120. O.V. Krylov. Yu. V. Maksirnov, and L. Ya. Margolis, J . Caiol.. 95. 289 (1985).

SELECTIVE OXIDATION I

199

121. T. S. R. Prasada Rao, and K. R. Krishnamurthy, J. Carol., 95, 209 (1985). 122. T. Ono, Y. Kubokawa, Bull. Chem. SOC.Jpn., 55, 1748 (1982). 123. T. 0x10, T. Ikehara, Y. Kubokawa, Bull. Chem. SOC.Jpn. 56, 1284 (1983). 124. Y. Amaud, J. Guidof J.Y. Robin, M. Romand, J. E. Germain, J. Chim. Phys.. 73, 651 (1976). 125. T. V. Andrushkevich. G.K. Boreskov. L.L. Kuznetsova, L.M. Plyasova, Y. N. Tyurin, and Y. M. Shchekochikhin, Kinet. Katal., 15,424 (1974). 126. J. C. J. Bart, and N. Giordano, J. Cafal.,64, 356 (1980). 127. J. C. J. Bart, N. Giordano, Ind. Eng. Chem. Prod. Res. Dev.,23, 56 (1984). 128, Ph. Jaeger, and J.E. Germain. Bull. SOC. Chim. France, 11-12, I(1982) 407. 129. J. Haber. Kinet. Carol.. 21, 100 (1980). 130. C. Mazzochia, P. Centola, R. Del Crosso, G. Terzaghi, and I. Pasguon, Chim. I d . , 55, 687 (1973). 131. P. Forzatti, P.L. Villa, D. Ercoli. G.Gasparini, and F. Trifiro, Id.Eng. Chem. Prod. Res. and Devel., 16, 26 (1977). 132. N. Giordano. A. Vaghi, J. Bart, and A. Castellan. J. Catal., 38, 11 (1975). 133. Y. Takita,Y. Moro-oka, and A. Ozaki, J. Catal., 52, 95 (1978). 134. J. Buiten. J. Catal., 10, 188 (1969); 13, 373 (1969); 27, 232 (1972). 135. M. Ai. J . Catal., 101, 473 (1986).

Chapter 12 SELECTIVE OXIDATION CATALYSIS I1

12.1 SELECTIVE OXIDATION OF BUTENES The selective oxidative dehydrogenation of butenes to butadiene (eq. 12-l), and the selective oxidation of butenes to maleic anhydride (eq. 12-2) are both industrially important reactions: C4Hg

+

C4Hg

+

300-4000c>

3 0 2

400-5000c)

c4H, + H20 C4H203

+

3H2O

(12-1) (12-2)

Other important reactions of butenes include the production of acetaldehyde and acetic acid catalyzed by promoted vanadium oxides supported on rutile Ti02 [l] and other vanadates, and the oxidation of 2-methylpropene (isobutene) to methacrolein (2-methylprop-l-en-3-al)or ammoxidation to methacrylonitrile (2cyanopropene) catalyzed by molybdates and cuprous oxide:

C4Hg

=@ 0 2

CH3COOH,

CH3CHO

(12-3)

( 12-4a)

(12-4b)

Selective oxidative dehydrogenation (eq. 12-1) can be carried out on many catalysts including molybdates. vanadates. ferrites, uranium-antimony oxide, tin200

SELECTIVE OXIDATION II

a01

antimony oxide and other antimonates. Selectivities for butadiene in excess of 85% have been obtained using these catalysts. Maleic anhydride can be produced selectively on catalysts based on vanadium oxide and/or molybdenum oxide, especially those promoted by phosphorous. A selectivity for maleic anhydride of higher than 70% has been reported in a number of patents [2].

Reaction Mechanism i) Oxidative Dehydrogenation: It is likely that the oxidative dehydrogenation of butene to butadiene and water involves stepwise abstraction of two hydrogen atoms: (12-5)

This mechanism has been proposed by analogy to propene oxidation in which dissociative adsorption of propene to form adsorbed x-ally1 is well established (see Chapter 11). Adsorbed x-bonded butene (C4H8(ad))and x-ally1 (C4H7(ad))on ferrites have been observed using infrared spectroscopy at room temperature [3,4]. It has also been shown that on ferrites, adsorbed butadiene is produced below 200°C. and the desorption of butadiene is rate limiting [5,6]. However, data at ordinary catalytic temperatures of about 350°C are not available. Butene isomerization accompanies dehydrogenation,but there are few studies on this reaction during oxidation. Isomerization by reversing the allylic hydrogen abstraction step (step I1 in eq. 12-5) is a possibility. On MgF%04 [8] and CoF%04 [9], it has been proposed that isomerization and dehydrogenation take place on separate sites [7] based on the fact that when a mixture of trans-2-C4D8 and trans-2-C4H8 is oxidized over the catalysts, there is no H/D isotope mixing found in butadiene. Such mixing is expected if step I1 is readily reversible on the oxidation sites. Isomerization also shows a smaller deuterium isotope effect than oxidation [8,9], but its implication on the mechanisms has not been explored. ii) Oxygenate Formation: The primary route for the oxidation of butene to maleic anhydride has been postulated to be [lo]: (12-6)

(ad or g)

\

/-nd

202

Other possible intermediates that have been suggested include 4 5 dihydrofuran and 2.5-dihydrofuran [11,121. Methyl vinyl ketone, acetic acid, acetaldehyde, and carbon oxides are among the common side products [13]. This mechanism is supported by the following results using V-P oxides (V/P 1. 1/1) as catalysts. When used as the reactant feed under similar reaction conditions. butene, butadiene, crotonaldehyde, and furan all yield substantial amounts of maleic anhydride [13,14]. The selectivity for maleic anhydrik increases in the sequence: butene c butadiene c furan. In the oxidation of 1-butene, substantial conversion begins at -220°C. The major product below 250°C is butadiene which becomes negligible at 280°C. As the temperature increases, the yield of maleic anhydride increases and reaches a maximum at 300-330°C. This supports butadiene as both an intermediate and a precursor for maleic anhydride. The CO and C02 production increases with increasing temperature continuously. Small amounts of furan, acetaldehyde, methyl vinyl ketone, and crotonaldehyde are also observed between 250-330°C [141. The product distribution on the V-P oxide also depends on the oxygen partial pressure. At low butene/02 ratios or at low conversions, maleic anhydride and COXare the only detectable products. At high 0 2 conversions, other intermediate oxidation products appear, including butadiene [141. In the absence of gaseous oxygen, butadiene is the major product until the catalyst is too extensively reduced to be active [15,161. Selective production of mdeic anhydride from butene, butadiene, and furan (with selectivity increasing in this order) has also been reported for NiMo04 containing 15% excess Moo3 [17,18]. The reaction sequence (12-6) probably applies to this catalyst.

Kinetics i) Oxidative dehydrogenation: On most catalysts, the oxidative dehydrogenation of butene to butadiene shows a positive order in butene and zero order in oxygen. In addition, a negative order in butadiene, indicative of product inhibition, is often reported, especially at lower temperatures. The order in butene depends both on the temperature and the pressure. It usually decreases with increasing butene pressure or decreasing temperature. For example, on bismuth molybdate, the rate is first order in butene over the temperature range 343-500°C [19]. The rate depends on the butene isomer. At 460°C. the rate for 1-butene is several times greater than for cis-2butene, which is greater than for trans-2-butene [19]. Below 4OO0C, butadiene inhibits the reaction. In contrast, water or carbon dioxide has little effect on the reaction. The same kinetics is observed on Fe-promoted Bi-Mo oxide [20]. Unlike bismuth molybdates or ferrites, the reaction on the scheelite Pbl -3xBi2x$x(Mo04) system shows a zeroth order in butene and a positive order in oxygen [211. The formation of butadiene shows a deuterium isotope effect. The ratio of rate constants kH/?.D is 3.9 at 300°C and 2.6 at 400°C on MgFe204 (8), and 2.4 at 430°C for CoFe204 [9]. The large isotope effects indicate that the breaking of C-H (C-D) bonds is involved in the slow step of the reaction. This conclusion is

SELECTIVE OXIDATION LI

203

supported by the relative reaction rates of various substituted butenes [22]. On aFe203, the relative rates at 270°C are 1.3 : 1.0 : 0.9 : 0.7 for

F

7

C I C=C-C

0

I

250

300

350

Temperature,

C

Figure 12-1 Effect of reaction temperature on: a. The selectivity of main products (butadiene BD, maleic anhydride MA, and COX)in 1-butene oxidation; b. The valence state of vanadium. Catalyst V-P-0 with V/P = 0.99 containing 100% VOV) before use; feed 0.6 % 1-butene, 12 % 02.(Ind. Eng. Chem. Prod. Res. Devel., 24, 221 (1985). copyright American Chemical Society).

VOP04 with HCl or alcohol, it contains vanadyl pyrophosphate [15]. p-VOP04 is a slightly less selective catalyst than a-VOP04. Both p-VOP04 and (V02)P207 have been tested in the pure form to produce maleic anhydride, and are found to interconvert slowly during butene oxidation [53]. It has been proposed that the active catalyst consists of p-(VO)2P2O7 and aVOP04 which can readily interconvert by reduction and oxidation under reaction conditions [52,57]. This interconversion is regarded as crucial in the reaction mechanism because it interconverts V(IV) and V(V). The active p-(VO)2P207 phase may be nonstoichiomeuic. At high temperatures under oxidizing

208

C4H

Figure 12-2 Catalytic cycle of butene oxidation on a V-P-0 catalyst.

conditions, a-VOp04converts to p-VOP04, which cannot be easily reconverted to the pyrophosphate phase. This causes a loss of the redox ability of the catalyst and the low selectivity. A more recent crystallographic study [56] of the V-P-0 system shows that vOHp04.#20 is the precursor prior to calcination. If VOHP04-#120 is obtained by reduction in an organic medium, a 6-VOPO, form is formed on calcination which successively converts to y and p-VOP04. Under nitrogen, yVOP04 is reduced reversibly to y-(V0)2P207. Repeated cycles of oxidationreduction slowly converts this y-VOP04/y-(VO)2P20,couple into the p-VOP04/p(V0)2Pz07 couple. Calcination of VOHP04.+ H20obtained in an aqueous medium produces aVOPo4 which exists in two forms: a1 and all.a1 is converted to p-(V0)2P20, by reduction at 76OoC,a11is converted to y-(V0)2P207 when heated under nitrogen at 700°C. Crystallographic considerations shows that the interconversion of y-VOP04 to y-(VO)2P207requires very small displacements of atoms [56]. The reduction of pVOW4 to form p-(VO)2P207 involves crystallographic shear planes, and would be more difficult. Thus, p-VOP04, once formed, makes the catalyst less selective. p-(VO)2P207contains double chains of edge-sharing VO6 octahedra linked by pyrophosphate (V2@4-) units. The axial V-O bond lengths within an octahedra are not equivalent. One of the V-O bonds can be viewed as a vanadyl ( V 4 ) group [59]. An IR band at 963 cm-' assigned to V=O has been observed on this compound. A V=O stretching band at 1001 cm-I has also been observed on p-VOP04 [60].

SELECTIVE OXIDATION II

209

The presence of V-0 might be important. There is a correlation between the density of V - 0 groups and the activity in butene and butane oxidation on unsupported V2O5 catalysts [47]. The activity per V=Q depends on the method of preparation of V205.Interestingly, butene is oxidized to maleic anhydride with about 45% selectivity on the catalysts, but butane is oxidized only to carbon oxides. In the NiMo04-Mo03system, excess Moo3 is found to be essential for the selective production of maleic anhydride from butene at 480°C. Pure Moo3 or pure NiMoO, produces butadiene at about 20-25% selectivity but no maleic anhydride. NiMo04 containing excess Moo3, however, produces maleic anhydride with up to 35% selectivity [17,611. A similar synergistic effect is also observed for the selective oxidation of butadiene to maleic anhydride [18]. Spectroscopic investigations lead to the conclusion that the selective catalysts are Moo3 particles whose surfaces are decorated with NiMoO, particles that effectively block the combustion sites of Moo3. Furthermore, NiMo04, being effective in producing butadiene, also serves to supply butadiene to Mo03 which produces furan and maleic anhydride more selectively from butadiene than from butene. The production of maleic anhydride on Moo3 is crystal-face specific [29,31]. A correlation has been found between the ratio of the (lOO)/(OlO) areas and the ratio of maleic anhydride to C 0 2 production in the oxidation of 2-methylpropene.

Effect of Promoters There are claims that many promoters are effective, and summaries of these are found in a number of review articles and monographs [2,7,61a,62]. Unfortunately, there is relatively little understanding of the promoting effects. In the oxidative dehydrogenation of butenes on ferrites, the promoting effects of Mg, Cr, Co, Zn, Bi, Mo, P, Sb, and V have been studied. The addition of Cr, Mg, or Zn increases the resistance of the ferrites to reduction [35]. In addition, Co, Zn and Mg form spinel ferrites with Fe2O3, which is a more desirable structure than corundum for this reaction. The promoting effect of Mo, V and P has been interpreted as due to changes in the acidity and basicity of the oxide [41,63]. This interpretation is based on the idea that butene and butadiene are both Lewis bases as indicated by their rather low ionization potentials. The Lewis acidity or basicity of a selective catalyst must not be too strong so that it is sufficiently acidic to adsorb butene, yet sufficiently basic to prevent strong adsorption of butadiene which would poison the active sites. Addition of Mo, V and P increases the acidity of iron oxide, which is presumably too basic. In the case of Sb and As, the promoting effect is a result of the formation of FeAs04 and FeSb04 [41]. The effect of the support in supported F%03 has been investigated [64]. Si02, Ti02, A1203 and their mixed oxides have been used as supports, and the activity per gram of catalyst varies widely over two orders of magnitude, but the selectivity for butadiene remains in the range 65-88%. The V-P oxide catalyst deactivates by the loss of phosphorous during operation. This problem can be reduced by operating at a lower temperature or by

210

stabilizing the composition of the catalyst with the addition of alkali metals. Li has been claimed to be effective for this purpose [2]. The rate-limiting step in butene oxidation on this catalyst is believed to be C-H bond breaking. Thus promoters such as Fe, Cr, Mg, and Zn that have dehydrogenation activities have been reported to enhance the catalytic acitivity. In a number of studies on vanadium oxide catalysts [65-681, it is reported that there is an optimal concentration of promoters beyond which the selectivity decreases rapidly. This may be due to possible changes in the crystalline phases and/or in the redox potential of the vanadium ions in the solid when too high concentrations of promoters are added. This interpretation is consistent with the fact that the effects of some promoters depend on the feed composition, especially the hydrocarbon to oxygen (or air) ratio and the conversion. Perhaps different gas phase compositions are needed to obtain the optimal distribution of oxidation states of vanadium in the presence of different promoters. Alkali promotion has also been reported for vanadium oxide. In one report, sodium addition significantly increases the selectivity for butadiene on a V205/SiOzcatalyst but decreases the selectivity to maleic anhydnde [69].

12.2 SELECTIVE OXIDATION OF BUTANE In all studies reported, butane is less reactive than butene and higher reaction temperatures are required. It is reasonable to expect that the oxidation of butane proceeds via dehydrogenation to form butene, which is then reacted to form butadiene, maleic anhydnde, or other oxidation products as discussed in the last scction. It is important to note, however, that because butane is less reactive and requires a higher temperature for reaction, some catalysts such as ferrites that are highly selective for butene oxidation are not as selective for butane oxidation. The most often studied reactions of butane are the oxidative dehydrogenation to butenes and butadiene (eq. 12-8) and oxidation to maleic anhydride (eq.12-9):

Nature of Catalysts A number of oxides have been reported to be quite selective in oxidative dehydrogenation. Among these are NiO, Moo3, and c0304 [70], and mixed oxides of Co-Mo, Ni-Mo, Mg-Ni [71], and Mg-V [72]. A selectivity for dehydrogenation in excess of 60% at a conversion of 20% or more has been obtained with some of these mixed oixdes. Among the mixed oxides, Mg-Mo oxide and Mg-V oxide have been studied quite extensively. High selectivities (60+%) for dehydrogenation (butenes and

SELECTIVE OXIDATION II

211

butadiene) on Mg-Mo oxide have been observed both in the presence [71] and in the absence of gaseous oxygen in the feed [73]. Analyses of the selective catalysts show that they are mixtures of MgMo04, MgO, and an unidentified phase. The selectivity for oxidative dehydrogenation on a Mg-V oxide is as high as 80% at low conversions [74]. It decreases to about 50% when the conversion increases to 40% [72]. Detailed analyses of the catalyst show that the active phase is magnesium orthovanadate (Mg3(V04)2). This conclusion is confirmed by the similarity in the catalytic behaviors of a stoichiomeuic Mg orthovanadate and a mixed Mg-V oxide. Interestingly, neither Mg pyrovanadate (Mg2VzG) nor Mg metavanadate (MgV206) is selective for this reaction [74]. On the other hand, other orthovanadates including LuV04 and SmV04 are about as selective as Mg3(V04)2 [751. The differences among the various vanadates have led to the suggestion that the bonding of oxygen in the oxide lattice is a factor that determines selectivity. The oxygen ions in an orthovanadate are bridged between a vanadium ion and another cation such as Mg, Lu, or Sm. If this other cation is not easily reducible, the bridging oxygen ion cannot be readily removed from the lattice, and the catalyst is selective. In a metavanadate or a pyrovanadate, there are oxygen ions bridging between two vanadaium ions. Since V5+ ions are easily reducible, the bridging oxygen ions are readily removable, and the catalysts show high combustion activities. The selective oxidation of butane to maleic anhydride is quite demanding and only a few catalysts have been reported to have high selectivities. Molybdenumphosphorous oxide and V-P-0 are among the few. Of these two, V-P-0 has been very extensively studied as described in section 12.1. Selectivities as high as 78% for maleic anhydnde at low conversions of butane and 65% at high conversions have been reported in the patents [2]. Although the active phases in V-P-0 in butane oxidation are believed to be the same as those in butene oxidation, some differences have been reported between the behavior of this catalyst in the two reactions. It has been found that although p-VOP04 transforms into (V0)2P207 during butene oxidation, the transformation is much slower during butane oxidation [53]. The areal activities of V-P-0 catalysts prepared either by aqueous reduction of V2O5 with H3P04or by reduction with H3P04 in an organic medium are the same in butene oxidation, but are different in butane oxidation [76]. This observation may indicate different rate-limiting steps in the two reactions. The participation of lattice oxygen in a V-P-0 catalyst in butane oxidation has becn demonstrated. It has been found that only the first one to five layers of lattice oxygen are involved [14,15]. This is very different from the behavior of bismuth molybdates in propene oxidation (see Chapter 11). It has been porposed that the active surface of V-P-0 is a layer of VOP04, and scrambling of lattice oxygen occurs rapidly, creating a pool of five exchangeable oxygen atoms per surface V ion [ 5 8 ] . When one of the five oxygen atoms is consumed in the reaction, the vacancy is filled by reoxidation by gaseous oxygen. The replenished oxygen atom scrambles rapidly with the remaining four oxygen atoms in the surface layer. This picture is consistent with the observation that in a pulse experiment, 90% of the

212

reaction products contain oxygen atoms already present in the catalyst, and 10% contains oxygen from gaseous oxygen. Finally, if the catalyst is first treated with D20, nearly 2f3 of the product maleic anhydride contains two deuterium atoms in a pulse experiment. Thus the reaction proceeds with exchange of H atoms in the surface intermediate (such as a x-allyl) and surface OD groups.

Mechanism The mechanism and the kinetics of butane oxidation to maleic anhydride on a V-P-0 catalyst have been studied. By studying the deuterium isotope effect at different positions of a butane molecule on the p-(VO)2P2@ catalyst, it has been demonstrated that the first step of the reaction is the production of adsorbed butyl species [58]. At 400°C. a sizable isotope effect, kH/kD of 2.1 to 2.2 is observed for C-H (C-D) bonds at the secondary carbons of a butane molecule, but there is almost no isotope effect at the two primary carbons. The k H / k D values of 2.1 to 2.2 are close to the theoretical maximum value of 2.27 for a homolytic dissociation of a C-H bond based on the difference in zero point energy of the C-H and C-D bonds, which is 4.4 kJ/mole. Thus the dissociation of a methylene C-H bond of butane to form an adsorbed sec-butyl species is the rate-limiting step in the reaction. There is no information on the subsequent steps. It is likely that the formation of an adsorbed butyl species is followed by a second C-H bond breaking which results in an adsorbed butene molecule. The butene molecule then reacts in the same manner as described in the last section. An in situ FTIR study identified maleic acid on the catalyst together with some unidentified alkenic species, in addition to butane, adsorbed maleic anhydride, and adsorbed carbon dioxide [1201. Little mechanistic information is provided by the kinetics [58,77-791.

12.3 SELECTIVE OXIDATION OF METHANE The selective oxidation of methane is one of the most challenging processes because the C-H bond in methane is the strongest among all hydrocarbons, and because the selective oxidation products are generally more reactive than methane such that high yields of products are difficult to obtain. This reaction has received considerable attention recently because of the abundant reserve of natural gas in the world, and two recent review articles have appeared on this subject [80,81]. In the review by Pitchai and Klier [81], different methods to oxidize methane are mentioned. These methods include partial oxidation to methanol or formaldehyde, oxidative dimerization to acetylene, ethene, and ethane, oxidative methylation, and others. The first two processes, oxidation of methane to methanol or formaldehyde and oxidative dimerization will be discussed here.

Formatwn of Methanol or Formaldehyde A large number of oxides can act as catalysts for the oxidation of methane to methanol or formaldehyde using oxygen [80,81]. Oxides of zinc, lead, nickel,

SELECTIVE OXIDATION 11

213

chromium, copper, iron, silver, cobalt, molybdenum, manganese, vanadium as well as mixed oxides containing these have been claimed to be active and selective. In general, however, the conversions are low, thus the yields of methanol and formaldehyde are low. Among these oxide, vanadium oxide, and especially molybdenum oxide are the more thoroughly investigated. The first extensive report on the use of Moo3 and mixed oxides of Moo3 with Fe2O3, ZnO, U02, or V 0 2 is in a patent by Dowden and Walker [82]. In a highly oxygen-lean feed of about 97% CH, and 3% oxygen, and at a low methane conversion of a few percent, a selectivity of methanol of up to 75% is achieved with Mo03-U02. 3MoO3-F~O3shows a methanol selectivity of 6396, and a formaldehyde selectivity of 8%. A later patent claims CuO-Moo3 to be about as effective as the other molybdates [83]. Selective oxidation to methanol and formaldehyde at a much higher conversion is reported by Iwamoto [84] and by Lunsford and coworkers on silicasupported Moo3 catalysts using N 2 0 as the oxidizing agent [85]. In the latter report, CH30H and formaldehyde are produced with 55.1% and 23.9% selectivity, respectively, at a conversion of 12.5%. The feed is 72 torr CH,, 277 torr N 2 0 and 266 torr H20, the reaction temperature is 580°C and the catalyst is 1.7 weight % Mo03/SiOz. Water is found to be essential for the high selectivity to the formation of partial oxidation products. Higher partial pressures of water increase the selectivity. At low water partial pressures, the activity increases with increasing water partial pressure, but becomes constant for P H > ~ 75 tom. ~ Later work show that below 5OO0C, methanol is formed almost exclusively [86.87]. Above 5OO0C, HCHO selectivity continues to increase and HCHO becomes the major partial oxidation product above about 550°C. Up to this temperature, no CO or C 0 2 is detected even at a methane conversion of 1.9%. At higher temperatures, CO formation becomes considerable. At 580°C. the selectivity to partial oxidation ratio in the feed, and reaches 100% products increases with increasing PcH4/P~20 when the ratio is 1.2. Although the conversion to methanol increases as PHz0 increases from 0 to 260 torr, the conversion to HCHO stops increasing for P H ~ ~ larger than 40 torr, and the conversion to CO decreases slightly. The activity of the catalyst also increases with PHZoup to 50 tom. beyond which there is no further effect. The steam reforming of methane and the water-gas shift reactions are negligible under these reaction conditions. Selective production of formaldehyde is also observed using oxygen as the oxidant over a M a 3 catalyst supported on silica-alumina [81]. Formaldehyde, CO and C02 are the only products reported. No methanol is formed. The selectivity for HCHO increases with increasing CH4/02 ratio, reaching a value of 46% at 600°C and a CH,/02 ratio of 17. As with N20 as the oxidant, addition of water to the feed is beneficial. The selectivity for HCHO increases with the amount of water added, although the CH, conversion decreases. At 6OO0C, a HCHO selectivity of over 90% is achieved with water in the feed and a C&/O2 ratio of 9. From these reports, a number of interesting differences is observed between using 0 2 and using N 2 0 as the oxidant. Methanol is not produced using O2 but is produced using N20. With increasing temperature, the selectivity for the partial

214

oxidation products increases using 0 2 , but decreases using N2O. The CH,J02 ratio required to obtain high selectivity to partial oxidation is much higher than the CH,/N20 ratio. The need to have water in the reaction mixture for high selectivity could limit the long term stability of the molybdenum oxide catalysts. Water reacts with Moo3 to form a volatile molybdenum hydroxide species 1811:

This could lead to loss of molybdenum under reaction conditions. At 540°C. the production of CH30H in CH, oxidation with N 2 0 on Moo3/ Si02 is roughly first order in CH, and H20 and zeroth order in N 2 0 [87]. The production of HCHO is zeroth order with respect to all reactants. At 580°C. the overall conversion of CH, is zeroth order in CHq and first order in N20 [86]. For some reasons, the rate constant is lower for P N ~ Ti02 (rutile) = zirconia > Ti02 (analase) [116,117]. In general, the selectivity decreases rapidly with increasing temperature because of increase in conversion. When VzOs is used in a solid solution with other oxides (A1203,Sn02, or Ti02), the selectivity for HCHO is very low at low vanadium concentrations. At high concentrations, the selectivity increases up to almost 100% [118]. V2O5 supported on A1203 or TiOz shows poor selectivity. On V2OS/Al2O3,dehydration dominates; on V205/ri02, combustion dominates [ 1191.

REFERENCES 1. W. E. Slinkard, and P. B. DeGroot, J. Cafal.,68.423 (1981). 2. R. K. Hodnett. Caul. Rev., 27, 373 (1985). 3. E. N. Amirbekov. F. V. Aliev, L. I. Lafer, A. M. Rubinshtein. and V. I. Yakevson, Izv. Akad. Nauk SSSR, Ser. Khim., 7. 1344 (1979). 4. E. N. Amirbekov. F. V. Aliev, L. I. Lafer. A. M. Rubinshtein, and V. I. Yakevson. Izv. Akad. Nauk. SSSR, Ser. Khim.. 11, 2250 (1979). 5. M. C. Kung, W. H. Cheng. and H. H. Kung. J. Phys. Chem., 83. 1737 (1979). 6. H. H. Kung. B. Kundalkar, M. C. Kung, and W.H. Cheng, J. Phys. Chem., 84. 382 (1980). 7. H. H. Kung and M. C. Kung. Adv. Cafal..33, 159 (1985). 8. M. A. Gibson. and J. W. Hightower, J . Cafal.,41, 420 (1976). 9. W. R. Cares, and J. W. Hightower. J. Cafal.,23, 193 (1971). 10. F. Cavani, G. Centi, I. Manenti, and F. Trifiro. Ind. Eng. Chem. Prod. Res. Dev.. 24. 221 (1985). 11. G . Busca, F. Cavani, G. Centi. and F. Trifire, J . Cafal., 99, 400 (1986). 12. G. Busca, G. Centi, and F. Trifiro, Appl. Cafal., 25. 265 (1986). 13. M. Ai. Bull. Chem. SOC. Jpn.. 43, 3490 (1970). 14. G . Centi, G. Fornasari, and F. Trifiro, J. Cofal.,89, 44 (1984). 15. F. Cavani, G. Centi, and F. Trifiro, Appl. Cafal.. 9, 191 (1984). 16. D. Brkic. and F. Trifiro, Ind. Eng. Chem. Prod. Res. Dev., 18, 333 (1979). 17. U. Ozkan, and G. L. Schrader, J. Cafal., 95, 137 (1985).

224 18. U. Ozkan, and G. L. Schrader. J. Catal.. M,147 (1985). 19. C. Adams, H. Voge, C. Morgan, and W. Armstrong, J. Catal., 3, 379 (1964). 20. Ph. A. Batist, C. G. M. van de Maesdijk, I. Matsuura, and G. C. A. Schuit, J . Catal., 20, 40 (1971). 21. A. W. Sleight, and W. J. Lm,Ann. New York Acad. Sci., 272, 22 (1976). 22. M. Misono, K. Sakata, F. Ueda. Y. Nozawa, and Y. Yoneda, Bull. Chem. Soc. Jpn., 53, 648 (1980). 23. W. J. Linn, and A. W. Sleight, J. Catal., 41. 134 (1976). 24. Ph. Batist. B. Lippens, and G. C. A. Schuit, J. Catal.. 5. 55 (1966). 25. I. Matsuma, and G. C. A. Schuit, J . Catal., 20, 19 (1971). 26. P. Sunderland, Ind. Eng. Chem. Prod. Res. Devel., 15, 90 (1976). 27. R. L. Varrna, and D. N. Saraf,J. Catal., 55, 361 (1978). 28. W. Ueda, Y. Moro-oka, T. Ikawa. J. Chem. Soc.Faraday Trans I . 78. 495 (1982). 28a. M. Ai, P. Bontry, R. Montarnal, and G. Thomas, Bull. Sm. Chim. Fr., (8-9), 2783 (1970). 29. J. M. Tatibouet, Ch. Phichitkul, and J. E. Germain, J . Catal., 99, 231 (1986). 30. F. Cavani, G. Cemti, I. Manenti, A. Riva, and F. Trifiro, Ind. Eng. Chem. Prod. Res. Devel.. 22, 565 (1983). 31. J. C. Volta, J. M. Tatibouet, C. Pitchitkul, and J. E. Germain, Proc. 8th Intern. Congr. Catal.,(Berlin 1984), Dechema, Frankfurt, IV,451 (1984). 32. B. L. Yang, M. C. Kung. and H. H. Kung, J . Catal.. 89, 172 (1984). 33. R. J. Rennard, and W. L. KeN, J. Catal., 21, 282 (1971). 34. M. Misono. Y. Nozawa. and Y. Yoneda, Proc. 6th Intern. Cong. Catal., 1976, I, 386 (1977). 35. F. E. Massoth, and D. A. Scarpiello, J. Cataf., 21, 294 (1971). 36. E. H. Lee, Catal. Rev., 8, 285 (1974). 37. B. L. Yang. F. Hong. and H. H. Kung. J. Phys. Chem., 88, 2531 (1984). 38. G. B. Raupp, and W. N. Delgass, J . Catal., 58, 337 (1979); M. C. Hobson. and H. M. Gager, J. Catal.. 16. 254 (1970). 39. J. A. Amelse, J. B. Butt, and L. H. Schwartz, J. Phys. Chem., 82, 558 (1978). 40. D. Patel, MS Thesis, Northwestern University , 1985. 41. I. Matsuma, Proc. 6th Intern. Congr. Catal., London, Paper B-21. 1976. 42. I. Kozlov, Izv. Akud. Nauk. SSSR, Ser. Khim. Nauk., 3, 10 (1983). 43. V. Fenin, V. Shvets, and V. Kazanski, Kinet. Katal., 20. 952 (1979). 44. A. Simakov, and S. Veniaminov. React. Kinet. Catal. Lett., 28, 67 (1985). 45. R. Fricke. W. Hanke, H. Jerschkewitz, B. Parlitz, and G. Ohlmann. Appl. Catal., 9, 235 (1984). 46. F. Cavani. G. Centi, I. Manenti, and F. Trifiro. Ind. Eng. Chem. Prod. Res. Dev.. 24, 221 (1985). 47. K. Mori, A. Miyamoto, and Y. Murakami, Appl. Catal., 6, 209 (1983). 48. B. K. Hodnett, and B. Delmon, Appl. Catal.. 9. 203 (1984). 49. B. K. Hodnett, Ph. Permanne. and B. Delmon. Appl. Catal., 6. 231 (1983). 50. E. Borde, and P. Courtine, J . Catal., 57, 236 (1979). 51. L. Morselli, A. Riva, F. Trifiro, M. Zucchi. and G. Emig. Chim. Ind. (Milan), 10. 791 (1979). 52. G. Poli, I Resta, 0. Ruggeri. and F. Trifiro, Appl. Catal., 1, 395 (1981).

SELECTIVE OXIDATION 11

225

T. P. Moser, and G . L. Schrader. J . Cataf., 104, 99 (1987). Y. N. Shyr, and G . L. Price, Ind. Ens. Chem. Prod. Res. Devef.,23, 536 (1984). R. L. Varam, and D. N. Saraf. Indian Chem. Eng., 20, T44 (1978). E. Bordes, and P. Courtine, J . Chem. Soc.Chem. Commun., 294 (1985). G. Centi, I. Manenti. A. Riva, and F. Trifiro, Appf. Cufal.,9, 177 (1984). M. A. Pepera, J. L. Callahan. M. J. Desmond, E. C. Milberger, P. R. Blurn, and N. J. Bremer, J. Amer. Chem. Soc.. 107, 4883 (1985). 59. Y. E. Gorbunova, and S . A. Linde, Sov. Phys. Dokl., 24, 138 (1979). 60. T.P. Moser, and G.L. Shrader, J. Cataf.,92, 216 (1985). 61. U. Ozkan, and G . L. Schrader. J. Cataf., 95, 120 (1985). 61a. D. J. Hucknall, "Selective Oxidation of Hydrocarbons", Academic Press, London, 1974. 62. D. B. Dadybiujor. S . S. Jewar. and E. Ruckenstein. Catal. Rev., 19, 293 (1979). 63. M. Ai. J . Cataf.,52, 16 (1978). 64. E. Rodenas, T. Iizuka, H. Katsumata, and K. Tanabe, React. Kinet. Cataf.Len., 19, 341 (1982). 65. M. Brutovsky, and S . Gerej, Czech. Chem. Commun., 47, 403 (1982). 66. M. Brutovsky, S. Gerej. F. Vasilco, and J. Gerejova. Czech. Chem. Commun., 47, 1290 (1982). 67. B. K. Hodnett, and B. Delrnon, J . Chem. Soc., Chem. Commun., 373 (1983). 68. B. K. Hodnett, and B. Delmon, Appf. Cataf..6. 245 (1983). 69. R. Fricke, W. Hanke. H.-G. Jerschkewitz, B. Parlitz. and G . i)hlmann, A p p f . Card., 9. 235 (1984). 70. V. A. Levin, T. V. Vernova, and A. L. Tsailingol'd, Kinet. Kataf., 13, 504 (1972). 71. G. A. Stepanov, A. L. Tsailingol'd, V. A. Levin, and F. S . Pilipenko, Stud. Surf. Sci. Catal., 7 (part B, New Horizon Catal.), 1293 (1981). 72. M. A. Chaar, D. Patel, M. Kung, and H. H. Kung. J . Cataf.,105, 483 (1987). 73. V. A. Doroshenko, L. P. Shapovalova, A. I. Zelenina, D. N. Tmenov. 0. M. Taranukha, and I. S . Lushnik, Kinet. Catal., 24, 749 (1983). 74. D. Patel. M. Kung, and H. Kung, Proc. 9th Intern. Cong. Cufaf..4, 1554 (1988). 75. D. Patel, PhD thesis, Northwestern University, Evanston, 1988. 76. G. Busca, G. Centi. F. Triliro, and V. Lorenzelli, J . Phys. Chem.. 90 1337 (1986). 77. A. Escardino. C. Sola, and F. Ruiz. An. Quim., 69, 385 (1973). 78. F. Cavani, G. Centi, and F. Trifiro, Appl. Cataf., 15, 151 (1985). 79. K. WoNfahrt, and H. Hofmann, Chem. Ing. Techn., 52, 811 (1980). 80. N. R. Foster, Appf. Cataf.,19, 1 (1985). 81. R. Pitchai, and K. Klier, Catal. Rev., 28. 13 (1986). 82. D. A. Dowden, and G . T. Walker, U.K. Patent 1,244,001 (1971). 83. H. J. F. Stroud, U.K. Patent 1,398,385 (1975). 84. M. Iwamoto, Jpn. Patent 58 92,629 (1983). 85. R. S. Liu, M. Iwamoto. and J. H. Lunsford. J . Chem. Soc., Chern. Commun., 78 (1982). 86. H. F. Lin, R. S. Liu. K. Y. Liew, R. E. Johnson, and J. H. Lunsford, J. Amer. Chem. SOC., 106, 4117 (1984). 87. M. M. Khan, and G. A. Somorjai. J. Cataf..91, 263 (1985). 88. K. J.Zhen. M. M. Khan, C. H. Mak,K. B. Lewis and G . A. Sornorjai, J . Cafaf., 94, 501 (1985).

53. 54. 55. 56. 57. 58.

226 89. M. Iwamoto, Jpn. Patent 58 92,630 (1983). 90. K. Otsuka, K. Jinno, and A. Morikawa, J. Catal., 1000, 353 (1986). 91. T. Ito, J.-X. Wang, C.-H. Lin, and J. H. Lunsford, J. Amer. Chem. Soc., 107, 5062 (1985). 92. K. Otsuka. K. Jinno, and A. Morikawa, Chem.Lett.,499 (1985). 93. C. H. Lin, K. D. Campbell, J.-X. Wang, and J. H. Lunsford, J. Phys. Chem., 90, 534 (1986). 94. I. Matsuura, Y. Utsumi, M. Nakai. and T. h i , Chem. Lett., 1981 (1986). 95. J. A. Sofranko, J. Leonard, and C. A. Jones, J. Catal.,103, 302 (1987). 96. C. A. Jones, J. J. Leonard, and J. A. Sofranko,J. Catal.. 103, 311 (1987). 97. R. Pearce, and W. R. Patterson, "Catalysis and Chemical Processes." John Wiley and Sons, NY 1981, p. 263. 98. C. J. Machiels, in "Catalysis under Transient Conditions," ACS Symposium series, no. 178, ed. A. T. Bell, L. L. Hegedus, 1982. p. 239. 99. C. J. Machiels, and A. W. Sleight, Proc. 4th Intern. Conference Chem. Uses of Molybdenum, H. F. Barry and P. C. H. Mitchell, ed., 1982, p. 411. 100. T.-J. Yang, and J. H. Lunsford, J. Catal., 103, 55 (1987). 101. N. Pernicone, F. Lazzeria, G . Liberti, and G.Lanzavecchia, J . Catal.. 14. 293 (1969). 102. C. J. Machiels, and A. W. Sleight, J. Catal., 76, 238 (1982). 103. U. Chowdhry, A. Ferretti. L. E. Firment, C. I. Machiels, F. Ohuchi. A. W. Sleight, and R. H. Staley. Appl. Surjme Sci., 19, 360 (1984). 104. M. Carbucchio, and F. Trifiro, J. Catal.,45, 77 (1976). 105. J. N. Allison. and W.A. Goddard, III, J. Catal.,92. 127 (1985). 106. M. Ito, Vib. Surf. [Proc. Intern. C o d . 2nd, 19801, publ., 1982, p.71. 107. R. P. Groff, J . Catal., 86, 215 (1984). 108. N. Pernicone. C. Liberty, and L. Ersini, Proc. 4th Intern. Cong. Catal. 1969 p.366. 109. B. Popov, K. Osipova. V. Malakhov, and A. Kolchin. Kinet. Catal., 12. 1464 (1971). 110. N. van Truong, P. Tittarelli, and P. Villa, Proc. 3rd Int. Conference Chem. Uses ofMolybdenwn. H.F. Bany, P. C. H. Mitchell, ed.. 1979, p.161. 111. G. Alessandrini, L. Cairati, P. Forzatti. P. Villa, and F. Trifiro, J . Less-Common Met.. 54. 373 (1977). 112. J. M. Tatibouet, and J. E. Germain, J. Catal., 72, 375 (1981). 113. J. M. Tatibouet. J. E. Germain, and J. C. Volta, J . Cafal., 82, 240 (1983). 114. J. M. Tatibouet, and J. E. Germain. J . Chem. Res.. ( S ) , 268 (1981); (M), 3070 (1981). 115. J. M. Tatibouet, and J. E. Germain. CR. Acad. Sci.. Paris C. 290, 321 (1980). 116. F. Roozeboom. P. D. Cordingley, and P. J. Cellings, J. Catal., 68. 464 (1981). 117. A. J. van Hengstrom, J. G. van Ommen, H. Bosch, and P. J. Gellings, Proc. 8th Intern. Congr. Catal., IV, 297 (1984). 118. P. J. Pornonis, and J. C. Vickerman, Faraday Disc.,Chem. SOC., 72, 247 (1982). 119. J. Kuenski. A. Baker. M. Glinski, P. Dollenmeier, and A. Wokann. J. Catal., 101, 1 (1986). 120. R. W. Wenig, and G.L. Schrader, J . Phys. Chem.. 90, 6480 (1986).

Chapter 13

CATALYTIC REACTIONS BETWEEN HYDROGEN AND CARBON OXIDES

13.1 INTRODUCTION CO and C02 can be hydrogenated on transition metal oxides to alcohols and hydrocarbons. Since the oxides of the Group VIII transition metals have lower hydrogenation activities, they may show higher selectivities for higher hydrocarbons and alkenes. Transition metal oxides also have a lower tendency to dissociate CO than their metals. Thus, they often produce alcohols more selectively. In addition, some transition metal oxides produce branched products quite selectively, especially 2-methylpropane, 2-methylpropene, and 2methylpropanol. However, in some of the systems that have been extensively studied, such as Cu-Zn oxide, rhodium or palladium oxide, it has been controversial as to whether the active components are the metal or the oxide. The water-gas shift reaction, which is the reaction between carbon monoxide and water to form hydrogen and carbon dioxide, also involves hydrogen and carbon oxides. This reaction is readily reversible, and the reverse reaction, known as reverse shift reaction, is between hydrogen and carbon dioxide. Various aspects of earlier work on these reactions catalyzed by oxides have been rcviewed [l]. For example, it has been discussed in these earlier reviews that catalysts based on ZnO, such as Zn-Cr oxide are effective in the hydrogenation of CO to methanol, which can also be catalyzed by copper catalysts. Higher alcohols can be produced by promoting these catalysts with alkali ions. On the other hand, Z r O 2 and Tho2 are effective for the production of branched products. Since these earlier reviews, new catalysts have been discovered and significant progress in understanding the reaction mechanisms has been made. The discussions here will emphasize the more recent developments. 221

228

13.2 ALCOHOL SYNTHESIS ON COPPER-ZINC OXIDE AND OTHER CU-BASED CATALYSTS Methanol Synthesis Copper-zinc oxides are very active and selective catalysts for the production of methanol from a feed mixture containing CO, C 0 2 and H2 [2-51. Highly selective catalysts can be made with a C a n ratio between about 3/7 to 7/3. Often the catalysts are suppported, and A 1 2 0 3 is the support in the commercial catalyst. At present, it is controversial as to whether the active phase is highly dispersed copper metal supported on ZnO or Cu(1) ion in a ZnO matrix. This will be discussed further later. Highly active and selective catalysts to produce methanol from a CO and H2 mixture have been reported using other copper compounds. In particular, rare earth-copper intermetallics appear to be quite effective 121. For these catalysts, the active phase is believed to be metallic copper. There has been a large volume of work reported on the study of this reaction, and a very comprehensive review has been published [2]. The readers should refer to this review for detailed discussions regarding this reaction. Methanol can be produced by the hydrogenation of CO or C02: CO CO2

+ +

2H2 3 H2

+ CH30H

(13-1)

+ CH30H +

H2O

(1 3-2)

Industrially a mixture of CO, C02 and H2 is used as the feed. The high activity of Cu-Zn oxide has made it possible to synthesize methanol at a lower temperature (c30O0C)and a lower pressure (c10 MPa) than those used for the older generation of catalysts such as Zn-Cr oxide.

Kinetics and Mechanism of Methanol Synthesis Typical industrial reaction conditions for a copper-zinc oxide catalyst are 200 to 275°C and 5 to 10 MPa total pressure. A typical feed consists of a mixture of CO (8 to lo%), C02 ( 5 to 6%), and the balance being Hz. At this temperature, the methanol synthesis catalyst is also an active catalyst for the water-gas shift reaction: CO

+

H20

. I - C02

+

H2

(13-3)

The rate of methanol production depends on the partial pressures of the various components. On Cu-Zn oxide, the rate from a C02/H2 mixture is faster than from a C0/H2 mixture [6-81. In a mixture of CO, C02, and H2, the ratc depends both on the CO/C02 ratio and the conversion. At low conversions, the rate increases with increasing C02 content [6]. At high conversions, the rate first increases rapidly with the addition of a small amount of C02 to a HdCO feed, but decreases on further addition beyond an optimal concentration [9,10]. This behavior is shown in Fig. 13-1. A small amount of water in the feed also enhances

CO HYDROGENATION

229

2

0 m

30

a

c

20

z W

0

a W

10

a

0

I

I

10

20

PERCENT C o p IN SYNGAS,

30 CO+co1=3O

Figure 13-1 Effect of COz partial pressure on the rate of methanol production on a Cu-Zn oxide catalyst. (From J. Catal.. 74, 343 (1982), copyright Academic Press).

the rate of methanol production from a mixture of Hz and CO (Fig. 13-2) [ll], although high concentrations actually suppresses the rate [6,11]. The behavior shown in Fig. 13-1 can be interpreted possibly by two factors. The first is that methanol is formed faster by the hydrogenation of COz than by CO. Thus addition of C02 enhances the rate at low C02 partial pressures. The apparent suppression of the rate at high C02 partial pressures is probably due to limitation by the approach to chemical equilibrium. However, it may also be due to poisoning of the catalyst by carbonate formation, oxidation of the catalyst, or the presence of a high partial pressure of water due to the reverse water-gas shift reaction and the hydrogenation of COz. The behavior shown in Fig. 13-2 can be similarly explained. In the absence of water, CO hydrogenation is the only reaction possible. Addition of a small amount of water results in the formation of C02 by the water-gas shift reaction, and the rate of methanol production is enhanced. Excess water results in suppression of the rate because water competes effectively for the active sites. That C 0 2 is the main source of carbon for methanol has been confirmed by isotope labeling experiments, first by Rozovskii and coworkers [12-141, and later by others [15-171. These results can be summarized as follows. If the carbon atom of a carbon dioxide molecule in a feed is labeled, and the feed is passed over a

230

0

20

40

Water Addition,

60

80

mol/h X l o 3

Figure 13-2The dependence of methanol synthesis rate on the quantity of water added to the binary H&O = 70/30vol% synthesis gas with GHSV = 6120 l(STP)/kg cat/hr at 7.5 MPa and temperature of 1: 190DC,2: 215"C, 3: 22SoC, and 4: 235°C. Catalyst weight = 2.45 g. (From F'roc. 8th Intern. Congr. Catal., II. 47 (1984), copyright Elsevier Scientific h b l . ) .

Cu-Zn oxide catalyst under conditions that are close to the industrial ones, the carbon label of the methanol molecule is identical to that of carbon dioxide when thc conversion is low (high space velocities). As the conversion increases (lower space velocities), the carbon label begins to appear in carbon monoxide, and the fractions of carbon dioxide and methanol in the reactor outlet that contain the label decrease. These results are shown in Fig. 13-3 [15,16]. Since water is formed in the experiments shown in Fig. 13-3 by reactions (13-2) and (13-3), these results suggest that the presence of a small amount of water has little effect on the main source of carbon for methanol. However, there are indications that a large amount of water suppresses the rate of hydrogenation of C02 such that the contribution from CO hydrogenation becomes significant [17]. The effect of C02 on the rate of methanol synthesis from CO differs on other catalysts. On CuCr02, addition of a small amount of C02 to a CO and H2 feed actually decreases the rate of production of methanol by about 25%,probably due to competition by C02 for the adsorption site of H2. Further addition does not show any more effect [18]. On a Cu/Th02 catalyst, addition of C02 to the C0/H2 feed

CO HYDROGENATION

231

a

0

5 0.8 m

CT

\

\

\

\

.

CO

,exit \

\

%

inlet CO

0

-- ----

-I

'1

1'0

2b

Space Velocity (h-lX Figure 13-3 Variation of specific radioactivity of the products with space velocity in the reaction of CO:C02:H2= 10:10:80 over Cu/Zn/Al/oxide. Total pressure = 5 MPa, 25O"C, and the feed contained 14C02. (From Appl. Catal., 30, 333 1987). copyright Elsevier Scientific h b l . ) .

has no effect on the rate of methanol production, although it prevents the deactivation of the catalyst that occurs at low HJCO ratios [19]. On 21-0~. methanol is produced as a minor product from a HJCO feed only if a small amount of water is also present [20]. In this case, it is believed that water is needed to form surface hydroxyl groups which react with CO to form surface formate species, which are believed to be intermediates in the formation of methanol on this oxide. The detailed mechanism of methanol formation on a Cu-Zn oxide catalyst has been investigated by a number of workers, but no agreements have been reached. An earlier infrared study has reported the observation of two bands at 2270 and 2661 cm-' which have been assigned to a formyl species [21]. In a later in situ study, these two bands were not observed. Instead, a surface formate, which can be hydrogenated to adsorbed formaldehyde and then to adsorbed methoxy was observed [22]. This surface formate could be formed by the hydrogenation of a surface carbonate. A carbonyl species with a vibrational frequency of 2090 cm-' was also observed which was assigned to CO adsorbed on copper. A surface formate adsorbed on copper metal has also been identified in the reaction of COz with H2 [231. The presence of surface formate has also been confirmed by chemical trapping with dimethyl sulfate [7]. The presence of surface formyl and formaldehyde species are inferred from results of chemical trapping with diethylamine, methyldiethylamine [ 111, and isopropylamine [24]. In these experiments, the chemical trapping agent is injected into the feed stream of CO and H2 under reaction conditions for methanol synthesis. The extra products are then

analyzed so as to deduce from them the nature of the adsorbed species that have reacted with the trapping agent, For example, in the study using isopropylamine as the trapping agent, methylisopropylamine is formed. Since condensation between methanol and the amine is slow under the reaction condition, it is proposed that methylisopropylamine is formed by the reaction of isopropylamine with adsorbed formaldehyde:

H R1R2N C h H I M

1.

R 1 R2NCH3

+

-H R1R2NCH20H

+

H2

M H20

H2O

____)

+

(13-4)

M

11-H20

The result of this experiment therefore supports the presence of surface formyl and/or adsorbed formaldehyde species, which is formed by the reaction of adsorbed CO and adsorbed hydrogen. Based on these results, a likely mechanism for the formation of methanol from C 0 2 is as follows:

* +

4 +

-

H

H

I

COz(ad)

+

*

+

-+H

CH3 I 0

J

*

+H 2 -*-OH

H > C& 4

(13-5)

+H -4

CH30H

Nature of Copper-Zinc Oxide Catalyst in Methanol Synthesis There are two proposals for the nature of the active component in copper-zinc oxide. Thcy arc Cu(1) ions stabilized in a matrix of ZnO, and highly dispersed copper metal on a ZnO support. In the former model, the Cu(1) ions are proposed to be sites for the adsorption of CO, whereas the ZnO matrix is for the adsorption of H2. This model was first advanced by Klier [3,25]. The data that support this

CO HYDROGENATION

233

model include the following: a) Optical diffuse reflectance results show that the active catalysts possess a new absorption band that is not characteristic of ZnO, copper metal, or cupric oxide. b) Electron microscopic and surface analyses show intimate contact between copper and ZnO. c) The active catalysts contain a copper phase that is amorphous to X-ray diffraction. d) The methanol synthesis activity has been reported to be directly correlated to the amount of strongly adsorbed CO [3], and only Cu(1) species are known to adsorb CO strongly [26]. In one in-situ infrared spectroscopic study of the catalyst using CO and H2, a carbonyl species at 2130 cm-I was observed whose intensity was linearly proportional to the methanol synthesis activity [27]. In another study, a band at 2093 cm-I was observed that appears to be involved in the synthesis reaction [lo]. These band frequencies suggest adsorption on an oxidized form of copper. e) XPS studies of an active catalyst show that Cu(1) is stabilized by ZnO [28]. f ) CuCr02, which contains Cu(I), is active in methanol synthesis [18,29]. In this catalyst, the activity could be correlated to the amount of Cu(1) detected by X P S . In the second model, copper metal is believed to be the active component. Zinc oxide serves to stabilize the copper metal in a state of high dispersion, thereby maintaining a large surface area of copper. Under high pressures, the reactants for methanol synthesis COz and H2 adsorb on copper. Participation of zinc oxide in the activation of H2 is also possible. The evidence that support this model include: a) The methanol synthesis activity has been found to be linearly proportional to the surface area of copper [15,16,30],which, for example, can be determined by the stoichiomeuic decomposition of nitrous oxide. b) Cu/Si02 is also an active and selective catalyst for methanol synthesis [27,31]. The activity is comparable to Cu-Zn oxide [27], but it decreases with time on stream due to sintering of copper [31]. Similarly, highly active Cu/A1203,Cu/MgO, Cu/MnO and C o h o 2 catalysts have been reported [15,32]. It is not known that Cu(1) can be stabilized by these supports. It is unlikely that the question of the active site can be resolved readily because at present it is quite difficult to characterize a catalyst under reaction conditions. Since the state of the catalyst, especially the oxidation state of copper, responds rapidly to the gas composition, there will be uncertainties in the applicability of the data from various characterization techniques to the working catalyst.

Effect of Promoters on Cu-Zn Oxide The performance of a Cu-Zn oxide catalyst can be improved by the addition of promoters. Aluminum is a structural promoter that significantly increases the thermal stability of the catalyst. Many other promoters have been mentioned, including boron, chromium, manganese, rare earth, and some alkaline metal ions

234

600

/- 7 I

E

6t

methyl formate

ethanol

Nominal Cs Conc. mole

%

Figure 13-4 Yields of methanol, methyl formate, and ethanol as a function of Cs loading on a Cu/Zn/O catalyst. Reaction conditions: Hz/CO = 2.333, P = 7.6 MPa, GHSV = 6120 l(STP)/kg cat-h, 523 K. (From J. Chem. SOC.Chem. Commun.. 193 (1986). copyright Royal society of Chemistry).

[2], but there is little understanding of their functions. Potassium, barium, and cesium ions are among the alkali and alkali earth promoters. The effect of the addition of Cs has been studied in detail [33,34]. It has been observed that the rate of methanol production by the hydrogenation of CO is greatly increased by adding a small amount of Cs to the catalyst. This is shown in Fig. 13-4. Addition of a large amount of Cs, however, blocks the catalyst surface and causes decline of the activity. It is proposed that Cs participates in the synthesis reaction in the following manner [34]:

(13-6)

CO HYDROGENATION

235

In addition to methanol, Cs promotion also enhances the production of C2 compounds such as ethanol and methylformate (Fig. 13-4). as well as higher alcohols [33,34]. Whether CO or methanol is the carbon source for ethanol and methylformate produced on a Cs-Cu-Zn oxide has been investigated by isotopic labeling experiments [34]. By using a mixture of 13CH30Hand 12C0/H2as the reactant, it is found that the 13C label in methylformate resides almost exclusively in the methyl group. In ethanol and propanol, the carbon atoms are about equally labeled. These labeling data can be explained by the following. Methylformate is formed by CO insertion into the surface-oxygen bond of a methoxide, which might be bound to a Cs ion:

CSOH

+

CH3 co -H20 I CH3OH e = = k '0 -+ + H 2 0 Cs@

+HzO CSOH

2

r--

-H20

+

YH3 w=O @cs

(13-7)

HCOOCH3

The nearly equal extent of labeling of the carbon atoms in ethanol and propanol suggests that CO insertion is not involved in their formation. It is proposed that they are formed by coupling of alcohol molecules. For example, the overall reaction for the formation of ethanol is: 2CH30H

CH3CH20H

+

H2O

(13-8)

which might proceed via an aldehydic intermediate:

(13-9)

Higher alcohols are then produced by analogous coupling reactions. Promotion of a Cu-Zn oxide catalyst by both Co and A1 results in a catalyst that produces almost as much ethanol as methanol at 563 K and 6 MPa pressure using a feed of HJCO ratio of two [35]. Higher alcohols are also produced the amounts of which decrease with increasing chain length.

236

Unlike Cs, promotion by potassium enhances the production of branched alcohols [36]. This will be described in the section on isosynthesis.

13.3 G R O W VIII METALMETAL OXIDE CATALYSTS A number of group VIII metal oxides have been reported to be active in the hydrogenation of carbon oxides. Rather extensive work has been performed on systems based on Pd, Rh, and Fe. These will be discussed here. However, in all of these cases, it is not yet settled whether the active component in the catalyst is the metal or the metal oxide.

Methanol Production on Palladium Catalysts It is well known that Pd catalyzes CO hydrogenation to methane. However, the production of methanol on Pd has also been known for a long time [37]. Recently, it has been found that methanol can be produced with high selectivity [38,39]. Since then, it has been found that the methanol selectivity depends on the support used. For example, methanol is produced almost exclusively on Pd supported on La2O3. On other supports, the selectivity for methanol decreases as: La203 > Si02(II) > ZrOz > ZnO - MgO > Ti02 > A1203 2 Si02(I), and the product on Pd/Si02(I) is practically exclusively methane [40]. It is interesting that the selectivities are quite different on the two different sources of silica, Si02(I) and Si02(II). This difference for different S O 2 supports has later been confirmed [41]. In general, basic oxide supports favor methanol production. Different promoters, such as Na [39,42], Mg, and L a [43] also enhance the selectivity for methanol. The reasons behind the dependence of the selectivity for methanol versus methane on the nature of the promoter or the support are not well understood. One proposal is that methanol is formed via a formate intermediate. The role of the alkali ion promoter is to stabilize the formation of the formate [42], which has been detected by infrared spectroscopy on a Na-promoted Pd catalyst. By isotopically labeling this formate species and following its reaction, its participation in methanol production is established. However, this proposal has not been supportcd by other evidence. In one study, chemical trapping experiments on a Mg-promoted catalyst did not detect any surface formate. Instead, a good correlation between the activity for methanol production and the concentration of a surface formyl species was obtained [44]. In another study, a surface formate species was detected only on a Pd/La203 catalyst but not on a Pd/SiO2 catalyst, although the kinetics of methanol synthesis on both catalysts were very similar P51. Another factor that might determine methanol selectivity is the morphology of the Pd crystalliks. It has been suggested that a Pd(100) surface in a Pd/La203 catalyst is more active than a Pd(ll1) surface, and for a fixed morphology, Pd/La203 is more active than Pd/SiO [461 . is- involved in methanol production. It Yet another proposal is that Pd3 ion has been found that the rate of methanol production can be correlated to the amount

CO HYDROGENATION

237

of Pd extractable by acetylacetone on the promoted catalysts [43]. Since acetylacetone extracts Pd' ions, this correlation strongly suggests that on these promoted catalysts, methanol formation is associated with an oxidized Pd center. The role of a promoter or a suitable support is to stabilize the Pd+ ion in a reducing atmosphere. Indeed, Pd' has been detected by EPR on MgO and La203-supported catalysts [471. This last proposal is also supported by the effect of oxygen treatment and isotope labeling results [42]. When Pd is supported on 2 3 0 2 or TiO2, the catalyst pretreated with oxygen at 723 K followed by a low temperature reduction results in a higher methanol production rate than one without the oxygen pretreatment. The effect is removed by a high temperature reduction. It is suggested that Zr02 and Ti02 can stabilize a certain form of surface oxygen at or near Pd that results in the active site for methanol production. The effect of oxygen treatment, however, is absent if the catalyst is promoted with Na, because the special active site is already stabilized by the alkali ion promoter. When C1*Ois used as a feed over a Na-promoted Pd catalyst, the methanol produced initially is CH3160H [42]. The amount of such l60product is much smaller on catalysts without Na promotion. Thus there is a surface oxygen species present on catalyts active in methanol production that is stabilized by the alkali ion promoter.

Rhodium Catalysts Some catalysts based on rhodium oxide have been found to hydrogenate CO to produce hydrocarbons, methanol and higher oxygenates. The product distribution depends on the reaction conditions, the presence of promoters, and the pretreatment of the catalyst 1481. Table 13-1 lists some examples where the production of oxygenates has been reported. Among the first rhodium oxide-based catalyst studied was a surfaceoxidized rhodium foil [49a]. At a pressure of 0.6 m a , and a HJCO ratio of 3 to 1/3, up to about 10% oxygenates were produced, which are primarily methanol, ethanol, and acetaldehyde. Decreasing the HJCO ratio or the reaction temperature shifted the product distribution towards higher molecular weight hydrocarbons, and higher ethylene to ethane ratios. Lowering the temperature also increased the oxygenate yield, while decreasing the HJCO ratio favored the production of acetaldehyde at the expense of methanol and ethanol. About the same product distribution was obtained using a H2/C02 feed instead of a HJCO feed [49b], but the reaction rate was higher. In comparison. the products from Rh metal that has not been preoxidized contain no oxygenates but a larger fraction of methane. Furthermore, the activity of the surface-oxidized rhodium, as measured by the methane turnover number, (that is the rate of methane production pcr exposed rhodium metal atom before oxidation), is about ten timcs that of rhodium [49a,b]. The use of a H2/C02 feed on Rh metal instead of HJCO results in a higher C b selectivity at the expense of longcr chain hydrocarbons [49b]. Rho2, Rh203, and Rh203.5H20have been studied as catalysts [50]. It has been found that under thc reaction conditions of 300°C, 0.6 MPa of 1:l CO/H2,

238 Table 13-1 Product Distribution in CO Hydrogenation over Rhodium Oxides

Products. wt % Catalyst

' P C H2/CO' CH, C2H, C,H, C,+ CH,OH C,H,OH MeCHO Oxyge- ref.b nates

oxidized Rh foil and Rh(ll1)

300 300 250 250

3/1 1/3 3/1 1/3

8 0 4 56 15 60 5 35 10

7 3 6 2

7 22 17

-

48

-

heavily 300 oxid. Rh( 111)

3/1

70

4

6

10

-

R$O,. 5Hz0

300

1/1

29

22(C2)

25

3

tr

16

11

LaRhO,

300 225

1/1 1/1

16

17 40

14 13

31 23

111

12

9(C,+) 6(C,+)

2

i

3

-

1

12

i

10

1

10

i

... ...

111

Footnotes: a) 0.6 MPa total pressure. b, References: i) D.G. Castner, R.L. Blackadar, and G.A. Somorjai. J. Catal., 66,257 (1980). ii) P.R. Watson, and G.A. Somorjai, J. Catal.. 72. 347 (1981). iii) P.R. Watson, and G.A. Somorjai. J. Catal., 74, 282 (1982).

only Rh203.5H20 resists complete reduction to the metal. The selectivity for methane on this catalyst is much lower than on a Rh metal or oxidized Rh metal (see Table 13-1). Almost all Cz+ hydrocarbons are alkenes, and a significant quantity of oxygenates is produced. In addition to the products shown, a small amount of propanal and butanal are produced. At the same temperature and total pressure, the C b yield increases with an increasing HJCO ratio, but the product distribution changes little with total pressure. Increasing the temperature increases the CH, selectivity. Addition of C 2 b to the feed enhances the production of C~HSCHO.This can be explained by the general scheme for the formation of aldehyde via carbonylation of surface hydrocarbon species: M-C,H,

+co ---+

+(z+l)H M - C O - C , H , A

M

+

C,H,+,CHO

(13-10)

CO HYDROGENATION

239

Addition of C2H4 to the feed increases the surface concentration of M-C2& (or MC2H5) which, upon carbonylation and hydrogenation, yields propanal. Results using LaRh03 as the catalyst follow the trend described thus far. At 225 and 30O0C, 0.6 MPa and using a 1/1 C0/H2 mixture, about 80% of the products were oxygenates, of which ovTr half were acetaldehyde and ethanol. Among the hydrocarbons, methane was the dominant product. The higher hydrocarbons were mostly alkenes. In comparison, LaRh03 supported on aA1203produced methanol very selectively [2]. The surface condition of W h o 3 during reaction is not known. In one study, surface analyses by A E S showed that the LaRh03 surface contained some Rh metal, but most of the rhodium was in the +1 state [51]. In another study, surface analyses by XPS showed the presence of Rh metal and Rh(II1). No Rh(1) was detected [52]. The higher rates of oxygenate and alkene production on rhodium oxides can be explained by the lower hydrogenation activity and CO dissociation ability of the oxides than the mctal. On going from heavily oxidized metal to rhodium oxide and LaRh03, the degree of oxidation and the stability of the oxide increases. (This has been shown by temperature programmed reduction studies [52]). Concurrently one sees an increase in the selectivity for higher hydrocarbons, alkenes, and oxygenates. The increase in oxygenate production is also a result of the higher concentration of adsorbed molecular CO on the oxides that leads to a greater rate of CO insertion into a surface M X H , species to produce Cz+ oxygenates, and to more direct hydrogenation of CO to methanol. It should be noted that the discussions above do not explain the higher activity due to surface oxidation, especially in methane production. This phenomenon is not yet satisfactorily explained. The role of oxidized Rh appears to be also important in other supported Rh catalyst. On a Rh/MgO catalyst that has a high selectivity for methanol and other oxygenates, a Rh(I1) EPR signal has been detected [47]. Isotope labeling experiments show that adsorbed oxygen and/or surface lattice oxygen on a Napromoted Rh/Si02 is incorporated into the product acetaldehyde, and that the C 4 bond is formed by CO insertion into a surface CH, species [421. However, the participation of rhodium oxide has not been universally accepted. Recently, it has been reported that the product distributions are similar on a Rh/Si02, reduced Rh2O3, and prereduced or unprereduced LaRh03 catalyst. This has led to the suggestion that Rh metal is the active component [52].

Iron Catalyst Recently, it has been suggested that iron oxide is active in CO hydrogenation because the catalytic activity of an iron foil is increased ten times when the foil is preoxidized in dry oxygen before exposure to a C0/H2 mixture [531. This high activity decreases with time as the surface oxide is reduced. Subsequently, it was reported that the pseudo-steady state CO hydrogcnation activity of an unsupported catalyst whose precursor is a-Fez03 is higher, if the iron oxide has not been prereduced, than if it is prercduced to a-Fe [54]. The bulk of the unprereduced catalyst is converted to Fe304 at pseudo-steady state. However, there is no information on the surface conditions. The a-Fe catalyst is converted to iron

240

carbide during reaction, and the used catalyst contains more surface graphitic carbon than the used unprereduced catalyst. These results show that bulk Fe304 is at least as active in CO hydrogenation as bulk metallic iron or iron carbide. Most of the observations above have been confinned in another study [55]. In particular, the pseudo-steady state bulk compound of a a-Fe203 catalyst is confinned to be Fe304. Some X-iron carbide is also present. If the a-Fe203 is prereduced to Fe before reaction, the pseudo-steady state catalyst contains xcarbide and small amounts of €'-carbide and a-Fe. However, the product distributions over these two catalysts are the same. This has led to the conclusion that neither F@03nor Fe304 are active for hydrocarbon synthesis. It is proposed that the superior catalytic acitivity of the unprereduced a-FezO3 catalyst is due to the formation of very small crystallites of a-Fe and X-iron carbide under reaction conditions, which are the active phases.

13.4 ISOSYNTHESIS REACTION It has been found that on some basic oxides, especially Zr02 and Tho2[56601, and some rare earth oxides such as La203 and Dy2O3 [61], branched alcohols and hydrocarbons are produced quite efficiently, particularly 2-methylpropane, 2methylpropene, and 2-methylpropanol. Table 13-2 shows some sample data for Zr02 [58]. At the low conversions shown, alcohols, ether, and hydrocarbons are the major products, and substantial amounts of 2-methylpropane and 2methylpropene are produced. These two branch-chain molecules are among the major hydrocarbon products also at much higher conversions of CO, although then the predominant reaction producct is COz, which is formed by the water-gas shift reaction [W]. The reaction that leads to the production of these branched products is called the isosynthesis reaction. In addition to the oxides mentioned above, some alkali-promoted ZnO [56] and potassium-promoted Cu-Zn oxide [36] also produce branched products. The formation of alcohols and hydrocarbons in the isosynthesis reaction are closely related. For example, on Th02, branched alcohols are produced in substantial yields below about 375°C. At higher temperatures (425-450°C), branched C4 hydrocarbons are formed. Above 50O0C, methane, ethane, and propane are the principal products [62]. A mechanism for chain growth and chain branching in the isosynthesis reaction on Zr02 is shown in Fig. 13-5 [58,63]. It is based on analogous compounds and reaction mechanisms involving coordination complexes. An important intermediate in this mechanism is the bound aldehyde (species I). The lowest member of the bound aldehyde is the bound formaldehyde (R=H), which can be formed by hydrogenation of adsorbed CO or surface formate. The hydrogenation of adsorbed CO may take place by insertion of CO into a metalhydrogen bond. A model for this step is the reaction of CO with [(Cp)2ZrHCIl, to give [(Cp)2ZrC1]2(p-CH20)[ a ] . This latter compound has a bridging formaldehyde moiety in which the oxygen atom of the formaldehyde bridges the two metal atoms and the mcthylene fragment is bound to one Zr and the 0 atom.

CO HYDROGENATION

241

Table 13-2 Product Distributions in CO Hydrogenation on Zr02 (425°C. 3.5 MPa total pressure, CO/H2 = l/l). (From J. Catal., 109, 284 (1988), copyright Academic Press).

catalyst 1

CO conversion % Product Selectivity (mole %) methane methanol dimethylether ethane and ethene propane and propene linear C4's branched C4's. linear C5's branched C5'sb C4 alkenes/C4 alkanes (mole ratio)

catalyst 2

0.6

0.4

42.8 4.1 16.8 11.8 3 .O 4.6 13.1 0.5 2.3

41.9 5.9 10.4

16.4

20.2

11.1

4.7 5.9 16.2 0.4 3.5

Foomotes: a) The ratio of 2-methylpropene/branchedC4's is greater than 0.93. b, Monomethylated butanes only.

C2 compounds are formed by insertion of CO into a bound formaldehyde to form a cyclic acyl species (I -+ TI). Hydrogenation of species II (I1-+ IV) would eventually lead to a linear alcohol that is one carbon longer. The acyl species @ has I an) isomeric structure (HI), which permits 1,2-H shift to form an enediolate (I11 -+V). The formation of enediolate by a similar process has been suggested in complexes of thorium, uranium, and hafnium [65,66]. Chain branching results when the enediolate is hydrogenated to form a bound diol (V -+ VI) which dissociates to form an enolate and a surface hydroxyl (VI4 VII). The latter step is similar to what would be involved in the catalytic dehydration of alcohols. Species VII has a resonance structure VIII which is a q3-enolate. Addition of H to VIII leads to a bound aldehyde structure OX). Insertion of CO into this species like the step I -+ I1 complctcs the cycle for chain branching. The validity of this mechanism, especically the importance of the bound aldehyde structure, has been tested by monitoring the incorporation of propanal and acetone into the isosynthesis products. It has been found that on ZrOz [58], addition of propanal to a feed of CO and H2 increases the yields of 1-butene, 2mcthylpropene, and 2-methyl-1-propanol. By using 13C-labeled propanal, it is confirmcd that these products can indeed be formed from propanal. Similarly, when acetone is added, the yields of I-butene, 2-methylpropene, and 2-methyl-l-

m

:T

In

N

I -

u

I 0,

II 0

/N

\ L

n

--

Y

0

0

II

u- =/ \u

I

242

I

1

In

I N

(Y

\

In I

u

I

/

u II

In I

u

In

M I

u

I

\

rJt x

N

0

S

n

uI I k5 N / b s o

0

\/o

N

I

[ I n

1 7

-

N

I

N2

'L

/

Y

F\o/

1

243

CO HYDROGENATION I,

C I H3

-

CH2 I CH

L

\

z;

+

- H

CH3 I 0

I

Zr

+H

(XI)

j o

1

(1)

*

CH 3

I

CH3-CH

I CH20H

+ H

*

CH 3

I

CH3-CH

1 7

- H

I CH

I )o

Zr (XI11

Figure 13-6 Aldol condensation mechanism in the formation of branched products on Z Q . (From the same source as Figure 13-5. copyright Academic Press).

butene are increased. The enhanced yields of these products support the participation of surface species of the bound aldehyde structure (Iand IX). A bound propanal would produced 1-butene by the reaction sequence I + I1 + I V 4butanol, which is dehydrated to 1-butene. The same reaction sequence for a bound acetone would produce 2-methylpropanol, which on dehydration would result in 2-methylpropene. The production of 1-butene from acetone is interesting. It can be accounted for if the steps leading from species II to IX are all rapidly reversible. It has also been shown that when 2-propanol is added to the reaction feed, there is little enhancement in the yields of the C4 products [%I. Thus 2-propanol cannot be converted easily into an intermediate for chain growth or chain branching. This rules out the possibility that acetone is first hydrogenated to 2propanol before it is converted to the C4 products. The enhanced production of 2-methylpropene or 2-methylpropanol from propanal, or 2-methyl-1-butene from acetone cannot be explained by this mechanism. However, these are important products. In the experiments where propanal was added, it was observed that the yields of 2-methyl-1-propanol and 2methylpropanal were both increased substantially in a parallel fashion. It was

244

concluded that these oxygenates were the primary products. They were subsequently converted to 2-methylpropene by dehydration and hydrogenation [%I. The enhanced yields of these products suggest the presence of a parallel mechanism, which is also suggested by other evidence. If the mechanism in Fig. 13-5 is the only process on the catalyst, it would predict a statistical distribution of products with respect to the number of carbon atoms in the molecules. In particular, the amount of each product would decrease in a proportional manner as the carbon number increases, (that is, it would follow the Anderson-Schultz-Flory distribution). The data in Table 13-2 shows that this is not the case. The amounts of C4 products are substantially more than those of C3 and C5. To account for these observations, an additional mechanism has been proposed. This mechanism is shown in Fig. 13-5. It involves alkylation of an q3enolate by a nearby methoxide group at the P-carbon. This is a condensation reaction, and it is assumed that during synthesis from CO and H2, this reaction is faster for the C2 species than for the C3 or C4 species, and has no equivalent in the C1 reactions [63]. The participation of this condensation reaction has been confirmed. It has been observed that when '3C-melhanol is added to the CO/H2 feed, 13C is incorporated into 2-methylpropene, 2- and 3-methyl-1-butene, and 2-methyl-2butene. Interestingly, no 13C incorporation is observed in the C4 or C5 linear products [ S S ] . Since it has been shown that methanol adsorbs as methoxide on the catalyst [67], the incorporation of 13C into the branched products supports the condensation mechanism. This mechanism explains the production of 2-methylpropanol and 2methylpropene from propanal. Condensation of a methoxide with a bound propanal results in these products. The production of 2-methyl-1-butene from added acetone can be understood also. Condensation of a methoxide with a bound acetone would produce 2-butanol or 2-butanone. CO insertion to a bound 2butanone followed by the reaction sequence leading from I to IV would produce 2methyl-1-butene.

13.5 WATER-GAS SHIFT REACTION The water-gas shift reaction is an industrial process to generate hydrogen from steam and carbon monoxide: CO

+

H20

+ C02 +

H2

The reaction is carried out between 1 to 3 MPa. The commercial catalysts are of two types. An iron-chromium oxide catalyst, sometimes promoted by other oxides such as A1203, or supported on Si02, is used at a higher temperature of 600725 K. A newer catalyst, based on copper-zinc oxide, promoted by Cr, Co, A!, and Ni, is used at a lower temperature of about 500 K. The advantage of the copper-zinc oxide catalyst over the iron-chromium oxide catalyst is that the catalyst is much more aclive and can be used at lower temperatures where the

245

CO HYDROGENATION

equilibrium is more favorable to the formation of H2. The disadvantages are sensitivity to poisoning by sulfur and chlorine, and shorter catalyst life. The copper-zinc oxide catalyst used for this reaction is essentially the same as the one used in methanol synthesis. In fact, during methanol synthesis, the watergas shift reaction proceeds at a comparable rate. Thus the readers are referred to section 13.2 for discussions of this catalyst. When this catalyst is used for watergas shift, methanol production becomes thermodynamically unfavorable because of the high water and low hydrogen partial pressures. It is believed that the mechanism of the reaction over copper-zinc oxide involves a surface formate that is formed by the insertion of CO into a surface hydroxyl group: CO(g) H20(g) CO(ad) + OH-(ad) HCOO-(ad)

+ CO(ad)

(13-1 la)

+ OH-(ad) + H+(ad)

(13-llb)

+ HCOO-(ad)

( 13-11C)

_L

C02(ad) + H-(ad)

(13- 1Id)

A correlation between the ability to form a surface formate and the water-gas shift activity of a catalyst has been observed [68]. On Cu-Zn oxide, this mechanism has been discussed for methanol synthesis (similar to reaction 13-6). A surface formate has been identified by chemical trapping [7] and by infrared spectroscopy for the reaction of H2 + C02 [22,23]. Further support comes from the fact that the rate for formic acid decomposition is comparable to that for watergas shift reaction [69]. In spite of the fact that a surface formate is probably involved as an intermediate in both the water-gas shift and the methanol synthesis reaction, there are no indications that these two reactions are necessarily correlated. In fact, the relative activities for the two reactions on different catalysts could be quite different. The iron-chromium oxide is a mixture of Fe304 and Cr203 under reaction conditions. It is generally agreed that the addition of Cr greatly helps retain the surface area of the catalyst. There is no evidence that chromium chemically affects the iron oxide. The mechanism with which the Cr promoter functions is not firmly established. It has been suggested that Cr2O3 forms a protective layer around the a-Fe203 particles (which has the same corundum structure as Cr2O3) during calcination, thereby prevcnting particle growth of the iron oxide [70]. Indeed, determination of the surface composition of an Fe203-Cr203solid solution has suggested the formation of a surface compound which might serve as the protective laycr [71]. It has been shown on another system of Fe304supported on

246

Si02 that the formation of a surface shell of iron-silicon oxide on a Fe304 particle stabilizes the particle from sintering during oxidation-reduction [72]. There are two proposed mechanisms for the iron-chromium oxide catalyst. One mechanism involves a surface formate as shown in eq. 13-11 173,741. The other is a redox mechanism in which a surface oxygen species is involved as is shown below:

(13-12) slow p M +CO2 L

-

s l o w 0,

-2

M

p

+H2

In this mechanism, the active site is proposed to be an anion-cation pair

O P site ( \M ). CO first adsorbs weakly on the coordinatively unsaturated metal cation of this site. This weakly adsorbcd molecule then reacts with surface oxygen in the site to form a bidcntate carbonate, which decomposes into C 0 2 leaving /O an anion vacancy in the site (M ). The anion vacancy is the site for the adsorption of water. The oxidized surfacc site is regcncratcd by thc desorption of H2 [751. Support for this redox mechanism (also called rcgencrative mcchanism) has been provided by chemisorption expcriments undcr catalytic conditions which show that: (1) H2 adsorbs dissociatively and H 2 0 adsorbs associatively, and the isotherm for concurrent adsorption of bolh molcculcs follow the form for competitive adsorption [75]; (2) whcn a Fe3O4 catalyst is exposed to a mixture of CO and C 0 2 , thc adsorption isotherms of thcse molecules also follow the form for competitive adsorption. Furthermorc, thc saturation coverage of CO corrcsponds directly to thc coverage of anion vacancics, while changes in the C 0 2 saturation coverage correlate with changcs in the covcragc of the reducible oxygcn spccics on thc surface 1761; (3) by following the CO-C02 isotopic exchange reaction using 13C-labeling 1771, it is concludcd that CO and C 0 2 arc indistinguishable once adsorbcd on thc catalyst as is prcdictcd by rcaction (13-12); (4) the rates of production of CO from thc rcaction of C02 with a catalyst surfacc, of C02 froin CO as mcasurcd h y thc chcinisorption cxpcrirncnts, and of 1-12 from H 2 0 and H2O

247

CO HYDROGENATION

0

100

50 Fe 0

3 4

150

200

Particle Size, nm

Figure 13-7 The F q 0 4 particle size dependence of the water-gas shift activity of magnetite. Open circles represent unsupported Fe304 (except that the smallest particle-size sample is a commercial catalyst) and solid circles represent silica-supported catalysts. (From J. Catal., 76,93 (1982). copyright Academic Press).

from H2 are comparable to the rate of th water-gas shift reaction under similar conditions [78]; and (5) it is observed that the rate of the wixr-gas shift reaction corresponds to the rates at which H 2 0 oxidizes and CO reduces the surface of Fe304 [79,80]. It has been further proposed that the active site consists of octahedrally coordinated Fe ions of magnetite that are exposed on the surface. The proposal is based on the following observations [75]. It is found that, the rate of the water-gas shift reaction per surface Fe ion (the amount of exposed surface Fe ions is dctcrmined by NO adsorption) is relatively ilJdependent of the crystallite size of unsupported Fe304 (see Fig. 13-7) 1811. Hovever, the rate per surface Fe ion decreases sharply with decreasing crystallite size on Si02-supported magnetite catalysts. On the other ha. neither unsupported nor supported catalysts show crystallite-six dependence when the number of active sites is determined by CO/C02 adsorption (which measures the amount of removable oxygm on the catalyst surface). It is further shown that Si ions substitute into the magnetite

248

lattice at the surface of the crystallites in the Si02-supportedcatalysts. The Si ion substitution occurs at the tetrahedral sites [72]. The data in Fig. 13-7 can be explained as follows. The catalytically active site consists of a surface-exposed Fe cation in an octahedral site and the surrounding oxygen anions. Then the apparent crystallite-size effect on the supported samples is due to the following. In a Si02-supported catalyst, Si4+ substitution in the magnetite surface causes all of the Fe ions to be in the octahedral sites as well as a decrease in the electron density at these ions. The latter effect perturbs the balance between Fez+ and Fe3+ at these sites and reduces their ability to undergo oxidation and reduction, as is required in the regenerative mechanism. This effect is manifested in a decrease in the amount of reducible oxygen on the surface as determined by CO/CO2 chemisorption, and a corresponding decrease in catalytic activity. Since the Si subsitution is confined to the surface, its effect is larger for smaller crystallites, and affects more strongly the more weakly adsorbed gases such as CO and C02. On the other hand, it does not affect the adsorption of more strongly adsorbed gases such as NO. This might be due to surface reconstruction induced by NO adsorption on cations that are inaccessible to more weakly adsorbed gases [81,82].

REFERENCES 1 . "Catalysis. vol. 3," P.H. Emmett ed., Reinhold, New York, 1955. 2. G. C. Chinchen, P. J. Denny, J. R. Jennings, M. S. Spencer, and K.Waugh. Appf. Cataf..36, 1 (1988). 3. K.Klier, Adv. Catal., 31, 243 (1982). 4. P. J. Denny, and D. A. Wan, "Catalysis". vol. 2. Chem. SOC.,London, 1978, p.46. 5. H. H. Kung, Cataf.Rev., 22. 235 (1980). 6. G. Liu, D. Wilcox, M. Garland, and H. H. Kung,J . Cataf.,90, 139 (1984). 7. R. Kieffer, E. Ramaroson. A. Deluzarche, and Y. Trambouze, React. Kinet. Cataf. Lett., 16, 207 (1981). 8. B. Denise, and R. P. A. Sneeden, J . Mofec. Cutaf.. 17. 359 (1982). 9. K. Klier. V. Chatikavanij. R. G. Herman, and G. W. Simmons. J. Cataf.. 74, 343 (1982). 10. R. Bardet, J. Thivolle-Cazat, and Y. Trambouze, J. Chim. Phys., 78, 135 (1981). 1 1 . G. A. Vedage, R. Pitchai, R. G. Herman. and K. Klier. Proc. 8th Intern. Congr. Catal.. vol. II, 47 (1984). 12. Yu. B. Kagan, A. Ya. Rozovskii. L. G.Liberov, E. V. Slivinskii, G. I. Lin, S. M. Loktev, and A. N. Bashkirov. Dokf.Akud. Nauk. USSR, Chemistry Section, 224, 598 (1975). 13. A. Ya. Rozovskii, Kinet. Cataf.,21, 78 (1980). 14. A. Ya. Rozovskii, Yu. B. Kagan, G. I. Lin, E. V. Slivinskii. S. M. Loktev. L. G. Liberov, and A. N. Bashkirov, Kinet. Cataf.,16, 706 (1975). 15. G.C. Chinchen, P. J. Denny, D. G. Parker, G. D. Short, M. S. Spencer, K. C. Waugh, and D. A. Wan, Preprint Div. Fuel Chemistry, A m r . Chem. SOC.. 29, 178 (1984).

CO HYDROGENATION

249

16. G. C. Chinchen. P. J. Denny, D. G. Parker. M. S . Spencer, and D. A. Whan, Appl. Catal., 30. 333 (1987). 17. G. Liu, D. Willcox, M. Garland. and H. H. Kung, J. Catal., 96, 251 (1985). 18. I. R. Monnier, G. Apai, and M. I. Hanrahan, J. Catal., 88, 523 (1984). 19. F. P. Daly. J. Catal.. 89. 131 (1984). 20. N. Jackson, and J. Ekerdt, J. Catal., 101. 90 (1986). 21. I. Saussey, 1.-C. Lavalley. I. Lamotte, andT. Rais, J . Chem. Soc. Chem. Commun., 278 (1982). 22. J. F. Edwards, and G. L. Schrader. J. Phys. Chem.. 88, 5620 (1984); I. Catal., 94, 175 (1985) 23. Y. Amenomiya, and T. Tagawa. Proc. 8th Infern. Congr. Catal., II,559 (1984). 24. G.A. Vedage, R. G. Herman, and K . Klier, J . Catal.. 95,423 (1985). 25. R. G. Herman, K. Klier, G. W. Simmons, B. P. Finn, J. B. Balko, andT. P. Kobylinski, J. Catal.. 56, 407 (1979). 26. Y. Y. Huang, J. Catul., 30, 187 (1973). 27. C. Visser-Luirink, E. R. A. Matulewicz, I. Hart. and J. C . Moi. J. Phys. Chem.. 87, 1470 (1983). 28. Y.Okamoto. K. Fukiro. T. Imanaka, and S. Teranishi, J. Phys. Chem., 87, 3740, 3747 (1983). 29. J. R. Monnier. M. Hanrahan. and G. Apai. J. Catal., 92, 119 (1985). 30. K. Fukino, T. Imanaka, and S. Teranishi, Proc. 8th Intern. Cong. Catal., 11. 159 (1984). 31. J. B. Friedrich. M. S. Wainwright, and D. J. Young,J. Catal., 80. 1 (1983). 32. E. G. Baglin. G. B. Atkinson, and L. J. Nicks, I&EC Proc. Res. Develop., 20, 87 (1981). 33. J. Nunan, K. Klier, C.-W. Young, P. B. Himelfarb, and R. G. Herman, J . Chem. Soc. Chem. Commun.. 193 (1986). 34. J. Nunan, C. Bogdan, K. Klier. K. Smith, C. Young, and R. Herman, J. Catal., 113, 410 (1088). 35. Ph. Courty, D. b r a n d , E. Freund, and A. Sugier. J. Molec. Cutal.. 17. 241 (1982). 36. K. J. Smith, and R. B. Anderson, C w d . J . Chem. Eng.. 61.40 (1983). 37. German patent 293787 (1913). 38. M.L. Poutsma. L. F. Elek, D. A. Ibarbia. A. P. Risch, and J. A. Rabo, J . Catal., 52, 157 (1978). 39. Y. Kikuzono, S . Kagami, S. Naito, T. Onishi, and K. Tamaru. Faraday Disc., Chem. SOC., 72, 135 (1981). 40. Y.A. Ryndin, R. F. Hicks, A. T. Bell, and Y. I. Yermakov, J. Catal., 70, 287 (1981). 41. F. Fajula, R. G. Anthony, and J. H. Lunsford. J . Cutal.. 73, 237 (1982). 42. S. Naito, H. Yoshioka. H. Orita, and K. Tamaru,Proc. 8th Intern. Congr. Catal., HI, 207 (1984). 43. J. M. Driessen. E. K. Poels, J. P. Hindermann, and V. Ponec, J. Catal., 82, 26 (1983). 44. J. P. Hinderman, A. Kienncmann, A. Chakor-Alami, and R. Kieffer, Proc. 8th Infern. Congr. Coral.. II. 163 (1984). 45. R. F. Hicks, and A. T. Bell, J. Catul.. 91, 104 (1985). 46. R. F. Hicks, and A. T. Bell. J. Catal.. 90,205 (1984).

250 47. E. K. Poels, P. J. Mangus, J. V. Welzen, and V. Ponec, Proc. 8th Intern. Congr. Cataf.,II. 61 (1984). 48. T. Wilson, P. Kasai. and P. Ellgen, J. Catal., 69, 193 (1981); M. Kawai, M. Uda, and M. Ichikawa, J . Phys. Chem., 89, 1654 (1985). 49a. D. G. Castner, R. L. Blackadar, and G. A. Somorjai. J. Cataf..66, 257 (1980). 49b. B. A. Sexton, and G. A. Somorjai, J. Cataf.,46, 167 (1977). 50. P. R. Watson, and G. A. Somorjai, J. Catal., 72, 347 (1981). 51. P. R. Watson, and G. A. Somorjai, J. Catal., 74, 282 (1982). 52. H. J. Gysling, J. R. Monnier, and G. Apai, J . Catal., 103, 407 (1987). 53. D. J. Dwyer, and G. A. Somorjai. J. Cataf.,52, 291 (1978). 54. J. P. Reymond, P. Meriaudeau, and S . J. Teichner, J. Cataf.,75, 32 (1982). 55. R. A. Dictor, and A. T. Bell, J. Cataf.,97, 121 (1986). 56. G. Natta, U. Columbo. and I. Pasquon, "Catalysis", P.H. Emmett ed.. Reinhold Publ., vol. V, 131 (1957). 57. H.Storch. N. Golumbic, and R. Anderson, 'The Fischer-Tropsch and Related Synthesis." Wiley, N.Y., 1951. 58. S. C. Tsang, N. Jackson, and J. Ekerdt, J. Cataf., 109, 284 (1988). 59. C. Chang. W. Lang. and A. Silvestri, J. Cataf.,56. 268 (1979). 60.T. Maehashi, K.4. M a y a . K. Domen, K. Aika, and T. Onishi. Chem. Lett. Chem. SOC. Jpn., 747 (1984). 61. R. Kieffer, J. Varela, and A. Deluzarche, J. Chem. Soc. Chem. Cornmum., 763 (1983). 62. H.Pichler, and K.-H. Ziesecke, Bur. Mines Buff.,448 (1950). 63. T. J. Mazanec, J. Cataf.,98, 115 (1986). 64. S. Gambarotta, C. Floriani, A. Chiesi-Villa, and C. Guastini, J. Amer. Chem. Soc., 105, 1690 (1983). 65. J. Manriquez. P. Fagan, T. Marks, C. Day, and V. Day, J. Amer. Chem. Soc., 100, 7112 (1978). 66. D.Roddick. M. Fryzuk. P. Seidler. G. Hillhouse, and J. Bercaw, Organometalfics, 4, 97 (1985). 67. N. Jackson, and J. Ekerdt, J. Catal., 101, 90 (1986). 68. D.G. Rethwisch, and J. A. Dumesic, Appl. Catal., 21. 97 (1986). 69.T. Van Henvijnen, and W. A. de Jong, J. Cataf.,63, 83. 94 (1980). 70. F. Domka, and M. 2. Laniecki, Anorg. Aflg. Chem., 435. 273 (1977). 71. M. Kung, and H. Kung,Surface Sci., 104. 253 (1981). 72. C.R. F. Lund, and J. A. Dumesic, J . Phys. Chem., 85, 3175 (1981); 86. 130 (1982). 73. S. Oki, and R. Mezaki. J. Phys. Chem., 77. 447 (1973). 74. S. Oki,J. Happel. M. Hinatow, and Y. Kaneko. Proc. 5th Intern. Congr. Catal., 1973,p. 173. 75. C. Lund, J. E. Kubsh. and J. A. Dumesic, ACS Symp. Series no. 279, "Solid State Chemistry in Catalysis," ed. R. K. Grasselli, and J. F. Brazdil, 311 (1985). 76. J. E. Kubsh, Y. Chen, and J. A. Dumesic, J. Cataf.,71, 192 (1981). 77. M.Tinker, and J. Dumesic, J . Cataf.. 103. 65 (1987). 78. J. E. Kubsh, and J. A. Dumesic. AlChE J., 28, 793 (1982). 79. G.K. Boreskov, T. M. Yur'eva. and A. S . Sergeeva, Kinet. Kafaf.,11, 1476 (1970).

CO HYDROGENATION

251

80. G . K. Boreskov, Kinet. Kufal.. 11, 374 (1970). 81. C. Lund, and J. Dumesic, J. Cafal., 76, 93 (1982). 82. K. Segawa, Y.Chen, J. Kubsh, W . Delgass, J. Dumesic. and W. K. Hall, J. Cnfal.. 76, 1 1 2 (1982).

Chapter 14 PHOTO-ASSISTED SURFACE PROCESSES

14.1 INTRODUCTION A general picture of the bulk electronic band structure of the valence electrons of transition metal oxides may be described as follows: the 0 2p orbitals overlap to form a filled valence band, and the s and d orbitals of the cations overlap to form the conduction band. The conduction band is at a higher energy than the valence band and separated from it for most transition metal oxides at their highest oxidation states. The separation between the two bands is the band gap (Fig. 14la). At thermal equilibrium, the Fermi energy (that is, the electrochemical energy) of the valence electrons in the solid is between the conduction band energy and the valence band energy. The conduction band is populated with only a few electrons that are thermally excited into the band from the valence band. This excitation also leaves holes (unoccupied electron orbitals) in the valence band. The density of electrons in the conduction band can be increased by incorporating donor ions in the solid that donate valence electrons to the conduction band, and the density of holes in the valence band can be increased by acceptor ions that accept electrons from the valence band. The presence of these donor or acceptor ions also causes a corresponding adjustment of the Fermi energy such that at thermal equilibrium, the density of electrons in the conduction band, n, and of holes in the valence band, p, are given by:

where E,, Ef.E, are the lowest energy of the conduction band, the Fermi energy, and the highest energy of the valence band, rcspectively. N, and N, are constant. N, is related to the effective mass and mobility of an electron in the conduction band, and N, is related to the effective mass and mobili,y of a hole in the valence 252

PHOTO-ASSISTED PROCESSES

1

253

/Ill/ conduction band

I- - - - - - - - band gap valence band

Figure 14-1 Energy level diagram of a semiconducting oxide. a. In vacuum; b. In contact with a solution containing a A-/A redox couple; c. As in b but under illumination. The photo-generated hole migrates to the surface to accept an electron from A-; the photo-generated electron migrates to the bulk of the solid and to a counter-electrode.

band. k is the Boltzmann constant, and T is the temperature. This simple picture applies to the bulk of an ideal oxide. As discusscd in Chapter 3, truncation of the solid to form surfaces might generate surface

254

electronic states. In addition, the position of the Fermi energy may be different under different situations such as when the oxide is in contact with another medium, when there are adsorbates on the oxide surface, or when the surface is partly reduced. These situations could result in a surface region of the solid that is very different from the bulk, and a potential gradient is developed between the surface and the bulk. Associated with this potential gradient is a distribution of densities of electrons and holes (Fig. 14-lb). Absorption of light of energy larger than the band gap energy by the solid also changes these populations. Depending on the solid, the light intensity and its wavelength, photo-excitation could increase the population of electrons and holes by thousands or millions times. When these photo-generated holes and electrons migrate to the surface, light-induced surface chemistry becomes noticeable. A photo-enhanced surface electron transfer process is schematically illustrated in Fig. 14-1. The energy level diagram of an oxide neglecting any surface effects is shown in Fig. 14-la. When this sample is placed in water which contains a redox couple A-/A, electron transfer between the solid and the solution occurs until the electrochemical potential of the valence electrons in the solid equals that in the solution. If the redox couple A-/A is the only species that can exchange electrons with the solid, the electrochemical potential of the solution is defined by the standard redox potential of this couple and the relative concentrations of A- and A. Fig. 14-lb shows the situation where before the solid contacts the solution, the Fermi energy of the solid is higher than the electrochemical energy of the elcctrons in the solution, and after contact, electrons flow out of the solid to the solution so as to establish equilibrium (that is, equalization of electrochemical potcntial) across the interface. The outflow of electrons from the solid results in a net positive charge near the surface in the solid, which eventually sets up a potcntial barrier to stop the electron flow. Equilibrium is then established. The region of the solid near the surface that has a net charge is called the depletion layer. When the solid absorbs a photon of light of an energy larger than the band gap, an electron is excited from the valence band to the conduction band (Fig. 14lc). Because of the upward-bending of the electron bands near the surface. the photo-generated hole tends to migrate to the surface, while the electron tends to migrate to the bulk. This results in an increased hole concentration at the surface, which facilitates surface reactions involving electron transfer to the solid, such as oxidation reactions. This situation persists as long as the solid is illuminated and there is a mechanism to remove the electron that has migrated to the bulk. Flow of electrons from the oxide via an external circuit to a counter electrode is one way to remove them. If there are no mechanisms to remove the electrons from the bulk, continuous illumination will result in accumulation of elcctrons and holes near the surface, to the extent that the bending of the electron bands is removed. The photo-generated elcctrons become available at the surface in significant concentrations, which facilitate surface reactions that involve transfer of elcctrons from the solid, such as reduction reactions. Recently thcrc has becn a strong intercst to replace the external circuit that

PHOTO-ASSISTED PROCESSES A

255

A-

D-

3

Pt

Ti O2

Figure 14-2 Schematic diagram showing a Ti02 particle deposited with Pt. Absorption of a photon generates an electron-hole pair. a. The hole migrates to the Ti02 surface to accept an electron from a donor species, and the electron migrates to thz Pt surface to be donated to an acceptor species. b. The energy level diagram of this system under illumination.

connccts the counter electrode and the irradiated oxide by direct contact of the two elcctrodes. Platinum deposited on TiOz is the most popular system that has been investigated. In this system, the photo-generated holes are consumed in the surface reaction at Ti02. The photo-generated electrons migrate to platinum where they are consumed in a reaction at the platinum surface. The complete oxidationreduction cycle can be accomplished on one composite particle. This is schematically shown in Fig. 14-2. It is seen from this description that the photo-effect is significant only when a large concentration of photo-gcnerated holes and/or electrons are present at the oxide surface. Since most of the light is absorbed by the region of an oxide up to a

256

few hundred nanometers from the surface, these holes and electrons must have sufficiently long lifetimes to reach the surface. Therefore, reduction of an oxide and the presence of lattice defects and impurities which shorten the lifetime of these holes and electrons will reduce the photo-effect. The lifetime also depends on the effective masses of these carriers, which depend on the band structure of the solid. Thus the magnitude of the photo-effect depends on the oxide. For example, photo-enhanced processes are readily observed on ZnO and Ti02,but are much weaker on V205. It is also apparent that light of an energy larger than the band gap energy is needed to produce photo-effects. It is possible to generate photoelectrons with light of a lower energy if there are discrete energy levels (commonly called traps) in the band gap. Absorption of a photon may excite an electron from the trap to the conduction band. Similarly, it is possible to generate photo-holes by exciting electrons from the valence band into the unfilled inter-band gap traps. However, these processes are normally not efficient. The concentration of these traps are limited so that the possible densities of photo-generated holes and electrons are also limited. These traps usually act as electron-hole recombination centers that shorten the lifetime of the photo-generated carriers. The description for the scheme shown in Fig. 14-1 is for an n-type semiconducting oxide. An analogous scheme can be readily drawn for a p-type semiconducting oxide, whose Femi energy is much closer to the valence band than n-type semiconductors. Because of the position of the Fermi energy, it is common that electrons flow into a p-type oxide from an external redox couple. This results in a bending of the electron bands in a direction opposite to those shown in the Fig. 14-1.

A side effect that occurs on illumination is heating of the oxide. A majority of electrons and holes recombine nonradiatively. The energy released heats up the oxide. In many experiments, light sources are used that give a broad spectral distribution. Some light of energy less than the band gap energy is absorbed by the phonon mode of the oxide and heats up the solid. The extent of heating depends on the oxide and the light source used. A temperature rise to 60°C from room temperature has been reported [19].

14.2 PHOTO-ASSISTED ADSORPTION AND DESORPTION As described in the last scction, absorption of light of energy larger than thc band gap energy increases the surface concentrations of electrons and holes of a semiconducting oxide significantly and enhances rate processes such as adsorption and dcsorption that involve electron transfer. Photo-assisted adsorption and desorption of oxygen is the most studied process in this category [ 1,23: Adsorption:

e-

+

O2

Desorption:

h+

+

02-(ad) --+02(g)

+

02-(ad)

(14-3) (14-4)

PHOTO-ASSISTED PROCESSES

257

Both adsorption and desorption can be enhance by irradiation. Which process dominates depends on the experimental conditions and the nature and pretreatment of the oxide. It is commonly found that the relative rates of these two processes depend on the pressure of oxygen. Band gap illumination of ZnO at room temperature results in photo-assisted adsorption of oxygen at low pressures. The effect decreases with increasing pressure, and eventually becomes photoassisted desorption at high pressures [3]. However, the effect of pressure is opposite at 400°C: photo-assisted desorption at low pressures, and adsorption at high pressures. The effect is not well understood. A similar pressure dependence has been reported on Ti02 at 500°C [41. At low pressures of oxygen, irradiation enhances desorption; at high pressures, it enhances adsorption. Detection of photo-assisted adsorption can be made by following changes in the conductivity of the sample. Since such adsorption changes the concentration of charged species on the surface and consumes photoelectrons or holes, it changes the conductivity of the sample. It has been found that the surface returns only slowly to the equilibrium state after the light is turned off, thus the enhanced adsorption due to irradiation must be retained for a long time, and these species are practically irreversibly adsorbed. Their desorption can be enhanced by heating the sample [51. The nature of the adsorbed species has been deduced from the dependence of photoconductivity 0 with pressure. Up to lo2 Pa of oxygen, 0 shows a -1/2 power dependence on the pressure on ZnO and ZrOz [ 10.1 11. There are indications that the order changes to -1 at higher pressures on ZnO [12]. On anatase TiOz, the order dependence changes from -1/2 to -1 as the pressure increases [lo]. An order of -1 has been reported on a single crystal rutile TiOz sample [13]. Furthermore, cr increases linearly with light intensity. The following processes have been proposed to explain the order dependence [lo]:

02(ad) O(ad)

+ e- + 02-(ad)

+

e-

+ 0-(ad)

(14-7) (14-8)

The order in oxygen changes from -1/2 to -1 when thc dominant reactions change from (14-6) and (14-8) to (14-5) and (14-7). Other processes have been excluded. The reaction 02-(ad) + e- +2 0-(ad) would lead to only a -1 order in oxygen, and the react. m: 02(ad) + 2 e- +022-(ad) +2 0-(ad) would lead to a half order dependence on light intensity. The surface condition is important. It has been shown that the amount of oxygen photo-adsorbed on Ti02 increases with the amount of hydroxyl groups on the surface [6]. A similar observation has been reportcd that dehydroxylation of

258

the surface reduces photo-adsorption of oxygen on TiO2, which can be restored by rehydrating the surface [7,14]. It is believed that hydroxyl groups are important in trapping photo-generated holes, making photo-electrons available for chemisorption: OH-

+

h+ -+ .OH(ad)

(14-9)

In this case, the fate of the hydroxyl radical is unclear. In the presence of hydrocarbon molecules, the radical may participate in oxidation reactions, as will be discussed later. In addition to photo-enhanced adsorption or desorption of oxygen, photolysis of the lattice resulting in evolution of lattice oxygen has been reported on ZnO [8,9]. On the other hand, photo-decomposition of Ti02 has not been observed [91. The photo-responseof a sample depends on its pretreatment, especially if the pretreatment results in a slightly reduced or fully oxidized sample. Reduction usually leads to n-type semiconductivity, and the extent of reduction determines the density of conduction electrons and the lifetime of photocarriers. For some oxides, severe oxidation may lead to excess lattice oxygen and p-type conductivity. Coupled with the effect of surface hydroxylation, it is not surprising that there are conflicting reports in the literature on the effect of illumination. Doping of an oxide changes its semiconducting properties and response to irradiation. Addition of Li to ZnO enhances photo-adsorption of oxygen, while addition of Ga or A1 reduces it [3,4]. Ti02 and ZnO are among those that demonstrate large photo-effects. Under the same conditions where these two oxides show photo-adsorption behavior, adsorption of oxygen on V205 does not show any response to light, and W 0 3 shows only a weak response [lo]. This behavior parallels their photo-enhanced catalytic oxidation activities which will be discussed next.

14.3 PHOTOCATALYSIS: GAS PHASE REACTIONS A large number of gas phase catalytic reactions have been found to be greatly enhanced by illuminating an oxide with light of band gap energy. These reactions include: oxidation of hydrocarbons oxidation of alcohols oxygen isotope exchange oxidation of CO, NH3 H2-D2exchange In general, these reactions proceed readily at elcvated temperatures in the dark. Thus most of the studies of light enhancement have bcen conducted near room temperature. Product selectivities are generally different between thermal and light-assisted reactions.

PHOTO-ASSISTED PROCESSES

259

Table 14-1 Partial Oxidation Products in the Gas Phase Photocatalytic Oxidation of Hydrocarbons at Low Conversions. Hydrocarbon

Oxide

Partial Oxidation F’roducts Major Minor

ethane propane

Ti02 Ti02 Ma3

ethanal acetone acetone acetone ethanal, acrolein

propene

wo3

Ti02

wo3

2-methylpropane 2-methylpropene 2,2-dimethylpropane butane

Ti02 Ti02 Ti02

ethanal, acrolein acetone acetone acetone

Ti02

butanone

2-methylbutane

Ti02

acetone

2,2-dimethylbutane

Ti02

3,3-dimethylbutan2-one

2-methyl-1-butene 3-methyl-1-butene 3-methyl-2-butene pentane

Ti02 Ti02 Ti02 Ti02

2-methy lpentane

Ti02

butanone, ethanal acetone acetone, ethanal pentan-2-one, pentan-3-one 2-methylpentan3-one, 4-methylpenta.n2-one, acetone phenol substituted benzaldehyde

toluene p-alkyltoluene

TiOz Ti02

propanal, ethanal propanal propene oxide, acetone, propanol

2,2-dimethylpropanal butanal, propanal, ethanal butanone, ethanal, 3-methylbutanone butanone, acetone, 3,3-dimethylbutanal, propanal, ethanal acrolein, acetone

pentanal, butanal, propanal, ethanal 2-methylpentanal, propanal, ethanal

Oxidation of Hydrocarbons Oxidation of hydrocarbons is among the most studied reactions. In particular, the oxidation of alkanes have been studied heavily. Table 14-1 lists thc partial oxidation products for a number of alkanes. It is clear from the data using

260

Table 14-2 Selectivities for Ketone and Aldehydes for Different Alkanes. (From Farad. Disc. Chem. Soc., 58, 185 (1974). copyright Royal Society of Chemistry). % Selectivity (excluding C o d

Alkanes n-alkanes alkanes possessing one or two tertiary carbon atoms neoalkanes

Ketones

Aldehydes

56 61

20 19

43

14

TiOz as the catalyst that photocatalytic oxidation produces a wide range of products. In addition, combustion is usually a significant reaction, especially at conversions higher than a few percent. However, there are some notable exceptions. The oxidation of 2-methylpropane to acetone on Ti02 proceeds with up to 75% selectivity [15], and the oxidation of 3-methyl-1-butene to acetone proceeds with 87% selectivity [16]. In general, ketones and aldehydes are the partial oxidation products for the hydrocarbons shown in Table 14-1, as well as for hexane, heptane, octane [17] and p-alkyltoluenes [52]. It should be noted that cracking of the carbon chain to products of lower carbon numbers is a common process. No dehydrogenation products (alkenes, dienes, etc.) have been reported as significant, although they have been suggested as reaction intermediates. The production of ketones is more preferred than aldehydes at low conversions. Teichner and coworkers have calculated the average selectivities for a large number of hydrocarbons, and grouped these values according to their structures [17]. These values are shown in Table 14-2. Apparently neoalkanes are more susceptible to combustion than other alkanes. Combustion is also the only reaction reported for butenes [15]. The reactivity also depends on the nature of the alkane. In particular, the reactivity of various types of carbon atoms on Ti02 follows the sequence:

The reactivity do not seem to follow the ionization potential of the hydrocarbons ~71. Although Ti02 has been the most extensively studied, and it shows the highest quantum efficicncy, it is not as sclcctive as W03, Moo3 or some nontransition metal oxides such as antimony oxide and tin oxide [ 18,191. In the formation of ketones and aldchydes on Ti02, it has been proposed that the first step of the reaction is the formation of an alcohol. If a primary alcohol is formed, it may dehydrogenate to an aldchyde. Dehydrogcnation of a secondary alcohol would form a ketone. A primary, secondary, or tertiary alcohol may also

PHOTO-ASSISTED PROCESSES

26 1

dehydrate to form an alkene, which could then be oxidatively cleaved to shorter chain ketones and aldehydes [161. These reactions are illustrated below for 2methylbutane:

CH3 I CH3CH2CHzCH3

CH3 I CH3CHCH2CH2OH

0

-H20

1

CH 3 I CH3CHCH=CH2 CH3 I CH3CHCH2CH3

5 -H2

CH3 I CH3CHCH2CHO (14- 10)

H2d

CH3

I 0CH3CHCHCH3 I

0

CH 3 I CH3CHCHO + COz CH 3 I CH3CHCCH3

6

OH

(14-11)

(14-12)

-H20 J CH 3 I CH3CHCH=CH2 : : & CH3CHCHO + COz 2

+ CH 3 I CH3C=CHCH3

0 2

CH3COCH3 + C02

(14-13a)

(14-13b)

OH

HZ CH3COCH3+ CH3CH0 This mechanism has been substantiated by studying the photo-oxidation of the appropriate alcohols and alkenes. Oxidation of 3-methyl-1-butanol yields 52% 3methylbutanal, 20% 2-methylpropanal and acetone: 3-methyl-2-butanol yields 39% ethanal, 34% acetone, 17% 3-methyl-2-butanone,and 10% 2-methylpropanal; and 2-methyl-2-butanol yields 50% acetone and 42% ethanal. Oxidation of 3methyl-I-butene yields 87% acetone and small amounts of 2-methylpropanal, acrolein and ethanal; and 2-methyl-2-butene yields 35% acetone, 29% ethanal, and some unidentified products. These results support the mechanism illustrated above.

262

Another mechanism for the production of alcohols and ketones involves a peroxy intermediate. This is illustrated for the oxidation of propane [20]: C3Hs

+

+

O-(ad)

0 2 +

+

C3H700.

OH-(ad)

(14- 15)

The peroxy species then further reacts to form an aldehyde. In the photocatalytic oxidation of propane on silica-supported Mo03, it is determined by varying the residence time that acetone, propanal, and propanols are primary products. As the residence time increases, the secondary products C02, CH3CH0 and methanol increase. The production rates of all primary products show a nearly first order dependence on light intensity, while the secondary products show an order dependence that is substantially larger than unity [19]. Since acetone appears as a primary product, its production cannot be via an alcohol intermediate as suggested earlier. The following mechanism has been proposed for Moo3: C3Hs

+

0-(ad)

-+

C3H70++ e-(lattice)

-+

C2HsCHO

n-C3H70.

+

O2

n-C3H70.

+

C3Hs

i-C3H70.

+

O2

+

-+ n-C3H70H

+

(CH3)2CO

(14- 16)

H02-

(14- 17)

+

(14-18)

C3H7

+ HO2.

(14- 19)

i-C3H70- + C3H8 -+ (CH3)2CHOH + C3H7

(14-20)

Since a C-H bond is stronger than a C&H bond, it is not clear how efficient are reactions (14-18) and (14-20) for the production of propanols. It is possible that there are other routes to produce these alcohols. Although most oxidation reactions produce a broad distribution of partial oxidation products, the oxidation of 2-methylpropane (isobutane) on anatase Ti02 produces almost exclusively acetone [21]. The following accounts for up to 95% of the reaction of 2-methylpropane: CH3 I CH3-CHXH3

+ 5/20 2

+

CH3COCH3

+

C02

+

2 H20

(14-21)

Since the activity in the photocatalytic oxidation of hydrocarbons is proportional to the light intensity, and the reaction only proceeds in the presence of oxygen, it has been assumed that photo-adsorbed oxygen and photo-generated holes are both important. The detailed roles of these species, however, have not been unequivocally identified. The photo-adsorbed 0-(ad) species is often assumed to be the species that activates alkanes by the abstraction of a hydrogen atom [21]: CnHzn+2

+ 0-(ad)

-+

.C,,H2"+,

+ OH-(ad)

(14-22)

PHOTO-ASSISTED PROCESSES or 0-(ad)

+

263

h+ -+ O(ad)

(14-23)

An alternate activation mechanism involves surface OH species that are formed when a surface hydroxyl group captures a hole:

or

OH-(ad)

+

02-(ad)

+ .OH(ad)

h+ + .OH(ad)

+ 0-(ad)

(14-25)

+

02H(ad)

(14-27)

Reactions (14-27) and (14-28) have been suggested because the reaction requires gaseous oxygen and does not take place with N 2 0 as the oxidant [17]. This has been taken to indicate that 02-(ad) is necessary. The participation of OH groups, however, may not be important on TiO2. It has been shown that a highly dehydroxylated anatase sample photo-oxidizes 2-methylpropane as actively as a partially dehydroxylated sample [22]. On the other hand, using isotopically labeled oxygen, it has been observed on silica-supported vanadia that lattice oxygen participates in the photo-oxidation of propene to acetaldehyde [32]. Since the photo-activity of vanadia is relatively poor, and the participation of lattice oxygen in thermal oxidation reactions on vanadia is well established, it is not clear if this observation applies to titania or other oxides. The fact that lattice oxygen does participate in CO oxidation on titania suggests that this may be important in some cases [33]. The participation of photo-adsorbed oxygen is also supported by the fact that the activity in photocatalytic oxidation over various oxides can be broadly correlated with their photo-adsorption properties [10,18]. Thus TiOz and Zr02, which are very active in photo-assisted adsorption of oxygen, are also very active in photocatalytic oxidation of propene. V205, which does not demonstrate photoassisted adsorption of oxygen, is quite inactive. The different selectivities on different oxides for partial oxidation products seem to parallel the trend in thermal oxidation. However, these correlations are at best qualitative. Changes in oxide morphology after reaction have been reported. On silicasupported molybdena, isolated tetrahedrally coordinated Mo6+ ions crystallizes to Moo3 crystallites after being used in the photocatalytic oxidation of cyclohexane 1231. Similar crystallization has also been reported for silica-supported V2O5 [24]. It is interesting that V2O5 is inactive in the oxidation of propene, but V205/Si02 is active in the oxidation of cyclohexane.

Oxidation of Alcohols Photocatalytic oxidation of alcohols at room temperature in the presence of oxygen proceeds readily on a number of oxides. For example, the oxidation of 2-

264

propanol on rutile Ti02 results almost exclusively in acetone [14,25]. A small amount of acetaldehyde is also produced which may be formed from the photooxidation of propene which is a dehydration product of 2-propanol. Continual exposure of the reaction mixture to illuminated Ti02 results in further oxidation of acetone to formic acid, and eventually C02 and water. However, the rate of oxidation of acetone is slow [171. Other selective oxidation of alcohols have been reported. On Ti02 butanol is oxidized to butanal, and 2-butanol to butanone [26], although there is a report that acetaldehyde is also a significant product from 2-butanol [30]. Thus oxidative cleavage of the carbon chain may occur. In some cases, this is the predominant reaction. For example, 2-methyl-2-propanol is oxidized to acetone quite selectively [26], 3-methyl-2-butanol is oxidized to ethanal and acetone, and 2methyl-2-butanol to acetone and ethanal. 3-Methylbutanol is oxidized primarily to 3-methylbutanal but there is a substantial production of 2-methylpropanal as well [16]. From the product distributions in the reactions of the appropriate alkenes, it has been suggested that oxidation of alcohol is the first reaction step for primary alcohols. For secondary and tertiary alcohols, the first step is dehydration to the corresponding alkenes, which are then oxidized by the photo-generated adsorbed oxygen 1271. Photo-enhanced oxidative dehydrogenation of 2-butanol and 2-propanol to the corresponding ketones have also been observed on MnOz and ZnO [30,31]. The photo-activity declines with time. Illumination has no effect on V205. Cr2O3, Fe2O3. Co3O4.NiO, or CuO. Photo-oxidation of methanol at 403 K on anatase Ti02 results in methyl formate up to about 60% conversion [281. It is proposed that an adsorbed methoxy is photo-oxidized to an adsorbed formate, which reacts with another methoxy to form methyl formate. The reaction rate at room temperture is very low, although formaldehyde is readily oxidized [22]. When the titania is impregnated with Moo3, the quantum efficiency decreases with increasing molybdena content to about 10%of the molybdena-free value when there is about a monolayer equivalent of molybdena. The rate of production of methyl formate is low, and the major product is dimethoxymethane. It is proposed that an adsorbed methoxy is first oxidized to an adsorbed formaldehyde which condenses with adsorbed methoxy groups to form dimethoxymethane [28]. The photooxidation of methanol to methyl formate has also been reported for ZnO [29].

Oxygen Isotope Exchange Light-enhance oxygen isotope exchange reaction has been studied in detail over Ti02 and ZnO. Both the homophasic exchange reaction: (14-29) and the hcterophasic exchange reaction: l8O2(g)

+

l6O(lattice) 4 l60l8O(g)

+

ls0(lattice)

(14-30)

PHOTO-ASSISTED PROCESSES

265

proceed readily at room temperature under band gap illumination. Exposure of oxygen to an illuminated Ti02 at 77 K results in the appearance of an EPR signal assigned to 03-(ad). This species is thermally unstable and decomposes to adsorbed 0- and O2 on warming to 137 K, but its intensity can be correlated with the catalytic activity for homophasic exchange of oxygen over the sample at 137 K [34]. Thus this species has been proposed as a reaction intermediate in the homophasic exchange reaction which proceeds as: l802(g) + l60-(ad)

+

03Jad)

-----)

l60l8O(g)

+

l80-(ad)

(14-31)

There is evidence that isotopic scrambling between adsorbed 0 2 (or 0 2 3 species does not contribute to this reaction. On a Ti02 sample previously exposed to 1802 and 1602,it has been found that equilibrium distribution of isotopic oxygen species in the gas phase can be obtained rapidly when the sample is illuminated. However, the isotopic composition of the adsorbed oxygen is very different from the gas phase composition at the end of the reaction [35]. In fact, it is similar to the composition of adsorbed oxygen from a sample that has not been used for the reaction. This has been found by thermally desorbing the adsorbed oxygen after reaction. Thus adsorbed O2 species are not involved in the homophasic exchange reaction. Monatomic oxygen species are also involved in the heterophasic isotope exchange reaction on Ti02. When anatase Ti02 is exposed to l S 0 2 under illumination, 160180 first appears in the gas phase followed by 1602 in a manner characteristic of stepwise reactions: 1802 * l60l8O+ 1602.Thus the reaction proceeds as [36]:

1802(g) + 160(ad) + l6OI8O(g) + lS0(ad)

(14-32a)

160180(g)+ 160(ad) + 1602(g) + I80(ad)

(14-32b)

Both homophasic and heterophasic reactions occur simultaneously on an illuminated oxide [37,38]. The homophasic reaction proceeds faster than the hctcrophasic reaction on ZnO pretreated with oxygen or prereduced. This contrasts the thermal reaction for which the homophasic reaction is greatly suppressed by pretreating the oxide with oxygen. The extent of surface hydroxylation may be important also. The photocatalytic activity of a partially dehydroxylated TiOz is higher than a fully hydroxylated sample [21]. Other photo-oxidation reactions compete with the photocatalytic oxygen isotope exchange reaction. In the presence of 2-methylpropane, oxygen isotope exchange is completely suppressed on Ti02 until the oxidation of the hydrocarbon is complete [36]. On ZnO, CO oxidation suppresses the oxygen isotope exchange reaction [33,39]. These results suggest that these reactions either proceed on the same sites or involve a common intermediate. Indeed, on TiO2, the activity per unit area for oxygen isotope exchange correlates with that for 2-methylpropane oxidation [36], and the participation of 0-(ad) in both reactions has been suggested.

266

Other Reactions Many other photocatalytic reactions have been observed [2,40]. Oxidation of C O [41] and NH3 [42], H2-D2 exchange [43], decomposition of NO [26] and N 2 0 [ a ] , and reduction of N2 [45] have been reported. In the oxidation of NH3 over Ti02, N2 and N20 are the products observed in addition to H20. From conductivity measurements, it appears that the species involved in this reaction, NH3, N2 and N 2 0 do not compete for photo-generated electrons with adsorbed oxygen. The decomposition of NO produces N2 as the product at low NO pressures, and N 2 0 at high NO pressures. Adsorbed oxygen is the other product at room temperature. The photo-enhanced H2-D2 exchange reaction has been analyzed in detail [2]. It is assumed to proceed by heterolytic dissociative adsorption of hydrogen. Participation of lattice oxygen in CO oxidation has been demonstrated for Ti02 [33]. Using a mixture of CO and the photo-enhanced production of C1602 is much larger than C 1 6 0 1 s 0at the beginning of the reaction. Photo-reduction of N2 to form NH3 has been successfully demonstrated by passing N2 saturated with water vapor over illuminated W 0 3 [45]. No N2H4 is formed. If the tungsten oxide is slightly reduced before the reaction, it is reoxidized during the reaction. The reverse reaction, decomposition of ammonia, is also photocatalyzed. Photo-reduction of N2 also occurs on illuminated anatase TiO2. Ammonia and traces of N2H4 are formed [461. Rutile Ti02 doped with Fe is quite effective in this reaction.

14.4 PHOTOCATALYSIS: LIQUID PHASE REACTIONS In gas phase photocatalytic reactions, charged reaction intermediates must remain adsorbed on the surface. On the other hand, charged species can readily exist in the liquid phase. Thus it becomes possible to remove an electron from an anion in the solution to form a neutral species by a photoelectrode, or inject an electron into a neutral species to form an ion. The resulting species could then further react in solution.

Oxidation Reactions The oxidation of oxalate and formate ions to C 0 2 by an illuminated ZnO in an aqueous suspension has long been reported [471. Hydrogen peroxide is a byproduct:

C2Od2- + HCOO-

+

0 2

0 2

+ 2 H2O

+

H20

---+

2 C02

C02

+

+

2 OH-

OH-

+ H202

+ H202

(14-33) (14-34)

Under similar conditions, 2-propanol is oxidized to acetone:

(CH3)ZCHOH + 0

2

(CHj)ZCO + H202

(14-35)

PHOTO-ASSISTED PROCESSES

267

The oxidative decarboxylation of acetate ions can be effected with illuminated mile TiOz. In the absence of oxygen, ethane is a product which is formed by coupling of two methyl radicals [48]:

-

CH3COz2CH3

+

h+

CH3

+

COz

(14-36)

Cz&

(14-37)

The production of ethane is greatly suppressed by the presence of oxygen. Presumably the methyl radicals react to form methanol and formaldehyde [49]. Photo-assisted oxidation has also been demonstrated at a TiO, electrode for hydroquinone, p-aminophenol, I-, Br-, C1-, Fez+, Ce3+, and CN- ions. M a n y of these species are oxidized at potentials negative of their standard redox potential

POI. It is interesting to note that in the photo-assisted decarboxylation of a dicarboxylate on TiOz, the reaction stops at the monoacid form. Perhaps this is due to the displacement of the monoacid from the surface by the diacid [Sl]: C02H O C 0 2 H

T i O z , hu CH3C

N

0-

COzH

+

COz

(14-38)

Photo-assisted oxidation of hydrocarbons has been studied rather extensively [511. In an aqueous suspension of TiOz, substituted toluene reacts to form both oxidation and coupling products [55,56]:

(14-39)

R

This contrasts sharply the relatively selective production of substituted benzaldehyde in the gas phase reaction [52]. Cyclohexene is likewise oxidized to a mixture of cyclohexenol and cyclohexenone [51]. Oxidative cleavage of 1 ,l-diphenylethylene to benzophenone can be achieved efficiently with TiOz [53]: (14-40)

268

Substituted naphthalene is photo-oxidized with cleavage on one of the benzene rings. For example [51]:

(14-41)

Amines can also be photo-oxidized. On ZnO, aniline is oxidized to azobenzene, and toluidines are oxidized to the corresponding azo compounds. The relative rates depend on the substituent, and the order meta > ortho > paratoluidine > aniline has been observed [54]. The mechanism of liquid phase photo-assisted oxidation of hydrocarbons with TiO, has been investigated in detail [51]. Several observations support the involvement of a radical cation intermediate:

RH

+

h+ ---+ RH+

(14-42)

Stilbene radical cation is observed spectroscopically after a Ti02 colloidal suspension in acetonitrile containing trans-stilbene has been flashed with a nitrogen laser [55]. Efficient photo-conversion of diphenylelhylene is achieved only in solvents of high dielectric constants. This is consistent with the involvement of a highly polar transition state. When para-substituted diphenylethylene is oxidized with Ti02, a linear Hammett plot with a negative slope is obtained between the relative rate of reaction and the substitution parameter 1561. The relative rates of reaction of other substituted diphenylcthylene molecules also show reduced reactivity with decreased K- electron density of the double bond. Thus these relative rates reflect the stability of the one-electron oxidation product, consistent with the proposed involvement of radical cations. Subsequent reactions of the radical cations may involve capture by triplet molecular oxygen. They may also involve attack by superoxide. In either case, the reactions are not photo-assisted. If the hydrocarbon molecule is an alkene, dioxetane is then formed which would be cleaved to a ketone or an aldehyde [51]. The oxidation of water has been studied extensively. Earlier work has been on the formation of hydrogen peroxide 12,471. The reaction mechanism can be either: H20

+

0-(ad)

HOO-(ad) or

H20

+

OH-(ad)

+

+

--+

HOO-(ad)

H+(ad)

+ H202

0-(ad) + OH-(ad)

+ OH(ad) ---+

+

H(ad)

(14-43) (14-44)

OH(ad)

Hz02+ e-(lattice)

(14-45) (14-46)

PHOTO-ASSISTED PROCESSES

269

Adsorbed electron acceptors retard the reaction. Thus adsorbed 0, [57] or adsorbed bicarbonate [58] slow the reaction. Recently, the interest has been shifted to the photo-assisted electrochemical oxidation of water to oxygen. It has been discovered that band gap illumination of a Ti02 anode substantially reduces the applied potential required for the evolution of oxygen, as shown in Fig. 14-3 [59]. Many other oxides have been subsequently shown to demonstrate a similar effect [60,61,68,69]. They include W03 [62], nFez03 [63,64], and mixed oxides such as SrTi03 [65], KTa03 [66] and others. In the absence of other more readily reducible species, proton is reduced to hydrogen at the cathode when a sufficiently high potential is applied. Thus the net reaction is the decomposition of water: 2 H20 + 2 H2

+

0 2

(14-47)

With the more efficient photo-anodes such as those listed above, the external potential needed to achieve this reaction is substantially less than the thermodynamic value of 1.23 V when the photo-anode is illuminated. Therefore, such a system serves to convert light energy into chemical energy. However, only the large band gap oxides such as SrTi03 and KTa03 can sustain the reaction without any applied external potential. In fact, it appears that the magnitude of the applied potential is roughly inversely correlated with the width of the band gap [611. The oxidation reaction at the anode may involve the photo-production of OH radicals which form H202 [671:

Hydrogen peroxide decomposes to water and oxygen in a basic solution. The reduced titania is reoxidized by water to regenerate the hydroxylated surface. In fact, if deuterated water is used in the reoxidation, D, is evolved 1671:

Reduction Reactions Photo-assisted reduction of metal complexes to metallic particles deposited on the oxide can be achieved in a number of systems. The reduction is accompanied by oxidation of another species. For example, copper ions can be reduced to copper metal deposited on illuminated Ti02 or W03 particles in water and oxygen is produced [701.

270

I

cu

E

I

0

+ c

2

1.0

L

3

0

.-U0 2

a

0

I

1I

V(SCE)

-0.5

Figure 14-3 Anodic current as a function of voltage (versus SCE) at a semiconducting anode in a pH 4.7 electrolyte. The potential for the evolution of oxygen from the oxidation of water is lowered by illumination. a. Anode illuminated with strong light; b. with weak light; c. No illumination. (After Nature, 238, 37 (1972), copyright Mcmillan Publisher).

Cu2+

+ H20

-T i 0 2 , hu

CU + 1/2 0 2 + 2 H+ (14-50)

In the presence of acetate, acetate is oxidized to C02: Cu2+

+ 2CH3COO-

Cu

+

C2&

+

2C02

(14-51)

Copper can also be deposited on KTa03 and SrTi03 [71]. The deposition of Pt from a solution of hexachloroplatinic acid and carbonate onto irradiated Ti02 has also been achieved. The carbonate is converted to C02 [72]. Other platinum complexes can likewise be used [73]. The reduction of chloroplatinic ion is accompanied by a release of proton that has been quantitatively detected. In the reduction of dinitrodiaminoplatinum, N20 and N2 are evolved. Transmission electron microscopy shows that platinum is initially deposited as very small particles distributed over the Ti02 particles. Deposition of other metals, including Pd, Ag, Rh. Au, and Ir on Ti02 and on other oxides including ZnO, Nb2OS and Tho2 have been reported [73, and references therein]. The rate of deposition depends on the strength of adsorption of the metal ion and the quantum efficiency for reduction. For the reduction of Ag+- the adsorption of the ion on ZnO is weaker than on Ti02, but the quantum efficiency for reduction is higher. The relative rates of deposition on the two oxides also depend on the concentration of the ions in the solution [74]. There are reports on the photo-assisted reduction of water by oxides. p-Type

PHOTO-ASSISTED PROCESSES

27 1

Fez03 has been successfully used with n-type F e 0 3 to decompose water. Water is reduced to H2 at the p-type iron oxide [75]. p-LuRh03 has also been reported as effective in this reduction [76].

14.5 PHOTOCATALYSIS BY METALMETAL OXIDE AND OXIDE/OXIDE COMPOSITES It can be Seen from the examples described above that some oxide particles or electrodes are quite efficient in effecting photo-assisted electron transfer. For an n-type semiconducting oxide, this has resulted in enhancement in many oxidation reactions. However, for an effective photocatalyst, every step of the catalytic cycle must proceed efficiently. This requirement may not be met by many oxides. For example, photo-oxidation on Ti02 is commonly observed; but photo-reduction on this oxide is much less known. Thus electron transfer out of Ti02 is more difficult than into the oxide. Therefore, attempts have been made to deposit metal or another oxide on a semiconducting oxide that will facilitate electron transfer in the difficult direction. Noble metals and RuO2 have been used as deposits for this purpose. Sometimes a composite of n- and p-type semiconducting oxides are used when a single oxide does not generate sufficient photoelectrochemical driving force to complete a desired reaction. Then a composite of two appropriate semiconducting oxides could be used to supply a combined photoelectrochemical driving force. Successful decomposition of water using visible light has been achieved using such devices. Such composite systems have been used in the oxidation of ethanol. Ethanol is oxidized in a water mixture by an irradiated Wi02to acetaldehyde and acetic acid. Hydrogen is also evolved [77]:

CH3CH20H hu, CH3CH0 CH3CHO

+

+

H2

H20 -!% CH3COOH

(14-52)

+

H2

(14-53)

After prolonged irradiation, methane and C02 are observed due to the decarboxylation reaction of acetic acid [78]. CH3COOH

C&

+ C02

(14-54)

A side reaction may also occur via the reaction of acetaldehyde with OH radical, which eventually leads to combustion:

CH3CHO

+

3 H2O

5 HzO

+

2 C02

(14-55)

Photo-assisted decarboxylation of acids takes place efficiently on Wanatase Ti02 in the absence of oxygen. Wanatase is more effective in this reaction than Wrutile, possibly because anatase has a slightly larger band gap than rutile. The

272

Table 14-3 Products in the Photo-assisted Decarboxylation of Acids on m i 0 2 Acid

Major Products

Minor Products

acetic propionic n-butyric n-valeric pivalic adamantane-1carboxylic adipic

cH4, co2 C2&. co2 propane, C02 n-butane, C02 isobutane. C02 adamantane, C02

C2&. H2 c2H4, H2 H2 H2 isobutene. H2

butane, C02

valeric acid

products of the reaction are listed in Table 14-3 [78,8S]. These products can be accounted for by a general reaction scheme:

h+

+

RC02-

e-

+

RCOOH

e-

+

R.

2 H(ad)

R.

+

+

-Re

H(ad)

RCOOH

--

H(ad)

+ CO2

(14-56)

+ RCOORH

+ RCOO-

H2

RH

It is of significance to note that H2 is a product in these reactions because Pt is an excellent catalyst for the recombination of H atoms. Pd or Ru02 can be used instead of R, but the efficiency in H2 evolution follows the order: W i 0 2 2 Pd/Ti02 > RuOJI’iO2 >> Ti02 1791. Complete decarboxylation of adipic acid, HOOC(CH3)4COOH,to C4H10 can also be achieved with illuminated m i 0 2 [85]. Oxidation of alkanes, including hexane, cyclohexane, heptane, nonane, and decane in an aqueous solution results in practically complete combustion [86]. The photo-decomposition of water has been studied extensively. Pt/SrTi03 is an efficient catalyst [80], more efficient than Wrutile Ti02. This is because SrTi03 has a smaller electron affinity and therefore a larger band bending at the surface [81]. For the Pt/TiO2 system, anatase is more efficient than mile. A small amount of Nb205 doping further enhances the efficiency [82]. Platinked

PHOTO-ASSISTED PROCESSES

273

KTa03 also decomposes water on illumination [711. Oxide/oxide composites have been used for this reaction. A NiO/SrTi03 composite, prepared by impregnation with Ni(NO3)2 followed by reduction and reoxidation at 500°C decomposes water effectively upon irradiation. Nearly stoichiometric amounts of H2 and O2 are produced [83]. The reduction step is necessary to produce metallic Ni to form a NiO/Ni/SrTi03 interface in which Ni serves as an ohmic contact. The activity in NaOH is higher than in water, perhaps because hydrogen peroxide decomposes more efficiently in basic than neutral solutions. Another system reported to be successful is a n-Fe203/pF%03 composite [75,84].p-Fe203 is made by Mg doping. The actual material in the composite is a mixture of MgO, p-Fe203 and MgFe204. The activity of the system declines slowly with time. It can be regenerated by oxygen treatment of the oxide.

REFERENCES 1. S. Baidyarov, W. Bottoms, and P. Marks, Surface Sci., 28, 517 (1971). 2. Th. Wokenstein, Adv. Catal.. 23, 157 (1973). 3. F. Romero-Rossi, and F. Stone, Proc. 2nd Infern. Cong. Catal., 1970, paper 70. 4. T. Kwan, in "Electronic Phenomena in Chemisorption and Catalysis on Semiconductors," p. 184, de Gruyter, Berlin, 1969. 5. K. Tanaka, and G . Blyholder, J. Phys. Chem., 76. 3184 (1972). 6. A. BoonstTa and C. Mutsaers. J . Phys. Chem., 79. 1694 (1975). 7. F. Stone, Coloq. Quim. Fis. Processos Superficies Solides, 1965, p. 109. 8. R. Collins, and D. Thomas, Phys. Rev., 112, 388 (1958). . Finn, and N. Samman, Farad. Disc. Chem. Soc., 58, 160 (1974). 9. J. C u ~ i n g h a m E. 10. J.-M. Herrmann. J. Disdier. and P. Pichat. J . Chem. Soc.Farad. T r m . I, 77, 2815 (1981). 11. W. Doerffler, and K. Hauffe. J . Catal., 3, 156 (1964). 12. H. Degn. and J. Nabbs, Appl. Phys. Leff., 26. 526 (1975). 13. R. Keezer, J. Appl. Phys., 35, 1866 (1964). 14. R. Bickley. and R. Jayanty, Faraday Disc. Chem. Soc.,58, 194 (1974). 15. F. Formenti, F. Juillet, P. Meriaudeau. and S. Teichner. Chem. Tech.. 1, 680 (1971). 16. N. Djeghri, and S. Teichner, J . Cafal., 62, 99 (1980). 17. N. Djeghri, M. Formenti. F. Juillet, and S . Teichner. Farad. Disc. Chem. Sm., 58, 185 (1974). 18. P. Pichat, J.-M. Herrmann, J. Disdier, and M.-N. Mozzanega, J . Phys. Chem.. 83, 3122 (1979). 19. K. Marcinkowska. S . Kaliaguine, and P. Roberge, J . Catal., 90, 49 (1984). 20. V. Kazansky. Kinet. Katal.. 19, 279 (1978). 21. J.-M. Herrmann. J. Disdier, and M.-N. Mozzanega, and P. Pichat, J . Catal., 60, 369 (1979). 22. M. Formenti. F. Juillet, P. Mariaudeau. and S. Teichner, Proc. 5th Intern. Cong. Cafal.,11. p. 1011 (1972). 23. A. Mahay, S. Kaliaguine, and P. Roberge, C a d . J . Chem. Eng.. 60. 2719 (1982). 24. D. Nguyen. P. Roberge. S. Kaliaguine. Canad. J . Chem. Eng., 57, 288 (1979).

214 25. R. Brickley, G. Munuera, and F. Stone, J. Cataf..31,398 (1973). 26. P. Pichat. H. Courbon, J. Disdain, M. Mozzanega, and J. Hemnann, Proc. 7th Intern. Cong. Catal.. XI. p. 1498 (1980). 27. A. Walker, M. Formenti, P. Meriaudeau, and S. Teichner, J. Catal., 50, 237 (1977). 28. Y. Liu, G. Griffin. S. Chan, and I. Wachs, J. Cataf..94. 108 (1985). 29. G.Schwab, F. Steinbach, H. Noller, and M. Venugopalan, 2. Naturforsch. A, 19, 445 (1964). 30. J. Cunningham, B. Hodrett, M. Ilyas. J. Tobin, and E. Leahy, Faraday Disc. Chem. SOC.,281 (1981). 31. J. Cunningham, and B. Hodnett, J. Chem. SOC. Farad. Tram. 1, 77, 2777 (1981). 32. S. Yoshida, T. Tanaka, M. Okada. and T. Fundabiki, J. Chem. SOC. Faraday Tranr. 1, 80, 119 (1984). 33. S. Sato. T. Kadowaki, and K. Yamaguti. J. Phys. Chem., 88, 2930 (1984). 34. V. Nikisha, B. Shelimov. and V. Kazanskii, Kinet.Cataf., 15,599 (1974). 35. K. Tanaka, J. Phys. Chem., 78, 555 (1974). 36. H. Courbon, M. Formenti, and P. Pichat. J. Phys. Chem., 81,550 (1977). 37. J. Cunningham, and E. Goold, J. Chem. SOC. Farad. Trans. 1. 78, 785 (1982). 38. J. Cunningham, E. Goold, and E. Leahy, J. Chem. SOC. Farad. Trans. 1, 75,305 (1979). 39. K. Tanaka, and K. Miyahara, J. Phys. Chem., 78, 2303 (1974). 40. S. Morrison, 'The Chemical Physics of Surfaces." Plenum Press. N.Y. 1977, chapter 9. 41. C. Schwab, F. Steinbach. H. Noller. and M. Venugopalan, 2.Naturforsch. A 19, 445 (1964). 42. H. Mozzanega. J.-M. Hemnann, and P. Pichat, J . Phys. Chem.. 83. 2251 (1979). 43. Th. Freund, J. Catal., 2, 289 (1964). 44. K. Tanaka, and G. Blyholder, J. Phys. Chem., 75, 1037 (1971). 45. E. Endoh, J. Leland, and A. Bard, J. Phys. Chem., 90, 6223 (1986). 46. G. Schrauzer, and T. Guth, J. Amer. Chem. SOC., 99. 7189 (1977). 47. Th. Freund, and W. Comes, Catal. Rev., 3, 1 (1969). 48. B. Kraeutler, and A. Bard, J. Amer. Chem. Soc.. 99. 7729 (1977). 49. B. Kraeutler, and A. Bard, J . h e r . Chem. Soc.,100,5985 (1978). 50. S. Frank, and A. Bard, J. Amer. Chem. Soc.,99,4667 (1977). 51. M. Fox,Acc. Chem. Res., 16, 314 (1983). 52. M. Mozzanega, J. Herrmann, and P. Pichat, TefrahedronLetf.. 2965 (1977). 53. M. Fox, and C. Chen, J. Amer. Chem. Soc., 103, 6757 (1981). 54. H. Kasturirangan, V. Ramakrishnan, and J. Kuriacose. J. Cataf.,69, 216 (1981). 55. M. Fox, B. Lindig, and C. Chen, J. Amer. Chem. SOC.. 104. 5828 (1982). 56. M. Fox, and C. Chen, Tefrahedron Left., 24, 547 (1983). 57. G. Korsunovsky, Dokl. Akad. Nauk. SSSR, 113,853 (1957). 58. G. Calvert. K. Theurer. T. Rankin. and W. MacNevin, J. Amer. Chem. SOC.. 76, 2575 (1954). 59. A. Fujishima, and K. Honda, Nature, 238,37 (1972). 60. H. Gerischer, in "Solar Energy Conversion, Topics in Applied Physics." vol. 31. B. 0. Seraphin ed.. Springer, NY,1979. 61. H. Kung, H. Jarrett, A. Sleight, and A. Ferretti, J. Appl. Phys., 48. 2463 (1977).

PHOTO-ASSISTED PROCESSES 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86.

M. Butler, R. Nasby, and R. Quinn, Solid Sfate Commun.. 19, 1011 (1976). K. Hardee, and A. Bard, J . Electrochem. Soc., 123, 1024 (1976). R. Nasby, and R. @inn, Mater. Res. Bull., 11. 985 (1976). M. Wrighton, A. Ellis, P. Wolczanski, D. Morse, H. Abrahamson, and D. Ginley, J . Amer. Chem. SOC.. 98, 44 (1976). A. Ellis, K. Kaiser, and M. Wrighton, J . Phys. Chem.. 80, 1325 (1976). G. Somorjai, H. Hendewerk, and J. Turner, in "Catalysis and Surface Science." H. Heinemann and G. Somorjai, ed., Marcel Dekker, NY, 1985, p. 381. L. Harris, and R. Wilson, Ann. Rev. Mot. Sci., 8, 99 (1978). A. Nozik. Adv. Hydrogen Energy. Hydrogen Energy Syst., 3, 1217 (1979). H. Reiche, W. Dunn, and A. Bard, J . Phys. Chem., 83, 2248 (1979). M. Wrighton, P. Wolczanski. and A. Ellis, J. Solid Stare Chem.. 22, 17 (1977). B. Kraeutler, and A. Bard, J . Amer. Chem. Soc., 100, 4317 (1978). J. Henmann, J. Disdier. and P. Pichat, J . Phys. Chem., 90,6028 (1986). P. Fleichaner, H. Kan, and J. Shephard, J . Amer. Chem. Soc., 94, 283 (1972). C. Leygraf. M. Hendewerk, and G. Somorjai. J. Cutul., 78, 341 (1982). H. Jarrett, A. Sleight, H. Kung. and J. Gillson. J . Appl. Phys., 51, 3916 (1980). T. Sakata, and T. Kawai, Chem. Phys. Lett.. 80, 341 (1981). B. Kraeutler. and A. Bard. J . h e r . Chem. Soc.,100, 5985 (1978). T. Sakata. K. Hashimoto, and T. Kawai, J . Phys. Chem., 88, 5214 (1984). F. Wagner, and G. Somorjai, J . Amer. Chem. Soc., 102, 5494 (1980). J. Mavroides. J. Kafalas, and D. Kolesar. Appl. Phys. Letf., 28, 241 (1976). E. Bargarello, J. Kiwi, E. Pellizzetti, M. Visca, and M. Gratzel, J . Amer. Chem. Soc.. 103, 6324 (1981). K. Domen, A. Kudo, and T. Onishi, J. Cutul., 102, 92 (1986). Ch. Leygraf, M. Hendewerk, and G. Somorjai, J . Phys. Chem., 86, 4484 (1982). I. Izumi, F. Fan, and A. Bard, J. Phys. Chem., 85. 218 (1981). I. Izumi, W. Dunn, K. Wilbourn, F. Fan, and A. Bard, J . Phys. Chem., 84. 3207 (1980).

275

This Page Intentionally Left Blank

INDEX

Acetone, adsorption, 66 Acid sites, formation, 74, 75-80 Acidity, 72 -, determination, 80, 83, 87 -, effect of pretreatment, 82 -, selective oxidation, 175 -, strength, 74. 78 Acrolein, from propene. 181 Acrylonitrile, from propene, 181 Adsorbed oxygen, 110 -, desorption, 111 -, detection, 111 -, IR, 114 -, photo-adsorption, 256 -, photo-desorption, 256 -, reactivity, 116, 118 -, selective oxidation, 176 -, superoxide, 114 -, types, 110 Adsorption, 61 -, abstractive, 62 and cus, 57 -, heterolytic dissociative, 3, 35, 57, 61 -, homolytic dissociative, 61 -, molecular, 57, 61 -, pretreatment effect, 63 -, reductive, 62 AES, 4 Aerogel, preparation, 124 -I

Alcohol, adsorption, 66

-, decomposition. 146 -, photo-oxidation, 261, 263 Aldol condensation, in isosynthesis,243 Alkane, activation, 178 -, deuterium exchange, 161 -, photo-oxidation, 259 -, reaction with adsorbed oxygen, 116 Alkene, activation, 178 adsorption, 64 -, deuterium exchange, 161 -, hydrogenation, 156 -, reaction with adsorbed oxygen, 118 Alkyne, adsorption, 65 Alumina, isoelecmc point. 85 Ammonia, adsorption, 81 -, photo-oxidation, 266 Anion vacancies, surface, 55-57 Anisotropy constant, 43 Antiferromagnetism, 35 --,crystallite-size effect, 39 -, temperature effect, 39 Antimonoy-tin oxide, propene oxidation, 181

-.

Band gap. 252 Bi-Fe-Mo-0, propene oxidation, 186 Bismuth molybdate, 189, 204 -, butene oxidation, 173, 204

277

278 -, propene oxidation, 173, 181. 182.

185, 187, 188 Brdnsted acid, 72 -, formation, 75 Butane oxidation, 210 Butadiene, from butane, 210 -, from butene, 200 Butene, adsorption, 65 .-, dehydrogenation, 201, 203 -, hydrogenation, 159 -, oxidation, 88, 189, 200 Carbon-oxygen bond, formation, 179 Carbanion, in isomerization, 143 Carbenium ion, in isomerization, 142 Carbonate, adsorbed, 62 Carboxylate, adsorbed, 62 Catalytic processes, 2 Chromia (see chromium oxide) Chromium-iron oxide, 209, 244 Chromium oxide, adsorbed oxygen, 116 -, butene hydrogenation, 160 -, butene isomerization. 141, 144 -, CO adsorption, 58, 61 -, dehydroxylation, 55, 58 -, deuterium exchange, 161 -, methanol decomposition, 149 -, NO reduction, 165 -, preparation, 126 -, propene hydrogenation, 159 CO, adsorption, 58, 67 -, copper catalysts, 228, 230, 232 -, hydrogenation, 228 -, in C02 hydrogenation, 228 -, iron catalyst, 239 -, Pd catalyst, 236 -, Rh catalyst, 237 Cobalt molybdate. propene oxidation, 188, 192 Cobalt oxide, ethene hydrogenation, 158 -, isomerization, 143 Conduction band, 252 Contracting sphere model, 95 COZ, hydrogenation, 228 Coordinative unsaturation, 53 .-, conjugate base, 72

-, effect on structure, 15 -, formation, 53. 55 Lewis acid, 72 Copper oxide, methanol decomposition, 149 -, NO reduction, 162 -, propanol decomposition, 151 -, reduction, 98 Copper-thorium oxide, 230 Copper-zinc oxide, 228, 232, 245 Corundum structure, 10 Cubic close-packing, 8 Curie temperature, 39 Cuprous oxide, propene oxidation, 182. 185 Cus (see coordinative unsaturation)

-.

Decarboxylation, 267, 271 Decomposition, methanol, 146 -, water, 269, 272 Decoration effect. 102 Dehydrogenation, butane, 210 -, butene. 200 Dehydroxylation, 54 -, C r 2 4 , 58 -, zno, 58 Diamagnetism, 36 Dimerization, 215 Dissociative adsorption, 3, 35 Electronic structure, bulk, 252 -, surface, 46

ELS,4 EPR, 4 Ethane, from carbon oxides, 235 -, from methane, 215 Ethanol, from carbon oxides, 235, 238 photo-oxidation, 271 Ethene. from carbon oxides, 235 -, from methane, 215 -, hydrogenation, 156 EXAFS. 4

-.

Fez& (see iron oxide) Fermi energy, 252 Femmagnetism. 38

INDEX Ferrites (see iron oxide) Formaldehyde, from methane, 212 -, from methanol, 217 Formate, 62, 147, 245 m, 4 Hexagonal close-packing, 8 HREELS, 4 Hydrogerb adsorption, 63 -, reaction. adsorbed oxygen, 118 Hydrogen-deuterium scrambling, 161 Hydrogenation, 155 Hydroperoxide, in propene oxidation, 186 Hydroxyl, Brdnsted acid, 72 -, in photo-assisted processes, 257 Impregnation, 130 Indicators, 87 Ion-exchange, in preparation, 129, 132 Ionic charge, effect on structure, 15 Ionicity, determination, 28 -, Pauling's, 33 -, Phillip's, 33, 34 -, values, 30 IR, 4 Iron-molybdate, methanol oxidation, 218, 220 Iron oxide, alcohol decomposition, 152 -, anisotropy constant, 43 -, butene dehydrogenation, 201, 203, 204 -, CO hydrogenation, 239 -, lattice dilation, 41 -, magnetic hyperfine field, 42 -, magnetic properties, 39 -, Morin transition, 41, 42 -, oxygen desorption, 111 -, preparation, 126, 131 -, quadrupole splitting, 43, 46 -, reduction, 98 -, structure, 10 -, water decomposition, 273 -, water-gas shift, 246 Isoelectric point. 83 Isomerization, akenes, 140 -, carbenium ion, 142 -, effect of reduction, 141

279 -, metathesis, 141 Isosynthesis, 240 ISS, 4

Lanthanum oxide, methane coupling, 217 Lattice oxygen, in selective oxidation, 170, 177, 190, 211 -, in photo-oxidation, 263 Lattice self-potential, 29, 32 LEED, 4 Lewis acid, 72 -, formation, 75 Magnetic properties, 36, 38 Magnetic hyperfine field, 4 1 4 3 Magnesium oxide, methane coupling, 216 Maleic anhydride, from butane, 210 from butene, 200 Manganese oxide, methane coupling, 217 Metathesis, 136 -, degenerative, 137 -, isomerization, 141 mechanism, 137 productive, 137 -, substituent effect, 138 -, support effect, 140 Methane, dimerization, 215 -, oxidation, 212 Methanol, decomposition, 146 -, from carbon oxides, 228, 231 -, from methane, 212 -, oxidation, 217 -, photo-oxidation, 264 Methyl formate, from carbon oxides, 235 Methyl radical, 116. 216 Mixed oxides, acidity, 75 -, alkane reaction, 118 -, reduction, 100 Molybdenum-nickel oxide, butene oxidation, 202, 209 Molybdenum oxide, adsorbed oxygen, 116 -, alcohol adsorption, 66 -, alkane reaction, 117, 118 -, alkene isomerization, 141. 143 -, butene oxidation, 204 -, ethanol decomposition, 152

-.

-. -.

-, isoelectric point, 85 -, methane oxidation, 213 -, methanol oxidation, 218, 221 -, metathesis, 138, 139 -, photo-oxidation 260 -, preparation, 124, 131 -, propene oxidation, 189, 192. 194 -, reduction, 96 -, shear structure, 12 -, structure, 10, 22 Molybdenum-tin oxide, propene oxidation, 173 Morin transition temperature. 41 Multicomponent oxidation catalyst, 187

Ne61 temperature, 38 Nickel oxide, acetone adsorption, 66 -, reduction, 95 -, structure, 6, 24 -, support effect, 99 -, surface phonon. 48 -, water decomposition, 273 Niobium oxide, ethanol decomposition, 152 -, reduction, 98 Nitric oxide, adsorption, 67 -, reduction, 162 Nitrous oxide. decomposition, 153 NMR. 4 Nonpolar surface, 15 Nucleation model, 93 Oxidation, photo-assisted, 259 -, selective (see selective oxidation) Oxide, properties, 3 -, structure, 6 Oxide catalysts, 2 Oxygen, adsorbed (see adsorbed oxygen) -, isotope exchange, 264 --,on chromia, 58, 116 -, on iron oxide, 111 -, on zinc oxide, 114, 116 -, photo-adsorption, 256 -, photo-desorption, 256 Palladium, CO hydrogenation, 236

Paramagnetism, 38 Perovskite. structure. 12 Phonon, 48 -, and ionicity, 35 Photo-assisted processes, 256 adsorption, 256 -, desorption. 256 -, oxidation, 259, 263 -, reduction. 269 Photoelectrons, 254 Photo-holes, 254 n-allyl, 64. 178, 182 -, radicals, 186 Polar surface, 15 Preparation, 121 -, complexation, 132 -, coprecipitation, 133 -, crystallographic phase, 124 -, freeze-drying. 132 -, impregnation, 130 -, ionexchange, 129. 132 -, morphology, 125 -, pH effect, 130 -, special techniques, 130 -, spraydrying, 132 Propene. adsorption, 64 -, ammoxidation, 181 -, hydrogenation, 158 -, metathesis, 137 Propene oxidation, 173. 181 -, kinetics, 185, 186 -, mechanism, 182 -, promoters, 193 -, water effect, 194 Promoters, butene oxidation, 209 -, COXhydrogenation, 233 -, propene oxidation, 193 Propanol decomposition, 149 Pyridine. adsorption, 8 1

-.

Quadruple splitting, 43, 46 Reconstruction, surfaces, 15 Reduced oxide, effect on metal, 102 -, properties, 100 Reduction, 91

28 1 -, bronze formation, 100

-, kinetics, 93 -, mechanism, 96 -, mixed oxides, 100 -, support effect, 98 -, thermodynamics, 92 Relaxation, surfaces, 18 Rhodium, CO hydrogenation, 237 Rhodium oxide, support effect in reduction of, 99 Rocksalt structure, 6 Rutile structure, 8, 10

Scheelites, in oxidation, 189, 203 Selective oxidation, 169 -, oxygen source, 173 -, redox mechanism, 170 -, selectivity in, 175 -, types, 170 -, water effect. 174 SEM, 4 Shear structure, 12 -, in selective oxidation, 175, 179 Solid solutions, surface composition, 28 Spinel structure, 8, 10 STEM, 4 Strontium titanate. surface phonon, 49 Structure, supported oxide. 21 Structure sensitivity, alcohol decomposition, 148 Supported oxide, preparation, 129 -, structure, 21 Surface, anion vacancies, 55-57 --,Composition, 27 -, electronic structure, 46 -, polar and nonpolar, 15 reconstruction, 15 -, relaxation, 18 -, structure, 15 Surface vibration, 48

-.

TEM, 4 Thorium oxide, isosynthesis, 240 Til03, surface structure, 55 Titanium oxide, acidity, 75 --,adsorbed oxygen, 114

-, alcohol, decomposition, 152 -, -, photo-oxidation, 261, 264 -, alkane, photo-oxidation, 260 ammonia oxidation, 266 -, decarboxylation, 267, 271 -, methanol decomposition, 148 -, 0, isotope exchange, 265 -, photo-adsorption. 257 photo-desorption. 257 -, photo-oxidation, 267 -, photo-reduction, 269 -, preparation, 124, 131 -, propanol decomposition, 151 -, reduced. decoration effect, 102 _ , - , reactivity, 101 -, reduction, 96 -, structure, 10 -, surface electronic structure, 47 surface structure. 18, 55 water decomposition, 269, 272 -, water oxidation, 268 Titration, acidity, 87 TPD, 4 TPR, 4 Tungsten oxide, alkane, photo-oxidation, 260 -, alkene isomerization, 143 -, reduction, support effect, 98 -, supported, structure, 24

-.

-.

-. -.

UPS, 4 uv-vis. 4 Vacancies, in selective oxidation, 175, 189 -, surface, 55-57 Valence band, 252 Vanadates, butane oxidation, 21 1 butene oxidation, 205 Vanadia (see vanadium oxide) Vanadium oxide, isoelectric point. 85 -, methane oxidation, 215 methanol oxidation, 223 -, NO reduction, 162 .-, propanol decomposition. 151 -, propene oxidation, 195 -, reduction, support effect, 99

-.

-.

282 structure, 10 - supported, structure, 21 -

V-P-0, 206 - butane oxidation, 21 1 - butene oxidation, 202, 203 ~

propene oxidation, 195

Water, adsorption, 57 - in COXhydrogenation, 230 - in selective oxidation, 174 - photodecomposition, 269 - photo-oxidation, 269 Water-gas shift, 244 WO, (see tungsten oxide) Wurtzite structure, 8

XANES.4 XPS, 4 Zinc oxide, acetone adsorption, 66, 150 -, acidity, 75 -, adsorbed oxygen, 114, 11 6 -, alcohol, adsorption, 66 -, -, decomposition, 152 -, alkane reaction, 116 -, alkene isomerization. 144 -, butene, adsorption, 65 -, -, hydrogenation, 160 -, CO adsorption, 60, 67 -, dehydroxylation, 58 -, deuterium exchange, 161 -, ethene hydrogenation, 158 -, hydrogen adsorption, 63 -, methanol decomposition, 146 -, O2 isotope exchange, 265 --,photo-adsorption, 257 -, photo-desorption, 257 -, photo-oxidation, 266, 268 -, preparation, 125 -, propanol decomposition, 149 -, propene adsorption, 64 -, propene hydrogenation, 158 --,surface electronic smcture, 47 -, surface phonon, 48 -, surface structure, 18

zirconium oxide, isosynthesis. 240 preparation, 124

-.

283

STUDIES IN SURFACE SCIENCE AND CATALYSIS Advisory Editors: B. Delmon, Universite Catholique de Louvain, Louvain-la-Neuve, Belgium J.T. Yates, University of Pittsburgh, Pittsburgh, PA, U S A .

Volume

Volume

Volume

Volume

Volume

Volume

Volume

Volume Volume

Volume

Volume

Volume

Volume Volume

1 Preparation of Catalysts I.Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the First International Symposium, Brussels, October 1417, 1975 edited by B. Delmon, P.A. Jacobs and G. Poncelet 2 The Control of the Reactivity of Solids. A Critical Survey of the Factors that Influence the Reactivity of Solids, with Special Emphasis on the Control of the Chemical Processes in Relation to Practical Applications by V.V. Boldyrev, M. Bulens and B. Delmon 3 Preparation of Catalysts II. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Second International Symposium, Louvain-la-Neuve, September 4-7, 1978 edited by B. Delmon, P. Grange, P. Jacobs and G. Poncelet 4 Growth and Properties of Metal Clusters. Applications t o Catalysis and the Photographic Process. Proceedings of the 32nd International Meeting of the Societe de Chimie Physique, Villeurbanne, September 24-28, 1979 edited by J. Bourdon 5 Catalysis by Zeolites. Proceedings of an International Symposium, Ecully (Lyon), September 9- 1 1, 1980 edited by B. Imelik, C. Naccache, V. Ben Taarit, J.C. Vedrine, G. Coudurier and H . Praliaud 6 Catalyst Deactivation. Proceedings of an International Symposium, Antwerp, October 13-15, 1980 edited by B. Delmon and G.F. Froment 7 New Horizons in Catalysis. Proceedings of the 7th International Congress on Catalysis, Tokyo, June 30-July 4, 1980. Parts A and B edited by T. Seiyama and K. Tanabe 8 Catalysis by Supported Complexes by Vu.1. Yermakov, B.N. Kuznetsov and V.A. Zakharov 9 Physics of Solid Surfaces. Proceedings of a Symposium, Bechyhe, September 29October 3, 1980 edited by M. LazniEka 10 Adsorption at the Gas-Solid and Liquid-Solid Interface. Proceedings of an International Symposium, Aix-en-Provence, September 21-23, 198 1 edited by J. Rouquerol and K.S.W. Sing 11 Metal-Support and Metal-Additive Effects in Catalysis. Proceedings of an International Symposium, Ecully (Lyon), September 14-1 6, 1982 edited by B. Imelik, C. Naccache, G. Coudurier, H . Praliaud, P. Meriaudeau, P. Gallezot, G.A. Martin and J.C. Vedrine 12 Metal Microstructures in Zeolites. Preparation - Properties - Applications. Proceedings of a Workshop, Bremen, September 22-24, 1982 edited by P.A. Jacobs, N.I. Jaeger, P. JirG and G. Schulz-Ekloff 13 Adsorption on Metal Surfaces. An Integrated Approach edited by J. Benard 14 Vibrations at Surfaces. Proceedings of the Third International Conference, Asilomar, CA, September 1-4, 1982 edited by C.R. Brundle and H . Morawitz

284 Volume 15 Heterogeneous Catalytic Reactions Involving Molecular Oxygen by G. I. Golodets Volume 16 Preparation of Catalysts Ill. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Third International Symposium, Louvain-la-Neuve, September 6-9, 1982 edited by G. Poncelet, P. Grange and P.A. Jacobs Volume 17 Spillover of Adsorbed Species. Proceedings of an International Symposium, LyonVilleurbanne, September 12-1 6, 1983 edited by G.M. Pajonk, S.J. Teichner and J.E. Germain Volume 18 Structure and Reactivity of Modified Zeolites. Proceedings of an International Conference, Prague, July 9-1 3, 1984 edited by P.A. Jacobs, N.I. Jaeger, P. JirP, V.B. Kazansky and G. Schulz-Ekloff Volume 19 Catalysis on the Energy Scene. Proceedings of the 9th Canadian Symposium on Catalysis, Quebec, P.Q.. September 30-October 3, 1984 edited by S.Kaliaguine and A. Mahay Volume 2 0 Catalysis by Acids and Bases. Proceedings of an International Symposium, Villeurbanne (Lyon), September 25-27, 1984 edited by 8. Imelik, C. Naccache, G. Coudurier, Y. Ben Taarit and J.C. Vedrine Volume 2 1 Adsorption and Catalysis on Oxide Surfaces. Proceedings of a Symposium, Uxbridge, June 28-29, 1984 edited by M. Che and G.C. Bond Volume 22 Unsteady Processes in Catalytic Reactors by Yu.Sh. Matros Volume 23 Physics of Solid Surfaces 1984 edited by J. Koukal Volume 2 4 Zeolites: Synthesis, Structure, Technology and Application. Proceedings of an International Symposium, Portoroi-Portorose, September 3-8, 1984 edited by B. Driaj, S.HoEevar and S.Pejovnik Volume 25 Catalytic Polymerization of Olefins. Proceedings of the International Symposium on Future Aspects of Olefin Polymerization, Tokyo, July 4-6, 1985 edited by T. Keii and K. Soga Volume 26 Vibrations at Surfaces 1985. Proceedings of the Fourth International Conference, Bowness-on-Windermere, September 15-1 9, 1985 edited by D.A. King, N.V. Richardson and S.Holloway Volume 27 Catalytic Hydrogenation edited by L. Cervenq Volume 28 New Developments in Zeolite Science and Technology. Proceedings of the 7th International Zeolite Conference, Tokyo, August 17-22, 1986 edited by Y. Murakami, A. lijima and J.W. Ward Volume 29 Metal Clusters in Catalysis edited by B.C. Gates, L. Guczi and H. Knozinger Volume 3 0 Catalysis and Automotive Pollution Control. Proceedings of the First International Symposium, Brussels, September 8-1 1, 1986 edited by A. Crucq and A. Frennet Volume 3 1 Preparation of Catalysts IV. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Fourth International Symposium, Louvain-la-Neuve, September 1-4, 1986 edited by B. Delmon, P. Grange, P.A. Jacobs and G. Poncelet Volume 32 Thin Metal Films and Gas Chemisorption edited by P. Wissmann Volume 33 Synthesis of High-silica Aluminosilicate Zeolites by P.A. Jacobs and J.A. Martens Volume 3 4 Catalyst Deactivation 1987. Proceedings of the 4th International Symposium, Antwerp, September 29-October 1, 1987 edited by B. Delmon and G.F. Froment

285 Volume 35 Keynotes in Energy-Related Catalysis edited by S.Kaliaguine Volume 36 Methane Conversion. Proceedings of a Symposium on the Production of Fuels and Chemicals from Natural Gas, Auckland, April 27-30, 1987 edited by D.M. Bibby, C.D. Chaney, R.F. Howe and S.Yurchak Volume 37 Innovation in Zeolite Materials Science. Proceedings of an International Symposium, Nieuwpoort, September 13-1 7, 1987 edited by P.J. Grobet, W.J. Mortier, E.F. Vansant and G. Schulz-Ekloff Volume 38 Catalysis 1987. Proceedings of the 10th North American Meeting of the Catalysis Society, San Diego, CA. May 17-22, 1987 edited by J.W. Ward Volume 39 Characterization of Porous Solids. Proceedings of the IUPAC Symposium (COPS I), Bad Soden a. Ts., April 26-29, 1987 edited by K.K. Unger, J. Rouquerol, K.S.W. Sing and H. Kral Volume 4 0 Physics of Solid Surfaces 1987. Proceedings of the Fourth Symposium on Surface Physics, Bechyne Castle, September 7-1 1, 1987 edited by J. Koukal Volume 4 1 Heterogeneous Catalysis and Fine Chemicals. Proceedings of an International Symposium, Poitiers, March 15-17, 1988 edited by M. Guisnet, J. Barrault, C. Bouchoule, D. Duprez, C. Montassier and G. Perot Volume 42 Laboratory Studies of Heterogeneous Catalytic Processes by E.G. Christoffel, revised and edited by 2 . Paal Volume 43 Catalytic Processes under Unsteady-State Conditions by Yu. Sh. Matros Volume 4 4 Successful Design of Catalysts - Future Requirements and Development. Proceedings of the Worldwide Catalysis Seminars, July, 1988, on the Occasion of the 30th Anniversary of the Catalysis Society of Japan edited by T. lnui Volume 45 Transition Metal Oxides: Surface Chemistry and Catalysis by H.H. Kung Volume 4 6 Zeolites as Catalysts, Sorbents and Detergent Builders. Applications and Innovations. Proceedings of an International Symposium, Wurzburg, F R G , September 4-8, 1988 edited by H.G. Karge and J. Weitkamp Volume 47 Photochemistry on Solid Surfaces edited by M. Anpo and T. Matsuura Volume 48 Structure and Reactivity of Surfaces. Proceedings of a European Conference Trieste. Italy, September 13- 16, 1988 edited by C. Morterra, A. Zecchina and G. Costa

This Page Intentionally Left Blank