Studies in Surface Science and Catalysis 64
NEW TRENDS I N CO ACTIVATION
This Page Intentionally Left Blank
Studie...
24 downloads
1864 Views
22MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Studies in Surface Science and Catalysis 64
NEW TRENDS I N CO ACTIVATION
This Page Intentionally Left Blank
Studies in Surface Science and Catalysis Advisory Editors: B. Delrnon and J.T. Yates Vol. 64
NEW TRENDS IN CO ACTIVATION Editor
L. GUCZl Institute ofIsotopes of the Hungarian Academy of Sciences, Budapest, Hungary
ELSEVIER
Amsterdam - Oxford - New York - Tokyo
1991
ELSEVIER SCIENCE PUBLISHERS B.V. Sara Burgerhartstraat 25 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.
Library o f Congress Cataloging-in-Publication Data
New t r e n d s in CO a c t i v a t i o n p.
/
e d i t o r . Laszlo G u c z i .
cm. -- ( S t u d i e s in s u r f a c e S c i e n c e a n d c a t a l y s i s
.
vol. 64) Includes bibliographical references. I S B N 0-444-88238-3 1 . C a r b o n m o n o x i d e . 2. F i s c h e r - T r o p s c h p r o c e s s . 3. C a t a l y s i s . . 1 1 . S e r i e s S t u d i e s in s u r f a c e s c i e n c e a n d I. G u c z i , L . . 1932catalysis 64. PD181.ClN48 1991 661'.81--dC20 91-8267 CIP
.
ISBN 0-444-88238-3
0Elsevier Science Publishers B.V., 1991 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 regulations for 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 t o the publisher. No responsibility is assumed by the Publisher for any injury and/or damage t o 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 t o conform t o 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
Preface Carbon monoxide activation has been an important subject for the last two decades. Since being discovered approximately half a century ago, Fischer -Tropsch synthesis has been commercialized in many countries having a gasoline shortage. Later, after World War II, the process died out due to the discovery of rich, new oil fields and to the introduction of economic reforming processes. The renaissance in carbon monoxide chemistry started right after the world’s first oil boom and it has continued to expand ever since. Due to the appearance of new, highly selective catalysts, scientists were able to use the possibilities posed by CO conversion in many different fields so that there is now a wide range of applications available to produce useful chemicals by using CO as a feedstock. Research in this field is, of course, dependent on the current fluctuation in the price of crude oil but according to long range forecasts we can expect that CO woill substitute crude oil as a feedstock for petrochemicals and that a further upswing in its chemical application can be foreseen. The underlying idea behind this book was to try to furnish scientists with a comprehensive summary of new research areas in the activation of carbon monoxide, as one of the most reactive molecules, and in its applications. In order to understand the variety of the reactivity of CO, a quantum-chemical approach helps the reader to understand the binding state of CO. to the solid surface (Chapter I). The structure of the adsorbed CO can be better understood by examining its reactivity towards single crystals in the absence and in the presence of promoters (Chapter 2). The first approach in the reactivity study is that of studying catalytic activity of single crystals and structure sensitivity which are summarized in Chapter 3. One of the most prominent effects in the CO activation process is ascribed to the presence of additives, promoters which, in a real catalyst system, are far more complicated than on single crystal surfaces (Chapter 4). The original FT process applied fused iron or cobalt catalysts which were suitable for producing mainly straight chain hydrocarbons. The two most important processes involving CO activation, the new FT process and alcohol formation are discussed in Chapters 5 and 7. The great variety of catalyst systems, as well as their parameters (support, metal dispersion, promoters, etc.), utilized in these two processes also ensured a great variety of products. An important type of catalyst, the bimetallic catalysts, are discussed in a separate chapter (Chapter 6). The role of hydrogen as one of the frequently used partners in CO activation is discussed in Chapter 8. The field of production of specialty chemicals is an excellent example of the homogeneous catalytic activation of CO (Chapter 9). In Chapter 10 there is an overview of the industrial applications of CO chemistry and these are illustrated by working processes. The final chapter gives the reader some hints about future progress in the field. We believe that thiz book will provide the reader with the present state-of-the-art survey in CO chemistry which will become even more important in the future. The Editor is very grateful to Dr. Heinz Heineman for his valuable advice during the preparation of the book and to Dr. Gibor Kisfaludi for his help in finalizing the manuscript. L. Guczi
v1
Table of Content CHAPTER 1 (Rutger A . van Santen and Ad de Koster) QUANTUM CHEMISTRY OF CO CHEMISORPTION AND ACTIVATION . . . . . . . . . 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 The Coordination of CO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Crystal Face Dependence -Promoter Effects . . . . . . . . . . . . . . . . . . . . 1.4 CO Dissociation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5 Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. .1 2 4 16 26 31 33
CHAPTER 2 (Maya Kiskinova) INTERACTION OF CO WITH SINGLE CRYSTAL METAL SURFACES . . . . . . . . . . . 37 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 2.2 Experimental Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.2.1 Dynamical Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 2.2.1.1 Thermal Desorption (TD) . . . . . . . . . . . . . . . . . . . . . . 39 2.2.1.2 Molecular Beam Technique . . . . . . . . . . . . . . . . . . . . 39 2.2.1.3 Electron and Photon Stimulated Desorption (ESD and PSD) and Electron Stimulated Desorption Ionangular Distribution (ESDIAD) . . . . . . . . . . . . . . . . . . . . . . . 40 2.2.1.4 Secondary Ion Mass Spectrometry (SIMS) . . . . . . . . . . . .40 40 2.2.2 Static Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2.1 Emission Spectroscopies . . . . . . . . . . . . . . . . . . . . . . 40 41 2.2.2.2 Absorption Spectroscopies . . . . . . . . . . . . . . . . . . . . . 2.2.2.3 Low Energy Electron Diffraction (LEED) . . . . . . . . . . . . 42 2.3 Electronic Structure of the Free CO Molecule and Molecular Orbital Model for CO Bonding to Metal Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.3.1 Free CO Molecule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.3.2 Molecular Orbital Model for CO - Metal Surface Bonding . . . . . . . . 43 2.4 Molecular CO Adsorption on Clean Single Crystal Metal Surfaces . . . . . . . . 44 2.4.1 CO Adsorption Probability and the Mechanism of CO Adsorption . . . . 44 2.4.2 CO Adsorption Binding Energy and Mobility of the CO Molecule in the Adsorption Phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 2.4.3 Surface Structure of the CO Overlayers . . . . . . . . . . . . . . . . . . . . 53 2.4.4 Orientation of the Chemisorbed CO Molecule with Respect to the Substrate Surface and the Corresponding Surface - C and C - 0 Interaction Distances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 2.4.5 CO Induced Work Function Changes and the Effective Charge Transfer during Formation of the Surface - CO Bond . . . . . . . . . . . 59 2.4.6 Influence of the Metal - CO Bonding on the CO Electron Core and Valence Level Binding Energies . . . . . . . . . . . . . . . . . . . . . . . . 62
vii Influence of the Metal - CO Bonding on the Energy Position of the Unfilled 2p CO Affinity Levels. Bonding and Antibonding States as a Result of the Metal/2p CO Coupling . . . . . . . . . . . . . . . . . . . . 2.4.8 Influence of the Metal - CO Bonding on the Vibrational Properties of the CO Molecule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.9 Dissociation Probability of CO Chemisorbed on Metal Surfaces . . . . 2.4.10 CO Induced Perturbations in the Surface Electronic and Geometric Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Effect of Some Additives on the CO - Transition Metal Surface Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5.1 Electronegative Additives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5.2 ElectropositiveAdditives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6 The Relevance of the CO/Single Crystal Metal Studies to the CO - Metal Catalyst Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.7
64 66
. 70 73 75 75 77 78 79
CHAPTER 3 (Jos A . Rodriguez and D . Wayne Goodman) CATALYTIC ACTIVATION OF CO OVER SINGLE CRYSTALS . . . . . . . . . . . . . . . . 87 88 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Methanation of CO on Single Crystal Surfaces . . . . . . . . . . . . . . . . . . . . 88 3.2.1 CO Methanation on Clean Metal Single Crystals . . . . . . . . . . . . . . 89 89 3.2.1.1 Monometallic Surfaces . . . . . . . . . . . . . . . . . . . . . . . 95 3.2.1.2 Bimetallic Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2 CO Methanation Chemically Modified Surfaces . . . . . . . . . . . . . . 9 8 99 3.2.2. I Electronegative Impurities . . . . . . . . . . . . . . . . . . . . . 103 3.2.2.2 Electropositive Impurities . . . . . . . . . . . . . . . . . . . . . 104 3.2.2.3 Related Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3 Metal - Support Interactions and CO Methanation . . . . . . . . . . . . 105 3.3 Water Gas Shift Reaction on Single Crystal Surfaces . . . . . . . . . . . . . . . 106 3.3.1 Kinetics over Cu( 110) and Cu( 111) Catalysts . . . . . . . . . . . . . . . 107 3.3.2 Sulfur Poisoning of Cu(ll1) Catalysts . . . . . . . . . . . . . . . . . . . . 109 3.3.3 Cesium Promotion of Cu( 110) and Cu(111) Catalysts . . . . . . . . . . .110 3.4 Methanol Synthesis on Single Crystal Surfaces . . . . . . . . . . . . . . . . . . . 111 3.4.1 KineticsonPd(ll0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 3.4.2 Studies on Cu(lll), ZnOx/Cu(lll) and Cu/Zn0(0001) Surfaces . . . . 112 3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 113 3.6 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 CHAPTER 4 (Vladirnir Ponec) SELECTIVITY IN THE SYNGAS REACTIONS: THE ROLE OF SUPPORTS AND PROMOTERS IN THE ACTIVATION OF CO AND IN THE STABILIZATION OF INTERMEDIATES . . . . . . . . . . . . 117 4.1 Selectivity Pattern in Syngas Reactions . . . . . . . . . . . . . . . . . . . . . . . . 118 4.1.1 Some Facts on Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 4.1.1.1 Metals for Synthesis of Hydrocarbons . . . . . . . . . . . . . . 118 4.1.1.2 Activity Pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 119 4.1.1.3 Selectivity Pattern . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2 Supports and Promoters . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
...
Vlll
Support Functions as a Promoter . . . . . . . . . . . . . . . . . 121 A Closer Inspection of the Promoter Function . . . . . . . . . 125 4.1.2.2.1 Transition Metals Pd. Pt, Ir and Co, Ru. Rh in CH30H Synthesis . . . . . . . . . . . . . . . . . . 125 4.1.2.2.2 Copper Catalysts . . . . . . . . . . . . . . . . . . . 127 4.1.2.2.3 Promoted Transition Metals as Catalysts for Cz+-oxygenates. . . . . . . . . . . . . . . . . . . . 128 130 Adsorption of CO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Adsorption of CO on Metals . . . . . . . . . . . . . . . . . . . . . . . . . . 130 4.2.2 Adsorption of CO on Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . 132 134 4.2.3 Alloy Based Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.3.1 Group VIII-Ib Metals . . . . . . . . . . . . . . . . . . . . . . 134 4.2.3.2 Group VIII -Group VIII Metals . . . . . . . . . . . . . . . . 135 4.2.3.3 VIII Group Metals -Early Transition Metals . . . . . . . . . 135 4.2.3.4 VIII Group Metals -Rare Earths . . . . . . . . . . . . . . . . 136 4.2.3.5 Copper -Early Transition Metals, Copper 136 Rare Earths . . . . . . . . . . . . . . . . . . . . . . . . . . . . Promotion of Metals. Theories and their Verification . . . . . . . . . . . . . . . 137 4.3.1 Some Physical Phenomena Relevant to the Promoter -Metal 137 Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1.1 Point Charge and Dipole Metal Interaction; 137 Image Forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1.2 Adsorption of Strongly Electrodonating Species . . . . . . . .138 4.3.1.3 A Through-the-MetalInteraction of Coadsorbed 139 Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1.4 Charge Transfer Between Phases. Metal - Metal and Metal - SemiconductorInteraction . . . . . . . . . . . . . . .140 4.3.2 Modern theories of Promotion Effects in the Syngas Reaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 145 4.3.3 Theories and their Verification . . . . . . . . . . . . . . . . . . . . . . . . 147 Related Reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 4.4.1 Water Gas Shift Reaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4.2 Hydrogenation of Unsaturated Aldehydes . . . . . . . . . . . . . . . . . . 148 Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 148 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.2.1 4.1.2.2
4.2
4.3
4.4
4.5 4.6
CHAPTER 5 (CalvinH .Bartholomew) RECENT DEVELOPMENTS IN FISCHER-TROPSCH CATALYSIS . . . . . . . . . . . . . 158 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 5.2 New Catalyst Developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 160 5.2.1 Chemical Modifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2.1.1 Additives and Promoters . . . . . . . . . . . . . . . . . . . . . . 163 5.2.1.2 Effects of Support, Metal Loading and Dispersion . . . . . . 169 5.2.1.3 Interstitial Compounds . . . . . . . . . . . . . . . . . . . . . . . 179 5.2.1.4 Bimetallics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 5.2.1.5 Effects of Pretreatment and Preparation . . . . . . . . . . . . . 186 5.2.1.5.1 General Developmentsin F T Catalyst Preparation/ Pretreatment . . . . . . . . . . . . . . 187 5.2.1.5.1.1 Preparation . . . . . . . . . . . . . . . 187 5.2.1.5.1.2 Pretreatment . . . . . . . . . . . . . . 190
ix Limitations of Chain Growth by Shape Selectivity . . . . . . . . . . . . .193 Interception of Intermediates . . . . . . . . . . . . . . . . . . . . . . . . . 194 5.2.3.1 Interception in Multifunctional Catalysts . . . . . . . . . . . . 194 5.2.3.2 Selectivity/StructureRelationships and Design Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 5.2.3.3 Recent Developments in Zeolite Catalyst Technology . . . . 198 5.2.3.4 Interception in Two-step Processes . . . . . . . . . . . . . . .199 5.3 New Developments in Reactor and Process Design . . . . . . . . . . . . . . . . . 199 5.3.1 Recent Developmentsin Reactor Design . . . . . . . . . . . . . . . . . . 199 5.3.1.1 Reactor Types and their Characteristics . . . . . . . . . . . . .199 5.3.1.2 Comparison of Attributes for Three Reactor Types . . . . . 202 5.3.1.3 Recent Experimental and Modeling Studies of FT Reactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 5.3.2 Second Generation FT Processes . . . . . . . . . . . . . . . . . . . . . . . 204 5.3.2.1 Second Generation Commercial Processes . . . . . . . . . . .204 5.3.2.2 Experimental/Conceptual Multi-Stage Processes . . . . . . . 206 5.4 Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . 208 5.4.1 Assessment of Current Technology and Conclusions . . . . . . . . . . . 208 5.4.2 Recommendations for Future Research and Development . . . . . . . . 211 5.5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 5.2.2 5.2.3
CHAPTER 6 (Johanna Schwank) BIMETALLIC CATALYSTS FOR CO ACTIVATION . . . . . . . . . . . . . . . . . . . . . . . 225 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226 6.2 The Genesis and Nature of Surface Sites in Bimetallic Catalysts . . . . . . . . . 227 6.3 The Effect of Catalyst Preparation on the Properties of the Support and the Consequences for Bimetallic Particle Formation . . . . . . . . . . . . . . . . . . 236 6.4 The Interaction of CO with Bimetallic Surfaces . . . . . . . . . . . . . . . . . . 242 6.5 The Interaction of Hydrogen with Bimetallic Surfaces . . . . . . . . . . . . . . . 244 6.6 CO Activation over Bimetallic Catalysts . . . . . . . . . . . . . . . . . . . . . . . 246 6.7 Effect of Second Metal Component on Catalyst Deactivation . . . . . . . 251 6.8 New Reaction Pathways in CO Activation . . . . . . . . . . . . . . . . . . . . . . 253 6.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255 CHAPTER 7 (Richard G . Herman) CLASSICAL AND NON-CLASSICAL ROUTE FOR ALCOHOL SYNTHESIS . . . . . . . 265 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266 7.2 Methanol Synthesis Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268 7.2.1 Active State of Copper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 7.2.2 Hydrogenation of CO vs COz . . . . . . . . . . . . . . . . . . . . . . . . . 272 7.2.3 Newer Methanol Synthesis Catalysts . . . . . . . . . . . . . . . . . . . . 274 7.2.3.1 Cs/Cu/ZnO Catalysts . . . . . . . . . . . . . . . . . . . . . . . . 274 7.2.3.2 ThfCu Alloy Catalysts . . . . . . . . . . . . . . . . . . . . . . . 275 7.2.3.3 Zr/Cu Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . 276 7.2.3.4 Ce/Cu Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . 277 279 7.2.3.5 Supported Pd Catalysts . . . . . . . . . . . . . . . . . . . . . . 7.2.3.6 NaH -RONa- M(0Ac)z Catalysts Suspended in Liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280 7.2.3.7 Homogeneous Methanol Synthesis Catalyst . . . . . . . . . . 280 7.2.4 Newer Methanol Synthesis Technology . . . . . . . . . . . . . . . . . . .281
X
7.2.4.1
7.3
7.4
7.5
7.6
New Reactor Systems for Gas Phase Methanol Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281 7.2.4.2 Liquid Phase Methanol Synthesis . . . . . . . . . . . . . . . . 283 7.2.5 Deactivation of Methanol Synthesis Catalysts . . . . . . . . . . . . . . . 285 Higher Alcohol Synthesis Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . 289 7.3.1 Rationale for Higher Alcohols as Fuels . . . . . . . . . . . . . . . . . . .289 7.3.2 Background of Higher Alcohol Synthesis Catalysts . . . . . . . . . . . .290 7.3.3 Historical Development of Higher Alcohol Synthesis . . . . . . . . . . .291 7.3.4 Newer Catalysts and Technology . . . . . . . . . . . . . . . . . . . . . . . 293 7.3.5 Higher Alcohol Synthesis 293 over Oxide-Based Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3.5.1 Alkali-Promoted Cu/ZnO Catalysts . . . . . . . . . . . . . . .293 7.3.5.2 Supported Alkali-Promoted Cu/ZnO/M203Catalysts . . . . 296 7.3.5.3 Other Oxide Catalysts Containing Transition Metal Additives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300 7.3.6 Alcohol Synthesis over Alkali-Promoted MoS2 Catalysts . . . . . . . . 303 7.3.6.1 Effect of Alkali Doping of MoS2 on the Activity and Selectivity for Alcohol Synthesis . . . . . . . . .305 7.3.6.2 Effect of Cesium Concentration on the Activity and Selectivity of Alcohol Synthesis . . . . . . . . . . . . . . 308 7.3.6.3 Effect of Reaction Temperature and Pressure on the Selectivity to Alcohols at Different Cs Loadings . . . . . . . 309 7.3.6.4 Effect of Reactant Contact Time . . . . . . . . . . . . . . . . . 310 7.3.6.5 Effect of C02, H2S, and Olefins in the Synthesis Gas . . . . 310 7.3.6.6 Effect of Adding Cobalt to the Alkali-Doped MoS2 Catalyst . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314 7.3.7 Mechanistic Implications of the Promotional Effect of Alkali . . . . . . 31.5 7.3.8 Research Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316 Mechanisms of Alcohol Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 317 7.4.1 Mechanistic Background of Higher Alcohol Synthesis over Oxide 317 Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.4.2 Formation of C2 Products over Cs/Cu/ZnO Catalysts . . . . . . . . . . . 319 7.4.3 Formation of C, and C, Alcohols over Cs/Cu/ZnO Catalysts . . . . . . 321 7.4.4 Formation of Oxygenates and Hydrocarbons over AlkalUMoS2 325 Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328 7.4.5 Mechanistic Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 Kinetic Models for the Synthesis of Alcohols . . . . . . . . . . . . . . . . . . . . 329 7.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.5.2 Development of Kinetic Models for Higher Alcohol Synthesis . . . . . 330 7.5.3 Kinetic Modelling of Alcohol Synthesis over Cs/Cu/ZnO Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335 7.5.4 Kinetic Modelling of Alcohol Synthesis over Alkali/MoS2-Based 336 Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 7.5.5 Kinetic Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CHAPTER 8 (Ldszld Guczi) EFFECT OF HYDROGEN IN CONTROLLING CO HYDROGENATION . . . . . . . . . . . 350 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 8.2 Hydrogen Adsorption On Metal Surface . . . . . . . . . . . . . . . . . . . . . . . 352
xi Quantumchemical Approach of the Hydrogen Bonding . . . . . . . . . . 352 Kinetics and Energetics of Hydrogen Adsorption on Metals . . . . . . . 355 8.2.2.1 Kinetics of Hydrogen Adsorption . . . . . . . . . . . . . . . . 355 8.2.2.2 Extent and Stoichiometry of Hydrogen Adsorption . . . . . . 358 8.2.2.3 Weak and Strong Chemisorption of Hydrogen . . . . . . . . . 359 Temperature Programmed Desorption of Hydrogen . . . . . . . . . . . . . . . . 362 362 8.3.1 Desorption of Hydrogen from Metals . . . . . . . . . . . . . . . . . . . . 8.3.2 Basic Knowledge about Temperature Programmed Desorption of Hydrogen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362 Effect of Hydrogen Bonding on the Selectivity in CO Hydrogenation . . . . . . 367 367 8.4.1 Hydrocarbon and Olefin Formation . . . . . . . . . . . . . . . . . . . . . 8.4.2 Hydrogen Effect in Alcohol Formation . . . . . . . . . . . . . . . . . . . 371 8.4.3 Effect of Promoters on the Activated Hydrogen . . . . . . . . . . . . . . 372 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376 8.2.1 8.2.2
8.3
8.4
8.5 8.6
CHAPTER 9 (Michael Riiper) CO ACTIVATION BY HOMOGENEOUS CATALYSTS . . . . . . . . . . . . . . . . . . . . . 381 382 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384 9.2 Mechanistic Implicationsof CO Activation . . . . . . . . . . . . . . . . . . . . . 9.2.1 Coordination of CO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384 385 9.2.2 Activation of the Reagent . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2.3 Conversion of Coordinated CO . . . . . . . . . . . . . . . . . . . . . . . . 387 9.2.4 Product Elimination and Catalyst Regeneration . . . . . . . . . . . . . . 388 9.3 Homogeneous Hydrogenation of CO . . . . . . . . . . . . . . . . . . . . . . . . . 389 9.3.1 Cobalt Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392 9.3.2 Rhodium Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393 9.3.3 Ruthenium Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395 9.4 Homogeneous Oxidation of CO . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396 9.5 Functionalizing Reactions of CO . . . . . . . . . . . . . . . . . . . . . . . . . . . 398 399 9.5.1 Carbonylation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5.1.1 Carbonylation of Alkynes . . . . . . . . . . . . . . . . . . . . . 400 402 9.5.1.2 Carbonylationof Alkenes . . . . . . . . . . . . . . . . . . . . . 9.5.1.3 Carbonylation of Alkadienes . . . . . . . . . . . . . . . . . . . 404 9.5.1.4 Carbonylation of Alkanes . . . . . . . . . . . . . . . . . . . . . 407 9.5.1.5 Carbonylation of Alkanols, Esters and Ethers . . . . . . . . . 407 9.5.1.6 Carbonylation of Organic Halides . . . . . . . . . . . . . . . . 411 9.5.2 Hydrocarbonylation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413 9.5.2.1 Hydrocarbonylation of Alkenes . . . . . . . . . . . . . . . . . 413 9.5.3.2 Hydrocarbonylationof Alkanols . . . . . . . . . . . . . . . . . 420 9.5.3.3 Hydrocarbonylation of Alkanals . . . . . . . . . . . . . . . . . 422 9.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423 9.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424 CHAPTER 10 (Helmut Papp and Manfred Baerns) INDUSTRIAL APPLICATION OF CO CHEMISTRY FOR THE PRODUCTION OF SPECIALTY CHEMICALS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430 431 10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.2 Carbonylation of Methanol and Related Processes . . . . . . . . . . . . . . . . . 431 10.2.1 Synthesis of Acetic Acid by Carbonylation of Methanol . . . . . . . . . 432
xii
10.3
10.4
10.5 10.6
10.2.1.1 Catalytic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 433 10.2.1.2 Industrial Importance of Acetic Acid . . . . . . . . . . . . . .434 10.2.2 Synthesis of Acetic Anhydride by Carbonylation of 435 Methylacetate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.2.2.1 Catalytic Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 435 10.2.2.2 Industrial Importance of Acetic Anhydride . . . . . . . . . . 436 10.2.3 Synthesis of Acetaldehyde and Ethanol . . . . . . . . . . . . . . . . . . . 438 10.2.4 Synthesis of Vinyl Acetate . . . . . . . . . . . . . . . . . . . . . . . . . . 440 10.2.5 Homologation of Carboxylic Acids and Esters . . . . . . . . . . . . . . 441 . 10.2.6 Oxidative Carbonylation of Alcohols and Production of Ethylene Glycol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442 Hydroformylation of Olefins (0x0 Process) . . . . . . . . . . . . . . . . . . . . . 442 10.3.1 Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443 10.3.2 Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445 10.3.3 Commercial Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 447 Reppe Carbonylation and Related Processes . . . . . . . . . . . . . . . . . . . . . 449 450 10.4.1 Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4.2 Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451 452 10.4.3 Commercial Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454 10.4.4 Carbonylation of Organic Halides . . . . . . . . . . . . . . . . . . . . . . Koch Reaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455 .457 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CHAPTER 11 (Gdbor A . Somorjai)
THE CATALYZED HYDROGENATION OF CARBON MONOXIDE: AN OVERVIEW AND FUTURE DIRECTIONS . . . . . . . . . . . . . . . . . . . . . . 462 11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463 11.2 The Chemisorption and Dissociation of Carbon Monoxide on Clean .464 Transition Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2.1 Alkali Metal Induced CO Bond Weakening and Dissociation . . . . . . 464 1 1.2.2 CO Dissociation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465 11.3 The Kinetics of the C0/H2 Reaction . . . . . . . . . . . . . . . . . . . . . . . . . 466 466 11.3.1 Evidence for Secondary Reactions . . . . . . . . . . . . . . . . . . . . . . 11.4 Promotion by the Oxide -Metal Interface . . . . . . . . . . . . . . . . . . . . . 468 11.5 Control of Secondary Reactions during CO Hydrogenation by Contact Time. Bimetallics and Zeolites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468 11.6 Future Directions of Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468 11.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .469
INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
471
...
Xlll
List of Contributors Manfred Baerns, Ruhr-University Bochum, P.O.B. 10 21 48, D-4630 Bochum, Gemany Calvin H . Bartholomew Department of Chemical Engineering, Brigham Young University, Provo, Utah 84602, USA D . Wayne Goodman Department of Chemistry Texas A&M University College Station, TX 77843-3255, USA Ldszlo' Guczi Surface Science and Catalysis Laboratory Institute of Isotopes of the Hungarian Academy of Sciences, H-1525 Budapest, P.O. Box 77, Hungary Richard G . Herman Zettlemoyer Center for Surface Studies, Sinclair Laboratory, NO. 7 Lehigh University, Bethlehem, Pennsylvania 18015 USA Maya Kiskinova Institute of General and Inorganic Chemistry, Bulgarian Academy of Sciences, Sofia 1040, Bulgaria Ad de Koster Laboratory for Inorganic Chemistry and Catalysis, Eindhoven University of Technology, P.O. Box 513,5600 MB Eindhoven, The Netherlands Helmut Papp Ruhr-University Bochum, P.O.B. 10 21 48, D-4630 Bochum, Germany Vladimir Ponec Gorlaeus Laboratorium, Leiden University, P. 0. Box 9502, 2300 RA Leiden, The Netherlands Jos A . Rodriguez Department of Chemistry Texas A&M University College Station, TX 77843-3255, USA Michuel Riiper BASF-AG, D-6700 Ludwigshafen, F.R. Germany Johannes Schwank Department of Chemical Engineering, The University of Michigan, H. H. Dow Building, Ann Arbor, MI 48109, USA Rutger A. van Santen Laboratory for Inorganic Chemistry and Catalysis, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands Gahor A. Somorjai Department of Chemistry and Center for Advanced Materials, Lawrence Berkeley Laboratory University of California, Berkeley, California 94720, USA
This Page Intentionally Left Blank
1
CHAPTER 1
QUANTUM CHEMISTRY OF CO CHEMISORPTION AND ACTIVATION R.A. van Santen and A. de Koster
Laboratory for Inorganic Chemistry and Catalysis, Eindhoven Universitv of Technolow. ", P.O. Box 513. 5600 M B Eindhoven. (The Netherlands)
2
1.1 INTRODUCTION Carbon monoxide chemistry is not only of significant interest to chemical practice, but also to fundamental catalysis. It is also a key chemical to the petrochemical process industry. Important chemicals as methanol and hydrocarbons can be produced from synthesis gas. Also many other oxygen containing hydrocarbons can be produced from methanol. Using the water gas shift reaction hydrogen can be produced from CO and water. The total oxidation of CO is an example of a reaction important to environmental considerations. Surface science and fundamental catalysis have chosen the study of CO chemisorption and reactivity as one of its key areas of research. As a result the interaction of CO with metal surfaces and bonding in metal complexes is currently reasonable well understood. Semi empirical (ref. 1) as well as first principle quantum chemical (ref. 2) theoretical methods have been used to establish the electronic features that govern the interaction. The result is a consistent picture of the electronic structure of chemisorbed CO. It appears that the Frontier Orbital method (refs. 3,4) describing the attractive part of the chemical bond in terms of the interaction between Highest Occupied Molecular Orbitals and Lowest Unoccupied Molecular Orbitals is a very useful approximate technique. Applied to a metal surface it relates the attractive part of the chemical bond to the local densities of state at the Fermi level of the surface orbital fragments (ref. 5 ) interacting with the Frontier Orbitals of the adsorbing molecule. The repulsive part of the chemisorptive bond is due to the interaction between doubly occupied orbitals of the interacting fragments. Two fragment orbitals form a bonding and an antibonding orbital. If both remain occupied repulsion results. These repulsive interactions are proportional to the number of neighbors (ref. 6) and hence significantly affect the geometry of adsorption sites. There are several es that enable reduction of these Pauli repulsion forces. These effects are very important to catalysis, since reduction of the repulsive interaction results in a lowering of activation energies (ref. 7). Since the antibonding orbital formed upon interaction between the fragments usually is antisymmetric with respect to the center of the bond, interaction with unoccupied metal orbitals of the same symmetry lowers the activation for the recombination or insertion. This has been elucidated very elegantly by Hoffmann et a1 (ref. 8) in a study of CO insertion in the metal-methyl bond in the coordination shell of organometallic complexes. For instance the interaction between the doubly occupied methyl o-orbital and the doubly occupied 5 0 (HOMO) orbital of CO is reduced because the antibonding orbital resulting from the two o-orbitals is stabilized by the interaction with an empty d-orbital. This may explain why insertion reactions into the metal CO bond require coordination to metal atoms with partially empty d atomic orbitals. It is the reason why only ions of the group VIII metals with a nearly filled d-valence band catalyse methanol formation. Complete release of repulsion may occur, if the energy of the occupied antibonding orbitals becomes higher than unoccupied unchanged orbitals in the metal complex or at the metal surface (refs. 9,lb), due to electron transfer from the occupied antibonding orbital to the unoccupied metal orbital. We will see that this may happen at metal surfaces. Antibonding orbital combinations may become pushed higher than the Fermi level, which will change repulsive interactions into attractive forces.
3
Whereas the interaction with the doubly occupied CO So-orbital may become repulsive or attractive, the interaction with the unoccupied CO 2n*-orbital (LUMO) is always attractive. Admixture of this orbital with interacting fragments leads to partial elecuonoccupation of this orbital. Since the 2n*-orbital is antibonding with respect to the CO bond, population of the 2n*-orbital tends to weaken the CO bond. This may result in a weakening of the CO stretch frequency, observable by Infrared Spectroscopy or Electron Energy Loss Spectroscopy. Since the electron population of the CO 2n* orbital is also approximately proportional to the number of atom neighbors (ref. lo), measurement of the CO stretching frequency provides information on the adsorption site geometry. The fate of the unoccupied 2n* CO orbital is also crucial to CO dissociation. CO is usually coordinated perpendicular to the metal surface, with the carbon atom directed towards the metal atom. The carbon on C has a smaller energy difference with respect to the metal Fermi level than the oxygen 40-orbital. This results in the stronger interaction with the So-orbital localized on carbon. Also the coefficient of the 2n*-orbital is larger on carbon than on oxygen, again favouring interaction with the carbon atom. The perpendicular mode of bonding is also due to the requirements to minimize repulsive interactions, in particular with the doubly occupied CO In orbital. On transition metals with low valence &electron band occupation, e.g. Fe, side on bonding of CO has been observed. Anderson (ref. 11) explained this as being due to a release of the repulsive interaction of the doubly occupied In-orbital with the metal surface orbitals, because the low d-valence elctron occupation results in a partial filling of antibonding metal surface orbital fragment resulting in an attractive interaction. As we will see CO bending, the first step towards dissociation, usually costs energy. This becomes partially compensated by an increased interaction with the 2n*-orbital, because now also oxygen will interact with the metal surface. The activation energy for dissociation rises steeply if the CO bond is stretched. The activation energy is determined by the 2n* occupation as well as the stzbilizing interaction of the 0 and C atoms with the metal surface. Shustorovich (ref. 12) has developed an elegant scheme, based on Bond Order Conservation, that explicitly relates the reaction energy for dissociation to the bond strength of the isolated adsorbed carbon and oxygen atoms. The method does not allow a prediction of the actual dissociation path to be followed. For such a prediction detailed quantum chemical approaches are required, as will be discussed below. The concepts described above will be illustrated by discussing in detail the electronic factors that determine the relative stability of CO chemisorbed atop or in higher coordination sites. Emphasis will be on the relation with valence d electron occupation and work function of transition metal surface. It will appear that the interaction with 50- and 2n*-orbitals follow different trends. The tendency to adsorb atop shifts to the right side of the group VIII metals on the periodic system and decreases from the third row to the fifth row of the periodic system. Next the relative bond strength of CO chemisorbed to different metal surfaces will be discussed. Whereas the interaction of the 5 0 - as well as 2n*-orbitals of CO with the metal surface is of great relevance to the coordination of CO, the difference in bond strength of CO chemisorbed
4
to different surfaces of the same metal mainly depends on differences in the interaction with the 50-orbital. In a subsequent section details of CO dissociation reaction paths will be presented. The results will be mainly based on calculations based on the use of Anderson's ASED method (ref. 13). Finally possible implications of the results presented to catalysis will be outlined.
1.2 THE COORDINATION OF CO We will discuss chemical bonding of CO based on results of Extended Huckel calculations. Such calculations have the disadvantage of being parameter dependent, because parameter choice determines the accuracy of the results. The parameter dependence may however also be
I EF
,"tep, -30
, ,
I -20
,
h._
Energy
?.
,
-10
(ev)
I
, ,
I 0
?tep, -30
, , I ,
k,
-20
Energy
, , ,
-10
I
0
feV)
Figs. l a and l b LDOS of the I n (a), 50 (b) CO molecular orbitals in the gas phase, 1-fold adsorbed on Rh( 11l), 3-fold adsorbed on Rh(ll1) and 3-fold adsorbed on stepped Rh( 111) (distance of CO to step: 1.551 A.) The Fermi level is indicated by EF.
5
advantageous because it enables to study properties by varying a single parameter. More important is the absence of explicit electrostatic interactions in the Extended Huckel method. Elsewhere (ref. 6) we have presented a detailed comparison of results of Extended Huckel, ab initio Hartree-Fock and Configuration Interaction Methods as well as the Hartree-Fock-Slater X,
P?(El
3-fold
-30
-20
Energy
-10
0
(ev)
Figs. l c LDOS of the 2n* CO molecular orbitals in the gas phase, 1-fold adsorbed on Rh(l1 l), 3-fold adsorbed on Rh( 111) and 3-fold adsorbed on stepped Rh( 111) (distance of CO to step: 1.551 A. The Fermi level is indicated by E p
method. Whereas quantitatively differences are found, the picture of chemical bonding that arises is essentially similar.
6
In Fig. 1.1 the local electron energy density of states pi(,?) of the In, 5 0 and 2n* states of CO are presented, for CO chemisorbed atop and three-fold (hcp site) to a 29 atom cluster of Rh simulating the (111) surface. The cluster models used are shown in Fig. 1.2. They have been chosen such that the nearest neighbor metal atoms of the chemisorbed CO will have the same metal atom nearest neighbor environment as on the transition metal surface. Compared to their gas phase positions the maxima of the I n and %-orbitals have shifted downward, but the second high energy maximum of the 2n*-orbital has shifted upwards. The CO In- and So-orbitals were doubly occupied before interacting with the metal surface, the CO 2a*-orbital was empty. As can be seen from the gross population table (Table l.la), the la-orbital remains doubly occupied, but the 50 has lost electrons, whereas the 2n*-orbital experiences an electron gain. The effects are larger in the bridge coordination site than in the atop position. These results correspond well to familiar notions of 50 to metal electron donation and metal to 2~c*-electronbackdonation usually applied to describe the chemisorptive bond of CO (ref. 14). The broadening of adsorbate orbital levels stems from the interaction of the orbitals with many metal surface molecular orbitals, with which they are nearly degenerate. One can consider the interactions as resonant. The downward shift of the la and Rh(100)
Rii(ll1)
Stepped Rh(ll1)
Rh(ll0)
I
[oio]
Fig. 1.2 Cluster models used for simulating the Rh surfaces. (a) (1 11) surface, simulated by a (18,ll) 2 layer cluster, (b) (100) Rh cluster, simulated by a (25,16) 2 layer cluster; (c) (1 10) Rh cluster, simulated by a (19,14,9,8) 4 layer cluster; (d) stepped (1) cluster, simulated by a (6,18,11) 3 layer cluster. The coordination site of CO in calculation of Table 1.1 is indicated.
7
5 0 orbitals indicates that they are part of bonding type molecular orbitals, the upward shift of the
upper part of the 2n* orbital density indicates it to be part of an antibonding orbital set. If the interaction between the fragments increases the corresponding bonding contributions to the electron density of the 2n* orbital may appear. This is already clear in Fig. l.lc where a lower maximum of the 2n* local energy density appears, corresponding to the bonding orbital set contribution. In Fig. l.lc also the changes in 2n*-electron density on a stepped and a nonstepped surface are illustrated. Because of the additional interaction of the 2n*-orbital with the step surface atoms, the ratio of bonding to antibonding peak maxima increases further. As a result the electron occupation of the 2x*-orbital of CO increases further. The resulting weakening of the CO bond gives quite low stretching frequencies, that indeed have been reported for stepped surfaces (ref. 15). In Fig. 1.3 the Extended Huckel contribution to the total energy and the ratio of Eat, of CO adsorbed 3-fold to CO adsorbed atop are presented, calculated as function of d-valence electron occupation (the parameters are those of Rh). The distances of CO to the neighboring metal atoms have been taken the same throughout the calculation. A very interesting trend is observed. Whereas at low to medium d-valence electron occupation three-fold coordination is considerably favoured, at a d-valence electron occupation corresponding to T , the difference between atop and three-fold coordination becomes very small and three-fold coordination is becoming more favoured at higher as well as lower electron occupation. Experimentally at low surface coverage CO prefers atop coordination on Ru (ref. 16) and Rh (ref. 171, but higher coordination sites on Pd and Ni (ref. 18). On W (ref. 19) again high coordination is preferred. In Fig. 1.4 the corresponding bond order overlap population densities: Xij(E) =
z Ci*k Cjk sij6(E-Ek)
(1.1) k are plotted. k labels the molecular orbitals, C? and C? the coefficients of fragment orbitals and Si, their overlap.
Table 1.la. Gross population Extended Huckel Rh( 111) adsorbate
atop 111
ads.
0.998
free
1.o
ads.
0.884
100
bridge
step
110
0.997
0.998
0.996
0.93
1.o
1.o
1.o
1.o
0.882
0.895
0.873
0.84
1 .O
1.o
1.o
1.o
0.12
0.13
0.265
0.37
0.0
0.0
0.0
0.0
__
-~
-.. .
__
~~
8
The changes in sign of xij reflect the changes from bonding to antibonding orbital contributions. As illustrated in Figs. 4, 5, 6 and 7 these changes are due to differences in overlap and local density of states. They can be understood on the basis of the group orbital concept.
Fig. 1.3 Attractive adsorption energy Eattof 1-fold adsorbed CO as a function of the occuuation of the metal valence electron band (dashed line), and the'ratio RgIl of Eat of CO adsorbed Mold to CO adsorbed 1-fold as a function ofthe occupation of the metal valence electron band (solid line). The elements , correspond to the total number of valence electrons according to the periodic system.
As an introduction Fig. 1.5 shows the local density of states pd,,(E) of the surface atom d-, orbital atom On the neighbor of the chemisorbing molecule before and after interaction L'
with CO. Whereas a significant shift upwards is observed for CO adsorbed
atop, a much smaller change is observed for CO chemisorbing in the bridge position. The d,, orbital is shifted considerably above the Fermi level for the atop chemisorbed situation, resulting in a significant lower dzzorbital occupation (Table 1.lb) Inspection of the bond order overlap densities shows that the fragment orbitals formed by the surface d,, orbitals have a significant antibonding contribution. The CO 50 orbital-metal d,, orbital interaction is less unfavorable in the atop position than in the bridge coordination site. In the atop position the 5 0 orbital interact with only partially occupied d-orbital (d,J, whereas in higher coordination sites it interacts with d-orbitals on several atoms (dxz,dx2-yJ.Pauli repulsion is approximately proprotional to the number of neighbour atoms (ref. 6). The differences in d local electron energy density of states behavior derive partly from the different o-symmetry orbital fragments that interact with the CO 50-molecular orbital dependent on CO coordination. Ignoring next nearest neighbor interactions in the atop coordination site the CO So-orbital interacts with one d,, atomic orbital on each neighbor metal atom. In the bridge coordination site it interacts with.the d,, orbital fragment = { 1/~(2+2S)}(~dz2(l)+~dZ2(2)) and in a three-fold coordination site it interacts with the d,, orbital fragment
@%*)
(1.2)
9
x 100
--+'
-20
-30
-10
Energy
0
(eVl
1 " " 1 " " I ' " ' l
n
-20
-30
Energy
(&)
Fig. 1.4b (1): CO 1-fold, dxZ-2r*.(2): CO 2-fold, [( 1/42)(d z(1)+dz(2))]-27t'; (3): CO 2-fold, [(l/~)(d2~(l)+d,,(2))]-27t .
,
-30
L
,
+
l
-
,
L
L
-20
Energy
, u 1 10
0
(eV)
Fig. 1.4 Bond Order Overlap Population densities of CO on Rh(ll1) a. (1): CO 1-fold, d,,-50; (2): CO 2-f0ld, [(1/d2)(d,,(1)+dz2(2))]-5o; (3): CO 2-fold, [(1/d2)(dx (1)+dx2(2))]-50;(4): CO 2-fold, [( 1/32)(d,~~2( l)+d,~~,(2))1-5O. In all plots the Femi level is indicated by EF.
t
-30
'
"
'
:
'
-20
"
'
l
Energy
"
"
-10
l
(eV)
Fig. 1 . 4 ~(1): CO 1-fold, s-50; (2): CO 2-fold, ([ 1/~(2+2s)](s(l)+s(2))}-5o; (3): CO 2-fold, { [ 1/4(2-2s)l(S(l)+s(2)))-2K*.
10
J
L
L
-30 CO -20
'0
-10
(eVl
Energy
0
Energy
1
i 30
lev)
lev)
lo
+-[:O ? - f o l d
CO 2-fold
-10
4
1
Energy
2-fold -20
Energy
-10
(eV)
Figs. 5 . LDOS of dz2(a); ( 1/d2)(dz2( 1)-dz2(2))(b); ( l/d2)(dz2(l)+d&) ( c ) and ( l/V~)(dxly,(l)+dx,y2(2)) (d) before and after adsorption of CO on Rh(l11).
11
-30
-20
Energy
- 10
Energy
(eVl
CO 2-fOld ' ' " ' ' I* I " -30
-20
Energy
-10
ieV)
0
-30
'
' ' I 1
-20
Energy
(eV)
'
+ ,
'
'
"
-10
lev)
Fig. 1.6 LDOS of d,, (a); ( l / ~ ~ ) ( d , ~ ( l ) + d , , ((b): 2 ) ) ( 1 / ~ 2 ) ( d ~ ~ ( 1 ) - ~C; , ~and (2)) ( l / ~ Z ) ( d ~ ~ - ~ * ( l ) + d ~(d) * -before ~ * ( ~and ) ) l )d t e r adsorption ot CO o n R h ( l l 1 ) .
'
II 0
12
Lo
-30 CO I - f o l d
Energy
-10
(ev)
I
Energy
EF
(evi
Fig. 1.7 LDOS o f s (a); {[l/d(2+2s)](s(l)+s(2))} (b) and {[1/~(2-2s)l(s(l)+s(2))} (c) before and after adsorption of CO on Rh(ll1).
13
bridge symmetry
atop
I after adsorption
I
before adsorption
111
100
110
0.67
0.65
0.67
0.92
0.91
0.88
0
I
x
0.86
0.91
0.93
0.92
1
These are called surface grouporbitals. Similar grouporbitals can be constructed for the other d, s orp orbitals. In Figs. 1.5, 1.6 and 1.7 the surface local density of states of the different dor s-group orbitals for atop and bridge adsorption are shown before and after chemisorption. In the atop position the dz2,d,, and dy, have significant overlap with adsorbed CO. In the bridge position group orbitals constructed from d,, and dxzyzorbitals are responsible for bonding. Also interaction with the s atomic orbitals is important. One observes differences in occupation of the antibonding molecular orbital fragments. They are found to arise from the differences in the position of the group orbital electron density maxima, already present before interaction, as well as overlap. The larger the CO coordination shell, the more the average energy of the d orbital group orbital with which it interacts decreases. The relative position of these energy values relates to their position found in the corresponding isolated cluster. Because at lower d-valence electron occupation only bonding orbital fragments become occupied d-electron occupation the two-fold coordination site becomes the most favored one (ref. 6). This is clearly seen comparing the atop d,, overlap population density and the group orbital population density of ( 1/d2)(dXzy2(1)+dXLy2(2)) (Fig. 1.4a). Similar but opposite results are found for interaction of the CO 2n* orbital in different coordination sites. In the atop position the 2n* orbital interacts with a single d,, or d,, orbital. In a bridge coordination site it interacts with group orbitals as:
or the corresponding dyz orbital. Interaction with the asymmetric group orbitals constructed from dX2-,,and dz2 can be ignored. It is small because of the very small overlap with the 2n* orbital in the bridge position (Table 1 . 1 ~ ) Clearly . with surface metal atom s orbitals the 2n* orbital only interacts in higher coordination sites where antisymmetric group orbitals can be constructed. The local density of states of orbital @(dxz)and group orbital Q2'(d,) are shown in Fig. 1.6. The corresponding bond overlap densities are shown in Fig. 1.4 and numerical values are listed in Table 1 . 1 ~Only bonding orbital fragments are occupied. In the bridge coordination site more bonding orbital fragments are found to be occupied than in the atop coordination site (compare (l),(2) and ( 3 ) in Fig. 1.4b). Because only bond orbital fragments are occupied, the interaction with the CO 2n* orbitals favors coordination in the bridge coordination site. In Figs. 1 . 4 ~and 1.7 also the corresponding behavior of the s metal group
14
Table 1. lc. Bond Order Overlap Population between CO and surface group orbitals atop
bridge symmetry
111
100
110
0
0.122
0.124
0.117
-0.0029
0.9 1
0.85
0.93
0.186
0.181
0.187
K
,
0.042 0.026
0.059 0.105
0.030 0.021
O.OOO8S1
orbitals is shown. One should remember that back-donation population of the 2 ~ orbital * becomes more favoured if the surface work function decreases, whereas donation depopulation of the 50 orbital has the reverse behavior. We have found that backdonation tends to favour bridge coordination but donation favours atop coordination for metals with a high d valence electron occupation and adsorbates that stronglu interact with the metal d-electrons. It is of interest to note that trends in stabilization of 0-and n-type interaction as a function of electron occupation are found to follow closely the trends as predicted solely from the corresponding surface group orbital densities of state at the Fermi level. Using second order perturbation theory we have demonstrated elsewhere that the following expressions for the attractive contribution to the bond energy of an adsorbate coordinated to a metal surface can be derived (ref. 10):
The repulsive contribution to the bonding energy is given by
Ere. = -4c PIcL,]S , i = -4 zp"
Q,l'
soc(,I
a,i pQj(EF)is the group orbital local density of states of valence band j at the Fermi level energy. The group orbital local density of states (LDOS) is given by: Paj(EF)
=c (@Qjwk)2 g(E-Ek) k
(1.7)
a metd surface orbital eigenfunction with corresponding eigenvalue E k p Q J .(EF) in averaged around its value at the Fermi level over an expression (I), is the value found for energy interval of the order = P2a,/Aa,j. This corresponds to the width of the broadened interacting adsorbate electron levels. v k is
15
In the derivation of expression (1) it is assumed that the reduced overlap matrix elements
(5):
are very different for the different valence orbitals. X , is an adsorbate orbital and H' the effective coupling mamx element between adsorbate orbital and surface. Since H' is totally symmetric, the symmetry and spatial extension of adsorbate orbital X, determines the symmetry of the surface metal orbital fragment Qj*, the group orbital. Aaj is the total band width of the metal valence electron band corresponding to @aj. and P a j is a measure of the electron occupation of that electron band.
@ is the surface dippole potential of the metal surface considered and the term -e2/(4ra+k,> represents the image potential interaction of ion state a,with effective adsorbate to metal distance
ra and screening length k, (ref. 20). Expression (2) relates the attractive component of the binding energy to: a. the surface group orbital local density of states at the Fermi level b. the effective energy difference between adsorbate orbitals and the Fermi level c. the surface metal orbital electron occupation d. orbital overlap Expression (2) is a interesting result. Whereas it has been speculated by many authors (ref. 21), that a relation between bond energy and density of states at the Fermi level should exist, expression (2) explicitly states this relation with the modification that the surface group orbital density of states at the Fermi level has to be used. Expression (1.3) gives the repulsive part to the bond energy. It is simply the repulsion between doubly occupied orbitals, calculated within the Extended Huckel method. Pa,? is the reduced overlap energy matrix element cc and S,,,O the overlap matrix element for 2 = 1. 2 is the number of neighbor atoms of the adsorbing molecule.
One observes that the repulsive contribution is proportional to 2.Norskov et al. (ref. 22) derived an alternative formula for the repulsive energy part based on free electron theory. The effective medium theory calculates the attractive part to the bond energy from electron density distributions that are a superposition of atom centered electron densities. As a consequence the dependence on the density of states at the Fermi level, the result found if interaction is weak, does not explicitly appear. The work function dependence according to expression (2) results on a similar dependence of the electrostatic field on chemisorption as found from first principle calculations (ref. 23),
effective medium theory (ref. 22) or adapted Extended Huckel theory (ref. 24). Lowering the effective ionization potential enhances electron backdonation between metal and adsorbate. Nieuwenhuys (ref. 25) has extensively documented experimental results indicating this correlation. One observes also in Fig. 1.3a a decrease in bond strength as the &valence electron band becomes
16
filled. This is due to the decreasing averaged LDOS at the Fermi level or alternatively the occupation of antibondingorbital fragments, The repulsive contribution to the bond strength is due to the interaction with doubly occupied core orbitals which is proportional to the number of neighboring atoms. As mentioned earlier repulsive interaction favours the atop configuration. The experimental observation that CO favours bridge coordination sites on Ni and Pd, but atop adsorption on Co, Rh and Ru can be understood on the basis of the arguments presented. The d-valence electron band width increases from Ni to Pd and Pt. As a result the repulsive interaction with the highly occupied d-valence electron orbitals increased in the same order. Secondly the workfunction tendsto decrease, so that backdonation also decreases. Both effects together result in the favored atop position of PtOn Co and Rh the decreased d-valence-electronoupation enhances the interaction with the d-valence electrons compared to that with the d-valence electrons. The attractive interaction with highly occupied d-valence orbitals also favours atop positions (ref. 6).
1.3 CRYSTAL FACE DEPENDENCE-PROMOTER EFFECTS Since the coordination possibilities of chemisorbed CO will vary according to crystal surface, it is impossible to derive general rules for changes in the chemisorptive bond strength with crystal face. Clearly changes in work function will affect the balance of the high coordination directing bonding contribution due to the backdonation of electrons, favoured by a low work function into the lowest unoccupied CO 2~*-orbitaland unfavorable to the atop directing contribution due to donation of electrons from the CO So-orbital into the unoccupied metal orbitals. This phenomenon is nicely illustrated by the effect of potassium coadsorption at the (1 1 1) face of Pt on the coordination of CO (refs. 22,23,24). The experiments of Garfunkel and Somorjai (ref. 26) show elegantly that CO changes from the atop to bridge coordination site under the influence of the work function lowering effect of potassium. First principle calculations of
Table 1.2a
Adsorption of C, 0 and CO on Rh( 111) ______
species C
0
co
_
site
_
_
~
1-fold 2-fold 3-foldfcd' 3-fold hcp5 1-fold 2-fold 3-foldf ~ c 3-fold hcp 1-fold 2-fold 3-foldf ~ c 3-fold hcp
hX(2) (A)
____--
-
-4.43 -4.88 -5.48 -5.58 -6.59 -6.09 -6.27 -6.30 -2.20 -2.27 -2.45 -2.39
-6.49 -6.85 -8.08 -8.19 -8.21 -8.12 -7.85 -7.59 -3.40 -3.60 -3.75 -3.69
1.80 1.40 1.20 1.20 1.45 0.70 0.40 0.50 1.90 1.50 1.40 1.40
d ~ ((A) ~ )
1.80 1.94 1.96 1.96 1.45 1.51 1.60 1.63 1.90 2.01 2.09 2.09
(1) bond energy contribution excluding the two-body repulsion term; (2) height of adsorbing species (X=C, 0 or CO) above the surface; (3) distance of adsorbing species (X=C, 0 or CO) to the nearest Rh atom; (4) no Rh present in second layer; (5) Rh present in second layer.
17
Table 1.2a' Adsorption of C, 0 and CO on stepped Rh( 111) surface species
site
hx(2) (A)
dx(3) (A)
C
1-fold 2-fold 3-fold fc& %fold hcp5 1-fold 2-fold 3-fold~ C C 3-fold hcp 1-fold 2-fold 3-foId hcp 3-fold hcp7 3-fold ~ C C
1.80 1.40 1.20 1.20 1.45 0.60 0.30 0.40 1.90 1S O 1.25 1.40 1.40
1.80 1.94 1.96 1.96 1.45 1.47 1.58 1.60 1.90 2.01 1.99 2.09 2.09
0
co
-5.34 -5.52 -6.60 -6.00 -6.32 -5.97 -2.22 -2.28 -0.78 - 1.94 -2.24
-7.96 -9.80 -8.21 -8.75 -8.19 -7.56 -3.86 -3.61 -5.62 -3.24 -3.56
(3) bond energy contribution excluding the two-body repulsion term ( 2 ) height of adsorbing species (X=C, 0 or CO) above the surface (3) distance of adsorbing species (X=C, 0 or CO) to the nearest Rh atom (4) no Rh present in second layer ( 5 ) Rh present in second layer (6) distance of the adsorption site to the nearest Rh atom (7) adsorbed on the step.
Freeman et a1 (ref. 23) show clearly the change in relative position of the CO molecular orbitals with respect to the surface Fermi level. The CO 2x*-orbital becomes closer to the Fermi level, whereas donation from the CO 5 0 orbital becomes more difficult because the energy difference with the surface Fermi level increases. Similar changes may be expected comparing dense faces with more open faces and chemisorption close to steps. Because according to classical electrostatics the electron distribution tends to smear itself out and positive charge tends to accumulate at edges, a dipole moment with the positive charge directed into the outward direction tends to develop. As a result the work function decreases at the more open surfaces and edges, favouring electron backdonation into the CO 2n* orbital. This tends to favours high coordination and as we will see later also dissoci ation. These electrostatic effects are not accounted for in the Extended Hiickel method, which has to be taken into consideration applying this technique. In Table 1.2 the energies of adsorption computed according to the ASED method for Table 1.2b Adsorption of C, 0 and CO on Rh( 110) species
site
C
1-fold 2-f0id(3) 2-f01d(4) 4-fold 1-fold 2-foldi3) 2-f0id(4) 4-fold 1-fold 2-f0id(3) 2-f0id(4) 4-fOld
____
0
co
dx(2)(A)
-4.33 -5.63 -5.05 -4.70 -6.09 -5.48 -6.29 -4.87 -2.04 - 1.95 -2.11 -2.04
-6.38 -8.47 -7.88 -5.96 -8.42 -7.25 -8.29 -6.57 -3.24 -2.80 -3.40 -2.86
1.80 0.30 1.30 0.60 1.40 -0.60 0.70 0.10 1.90 0.90 1S O 0.70
1.80 1.92 1.87 2.40(5) 1.40 1.99(6) 1.51 2.33(7) 1.90 2.10 2.01 2.43@)
18
Table 1 . 2 ~ Adsorption of C, 0 and CO on Rh(100) species
site
C
1-fold 2-fold 4-fold 1-fold 2-fold 4-fold 1-fold 2-fold 4-fold
0
co
h#) -6.57 -8.07 -7.69 -8.87 -8.81 -5.50 -3.51 -3.64 -3.19
-4.52 -5.24 -5.68 -6.54 -6.82 -4.23 -2.31 -2.34 -2.23
(A)
1.80 1.30 0.80 1.40 0.70 -0.40 1.90 1.so 1.10
1.80 1.87 2.06 1.40 1.51 1.94(9) 1.90 2.01 2.19
(4) Height of adsorbents to the surface in A; (2) Rh-X distance in A; (3) Bridge in between 2 next-nearest neighbors in the (001) direction; 4) Bridge in between 2 nearest neighbors in the (110); (5) Distance to second layer Rh: 1 . 9 4 ~; (6) Distance to second layer Rh: 1.53=A; (7) Distance to second layer Rh: 1.44=A; (8) Distance to second layer Rh: 2.04=& (9) Distance to second layer Rh: 1SO=&
a
different adsorption configurations on the Rh clusters shown in Fig. 1.2 are presented. According to the ASED method the total energy consists of an attractive term, computed according to the Extended Hiickel method and a repulsive term, defined by Anderson (ref. 13) on the basis of an approximation to the Hellman-Feynman theorem. This method allows the prediction of bond distances and the calculation of potential energy curves. The parameters used are given in Appendix 1 and follow from the work of Hoffmann (ref. 27a) for Rh and the work of Anderson (ref. 27b) for C and 0. Experimental results of CO adsorption on Rh single crystal surfaces are rather scarce. CO is reported to adsorb linearly (1-fold) on Rh with adsorption energies of respectively -31.6 kcal/mol (ref. 28), -32 kcal/mol (ref. 29) and -31 kcal/mol (ref. 30) for the (11l), (100) and (110) surfaces. The difference in adsorption energy of linearly and bridge bonded CO is reported to be 4 kcaVmol for Rh(ll1) (ref. 17) and 1.1M.06 kcal/mol for Rh(100) (ref. 29). In a series of LEED studies van Hove and Somorjai (ref. 31) found a trend of an increasing Rh-C distance: CO is adsorbed with an Rh-C distance of 1.94rtO.1 A (1-fold), 2.03a.07 8, (2-fold) and 2.02k0.04 A @fold, on coadsorption With CgHg). The desorption energy of 0 2 is reported to be -56k2 kcaUmol on Rh(ll1) (ref. 32) and -85 kcal/mol (ref. 33) for Rh(100). If we combine the adsorption energy of e.g. three-fold hcp oxygen (Ead=-6.31 eV) with the calculated bond strength of molecular oxygen, we obtain a desorption energy for 0 2 of -9.65 eV (-222.6 kcavmol). This overestimation of the desorption energy is partly due to an overestimation of the adsorption energy of an oxygen atom, and a 50 % underestimation of the 0 2 bond strength (-2.97 eV vs. an experimental value of -5.16 eV). The calculated oxygen distances are rather small, e.g. oxygen on a 3-fold hollow site has a R h - 0 distance of 1.58 A, which indicates a much smaller distance as within the metal oxide (Rh-0 = 2.03 A) (ref. 34). Wong et al. (ref. 35) found a Rh-0 distance of 1.98 site for the Rh(l11)-(2x2)0 surface.
for oxygen on afcc hollow
19
Oed et a1 (ref.36) used LEED intensity analysis and found an oxygen height of 0.95k0.048, for Rh(100)-(2x2)0. Oxygen on this surface is found to adsorb in hollow (4-fold) sites. In agreement with theoretical models (ref. 12) carbon is most stable on a hcp hollow site on Rh( 111). Carbon does not form a stable carbide on Rh. It is important to remember the difference between computed and observed oxygen bond strength, because it is mainly responsible for the too low values of the activation energy that we predict later on. In order to distinguish the effects of changing the electronic structure and adsorption site topology in Fig. 1.8 the ratio's of the attractive contribution to the bond energy for CO adsorbed atop to the (11l), (100) and (1 10) faces of the samefcc transition metal clusters are presented as a function of the valence electron occupation. Only at medium to low band filling one finds that bonding in the atop position is strongest to the most open surface. Very interesting is the observed change in sequence at high d-valence electron occupation. From a comparison of the attractive bond strengths one finds for a metal containing 9 electrons in the valence electron band that as far as the atop positions are concerned the (100) surface becomes most reactive and both the (1 11) and (1 10) surfaces have lower reactivities. The results strongly depend on the d-valence electron occupation. It not only is a function of the total number of valence electrons, but also depends on the distribution of electrons over the s-, p- and d-valence electron subbands. In the present case the s-, p-band occupation is low (0.2 and 0.0
1
R
1.1--
respectively with the particular parameters used). As will be shown the stronger interaction
110 ----.._..
1.0
of the (110) face at low electron occupation has a direct relation to it's d-valence electron distribution. Therefore the stronger interaction of the (100) face might occur at higher total valence electron occupation, if more electrons were located in the s-, p-valence electron bands. This distribution is obviously a function of the difference in s and d atomic orbital energies.
4
;-._ ._
,'
-._
0.9-
--._ ..,,--Nb I
0.8 -6
'
"
j
-4
"
' '
1-c I ' -2
Rh
' ' ' I ' 0
A
2
Fig. 1.8 The ratio's R of Eattof CO adsorbed 1-fold on Rh(100) (solid line) and Rh(ll0) (dashed line) to CO adsorbed I-fold on Rh(l11) as a function of the occupation of the metal valence electron band.
20
Analyzing the 0- and n- changes in the M-C bond orders one observes that the differences in behavior arise from changes in the interaction of the 2n* as well as 5 0 C o molecular orbitals with the d metal atomic orbitals. Changes in bonding with the 50
1
p,(E)
I
Hh 13001
Rh 11101 I " " 1 ' -30
orbital dominate as becomes apparent from the correlation with the differences in bond strength computed in the atop position as shown in Fig. 1.3. As follows from expression ( 2 ) the differences in bonding with the d,, orbital can be rationalized on basis of the changes in relative LDOS at the EF level. It can be seen from Fig. 1.9 that for Rh the LDOS at the Fermi level is highest at the (100) surface and close for the (111) and (1 10) faces. At lower valence electron band
" -20
'
"
"
Energy
'
I -10
0
(eV)
Fig. 1.9 LDOS ofdz2ofthe Rh(l1 I), Rh(1W) and Rh(ll0) faces. EF indicates the Fermi level.
occupation the LDOS at the Fermi level of the (110) surface increases and dominates. At higher band occupation the LDOS at the Fermi level of the (111) face
dominates. It is of interest to note that changes in surface reactivity as predicted by the EHMO calculations have been reported by Nieuwenhuys (ref. 37) and Banholzer (ref. 38) for chemisorption of NO. Banholzer et a1 explained the higher reactivity of the (100) surface found for
Z
b
Fig. 1.10 dxyorbitals in thefcc crystal
Pt on the basis of the Bond surface dangling bond model (ref. 39). Early work by Bond (ref. 39) and Weinberg (ref. 40) used the d orientation of orbitals at a surface and Goodenough's band theory (ref. 41) to study the interaction of molecules and atoms with a surface. Application of this
21
model to Rh results in similar predictions. Here we find that changes in electron occupation of the proper symmetry orbitals explain the differences in behavior of the Rh and Pt surfaces, implying that the Bond model requires modification. Such an approach can be developed based on more recent band models (ref. 42). We will shortly discuss this. Whereas highly simplified, it provides a striking insight confirmed by the earlier presented Extended Hiickel calculations.
In the bulk the face centered cubic lattice metal atoms have 12 nearest neighbors. As sketched in Fig. 1.10, the 3 dxy, d,,, and d,, orbitals each have 4 nearest neighbors, which are not shared. This leads to a symmetric density of states, three fold degenerate in the three perpendicular planes. The d,, and dxLyz have nearest neighbors at a 42 larger distance than the d,,, d,,, d,,
Fig. 1.11 Schematic sketch of the tight binding bulk valence electron distribution of a fcc transition metal.
orbitals. They will form a two-fold degenerate electron band of much narrower band width than the d,,, d,, and d,, orbitals. The s orbitals have a much larger overlap than the d-orbitals and have 12 nearest
neighbors, they will form a broad band, usually overlapping the much smaller d-valence electron energy band. The resulting electron distribution is sketched in Fig. 1.1 1. For group VIII metals to a good approximation 1 electron per atom is present in the s-band valence electron band and the other valence electrons are located in the d valence electron band, with varying electron occupancy. So for metals Ni, Pd and Pt with 9 d-valence electrons one expects the holes in the d-valence electron band to have considerable d,, d,, and d,, character and little d,, and dxly2character. Thus in the spirit of second order perturbation theory in the following only interaction with the d,,,
d,, and d,, orbitals will be considered. At the (111) face each of the
d,,, d,, and d,, surface orbitals looses one neighbor. As pointed out by Kahn and Salem (ref. 43), the resulting three degenerate dangling orbitals will rehybridize according to the local symmetry of the surface atoms. As a result two degenerate and one symmetric surface orbital is formed. The degeneracy is lifted and the resulting d electron density of states at the Fermi-level is sketched in Fig. 1.12. At the (100) face, the dxyorbital in the (100) plane does not lose any neighbors. However the d,, and d,, orbitals each loose two neighbors. As a result the local density of states is split into two bands, a broad band corresponding to the dxyorbitals and two more narrow bands dyz and d,, that get a higher electron occupation (Fig. 1.13). At the (1 10) face two different atoms are generated. The edge atoms loose 5 neighbors, the other atoms only one, resulting in a dangling orbital of sigma type. At the edge atom two n-symmetry orbitals are generated, from d orbitals loosing 2 neighbors and 1 sigma symmetry dangling orbital is generated with loss of 1 neighbor. The resulting dangling orbitals and density of states curves are sketched in Fig. 1.14. Let us first discuss the consequences of this orbital schema for the interaction with an orbital of sigma symmetry (H atom, sigma orbital of CO etc). In the atop adsorbed state a o type orbital will only
22
interact with (1 11) and (1 10) face d-orbitals, because at the (100) face no d orbitals with sigma symmetry with a finite density of states at the Fermi level are available. Since the o-dangling bond width at the (1 11) and (1 10) face are comparable, the interaction with the d-valence electron band will be the same. This implies a bonding interaction with an adsorbate orbital that is half filled, but a bonding or repulsive interaction with an adsorbate orbital of sigma symmetry that is completely filled. At the end of the row of group VIII metals the surface d-orbitals will be nearly filled and repulsion tends to dominate. When the surface d orbital band depletes, repulsion will become converted into attraction (the LDOS at EF). We will next consider the interaction with an adsorbate orbital of n-symmetry. All these faces contain dangling bond orbitals of n-symmetry. At the (1 11) faces these orbitals are broader than at the (110) and (100) face. Since the bulk Fermi level does not change one expects a higher n-dangling bond electron occupation for the (110) and (100) faces than for the (1 11) face. We have argued earlier (ref. 24) that band occupation is such that the LDOS at EF for the (110) and (100)
Fig. 1.12 d-valence electron distribution at the (1 11) face; (a) the out of plane lobes of degenerate dxy,dyzy d dx,zatomic orbitals; (b) linear combination of the out of plane lobes of the d,,, dyzand dx, atomic orbitals symmetry adapted to the (1 11) surface; (c) schematic surface d-electron density of states at the (1 11) surface
nlE)
E’i-
Fig. 1 . 1 3 d-valence electron distribution at the (100) face (5) d,, and d, lobes at the (100) face; (b) Schematic sketch of surface electron distribution at the (100) face
lobes is higher than in the (1 11) face. So for CO and NO n-backdonation into the empty adsorbate n-symmetry orbitals is larger at the (1 10) and ( 1 0 0 ) faces than the (1 11) face. Feibelman et a1 (ref. 44) as well as Joyner (ref. 45) have shown that S adsorption to a transition metal surface causes changes in the local density of states extended over several atom distances measured from the sulphur adsorption site. At neighboring atoms significant decreases in the LDOS are computed. Application of expression (2) predicts a decrease in the CO bond strength on such a site. This will lead to a decrease in surface coverage of CO with S much faster than expected on the basis of site blocking. Such enhanced effects have experimentally been observed (ref. 46). z (4 In Table 1 . 3 changes in the total energy and attractive energy induced by the presence of steps are shown. Only configuration with CO perpendicular to the dense surface phase are considered. The surface step studied is shown in Fig. 1.2d. It is in the (1 10) direction on the (11 1) surface. Table 1 . 4 summarizes the changes in maximum bond strength as computed for the different Rh surfaces.
dTt
Fig. 1 . 1 4 d-valence electron distribution at the (1 10) face. (6) (1 10) face d dangling bonds; (b) schematic d valence electron density of states at the (1 10) face.
24
I
&\
&\
1-fold 2-fold 3-fold~ C C 3-fold hcp
-2.20 -2.27 -2.45 -2.39
-3.40 -3.60 -3.75 -3.69
Rh(100)
1-fold 2-fold 4-fold
-2.31 -2.34 -2.23
-3.51 -3.64 -3.19
Rh(ll0)
1-fold 2-fold (001) 2-fold (1 10) 4-fold
-2.04 -1.95 -2.11 -2.04
-3.24 -2.80 -3.40 -2.86
surface Rh(ll1)
site
Fig. 1.15 shows the corresponding adsorbate valence local density of states of CO adsorbed
on 2-fold (in the (1 10) direction) and 4-fold sites on Rh(1 lo), and of CO adsorbed on a 2-fold site on Rh(100). It is of interest to note that the hollow sites of the (110) surfaces have a very similar interaction with the 2n*-orbitals as CO adsorbed close to the steps on the (111) surface. One observes that the attractive contribution to the bond energy increases considerably if CO is located such that the step atom wave functions are able to overlap with the CO 2a*-orbitals.The increased interaction between 2n*-orbital and step atoms becomes also clear from the increased broadening of the 2x* CO LDOS at the step sites (Fig. 1.1~).As pointed out earlier, these increased interactions with the 2n* orbital result in an increased 2n* electron occupation and a corresponding low CO stretch frequency. One also notes that there is a significant change in the repulsive energy contribution. This increase is such that the overall energy change results in an unfavorable interaction of the CO in contact with the steps compared to CO bonding to the non-stepped surface. Experimental evidence so far is contradictory (ref. 47). Some authors (refs. 47a,c) report increased
CO dissociation to the presence of steps. The increased 2n* stabilization seems to agree with this. Others do not find any changes on the bond strength of chemisorbed CO (ref. 47d), which seems to agree with our results. More rigorous calculations than the ASED method to resolve this issue are necessary.
5u
-30
-20
Energy
-30
-10
lev)
-20
Energy
CV)
Ln'
.c
-20
-10
0
E n e r g y (evl Fig. 1.15 LDOS of CO 50 and 2n* orbitals of CO adsorbed on 2-fold (1 10 direction) (a) and @old (b) sites of the (1 10) surface, and on 2-fold (c) site of the (100) surface.
26
1.4
CO DISSOCIATION
The advantage of using a simple semi-empirical method to study chemical bonding is that it enables the computation of many different reaction paths. Based on the insights that can be generated in this way a few selected paths can be chosen for furtherevaluation by more sophisticated techniques. Empirically the activation energy for dissociation appears not to be related with the heat of adsorption of carbon monoxide. For instance the strength of the chemisorptive bond of CO to Pt is larger than that with Ni, but CO dissociation occurs much more readily on Ni than on Pt (ref. 42). It also is not related with the tendency to chemisorb preferentially in the atop or bridge position. Early UPS data (ref. 48) enabled the conclusion that the activation energy for dissociation relates to the degree of population of the antibonding CO 2n*-orbital. Inverse photoemission studies seem to confirm this (ref. 49). The higher work function of Pt compared to Ni then explains the ease of dissociation on Ni. We will see that theoretical studies tend to confirm this. Shustorovich (ref. 12) derived rules for the calculation of activation energies based on the Bond Order Conversation (BOC) postulate. According to Shustorovich's expressions the activation energy for dissociation relates to the bond strength of the resulting C and 0 atoms. The trends predicted according to this theory appear quit reasonable. According to the BOC model CO dissociates more readily on Ni than Pt because of the increased bond strength of oxygen atoms to Ni compared to Pt. Strictly dissociation of CO is only thermodynamically allowed if the free energy of the chemisorbed fragments Cad and 0, is larger than that of gas phase CO. Chemisorbed CO is a stable intermediate, if there is an activation energy between chemisorbed CO and dissociated CO. This appears to be the case for all transition metals. An interesting thermodynamic analysis has been presented by Benziger (ref. 50). Whereas Shustorovich's BOC method predicts in general correct trends for the activation energy of dissociation, its main shortcomings are that usually its predicted activation energies are too high, and secondly it cannot be applied to a particular dissociation path. According to the ASED method is the two important variables are: - occupation of the CO ZK*level - bonding of the dissociated C and 0 atoms.
Table 1.5a. Results of CO dissociation on Rh( 111)
I
n
111
IV V VI
-2.20 -2.39 -2.45 -2.39 -2.39 -2.39
-1.58 -2.17 -1.98 -2.17 -2.17 -2.52
I
(1): energy of adsorbed CO before dissociation (2): energy of atomic C and 0 after dissociation (3): calculated activation energy
2.99 3.01 2.30 3.25 1.98 2.01
-3.40 -3.75 -3.69 -3.69 -3.69 -3.69
I
1
-2.83 -3.62 -3.52 -3.85 -3.70 -4.80
4.85 3.39 2.14 3.07 1.98 2.00
27
In Table 1.5 the attractive contribution to the bond energies and total energy contribution are separately listed. Dissociation on the unstepped and stepped (1 11) surface has been studied, as well as the dissociation reaction paths at the (100) and (110) surfaces. In all calculations, the same general approach is followed. In a first step CO is adsorbed on a particular site, this yields the starting energy. The carbon atom is fixed on this position, while the oxygen atom is allowed to move according to the reaction path considered. At several intermediate steps in each reaction path, the height of the oxygen atom to the surface is optimized and the energy is calculated. The final energy (denoted as “end” in Table 1.5) is obtained when CO is dissociated and the oxygen atom has reached its final position. The activation energy is the difference between the maximum in energy found during this reaction path, and the starting value. Tables 1.5 summarizes the reaction paths for dissociation considered as well as computed activation energies. As the optimal height of adsorbed carbon is lower than the height of carbon in carbon monoxide adsorbed on the same site, the carbon height will decrease during the process of CO dissociation. The introduction of a free carbon height
Table 1Sb. Results of CO dissociation on stepped Rh( 1 1 1) reaction path
s-I s-I1 s-I11 s-IV s-v s-VI s-VII
start(’)
-2.24 -2.24 -0.92 -0.92 -2.24 -2.24 -0.92
-1.68 -1.68 - 1.47 - 1.47 - 1.68 -1.33 -0.47
3.46 2.44 2.37 2.02 3.17 2.26 2.40
Eatt
-3.56 -4.57 -4.57 -3.56 -3.56
end(2)
-3.23 -4.68 -5.45 -3.32 -3.80 _
3.54 2.03 3.23 2.55 2.88 2.48 2.54 _
~
(which has to be optimized for every intermediate step in a reaction path) results in a large increase in the number of calculations. Therefore, we performed two series of calculations for a few selected reaction paths with the carbon height set to the limiting values of adsorbed carbon and of adsorbed carbon monoxide. As the maximum energy found with the latter always exceeds the maximum energy found with the former, the value found with the carbon height set to the value of
Start(’)
V
- 1.98 -1.98 -2.05 -1.89 -1.89
Etot
end(2)
-0.43 -2.43 0.30 -2.44 -0.87
start(’)
3.05 3.00 2.87 1.64 2.95
___-___
-3.28 -3.28 -3.25 -2.54 -2.54
Eatt
end(2)
-3.05 -5.35 -1.15 -4.62 -4.32
2.53 2.57 3.22 1.95
____
28
Table 1.5d reaction path
IV V
VI
Results of CO dissociation on Rh(1lo), simulated by a (19,14,9,8) 4 layer cluster
Start(')
-1.95 -1.95 -2.04
Em,
end(2) -2.46 -1.10 -1.08
Ed3)
start(')
Eatt
-4.64 -3.04 - 1.74
-2.80 -2.80 -2.86
1.90 2.94 2.41
end(2)
Eacd3) 1.53 2.13 3.27
adsorbed carbon is used for obtaining the activation energy. The reaction paths found most favorable according to Table 1.5 are pictured in Fig. 1.16. A detailed reaction sequence has been sketched in Fig. 1.17. During the first stage of CO dissociation, the C-0 distance remains constant and the height of the oxygen atom is decreasing. This indicates that CO bends first without stretching the CO bond. After bending the CO axis is stretched: the C - 0 distance is increased. From many reaction paths considered we have found a clear pattern for the lowest activation energy dissociation paths. Dissociation is favoured with the dissociated atoms
Table 1.5e Results of CO dissociation on Rh(100) reaction path I I1 111 IV V
start(')
-2.23 -2.34 -2.3 1 -2.23 -2.23
-2.51 -1.35 - 1.43
1.48 3.06 2.66 1.48 1.80
Eatt
end(2)
i -3.19
-3.51 -3.19 -3.19
-3.57 -4.68 -3.08 -2.91 -2.32
1.43 3.06 3.64 1.43 1.12
ending in high coordination sites and sharing the least number of surface metal atoms. Secondly for the dense faces crossing of the CO bond over a surface metal atom (reaction paths V for (1 1l), S-11 for the stepped ( l l l ) , I for (100) and IV for the (110) face), when it stretches to dissociate, is considerably favoured. Computation of the bond overlap populations between the initially unoccupied CO 2 ~ orbitals * and the surface metal atom d orbitals concerned (Fig. 1.18a), clearly show the importance of stabilization of the stretched CO bond by interaction with the surface d orbitals. The higher the antibonding CO 2n* molecular orbital electron density (Figure 18b) becomes, the more the activation energy for CO dissociation becomes lowered (Table 1.6). This observation may explain why it is experimentally found that the activation energy for CO dissociation behaves parallel to the bond energy of CO in the atop position (refs. 37,38). As discussed earlier this relates also partly with the 2 ~ orbital * interaction. The promoting action of work function lowering coadsorbents as alkali metals (ref. 26) or oxides (ref. 51) that enhance dissociation, agrees with the need to populate the antibonding CO 2n* orbital in order to lower the activation energy for dissociation.
29
RhC100) RhCllO)
’
IV
Fig. 1.16 Most favorable dissociation reaction paths for CO on Rh(l1 l), stepped Rh(1 l l ) , Rh(100) and Rh( 110).
0.0
1
I
I
I
I
Rh (111) Mechanism V
Projected C-0 d i s t a n c e
i
(A)
Fig. 1.17a Bond energy as a function of CO distance projected on the surface (reaction coordinate).
30
Table 1.6
Summary of lowest dissociation energies of CO
surface population(') Rh(ll1) Rh(100) Rh(ll0)
1.98 1.48 1.90
1.43 1.53
0.55 0.65 0.56
(1): 2x* orbital gross population at transition state
On the other hand in order for a CO molecule to dissociate relative large ensembles consisting of at least 5 or 7 Rh atoms are required. Dissociation will be suppressed for geometric reasons if by coadsorption of nonreactive coadsorbates (e.g. S ) the ensemble size is diminished or the sites for favorable dissociation (high coordination sites) are occupied by the adsorbate.
O 0
-
17
P r o j e c t e d C-0 d i s t a n c e
(A)
Fig. 1.17b CO bond length (solid line) and surface-oxygen distance (dashed line) as a function of CO distance projected on the surface (reaction coordinate).
31
11111
11001
(1101
-30
-20
Energy
(eV)
Fig. 1.18 LDOS of 2n* of CO at point of highest energy during CO dissociation on Rh(l1 l), Rh(100) and Rh(ll0) respectively.
1.5 DISCUSSION AND CONCLUSIONS We have presented a detailed discussion of the surface chemical bond of chemisorbed CO and analyzed CO dissociation. Also some indications were given how these aspects relate to the effect of promoters on the chemical reactivity of CO. The analysis of CO dissociation highlighted the need to activate the CO 2n* bond and the importance of stable C and 0 bonds. We shortly mentioned the electronic details of CO insertion reactions based on Hoffmann's analysis. His analysis has shown that the repulsive barrier due to the interaction of a doubly occupied o-type orbital of an inserting fragment (H or CH3) with the doubly occupied CO 5 0 orbital can be reduced by backdonation of electrons from the corresponding antibonding orbital into empty orbitals of proper symmetry. On atoms d-orbitals provide the proper symmetry, so that atoms with empty d-atomic orbitals are required. Essentially the same point of view has been expressed by Koga and Morokuma (ref. 52) on the basis of ab initio calculations. This may explain why on Pd metal ions are favoured sites for methanol or oxygenate production, if the rate limiting step is a CO insertion
32
step. Based on experimental evidence (ref. 53) as well as calculations (ref. 54) this point of view has been presented by Ponec (ref. 53). No calculations on CO insertion are available for surfaces. Recently Baetzold (ref. 55) and Hoffmann et a1 (ref. lb) published results in Fisher-Tropsch chain growth mechanism. The results confirm the earlier concepts. There is however a difference in detail if recombination or insertion reactions occur on metal surface sites consisting of several metal atoms. The antisymmetric orbital required to release electrons from the repulsive antibonding orbital formed by interacting sigma-type orbitals, can also consist of an antisymmetric combination of orbitals centered on different atoms (ref. 9). Because of the small dimensions of the d orbitals, on Ni and Cu surfaces this implies that antisymmetric group orbital fragments of s-type orbitals may become important. In view of this, it is of interest to return to the question of top versus bridge site banding. It has been pointed out that if the interaction with atomic d orbitals is significant CO tends to adsorb atop of surface metal atoms. If the interaction with the d valence electrons is relatively unimportant coordination to high coordination sites results. For theses reasons CO prefers high coordination sites on Ni but atop coordinations to Pt. It has shown by Minot, van Hove and Somorjai (ref. 56) that when adsorbed to a metal with extended d-valence atomic orbitals hybridization of CH, prescribes a preference for a particular metal atom coordination. For instance a CH, fragment prefers atop coordination, a CH2 fragment bridge and a CH fragment threefold coordination. A first condition that has to be satisfied in a Fischer-Tropsch catalyst is that CO dissociates (ref. 57). For this reason metals that only weakly promote the CO 2n* orbital are not good Fischer-Tropsch catalysts. Cu, Pt and Ir belong to this category. Alternative synthesis gas conversion pathways, as methanol synthesis, then become relatively more favoured (e.g. Cu). One expects an optimum metal-carbon interaction to favor formation of higher hydrocarbons. If the metal-carbon bond is too strong, carbide formation will occur at the expense of carbon-carbon bond formation. If the metal-carbon bond is too weak formation of nonreactive graphite will occur at the cost of the availability of reactive carbon. Depending on the hydrogen coverage methanation or graphite formation will compete. An intermediate metal-carbon interaction will lower methanation, suppress carbide formation and enhance the propagation reaction. We have shown elsewhere (ref. 57) that the activation energy for C-C bond formation is only a weak function of the metal-carbon interaction. We will conclude with a short discussion of the effect of alloying on chemisorption of CO. A classic experiment is that of Soma-Nota and Sachtler (ref. 58). They studied the effect of silver alloying on the chemisorption of CO on Pd with infrared spectroscopy. CO favours bridge coordination on Pd and does not chemisorb on Ag at room temperature. The explanation of the latter derives from two effects. The CO 5 0 orbital has a repulsive interaction with the filled d valence electron band of silver. Since this repulsive interaction is proportional to the number of silver neighbor atoms it directs CO to the atop position. Because of the low energy of the d valence electrons, backdonation into the empty CO 2n* orbital results in only a weak interaction. The analogous case of CO chemisorption to Cu has been discussed extensively in ref. 5. In the Pd-Ag alloy the atoms keep their chemical identity. Notwithstanding some small changes that do occur in the valence electron band structure. According to the coherent potential approximation the average
33
position of the d electrons around Ag remains the same as in the unalloyed metal, the same holds for Pd (ref. 60). The bond strength of CO chemisorbed atop of a Pd atom changes very little upon alloying. There is however a significant change in bond strength of bridge coordinated CO. 2n*-backdonation is significantly decreased because of the low position of the silver d orbitals. The interaction between the CO 50 orbital and the silver d orbitals is repulsive. As is also the case for Cu these effects are not compensated for by increased backdonation into the CO 2n* orbitals by s-p valence electrons, notwithstanding the lower work function of Ag compared with Pd. The change in coordination on alloys of a reactive and an unreactive metal has been called the secondary ensemble effect. It can be considered a geometric effect and the description given privides its electronic basis. Calculations that confirm this picture can be found in ref. 6.
1.6 REFERENCES la lb lc Id le If lg 2a 2b 2c
S . Sung, R. Hoffmann, J. Am. Chem. Soc., 107 (1984) 2006 Ch. Zeng, R. Hoffmann, J. Am. Chem. SOC.,110 (1988) 749 N.K. Ray, A.B. Anderson, Surf. Sci., 119 (1982) 35 W. Andreoni, C.M. Varma, Phys. Rev., B25 (1981) 437 G. Doyen, G. Ertl, Surf. Sci., 69 (1977) 157 S.P. Mehandru, A.B. Anderson, Surf. Sci. (1988),201, 345 A. de Koster, R.A. van Santen, J. Vac. Sci. Techn., A6,(1988) 1128
2g 2h 3 4 5 6
D. Post, E.J. Baerends, J. Chem. Phys., 78 (1983) 5663 F. Raatz, D.R. Salahub, Surf. Sci., 176 (1986) 219 K. Hermann, P.S. Bagus, C. Nelin, Phys. Rev., B35,(1987) 9467; (7) Hermann,P.S. Bagus,C.W. Bauschlicher, Phys. Rev. B31 (1985) 6371 J.N. Allison, W.A. Goddard 111, Surf. Sci., 110 (1981) L615 G. Pacchioni, J. Koutecky, J. Phys. Chem., 91 (1987) 2658 R.A. Blomberg,C.B. Rebrilla,P.M. Siegbahn, Chem. Phys. Lett 150 (1988) 522 K. Brownfield, R.M. Lambert, Surf. Sci., 195 (1988) L156 R.V. Kasowski, T. Rhodin, M.-H. Tsai, Appl. Phys., A41,(1986) 61 R.B. Woodward, R. Hoffmann, Ang. Chem. Int. Ed., 8 (1969) 781 K. Fukui, Science, 218 (1982) 747 R.A. van Santen, J. Mol. Struct., 173 (1988) 157 R.A. van Santen, E.J. Baerends, in Theoretical Models of Chemical Bonding part 4, Ed.
7 8 9 10 11 12
Z.B. Maksic, Springer-Verlag, in press R.A. van Santen, J. Mol. Catal., 54 (1990) 288 D.L. Thorn, R. Hoffmann, J. Am. Chem. Soc., 100 (1978) 7224 R.A. van Santen, Progr. Surf. Sci., 25 (1987) 253 R.A. van Santen, J. Chem. Soc.Far. Trans. I, 83 (1987) 1915 A.B. Anderson, R.W. Grimes, S.Y. Hong, J. Phys. Chem., 91 (1987) 4245 E. Shustorovich, Surf. Sci. Rep., 6 (1980) 1; Acc. Chem. Res., 21 (1988) 183
2d 2e 2f
34
13 14 15a 15b 16 17 18 19 20 21 22
23 24 25 26 27a 27b 28 29 30 31
32 33 34 35 36 37 38
39 40
A.B. Anderson, J. Chem. Phys., 62 (1975) 1187 G. Blyholder, J. Phys. Chem., 68 (1964) 2772 H. Hopster, H. Ibach, Surf. Sci., 77 (1978) 109 W. Erley, H. Ibach, S. Lehwald, H. Wagner, Surf. Sci., 83 (1979) 585 H. Papp, Surf. Sci., 129 (1983) 205 L.H. Dubois, G.A. Somorjai, Surf. Sci., 91 (1980) 514 J.C. Bertolini, B. Tardy, Surf. Sci., 102 (1981) 131 C. Benndorf, B. Kriiger, F. Thieme, Surf. Sci., 163 (1985) L675 R.A. van Santen, J. Chem. Phys., 83 (1985) 6039 T.B. Grimley, Phil. Trans. Roy, Soc. (Lond), A318 (1986) 135 J.K. Norskov, S. Holloway, N.D. Lang, Surf. Sci., 137 (1984) 65; J.K. Norskov, Physica, 127B (1984) 193; N.D. Lang, S. Holloway, J.K. Norskov, surf. Sci., 150 (1985) 24 E. Wimmer, C.L. Fu, A.J. Freeman, Phys. Rev. Lett., 55 (1985) 2618; M.Y. Chou, J.R. Chelikowski, Phys. Rev. Lett., 59 (1987) 1737 R.A. van Santen, in Proc. 8th International Congress on Catalysis (Springer-Verlag, Berlin, 1984), Vol.IV, p 97 B.E. Nieuwenhuys, Surf. Sci., 105 (1981) 505 E.L. Garfunkel, M.H. Farias, G.A. Somorjai, J. Am. Chem. Soc., 107 (1985) 349 J-Y Saillard, R. Hoffmann, J. Am. Chem. SOC.,106 (1984) 2006 A.B. Anderson, J.A. Nichols, J. Am. Chem. S o c . , 108 (1986) 1385; A.B. Anderson, Inorg. Chem., 15 (1976) 2598 P.A. Thiel, E.D. Williams, J.T. Yates Jr., W.H. Weinberg, Surf. Sci., 91 (1980) 514 B.A. Gurney, L. Richter, J.S.Villarrubia, W.Ho, J. Chem. Phys., 87 (1987) 6710 R.A. Marbrow, R.M. Lambert, Surf. Sci., 67 (1977) 489 H. Ohtani, M.A. van Hove, G.A. Somorjai, in "The structure of Surfaces: Proceedings of the ICOSS-2, Amsterdam, 1987, ed. J.F. van der Veen and M.A. van Hove; H. Ohtani, M.A. van Hove, G.A. Somorjai,J. Chem. Phys., 92 (1988) 3974 P.A. Thiel, E.D. Williams, J.T. Yates Jr., W.H. Weinberg, Surf. Sci., 82 (1979) 22 G.B. Fisher, S.J. Schmieg, J. Vac. Sci. Technol., A1 (1983) 1064 J.M.D. Coey, Acta Cryst., B26 (1970) 1876 P.C. Wong, K.C. Hui, M.Y. Zhou, K.A.R. Mitchell, Surf. Sci., 165 (1986) L21 W. Oed, B. Dotsch, L. Hammer, K. Heinz, K. Muller, Surf. Sci., 207 (1988) 55 H.A.C.M. Hendricks, A.P.J.M. Jongenelis, B.E. Nieuwenhuys, Surf. Sci., 154 (1985) 503; H.A.C.M. Hendricks, B.E. Nieuwenhuys, Surf. Sci., 175 (1986) 185 W.F. Banholzer, P.O. Park, K.M. Mak, R.I. Masel, Surf. Sci., 128 (1983) 176; P.O. Park, W.F. Banholzer, R.I. Masel, Surf. Sci., 119 (1983) 145; P.O. Park, W.F. Banholzer, R.I. Masel, Surf. Sci., 155 (1985) 341,653 G.C.Bond, Disc. Far. Soc.,41 (1966) 200 W.H. Weinberg, R.P. Merniel, J. Catal., 40 (1975) 268
35
41 42 43
44 45 46a 46b 46c 46d 47a 47b 47c 47d 48 49 50 51a 51 52 53a 53b
J.B. Goodenough, in Magnetism and the Chemical Bond, Krieger Publishing Co., Huntington, New York, 1976 R.A. van Santen, Rec. Trav. Chim. Pays-Bas, 101 (1982) 121 0.Kahn, L. Salem, in Proc. 6th International Congress on Catalysis (The Chemical Society, London, 1977), Vol. I, p 101 P.J. Feibelman, D.R. Hamann, Phys. Rev. Lett., 52 (1984) 61; P.J. Feibelman, Phys. Rev. Lett.,24 (1985) 2627; P.J. Feibelman, D.R. Hamann, Surf. Sci., 149 (1986) 84 R.W. Joyner, J.B. Pendry, D.K. Saldrin, S.R. Tennism, Surf. Sci., 138 (1984) 84 D.W. Goodman, M. Kiskinova, Surf. Sci., 105 (1981) L265 R.J. Madix, S.B. Lee, M. Thornburg, Surf. Sci., 133 (1983) L441 J. Benziger, R.J. Madix, Surf. Sci., 94 (1980) 119 S.W. Jorgensen, R.J. Madix, Surf, Sci., 163 (1985) 19 W. Erley, H. Wagner, Surf. Sci., 74 (1978) 333 D.G. Castner, G.A. Somorjai, Surf. Sci., 83 (1979) 60 P.W. Davis, R.M. Lambert,Surf. Sci.,,lll (1981) L671 B.E. Nieuwenhuys, Surf. Sci., 126 (1983) 307 G. Broden, T.N. Rhodin, C.F. Brucker, R. Benbow, Z . Hurych, Surf. Sci., 59 (1976) 593 C. Benndorf, E. Bertel, V. Dose, W. Jacob, N. Memmel, J. Rogozik, Surf. Sci., 191 (1987) 455 J.B. Benziger, Appl. Surf, Sci., 6 (1980) 105 W.M.H. Sachtler, M. Ichikawa, J. Chem. Phys., 90 (1986) 4752 T. Mori, A. Miyamoto, N. Takahashi, M. Fukagaya, T. Hattori, T. Murakami, J. Phys. Chem.,,90 (1986) 5197 N. Koga, K. Morokuma, J. Am. Chem. Soc.,108 (1986) 6136 G. van der Lee, A.G.T.M. Bastein, J. van den Boogart, B. Scheller, H. LUO,V. Ponec, J. Chem. Soc. Far. Trans. I., 83 (1987) 2103 A.G.T.M. Bastein, W.J. van den Boogart, G. van der Lee, H. Luo, B. Schuller, V. Ponec, Appl. Catal., 29 (1987) 243
54a
N.A. Anikin, A.A. Bagatur'yants, G.M. Hidomirov, V.B. Kazanskii, Zhur. Fiz. Khim., 57 (1983) 653
54b 55 56 57 58 59 60 61
V. Bonacic-Koutecky, J. Koutecky, P. Fantucci, V. Ponec, J. Catal., 111 (1988) 409 R.C. Baetzold, J. Phys. Chem., 88 (1984) 5583 C. Minot, M.A. van Hove, G.A. Somorjai, Surf. Sci., 127 (1982) 441 P. Biloen, W.M.H. Sachtler, Adv. Catal., 30 (1981) 165 Y. Soma-Nota, W.M.H. Sachtler, J. Catal, 32 (1974) 315 R.A. van Santen, A. de Koster, T. Koster, to appear W.M.H. Sachtler, R.A. van Santen, Adv.Catal., 26,(1977) 69 R.A. van Santen, W.M.H. Sachtler, Surf. Sci., 63 (1977) 358
36
APPENDIX 1 Atomic parameters: principal quantum number (n),ionization potential (VSIP), orbital exponents (<) and respective coefficients (Ci) - d only - used. Atom
S
n
VSIP
(6)
C 0 Rh
2 2 5
20.00 28.48 8.09
1.658 2.246 2.135
Rh
4
12.50
4.290
P
n
VSIP
(5)
2 2 5
11.26 13.62 4.57
1.618 2.227 2.100
0.5685
1.97
0.5807
37
CHAPTER 2
INTERACTION OF CO WITH SINGLE CRYSTAL METAL SURFACES M. Kiskinova lnstitute of General and Inorganic Chemistry, Bulgarian Academy of Sciences, Sofia 1040, Bulgaria
38
2.1 INTRODUCTION The interaction of CO with metal surfaces is one of the most extensively studied processes. Numerous experimental and theoretical efforts have been dedicated to describe the different aspects of interactions in CO-metal systems. They still serve as typical models for testing and deepening our fundamental knowledge about phenomena such as chemisorption, dissociation, surface diffusion, surface reactions, etc. This knowledge is closely related to understanding the mechanism of the technologically important catalytic reactions, e.g. CO hydrogenation over different transition metal catalysts and synthesis of various hydrocarbon products, water gas shift reaction, catalytic oxidation of CO from the automobile exhaust gases, etc. The development of a great number of surface sensitive techniques in the last decade has supplied scientists with extremely powerful tools for studying the interactions taking place on solid surfaces at the atomic and molecular level. Most of the recent studies of the CO interactions with metal surfaces were performed under ultrahigh vacuum conditions using single crystal metal substrates. This ensures studies of well defined crystallographic planes, the composition, geomemc and electronic surface structure of which can be characterized and changed in a controllable way. The main purpose of these studies is to describe the nature of the metal surface - CO bonding and the factors contributing to the observed changes in the electronic and geometric structure of the adsorbate and the substrate surface atoms involved in the adsorption complex. Since a comprehensive review summarizing in detail the current data for CO adsorption on transition metal surfaces (ref. 1) will appear soon, the primary goal of the present review is to illustrate the power of the model surface science studies in gaining fundamental knowledge about the following properties of the CO-transition metal systems: the adsorption and desorption kinetics and energetics; (2) the degree of ordering in the CO overlayers and its dependence on the CO surface coverage and the presence of modifier adatoms; (3) the orientation of the adsorbed CO molecule with respect to the surface and its dependence on the substrate surface structure, CO coverage and the presence of foreign adatoms; (4) the CO adsorption site configuration and its dependence on the CO coverage; ( 5 ) the perturbations in the electronic structure of the CO molecule as a result of formation of the metalLC0 chemisorption bond; (6) the factors determining the CO dissociation probability on the substrate surface; (7) CO induced perturbations in the substrate electronic and geometric structure; ( 8 ) the influence of additives (acting as poisons and promoters in some important CO catalytic processes) on the CO adsorption behavior. This review will introduce the reader to the present state of knowledge about CO interaction with surfaces of metals from groups IB, VIB, VIIB and VIII, emphasizing on the results, obtained with different well defined substrate crystallographic planes. As outlined above, this is the best starting point in the attempt to describe the processes occurring at atomic and molecular level. The effect of the chemical nature and the geomemcal configuration of atomically clean or modified substrate surfaces on the CO adsorption behavior and dissociation propensity (1)
39
will be discussed considering the most recent results obtained by using modem experimental and theoretical methods.
2.2 EXPERIMENTAL TECHNIQUES For the sake of better understanding in this subsection the reader can become familiar with the sophisticated surface techniques which are nowadays commonly applied in studying surfaces and surface interactions at molecular level. Since a great number of books has already appeared where a detailed description of the surface science techniques can be found, e.g. ref. 2, here only a brief classification of the most widely used techniques will be offered.
2.2.1
DYNAMICAL METHODS
These methods involve techniques where the adsorbed species (or fractions of them) are desorbed from the surface and analyzed in the gas phase, most often by means of a mass spectrometer. The desorption of the surface species in invoked in different manners, e.g. programmed increase of the substrate surface temperature or irradiating the surface layer by electrons, photons, atoms or ions. 2.2.1.I T h e m 1 Desorvtion (TDJ This widely used method is based on measuring the flux of the desorbing species for a given adsorption system (refs. 3-7). The desorption is induced by heating the sample using an
adequate temperature ramp (temperature programmed desorption, TPD) (refs. 3-6), or pulsed laser beam (laser induced desorption, LID) (ref. 7). The desorption flux is usually recorded and analyzed by a mass spectrometer. The thermal desorption spectra provide information about the adsorbate surface coverage, the existence of different adsorption states, and the desorption characteristics such as desorption energies, pre-exponential factors and order of desorption. In order to obtain more reliable information it is necessary to avoid side effects, such as recording desorption from the sample supports, temperature gradient across the crystal, the possible laser induced damages of the surface, changes in the pumping speed, erc. (6,7).When interpreting the data, especially the multipeak TD spectra, care must be taken, because in some cases the raise of the temperature might cause interconversion between the adsorption states. Consequently, the distinguished different peaks in the TPD spectra do not necessarily reflect a coexistence of different adsorption states at the adsorption temperature. 2.2.1.2 Molecular Beam Techniaue Molecular beam experiments have attained rapid progress recently because they represent a good approach to the heterogeneously catalyzed surface reactions (refs. 8- 10). In these experiments the sample of the detector (or both) can be moved so that the molecular beams of the studied molecules or atoms can be directed and reflected elastically or inelastically from the sample surface at different angles. This ensures to vary the angle of incidence of the molecular beam, to change independently the gas and surface temperature, and analyze directly the reaction products by means of a mass spectrometer. The use of a mass spectrometer as an analyzer and lock-in phase
40
sensitive detection restricts the measurements only to the primary reaction products of interest. Additionally, by using a modulated molecular beam, via analysis of the phase shift between the input and output signal, it is possible to obtain an information about the residence time of the adsorbing species at different sample temperatures and deduce the adsorption and desorption parameters and mechanism.
Electron and Photon Stimulated DesorDtion (ESD and PSD and ElectroQ -d Desomtion Ionaneular Distribution (ESDIAD 1 This method is based on analysis of the desorbed particles (ions and neutrals) as a result of electronic excitations of the adsorbed surface species, induced by electron or photon irradiation (refs. 11-14). A variety of molecular and fragment ions and neutrals (either in ground or electronically excited states) have been detected from the adsorbed overlayers as electron stimulated products. Depending on the type of the excitation process involved, the cross section of the different ESD products can vary substantially with the changes of the adsorption state and the adsorbate coverage. This changes carry an information about the nature of the surface-adsorbate interactions and bonding. On assumption that the directions of the desorbing beams are exclusively determined by the initial state bond orientation, the angular distribution measurements of the species desorbing by electronic stimulation developed the ESDIAD method (ref. 14). This method turns out to be a sensitive probe of the chemisorption bond angles, site location, amplitudes of the soft bonding molecular vibrations parallel to the surface and their temperature and coverage dependence. 2.2.1.3
2.2.1.4
Secondan, Ion Mass SDectrometn, (SIMS) This method is based on mass analysis of the species sputtered from the surface as a result of bombardment with highly energetic particles, usually Ar ions with energies in the keV range (15). The emission of the surface particles is induced by the energy transfer from the impact ions to the substrate lattice atoms or to the adsorbates. The mass analysis of the sputtered particles gives information about the surface composition. It also carries information for the local structure using as fingerprints the kind of the fragments monitored by the mass spectrometer and the angular dependence of the ion emission. Depending on the density of the incident beam current (ranging from 1 nA cm-2 to 1 mA cm-2) this technique can be used for accurate surface and bulk analysis, providing information on a depth scale.
2.2.2
STATIC METHODS
These methods involve analyses of the stationary adsorbed phase by means of various spectroscopies based on the interactions of electrons and photons with the surface layer.
2.2.2.I Emission SaectroscoDies (ref.2 ) Depending on the kind of emission which is recorded and used for characterization of the overlayer the emission spectroscopies can be classified into two general groups:
41
(a)
Spectroscopies based on the measuring of the intensities and the kinetic energies of the electrons emitted from various electron levels of the matter under investigation - electron emission spectroscopies. Depending on the primary irradiation that causes the electron emission, these
spectroscopies can be divided as follows: (i) photoelectron spectroscopies, where the primary irradiation is electromagnetic (photons). In this class are: X-ray photoelectron spectroscopy (XPS), ultraviolet photoelectron spectroscopy (UPS), extended X-ray absorption fine structure (EXAFS), X-ray excited Auger electron spectroscopy (XAES), erc.; (ii) Electron spectroscopies, where the primary irradiation is performed by electrons. In this class is Auger electron spectroscopy (AES); (iii) Spectroscopies where the primary irradiation is performed by noble gas ions and metastable atoms. In this class is the metastable quenching spectroscopy (MQS) or penning ionization electron spectroscopy (PIES). (b) Spectroscopies based on measuring the intensities and the energies of the electromagnetic radiation emitted as a result of the interaction of slow electrons with the surface layer - (photoemission spectroscopies). In this class are the appearance potential spectroscopy (APS) and the inverse photoemission spectroscopy (IPS) (refs. 2, 16). It should be pointed out that whereas the electron emission spectroscopies carry an information exclusively for the occupied electronic states (below Fermi level) of the substrate and the occupied molecular or atomic orbitals of the adsorbates, the photoemission spectroscopies provide an information about the empty electron density of states above the Fermi level and the unoccupied adsorbate electron orbitals.
2.2.2.2 Absorution Suectroscquies (ref.2 1 These spectroscopies are based on monitoring of the inelastically scattered primary electrons (electron loss spectroscopies) or reflected electromagnetic radiation (infrared spectroscopy). Depending on the characteristic energy losses two categories of electron loss spectroscopy are distinguished. When the electron losses are caused by the excitation of vibrational modes at the surface or in the adsorbed species we are dealing with high resolution electron energy loss spectroscopy (HREELS), where the losses are only of the order of some 100 meV. H E E L S requires very high resolution and primary electron energies of the order of a few eV. Together with the infrared spectroscopy (IR), where the vibrational excitations are induced by electromagnetic irradiation, HREELS is known also as a vibrational spectroscopy. The normal electron energy loss spectroscopy (EELS) where the primary electron energies are of the order of l W e V , is successfully applied for studies of the excitations of the intraband and ilfterband transitions, plasmon and core electron excitations and one electron transitions in the adsorbates. These losses are typically observed in the range 1-50 eV.
42
2.2.2. This method is based on recording and spatial analysis of the elastically back scattered low energy (15-350 eV) primary electrons from the surface. According to Lue de Broglie equation, describing the interference phenomena in electrons scattered by a crystal, the pronounced maxima in the angular distribution of the backscattered electrons displayed on the detector reflects the periodicity of the surface and the possible variations as a result of reconstruction or fornation of ordered adsorbate super-structures. A complete structural analysis is possible applying the kinematic theory and measuring the intensity of the diffracted beam as a function of the direction and energy of the primary electron beam (ref. 17). This ensures determination of the site location of the surface species within the unit cell, and the corresponding bond lengths.
2.3 ELECTRONIC STRUCTURE OF THE FREE CO MOLECULE AND MOLECULAR ORBITAL MODEL FOR CO BONDING TO METAL SURFACES 2.3.1 FREE CO MOLECULE Free CO molecule possesses 14 electrons. Four of them are nonbonding, located in the molecular orbitals (MO) lo (0 1s) and 20 (C Is), which are close to the nuclei and have ionization potentials of 543 and 296eV, respectively (ref. 18). The remaining ten valence electrons are distributed in the following five molecular orbitals - 30,40, Iny In, and 50 with molecular orbital energies 38.4, 19.7, 16.9 and 14eV, respectively (refs. 18-20). Obviously 5 0 is the highest occupied molecular orbital (HOMO). The lowest unfilled molecular orbital (LUMO) is the 27~MO, lying at 1.5-2 eV above the vacuum level, E,. It consists of two antibonding orbitals with a larger coefficient at the carbon atom (ref. 19). Drawings of the spatial distribution of the CO molecular orbitals are given by W. Jorgensen and L. Salem (ref. 20). The bonding between C and 0 involves 30, Iny and la, molecular orbitals and consists of one 0 and two n-bonds, slightly asymmetric towards the oxygen atom. The remaining occupied molecular orbitals 4 0 and 5 0 are essentially nonbonding. 40 MO is located near the oxygen atom, whereas 5 0 MO is situated mainly on the carbon atom and is directed away from the oxygen atom. The length of the C - 0 bond in the free molecule is 2.08 bohr (1.1 A) (ref. 21). The measured dipole moment of the free CO molecule is 0.1 12 D with the oxygen atom slightly positively charged [22). The polarity of the CO molecule, which appears to be contrary to the expected from the C and 0 Pauling electronegativity values (2.5 and 3.5, resp.), is explained by the fact that the center of the 5 0 orbital is outside the intemuclear distance. As shown in some theoretical works (refs. 23, 24), the polarity and the energy positions of the 5 0 and In levels can easily change when the C - 0 distance lengthens. The absorption stretch frequency of the C-0 bond in the free CO molecule was measured to be 2143 cm-l (ref. 25). It is worth noting that the molecular orbital structure of the free CO molecule possessing low lying unoccupied 2n levels (affinity levels) makes this molecule a good basis for model theoretical studies, because similar unoccupied or partly occupied affinity levels are possessed by other molecules of importance for the heterogeneous catalysis, such as NO, N2, 0 2 , etc.
43
2.3.2 MOLECULAR ORBITAL MODEL FOR C O - METAL SURFACE BONDING The most exploited model for explaining the CO bonding on the transition metal surfaces is the synergic mechanism proposed by Blyholder for description of the metal - CO bond in transition metal carbonyls, where CO is known as a ligand in the transition metal inorganic chemistry (ref. 26). According to Blyholder's model the metal-CO bonding involves donation from the strongly directed 50 HOMO to the unoccupied or partly occupied s, p or d, metal orbitals and back donation from the metal (most often dn occupied states) to the CO 2x LUMO. As a result: (i) CO 5 0 HOMO forms a bonding orbital, which according to the simple chemical bond model should cause an increase of the 50 binding energy (BE); (ii) the mixing of the unoccupied CO 2 x levels with the metal dn electrons will form a bonding configuration with a large amount of d , character. As will be illustrated in the forthcoming sections the degree of backdonation depends on the availability of suitably oriented metal orbitals and it ensures the major contribution to the bonding (ref. 27). The most recent theoretical approaches (refs. 28, 29) have shown that electrons from the other orbitals (e.g. sp) can be backdonated to a lesser extent or hybridization with resonant unoccupied metal states might conmbute as well. Since the 5 0 HOMO is essentially nonbonding and the 2n LUMO is strongly antibonding, the formation of the chemisorption bond leads to weakening of the C - 0 bond, an increase of the C - 0 bond length and a drop of the C - 0 stretching frequency below the free molecule value. A schematic of the CO molecular orbital perturbations during the formation of the metal - CO chemisorption bond is shown in Fig. 2.1.
co
Molecular Orbitals Energy Levels
E,=O
3
EF
v
> 10
En 0)
c W
20
CO
gas phase
CO on
Metal (no interaction)
Strongly Chemisorbed CO on Transition M e t a l
Fig. 2.1 Schematics of the CO molecular orbital energy perturbations as a result of the CO interaction with a metal surface.
44
The relation of the CO chemisorption bond to the complex inorganic chemistry is possible because all experimental data provide an evidence that the chemisorption process can be regarded as a highly localized phenomenon, i.e. the chemisorbed species is strongly coupled to a finite small number of neighboring surface atoms. One of the first experimental evidence in this respect are the IR vibrational spectra (ref. 30). They revealed that the differently coordinated adsorbed CO molecules exhibit stretching frequencies in the same range as the analogical carbonyl compounds. After this introductory remarks about the CO chemisorption model our attention in the forthcoming sections will be focused to describe the different aspects of the CO interaction with various single crystal metal surfaces.
2.4 MOLECULAR CO ADSORPTION ON CLEAN SINGLE CRYSTAL METAL SURFACES 2.4.1
CO ADSORPTION PROBABIUTY AND THE MECHANISM OF CO ADSORPTION
Usually, as a characteristic of the CO adsorption probability rate serves the adsorption sticking coefficient. It reflects what fraction of the molecules hitting the substrate surface ischemisorbed. The temperature and the coverage dependence of the sticking coefficient carries information for the adsorption mechanism as well. A great number of the available surface sensitive techniques (e.g. thermal desorption, some emmision spectroscopies, such as XPS and AES, measurements of the CO induced work function changes, etc.) can be used for measuring the CO surface coverage, (Oco), and obtaining the CO adsorption uptake plots - CO coverage as a function of CO exposure. From the initial slope of these plots one
I
@CO
0.6 -
CO E X PO SuR Eftorr. s x l b )
Fig. 2.2 (a) CO uptake curves on Pd(l11) at different adsorption temperatures, T,: A - 250 K, A - 300 K; 0 - 358 K, 0 - 390K (b) The dependence of the sticking coefficient on the CO coverage, OCO,for the same T , as in (a) (from ref. 32).
can estimate the initial sticking coefficient, whereas the change of the slope with increasing O,-O reflects the coverage dependence of the sticking coefficient (ref. 31). Fig. 2.2a presents typical CO uptake plots obtained at four different adsorption temperatures, Ta for the adsorption system CO/Pd(ll 1) (ref. 32). As is evident for the chosen temperature range the initial slope of the four uptake plots remains invariant. It gives an initial sticking coefficient value, SOz0.9, independent on the adsorption temperature up to T,=400 K. Fig. 2.2b illustrates the dependence of the sticking coefficient, S , on OCO for the four different adsorption temperatures. This S(Oco) shape with invariant S up to certain CO coverages is the usual CO coverage dependence of the sticking coefficient
45
observed for the most of CO/transition metal adsorption systems. This behavior was explained in the framework of the precursor state model for CO adsorption supposing that CO trapping on the surface from the gas phase can occur into a precursor state over an already occupied site, with subsequent diffusion into an empty chemisorption site (ref. 33). The experimentally measured S(Oc0) plots in Fig. 2.2b for different T , fit satisfactorily to the relationship, based on the Kisliuk precursor kinetic model (ref. 34): S = Sd(1+kW( 1-0))
where k=PdIP, is the ratio determined by the probability for desorption of the precursor over occupied site, Pd and the probability for chemisorption from the precursor state, P,. The increase of k with increasing the adsorption temperature indicates more likely an increase of Pd i.e. a decrease
of
-
c o /Nl(111)
v)
c-
C , (hcol/molc)
5 >
A
LL
a
a
i
v
+
1.0 3.7 8.7 18.7 28.5
6-
A
A.
a
EXPOSURE (MOLECULES/CM')
the
lifetime of the precursor over an occupied site. A survey over the available data about CO adsorption kinetics on transition metal surfaces have shown that for all transition metals from groups VIB, VIIB and VIII. the initial sticking coefficient, So is rather high. It lies within the range 0.5-1 and does not show a
0.I
0
2
4
6
8
10 12
14
16 18 20 22 24 26 28
CO INCIDENT ENERGY (kcol/mole)
Fig. 2.3 (a) The CO coverage vs exposure plots for CO molecular beam energies, E l ; below and above the transition region of 4-7 kcal/mol. (b) The initial sticking coefficient as a function of the CO gas translational energy (from ref. 42).
46
substantial sensitivity to the crystallographic plane orientation (see e.g. the data bank in ref. 35. Recently, the same trend in the CO sticking coefficient behavior as a function of CO coverage and adsorption temperature was reported for CO adsorption on metals from group IB (Cu) (ref. 36). These CO adsorption behavior on the metal surfaces under consideration is believed to be determined by the high adsorption probability of the precursor state, so that the adsorption rate (sticking coefficient) changes negligibly up to temperatures near to the CO desorption temperature, as illustrated in Fig. 2.2. More accurate data for So, So(Ta) and S(Oco) dependence and the details of the CO adsorption mechanism have been obtained by means of the molecular beam studies. Most of the studied adsorption systems as CO/Pt(lll) (refs. 37,38), CO/Pt( 110) (ref. 39), COPd(111) (ref. 40), CO/Ni(lll) (refs. 41-43), and CO/Ni(llO) (ref. 44) showed S(Oc0) dependence described with a precursor state model, where. the lifetime of the CO precursor over the occupied sites (extrinsic CO precursor) is of importance. The existence of a CO precursor state over an empty site (intrinsic precursor) was questioned by most of the authors (refs. 37-39, 44), because the observed temperature dependence of the initial sticking coefficient and the modulated beam angular distribution suggested direct chemisorption on the unoccupied sites. However, the most recent accurate molecular beam studies of the CO/Ni(lll) adsorption system (refs. 41-43) have shown substantial changes of the CO adsorption probability (as reflected by the initial sticking coefficient) on the empty sites when the translational energies of the incident CO molecules, Ei exceeded 7 kcal/mol (see Figs 3a, 3b). This was explained assuming the coexistence of two channels for chemisorption on an empty site: (i) direct chemisorption from the gas phase with an activation barrier of the order of 4-7 kcal/mol, and (ii) CO intrinsic precursor channel which ensures nonactivated pathway to chemisorption. Consequently, the low energy molecules (Eiless than 4-7 kcal/mol) chemisorb exclusively via a precursor along a low activation energy pathway, whereas the high energy CO molecules can chemisorb directly from the gas phase (Fig. 2.4). The reduced adsorption probability of the high energy molecules displayed in Fig. 2.3 is related to the decreased adsorption probability into the precursor state, because the direct chemisorption requires correctly oriented incident molecules in order to be chemisorbed (ref. 42). It has been estimated that the lifetime of the intrinsic CO precursor is between and s at 300 K (refs. 42,43), the desorption energy of the precursor - 6-10 kcal/mol aid the activation energy for conversion form the precursor to the chemisorption state - 0.2-0.3 kcaVmol (ref. 43). Direct detection of the trapped physisorbed CO which is believed to be the precursor for the molecular chemisorption has been reported for CO on Cu films, where the UPS spectra showed that the 50, l x and 4 0 CO molecular orbitals of the physisorbed state have almost the same energy separation as the free CO molecule (ref. 45). This indicates that the precursor state does exist and it is different from the chemisorbed state. Considering the Boltzman energy distribution (=1 kcal/mol) of the gas molecules under the usual conditions of the catalytic reactions (temperatures in the order of 300-800 K) it can be concluded that in these systems CO adsorption occurs via a precursor and is a nonactivated process.
47
)r
F
direct chemisorption
a,
c W
6-\$ I
E O .-
X
. I -
c
a, 0
. I -
a
I
W
Fig. 2.4 Schematics of the potential energy diagram for a direct and precursor mediated adsorption.
2.4.2
C O ADSORPTION BINDING ENERGY AND MOBILITY OF THE CO MOLECULE IN THE ADSORPTION PHASE
The description of the CO chemisorption state usually starts with evaluation of the adsorption binding energy which is a measure of the strength of the interactions between the adsorbed CO molecule and the substrate surface atoms. Since the adsorption kinetic results have shown that the molecular CO adsorption is a nonactivated process, the heats of adsorption AH, or the activation energies for CO desorption, Ed, can be used as a direct measure of the CO adsorption binding energy, Eb CO adsorption isobars and isosteres are very commonly used for the evaluation of AHc-,. This method requires a series of plots of the CO coverage versus: (i) the adsorption temperature, T,, at constant CO pressure (isobars); or (ii) the CO pressure at constant temperature (isosteres). From these plots the isosteric heat of adsorption can be evaluated using the Clausius-Clapeyron equation (ref. 46): dlnpld( 1/T) = AHIk
(2.2)
Most important for obtaining correct AHco values is the choice of a suitable and precise surface analytical method for determination of the CO coverage. In the case of the COImetal adsorption systems, the most often used methods for measuring the relative CO coverage are: (i) thermal programmed desorption, where the area under the CO TPD spectra is proportional to the CO coverage; (ii) laser induced thermal desorption, where the laser induced thermal desorption signal is proportional to the CO coverage; (iii) X-ray photoelectron spectroscopy, where the intensity of the XPS 0 Is or C 1s core peaks is proportional to the CO coverage; (iv) CO induced
48
Eeff or
100
t
it 1
0.l
0.2 Q3
0,4 0.5 0,s
8
Fig. 2.5 The dependence of the CO adsorption binding energy, E ,on Ru(0oOl) as a function of the CO coverage, Oco, obtained using different methods: circyes - isothermal desorption; diamonds - A@ - TPD evaluated via desorption isosteres; triangles and squares - from the peak temperatures in A@ and AP - TPD (from ref. 47).
work function changes, A@, when it varies linearly with Oco, erc. In order to obtain the absolute CO coverage, usually given as a number of CO molecules per surface atom (ML), one can use the LEED data, when ordered CO overlayers are formed, or calibrate the intensities of the TD or XPS spectra with that for adsorbates with known surface concentrations. Thermal desorption data, on the other hand, can be used directly for determination of the adsorption binding energy, E,, because, as outlined above, for nonactivated adsorption the desorption energy, Ed, is equal to the adsorption binding energy, E, (refs. 5-7, 31). Most often Ed is determined by analysis of the TPD peak temperature (ref. 4), assuming invariant preexponential factor of desorption, vd However, as discussed in detail in refs. 47,48, both Ed and vd are usually coverage dependent, so that more accurate methods are required, such as fitting by selection ECvd pairs to the experimental curves, or constructing ln(O/dO/dr) versus 0 curves with l/T as a parameter (refs. 5, 6, 49). Examples for the observed dependences of the CO adsorption binding energy (desorption energy) on the CO coverage for CO/Ru(OOOl) (ref. 47) and CO on flat and stepped Pt(ll1) (ref. 50), measured by the applying of different methods are shown in Figs. 5 and 6. The values of the reported CO adsorption binding energies for the molecular CO state on the group VIII transition metal surfaces in the limit of low CO coverages (initial heats of adsorption, AH'co, are ranging from 30 to 40 kcal/mol. As illustrated in Fig. 2.6 for the same substrate the defect (e.g. stepped) surfaces exhibit higher CO adsorption binding energies. In the case of metals from groups VIB and VIIB the picture is more complicated because one should discriminate between molecular and dissociated CO coexisting on the surface. The TPD data, however, have shown that from the
49
0.0
0.2
0.4
0.6
0.8
1.o
1
e/%lax
Fig. 2.6 The dependence of the CO desorption energy, E on the CO coverage for Pt(l1 l), Pt(557) and Pt( 112) (from r e t 50).
VIB and VIIB metal surfaces the nondissociated CO desorbs at the same temperature range as CO
from the group VIII metals, indicating that the molecular CO adsorption binding energies are similar (refs. 1,35). The metals from group IB exhibit a completely different behavior showing a rather weakly bound molecular CO state with Eb of the order of 6 kcaVmol for Ag (ref. 51) and 10-14 kcal/mol for the different Cu planes (refs. 1,36). An attempt to correlate the molecular CO adsorption binding energies (determined from the CO 'TPD peak maxima) and the corresponding energetic position of the CO 2x molecular orbitral (measured by inverse photoemission spectroscopy) was made in ref. 16. It was assumed that if the metal/2x backdonation is the major contributor to the metal-CO bonding, than the lowering of the CO 271 level below the vacum level, AEK (used as a measure of the backdonation) should be linearly proportional to the corresponding CO adsorption binding energies. Fig. 2.7 shows the plot A E 2 K versus the temperature of the CO TPD maximum taken from ref 16. Indeed, there is a satisfactorily agreement between the measured magnitude of A E 2 c E 2 z - E V and the corresponding desorption energies if one compares CO/Cu with the CO/transition metal systems. Close inspection of the plot in Fig. 2.7, however, shows that the suggested existence of a simple linear relationship between Ed and AE2n should be taken with reservation, because even the data for the chosen transition metals are rather scattered. One possible reason for that is the fact that the 2x energy position depends not only on the degree of overlapping with the suitable metal orbitals (backbonding), but also on the site potential (determined by the work function of the substrate) as is shown in Fig. 2.1. In addition, when describing metal-CO adsorption system, besides the metalKO 2n backdonation, the CO %/metal donor bond should be considered as well. The latter
50
I
I
I
I
I
500
600
-
0
r
n
w
* 100
200
300
400
Temperature I K
Fig. 2.7 The energetic position of the CO 2n level relative to the vacuum level AE e ( E 2 - Ev), as a function of the CO desorption peak temperature for various metal surfaces (gom re? 16).
might contribute to a different extent to the metal - CO bonding on the different metals, depending on the particular surface band structure and the adsorption site coordination. More discussions concerning the 2n-energy level will be given in Section 4.7. Generally the dependence of the CO adsorption energy on OCO shows a decrease of the
Fig. 2.8 AP - TPD traces for various CO initial coverages on Ru(0001): (a) 0.66 (b) 0.63 (c) 0.58; (d) 0.54 (e) 0.49 (f) 0.45; (g) 0.43; (h) 0.38 (from ref. 47).
51
adsorption binding energy at high CO coverages. This behaviour is attributed to arisement of repulsive interactions between the adsorbed CO molecules when the intramolecular distance is reduced. The E&o)
dependence is determined by several factors, e.g. the actual structure of the
co overlayers, the CO bonding configurations and their changes with increasing CO. This justifies the variety of the Eb(@C,) shapes observed for the different CO/metal adsorption systems. For example, the unusual rise of E b at Qcof0.33 ML on Ru(0001) in Fig. 2.5 is due to the arisement of attractive interactions between the next near neighbour CO molecules, leading to the formation of an ordered (./3xd3)R3Oo structure. Further increase of Oco causes the usual Eb fall, because of the destruction of the 43 order at higher Oc0 as a result of the stronger CO-CO repulsive interactions at higher coverages. As illustrated in Figs. 8-10 the increase of the CO coverage affects substantially the shape and the maxima positions of the CO TPD curves. Usually, the decrease of Eb is reflected by the appearance of lower temperature TD peaks or shift of the peak maxima to lower temperatures. As outlined above and will be described in detail in the forthcoming sections, the variety of the CO coverage induced changes in the TPD spectra can be also associated with the occupation of different adsorption sites on the different substrate surfaces and crystallographic planes, and formation of various ordered structures characterized by different interadsorbate separations. The second important characteristic of the CO adsorption state is the preexponential factor of desorption, vd. According to the statistical thermodynamics, it carries information about the mobility of the adsorbed species on the surface (ref. 52). For the case of nonactivated co adsorption in the limit of systems where measurements of vd are reported, e.g. k ( l 1 I), Pd(l11), Ni(l1 l), Ni(100) (refs. 35,481,Ru(0001) (ref. 47), Pt(ll1) (refs. 48, 53), the numerical values of s-’ vd range between 1014 to 1019 s-l, i.e. exceed by more than an order of magnitude the value derived from the statistical mechanic equation for a mobile two dimentional ideal gas
I “R
2
Fig. 2.9 CO TPD spectra from Pd(ll1) for different adsorption temperatures, T,; (a) 250 K; (b) 300 K; (c) 390 K. CO exposure in s torr 10-6(from ref. 42).
52
(ref. 52). This high v d values are consistent with the case of a localized adsorption state, which means a strongly reduced mobility of the CO molecule in the adsorbed phase. As discussed in delail in refs. 47,48, 53 the only vibrational mode of the adsorbed CO that contributed to v d is the frustrated translation parallel to the surface. Consequently, for the case of nonactivated CO adsorption in the limit of low Oco when the sticking coefficient is of the order of unity, the preexponential factor can be described by the relationship:
(2.3)
whereft andf, are the partition functions for the translational and rotational degrees of freedom of the gas molecule, fv is the partition function for the frustrated translational vibration of the chemisorbed molecule and Ns is the number of the substrate surface atoms. Using the gas phase values for ft and f, at 300 K and Np1.6. 1015 atoms for the densely packed fcc crystallographic planes, the calculated value of v d is 8.1Ol6/f, (ref. 53). There are few experimental data I A Pt (112) where the stretch frequencies of the CO molecule frustrated translational modes on transition metal surfaces are measured, but they are usually in the range 50-100cm-' (refs. 54, 55). This gives fv values of the order of a few tenths. Obviously, the fv contribution is small and usually the CO v d values remain considerably larger than 10-13s-'. As in the
250
500
550
400
450
500
550
TEMPERATURE (K)
Fig. 2.10 CO TPD spectra from flat and stepped Pt(ll1) surfaces. Ta=220 K (from ref. 50).
I
case of E,, the same reasons-arisement of repulsive or attractive interactions between the adsorbate molecules are responsible for the observed coverage dependence of the preexponential factores (Figs. 11, 12). The drop of the sticking coefficient below unity above certain CO coverages should be condiered also (refs. 47,48b). Usually, the preexponential factor changes in the same direction as the adsorption binding energy (compare Figs. 5, 6, 11 and 12). The adsorbate is expected to become more mobile when the degree of coupling with the surface reflected by Eb decreases, but as is discussed in ref. 48b there are some exceptions, e.g. CO/Pd( 100) and Pd( 111) (refs. 56,57), where
53 Vd
remains invariant with
ec0,and CO/Ni(lOO) (ref. 58), where vd rises and Ed decreases with
QCO At the present state of knowledge there is no complete theory which can explain satisfactorily
the varieties of the vd Oc0 behavior because numerous factors can contribute to a different extent, e.g. metal-CO bonding strength, CO surface structure, CO-CO lateral interactions, changes in the CO surface mobility in the different adsorption sites, CO induced substrate surface reconstruction or perturbations in the surface phonon spectra, etc.
2.4.3 SURFACE STRUCTURE OF THE CO OVERLAYERS LEED data have shown that with exception of the weakly bound CO on Ag and Au, the CO overlayers form ordered structures on all other single crystal metal surfaces under consideration. The data available for CO adsorption on the low index single crystal planes (refs. 59-81) show that first always ordered structures related to the unit cell of the surface lattice (in registry structures) are detected. They can be summarized as follows: (i) For crystallographic planes with a hexagonal lattice such as hpc(OOO1) Ru (ref. 60),Co (ref. 61) and 0 s (ref. 62) and fcc(ll1) Pt (ref. 63), Ni (ref. 64), Pd (ref. 65), Rh (refs. 66, 67), Ir (ref. 68), Cu (refs. 36, 69) and others (ref. 59), (d3xd3)R3Oo LEED structure, corresponding to @~0=0.33h4L and ~(4x2)LEED structure, corresponding to Oco=0.5 ML have been reported (Figs. 13, 14); for the crystallographic planes with a square (Fig. 2.15) lattice such as (100) Pd (refs. 70,72), Pt (ref. 71), Ni (refs. 58,73-74), Cu (refs. 75-77), Rh (ref. 78), and Fe (ref. 79) - ~(2x2)LEED pattern, corresponding to @c0=0.5is observed. Different c(nx2) LEED structures are observed on thefcc(ll0) surfaces (ref. 59) but the picture is sometimes complicated because of possible CO induced reconstruction of the substrate surface. More information about the different ordering in the CO overlayers can be found in
(s-1 1 1020A
I
0.1
m
I
0,3
Q5
0
Fig. 2.1 1 The dependence of the preexponential factor, vd, for CO desorption from Ru(0001) on the CO coverage, 0 ~ 0obtained , by the same methods as described in Fig. 2.5 (from ref. 7).
54
t
Pt(112)
-I
Fig. 2.12 Dependence of vd on the CO coverage on flat and stepped Pt( 111) (from ref. 50).
refs. 59,72, 80, 81. The orientation of the CO molecules building (d3x43)R3O0, ~ ( 4 x 2 and ) 42x2) ordered overlayers is always with the C-0 axis normal to the surface, as will be discussed in the forthcoming section. As is obvious from Figs. 13-15 the CO site occupation varies for the different substrates and
Ru ( O W )
b
C
Fig. 2.13 CO adsorption structural models on: (1) Ru(0001) (a) (d3xd3)R30 superlattice; (b) arid (c) (243x243) R30° and (43x4) rect superlattices interpreted with a compact hexagonal model; (2) Rh(ll1) - (a) (d3xd3)R3O0; (b) (2x2); (c) as (b) with relaxations. (3) Ni(ll1) - (a) (d3xd3)R30° (b) ~(4x2);(c) (47xd7)Rl9.l0 (from ref. 81).
55
High coverage CO overlayers on Pt(ll1) surface
Model structure for
e=o.ee
Fig. 2.14 The structural model for ~ ( 4 x 2 and ) compressed CO layers on Pt(ll1) and schematics of the tilted terminal CO molecules building the edges of the fault lines (from ref. 85). different surface orders. At higher CO coverages, exceeding 0.5 ML, more complicated out of registry LEED patterns are usually detected. For explaining these complicated structures various geometric models have been proposed, including tilting of the CO molecule and/or shifting of CO out of its equilibrium high symmetry adsorption site (refs. 59,77, 80-84). For the fcc(ll1) and hpc(OOO1) planes (refs. 59, 81) and the fcc(100) planes (refs. 59,72,80), a variety of complex , hexagonal structures are observed at , @ exceeding 0.5 (see Figs. 13-15). Two different interpretations of these structures have been proposed. The first assumes an existence of incommensurate CO overlayers described by a
compressed
hexagonal
lattice.
This
Fig. 2.15 CO/Pt(lOO): (a) ~ ( 2 x 2 structure ) with bridge site adsorption; (b) p(342x42)R45O with top and bridge site adsorption; (c) and (d) ~ ( 4 x 2 )structure with top and bridge sites; (d) and (f) same as (c) and (d) after relaxations leading to compact models (ref. 80).
56
-
VI a C,
C -
0
11101
30
20
10
0
10
20
Polor angle [degrees;
30
t I-
11001
Fig. 2.16 (a) ESDIAD of O+ from (2x1) CO/Ni(llO) obtained with channel plate array (top) and with mass spectrometer (bottom) Ep=lOOO eV, T=100 K (from ref. 83) (b) Structural model for (2x1) p2mg CO/Ni(llO) (top and side view).
interpretation predicts occupation of many kinds of adsorption sites, which is in contradiction with the vibrational spectroscopy data, where no more than two kinds of CO coexisting surface configurations were detected. The second model assumes that the hexagonal structures consist of ordered domains composed of simple superstructures such as (d3xd3)R3Oo and ~ ( 4 x 2 )for the hpc(O001) and fcc(l11) planes and 42x2) and p(.\/2~2xd2~2)R45' for the fcc(100) surfaces. In refs. 72, 80, 81, 84 M. Van Hove et al., considering the available information about the CO bonding configuration (evidenced by the CO vibrational spectra), have concluded that the observed hexagonal LEED patterns at high Oco consist of regular antiphase domains, where CO molecules occupy well defined sites - mostly on top and/or bridge configuration. A slight shift off equilibrium position and tilting no more than 10' is supposed in some cases (refs. 83,84) and confirmed by the most recent ESDIAD studies (ref. 85). On more open crystallographic planes such as Ni(ll0) and Pt(ll0) the 42x2) CO superstructure at Oc0=0.5 ML is replaced by a series of LEED patterns ending with (2xl)plgl structure at the maximum CO coverages achieved at Oco=l (refs. 82-84,86-88). AS will be demonstrated in the forthcoming section on these planes at Oco exceeding 0.75 the CO molecules are tilted away from the normal in the plane perpendicular to the rows of the substrate atoms (Fig. 2.16). Obviously in these crowded CO overlayers the distances between the chemisorbed CO molecules are much smaller than that expected when the van der Waals diameter of the CO molecule is considered - 1.7 8, (refs. 85). As is outlined in refs. 80, 81, 84, 85 the distances between the CO molecules at high Oco in the crowded overlayers become of the order of the
57
substrate surface lattice constant, and turns out to be comparable with the CO-CO distances measured for some metal carbonyls (ref. 89).
2.4.4 ORIENTATION OF THE CHEMISORBED CO MOLECULE WITH RESPECT TO THE SUBSTRATE SURFACE AND THE CORRESPONDING SURFACE-C AND C - 0 INTERACTION DISTANCES All available data obtained by various surface sensitive techniques, such as near edge X-ray absorption fine structure (NEXAFS) (refs. 90-92), X-ray photoelectron diffraction (XPD) (refs. 93-96), angle resolved ultraviolet photoelectron spectroscopy (ARUPS) (refs. 65,66), Electron Stimulated Desorption Ion Angular Distribution (ESDIAD) (refs. 82, 83,85,99-102) and Angle Resolved Auger Electron Spectroscopy (ARAES) (ref. 103) have shown that on the most closed packed metal surfaces up to moderate O,, the CO molecule is adsorbed via the C atom with a C - 0 axis normal to the surface. Existence of off-normal bonded CO molecules has been observed: i) on more open crystallographic planes, e.g. Mo(100), W(111) (ref. 104), (lx2)Pt(llO) (ref. 105), W(100) (ref. 106), Ni(l10) (refs. 82, 83,96, 107) where tilting angles greater than 10" are measured at high CO coverages when compressed ordered structures are developed. Fig. 2.16 presents the ESDIAD data for CO on Ni( 110) for the (2x2)plgl compressed overlayer with CO tilting angles of the order of 19" in (001) direction; ii) stepped surfaces, e.g. Pd(210) (ref. 108) and stepped W(111) (ref. log), where tilt angles 18" are reported; iii) from compressed CO overlayers on densely packed substrate surfaces, e.g. Ru(0001) (ref. 83) and Pt(ll1) (ref. 85), where the tilt angles are less than 10". Fig. 2.17 illustrates a development of tilted CO species on Pt(111) observed at Oco>0.5 ML when the crowded fault lines (Fig. 2.14) composed of terminal CO species are built and separate the ~ ( 4 x 2 )domain structures. The CO tilt angle measured from the polar angles of the six off-normal ESDIAD beams maxima is of the order of 6" and the azimuthal angle of the ESDIAD beams indicates that the CO molecules are tilted in equivalent (1 10) plane directions. Recently, it has been reported that CO exhibits a unique behaviour on the closed packed Cr(ll0) (refs.110, 111) and theopenFe(100) (refs.112-114) andFe(ll1) (refs.115, 116) surfacesat low CO coverages, expressed by a lack of any ESD emission and unusually low C - 0 stretching frequencies (below 1350 cm-l). These results are unexplainable in the framework of the donoracceptor Blyholder mechanism of bonding. For the Fe( 111) surfaces some authors (ref. 116) proposed a model where the CO molecule is located (partially imbedded) in four-fold symmetric hollow sites. In this position possible interactions of the nearby Fe atoms with the CO Iny 1xZ molecular orbitals might account for the strong reduction of the C - 0 stretching frequency and the reduced cross section of ESD desorption. NEXAFS data for CO on Fe(100) (ref. 114) have shown that the low coverage CO adsorption state is strongly inclined which facilitates the involvement of In and 40 molecular orbitals in the metal-CO bonding. For the closed packed Cr(l10) surface even CO lying down with both C and 0 ends coordinated to the metal atoms is supposed
58
(refs. 110, 111). Such strongly inclined or, "lying down" K and 4 0 bonding configuration is possible in the metal carbonyls (ref. 117) and it is supposed to precede CO dissociation on the metal surfaces (refs. 115, 118). The existence of this non-Blyholder type CO bonding configuration depends not only on the nature of the substrate, but also on the smcture of the crystallographic plane. For example no similar CO adsorption state was found on the closed packed Fe(ll0) plane. M-C interatomic distances, dM-c, associated with the molecular CO adsorption state in normal configuration, were measured to be within =18, of the single bond length, as predicted by the Shomaker-Stevenson equation (ref. 119) and lies in the range 1.7-1.9 8, for the different substrate metals. According to the recent experimental results the most favorable value for the C-0 bond length, 4-0, is 1.15fl 8, (ref. 74). This value is within the dc, bond values ranging from 1.14 to 1.21 8, measured for the metal carbonyls, where CO is bound via the C atom (ref. 120).
CO" ESDIAD SIGNAL FROM CO/Pt(lll) SURFACE
b.
8 =0.50
h.
8 =0.66
Fig. 2.17 CO ESDIAD patterns - three dimensional and contour (azimuthal) display from CO/Pt(111) at T=90 K (from ref. 85).
59
2.4.5 CO INDUCED WORK FUNCTION CHANGES AND THE EFFECTIVE CHARGE TRANSFER DURING FORMATION OF THE SURFACE-CO BOND The work function changes, A@, induced upon adsorption are very often used as a measure for the degree of the electron charge transfer between the substrate and the adsorbate. The amount of the charge transfer is usually obtained using the simple relationship A@ = 4 nod
(2.4)
where e is the electronic charge and d is the dipole length (the component of the bond length perpendicular to the surface). (r is the surface charge density, related to the adsorbate surface concentration, N,, as follows: o=pONa, where p 0 is the dipole moment of the CO species. p(0) changes with 0 ~ mainly 0 because of dipole depolarization effects (ref. 121). The available data show that with exception of Pt( 11l), Pt( 100) and Pt(l1 l), CO adsorption always causes positive changes (an increase) of the work function on the metals of groups VIB, VIIB and
1.5
- 1.0 ->, 8-
a
0.5
n
v
0
f 0.2
8,,
0.4
0.6
( C O / Ni
Fig. 2.18 Work function changes, A@, during CO adsorption on Ni(ll1) at 288 and 90 K. The open circles show the initial and final A@ values obtained upon cooling the surface from 288 to 90 K at constant 0 ~ (from 0 ref. 180).
60
VIII. Comparing the results for the metals from group VIII (ref. 122) it becomes obvious that with exception of Pt and Ir CO adsorption on all metals leads to maximum A@ values of the order of 1-1.5 eV. The CO induced work function changes on Ir are also positive but rather small ranging between 0.18 and 0.23 eV on the different Ir planes (ref. 122). CO adsorption on Pt always causes negative work function changes (refs. 122-124), the Oco plots passing through a minimum at =-0.15 - -0.2 eV. Figs. 18 and 19 illustrate two extremes of the CO induced work function changes measured for Ni(ll1) and reconstructed and nonreconstructed Pt(ll0). On Ni(ll1) the positive A@ value reaches ~ 1 . eV 3 at high Oco, the terminal and bridge CO contributing to a different extent to A@. The reversible temperature dependence reflects the reversible bridge - terminal site interconversion.
In the case of Pt(l10) (as well as for Pt(ll1) (ref. 123)) the negative work function change is associated with terminal CO and the positive with bridge CO adsorption. The turning point is close to the onset of the bridge site population as evidenced by the vibrational and photoelectron spectroscopies data. The various breaking points in the A@(Oco) plots of the COPt(l10) system reflect the different type of terminal sites on the (110) rows and the (111) facets building the reconstructed Pt(ll0)-(1x2) surface. The negligible rise of A@ after the turning point of Pt(llO)(lxl) is associated with the different type of bridge CO on the (1 10) atomic rows, compared to that on flat (111) plane. These two figures are a good example of how in many cases the work function changes can be related to the CO site configuration on the surface. All metals from Group IB exhibit negative work function changes upon CO adsorption (refs. 1, 122). It should be pointed out that the A@(Oco) plots for COPt and CO/Cu adsorption systems are very similar, the absolute maximum work function change for the latter being somewhat larger, =-0.4 eV. Assuming that the A@ sign reflects the direction of the net charge transfer, it was supposed that by measuring A@ values one can deduce the extent of the metal/C02, backdonation (ref. 125). Consequently, on surfaces where CO is bonded with its molecular axis perpendicular to the surface, the effective dipole moment is expected to be equal to the bonding dipole moment. Since the experimental data (e.g. see the slope of the Oco plots in Fig. 2.18) show that due to depolarization effects or change in the site occupation, the charge transfer per molecule changes with increasing OCO, the initial dipole moment LO, measured in the limit of very low OC, is considered as the most relevant with respect to comparing the metal to CO charge transfer. Following this considerations the experimental A@(@,,) data suppose that the degree of backdonation for the CO/Ir and CO/Pt adsorption systems is considerably smaller and even reverse than for the other metals from group VIII (ref. 125). On the other hand, the measured CO adsorption binding energies are very similar for Pt, Ir and the other group VIII metals. This contradicts the concept that the backdonation is the main contributor to the M-CO bond strength. A possible explanations are either larger contribution of the %/metal component or lack of a direct relationship between the measured A@ values and the bonding dipole moment. The second explanation seems more reasonable, because the measured A@ values depend on two contributions: (i) A@ external which is due to the induced bonding dipole and is determined by the substrate-adsorbate charge transfer and (ii) A@ internal - an extra term due to the internal
61
Fig. 2.19 Work function changes due to CO adsorption on (a) Pt(llO)-(lx2) at 120 K, (b) Pt(ll0)-(1x1) at 120 K and (c) Pt(ll0)-(1x2) at 300 K. The LEED patterns at the maximum CO coverages are: (a) ~(8x4);(b) (2xl)plgl and (c0 disordered (1x1) (from ref. 124).
polarization dipole moment (ref. 126). As is discussed in refs. 126, 127 the second term comes from the fact that all adsorbates cause surface polarization changes leading always to a decrease of the metal surface work function. Consequently, in the case of CO, which behaves as an acceptor these two terms doextand AQint have opposite signs, i.e. AQDext>O, AQint
62
2.4.6 INFLUENCE OF THE METAL-CO BONDING ON THE CO ELECTRON CORE AND VALENCE LEVEL BINDING ENERGIES Generally, when chemisorption occurs the adsorbate electronic levels suffer energy shifts, the magnitude of which is determined by their spatial distributions, the substrate nature and the extent of their involvement in the formation of the substrate-adsorbate bonding. The electronic characteristics of the adsorbed molecules are usually determined using photoelectron spectroscopies (UPS/XPS) for measuring the binding energies (B.E.) of the electrons in the molecular orbitals and comparing with the corresponding values for the free molecule. From the observed electron binding energy shifts one can judge to what extent the different molecular orbitals are involved in the formation of the chemisorption bond. The influence of the metal-CO bonding on the energy and space positions of the filled CO molecular orbitals have been already measured by means of ARUPS, where CO/metal adsorption systems have been extensively used as test systems. Since, according to the Blyholder model summarized in section 2.3.2., only 5 0 HOMO is directly involved in the bonding, the remaining lower
MO-EV, 20 'CT
lying
(ev) 8
14 ' 7 ~
molecular
orbitals
are
expected to be less perturbed by
CO gas phase
the chemisorption. Using p and s polarized light and angle resolved detection the energy positions of the CO molecular orbitals are measured first for the system CO/Ni(lOO) (ref. 128). Later,
I
CO/Ni(ioo)
:% 12 8
4
O=E,
Fig. 2.20 ARUPS data for CO adsorbed on Ni( 100) measurements are performed with p and s-polarized radiation in order to resolve In and 50 molecular levels (from ref. 98).
considering the symmetry rules, the geometric configurations and the energy positions of the CO molecular orbitals are determined for most of the metals under consideration (ref. 122). Fig. 2.20 shows the ARUPS spectra for CO in gas (ref. 129) and chemisorbed state on Ni(100) (refs. 97, 98, 130). As is obvious the hybridization with the substrate surface orbitals causes change in the ordering of the CO molecular orbitals. The energy positions of the 50 and In MOs is reversed, because the 5 0 MO is
63
Table 2.1
CO MO binding energy (in eV) referenced to the Fermi level
ni(100)(5.22) ni(111)(5.35) ni(l10)(5.04) pd(l11) pt(ll1)(5.93) Pt(110) ir(ll1)(5.76) ir(100) co(oO01) rh( 111) fe(110)(5.05) cu( 1OO)(4.59) al(ll1)(4.24)
f e5o
el2
e4of
8.3
7.8 7.1 7.7 7.3 8.6 8.2 8.6 8.6 7.7 7.8 6.9 8.6 11.9
10.8 10.7 10.8 10.7 12.0 11.7 11.7 11.7 10.7 11.2 11.0 11.8 14.6
8.0 8.0 7.8 9.6 9.2 9.2 9.2 7.9 8.3 8.1 8.2 8.9 14
6e 3.0 3.6 3.1 3.4 3.4 3.1 3.1 3.1
3.0 3.3 4.1 3.2 2.7
r
ref.
1.2 1.33 1.11 1.17 1.4 1.24 1.24 1.24 1.07 1.13 1.4 0.84 0.47
97 137 98 138 139 140 141 141 142 143 144 131 145
16.9
stabilized with respect to much less affected 1x and 4 0 orbitals, not directly involved in the formation of the chemisorption bond. Very similar energy order for 17c, 5 0 and 4 0 CO MOs has been measured for the other metals from groups VIB, VIIB and VIII, where ARUPS studies were performed (ref. 122). It should be reminded that one of the disadvantages of the electron spectroscopies is the influence of the initial and final state effects on the measured binding energies values. This is the reason for the different energy shifts experienced by the CO MOs not involved directly in the bonding. As discussed in details in ref. 98 the initial state shifts for these CO orbitals are similar, but the final state (relaxation) shifts are different and depend on the extent of the screening effects experiences by the particular CO orbital. For the weakly bound CO on Cu (refs. 131, 132), as well for the other CO/ (group IB metals adsorption systems (ref. 133), the ordering of the CO molecular orbitals remains the same as for the free gas molecule. This indicates smaller energy shift experienced by the 50 orbital, i.e. reduced 50) metal coupling compared to the other transition metals under consideration. Possible explanation for these CO adsorption behavior on the metals from group IB is proposed recently by Messmer et al. (refs. 134, 135). It was supposed that the 5 0 backdonation requires the presence of an adequate empty metal state, not available either in the d or s levels of the IB metals with configuration dl0s. Consequently, in the framework of the synergic bonding model the CO bonding for these metals should be weak and the 50 level will suffer a smaller energy shift. It is worth to mention that in the case of a simple sp metal, such as Al, the presence of a large number metal valence electrons leads even to a strong repulsion between the superimposed 5 0 CO and A1
64
valence electrons (ref. 136). As a result only physisorbed CO is observed on A1 with CO molecular orbital energy positions almost the same as for the free gas molecule. Table 2.1 presents some data for the energy positions of the CO molecular orbitals for the free molecule and CO adsorbed on different metal surfaces. Considering the CO valence levels energy shifts as a result of the So 2n synergic bonding, it can be expected that the o/metal coupling will lead to a decrease of the So-4 0 energy separation, whereas the metaV2n coupling should cause an increase of the 40-1n separation due to the elongation of the C-0 bond. There are several attempts to correlate the observed changes in the energy separation between the CO molecular orbitals with the CO dissociation propensity (ref. 146) or the strength of the metal-CO bonding (ref. 98). First Broden et al. (ref. 146) proposed to use the energy difference hE=E(4o)-E(ln) as a measure of the C-0 bond weakening and CO dissociation propensity. However, the results obtained during the last decade do not support such a simple relationship. Considering the AE values in Table 2.1 we can immediately give an example for the lack of correlation AE-dissociation propensity comparing the data for Ni, Ir and Cu. It is well known that the CO dissociation probability on Ni and Ir is much higher than on Cu but the AE values for the three metals turn out to be the same. Another correlation was proposed by Plummer et al. (ref. 98). as an indication for the metal-CO bond They used the values of the ratio r=AE(e-lxjAE(40-5G) strength, Eb It turned out that for a couple of metals (e.g. Cu and transition metals in Table 2.1) the correlation requirement lower r - lower Eb is satisfied (ref. 98). Although still there are not enough ARUPS - Eb data in order to make a good statistic, this correlation seems more plausible, because as will be shown in section 4.9. the dissociation probability is not directly related to the strength of the metal-CO bonding. At high CO coverages the molecular valence level broadens as a result of the CO-CO interactions (ref. 147). Since in most cases the CO axis deviates from normal only at the highest CO coverages (see section I1 4. 4.) the highest surface dispersion was observed for the In CO states (ref. 147).
2.4.7 INFLUENCE OF THE METAL-CO BONDING ON THE ENERGY POSITION OF THE UNFILLED 2 p CO AFFINITY LEVELS. BONDING AND ANTIBONDING STATES AS A RESULT OF THE METALI2p C O COUPLJNG As pointed out in section 2.3.2 metal to CO backdonation has a larger contribution to the bonding than CO to metal donation. The extent of the backdonation is different for the different metals and when it is lacking (e.g. CO/AI (ref. 136)) only a very weakly bound physisorbed CO state exists. That is why one should expect a considerable variation in the 2n affinity level of the chemisorbed CO with the different metals and crystallographic planes. A number of attempts have been made to identify the energy positions of the bonding 2nb (dn-like) levels, which result from the CO-2n/metal-dn mixing (refs. 148-1.53). Only in the most recent ARUPS studies (ref. 153) unambiguous identification of the n-d surface band induced by the CO 2n-Ni-3d interactions was made. The C0(2n)-Ni(d) bands with a primarily Ni 3d character were located between 1 and 2.7 eV below the Fermi level. As discussed in ref. 153, the extent of the metal/2n overlap is a
65
matter of symmetry and is determined by the available surface metal bands. That is why the degree of the metal/2n coupling will depend on the nature of the substrate metal, the surface crystallographicorientation and the structure of the CO overlayer. Together with ARUPS which allows to determine the energetic positions of the occupied electronic levels, recently the inverse photoemission has offered the opportunity to gain an information about the energetic positions and the dispersion of the unoccupied antibonding 2na (2n-like) CO levels (refs. 16, 154). By means of IPE the following values for the maxima of the 2n CO resonance state above the Fermi level are reported - 2.7,3.4, 3.8 and 4.8 eV for CO/Ni, CO/Cu, CO/Ru and CO/Pd adsorption systems, respectively (ref. 16). P E studies for CO adsorption on the different planes of the same substrate showed the following differences in the 2na energy position - CO/Ni(110) - 3.7 eV, CO/Ni(lll) - 4.0 eV and CO/Ni(100) - 4.5 eV (ref. 154). In the latter system the 2na feature consists of two peaks. These recent results for the 2na energy position on the different Ni planes, as well as the 2na energy values given above for Ni, Cu, Ru and Pd question the existence of a simple general correlation between the IPE 2na values and the corresponding metal-CO bond strength in Fig. 2.7, because the CO Eb values are the same for the three different Ni planes, very similar for Ni, Ru and Pd, and considerably smaller for CO/Cu. The lack of general trends in the IPE measured 2na energy positions for the different CO/metal adsorption systems can be due to the following reasons:
(1)
different local surface electrostatic potential for the different metal surfaces and CO adsorption sites;
involvement of energetically different electronic levels in the 2n/metal coupling on the different substrate surfaces; (3) possible hybridization of the CO-2n levels with near resonant unoccupied metal states prior to the interaction with the occupied dn states (ref. 29). Recent theoretical calculations of Freeman et al. (ref. 155) allow us to determine the most probable energetic positions of the 2na orbital for the chemisorbed CO. They found a density from 1.5 to 5.0 eV with a maximum at =3 eV above the Fermi level (ref. 155). Following the above considerations it seems that in the framework of the Blyholder model the energetic splitting between 2xb and 2n,, AE2nb-2m, rather than the energy location of the 2na level should be a more reliable measure for the extent of the metaV2aCO coupling. It should be expected that AE2nb-2na for strong M-CO coupling will be larger than for weak coupling. Measurements of the energies of the 2nb+2na transitions have been performed by means of electron energy loss spectroscopy. The available EELS data show indeed that going from strong chemisorption (e.g. CO/Ru (refs. 156, 157), CO/Ir (ref. 158), CO/Pt (ref. 159), CO/Ni (ref. 160)) to weak adsorption (e.g. Co/Cu (ref. 161), CO/Ag (ref. 162)) AE2nb2m values drop from 6.5-7.5 eV for the strongly bound CO below 5 eV for the weakly bound one. Summarizing the results presented in this section we can conclude that they give an evidence that the metal/2nCO coupling results in two (bonding and antibonding) states with n symmetry. The bonding state, 2nb is located most probably at 1-2 eV below the Fermi level and the antibonding one 2na - in the region 3 - 6 eV above the Fermi level. The energy separation between (2)
66
these states is of the order of 7 eV for strongly chemisorbed CO on the metals from groups VIB, VIIB and VIII and less than 5 eV for the weakly adsorbed CO on the group IB metals.
2.4.8 INFLUENCE OF THE METAL-CO BONDING ON THE VIBRATIONALPROPERTIES OF THE CO MOLECULE As outlined in the previous Sections, the extent of the metal surface C o coupling determines the strength of the chemisorption bond and affects the intramolecular C - 0 bond
strength and bond length. This is reflected by the changes in the frequencies and the intensities of the metal-C and C - 0 vibrational modes (refs. 163-166). In their comprehensive review, N. Sheppard and T.T. Nguen (ref. 165) describe how the vibrational data for the complex transition metal compounds can be used for identification of the CO adsorption sites (surface metal atoms coordination number) on a single crystal metal surfaces. The following wave number ranges are predicted for the CO stretch frequencies, wc-o, of the differently coordinated adsorbed CO in upright configuration (ref. 165): (1)linear (on top, terminal or one fold coordinated) MCO-wc-0 = 2000-2300 cm-l; (2)bridged (two fold coordinated) M2C0 - W C - ~= 1880-2000 cm-l; (3) Three or four-fold coordinated, M3C0, M4C0 - wc-o = 1650-1880 cm-l. An inspection of the available vibrational data for the CO/metal adsorption systems shows indeed rather similar frequencies for the identically coordinated sites on the different metals, e.g. 2000-2100 cm-l for terminal CO, 1880-2000 cm-* for bridge CO and less than 1880 cm-l for the three and four coordinated CO (refs. 122, 167). The observed negative shift of the C - 0 stretch frequencies compared to the gas molecule ( ~ ~ - ~ = 1cm-l) 1 4 3 is due to several additive effects - mechanical coupling, self image shift and chemical bonding, the latter being considered as the major contributor (refs. 166, 167). To a lesser extent the coupling of the M-C and C - 0 vibrations with the substrate phonons (ref. 168) might contribute as well. At high CO coverages the dipole coupling with the vibrations of the other CO molecules in the overlayer also introduces values (refs. 164-167). changes in WC-0 The importance of the chemical effects and especially the contribution of the metal/2n backdonation to the observed negative oc-O shifts has been extensively discussed in many reviews (refs. 165-167, 169, 170). It is generally accepted that in the case of strong CO chemisorption the positive frequency shift due to the %/metal bonding contribution is overcome by the stronger metal/27t negative contribution and results in wc-o values lower than that of the free CO molecule. Depending on the chemical nature (electronic structure) of the substrate metal surface, differently coordinated adsorption sites are preferentially occupied on the various single crystal metal surfaces. In order to exclude the geometrical factor (it will be discussed later), first we will consider the stretching frequencies, measured for CO adsorbed on the different close packed metal surfaces with similar surface structure. The suggested correlations between the vibrational frequencies and the metal-CO bonding phenomena will be discussed. Miiller et al. (ref. 171) tried to correlate the direction and the magnitude of the negative oc-oshift, Awe and the effective charge, e*, for the adsorbed terminal CO on Cu, Ni and Fe with the extent of the metal dn to CO 27t charge donation.
67
As a result they showed that both Awe
and e* increase in the order CO/Cu, C O N CO/Fe, as expected from the increase of the degree of the dn/2n
lonnealcd T = 9to0 K2 4 0 K 205,8
donation and the M-CO bond strength in
1917
!PO5
-
the same order. However, this is the ideal case, where the same CO site occupation and the most appropriate CO/metal systems (with well expressed difference in their adsorptive properties) are chosen. Unfortunately, such correlation can not be generalized for all CO/metal systems under consideration. For example, if we account the C - 0 stretch frequencies for terminally bonded CO on: Pt(l1 l)-q-o(Pt)= 2065-2081 cm-l (refs. 172-173) and on Cu( 111) wc_o(Cu)= 2078-2070 cm-l (refs. 174-176), obviously the shift Awe is almost the same whereas the dx/2n coupling and M-CO bond strengths are very different for Pt and Cu.
I
I
8CO 0.57
0.545 0.525
1
30
2020 1950
1850
Consequently, the simplified Wavenumber Y (cm-’) interpretation of the absolute values of Fig, 2.21 IRAS following CO adsorption on the C - 0 stretching frequencies in terms Ni(ll1) at 90 K with annealing to 240 K. All spectra of bonding arguments only, i.e. relating were measured at 90 KL (from ref. 180). wc-o to the absolute magnitude of backdonation (ref. 171) or to the M-CO bond strength (ref. 125) can not be generalized over all CO/metal adsorption systems. One of the reasons is that we do not know what is the size of the induced image dipole effect on the W c - 0 value, but obviously it changes with the different substrate metals (refs. 177, 178). Another correlation, proposed by S . Ishi et al. (ref. 122) relates the variations of the M-C stretch frequencies to the M-CO bond strength. Indeed, wMvI-cis more directly affected by the metal-CO coupling and if we again compare the C O p t ( l l 1 ) and CO/Cu( 1 1 1) systems, the corresponding wpt-c=486 cm-l (ref. 173) and wcU-c=330 cm-l (ref. 179) values correlate with the measured CO Eb in the limit of low 0,- - 32 and 14 kcal/mol, respectively. Consequently, it seems that the correlation between Eb and o ~might - exist ~ for the CO/metal adsorption systems where CO is in the same bonding configuration at low CO coverages. Unfortunately, the different metals exhibit a great variety of the CO site occupation sequence, e.g. at low Oco CO occupies first:
68
(1)
three foldand bridgesitesonNi(ll1) andPd(ll1); terminal sites on Pt( 11l), Rh( 11l), Cu( 11l), Ru(0001). At high CO coverages a mixture
(2)
of terminal and bridge (or off equilibrium) bonded species usually coexist (ref. 167) (see e.g. Figs. 12-14). Figs. 21 and 22 show two examples for the CO site occupation sequence with increasing 0,- In the case of Ni(ll1) (Fig. 2.21) a subsequent occupation of three fold, bridge, and terminal co is observed (ref. 180). The splitting of the bridge IR band at high 0,-0 is associated with the structural changes in the CO overlayer. In the case of stepped Pt(533) surface (Fig. 2.21) (ref. 181) the two CO vibrational bands are located at wc-d2000 cm-1 (ref. 181). They are associated with terminal CO configurations and first occupation of the outside sites at the steps, followed by adsorption on the (111) terraces. These spectra are an example for the sensitivity of the C-0 vibrational frequencies on the particular type of the adsorption sites. They are in .___
I R A S of
CO
adsorbed on Pt(533) at 8 5 K
I
A /I
excellent correlation with the observed differences in the Eb for CO on these different sites (see Figs. 6, 10). At the present stage of knowledge the influence of the chemical nature of the substrate on the CO site selectivity is not understood. Many factors can be supposed to contribute to the choice of the coordination, such as the energy positions of the surface states, the energy positions, spacing and densities of the surface bands, the surface lattice constant, etc. They affect the degree of coupling between the metal and CO at the different adsorption sites and determine the most favorable one. More thorough theoretical analysis of the
-7
0
-
Fig. 2.22 IRAS in the CO stretch region for the system COPt(533) as a function of increasing relative CO coverage (from ref. 181).
factors affecting the CO coordination number on the transition metal surfaces is given in Chapter I of this book. Generally, the crystallographic plane orientation always affects the sequence of the site occupation, e.g. CO adsorbs initially in three fold and bridge sites on Ni(l11) (refs. 122, 180), bridge sites on Ni(ll0) (ref. 182), and terminal sites on Ni(100) (ref. 183). For the same coordination number the C-0 stretch
69
EEL Spectra of CO/Mociio) Recorded as a Function of CO Coverage
,
2055
II CO exposure (molecules/cm')
(91 6 . 3 ~ 1 0 ' ~
( f ) 2.6X1015
X
600
I .r x 1015
8.6 x 10l4
4 . 2 ~ 1 0 ~ ~
i 2.1 x 1014
x
1000
Clean
0
1000
2000
Energy Loss (cm-')
Fig. 2.23 H E E L S vibrational spectra for increasing CO coverages on Mo( 110). Adsorption temperature 120 K (from ref. 188).
frequency depends also whether CO resides in a hollow (1818 cm-I) or non hollow (1831 cm-') site on Ni(ll1) (ref. 180), long bridge (1855 cm-') or short bridge (1960cm-') on Ni(ll0) (ref. 182). Following this short summary of the different site occupation sequence on the different metal surfaces and planes it is obvious that a general correlation between CO E b and WM-c values involving all CO/metal adsorption systems is impossible. Let us now summarize briefly the CO coverage effect on the vibrational properties the adsorbed molecule. For the strongly chemisorbed CO on the metals from groups VIB, VIIB and VIII for the same site occupation the increase of Oc0 always causes an increase of 0c-0 and WM-C i.e. a positive ~ ( O C Oshift. ) The magnitude of this shift varies with the different metal substrates and it is attributed to several possible contributions:
70
i) ii)
direct or via metal vibrational coupling; chemical shifts due to changes in the
iii)
metal-CO bonding; structural changes in the overlayer. As postulated in refs. 174,176 the magnitude of the positive vibrational coupling shifts is similar for all metals. Consequently, the main reason
0
and K bonding components of the synergic
for the variations in the o(Oco) dependence for the different CO/metal systems is likely to be due to: (1) changes in the balance of the 5 0 / 2 ~bonding components (they account (Qco) on Ag, Au (refs. 184, 185) and Cu (ref. 176)); or (2) for the total negative WC-0 structural changes (formation of ordered islands, coexistence of different domain structures, etc. (ref. 167)). Up to here, we have considered only systems where the chemisorption bond is formed via the C atom. However, as has been mentioned in section 2.4.4, there are systems where the chemi-sorption bond can not be described in terms of the Blyholder model. For these CO/metal systems C-0 stretching frequencies lower than 1600 cm-l are measured. The reported Wc.0 values are: CO/Fe(100) - 1245 cm-l (ref. 112), CO/Fe(lll) - 1550 cm-l (ref. 116), CO/Cr(llO) - 1200 cm-l (refs. 110,186), CO/Mo(lIO) - 1345 cm-1 (refs. 187-189). These CO bonding configurations are usually observed at low CO coverages and are assumed to be a precursor for CO dissociation. Fig. 2.23. presents the evolution of the CO vibrational spectra with increasing CO coverage on Mo(l10). Three types of adsorbed CO with different vibrational frequencies WC-0 are distinguished. The CO species with oc-oranging from 1920 to 2055 cm-l are adsorbed in the conventional terminal bonding configuration. The CO species with the unusually low WC-0 - 1345 cm-l are supposed to be inclined multi-coordinated ones. As will be discussed in the forthcoming section these species can be completely dissociated at T>300 K.
2.4.9 DISSOCIATION PROBABILITY OF CO CHEMISORBED ON METAL SURFACES The dissociation of a molecule is thennodynamically favoured when the sum of the binding energies of the atomic constituents exceeds the gas phase dissociation energy, D.However, this requirement is necessary but not sufficient because the dissociation can be kinetically hindered by a substantial activation banier, E l . According to the recent views, supported by some theoretical calculations (ref. 127), E,* is determined exclusively by the binding energies (heats of adsorption) of the atomic constituents, whereas the molecular adsorption energy is of minor importance as far as E c values are considered. Fig. 2.24. and Table 2.2 illustrate the energetic situation for CO and its constituents - 0 and C on some single crystal metal surfaces where data for the surface binding energies of 0 and C, EM-O and EM<, are available. The energy level of the molecular CO is determined on the basis of the Eb values measured at low Oc0 As shown in Section 2.4.2, Eb values for the metals considered in Fig. 2.24. lie in the range 30-40 kcal/mol. The energy level of the dissociated CO, Ec.0,is estimated using the well known relationship:
71
250h
,
-
aJ -
2
Dco
\
Fig. 2.24 Potential energy diagram for CO adsorption on Pd( 11l), Pt( 11l), Ru(0001), Ni( 11l), Fe(l10) and W(l lo), representing the molscular and dissociated CO states and the activation energy, E, ,for dissociation.
k.0 =
+ EM-O
EM^
where EM-C and
(2.5)
- DCO
are the metal-carbon and metal-oxygen binding energies on the metal
surface. EM-O is related to the atomic oxygen heat of adsorption, AHo,by the equation:
EM-O =
112 (Lwo
+ Ds)
(2.6)
D ( C 0 ) and D ( 0 2 ) are the dissociation energies of CO and 0, gas molecules - 256 and 117 kcal/mol, respectively. The activation energy for dissociation of the chemisorbed COaE,*, is Table 2.2
M-C, M - 0 binding energies and the calculated EC,O and E,* values (in kcal/mol using eqs. 5-7) r
metal Pt(ll1) Pd(ll1) Fr(ll1) Ru(0001) Ni(ll1) Fe(ll0) W(110)
I
II EM-C (ref. 190) 136 117 152 144 152 175 198
L EM-O
84(ref. 191) 87(ref. 190) 87(ref. 190) 106(ref. 192) 13O(ref. 190) 127(ref. 190) 129(ref. 190)
-
Ec,o __-____
-36 -52 -12 -6 24 46 70
E,* 106 121 83 74 36 22 0.1
I
12
estimated using the relationship, derived in ref. 193:
As is demonstrated in Table 2.2 and Fig. 2.24, the dissociation of CO adsorbed on Pt, Pd, Ir and RU single crystals is a thermodynamically unfavored endothermic process. Indeed, the expenmental data on these metal surfaces show no evidence for CO dissociation even upon heating the substrate. AS can be judged from Fig. 2.4 the dissociation process is kinetically hindered because the activation barrier for dissociation, E,* is considerably higher than that required for CO desorption, Eb (equal to Ed). In the case of Ni( 111) the dissociation is thermodynamically favoured and E,* does not exceed substantially E,. That is why some Thermal Behavior of a Low Coverage CO/ Mo ( 1 1 0 ) Layer dissociation of was
co
I
x'ooo
'Oo0
I
CO/Mo
\
/
(110)
Heated to:
GL-----.
(f)
600 K
1130
l
I
l
I
observed on stepped surfaces, where as outlined in Section
6.2., the molecular state is stabilized on the step sites and its desorption energy is higher than that of the CO molecules adsorbed on the flat surfaces. The larger binding energy of the CO molecular state means an increased lifetime of the CO molecular
-
precursor
for
dissociation. In the case of Fe(ll0) and W(110) it is obvious that even on the flat
II
j,\ 0
I, i
,
I
' L
7& 1000
surfaces of these two metals the dissociation of the chemisorbed CO is both thermodynamically and
(a)
120 K
2000
Energy Loss (cm-'1 Fig. 2.25 Evolution of the low frequency CO HREELS spectra on Mo(l10) upon heating after adsorption at 120 K. All measurements are performed after cooling down to 120 K (from ref. 188).
kinetically favoured, which is in complete agreement with the experimental data (ref. 1). The above considerations illustrate quite well that the dissociation propensity of the chemisorbed CO cannot be simply related to the M-CO adsorption bond strength. Obviously, the
73
dissociation probability depends strongly on the affinity of the substrate to the atomic constituents C and 0. This explains why the different substrate metals, where the M-CO molecular binding energy is quite similar, exhibit rather different behavior with respect to CO dissociation. An interesting question is what is the detailed mechanism of CO dissociation on the surface. Indeed the molecular adsorption is required as the first step, but the classical Blyholder type bonding configuration with CO axis normal to the surface seems rather unfavourable with respect to the M - 0 coupling. As mentioned in the previous sections, flat lying or strongly inclined CO configurations are suggested to exist on the metal surfaces where CO dissociation takes place Fe(100) (ref. 112), Fe( 111) (ref. 116), Cr( 110) (ref. 1lo), Mo(100) (refs. 110, 186) and Mo(l10) (refs. 187-189). This bonding configuration is characterized with very low C - 0 stretch frequencies and it always dissociates upon heating, whereas the coexisting CO in normal configuration desorbs. Recently, the dissociation pathway of the inclined CO species on Mo(ll0) has been described (refs. 188, 189) considering the changes observed in the CO vibrational spectra upon heating. AS illustrated in Fig. 2.25 the inclined CO with ~ ~ - ~ = 1 cm-1 3 4 5is converted to C and 0 via an intermediate CO* state characterized with W C - ~ = 130 I cm-l. This CO* state is assigned as flat lying where both the 0 and C atoms of the molecule are bonded to the Mo surface atoms. Consequently, the dissociation of CO requires a favorable CO bonding configuration, which allows not only an increase of the degree of the metal dx/2%*CO coupling, but also involvement of the other (1% and 40) CO molecular orbitals in the bonding.
2.4.10
C O INDUCED PERTURBATIONS IN THE SURFACE ELECTRONIC AND GEOMETRIC STRUCTURE
The formation of the CO chemisorption bond, which is considered as a chemical type of bonding should result in changes of the distribution of electrons at the surface. Analyses of the angular dependence of the metal valence band photoemission spectra permits US to separate the contributions form the surface layer and the bulk and study the adsorbate induced changes in the electronic structure of the surface metal layer (ref. 194). The ARUPS data (refs. 148, 153, 195-199) for CO adsorption on transition metal surfaces have shown that CO causes a strong attenuation of the intensity of the metal surface d-band at Fermi level, a shift of the d band peak to a larger binding energy and a broadening of the d-band. The comprehensive ARUPS study of CO on Pd( 111) has shown that the Pd d-band is not attenuated uniformly upon CO adsorption and the features suffering strongest changes are associated with surface states or resonances of the Pd(l11) surface (ref. 197). As has been already mentioned in Section 2.4.7, together with the alteration of the metal d-band emission spectra a new surface resonance appears in the region 1-2.7 eV below the Fermi level. It is interpreted as co 2dd-metal bonding configuration with dominant metal-drc character (ref. 153). The changes in the charge distribution in the metal valence band occumng in the presence of CO lead also to binding energy shifts of the core levels. In the case of negligible final state effect the direction of this shift can be used as an indication of the direction of the charge transfer taking place in the CO/metal adsorption system (refs. 198, 199). The available data for CO adsorption on transition metals as Pt(ll1) (ref. 200), W and Ta (ref. 198) show that as expected CO
74
induces a shift of the surface core levels to higher energies, a direction observed for other acceptor adsorbates as well. Here, it is worth stressing that the observed CO induced downward 4fcore level shift on Pt(ll1) (ref. 200) is in contradiction with the sign of the CO induced work function changes. This supports the view, discussed in Section 4.5. that the measured work function changes do not always reflect the direction of the net charge transfer.
CO adsorption in some cases can cause also structural surface changes. This concerns (1 10) single crystal surfaces of the third row transition metals (Pt, Fr, Au). In their riyal state they exhibit (1x2) reconstruction (refs. 201,202), where every second atomic row in the (1 10) direction is missing such that microfacets
(b)of
(111) orientation exist
085 065
OLS 03
1
015
Il.21 T=mK
0
E, lev)
Pt(llOl(1rlt + co
O M hv = 1253 6 eV
T=120K
A5327
NIE
535
530 E, lev1
Fig. 2.26 0 1s spectra of CO adsorbed on: (a) Pt(1 lo)( 1x2) at 120 K, (b) Pt(llO)(lx2)at 3 0 K , (c)Pt(llO)(lxl)at 120K(fromref. 124).
on either side of the remaining (110) rows (refs. 203, 204). It turns out that the co adsorption on reconstructed (1x2) surface at 250K causes lifting of the reconstruction and restoration of the (1x1) order (refs. 124, 205-2 10). This phase transition is both coverage and temperature driven. The changes in the surface structure leads to changes of the specific CO adsorption as sites, reflected by the site sensitive
75
XPS O(1s) spectra in Fig. 2.26, work function changes (Fig. 2.19) and the vibrational spectroscopy data (ref. 210). For CO adsorbed on reconstructed Pt(llO)-(lx2) at 120 K these sites turned out to be terminal tilted CO on the (1 10) rows (ref. 210), and on top and bridge CO on the (1 11) facets, similar to that on Pt( 111) (ref. 123). For CO adsorbed on nonreconstructed Pt( 1lo)-( 1x1) at 120 K they are terminal and bridge, but the environment of the bridge species is different because they reside now on the (110) atomic rows. This difference is reflected by both XPS and vibrational spectroscopy, where the bridge CO species on Pt(1 lo)-( 1x1) and Pt(l10)-(1x2) show different O(ls) binding energies (ref. 124) and C - 0 stretching frequencies (refs. 206,207). The adsorption of CO on Pt(llO)-(lx2) at 300 K shows only terminal CO, because at this temperature CO induces (1x2) (1x1) conversion and the (1 11) facets are removed the bridge CO species on (Pt(1 10)-(lxl) are weakly bound and they are not present at 300 K). This behavior of the CO/Pt(llO) system is a model example for the variety of the structural changes that might take place on the real metal catalyst surfaces, consisting of various oriented microcrystallites. The possible structural changes can affect the CO adsorption state in a rather complex way (bond and site configuration) and will be reflected by the activity and selectivity of the catalytic systems.
2.5 EFFECT OF SOME ADDITIVES ON THE CO-TRANSITION METAL SURFACE INTERACTIONS The interest in the studies of CO adsorption on metal surfaces modified with different kind of additives is closely related to fundamental understanding of two very important phenomena - poisoning and promotion in heterogeneous catalysis. Since CO adsorption is always the first step of the catalytic reactions the different aspects of the additive effect on the chemisorptive properties of the surface with respect to CO has to be known.
2.5.1
ELECTRONEGATIVE ADDITIVES
Since CO exhibits electron acceptor behavior, the elements of groups IVA, VA, VIA and VIIA, which are electronegative with respect to the metal catalysts, act as poisons in the surface reactions involving CO, e.g. Fischer-Tropsch synthesis, CO oxidation, CO/NO neutralization, steam reforming, etc. Well known poisons for these catalytic processes are S, Se, AS, P, C, N. Recently many studies have been dedicated to CO adsorption on metal surfaces modified with electronegative adatoms (refs. 21 1-216). The adsorption of the electronegadve modifiers always perturbs substantially the properties of the surface metal atoms involved in the bonding and causes the following major changes in the CO adsorption kinetics, energetics, site occupation, surface dynamics and adsorption propensity (ref. 214): (1) Sequential elimination of the original CO adsorption sites starting with the most tightly bound one. It is illustrated in Fig. 2.27 where the reduction in the occupation of the most tightly bound Pz-CO state on Ni(100) and the total adsorbed amount CO are plotted as a function of the additives coverage. (2) Reduction of the total adsorption capability of the surface (Fig. 2.27).
Reduction of the CO adsorption rate due to the decrease of the lifetime of the
co
precursor on the surface sites affected by the modifier. Appearance of new less strongly bound CO states. They are associated with the occupation of different adsorption sites in the close vicinity of the modifier. They exhibit different vibrational modes and reduced CO adsorption binding energy. Stiffening of the CO frustrated translational vibrational motions parallel to the surface, which indicates that the modifier induced changes in the surface potential energy contour affects the CO mobility as well (ref. 216). As is well known the adsorbate surface dynamic is an important factor in the surface reactions. Decrease of the CO dissociation probability and complete inhibition of the CO dissociation above certain modifier coverages. This effect on the CO dissociation propensity is due to several factors - (i) decrease of the lifetime of the molecular CO state (Eb) which is a precursor for dissociation; (ii) possible increase of the activation barrier
for dissociation if the binding energies of the C and 0 atoms are affected in the same manner by the presence of a modifier; (iii) blocking of the energetically most favorable adsorption sites for the precursor and the dissociation products; (iv) hindered CO diffusion, etc. (ref. 214). The main reason for the observed effects is supposed to 0.4
be the existence of strong
CI
* S 0.3
repulsive interactions between
AP
0.2 0.1
0 06 0.5
0.4
0.3 0.2 0.1 0 0
0.1
0.2
0.3
0.4
0.5
06
0.1
ADDITIVE COVERAGE (ML)
Fig. 2.27 Dependence of the occupation of the most tightly bound 02-CO state and the total CO uptake on the additive precoverage at 120 K (from ref. 215).
CO and the coadsorbed electronegative adspecies. They can involve substrate mediated or direct repulsion between some spatially extended (e.g. 271*) and/or energetically suited (e.g. 50 CO and np-modifier) CO and modifier electron orbitals and electrostatic interactions. The substate mediated effects due to the modifier induced changes in the surface electron local density of states also contributes on a local scale (at distances less than 5 8).As a result both the metal/2x CO and 50/metal contributions to the co adsorption bond are reduced.
All these effects can be directly related to the observed undesired changes in the activity and selectivity of the metal catalysts for surface reactions involving CO.
2.5.2 ELECTROPOSITIVE ADDITIVES Typical electropositive additives are the alkali metals. They are introduced as alkali salts during preparing the catalyst for reactions requiring CO dissociation as an intermediate step. Alkali metals serve as promoters, in these reactions and change in a desired direction the activity and/or selectivity of the catalytic reaction. Generally speaking, alkali effect on the CO adsorption behavior and dissociation propensity is in a reverse direction compared to that of the elecuonegative additives (refs. 212,217,218). Alkali species on the surface create local changes in the CO adsorption state. These promoted adsorption sites for CO are characterized with: Higher CO adsorption binding energy. Fig. 2.28 presents the alkali induced changes in the
CO TPD spectra and the initial heats of CO adsorption A H C O O , associated CO occupying first the preferred, promoted, sites. The increase of the CO binding energy on the promoted sites with an increase of the alkali coverage is associated with changes in the adsorption state of the alkali species (becoming less ionic-like at higher alkali coverages). These changes in the net population of the alkali valence ns states enhances the contribution of the direct alkali ns/2x CO coupling to the stabilization of the CO adsorption state on the surface (refs. 157,218-220). Promoted CO has always a lower C - 0 stretching frequency. The magnitude of the alkali induced downward frequency shift increases from 100 cm-l up to 4 5 0 cm-’ going from low to high alkali coverages (refs. 217,221). This effect confirms the supposed above
increase of the degree of coupling between CO and alkali species at higher alkali coverages and reflects the weakening of the C - 0 bonding. LEED, H E E L S , TPD and WF data (refs. 157, 217-221) give evidence that as a result of the direct alkali-CO interactions mixed alkali-CO patches are formed on the surface, resembling a surface “compound-like” complex with fixed stoichiometry. The dissociation propensity of the promoted CO is enhanced. This can be rationalized considering the stabilization effect on the CO molecular state and the possible similar effect on the adsorption state of the dissociation products. The increased electron density in the 2n* CO antibonding orbital in the presence of alkalis facilitates the breakage of the C - 0 bond. Since the horizontal CO configuration is believed to precede the dissociation, for this configuration coadsorbed alkalis are expected to decrease the activation barrier for dissociation as well (ref. 222). [t can be summarized that contrary to the electronegative additives, the observed alkali induced effects on the CO adsorption behavior are due to the existence of strong attractive interactions between the coadsorbates. These interactions include direct coupling between the alkali species and CO and electrostatic interactions, the former being of major importance.
As I
,
,
,
,
,
it
concerns
the
alkali
,
promotion effect in the catalytic reactions involving CO dissociation, one can predict, on the basis of the CO
4C ll 0.7
0.4
0.2 0.1
I
I
t
300 500
I
I
700
T(K)
b h
I
0.0 900 I
adsorption studies, that the increased CO dissociation propensity will change the adsorption activity and selectivity towards production of unsaturated hydrocarbons with higher molecular weight (refs. 217-231).
2.6 THE RELEVANCE OF THE CO/SINGLE CRYSTAL METAL STUDIES TO THE CO-METAL CATALYST INTERACTIONS
The fundamental knowledge for CO interaction with clean and modified E transition metal single crystal surfaces \ has been successfully applied recently 0 40 to describe the intermediate steps in v catalytic reactions involving CO. An 0 o v I advantage is that most of the important A Cs reactions, e.g. CO oxidation and CO 30t hydrogenation are structurally 0.2 0.4 0.6 0.8 insensitive and the results obtained on Alkali C o v e r a g e ( M L ) single crystals can be successfully correlated to the real metal catalysts. Fig. 2.28 (a) CO TPD spectra from Ru(1010) The latter consist of small differently modified with increasing amount of Na (b) The oriented microcrystallites with a large dependence of the binding energy of the promoted CO species on the alkali coverage estimated from proportion of edge sites. The closest in the TPD spectra at low 0 0 (initial heat of structure to the real catalyst surfaces are adsorption) (from (ref. 1.55) and (ref. 219)) the single crystal stepped surfaces and the reconstructed surfaces, where the environment and possible coordination of CO on the step sites, on the rows, and on the flat facets are different. As discussed in the previous sections, this difference is reflected by the CO Q)
2
a
50
79
vibrational spectra, the XPS 0 1s and C 1s binding energies, the CO induced work function changes, the CO adsorption binding energies, and the CO dissociation propensity. In their comprehensive review N. Sheppard and T.T. Nguyen (ref. 165) have shown that the single crystal results can be successfully applied to describe CO surface configurations on supported metal particles when they are large enough and well reduced. it has been found that the trend in the site occupation observed for the supported catalysts can be correlated with the preferential crystallographic planes exhibited by the catalyst microcrystallites. Recently, a very good correlation between CO IR bands on silica supported Pt particles and IRAS data for CO adsorbed on flat, stepped, and step-kinked single crystal surfaces has been reported. it was shown that the step atoms mimic the edge atoms and the kink atoms mimic the corner atoms of the Pt supported particles (ref. 223). However, for highly dispersed catalysts consisting of very small metal clusters new CO bands are observed as well, e.g. due to M(C0)z (ref. 165) coordination, which is an indication of the existence of atomically dispersed metals. The knowledge about the CO adsorption behavior on clean and modified single crystal metal surfaces has already been successfully used to describe the CO oxidation rate on Pt,Fr and Rh catalysts (refs. 224-227), CO hydrogenation reactions on clean and modified with S, P, C, or K, Ni (refs. 211,229), Fe (ref. 230), Co (ref. 231), Mo (ref. 232), and W (ref. 233) catalysts. AS will be discussed in the forthcoming chapters, in situ reaction studies on model single crystal metal catalysts in combination with studies of the adsorption behavior of the reagents and the products on the same surfaces, have already made the first successful steps towards fundamental understanding of the mechanism of the surface processes.
2.7 REFERENCES 1
J.C. Campuzano, The Adsorption of Carbon Monoxide by Transition Metals, to be published in The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis (eds. D. King and D.P. Woodruff, Elsevier)
2a 2b 3 4 5
6
7
8 9
G. Ertl and J. Kiippers, Low Energy Electrons and Surface Chemistry, VCH, 1985 and the references therein; D.P. Woodruff, T.A. Delchar, Cambridge Solid State Sciences Series (eds. R. Cahn, E. Davies and I. Ward, Cambridge) 1986. G. Erlich, J. Appl. Phys. 32 (1961) 4; Adv. Catalysis, 14 (1963) 255. P. A. Redhead, Vacuum, 12 (1962) 203. J.T. Yates, Jr., The Thermal Desorption of Adsorbed Species, in Methods of Experimental Physics, Vol. 22 (ed.R.L. Park, Academic Press, 1985, p. 425). D. Menzel, in Chemistry and Physics of Solid Surfaces IV, Springer Series in Chemical Physics, (ed.R. Vanselow and R. Rowe 1982, p. 389). L.P. Levine, J.F. Ready and E. Bemalg, J. Appl. Phys., 38 (1967) 531; IEE J. Quantum Electron. QE-4 (1968) 18. P.M. Menill, Cat. Rev., 4 (1970) 115. J.A. Barker and D.J. Auerbach, Surface Sci. Reports, Vol. 4 (1984) 1 and the references therein.
80
10
M.P. D,Evelyn and R.J. Madix, Surface Sci. Reports, Vol. 3 (1983) 413.
11
D. Menzel and R. Gomer, J. Chem. Phys., 41 (1964) 331 1.
12 P.A. Redhead, Can. J. Phys., 42 (1964) 886. 13 M.L. Knotec and P.J. Feibelman, Phys. Rev. Lett., 40 (1978) 904. 14 T.E. Madey,. P.L. Doering, E. Bertel and R. Stockbauer, Ultramicroscopy I1 (1983) 187 and the references therein. 15 A. Benninghoven, Z. Phys., 230 (1970) 403; Surface Sci., 57 (1975) 596. 16 V. Dose, Surface Sci. Reports, 5(8), 1985, 337 and the references therein. 17 J.B.Pendry, Low Energy Electron Diffraction, in Techniques of Physics, (eds. G.M. Conn 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41
42 43
and K.R. Coleman, Academic Press, London, 1974). E.W. Plummer, W.R. Salaneck and J.S. Miller, Phys. Rev. B, 18 (1978) 1673. G.J. Schulz, Rev. Mod.Phys., 45 (1973) 423. W.L. Jorgensen and L. Salem, The Organic Chemists Book of Orbitals (Acad. Press, New York 1973, p. 28). Handbook of Chemistry and Physics, 63rd Edition, 1982-83, CRC Press, p. F-182. B. Rosenblum, A.H. Nethercot, C.H. Townes, Phys. Rev., 109 (1958) 400. F.B. Billingsley and M. Krauss, J. Chem. Phys., 60 (1974) 4130. K. Kirby-Dockman and B. Liu, J. Chem. Phys., 66 (1977) 4309. K.P. Huber and G. Herzberg, Molecular Spectra and Molecular Structure IV, Constants of Diatomic Molecules (Van Nostrand Reinhold, New York 1979). G. Blyholder, J. Chem. Phys., 68 (1964) 2772. P.S. Bagus, C.,L. Nelin, C.W. Bauschlicher, J. Vac. Sci., Technol., A2 (1984) 905. D. Post, E.J. Baerends, Surface Sci., 109 (1981) 167. Ph. Avouris, P.S. Bagus, C.J. Nelin, J. Electr. Spectr. Rel. Phenom., 38 (1986) 269. R.P. Eischens, W.A. Pliskin and S.A. Francis, J. Chem. Phys., 22 (1954) 1786. D.A. King, in Chemistry and Physics of Solid Surfaces I1 (ed.R. Vanselow) CRC 1979 p. 82. M. Kiskinovaand G. Bliznakov, Surface Sci., 123 (1982) 61. P.J. Kisliuk, J. Phys. Chem. Solids, 3 (1957) 95. R.L. Clavenna and L.D. Schmidt, Surface Sci., 22 (1970) 365. M.A. Morris, M. Bowker and D.A. King, in Comprehensive Chemical Kinetics Vol. 19 (eds. C.H. Bannford, C.F.T. Tipper and R.G. Comption, Elsevier, 1984 p. 42 C. Harendt, J.W. Goschnick and H. Hirschwald, Surface Sci., 152/153 (1985) 453. C.T. Cambell, G. Ertl, H. Kuipers and J. Segner, Surface Sci., 107 (1983) 207. T.-H. Lin and G.A. Somorjai, Surface Sci., 107 (1980) 573. J. Fair and R.J. Madix, J. Chem. Phys., 73 (1980) 3480. T. Engel, J. Chem. Phys., 69 (1978) 373. S.L. Tang, M.B. Lee, J.D. Beckerle, M.A. Hines and C.T. Ceyer, J. Chem. Phys., 82 (1985) 2826. S.L. Tang, J.D. Beckerle, M.B. Lee and S.T. Ceyer, J. Chem. Phys., 84 (1986) 6488. J.D. Beckerle, O.Y. Yang, A.D. Johnson and S.T. Ceyer, Surface Sci., 195 (1988) 77.
81
44 C.R. Helms and R.J. Madix, Surface Sci., 52 (1975) 677. 45 46 47 48 49
P.R. Norton, R.L. Tapping and J.W. Goodale, Surface Sci., 72 (1978) 33. J.P. Tracy, P.W. Palmberg,. Surface Sci., 14 (1969) 274. H. Pfnur, P. Feulner and D. Menzel, J. Chem. Phys., 79 (9)( 1983) 4613. E.G. Seebauer, A.C.F. Kong and L.D. Schmidt, a) Surface Sci., 176 (1986) 134; b) Surface Sci., 193 (1988) 417. J.B. Miller, H.R. Siddiqui, S.M. Gates, J.N. Russell, Jr., J.T. Yates, Jr., J.C.Tully and M.J. Cardillo, J. Chem,. Phys., 87 (1987) 6725.
50
H.R. Siddiqui, P. Chen, X. Guo, 1. Chorkendorff, J.T. Yates, Jr. and J.C.Tully, to be published in Surface Sci.,
51 52
G. McElthiney, H. Papp and J. Pritchard, Surface Sci., 54 (1976) 617.
53 54 55 56 57 58 59
e.g. J.H. de Boer, The Dynamical Character of Adsorption (Clarendon, Oxford, 1953); Adv. Catal. 8 (1956) 82. H. Ibach, W. Erley and H. Wagner, Surface Sci., 92 (1980) 29. A.M. Lahee, Y.P. Toennies and Ch. Woll, Surface Sci., 177 (1986)255. N.V. Richardson and A.M. Bradshaw, Surface Sci., 88 (1979) 255. C.M.A.M. Mesters, A.F.H. Welers, O.L.J. Gijzem, J.W. Geus and G.A. Bootsma, Surface Sci., 115 (1982) 237. H. Conrad, G. Ertl, J. Koch and E.E. Latta, Surface Sci., 43 (1974) 462. J.C. Tracy, J. Chem. Phys., 56 (1972) 2736.
H. Ohtani, E.-T. Kao, M.A. Van Hove and G.A. Somorjai, Progress in Surface Science, Vol. 23 (1986) 155. 60 E.D. Williams and W.H. Weinberg, Surface Sci., 82 (1979) 93. 61 H. Papp, in Proc. IVC-8, ICSS-4, ECOSS-3, Cannes 1980, p. 245. 62 N. Vennemann, E.W. Schwarz and M. Neumann, Surface Sci., 126 (1983)273. 63 G. Ertl, M. Neumann and K.M. Streit, Surface Sci., 64 (1977) 393. 64 J.C. Campuzano, R. Dus and R.G. Greenler, Surface Sci., 102 (1987) 172. 65 G. Ertl and J. Koch, Z. Naturforsch. 25a (1970) 1906. 66 P.A. Thiel, E.D. Williams, J.T. Yates, Jr. and W.H. Weinberg, Surface Sci., 84 (1979) 54. 67 R.J. Koestner, M.A. Van Hove, G.A. Somorjai, Surface Sci., 107 (1981) 439. 68 J. Kuppers and A., Plagge, C.M. Comrie and W.H. Weinberg, J. Vacuum Sci., Technol., 13 (1976) 259; 264. 69 J. Kessler and F. Thieme, Surface Sci., 67 (1977) 405. 70 S.D. Bader, J.M. Blakey, M.B. Brodsky, R.J. Friddle and R.L. Panosh, Surface Sci., 24 (1978) 425. 71 G. Broden, S. Pirug, H. Bronzel, Surface Sci., 72 (1978) 45. 72 J.P. Biberian and M.A. Van Hove, Surface Sci., 122 (1982) 600. 73 S. Andersson and J.B. Pendry, Surface Sci., 71 (1978) 25. 74 S.Y. Tong, A. Maldonado, C.H. Li and M.A. Van Hove, Surface Sci., 94 (1980) 73. 75 J.C. Tracy, J. Chem. Phys., 56 (1972) 2748. 76 S. Andersson, Surface Sci., 89 (1979) 477.
82
77 J. Pritchhard, Surface Sci., 79 (1979) 231. 78 P.G. Castner, B.A. Sexton and G.A. Somorjai, Surface Sci., 71 (1978) 519. 79 J. Benziger and R.J. Madix, Surface Sci., 94 (1980) 119. 80 J.P. Biberian and M.A. Van Hove, Surface Sci., 118 (1982) 443 and references therein. 81 J.P. Biberian and M.A. Van Hove, Surface Sci., 138 (1984) 361. 82 M.D. Alvey, M. Dresser, J.T. Yates, Jr., Surface Sci., 165 (1986) 447. 83 W. Riedl and D. Menzel, Surface Sci., 163 (1985) 39. 84 M.L. Ban, M.A. Van Hove and G.A. Somojai, Surface Sci., 185 (1987) 355. 85 M. Kiskinova, A. Szabo and J.T. Yates, Jr., Surface Sci., (1988), in press. 86 R.J. Behm, G. Ertl, V. Penka, Surface Sci., 160 (1985) 387. 87 T.E. Jackman, J.A. Davis, D.P. Jackson, W.N. Unertl, P.R. Norton, Surface Sci., 120 (1982) 389. 88 C.M. Lambert,J. Chem. SOC.Far. Trans. I, 7 (1976) 1659. 89 H. Vahrenkamp, Chem. Ber., 111 (1978) 3472. 90 J. Stohr and R. Jaeger, Phys. Rev. B, 26 (1982) 4111. 91 K. Stohr, J.L. Gland, W, Eberhardt, D. Outka, R.J. Madix, F. Sette, J.R. Koestner and u. Doebler, Phys. Rev. Lett., 51 (1983) 2414 and the references therein. 92 W. Wiirth, C. Schneider, E. Umbaqch and D. Menzel, Phys. Rev. B,. 34 (1986) 1336. 93 L.D. Petersson, S. Kono, N.F. T. Hall, C.S. Fadley and J.B. Pendry, Phys. Rev. Lett., 42 (1979) 1545. 93 L.D. Petersson, S. Kono, N.F.T. Hall, C.S. Fadley and J.B. Pendry, Phys. Rev. Lett., 42 (1979) 1545. 94 95 96 97 98 99 100
101 102 103 104a 104b 105a
K.C. Prince, E. Holub-Krappe, K. Horn and D.P. Woodruff, Phys. Rev. B, 32 (1985) 4249. D.A. Wesner, F.P. Coenen and H.P. Bonzel, Phys. Rev. B, 33 (1986) 8837. D.A. Wesner, F.P. Coenen and H.P. Bonzel, Surface Sci., 199 (1988) L419. C.L. Allyn, T. Gustafsson, E.W. Plummer, Chem. Phys. Lett., 47 (1977) 127. A.W. Plummer, W. Eberhardt, Adv. Chem. Phys., 49 (1982) 533 and the references therein. T.E. Madey, Surface Sci., 79 (1979) 575. T.E. Madey, F.M. Netzer, J.E. Houston, D.M. Hanson, R. Stockbauer, DIET I (Eds. N.H. Tolk, M.H. Traum,. J.C. Tully and T.E. Madey, Springer Verlag, Berlin 1983) p. 120. W. Riedl, D. Menzel, DIET I1 (Eds. W. Brenig and D. Menzel Springer-Verlag, Berlin 1985) p. 136. M. Kiskinova, A. Szabo, A-M. Lanzzillotto and J.T. Yates, Jr., Surface Sci., 202 (1988) L559. W. Wurth, J.J. Weimer, E. Hudeczek and E. Umbach, Surface Sci., 173 (1986)L615. H. Niehus, Surface Sci., 92 (1980) 88; Surface Sci., 80 (1979) 245: 87 (1979) 561. D. Rieger, R.D. Schnell and W. Steinmann, Surface Sci., 143 (1984) 157;
83
105b
G.R. Bare, K. Griffiths, P. Hofmann, D.A. King, G.L. Nybnerg and N.V. Richardson, Surface Sci., 120 (1982) 367. 106 R. Jaeger and D. Menzel, Surface Sci., 93 (1980) 7 1. 107 H. Kihlenbeck, M. Neumann and H.J. Freund, Surface Sci., 173 (1986) 194. 108 T.E. Madey, J.T. Yates, Jr., A.M. Bradshaw and F.M. Hoffmann, Surface Sci., 89 (1979)
370. 109 110 111 112
T.E. Madey, J. Houston, S. Dahlberg, Le Vide, Suppl. 1 (1980) 205. N.D. Shim and T.E. Madey, Phys. Rev. Lett., 53 (1984) 2481. N.D. Shim and T.E. Madey, Surface Sci., 180 (1987) 615. C. Benndoxf, B. Kruger and F. Thieme, Surface Sci., 163 (1985) L625.
113 R.W. Moon, D.J. Dwyer, S.L. Bemasek, Surface Sci., 163 (1985) 215. 114 D.W. Moon, S. Cameron, F. Zaera, W. Eberhardt, R. Carr, S.L. Bemasek, J.L. Gland and D.J. Dwyer, Surface Sci., 180 (1986) L123. 11.5 C. Benndorf, B. Nieber and B. Kriiger, Surface Sci., 177 (1986) L907. 116 U. Seip, M.C. Tsai, K. Christmann, J. Kiippers and G. Ertl, Surface Sci., 139 (1984) 29. 117 C.K. Rofer-De Porter, Chem. Rev., 81 (1981) 447. 119 A.B. Anderson and D.P. Onwood, Surface Sci., 154 (1985) L261. 119 K.A.R. Mitchel, Surface Sci., 149 (1985) 93 and the references therein. 120 P. Chini, Inorg. Chim. Acta Rev., 2 (1968) 31. 121 J. Holzl and F.K. Schulte, Solid Surface Physics, (Springer-Verlag, Heidelberg 1979, p.
1). 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139
Shin-ichi Ishi, Yuichi Ohno and B. Viswanathan, Surface Sci., 161 (1985) 349 and the references therein. M. Kiskinova, G. Pirug and H. Bonzel, Surface Sci., 133 (1983) 321. N. Freyer, M. Kiskinova, G. Pirug and H. Bonzel, Applied Physics, A39 (1986) 209. B.E. Nieuwenhuys, Surface Sci., 105 (1981) 505. E. Shustorovich and R.C. Baetzold, Appl. Surface Sci., 11/12 (1982) 693. E. Shustorovich, Surface Sci. Reports, 6 (1986) 1. R.J. Smith, J. Anderson and G.J. Lapayre, Phys. Rev. Lett., 37 (1976) 1081. E.W. Plummer, T. Gustafsson, W. Gudat and D.E. Eastman, Phys. Rev., A15 (1977) 2339. C.A. Brundle, P.S. Bagus, D. Menzel and K. Hermann, Phys. Rev. B, 24 (1981)7041. C.L. Allyn, T. Gustafsson and E.W. Plummer, Solid State Commun. 24 (1977) 531. P. Heskett, E.W. Plummer and P.R. Messmer, Surface Sci., 139 (1984) 558. D. Schmeisser, A. Gutman and K. Jacobi, 8, ECOSS6, York, 1986. R.P. Messmer, Surface Sci., 158 (1985) 40. C.M. Kao and R.P. Messmer, Phys. Rev. B, 31 (1985) 1835. P.S. Bagus, C.J. Nelin, C.W. Bauschlicher, J. Vac. Sci., Technol., A 2 (1984) 90.5. C.L. Allyn, Ph.D. Thesis, University of Pennsylvania, 1977. R. Miranda, K. Wandelt, D. Rieger and R.D. Schnell, Surface Sci., 139 (1984) 430. M. Trenary, C.L. Tang, R.J. Simonson and F.R. Mcfeely, Surface Sci., 124 (1983) 555.
84
140 C.R. Bare, P. Hofmann and D.A. King, Vacuum 31 (1981) 463. 141 C.W. Seabury, T.N. Rhodin, M.M. Traum, B. Benbow and Z. Hurych, Surface Sci., 97 (1980) 363. 142 T. Grenter, D. Heskett, E.W. Plummer and F.J. Freund, Phys. Rev. B, 27 (1983) 71 17. 143 W. Braun, N. Neumann, M. Iwan and E.E. Koch, Phys. Status Solid (B) 90 (1978) 525. 144 E.S. Jensen and T.N. Rhodin, Phys. Rev. B, 27 (1983) 3338. 145 T.C. Chiang, G. Kaidl, D.E. Eastman, Solid State Commun. 36 (1980) 25. 146 G. Broden, T.N. Rhodin, C. Brucker, R. Benbow and Z. Hurych, Surface Sci., 159 (1976) 593. 147 D. Bullett, Surface Sci., 152/153 (1985) 654. 148 R.J. Smith, J. Anderson, C.JU. Lapeyre, Phys. Rev. B, 15 (1980) 632. 149 C.F. McConville, C. Somerton and D.P. Woodruff, Surface Sci., 139 (1984) 75. 150 F. Bozso, J. Arias, J.T. Yates, Jr., R.M. Martin and H. Metiu, Chem. Phys. Lett., 94 (1983) 243. 151 H. Conrad, G. Ertl, J. Kuppers, S.W. Wang, K. Gerard and H. Haberland, Phys. Rev. Lett., 42 (1979) 1082. 152 F. Micchielli, C. Boizian, R. Nuvolon and J. Roussel, J. Phys. C, 13 (1980) 2441. 153 H. Kuhlenbeck, H.B. Saalfeld, M. Neumann, H.-J. Freund and E.W. Plummer, Appl. Physics, A 44 (1987) 83. 154 W. Reimer, Th. Fink and J. Kuppers, Surface Sci., 186 (1987) 55. 155 A.J. Freeman, C.L. Fu andE. Wimmer, J. Vac. Sci., Technol., A 4 (1986) 1265. 156 R. Hesse, P. Staib and D. Menzel, Appl. Phys., 18 (1979) 227. 157 M. Kiskinova, M. Tikhov, Surface Sci., 194 (1988) 379. 158 F.P. Netzer, E. Bertel and J.A.D. Matthew, J. Electr. Spectr. Rel. Phenom., 18 (1980) 199. 159 F.P. Netzer and J.A.D. Matthew, Surface Sci., 81 (1979) L651. 160 B.F. Koel, D.E. Peeble and J.M. White, Surface Sci., 125 (1983) 739. 161 J. Kessler and F. Thieme, Surface Sci., 67 (1977) 405. 162 G. McElhiney, H. Papp and J. Pritchard, Surface Sci., 54 (1976) 617. 163 H. Ibach and D.L., Mills, Electron Energy Loss Spectroscopy and Surface Vibrations (Academic Press, New York, 1982). 164 Vibrations at Surfaces (ed. C.R. Brundle and H. Morrow, Elsevier, Amsterdam, 1983). 165 N. Sheppart and T.T. Nguyen, Advances in Infrared and Raman Spectroscopy, Vol. 5 (eds. R.J.H. Clark and R.E. Hester, Hayden, London, 1978, p. 67). 166 R.F. Willis, A.A. Lucas, G.D. Mahan, The Chemical Physics of Solid Surfaces and heterogeneous Catalysis, Vol. 2, Chapter 2 (ed. D.A. King, D.P. Woodruff, Elsevier, Amsterdam, 1983). 1 67 F. Hoffmann, Surface Sci. Reports, 3 (1985)107, and the references therein. 168 T.S. Rahman, D.L. Mills and J.E. Black, J. Electr. Spectroscopy Rel. Phenom., 29 (1983) 199. 169 Vibrational Spectroscopy of Adsorbates, Chapters 8, 10, and 11 (ed. R.F. Willis, Springer Verlag, 1980).
85
170
Vibrational Spectroscopy of Molecules on Surfaces (eds. J.T. Yates, jr. and T. Madey, Plenum Press, New York, 1987).
W. Muller and P.S. Bagus, J. Vac. Sci., Techhnol. A 3 (1985) 1623. B.E. Hayden and A.M. Bradshaw, Surface Sci., 125 (1983) 787. 173 H. Steiniger, S. Lewald and H. Ibach, Surface Sci., 123 (1982) 264. 174 P. Hollins and J. Pritchhard, Surface Sci., 89 (1979) 486. 175 B.F. Hayden, K. Kretzschmar and A.M. Bradshaw, Surface Sci., 155 (1985) 553. 176 P. Hollins and J. Pritchhard, in Vibrational Spectroscopies of Adsorbates, Vol. 15 (Ed. R.F. Willis, Springer Verlag Berlin 1980). 177 S. Efrima and H. Metiu, Surface Sci., 92 (1980) 433. 178 M. Sheffler, Surface Sci., 81 (1979) 502. 179 J. Paul andF.M. Hoffmann, Surface Sci., 122 (1986) 151. 180 L. Surnew, Z. Xu and J.T. Yates, Jr., Surface 201 (1988) 1;14. 181 B.E. Hayden, K. Kretzschmar, A.M. Bradshaw and R.G. Greenler, Surface sci., 149 (1985) 394. 182 M. Nishijima, S. Masuda, Y. Sakisaka and M. Onchi, Surface Sci., 107 (1981) 31. 183 J.D. Feduk and M.J. Dignam, in ACS Symp. Ser. No. 137, eds. A.T. Bell and M.L. Hair (Amer. Chem. Soc. 1980) Ch. 5. 184 P. Dumas, R.G. Tobin and P.L. Richards, Surface Sci., 171 (1986) 555. 185 M.A. Chester, J. Pritchard and M.L. Sins, Adsorption Desorption Phenomena, ed. F. Ricca 171 172
186 187 188
(Academic Press, London, 1972), p. 2. N.D. Shinn, T.E. Madey,. J. Chem. Phys., 83 (1985) 5928. F. Zaera, E.B. Killin, J.L. Gland, J. Chem. Phys. Lett., 121 (1985) 464. J.G. Chen, M.L. Colaianni, W.H. Weinberg and J.T. Yates, Jr., submitted to Phys. Rev.
Lett., 189 190 191 192
M.L. Colaianni, J.G. Chen, W.H. Weinberg and J.T. Yates, Jr., to be published. K.W. Fresse, Surface Sci., 182 (1987) 85 and references therein. T. Matsushima, Surface Sci., 157 (1985) 297.
L. Surnew, G. Rangelov and G. Bliznakov, Surface Sci., 159 (1985) 441. 193 E. Shustorovich, Surface Sci., 150 (1985) L115. 194 P.H. Citrin, G.K. Wertheim and Y. Baer, Phys. Rev. Lett., 41 (1978) 1425. 195 R. Miranda, K. Wandelt, D. Rieger and R.K. Schnell, Surface Sci., 139 (1984) 430. 196 W.E. Egelhoff, J. Vac. Sci., Technol., A 2(2) (1984) 932. 197 A.B. Rives and R.F. Fenske, J. Chem. Phys., 75 (1981) 1293. 198 D. Spanjaard, C. Guillot, M.C. Desjongueres, G. Treglia and J. Lecante, Surface Sci. Reports, (1985) 1. 199 W.E. Egelhoff, Surface Sci. Reports, Vol. 6 (1986) 253. 200 G. Apai, R.C. Baetzold, E. Shustorovich and R. Jaeger, Surface Sci., 116 (1982) L191. 20 1 H.P. Bonze1 and R. Ku, J. Vac. Sci., Tech. 9 (1972) 665. 202 A.M. Lahee, W. Allison, R.F. Willis and K.H. Riedel, Surface Sci., 126 (1983) 654. 203 H. Niehus, Surface Sci., 145 (1984) 407.
86
204 205 206 207 208 209 210 211 212
213 214 215 216 217 218 219 220 221 222 223
G. Binnig, H. Rohrer, Ch. Gerber and E. Weibel, Surface Sci., 131 (1983)L379. H.P. Bonzel and S. Ferrer, Surface Sci., 119 (1982)234. S.R. Bare, P. Hofmann and D.A. King, Surface Sci., 144 (1984)347. P. Hofmann, S.R. Bare and D.A. King, Surface Sci., 117 (1982)245. P.R. Norton, J.W. Goodale and E.B. Selkirk, Surface Sci., 83 (1979)189. L.A. De Louis, E.J. White and N. Winograd, Surface Sci., 147 (1984)252. P. Hofmann, S.R. Bare and D.A. King, Physica Scripta, T4 (1983)118. D.W. Goodman and M. Kiskinova, Surface Sci., 105 (1981)L265. Metal-Support and Metal-Additive Effects in Catalysis, (eds. B. Imelik, C. Naccache, G. Coudurier, H. Praliaud, P. Meriaudeau, P. Gallezot, G.A. Martin and J.C. Vedrine, Elsevier, Amsterdam, 1982). C.H. Bartholomew, P.K. Agranal and J.K. Katzer, Adv. Catalysis, 31 (1982)135. M. Kiskinova, Surface Sci. Reports, 8 (1988)359 and the references therein. M. Kiskinova and D.W. Goodman, Surface Sci., 108 (1981)64,Surface Sci., 109 (1981) L555. M.Kiskinova, A. Szabo and J.T. Yates, Jr., J. Chem. Phys., 89(1988)7599. H.P. Bonzel, Surface Sci.Reports, 8 (1988)43 and Surface Sci. 226(1990)237. M. Kiskinova, J. Vac. Sci., Technol., A 5 (1987)852. M. Kiskinova, Surface Sci., 182 (1987)150. M. Kiskinova, Surface Sci., 11 1 (1981)584. G. Pirug and H. Bonnzel, Surface Sci., 199 (1988)371. P. Tomanek and K.H. Bennemann, Surface Sci., 127 (1983)Llll. R.G. Greenler, K. Burch, K. Kretzschmar, R. Klauber, A.M. Bradshaw and B.E. Hayden, Surface Sci., 152/153(1985)338. J. Segner, C.T. Campbell, G. Doyen and G. Ertl, Surface Sci., 138 (1984)505. P.J. Berlowitz, F.H.C. Peden and D.W. Goodman, J. Phys. Chem., 92(1988)5213.
224 225 226 C.H. Peden, D.W. Goodman, D.S. Blair, P.J. Berlowitz, G.B. Fisher and S.H. Oh, J. Phys. Chem., 92(1988)1563. 227 S.H. Oh, G.B. Fisher, J.E. Carpenter and D.W. Goodman, J. Cat. 100(1986)360. 228 D.W. Goodman, Appl. Surf. Sci., 19 (1984)1. 229 C.T. Campbell and D.W. Goodman, Surface Sci., 123 (1982)413. 230 H.P. Bonzel and H.J. Krebs, Surface Sci., 117 (1982)639. 231 P.A. Wesner, G. Linden and H.P. Bonzel, Appl. Surf. Sci., 26 (1986)335. 232 M. Logan, A. Gellman and G.A. Somorjai, J. Catalysis, 94(1985)60. 233 P.D. Szuromi, R.D. Kelley and T.E. Madey, J. Chem. Phys., 90(1986)6499.
87
CHAPTER 3
CATALYTIC ACTIVATION OF CO OVER SINGLE CRYSTALS
Jos A. Rodriguez and D. Wayne Goodman Department of Chemistry Texas A&M University College Station, TX 77843-3255 (USA)
88
3.1 INTRODUCTION One of the important challenges in basic and applied science is to understand how the atomic structure and composition of the surface of a catalyst determine catalytic activity and selectivity. Over the past 20 years many electron spectroscopies have been developed that permit the study, on the molecular level, of surfaces and adsorbates in ultrahigh vacuum (UHV)conditions (refs. 1,2). This has furthered our understanding of the chemistry of metal surfaces (refs. 1-3). Unfortunately, the conditions under which the analytical techniques of surface science can be applied are highly idealized. The very low pressures required for surface science studies are typically many orders of magnitude below the pressures used in practical catalytic processes. In order to overcome this problem, several laboratories have developed experimental systems which combine a high pressure reactor system with an UHV analysis chamber (refs. 4-9). The high pressure reactor allows the kinetics of catalytic reactions to be measured on a given surface, while analysis of the structure and composition of the surface both before and after reaction can be accomplished in the UHV analysis chamber. In many high-pressure/UHV studies a clean and well-defined, single-crystal plane is used to model a site or set of sites expected to exist on practical high-surface-area catalysts. This approach allows dircct comparison of reaction rates measured on single crystal surfaces with those measured o n more realistic supported metal catalysts, and also allows detailed study of structure sensitivity, the effects of promoters and inhibitors on catalytic activity, and, in certain cases, identification of reaction intermediates by post-reaction surface analysis. In this chapter, we review existing literature that deals with the catalytic activation of CO on single-crystal surfaces. The following reactions are considered: -
methanation
3H2 + CO + CH4 + H2O -
H2O + CO + H2 + C02 -
(3.1)
water-gas shift
(3.2)
methanol synthesis
CO + 2H2 -+ CH3OH
(3.3)
Examples are provided which demonstrate the relevance of single Crystdl studies for modeling the behavior of high surface area supported catalysts.
3.2 METHANATION OF CO ON SINGLE CRYSTAL SURFACES The reaction of low concentrations of CO in a mixture with H2 to form CH, was developed
as a gas-purification process in the 1950s (ref. 10). At the present time, the methanation reaction has a critical role in the production of synthetic natural gas from hydrogen-deficient carbonaceous materials (refs. 10-12). In addition, the reaction is an obvious starting point in studies of fuel and chemical synthesis from carbon sources (refs. 10-12). In the last ten years the methanation reaction
89
over single crystal surfaces has been the subject of many investigations. Here we review the results of these investigations. We begin with a discussion of the studies on monometallic (section 3.2.1.1) and bimetallic (section 3.2.1.2) single crystal surfaces. Next the results of studies dealing with the effects of electronegative (section 3.2.2.1) and electropositive (section 3.2.2.2) impurities on the kinetics of the methanation reaction are presented. Finally, we show studies concerned with metal-support interactions and the methanation reaction (section 3.2.3).
3.2.1
CO METHANATION O N CLEAN METAL SINGLE CRYSTALS
3.2.1. I
3.2.1.1 .I
Monometallic surfaces
Ni(lO0) and N i ( l l 1 )
The data in the Arrhenius plot of Fig. 3.la represent steady- state specific methanation rates (CH4 molecules/site-s) on both the Ni( 1 11) and Ni( 100) surfaces (ref. 13). At a given temperature the rate of production of CH4 over an initially clean catalyst crystal was constant, with no apparent induction period (ref. 13). The atomic configurations of the Ni(100) and Ni(ll1) surfaces are shown in Figs. l b and l c , respectively. The similarity between the data for the close-packed (1 11) and for the more open (100) crystal plane of Ni is evident in both the specific rates and activation energy (103 kJ/mol). The single crystal results are compared in Fig. 3.la with three sets of data taken from ref. 14 for alumina supported nickel catalysts. This comparison shows extraordinary similarities in kinetic data taken under nearly identical conditions. Thus, for the H2+C0 reaction over nickel, there is no significant variation in the specific reaction rates or the activation energy as the catalysts change from small metal particles to bulk single crystals. These data provide convincing evidence that the methanation reaction rate is indeed structure insensitive on nickel catalysts.Post-reaction analysis of the surface of the Ni crystal catalysts with Auger electron spectroscopy (AES) showed a low level of a carbon species and the absence of oxygen (refs. 6,13). The Auger lineshape for the carbonaceous residue was similar to that of nickel carbide, indicating that the carbon was in a “carbidic” form (ref. 6). Experiments were carried out studying the interaction of CO (24 torr, 0-1000 s exposures) with Ni(100) at different temperatures (450-800 K) (ref. 15). AES data showed the deposition of carbon on the surface and the absence of oxygen. Two kinds of carbon were formed on the surface: a carbidic type which occurs at temperatures c650 K and a graphite type at temperatures >650 K. The carbide type saturates at 0.5 monolayers, and can be readily removed from the Ni(100) surface by heating the crystal to 600K in 1 atm of H, with methane formed as the product. In contrast, the graphite type is a poison. The deposition of an active carbon residue and the absence of oxygen on the nickel surface following heating in pure CO is consistent with a well-known disproportionation reaction, the Boudouard reaction,
90
which has been studied on supported Ni catalysts (refs. 16,17) and on Ni films (ref. 18). On Ni(100), the carbon formation data from CO disproportionation indicates a rate equivalent to that observed for methane formation in a H&O mixture. Therefore, the surface carbon route to product is sufficiently rapid to account for methane production with the assumption that kinetic limitations are not imposed by the hydrogenation of this surface carbon. A set of experiments was performed (ref. 15) in which a Ni(100) surface was precarbided by exposure to CO and then treated with hydrogen in the reaction chamber for various times. This study showed that the rate of carbon removal in hydrogen compared favorably to the carbide formation rate from CO and to the overall methanation rate in H2/C0 mixtures. Thus in a H2+C0 atmosphere the methanation rate is determined by a delicate balance of the carbon formation and removal steps and neither of these is rate determining in the usual sense (refs. 6,13-15). More recent studies (ref. 19) using isotopically labeled CO have shown that the CO dissociation step is essentially unidirectional in that the rate of C(a) and O(a) recombination is insignificantly slow compared to the C(a) hydrogenation rate. Figure 2a shows the changes in the methanation reaction rate as the total pressure is increased from 1-120 torr at a fixed H,:CO ratio (ref. 13). At low temperatures the rates fall on the same straight line at all pressures. As the temperatures is increased, a deviation from linearity is seen - the higher the pressure the higher the deviation temperature. Accompanying this non-linear rate behavior is an increase in the active carbon level on the surface of the catalyst crystal (ref. 13).
a
Temperature a 0 0 7 0 0 600
(K)
5 0 0 450
N i t 100)0
single c r y s t a l
Nit11 1) 0
single c r y s t a l
lo4[
I
'
'
'
'
' ' ' ' '"1
1.2 1.4 1.6 1.8 2.0 2.2
Reciprocal temperature x
lo3
Fig. 3.1 (a) A comparison of the rate of methane formation (CH4 molecules/site-s) over single crystal nickel catalysts and supported Ni/A1203. Reaction conditions: 120 torr, H2/CO ratio = 4 (from ref. 13). (b) Atomic configuration of a Ni(100) surface. (c) Atomic configuration of a Ni(l11) surface.
91
I / T x 103 ( ~ - 1 )
I/T x 1 0 3 ( ~ - l )
Fig. 3.2 (a) Arrhenius plot of CH, synthesis on a Ni(100) catalyst at total reactant pressures of 1, 10 and 120 tom. The ratio H2/CO is 4 (from ref. 13). (b) Arrhenius plot of CH4 synthesis on a Ru( 110) catalyst at total reactant pressures of 1, 10 and 120 torr. The ratio H2/CO is 4. Data at two temperatures for a Ru(001) catalyst at 120 torr are plotted with the symbol X (from ref. 13).
It has been proposed (ref. 13) that this departure from the linearity of the rate in Fig. 3.2a and the accompanying increase in the surface carbon level is due to a decrease in the surface
Methanation Rate Versus Surface Carbon Level PI
Na 1 2
3
1
H2'CO
R.110 44 40 44
PIC.. (Torr) 1 14
36
4
40
5 6
40
61 54
40
120
12
14
504
10
I
100
Methanation Rate (N x CH4
Fig. 3.3 Methane production rate (molecules/Ni surface atom-s) at 625 K over a Ni(100) catalyst versus surface carbon coverage (under steady-state conditions). The H2/CO ratio and the total pressure (tom) for each point are indicated in the insert (from ref. 13).
92
coverage of hydrogen and thus a decrease in the rate of hydrogenation of surface carbon. According to the mechanism proposed above for CO methanation, if reaction conditions are altered such that the surface hydrogen concentration decreases (e.g. low H2 pressure and high temperature) then a correlation between decreasing methane yield and increasing surface carbide should be observed. This correlation holds very well as evidenced by the data in Fig. 3.3. Thus, the proposed reaction mechanism involving the dissociation of CO and the subsequent hydrogenation of the resulting carbon species (C(a)) accounts quite satisfactorily for the effect of pressure on the methanation rate, for the variation in the measured surface carbon level as reaction parameters are changed, and for the formation at characteristic temperature and pressure conditions of a catalyst-deactivating graphitic carbon.
3.2.1.1.2
R u ( l l 0 ) and Ru(001)
Figure 2b displays steady state specific rates for CO methanation on two faces of ruthenium: the zigzag, open (110), and the close-packed (001) (ref. 13). While the comparison is limited, it is clear that the H2+C0 reaction is quite similar in regard to the specific reaction rate and the activation energy for these two crystal planes of ruthenium. Thus, it appears that CO methanation is structure insensitive on ruthenium surfaces (ref. 13). Post-reaction surface analysis of the Ru crystal catalysts with AES showed the presence of carbidic carbon (ref. 13). The hydrogenation of this carbonaceous residue can be followed readily (ref. 13, 20). Furthermore, the specific rates of carbide formation on ruthenium surfaces from CO decomposition are equal to the rates of methane formation in CO+H2 mixtures (refs. 13,20). This
800K 700K
600K
500K
100 Torr HI
=-. Ni(100)
10
450
-
lo-' 5OOK
0.1 Ton CO
0.1 Ton CO
W(110)
1
: / -2-o:= 10-1
W(llOj.\
'\\
2 0 10-3-
10-2 :
lo4-
I
I
I
'\
1o
- ~
~~
1.2
1.4
1.6
~
1.8
2.0
2.2
IIT x lo3 (K-')
Fig. 3.4 (a) The methanation activity for W(110) compared to that for Ni(100), plotted in an Arrhenius fashion (from ref. 22). (b) The H, dependence of the methanation activity for W(110) compared to that of Ni(100) (from ref. 22).
93
experimental evidence suggests that the reaction mechanism for CO methanation on ruthenium surfaces is similar to that mentioned above for methanation over nickel surfaces. In Fig. 3.2 the variation of the reaction rate with pressure is very similar for the Ni(100) and Ru(l10) crystals. In both cases, the non-linear rate behavior is accompanied by an increase in the active carbon level on the surface of the catalyst crystal. It has been proposed (ref. 13) that the departure from linearity of the rate in Fig. 3.2 is due to a decrease in the surface coverage of hydrogen, which causes a decrease in the rate of hydrogenation of surface carbon. In fact, since the binding energy of hydrogen on Ru is lower than on Ni (refs. 13,21), the deviation from linearity should be expected at lower temperature for ruthenium. This is particularly evident in the 1 torr data of Figs. 2a and 2b.
3.2.1.1.3
W(II0)
The methanation activities for W(110) (ref. 22) and Ni(100) (ref. 12) are compared in Fig. 3.4 over a range of temperatures (Fig. 3.4a) and H2 partial pressures (Fig. 3.4b). The data clearly indicate that W(110) is an active methanation catalyst, with an activity that in some cases can surpass the activity of Ni(100). Plotting the data in an Arrhenius fashion (Fig. 3.4a) yields an apparent activation energy of 56 H mol-1 for W(l lo), as compared to 103 kJ mol-I for Ni(100). The activation energy over W(110) is in reasonable agrezment with the value of 63 !d mo1-I observed on catalysts prepared by decomposition of W(CO)6 on alumina (ref. 23). Auger electron spectra of the W(110) surface after steady- state reaction conditions indicated that the active methanation surface was highly carbidic (ref. 22), in contrast to the case of nickel, where the active methanation surface is the metal itself with only a low surface coverage (0.05-0.1 monolayers) of carbidic carbon species present (refs. 6,13). The idea that the active W surface is carbidic is consistent with the significant hydrogenation activity reported for W carbide catalysts (ref. 22).
3.2.1.I.4 Rh(ll1) CO hydrogenation over clean Rh(ll1) was studied at a temperature of 573 K and partial pressures of 4.5 atm of H2 and 1.5 atm of CO (ref. 24). Under these reaction conditions the Rh catalyst produced primarily methane (90 wt%) at an initial rate of 0.15 molecules/site-s. Small amounts of C, and C3 hydrocarbons were also formed, but no oxygenated hydrocarbons were detected. The rates of formation of all the products were found to be the same on the Rh( 111) single crystal and on a polycrystalline Rh foil suggesting that CO hydrogenation is structure insensitive on these surface (ref. 24). The results of AES showed the presence of =1 monolayer of carbon on the surface of the Rh catalysts after 3 hours of reaction (ref. 24). The close proximity of the Rh 256 and 302 AES peaks to the C 272 AES peak prevented an analysis of the lineshape of the carbon peak in order to determine the chemical nature of the carbonaceous residue.
3.2.1.1.5 F e ( l l 1 ) The hydrogenation of CO on the (1 11) face of iron was examined at partial pressures of 3.5 atm of H2 and 1.5 atm of CO and at a temperature of 573 K (ref. 25). The major product of the
94
reaction was methane (=70 wt%), which was formed at an initial rate of 1.35 molecules/site-s. The formation of C2 (=20 wt%), C3+ and C4 products was also observed. 32.1.1.6
Mo(I00) Figure S shows the effects of temperature and pressure on the rate of CO methanation over Mo(100). On this surface, the hydrogenation of CO produced primarily methane (=90mol%), ethene and propene (ref. 26). An activation energy of =lo0 kJ/mol was found for the methanation reaction on Mo( 100). The observed rate law for methanation (see fig Sb) is given by:
r(CH4) = kP(C0) + 0.32 P(H2)l.O
(3.5)
The positive power of rate dependence on the pressure of CO is unusual since the methanation rate has a negative-order dependence on CO partial pressure over Ni, Ru, Fe and Co catalysts (refs. S,26,27). Auger electron spectra taken after the hydrogenation reaction indicate that the "active" surface is covered by a submonolayer of a carbidic carbon species. The reaction is poisoned as the carbidic species is converted to graphitic carbon. The rate of poisoning is determined by the ratio of co to H2 in the reaction mixture and by the reaction temperature (lower CO/H2 ratios and lower temperatures prolong the lifetime of the active catalyst). The following set of elementary steps was proposed for the methanation reaction on Mo(100) (ref. 26):
.
l\ - II \\ \
COtH2 1
3
1320 Tor,
I0"t
-
\
< E,-
24 kcol mole
I
I 16
ia
,
inx i o 3 ( ~ ' )
Id' \I
,
@-
I
Partial Pressure (atm)
2 0
Fig. 3.5 (a) Arrhenius plot for CO methanation on Mo(100) (from ref. 26). (b) Rate of formation of methane over Mo( 100) versus the partial pressure of each reactant. Constant H2 pressure of 3 atm for determination of CO dependence, and constant CO pressure of 3 atm for determination of H2 pressure dependence. In all the cases the total pressure was 6 atm, with nitrogen or argon used as a buffer gas (from ref. 26).
95
(3.6) (3.7)
(3.8) (3.9) (3.10)
In terms of this model the rate-determining step is reaction (9) and all the steps preceding it are in quasiequilibrium. A mathematical analysis of this kinetic model (ref. 26) leads to a rate expression of the form:
which is in reasonable agreement with the experimental results of Fig. 3.5b. Reactions (6) to (10) probably take place on top of a carbidic overlayer (ref. 26). This overlayer will deactivate by forming graphite on the surface, which will block the reaction sites. No differences in either rates or product distributions were observed between CO hydrogenation on Mo(100) and over polycrystalline Mo foils (ref. 26). Thus, the reaction does not appear to be structure sensitive o n molybdenum surfaces. 3.2.1.1.7 Summary Table 3.1 shows the apparent activation energies observed for CO methanation on different single crystal surfaces. In general, the energies are close to the value of 110 kJ/mol. The only exception is W( 1lo), with an apparent activation energy of 56 kl/mol. The studies reviewed in this
section provide convincing evidence that the methanation reaction is structure insensitive on surfaces of Ni, Ru, Rh, Fe and Mo. On these metals the methanation of CO occurs in the presence of an active carbidic overlayer. The transformation of this overlayer into graphite leads to a decrease in the catalytic activity of the metal surfaces. 3.2.1.2
Bimetallic surfaces
Catalytic properties of metal surfaces can be altered greatly by the addition of a second transition metal (ref. 28). In many cases, mixed-metal systems are superior over their single-metal counterparts in terms of catalytic activity and/or selectivity (ref. 28). Many fundamental studies have focussed on trying to understand the roles of “ensemble” and “ligand” effects in bimetallic catalysts (refs. 28,29). Ensemble effects are defined in terms of the number of surface atoms needed for a catalytic process to occur. Ligand effects refer to those modifications in catalytic activity or selectivity that are the product of electronic interactions between the components of a bimetallic system. In gathering information to address these issues, it has been advantageous to simplify the problem by utilizing models of bimetallic catalysts such as the deposition of metals onto single-crystal substraks. Work on ultrathin metal films supported on well-defined metal surfaces
96
(Cu on Ru(0001) (refs. 30-33); Cu, Ni, Pd and Pt on W(110) and W(100) (refs. 34-38); Fe, Ni and Cu on Mo(l10) (refs. 39,40); and Fe and Cu on Re (0001) (refs. 40,41)) has shown that a metal atom in a matrix of a dissimilar metal can be significantly perturbed, and that this perturbation can dramatically alter the chemical and electronic properties of both constituents of the mixed-metal system. The studies reviewed here are part of a continuing effort to identify those electronic and structural properties of bimetallic systems which can be related to their superior catalytic abilities.
3.2.1.2 .I
NiIW(II0) and NiIW(IO0)
The N N ( 1 0 0 ) and N N ( 1 1 0 ) systems are particularly interesting because they involve the
Table 3.1
Metal Single Crystal Catalysts: Specific Activities and Activation Energies in CO Methanation
Catalyst
Ref.
T
K Ni(100) Ni(ll1)
450-700
96
24
550-700
96
24
440-540
60
Ru( 110)
500-700
96
20 24
Ru(001) W(110)
550-650
96
24
480-750 510-640
1100
Mo( 100)
100
0.1 220
0.0002-a 0.04-5 0.0002-0.05 0.002- 10 0.01-0.8 0.002-0.5 0.00 1-0.2
1 I
13 103
13
111 123 =120
84 13 13 22 26
aTOF = turnover frequency = CH4 molecules/site-s
addition of an active metal for CO methanation (Ni) to relatively inactive tungsten surfaces. At =lo0 K, Ni is adsorbed layer by layer on W(110) and W(100) (ref. 36). The results of low-energy electron diffraction (LEED) indicate that at coverages up to 1 ML the Ni films grow pseudomorphically with respect to the W(110) and W(100) substrates (ref. 36). This growth pattern leads to Ni monolayer densities on W(110) and W(100) which are 21% and 38% less than the corresponding monolayer densities for Ni(ll1) and Ni( loo), respectively. Specific rates of CH4 production, expressed as turnover frequencies (CH4 molecules/Ni atom-s), over Ni covered W(110) and W(100) surfaces are shown in Fig. 3.6 (ref. 38). Under the experimental conditions of this figure, the reaction rates on the clean W surfaces were =lo2 times lower than on the Ni covered surfaces. Fig. 3.6a shows that for a total pressure of 120 torr, Arrhenius behavior is observed over the entire temperature range studied (450-700 K) as the CH,
97
101,
,
,
CO
f
,
,
-
I
, . ,
,
H2-CH.
, . , .
..
3
0.1 YL M I
.
0 0 . 4 YL MI
>
.
a W >
.
I- -
:
w
.
0
lo-'
'
1.0
0.5 YL NI 0.11 Y L NI
+ 1.0 YL M I 0
0
" 1.2 1.4
'
I
1.6
.
I
.
1.8
,
N I I W l110l
I
I
2.0
2.2
.
.
'
2.4
1000/T ( K - ' )
Fig. 3.6 (a) Arrhenius plot for CH4 synthesis over several different Ni coverages on W(110) and W(100) at a total reactant pressure of 120 torr (H2/CO=4) (from ref. 38). (b) Arrhenius plot for CH, synthesis over several different Ni coverages on W(110) at total reactant pressures of 1, 10 and 120 torr (HZ/CO=4) (from ref. 38).
production rates v x y by almost 3 orders of magnitude. The similarity between NiN(110) and Ni/W(100) at all the coverages studied is evident in both the turnover frequencies and activation energy, 77+4 kJ/mol. The activation energy for the Ni covered W surfaces is lower than the value of 103 kJ/mol reported for Ni(100) and Ni(ll1) (ref. 13). However, the specific rates for CO methanation on Ni/W(110) and Ni/W(100) correlate well with those observed on supported Ni catalysts, Ni single crystals and Ni films (ref. 38). These results are further manifestations of the structure insensitive behavior of the CO methanation reaction and suggest that the mechanistic steps which control the rate of CO hydrogenation are the same in all these surfaces. Lowering the total pressure has a significant effect on the. rate of methane production (Fig. 3.6b) for Ni supported on W(110) (ref. 38). A similar effect was observed for Ni(100) (see Fig. 3.2a), and was attributed to a decrease in the concentration of atomic hydrogen on the surface as the pressure was lowered and the temperature was increased (see section 2. 1. 1 ) (ref. 13). The departure from Arrhenius behavior occurs at lower temperature for Ni/W(l 10). This correlates (ref. 38) with the fact that the activation energy for H2 desorption from NinV(ll0) (=71 kJ/mol (ref. 36))is lower than that from Ni(100) (96 kJ/mol) (ref. 42). 3.2.1.2.2
CulRu(OO1)and AglRh(ll1)
A bimetallic system that has been extensively studied in supported catalyst research is copper on ruthenium (refs. 28,32,33). The immiscibility of copper in ruthenium circumvents the complication of determining the three dimensional composition. The adsorption and growth of
98
copper films on the Ru(0001) surface have been examined (refs. 30,31,43,44) by work function measurements, LEED, AES and TPD. The experimental evidence indicates that for submonolayer depositions at 100 K the Cu grows in a highly dispersed mode, forming 2-D islands pseudomorphic to the Ru(001) substrate upon annealing to 300K. The pseudomorphic growth implies that the copper-copper bond distances are strained approximately 6% beyond the equilibrium bond distances found for bulk copper. Copper surfaces are inactive catalysts for CO methanation. A study of the rate of CO methanation over Cu/Ru(OOl) (ref. 33) indicates that copper merely serves as an inactive diluent, blocking the active sites of the ruthenium surface in a one-to-one basis. Similar results have been found in analogous studies (ref. 45) introducing silver onto a Rh( 111) methanation catalyst. 3.2.1.2.3 ColW(l10)and ColW(IO0) Cobalt forms pseudomorphic monolayers on W(110) and W( 100) which are thermally stable to 1300 K (ref. 46). The cobalt overlayers are geometrically strained with respect to bulk cobalt surfaces. The pseudomorphic monolayer of Co/W(llO) has an atomic density 21% less than Co(0001), while the pseudomorphic monolayer of Co/W(lOO) has an atomic density 45% less than Co(O001) (ref. 46). However, the CO hydrogenation activity of these Co/W surfaces is very similar (ref. 46). A fact that suggests that CO hydrogenation is structure insensitive on cobalt surfaces (ref. 46).
AES spectra show the after-reaction Co/W surfaces to have high coverages of both carbon and oxygen, with carbon lineshapes characteristic of carbidic carbon (ref. 46). The catalytic activity is apparently not correlated with surface carbon level (ref. 46).
3.2.2 CO METHANATION CHEMICALLY MODIFIED SURFACES The addition of impurities to a metal catalyst can produce dramatic changes in the activity, selectivity and resistance to poisoning of the catalyst. For example, the selectivity of some transition metals can be altered greatly by the addition of light metals such as potassium, and the activity can be reduced substantially by the addition of electronegative species such as sulfur. Although these effects are well-recognized in the catalytic industry, the mechanisms responsible for chemical changes induced by surface additives are poorly understood. An important question concerns the underlying relative importance of ensemble (steric or local) versus electronic (nonlocal or extended) effects. A general answer to this question will improve our ability to design efficient catalysts. Catalyst deactivation and promotion are extremely difficult questions to address experimentally (ref. 47). For example, the interpretation of related data on high-surface area supported catalysts is severely limited by the uncertainty concerning the structural characterization of the active surface. Specific surface areas cannot always be determined with adequate precision. In addition, a knowledge of the crystallographic orientation, the concentration and the distribution of impurity atoms, as well as their electronic states is generally poor. The use of metal single crystals in catalytic reaction studies essentially eliminates the difficulties mentioned above and
99
allows, to a large extent, the utilization of a homogeneous surface amenable to study using modem surface analytical techniques. In this section we review studies dealing with the effects of electronegative and electropositive surface impurities on the rates of CO methanation over single crystal catalysts. Although the studies to date are few, the results appear quite promising in addressing fundamental aspects of catalytic poisoning and promotion.
3.2.2.1 Electronerative imourities Impurities whose elecconegativities are greater than those for transition metals generally poison a variety of catalytic reactions, particularly those involving H2 and CO. Of these poisons sulfur is the best known and is technologically the most important (refs. 10,47). 3.2.2.1 .I Atomic chlorine, sulfur and phosphorus on Ni(100) The effects of preadsorbed C1, S and P atoms on the adsorption-desorption of H2 and CO on Ni(100) have been extensively studied (refs. 48-52) using Auger electron spectroscopy, low-energy electron diffraction and temperature programmed desorption. Fig. 3.7 shows the variation of the saturation coverage of H2 and CO on Ni(100) with the coverage of C1, S and P. Both CO and H, adsorption decrease k 3 CI markedly in the presence of surface tu)0 0 . 3 1 0 s impurities. The effects of P, however, AP 0 are much less pronounced than for C1 and S . The similarity in the atomic radii of C1, S and P (0.99, 1.04 and 1.10, respectively (ref. 53)) suggests a 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 A O D I T I V E COVERAGE fMLl relationship between electronegativity and the poisoning of chemisorptive properties by these surfaces CI impurities (refs. 48-50). Impurities .S A p that are strongly electronegative with respect to nickel, C1 and S , modify the chemisorptive behavior far more strongly than would result from a simple site blocking model. The initial effects of these impurities as shown in Fig. 3.7 indicate that a 0 0.1 0.2 0.3 0.4 n single impurity atom can successfully ADDITIVE COVERAGE (ML) poison more than just its Fig. 3.7 Variation of the saturation coverage of CO nearest-neighbor nickel atoms. This (part a ) and H2 (part b) on Ni(100) with the precoverage of C1, S and P (from ref. 50).
100
type
of
poisoning
supports
an
interaction that is primarily electronic in nature. The (refs. 48-50)
experimental indicate
results
that
the
presence of electronegative C1, S and P atoms causes a reduction of the SULFUR ON Ni(100) AT 0.5 MONOLAYERS
SULFUR ON Ni(100) AT 0.25 MONOLAYERS
(b) w I-
< a z
10‘
g<
adsorption rate, the adsorption bond strength and the capacity of the Ni(100) surface for CO and H2 adsorption. In general, these poisoning effects become stronger with increasing electronegativity of the adsorbed impurity. Kinetic studies (refs. 48-50) have been camed out for
i
B
CO
methanation
over
Ni(100)
surfaces covered with chlorine, sulfur and phosphorus impurities. Fig. 3.8 shows the rate of CO methanation as ,
10’ 0
1
0.1
,
1
,
0.2
ADATOM COVERAGE
1
0.3
,
1
,
0.4
(
IN MONOLAYERS
Fig. 3.8 Plot -of the rate of methanation (CH, molecules/site-s) over a sulfide and phosphided Ni(100) catalyst at 120 torr and a pressure ratio of H2 to CO of 4 (from ref. 50).
a function of sulfur and phosphorus coverage over a Ni(100) catalyst at 120 torr and a H2/C0 pressure ratio equal to 4. In the case of C1 overlayers
no
change
in
the
methanation rate was observed. At the reaction conditions of Fig. 3.8, H, reacts with the adsorbed C1 forming HCl.
This reaction “cleans” the Ni(100) surface and prohibits the study of the effects of C1 on the methanation kinetics. The results presented in Fig. 3.8 correlate very well with the studies on the effects of S and P overlayers upon the adsorption of H2 and CO on Ni(100) ((refs. 48-50) and Fig. 3.7.) Fig. 3.8 shows a non-linear relation between the sulfur coverage and the methanation rate. A steep drop in catalytic activity is observed at low sulfur coverages, and the poisoning effect maximizes quickly. A similar reduction of methanation activity by sulfur poisoning has been observed for alumina supported nickel catalysts (ref. 54). The initial attenuation of catalytic activity by sulfur suggests that ten or more equivalent nickel sites are deactivated by one sulfur atom. There are two possible explanations for this result: (1) an electronic effect that extends to the next-nearest-neighbor sites or (2) an ensemble effect, the requirement being that a certain number of surface atoms is necessary for a reaction to occur. If extended electronic effects are significant,
101
0
0
0.1
0.2
0.3
0.4
SULFUR COVERAGE (monolayer)
0.5
0
0.1
0.2
0.3
0.4
0.5
SULFUR COVERAGE (monolayer)
Fig. 3.9 Methanation rate as a function of sulfur coverage on: (a) Ru(001) (from ref. 55), and (b) Rh( 111) (from ref. 45) catalysts. P=120 torr, H2/CO=4, reaction temperature: 600 K.
then the reaction rate is expected to be a function of the relative electronegativity of the poison. In contrast, if an ensemble of ten nickel atoms is required for the critical step of methanation, then altering the electronegative character of the poison should produce little change in the poisoning of the reaction. Substituting phosphorus for sulfur (both atoms are approximately the same size (ref. 53)) results in a marked change in the magnitude of poisoning at low coverages as shown in Fig. 3.8. Phosphorus, because of its less electronegative character, effectively poisons only the four nearest-neighbor metal atom sites. These results support the conclusion that extended electronic effects do play a major role in catalytic deactivation by sulfur.
3.2.2.1.2 Sulfur on Ru(OOl), Rh(ll1) and W ( I I 0 ) Fig. 3.9 presents the effects of sulfur coverage on the rate of CO methanation on Ru(001) and Rh(ll1) catalysts (refs. 4535).As for the case of Ni( loo), a precipitous drop in the catalytic activity is observed for low sulfur coverages. The initial changes in the rates suggest that more than ten Rh or Ru atom sites are deactivated by one sulfur atom. Kinetic data for sulfur-covered W( 110) surfaces indicate that the activation energy of the methanation reaction does not change with sulfur coverage (ref. 56). In this respect tungsten is similar to nickel (ref. 48). Sulfur decreases the rate of CO methanation on W(110) (ref. 56). In Fig. 3.10 the relative change in rate is plotted as a function of sulfur coverage. For comparison, the data for the Ni(100) surface (ref. 48) are included. While sulfur clearly exhibits long-range effects on nickel, the operation of long-range effects over tungsten occurs only at the lowest coverages. At very low sulfur coverages (~0.03 ML), the decrease in activity is quite steep and extrapolates to between 10 and 12 atoms sites deactivated per sulfur atom adsorbed (ref. 56). Apparently the adsorption of sulfur occurs initially in a random, disordered fashion, so that there is little overlap of the inhibiting effect of sulfur atoms on open sites. As the sulfur coverage increases, clustering into islands occurs (ref. 56).
102
1.o
0 I
750 K -Reaction A 700 K PMz =
100 Torr
=
1 Tom
0 500 K
0.8
Pco
Temperatures
Q) e
m
Inhibition of Methanation mer W (1101 by Adsorbed
a .- 0.6
-
Sulfur
0
m
Q
x Similar Data for Ni (1001
a
.-2
600 K
.m -
0.4
P H2
=
Pco:
96 Torr 24 Torr
Q
a 0.2
--+-
0.0 0.0
0.1
0.2
0.3
0.5
0.4
0.6
Sulfur Coverage
Fig. 3.10 Relative change in methanation rate over W(110) as a function of sulfur coverage (from ref. 56. For comparison the data for Ni(100) is also shown (from ref. 50).
-\ H2 / C 0 = 4 / 1
P= 120 TORR
0
0.05
0.10
FolAsslM c
0.15
o m eN
0.20
I
12
'
1.4
l!S
,
l!S ' 20
&
1TT x l$(K-l)
Fig. 3.1 1 (a) Relative methanation rate as a function of potassium coverage at various reaction conditions: A P(CO)=l.O torr, P(H2)=99.9 torr, T=600 K, 0 P(C0)=24 tom, P(H2)=97.6, T=538 K 0 P(C0)=24 tom, P(H2)=96 tom, T=600 K; 0 P(C0)=24 tom, P(H2)=96torr, T=594K (from ref. 57). (b)A comparison of the rate of methane synthesis over a clean single crystal Ni(100) catalyst with the rate over a potassium doped catalyst. Total reactant pressure is 120 torr, H2/CO=4 (from ref. 57).
103
Thus, the inhibiting effect of additional sulfur atoms is diminished due to overlap with the effect of previously adsorbed sulfur atoms.
3.2.2.2
Electrooositive impurities
We have discussed above the role of electronegative impurities in poisoning Ni( IOO), Ru(001), Rh(ll1) and W(110) toward methanation activity. These results have been ascribed, to a large extent, to an electronic effect. In the context of this interpretation it is expected that an electropositive impurity might have the opposite effect, i.e. to increase the methanation activity of a metal surface. A study of CO hydrogenation over potassium covered Ni(100) (refs. 50,57) has shown that this is not the case, although certain steps in the reaction mechanism are strongly accelerated by the presence of the electropositive impurity. 3.2.2.2.1 Potassium on Ni( 100) Fig. 3.1 1 shows kinetic measurements of CO methanation over a Ni(100) catalyst containing well-controlled submonolayer quantities of potassium adatoms (ref. 57). These data indicate a decrease in the steady-state rate of methanation with potassium coverage. A coverage of about 0.22 MJ- of potassium would be sufficient to terminate the reaction completely. The presence of K did not alter the apparent activation energy associated with the kinetics, as shown in the Arrhenius plot of Fig. 3.1 lb. However, the potassium did change the steady-state coverage of active carbon on the catalyst. This carbon level changed from 10% of a monolayer on the clean catalyst to 30% of a monolayer for a catalyst covered with 0.1 ML of potassium (ref. 57). As shown in Fig. 3.12, adsorbed potassium caused a marked increase in the steady-state rate and selectivity of Ni(100) for higher hydrocarbon (MW>16) synthesis (ref. 57). At all the temperatures studied, the overall rate of higher hydrocarbon production was faster on the potassium-dosed surfaces, so that potassium may be considered a true promoter with respect to this reaction, Fischer-Tropsch synthesis. The effects of potassium upon the kinetics of
co
hydrogenation over Ni( 100) (i.e. a decrease in the rate of methane formation and an increase in the rate of higher hydrocarbon production) are similar to those reported for high-surface-area supported Ni catalysts (refs. 58-59). This agreement between bulk, single crystal Ni and supported Ni indicates that the major mechanism by which potassium additives alter the activity and selectivity of industrial catalysts is not related to the support material, but that it is rather a consequence of direct K-Ni interactions. Adsorbed potassium causes a marked increase in the rate of CO dissociation on a Ni(100) catalyst (ref. 57). The increase of the initial formation rate of “active” carbon or carbidic carbon via CO disproportionation is illustrated in Fig. 3.13. The relative rates of CO dissociation were determined for the clean and potassium covered surfaces by observing the growth in the carbon Auger signal with time in a CO reaction mixture, starting from a carbon-free surface. The rates observed in Fig. 3.13 are the observed rates of carbon formation extrapolated to zero carbon coverage. The presence of K adatoms leads to a reduction of the activation energy of reactive carbon formation from 96 kJ/mol on clean Ni( 100) to 42 kJ/mol on a 10% potassium covered surface (ref. 57).
104
In spite of increasing
PROWCT DlSTRlUTlON OVER A Nl(100) CATALYST
nl l
the rate of CO dissociation or carbide buildup, potassium decreases the overall rate of methanation. This reduction in methanation activity must be related to a poisoning of either the hydrogen adsorption or the hydrogen addition steps (ref. 57). The enhancement of
-
ETHYLENE
steady-state carbide coverage caused by potassium favors C-C bond formation and the synthesis of heavy hydrocarbons (MW>16).
ETHANE
Fig. 3.12 A comparison of the product distributions (weighi percent) observed for clean and K-doped catalysts at T=500 K, H2/C0=4, and a total pressure of 120 torr. Potassium coverage=O.10 ML (from ref. 57).
3.2.2.3 Related Theorv CO is generally thought to be adsorbed on transition-metal surfaces by the Blyholder mechanism (refs. 60,61), which involves e-donation of electron density from CO into the unoccupied metal orbitals and n-back-donation of electron density from occupied metal orbitals into the lowest unoccupied
z 0 t-
z *
4
Pz
0 m U
3
a
u w
5 c
u 4
2
-1 w
=
o
I
I
.02
.04
I
.06
I
.00
I
.10
POTASSIUM COVERAGE (ML)
Fig. 3.13 The relative initial rate of reactive carbon formation from CO disproportionation as a function of potassium coverage. P(C0)=24 torr, T=500 K. (From ref. 57).
molecular orbitals (2n*) of the CO molecule. The mechanism is similar to that observed for CO bonding in transition metal compounds (ref. 62). Theoretical results with the constrained space orbital variation (CSOV) method show that, in this type of synergistic bond with e-donationln-back-donation, n-back-donation is energetically more important in determining the character of the bond than is e-donation, at least for the case of Cu surfaces (refs. 63-65).
105
Recently, inverse photoemission results have supported the predominant importance of 2x* back-donation in CO chemisorption on Pd and Ru (ref. 66). The thought that the antibonding 2 x orbitals of CO are populated upon adsorption is consistent with the results of HREELS (refs. 67,68), which show that the C - 0 stretching frequency and the C - 0 force constant of CO adsorbed on metals are lower than those of free CO. The standard picture used to describe the effects of electron-transferring species upon CO chemisorption is an extension of the basic Blyholder model. Electropositive impurities donate charge to the metal. This excess charge is partially accommodated in increased n-back-donation to the 2n* orbitals of CO. This increases the metal-CO bond strength, while decreasing the C - 0 bond strength. Opposite effects are expected for coadsorption with electronegative impurities. In a few coadsorption cases, this general picture has been to some extent substantiated by calculations with different quantum-chemical methods (refs. 69-75,107). Theoretical work has been undertaken to address directly the predicted magnitude of the near surface electronic perturbations by impurity atoms. Early work was concentrated on the indirect interactions between adsorbates which occur via the surface conduction electrons (refs. 76-78). These calculations suggested that atom-interactions through several lattice spacings can occur. Recent theoretical studies have expressly addressed the surface electronic perturbations by sulfur (ref. 79) as well as by C1, P and Li (ref. 80). The sulfur-induced total charge density vanishes beyond the immediately adjacent substrate atom site. However, the Fermi-level density of states, which is not screened, and which governs the ability of the surface to respond to the presence of other species, is substantially reduced by the sulfur even at nonadjacent sites. The results for several impurities indicate a correlation between the electronegativity of the impurity and its relative perturbation of the Fermi-level density of states (refs. 79,80), a result which could be very relevant to the poisoning of CO methanation by S and P as discussed above. Finally, an alternate model, which produces the same final results, involves an electrostatic, through-space (as opposed to through-metal) interaction between the charge distribution of the coadsorbed species (refs. 81,82). Both sets of theories, that is, “through-metal” or “through- space”, are consistent with adsorbate perturbations
sufficiently large
to effect chemically
significant
changes at
next-nearest-neighbor metal sites. This perturbation length is sufficient to adequately explain the observed poisoning of catalytic activity by surface impurities discussed above.
3.2.3
METAL-SUPPORT INTERACTIONS AND CO METHANATION
The early concept of a support or a carrier was that of an inert substance that provided a means of spreading out an expensive catalyst over a large surface area. However, the support may actually modify the activity of the catalyst, depending upon the reaction and reaction conditions (refs. 10,83,106). The oxidic support materials (e.g. La203, Cr2O3, ZnO, MgO, Ti02, ZrO2) can favorably or adversely influence the performance of a metal in a particular catalytic process (refs. 83,106). Titania (Ti02) is a typical example of an “interacting” support (refs. 83,106).
106
0 w
e
Ni/TIO=.ltOOL
l+/CO=60Tar/20Ton
0
8
16
24
NI THICK N E S S ~ ~ I
1
20
22
J
103/TlK)
Fig. 3.14 (a) The methane yield from Nfli02(100) as a function of the average Ni thickness. P (H2)=60tom, P(C0)=20 tom, T=l9O0CLfromref. 84).(b) Anhenius plots of the methane turnover number (CH, molecules/site-s) for Ni(ll1) and 5 Nf1102(100),P(H2)=60 torr, P(C0)=20 tom (from ref. 84). Surface science methods have been applied to study the CO hydrogenation activity of nickel overlayers on the Ti02(100) surface (ref. 84). Results of ultraviolet photoelectron spectroscopy indicate that there is an electron transfer from Ti02(100) to Ni when Ni is deposited onto a reduced Ti02(100) surface (ref. 84). When the Ni/Ti02(100) surface was used as a methanation catalyst, the
CH4 yield varied as a function of the Ni coverage, as shown in Fig. 3.14a. At a temperature of 19OoC, an average Ni thickness of =5A gave optimum activity. An Arrhenius plot of the specific rate of methane formation over the 5 w Ni-covered Ti02(100) surface is included in Fig. 3.14b. The methane yields from the Ni/Ti02( 100) catalyst are 3.3-3.7 times that from a pure Ni(ll1) catalyst. The apparent activation energy for methane production over the Ni/Ti02(1OO) surface (105.5f2.5 Id/mol) is very close to that seen over Ni(ll1) (1 11.8k3.8 kJ/mol) (ref. 84). A study of the methanation activity of a Ni(ll1) surface containing controlled amounts of TiO, (x=1.0-1.5), showed that the nickel catalyst is optimally promoted at a titanium coverage of 4 . 1 monolayer, displaying an activity and product distribution similar to those seen over 8 A Ni-covered Ti02(100) catalysts (ref, 85). The fact that TiOx/Ni(lll) and Ni/Ti02(100) are so similar in their catalytic behavior, suggests (ref. 85) that in high-surface area Ni/TiO2 catalysts, TiO, species diffuse from the support material to the nickel. Dispersion of the oxide on the metal and the formation of nickel-titanium bonds modify the catalytic properties of the surface.
3.3 WATER-GAS SHIFT REACTION ON SINGLE CRYSTAL SURFACES The water-gas shift reaction (CO+H20+H2+C02) is widely used industrially in various hydrogen production or enrichment processes (refs. 10,86). Many materials are able to catalyze this reaction (refs. 10,86). Originally, the most commonly employed industrial catalysts were based on iron oxides and operated at high temperatures (570-820°C) (refs. 10,86). A substantial improvement came about by the deveIopment of Cu/ZnO-based catalysts, which operate at relatively low temperatures (470-530°C) and allow higher thermodynamic conversions (refs. 10,86). In these
107
(refs. 10,86). In these catalysts, copper is the active species, and the principal role of the zinc oxide is to act as a support for the copper (refs. 10,87-91). It is not clear why ZnO is a superior support (refs. 86-91). In this section we review the results of studies in which the kinetics and mechanism of the water-gas shift reaction have been investigated using the modem techniques of surface science and copper single crystals.
3.3.1 KINETICS OVER Cu(l10)and Cu(ll1)CATALYSTS Figure 15 illustrates the effect of temperatures on the rate of the water-gas shift reaction over Cu( 110) (ref. 88) and Cu(ll1) (ref. 87) surfaces. For the same geometric area, Cu( 110) is 2.5-7.0 times more active than Cu(ll1) between 550 and 650 K (ref. 88). On a “per copper surface atom” basis, this difference is 1.63 larger due to the higher surface atom density of Cu(ll1). The slopes in Fig. 3.15 give apparent activation energies of 10 and 17 kcal/mol for the water-gas shift reaction on Cu( 110) and Cu(ll1). These results indicate that the reaction is structure sensitive on copper surfaces. A fact that is consistent with data for high-surface area supported catalysts, which show an increase in the catalytic activity as the TEMPERATURE / K Cu particle size decreases 700 650 600 550 (ref. 91). High area 0.4 I I I supported and unsupported catalysts show apparent activation energies in the range of 13 to 16 kcal/mol (ref. 87), well within the range between Cu( 110) and Cu( 111) (10-17 kcal/mol). A study of the influence of CO and H 2 0 partial pressures on the reaction rate over Cu( 110) (ref. 88) and Cu(ll1) (ref. 87), revealed that on both surfaces the rate is nearly independent of P(C0) (order 0) and strongly positive order in H20 (0.5-1.0). Analysis of the surface of the catalysts with AES and
-
0.2 -
0.0 C A
2 -0.2 a u
-
-0.4
-
v)
-0.6
-
-0.8
-
W
E“
-.
Ea = 17kcaVmole
0 v 1
-.W & 4 Y
-1.2 -
-1.0
0
H2
A CO,
-1.61 1.4
I
1.5
I
1.6
I
I
1.7
1.8
1.9
1000 K / TEMP. Fig. 3.15 Variation of the water-gas shift reaction rates on Cu(ll0) and Cu( 111) with temperature, in Arrhenius form. P(H20)=10 torr and P(C0)=26 torr. (From ref. 88 and ref. 87).
108
XPS after reaction conditions showed essentially oxygen-free copper surfaces (refs. 87,881. To further prove that Cu surfaces are fully reduced under water-gas shift conditions at low conversion, experiments were carried out using a heavily preoxidized Cu(ll1) crystal (ref. 87). After a few minutes reaction time, the surface displayed a fully metallic Cu(2p) XPS spectrum, and gave no oxygen signal in AES or XPS (ref. 87). This suggests that metallic Cu is the active ingredient for high-surface area Cu/ZnO or Cu-based catalysts (ref. 87). The following mechanism has been proposed (ref. 88) for the water-gas shift reaction on Cu( 110) and Cu( 111): (3.12) (3.13) (3.14) (3.15) (3.16) (3.17) In this mechanism the rate-determining step involves 0 - H bond cleavage in H20(a)
(reaction (13)). The enhanced catalytic activity of Cu( 110) compared to Cu( 111) has been attributed (ref. 88) to a lower barrier for 0-H bond cleavage on the Cu( 110) surface. Reaction (14) may not necessarily be an elementary step, but a consequence of the process (ref. 88):
WGS REACTION COORDINATE --->
Fig. 3.16 Potential energy diagram of the water-gas shift reaction on Cu(l10) (from ref. 88).
109
(3.18) (3.19) (3.20) (3.21) net OH(a) + O(a) + H(a)
(3.14)
Reaction (20) provides an easy pathway for converting OH(a) to O(a), which proceeds rapidly even at 290 K on Cu(l10) (refs. 92,93). Fig. 3.16 shows a potential energy diagram for the water-gas shift reaction on Cu(ll0). The diagram was constructed (ref. 88) using kinetic and thermochemical data for reactions (12) to (17). The activation energy for the rate determining step (H20(a)+OH(a)+H(a)) is =20 kcal/mol. A value of 27 kcal/mol has been estimated for this step on Cu(ll1) (ref. 87). This difference is probably due to the fact that the (1 10) plane is much more open, offering Cu surface atoms which are much more coordinatively unsaturated and hence more active for breaking 0 - H bonds.
3.3.2 SULFUR POISONING OF Cu(ll1) CATALYSTS The poisoning of Cu/ZnO catalysts by sulfur is one of the most serious problems in the water-gas shift process (refs. 86,94). Fig. 3.17 shows the effect of pre-adsorbed sulfur atoms upon the rate of the water-gas shift reaction over C u ( l l 1 ) (ref. 94). Sulfur addition causes a linear decrease in the rate of the reaction, with the rate going to zero at saturation sulfur coverage (0,=0.39). The decay in the rate with 0,can be easily understood in terms of a simple site-blocking model, where sulfur adatoms sterically prevent the dissociation of water (ref. 94). The rate decay of Fig. 3.17 is well fit by the expression: (1-2.6 0&. According to a statistical analysis (ref. 94), this 0,
8s 0.0
0.3
0.2
0.1
0.0
0.4
0.1
0.3
0.2
.
CU(II0 TOR“
,0111(
0.4
-
*a0
an m i i co
10 TOR8
3
.
7
2s
W
TORR
n2
ni.coa o
.
C W I I I )
.
.,2<
“20
co COI
0
0
‘.
N
1 0 -
v
0
L
0.0
3
-
W
9
v
2
-
I
0.1
0.2
S/CU
AES
0.3 RATIO
0.4
0.5
0.0
0.1
0.2
S/CU
AES
0.3
0.4
0.5
RATIO
Fig. 3.17 Rate of water-gas shift reaction over Cu(l11) as a function of sulfur coverage (from ref. 94).
110
linear function indicates that each sulfur atom blocks about 2.6 Cu atoms, and that the ensemble required for H 2 0 dissociation is rather small (1 or 2 Cu atoms). This is not unexpected, given the small size of the H 2 0 molecule and its dissociation products (OH(a)+H(a)).
3.3.3 CESIUM PROMOTION OF C u ( l l 0 )and C u ( l l 1 ) CATALYSTS Cesium has been shown to promote the water-gas shift reaction over CuEnO catalysts (refs. 9596). Kinetic data for the reaction on Cs promoted Cu(ll0) surfaces (ref. 88) are displayed in Fig. 3.18. The optimum Cs coverage to promote the reaction is 0~,=0.25ML. At this coverage, the reaction rate is five times faster than the clean surface rate at any reaction temperature (ref. 88). The coverage dependence and optimal coverage for cesium promotion shown in Fig. 3.18 for Cu(l10) are similar to those observed over Cu( 111) (ref. 97). An apparent activation energy of 11 kcal/mol was determined for the water-gas shift reaction on an optimally-promoted Cs/Cu(l 10) surface (0,,=0.27) (ref. 88). This activation energy is very close to the value of 10 kcal/mol found for Hz + COz H20 + C O CS/CU(llO) Pco = 26 torr
PH~O = 10 torr
CslCu AES Ratio
Cesium Coverage ( @ c J Fig. 3.18 The water-gas shift reaction rate as function of Cs coverage ( 0 ~ on ~ )Cs/Cu(llO) at 493, 523 and 573 K. P(H20)=10 torr and P(C0)=26 tom (from ref. 88).
the clean surface. A tendency toward a slightly higher apparent activation energy was also observed in experiments with Cu( 11l), where the optimally-promoted Cs covered surface had an activation energy of 20 kcal/rnol versus 17 kcal/mol for the clean Cu(l11) surface (ref. 97). A study of the pressure of the dependence water-gas shift reaction over optimally-promoted Cs/Cu(l lo), shows that on this surface the reaction orders in P(C0) and P(H20) are very different from those on clean Cu( 110) (ref. 88). This fact rules out the possibility that the same
111
elementary step is rate determining with Cs present or absent. In order to understand the role of Cs promoters in the water-gas shift reaction, detailed studies were carried out investigating the effects of Cs upon the surface chemistry of HzO, CO and C 0 2 on Cu( 110) (refs. 93,98,99). Post-reaction surface analysis of the Cs-promoted Cu(l10) catalysts show that the dominant form of Cs is a surface cesium-carbonate complex (CsCO3) (ref. 100). This same species can be produced by dosing C02 to Cs/Cu(l 10) under ultra-high vacuum (UHV) conditions (ref. 98). It decomposes in UHV at 450-600 K (for OcS<0.25) via: CsC03(a)+CsO(a)+C02(g) (ref. 98). The equilibrium: C02(g)+CsO(a)
+ CsC03(a)
(3.22)
occurs rapidly under water-gas shift reaction conditions (refs. 99,100). The CsO(a) species is readily reducible with CO gas via: CO(g)+CsO(a)+C02(g)+Cs(a) (refs. 88,100). Thus, a mechanism involving the species CsO(a) (produced by the reaction of H 2 0 and Cs (refs. 93,99,100) rather than simply O(a) may be operative on the Cs promoted catalysts. The species CsO(a) should be thought as a cesium-stabilized oxygen adatom, with a dominant bond to Cu and some degree of bonding to a neighboring Cs adatom (refs. 93,99).
3.4 METHANOL SYNTHESIS ON SINGLE CRYSTAL SURFACES Methanol is synthesized from carbon monoxide and hydrogen by the reaction:
C0+2 H2 + CH3OH. The selective synthesis of methanol is a process of major industrial importance because of the use of methanol as a chemical intermediate, its potential use as a starting material for fuel production, and many other applications (refs. 10,101,102). The most commonly employed industrial catalysts are based on Cu supported on ZnO, or a mixture of ZnO and other oxides such as Cr203 (refs. 10,101,102). It has been demonstrated that under certain conditions Pd is also an active methanol synthesis catalyst, and that the nature of the support has great influence
on both the rate and selectivity (ref. 103). Here we review studies that deal with methanol synthesis on Pd(1 lo), Cu(l1 l), ZnO,/Cu(ll 1) and Cu/znO(OOOl) single crystals.
3.4.1
KINETICS ON P d ( l l 0 )
Fig. 3.19a illustrates the effect of temperature on the rate of methanol production over a Pd(l10) catalyst. The rate shows approximately Arrhenius behavior with an apparent activation energy of 18.4k1.9 kcal/mol and a pre-exponential factor of 8,104 s-l over the temperature range of 493 to 553 K (ref. 104). The activation energy is in good agreement with values reported for Pd on “noninterating” supports such as Si02 and basic supports such as ZnO and La203 (refs. 104,108,109). In addition, the pre-exponential factor determined for Pd(ll0) agrees well with reported values for Pd/Si02 (refs. 104,109). The activation energy for Pd(l10) is, however, significantly different from values obtained on acidic AI203 and Ti02 supports (refs. 104,108). Figure 19b shows the variation of the reaction rate with pressure at 553 K. Methanol production was roughly first order (1.2M.2) in total pressure. The specific rates seen on Pd(1 lo),
112
Fig. 3.19 (a) Turnover frequency (CH30H molecules producedPd surface atom-s) versus inverse temperature at a total pressure of 244 kPa and a 3.9:l H2/CO mixture (from ref. 104). (b) Turnover frequency plotted as a function of total pressure at 553 K, for a 3.9:l H-JCO mixture (from ref. 104).
extrapolated to pressures typical of most methanol synthesis works, are in good agreement with rates observed for Pd dispersed on SiOz (refs. 104,108,109). Thus, Pd metal is an active methanol synthesis catalyst, and no specific support interaction is required. Highly interacting supports exhibit rates and/or selectivities substantially different from those of Pd(ll0) (ref. 104). The rates on Pd(ll0) are higher than those seen over Pd supported on zeolite and acidic supports (e.g., PdNaY and A1203) (refs. 104,108,110), but much lower than those observed on highly active La203-supported Pd (refs. 104,108,109). After reaction on Pd(1 lo), submonolayer quantities of a carbonaceous residue (generally between 0.05 to 0.25 carbon monolayers) were detected by Auger electron spectroscopy on the surface of the catalyst (ref. 104). Significant adsorption of hydrogen into the bulk, determined by post-reaction temperature programmed desorption, was not observed (ref. 104).
3.4.2 STUDIES ON C u ( l l l ) ,ZnOx/Cu(lll)and CulZnO(OOO1) SURFACES The synthesis of methanol from CO and H2 was attempted on Cu( 11 1) and over model catalysts consisting of well-defined Cu overlayers on ZnO(0001) (ref. 105). For temperatures between 500 and 600 K, H2+CO total pressures up to 1500 torr, and reaction times of 40 minutes, no production of methanol was observed. The same result was obtained when ZnO, films on Cu( 111) were used as catalysts (ref. 105). On the basis of these data, an upper limit of <2. CH3OH molecules/site-s was obtained for the catalytic activity of Cu(l1 I), ZnOx/Cu(lll) and Cu/Zn0(0001) surfaces (ref. 105). This limit is consistent with the rates expected for high-surface area Cu/ZnO catalysts (extrapolate from a somewhat higher pressure regime) (ref. 105)
113
3.5 CONCLUSIONS The use of a high pressure reactor combined with an ultra- high vacuum surface analytical chamber to study catalytic reactions can provide detailed information about the relation between surface structure and catalytic activity. Using these techniques, the concepts of structure sensitivity and structure insensitivity can be investigated. For CO methanation (a structure insensitive reaction), excellent agreement is obtained between studies on single crystal surfaces and studies on high surface area supported catalysts, demonstrating the relevance of kinetics measured on well-ordered single crystal surfaces for modeling the behavior of practical catalysts. Model studies on single crystal surfaces are also helpful in developing a better understanding of the effects of surface additives (poisons and promoters) on catalyst performance. The influence of additives on the surface chemistry of adsorbed reactants, products and intermediates can be studied using UHV techniques, and this information can be related to the effects of the additives upon the activity and selectivity of a catalyst. Of particular interest is the possibility that these types of studies will help to clarify the relative importance of electronic and geometric effects in determining additive effects. The types of studies reviewed here, in conjunction with studies on high-surface are supported catalysts, hold great promise for contributing to an overall understanding of CO activation.
3.6 ACKNOWLEDGEMENTS We acknowledge with pleasure the partial support of this work by the Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences. We also would like to thank Prof. Charles T. Campbell for communication of results prior to publication.
3.7 REFERENCES 1
G.A. Somorjai, Chemistry in Two Dimensions: Surfaces, (Cornell University Press, Ithaca, 1981).
2
G. Ertl and J. Kippers, Low Energy Electrons and Surface Chemistry, (VCH Verlagsgesellschaft, Weinheim, 1985). M.R. Albert and J.T. Yates, The Surface Scientist's Guide to Organometallic Chemistry, (American Chemical Society, Washington, 1987). D.R. Kahn, E.E. Petersen and G.A. Somorjai, J. Catal., 34 (1974) 294. B.A. Sexton and G.A. Somorjai, J. Catal. 46 (1977) 167. D.W. Goodman, R.D. Kelley, T.E. Madey and J.T. Yates, J. Catal., 63 (1980) 226. H.P. Bonze1 and H.J. Krebs, Surf. Sci., 91 (1980) 499. C.T. Campbell and M.T. Paffett, Surf. Sci., 139 (1984) 396. C.T. Campbell, Adv. in Catal., 36 (1989) 1. C.N. Satterfield, Heterogeneous Catalysis in Practice, (McGraw-Hill, New York, 1980). M. Vannice, Catal. Rev., 14 (1976) 153.
3 4 5
6 7 8 9 10 11
114
12
R.D. Kelley and D.W. Goodman, in: The Chemical Physics of Solid Surfaces and
Heterogeneous Catalysis, Vol. 4, Eds. D.A. King and D.P. Woodruff, (Elsevier, Amsterdam, 1982). 13 R.D. Kelley and D.W. Goodman, Surf. Sci. 123 (1982) L743. 14 M.A. Vannice, J. Catal., 44 (1976) 152. 15 D.W. Goodman, R.D. Kelley, T.E. Madey and J.M. White, J. Catal., 64 (1980) 479. 16 P.R. Wentrcek, B.J. Wood and H. Wise, J. Catal., 43 (1976) 363. 17 J.A. Rabo, A.P. Risch and M.L. Poutsma, J. Catal., 53 (1978) 295. 18 M. Araki and V. Ponec, J. Catal., 44 (1979) 439. 19 D.W. Goodman and J.T. Yates, J. Catal., 82 (1983)255. 20 D.W. Goodman and J.M. White, Surf. Sci., 90 (1979) 201.
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
D.W. Goodman, T.E. Madey, M. Ono and J.T. Yates, J. Catal., 50 (1977) 279. T.J. Udovic, R.D. Kelley and T.E. Madey, Surf. Sci., 150 (1985) L71. A. Cichowlas, E.P. Yesodharam and A. Brenner, Appl. Catal., 11 (1984) 353. D.G. Castner, R.L. Blackadar and G.A. Somorjai, J. Catal., 66 (1980) 257. D.J. Dwyer, K. Yoshida and G.A. Somorjai, Adv. in Chemistry Series, American Chem. Soc.,178 (1979) 65. M. Logan, A. Gellman and G.A. Somorjai, J. Catal., 94 (1985) 60. M.A. Vannice, J. Catal., 37 (1975)449. J.H. Sinfelt, Bimetallic Catalysts, (Wiley, New York, 1983). W.M.H. Sachtler, Faraday Discuss. Chem. SOC.,72 (1981) 7. J.E. Houston, C.H.F. Peden, D.S. Blair and D.W. Goodman, Surf. Sci., 167 (1986) 427. J.E. Houston, C.H.F. Penden, P.J. Feibelman and D.R. Hamann, Phys. Rev. Lett., 56
(1986) 375. C.H.F. Peden andD.W. Goodman, J. Catal., 104 (1987)347. C.H.F. Peden and D.W. Goodman, Ind. Eng. Chem. Fundam., 25 (1986) 58. R.W. Judd, M.A. Reichelt and R.M. Lambert, Surf. Sci., 198 (1988) 26. I. Hamedeh and R. Comer, Surf. Sci., 154 (1985) 168. P.J. Berlowitz and D.W. Goodman, Surf. Sci., 187 (1987) 463. P.J. Berlowitz and D.W. Goodman, Langmuir, 4 (1988) 1901. C.M. Greenlief, P.J. Berlowitz, D.W. Goodman and J.M. White, J. Phys. Chem., 91 (1987) 6669. J.W. He and D.W. Goodman, submitted for publication. J.W. He, W.L. Shea, X. Jiang and D.W. Goodman, 36th National Symposium of the Amer. Vac. Soc.,in press. J.W. He and D.W. Goodman, J. Phys. Chem., 94 (1990) 1496. K. Christmann, 0. Schober, G. Ertl and M. Neumann, J. Chem. Phys., 60 (1974) 4528. K. Christmann, G. E d and H. Shimizu, J. Catal., 61 (1980) 397. C. Park, E. Bauer and H. Poppa, Surf. Sci., 187 (1987) 86.
115
45 46 47
D.W. Goodman, in: Heterogeneous Catalysis (Proceedings of IUCCP Conference), Texas A&M University (1984). B.G. Johnson, C.H. Bartholomew and D.W. Goodman, J. Catal., submitted for publication. B. Imelik, C. Naccache, G. Couduier, H. Praliud, P. Meriaudeau, P. Gallezot, G.A. Martin
48 49
and J.C. Vedrine, Eds., Metal-Support and Metal-Additive Effects in Catalysts, (Elsevier, Amsterdam, 1982). D.W. Goodman and M. Kiskinova, Surf. Sci., 105 (1981) L265. M. Kiskinova and D.W. Goodman, Surf. Sci., 108 (1981) 64.
50 51 52 53 54 55
56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78
D.W. Goodman, Appl. Surf. Sci., 19 (1984) 1. S. Johnson and R.J. Madix, Surf. Sci., 108 (1981) 77. E.L. Hardegree, P. Ho and J.M. White, Surf. Sci., 165 (1988) 488. C. Kittel, Introduction to Solid State Physics, 6th ed., (Wiley, New York, 1986) p. 76. J.R. Rostrup-Nielsen and K. Pedersen, J. Catal., 59 (1979) 395. C.H.F. Peden and D.W. Goodman, ACS Symposium Series, Proceedings of Sym. on the Surface Science of Catalysis, M.L. Devinny and J.L. Gland, Eds.; Philadelphia 1984. P.D. Szuromi, R.D. Kelley and T.E. Madey, 1. Phys. Chem., 90 (1986) 6499. C.T. Campbell and D.W. Goodman, Surf. Sci., 123 (1982) 413. G.A. Mills and F.W. Steffgen, Catal. Rev., 8 (1973) 1.59. P. Schoubye, J. Catal., 14 (1969) 238. G. Blyholder, J. Phys. Chem., 68 (1964) 2772. G. Blyholder, J. Phys. Chem., 79 (1975) 756. F.A. Cotton and G. Wilkinson, Advanced Inorganic Chemistry, 4th ed. (Wiley, New York,
1980) p. 82-86. P.S. Bagus, K. Hermann and C.W. Bauschlicher, J. Chem. Phys., 81 (1984) 1966. P.S. Bagus and K. Hermann, Phys. Rev. B, 33 (1986) 2987. K. Hermann, P.S. Bagus and C.J. Nelin, Phys. Rev. B, 35 (1987) 9467. J. Rogozik and V. Dose, Surf. Sci., 176 (1986) L847. K.L. D'Amico, F.R. McFeely and E.I. Solomon, J. Amer. Chem. Soc.,105 (1983) 6380. H. Ibach and D.L. Mills, Electron Energy Loss Spectroscopy and Surface Vibrations, (Academic Press, New York, 1982). J.A. Rodriguez and C.T. Campbell, J. Phys. Chem., 91 (1987) 2161. J. Benziger and R.J. Madix, Surf. Sci., 94 (1980) 119. N.K. Ray and A.B. Anderson, Surf. Sci., 125 (1983) 803. R.C. Baetzold, Phys. Rev. B, 30 (1984) 6870. M.C. Zonnevylle and R. Hoffman, Langmuir, 3 (1987) 452. A.J. Freeman, C. Fu and E. Wimmer, J. Vac. Sci. Technol. A, 4 (1986) 1265. J.M. MacLaren, J.B. Pendry, R.W. Joyner, and P. Meehan, Surf. Sci., 175 (1986) 263. T.B. Grimley, Proc.Phys. Soc.,90 (1967) 751. T.B. Grimley and M. tomni, J. Phys. Chem., 87 (1973) 4378. T.E. Einstein and J.R. Schrieffer, Phys. Rev. B, 7 (1973) 3629.
116
79 80 81 82 83 84 85
P. Feibelman and D. Hamman, Phys. Rev. Lett., 52 (1984) 61. P. Feibelman and D. Hamman, Surf. Sci., 149 (1985) 48. J.K. Nrskov, S. Holloway and N.D. Lang, Surf. Sci., 137 (1984) 65. N.D. Lang, S. Holloway and J.K. Norskov, Surf. Sci., 150 (1985) 24. S.J. Tauster, Accounts of Chemical Research, 20 (1987) 389. C.C. Kao, S.C. Tsai andY.W. Chung, J. Catal., 73 (1982) 136. Y.W. Chung, G. Xiong and C.C. Kao, J. Catal., 85 (1984) 237.
86 87
D.S. Newsome, Catal. Rev. Sci. Eng., 21 (1980) 275. C.T. Campbell and K. Daube, J. Catal., 104 (1987) 109. J. Nakamura, J.M. Campbell and C.T. Campbell, submitted for publication. G.C. Chinchen, M.S. Spencer, K.C. Waugh and D.A. Whan, J. Chem. SOC.,Faraday Trans. 1,83 (1987) 2193. S. Kinnaird, G. Webb and G.C. Chinchen, J. Chem. SOC., Faraday Trans. 1, 83 (1987) 3399. E.G.M. Kuijpers, R.B. Tjepkema, W. Van der Wal, C.M. Mesters, S.F. Spronck and J.W. Geus, App. Catal., 25 (1986) 139. K. Bange, D.E. Grider, T.E. Madey and J.K. Sass, Surf. Sci., 137 (1984) 38. W.D. Clendening, J.A. Rodriguez, J.M. Campbell and C.T. Campbell, Surf. Sci., 216 (1989) 429. C.T. Campbell and B.E. Koel, Surf. Sci., 183 (1987) 100. D.C. Bybell, P.P. Deutsch, R.G. Herman, P.B. Himelfarb, J.G. Nunan, C.W. Young, C.E.
88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
Bogdan, G.W. Simmons and K. Klier, Prep. Petrol. Div., Am. Chem. SOC.,31 (1986) 282. K. Klier, C.W. Young and J.G. Nunan, Ind. Eng. Chem. Fundam., 25 (1986) 36. C.T. Campbell and B.E. Koel, Surf. Sci., 186 (1987) 393. J.A. Rodriguez, W.D. Clendening and C.T. Campbell, J. Phys. Chem., 93 (1989) 5238. J.A. Rodriguez, W.D. Clendening, J.M. Campbell, W. Min and C.T. Campbell, J. Vac. Sci. Technol. A, 7 (1989) 2118. J. Nakamura, J.M. Campbell and C.T. Campbell, in preparation. H.H. Kung, Catal. Rev. Sci. Eng., 22 (1980) 235. K. Klier, Adv. in Catal., 31 (1982) 243. J. H. Lunsford, Chem. Ind., 22 (1985) 95. P.J. Berlowitz and D.W. Goodman, J. Catal., 108 (1987) 364. C.T. Campbell, K.A. Daube and J.M. White, Surf. Sci., 182 (1987) 458. G.L. Haller and D.E. Resasco, Adv. in Catal., 36 (1989) 173. J.M. MacLaren, D.D. Vvedensky, J.B. Pendry and R.W. Joyner, J. Chem. Soc., Faraday Trans. 1,83 (1987) 1945. Y.A. Ryndin, R.F. Hicks, A.T. Bell and Y.I. Yermakov, J. Catal., 70 (1981) 287. R.F. Hicks and A.T. Bell, J. Catal., 91 (1985) 104. F. Fajula, R.G. Anthony and J.H. Lunsford, J. Catal., 73 (1982) 237.
117
CHAPTER 4
SELECTIVITY IN THE SYNGAS REACTIONS: THE ROLE OF SUPPORTS AND PROMOTERS IN THE ACTIVATION OF CO AND IN THE STABILIZATION OF INTERMEDIATES
V. Ponec Gorlaeus Laboratonurn, Leiden University, P. 0. Box 9502, 2300 RA Leiden, (The Netherlands)
118
4.1 SELECTIVITY PATTERN IN SYNGAS REACTIONS SOME FACTS ON METALS
4.1 .I
; Metals used first in catalytic hydrogenation of CO into hydrocarbons were Ni (refs. 1,2) and Co (ref. 3). Hydrogenation to methane was observed first and later also the formation of higher hydrocarbons has been described (refs. 2,3). After a search for a cheaper catalyst than cobalt, iron has been added to the family of the Fischer-Tropsch Synthesis (FTS) catalysts (ref. 4).A systematic study performed by Vannice (ref. 5 ) helped to bring most of the other VIII group metals in one picture of activities as can be seen in Fig. 4.1. Starting on the left in the group VIIIa (8th group, new count) the activity drops when going to the right in the tabel. However, according to the information available from other papers it drops also when going to the left (not shown). The reason is - most likely - the behavior of a metal in the CO adsorption or dissociation (refs. 6,7). All metals can be divided according to their activities in synthesis of hydrocarbons into the 5 groups, as shown below. 4.1.1.2
Activim mttern
1) Metals of the groups 3-7 of the periodic table are active in CO adsorption and adsorb co dissociatively However, the surface oxides of these metals are of low activity and they are irreducible under the temperatures of the FTS (i. e. under the
-1
conditions that chain growth
- 10000
thermodynamically
0,
ul
the is
n
100-
sufficiently favoured). Some of the surface
=
10-
carbides are probably
1-
inactive, some are active (Mo, W) (ref. 6). 2) Metals: Fe,
z
0 l-
L
Q,
>
=0
0.1 -
V
0 0
00.1I
I
1
8
9
10
I
11
I
12 group number
Fig. 4.1 Activity pattern in FTS metal/Si02 catalysts (ref. 5). T. 0. N. of the CO molecules converted per site and second. Number of sites is determined by CO adsorption, as a function of the position in the periodic system.
Co, Ni, Ru, 0 s (i. e. metals above the diagonal through the table of the VIII group metals). Surface carbon is with this group of metals active and adsorbed oxygen is reducible under FTS
119
conditions (ref. 8). These metals adsorb CO dissociatively under mild conditions (from the ambient temperature up to that of the FTS) and for these metals sufficient evidence is available that surface carbon is converted into building units of the hydrocarbon chain growth process (see for the evidence refs. (refs.6,7) and other chapters of this book)
3) Metals of a low-activity Pd, Ir, Pt (under the diagonal of the Table of VIII group metals). Their activity can be enhanced considerably by promotion. It is possible but not yet proven that on these metals CO is first hydrogenated and only than the C - 0 bond is broken (refs. 5-9). A CO dissociation step is not excluded either but molecular adsorption of CO is in any case rather strong. 4) Rh, which is in all respects an intermediate case between 3 ) and 2). 5 ) Metals of 1lth and 12th group are of a very low or zero activity. They adsorb CO weakly
or they do not adsorb it at all. The most important factor underlaying the division of metals in the five groups above is undoubtedly the thermodynamics of CO adsorption and that of surface carbide and oxide formation. If the adsorption of CO is strongly exothermic and the formation of surface carbide and oxide together only weakly exothermic or endothermic, the situation is set for a kinetically unfavoured CO dissociation (high activation energy, low preexponential factor). This could be the case of metals Pd, Ir and Pt. However, the data are not yet available to make definitive conclusions possible. 4.1.1.3
Selectivitv pattern
Metals differ not only by their overall activities (Fig. 4.1) but also by their selectivities. It has been shown that the main hydrocarbon chain growth mechanism is a coupling of CH, units with C,H, hydrocarbon fragments (y=l to n) (refs. 10-12) and therefore it can be reasonably expected that the selectivity to the formation of higher hydrocarbons depends on the surface concentration of CHx and CYH, units and their reaction with adsorbed hydrogen. Reaction with adsorbed hydrogen leads finally to the chain growth termination and desorption. The steady state concentration of CH, is in its turn dependent on the rate of CO dissociation and rates of hydrogenation of C and CH, species. The selectivity to the formation of higher hydrocarbons is usually characterized by the parameter, a, of the Flory-Schulz-Anderson product distribution function:
where Fi is the molar fraction of a hydrocarbon with i C atoms. Parameter Q is the probability Pp of the growth divided by the sum of two probabilities: that of growth (propagation) P p and that of termination P,. It is impossible at the present state of knowledge to explain quantitatively the differences in a (i.e. i n selectivities) amongst the metals. However, qualitatively the selectivity pattern can be well rationalized. The metals Fe, Co and Ru can produce a high surface concentration of the buildings blocks (C, CH,) (ref. 13) so that the ratio PAPt is favorable for the high hydrocarbon formation. The most favorable conditions are in further details determined by the
120
Table 4.1 111
0
Hydrocarbon synthesis, activity pattern
IV
V
VI
IIb
VIII
VIII
H
Ni
n
Rh
L
-0
M
M
M
8
9
10
-0 11
I T 2
11
I
VIII
VII
IB
H
l o ~
0 O
0
0
0 3
0 4
I ~
O
C
0
C
0 5
C 6
H
~
7
_
_
J
0
0
_
0
Note:
H, M, L C
0
0
Rh Ni
111
I
- high, medium, low activity - active, when partially carbided - zero activity, deactivation by CO dissociation products - zero activity, because of lack of CO adsorption - a medium position in all respects - a high methanation activity
IV
V
+ + 3
4
+
+
+
*
+ *
+ 5
6
7
1
s
+ lr
Pt
9
10
I
12
f cdfH2 ratio which dictates the surface coverage by hydrogen (gH) and influences by that (mainly) the termination rate. Nickel is a good hydrogenation catalyst and the steady state surface concentration of buildings blocks on its surface is under most conditions rather low (ref. 14) and that of hydrogen high, with as a consenquence that a of Ni is low and Ni forms mainly methane. Metals which dissociate CO reluctantly or (which possibly, even form CH, units by an reversed sequence of steps: CO hydrogenation
+ C-0
bond dissociation) and are good hydrogenation
catalysts, usually show low CI values as well (ref. 5). From the active VIII group metals (point 2, above), only Fe seems to produce at high pressures or after modification (refs. 3,15) some oxygen containing molecules (“oxygenates”). It is probably relevant to mention in this relation that under a steady state of a commercial process, iron is to a high extent converted into oxides and carbides (ref. 17). Also Rh when being in contact with Rh-oxide(s) can produce some oxygenates (mainly CH30H) (ref. 18). On the other hand at low
121
pressures and in an unpromoted and not pre-oxidized state, various metals (Fe, Co, Ni, Ru, Rh) produce only hydrocarbons. Some oxides of them can produce CH30H apparently also without any metal (ref. 19). Metals Pd, Ir, Pt and Cu produce CH30H when they are combined with a suitable carrier (refs. 20,21) and there are again indications available that the role of the carrier is to stabilize positively charged forms of these elements (see below). However, some authors (ref. 22) believe that the metallic (zero valency) surface of the transition metal is also able to catalyse CO hydrogenation into CH30H (or other oxygenates). It can indeed, be so since Pd, Ir, Pt (and also Cu)
are known as hydrogenation catalysts, but it is not easy to bring the definitive evidence for the above mentioned statement. When working with single crystals, the sample holders usually contain enough of various oxides which could stabilize some active transition metal ions to make CH3OH synthesis observable by them. With Cu, there are several authors (for review see ref. 21) who derive from the linearity of the relationship; activity vs. Cu metal surface that Cuo is the only site of catalytic activity in the CH30H formation. However, this cannot still be accepted as the final evidence either, since the measured dispersion of Cu (that is the Coo surface area) could be correlated to the (not determined) concentration of Gun+ in the system. For example in such a way that a constant fraction of the Cuo surface area is under reaction conditions constantly covered by
Cun+O, species, the potential centres of the CH30H synthesis, etc. The question has to be thus considered at the moment as being still open. The author's experience with Pd is that Pd on a very clean SiO2 (in the quality needed for fabrication of fibres, samples of which were kindly supplied by Philips/Netherlands) does not produce any CH30H (ref. 23). We shall turn to this point in the next section. TFor the benefit of the reader, the conclusion just fomiulated are graphically illustrated by the Tables 4.1 and 4.2. Alloys (bimetallics) will be discussed below.
4.1.2
SUPPORTS AND PROMOTERS
4.1.2.1
Supuort functions as a promoter
The problems concerning the role of supports and promoters can be best illustrated in Figs. 2 and 3, which at the same time represent the pioneering research performed in this field by Ichikawa and his coworkers (ref. 42). Platinum catalysts of the CH30H synthesis shown in Fig. 4.2 have been prepared with the indiacted oxides as supports and essentially the same results have been obtained with the Pd catalysts. Rhodium catalysts shown in Fig. 4.2 have been prepared from a Rh/SiO* catalyst promoted by a compound yielding the indicated oxide. Later investigations by Van der Lee, Bastein, LUOet al. (refs. 24-29) revealed, that essentially the same result can be obtained when one uses the oxide in question (in their research V203 was used) either as a support or as a promoter added to Rh/Si02
122
For further analysis two points from the above mentioned are particularly relevant: 1) Different oxides are the optimal support (or promoter) for the CH30H on one side and C2H50H synthesis on the other side (see e. g. the position of MgO or La203 and TiO,). 2) Obviously, it does not matter too much if a certain oxide is used as a promoter or as a support. The two points just mentioned have been analyzed in more details by the papers published from Leiden (refs. 24-29). First, it has been shown that one and the same metal, Rh, can be made a catalyst with a rather high selectivity to either methanol or to ethanol. In the first case one has to take preferably a basic oxide (MgO, La203) as a support and to reduce the catalyst at low temperatures (under 500 K). In the catalysts prepared in this way Rhn+ can be detected (refs. 24,26).
EtOH mmol/h CO,ds, site mmol
i
La203
1
0
Fig. 4.2 Specific activities in methanol (0) and methane (A) formation, with Pt catalysts prepared from Pt carbonyls and oxides, as indicated. The same for the catalysts prepeared from H2PtC16, ( 0 ) CH30H, (A) cH4.
0
0 0
0
7 0.5 0
c
c 3
,E 0.L 0 LL
&
-
0.3
c
W
x
-
0.2
> .c _
2
0.1 0
Fig. 4.3 Activities and selectivities in ethanol synthesis obtained with promoted Rh/Si02 catalysts (RhCldSi02 with an oxide precursor calcined at 770 K and reduced at 670 K by H2 (ref. 42)).
A harder reduction leads to the decrease of Rhn+ concentration and CH30H synthesis activity. On the other hand, in the catalysts most active in the ethanol synthesis Rh/V203, virtually zero concentration of Rh"+ was found (refs. 24,26). Rhodium (I, 11, 111) ions were detected not only by chemical extraction (refs. 24,26) but Rh2+ also by ESR (ref. 25). The suggestion made by the second point above has been proven to be true by the following experiments (ref. 26). Rhodium has been brought as chloride onto a V 2 0 3 support and reduced at either 473 K or 673 K. Catalyst
123
prepared at the higher temperature was less active but
more
selective
to
the
formation
of
0 CZf 0 CrOXY
C2-oxygenates (ethanal and ethanol) upon 1 atm. synthesis (if not indicated otherwise all synthesis experiments have been made at 1 atm and at 473 K). It is known (ref. 30) that V203 is one of the oxides which most easily create a state of “SMSI” (strong metal support interaction, manifesting itself by a
1
10
e *OI I
I
I
200
300
LOO
Tred I O C 1
strong suppression of adsorption of gases as H2 or CO) and it is also known that the main reason of the
“SMSI” state is the coverage of a metal by species created by the high temperature reduction from the oxidic support and brought by migration on the metal (refs. 31-35). Similar observation - that a creation of an SMSI state improves the selectivity or the activity in the syngas reactions (refs. 36-37) - has been made already earlier by other authors. However, as has been correctly pointed out by Burch (ref. 37) and later quantitatively confirmed by others (refs. 25,28) under a running syngas reaction, the SMSI state (suppression of adsorption capacity) is almost completely abolished but small amounts of the oxide stay on the metal surface and can influence
200
300
LOO Tred I ’ C I
Fig. 4.4 Rate of the formation (470 K, 1 atm.) of various products as a function of the reduction temperature. Rh/SiO, catalysts promoted by a VOC12 added. All samples with the same V/Rh atomic ratio (above). Selectivities as a function of temperature, for the same catalysts (below) (ref. 28).
the activity and selectivity. The mechanism of this effect will be discussed in section 4.3. Important point to remember is that migration can bring the support material on the metal to act as a promoter there. It has been observed that the solution with the metal precursor - for example a Rh saltdissolves a support like V203 when pH is sufficiently low. So is in the presence of HCI and air, V(rv)OC12very easily formed. Upon drying and reduction above 600K, VOc1, is converted into a stable V(lll)OC1and a volatile hichloride. In the presence of HN03, V(1V)O(N03)2can be formed etc. Formation of V(Iv) ions can be followed by ESR or UV/vIS spectra (refs. 24,27). After drying and reduction, V0,-species appear on the metal and act as a promoter or as precursor of the promoter (refs. 24-29). It should be stressed already on this place that the just mentioned mechanism of converting support into species modifying the metal support is quite general with many systems. A direct evidence of dissolution of the support by metal precursor is available for systems like Pt/A1203 (refs. 38,39) Cu/A1203 (ref. 40) and an indiiect evidence exists for many others, including Pt/ex ammonia complexes/Si02 (ref. 41). Bastein and Luo (ref. 28) have confirmed that both mechanisms: the high temperature migration or “volatalization” of the support during the wet steps of catalyst preparation, lead to very similar
124
a
results as can be seen in the Figs 4 and 5. The authors (ref. 28)
Tred=200O C
prepared
/
\
b.
Rh/Aerosil-SiO,
with
VOCl2 as a promoter and with one given promoter concentration, they of varied the temperature reduction. In other experiments
Tred = LOG “C
they reduced always at the same temperature but varied the V/Rh V/Rh
2
1
6
8V/Rh
Fig. 4.5 (a) Rate of formation (470 K, 1 atm.) of various products as a function of the V/Rh atomic ratio. Rh/SiO, catalysts promoted by VOCI, added to the Rh/SiO, catalyst (post-impregnation). (b) Selectivities of the same catalysts (ref. 28).
ratio. In both cases the curves activity/g vs. temperature or V/Rh showed a maximum as one would expect if the metal surface were increasingly covered by a promoter which itself is inactive for the given reaction.
Obviously, it is not essential in which way the promoting species come on the metal surface, behavior of the catalysts is very similar. Further, it is not essential whether a certain oxide is introduced as a support or as a promoter, by a “suitable” procedure the support material can always be converted into a promoter. However, the condition of functioning as promoter is that the promoting compound is on the metal-gas interface and not only under the metal particle. This has been proven by the following experiment (refs. 24,29). Catalyst 1 has been prepared by contacting in air V,03 with HRhCIS.H,O, catalyst I1 by doing it with Rh(N03)3. In the first case VOCl, is formed in a high concentration, in the second case no vanadyl nitrate was actually seen but its formation is not excluded (low detection sensitivity). Upon drying chemical changes occur: with catalyst I VOC1, (and after reduction VOCI) is formed. As far as the catalyst I1 is concerned it is known that VO(N03), decomposes into V205 which does not adhere well to the metal surface and is a bad support/promoter for formation of oxygenates (refs. 2427). After a reduction at 740 K, catalyst I showed Soxygenof 4095, catalyst I1 did not produce oxygenates at all. However, catalyst 11 could be promoted to produce some oxygenates (+lo%)when treated by HC1 or when reduced at high temperature, when a conversion V,O, + VzO, takes place. In the latter case Soxyyen. was 27%. In both catalysts Rh was in close contact with V203 but only with catalyst I it had been garanteed that V promoter appeared on the metal surface. This observation that a promoter cannot function through the metal is important for the theory of the promoter action and we will turn to it once more in section 4.3. The mechanism by which the support material is brought on the surface of the metal component of the catalyst (high temperature reduction or dissolution of the support upon wet steps of preparation) operate obviously quite generally and they modify metals to become selective for oxygenate formation. One would even tend to conclude, on basis of the existing data, that with Co,
Ru and Rh, this is a condition sine qua non to make these metals active in the production of
125
oxygenates. However, there is a problem of silica. Silica is not a material which dissolves in acids and it also does not migrate easily and yet some silica supported Rh catalysts produce ethanol (refs. 42-44). It is not clear why in different experiments by the same authors in some cases Rh/Si02 is and in some others (compare refs. 42 and 43) it is not a support leading to production of catalysts for oxygenates. One can only speculate what would solve this problems and the possible ideas are as follows. i) Upon preparation metal (Rh) silicates are formed and act as promoters on the metal surface. ii) Si02 contains minute but important impurities which are leached out of Si02 during the catalyst preparation and appear finally as promoters on the metal surface. Bastein and Nonneman (refs. 25,45) observed that Rh on the ultraclean SiO, (Philips) is only a bad catalyst for ethanal while no (or almost none) ethanol is formed by this catalyst. The same holds for Rh/Si02 when a common Si02 carrier has been first well leached by strong acids (HCI) Leaching solutions containing alkali and iron ions can however promote the catalyst producing only ethanal to a catalyst with a selectivity for ethanol when the solution is added to it. This is in favor of the idea under ii), but does not exclude i) completely. Leaching of the support is the better the higher the concentration of the acidic metal-precursor and this can obscure the results of studies on particle size effects or on effects of various precursors on the properties of the catalysts. 4.1.2.2
A closer inspection of the uromoter function
4.1.2.2.1
Transition metals Pd, Pt, Ir and Co, Ru, Rh in CH3OH synthesis Isotopic labelling has shown that the original C - 0 (N. B. this can only be observed when the apparent contact time of the gas mixture is sufficiently low) is preserved upon formation of CH30H. Therefore, the second group of metals above - Co, Ru, Rh - can only have a high selectivity towards CH30H if the methanation and FTS reactions are strongly suppressed by the presence of a promoter. This can most easily happen by bringing the support material on the metal (see above) for blocking the sites for G O dissociation. This is a reaction which to occur requires a large ensemble of contiguous surface atoms (ref. 47). Indications that this blocking by a carrier occurs has been published (ref. 48). If we accept this idea we are still left with the problem, how does the active site on the title metals look out. The ideas suggested in the literature can be subdivided in several cathegories, ( i) - iii) below). i) It is the zero-valent metal (e.g. Pdo) which alone fomis the centers for hydrogenation of CO into CH3OH (ref. 204. However the properties of the metal are to be modified in one or another way: (a) the metal should be present in a certain particle size (ref. 49) and (b) there should be a certain metal-support interaction operating (most explicitely formulated in ref. 50) or (c) the factors (a) and (b) operate together. The particle size is assumed to be important as such or to act through variations in the morphology (ref. 51) (for example the population of (100) and [ l l l ) planes in the surface of the fcc metal particles should vary).
126
If the zero-valent metal atoms were indeed the active sites, the suggestions formulated under (a) (c) would not offer a complete explanation. Transition metal particles can be made active for CH30H synthesis and yet be of such a large size that effects under (a) - (c) could probably not play any significant role. Moreover, a promotion which does not change the metal particle size or form, can change the activity and selectivity, as can be derived from data in several papers (refs. 52,53). ii) Material of the support participates in the formation of the intermediates of the conversion CO -+ CH30H, or it activates CO for this pathway. The group in Lyon favors the idea, that with some carriers (Tho2, CaO, La203 and Cr203; but not MgO or SiO2) CO is being activated on the carrier and H2 (H2 -+ 2HdJ on the metal and the authors call this mechanism a bifunctional one (refs. 543.5). Other authors prefer a picture in which CO is bound to two centres: with it's "C"-end to a metal atom and with it's "0'-end to the oxide. This second oxide site is either at the perimeter of the metal particles embedded into the oxidic support or at the perimeter of the oxide patches on the metal surface particles (brought there during the catalyst preparation) (see e.g. refs. 37,558)). This picture better than the foregoing one, would explain why also MgO or alkali's can promote the metal to a high selectivity to methanol. Kakuzono et al (ref. 59) suggest that the intermediate - the formiate - is always bound to the support and the oxygen of CO can be exchanged, by a process of forming and decomposing of the formiate, for oxygen of the support. The latter is in contradiction with the results (ref. 46) but the idea of formiates being formed on the support is still an open possibility. iii) Support/promoter creates new centres from the transition metal element, for example Pdn+ (n being most likely 1) centres. Driessen et a1 (ref. 60) found that the activity of Pd/SiO2 catalysts, promoted by MgO/MgC12 was linearly proportional to the amount of extractable Pd. Extractable (by acetonyl-acetone) Pd is most likely an ionic species. Therefore, the authors suggested (see Fig. 4.6) that Pd*+ are the active sites of the reaction and are stabilized by the support/promoter. Unlike the correlation with the Pdn+ concentration, the activity as a function of promoter content is a function with a maximum. The assumption made in (ref. 60) has got two pieces of additional support. Anikin et a1 (ref. 61) and Pacchioni et al (ref. 62) showed that with Pd, the formation of Pd formyl complex (potential intermediate) is easier (lower activation energy) and formyl is more stable, when Pd is positively charged. Hindermann et a1 (ref. 63) applied the chemical trapping method and when using CH3I as a trapping agent, the authors could show that the activity in CH,OH synthesis is proportional to the concentration of the formyl groups formed on different Pd/MgO, MgC12/SiO, catalysts under standard conditions (such correlation was not found with Pd with respect to formates). An indirect support for the picture iii) is also given by the fact that the compounds active as promoter, slow down the reduction of the transition or noble metal atom (refs. 48,64-66).A qualitative evidence for the presence of unreduced transition metal (Pd+, Rh2+) has been obtained by ESR spectra (refs. 47,67). Also the UV/VIS spectra (ref. 67) indicate stabilization of Pd ions by the promoter.
127
With Ru (refs. 21,68) and Rh (refs. 26,69) the situation is very much similar. Again, oxides which can be suspected to stabilize Ru or Rh ions (MgO) increase the selectivity to methanol and oxidative treatments enhance the selectivity to methanol.
4.1.22.2
Copper catalysts
A catalyst, which is at the moment the most important commercial catalyst for syngas reactions is based on CuEnO. This couple is further combined with A1203 and sometimes, with additional oxides (Cr2O3) and the resulting catalyst is not only an active (this is less difficult to achieve than the other properties) but also a very stable catalyst. The stability is required also under potential variations in temperature and in the CO/CO, ratio. Catalysts reduced under too rigorous conditions or catalysts in which an incipient alloy formation ( C a n ) occured, loose their activity. Addition of alkali oxides or of oxides of other nansition metals,
Fig. 4.6 Activity (yield) in CH3OH formation at 488 K (defined as total conversion x selectivity x 10-2) as a function of the relative (to the total Pd content) content of Pd extractable by acetylacetone under standard conditions (most likely Pdn+); $) Mg promoted, chlorine containing Pd/Si02 catalysts; (m) Mg promoted, chlorine free ; (0)La promoted yield of all products (0)La promoted, methanol yield.
can lead to a simultaneous production of higher alcohols (refs. 19,21). Ideas about the site active in CH30H synthesis are very diversified. Both basic components (Cu and ZnO) adsorb H2 (refs.70,71) and CO, but none of them very well; Cu+ adsorbs CO stronger then Cuo (ref. 72), Hermann et a1 (ref. 73) suggested that Cu+ survives the reduction being embedded in ZnO, and it serves as a site for CO while H2 is adsorbed on ZnO in a heteropolar way. Later, it has been speculated that Cu+ might also be a center of H2 adsorption (ref. 74). In this picture Cuo serves as reservoir for the formation of additional (after oxidation) Cu+ ions; mainly formed later in the presence of C 0 2 or water. A good correlation is found between the activity and the concentration of the assumed active centers (ref. 19). The second idea, being the other extreme, does not ascribe any role to ZnO in the formation of the active surface complexes (intermediates) and assumes that Cuo surface is the site of the catalytic activity (an illustrative review is presented by the work of (ref. 75a); see also the refs. therein). ZnO is then just a suitable support stabilizing the active form (whatever it could be) of Cu0 metal. However, the same authors (ref. 75a) admit that oxygen makes Cuo more active (as it does with Pd too) (ref. 75b) and they show that a constant fraction f of the Cuo surface is in the presence of gas mixtures with C 0 2 always covered by oxidic (or perhaps also carbonate) species. The fractionfis a function of the gas composition. It is obvious that the oxidic species Cun+formed in this way can appear in the neighbourhood of Cuo particles or on these particles, where they can
128
be stabilized by ZnO, AI2O3 or Cr203 (ref. 7%). The latter oxides can easily appear on Cuo surface
in the process of catalyst preparation (see above). The importance of the oxidized surface leads to speculations that the potential centers for CH30H synthesis are, indeed, Cu+ ions which in reducing gas mixtures (H2/CO) only survive in ZnO or in another oxide present, while in gas mixtures with COz or H20, additional Cu+ centers appear on the Cuo surface. This picture would lead to the mentioned linear relation between the Cuo surface area and CH30H-activity. However, the authors (ref. 75a) reject this interpretation (see the discussion following the ref. 75a). Additional useful information on this matter can be find in (ref. 21) and the discussion cannot be considered at the moment as definitively closed. The mechanism of the formation of higher alcohols on Cu catalysts promoted with alkali is probably different from the mechanism operating on promoted transition metals. While with transition metals the CH, unit formation probably follows a CO dissociation with Cu - which does not dissociate CO easy - other mechanisms have been suggested (ref. 19) and the aldol-condensation mechanism seems to have the best support (see other parts of this book). Aldol condensation is catalysed by bases what would be the role for the alkali additives to the Cu catalyst. 4.1.2.2.3 Promoted transition metals as catalysts for C~+-oxygenates. We have already seen in 4.1.2.2.2 that Cu catalysts can be promoted to produce higher
oxygenates (C2+-oxygenates). Obviously, with Cu one has to promote the reactions of C-C bond formation while this step is easy with the group of transition metals: Fe, Co, Ni, Ru and Rh discussed here. By using a proper support/pronioter all those metals can be made producing some CZ+-oxygenates (refs. 18,26,29,42-44,7683). Iron can be also promoted by nitridification (ref. 15) and it has from the metals mentioned, the strongest tendency to form the highest molecular weight oxygenates. On the other hand, the Cz+-oxygenate selectivities are usually lowest with Co and Ni and in some cases it is not clear whether we should consider the catalyst “Co-Cu-Ni-alkalis” as e.g. a promoted Cu catalyst or a promoted Ni or Co catalyst. The most versatile is Rh which can be promoted to either a rather high selectivity to CH30H or C2+-oxygenate (ref. 26) and in a pure state or with an inert carrier (quartz) it is a catalyst for the formation of hydrocarbons (ref. 18). The best promoters for Cz+-oxygenate formation are the group 111-VII transition metal single oxides (refs. 42,7643) or combinations of them. Sometimes, alkalis are reported to improve further the selectivity already achieved and there are claims that some other oxides (MnO) are mainIy promoters of activity (refs. 82,83) while others modify the selectivity. Therefore, classifications appear repeatedly in the literature, cathegorizing promoters as the activity and the selectivity promoters respectively. However, one can find in the literature so many exceptions to these rules that doubts about the validity still exists ((ref. 83) see also discussion on this paper]). Oxides of transition metals as well as alkali oxides can interact with some camers and appear finally on the metal surface (ref. 84), where the compounds formed from carriers and added oxides can play a promoting role. Little attention has been paid up to now to the fact that
129
tommercial silicas and other oxides contain often contaminations in minute amounts but nevertheless amounts sufficient to promote the metal considerably when they had been leached out
of silica of an other oxide. The leaching is more complete the higher the concentration in the solution of the acidic metal precursor. However, alkali compounds react with Si02 eagerly and they can thus also make the contaminations mobile and promoting, by displacing them out of silica. Because of these potentially very complicating aspects of promotion, a lot of information in literature has to be accepted only after a very critical reading. One has also critically approach the data concerning the size-dependent effects, when the variations in the metal particle size had been achieved solely by variations in the metal loading. A warning of a similar content has been issued by Burch, during the Faraday Symposium in Bath, 1986. Two points put forward above are worthwhile repeating here
1) 2)
Oxides which modify the metal can either be used as a support or as a promoter. Whatever the way of preparation, and irrespective whether the essential oxidic component of the catalyst is a carrier or a promoter, the beneficial effect is only achieved when the promoting species are on the metal. Their presence under the metal particles or on a distance on the support is not sufficient.
Let us now turn to the problem “what is affected by the promoter”. Metals of low or zero activity in CO dissociation like Pd, Ir and Pt show from all oxygenates the highest or exclusive selectivity to methanol. However, when CH, units are added from gas admixtures to the syngas (e.g. in the form of CH,CI,.,,
withx< 3 ) , Pd catalysts produce more ethanol than methanol
(ref. 85a). The propensity of Pd to produce C2+-oxygenates can be also enhanced by V- promoters. It points (but does not prove, of course) to that that the C - 0 bond dissociation is promoted by V-promoters (refs. 25,85b). This conclusion is supported by the temperature programmed reaction experiments, with vanadium and other promoters (ref. 86). Also some other papers demonstrate this function of V-promoter, which seems to be the best promoter of the C2+-oxygenate synthesis. Fig. 4.5 qualitatively indicates and calculations confirm it that in comparison with the unpromoted catalyst, the formation of C2+-oxygenate is promoted by the catalyst with the optimal V E h ratio more than the total production of all CH, groups together (ref. 25). Unlike the cu catalysts, the transition metal based catalysts are usually supposed to operate by a mechanism consisting of the formation of CH, units, recombination of these units into an hydrocarbon chain and termination of this polymerization by CO insertion into a metal alkyl or methyl-carbene bond. Aldehydes are, according to most of the authors, easily hydrogenated to alcohols. Strassbourg laboratory showed (ref. 78) that oxidic promoters promote such hydrogenation. Data obtained with (ultra) pure silicas (ref. 45) revealed that unpromoted supported catalysts produce only very little of ethanal but practically no ethanol. This indicates the following problem: are aldehydes and alcohols formed by two consecutive steps (refs. 56,88-90) or by two parallel running reactions? There are indications that the second charateristics is true (refs. 25,87-90), but the question must be (for several reasons) considered as still open. One can only speculate at the moment in which step or steps a promoter exerts its improving action. The possibilities are:
130
i) ii) iii) iv)
influence on the CO adsorption; for example by activating CO to CO-insertion or CO-dissociation, respectively; beneficial stabilization of intermediates like formates, acetates or generally - carboxylates; improving the rate of hydrogenation of aldehydes formed on a metal (ref. 78); suppression of hydrogen adsorption (ref. 91) or CO dissociation (blocking ensembles).
The influence of various oxides (potential or real promoters) on CO adsorption can be followed conveniently by IR spectra (refs. 94-98). Two kinds of effects are usually observed: a) a modest shift in the (CO) of the single, double or triple coordinated CO adsorptions; b) one absorption band is shifted much more or it should be perhaps better considered as a new band. This band is with some metals and promoters in the region where also adsorbed carboxylates show the strongest absorption band (characteristic of assymetric vibrations of “CO,” groups). The band is formed on surfaces rich in OH groups or in the Corn, mixtures. The carboxylates can be hydrogenated away from the surface. They are usually active, but their activity’is considered by most authors to be low to suspect the IR visible carboxylates to be the reactive intermediates. Most of the carboxylates is bound to the oxidic component of the catalyst (refs. 42,92). Yet, these carboxylates are with silica supported catalysts only seen when the catalyst is found to be active in the formation of alcohols (refs. 25,92) so that there is some relation between their appearance and the CToxygenate activity and selectivity. In conclusion the available data make all ideas on promotion quoted under i)-iii) above plausible, but still they do not say enough on the relative importance of the individual effects in the total effect of a promoter. At the moment one has to consider them all as acting simultaneously.
4.2
ADSORPTION OF CO
The adsorption of CO is very important for the activity and the selectivity of the syngas reaction catalysts, therefore, it will be discussed in more details.
4.2.1 ADSORPTION OF CO OM METALS Carbon monoxide is a molecule with an extremely strong C - 0 bond (1.06 MJ/mol dissociation energy) and only due to the fact that it posseses “reactive” orbitals (50, 2x7 by which it can very easily interact with metals and ions, the CO-chemistry offers such a variety of reactions. Among them also the CO bond breaking by many metals, since the affinity of many metals (see above) to oxygen and carbon is high enough to make the breaking of the CO bond energetically possible (refs. 11,99). Carbon monoxide has a dipole with a negative end on carbon. The existence of the almost non-binding So-orbital and of electron-electron interaction is responsible for the dipole orientation and size (refs. 100,101). With transition metals, CO interacts with C-oriented to the metal and 0 away from it. According to all available information, the orientation of the molecularly bound CO is always perpendicular to the surface, within a small angle (for review see (ref. 101)). This orientation is - actually against the intuitive expectation-preserved also when CO is adsorbed o d o r
131
in the neighburhood of “steps” on the metal surface or next to an atom of another metal, as e.g. alkali atom (refs. 101-103). When CO has to undergo a dissociation on a metal (a pure or promoted one) it is very likely that CO must tilt (an activated movement (ref. 104)) and bring 0 in contact with the surface. However, there are only very few pieces of information on the - possibly - horizontal laying CO. This is a very elusive hard to detect transient state (ref. 105). Adsorption of CO on a zero valent metal is accompanied by a decrease of the C - 0 bond strength. A delocalization or partial removal of 5 0 electrons increases slightly the CO bond strength but this effect is overweight by the effect of acceptance of electrons into the 2n* (antibonding) CO orbitals, by which step the CO bond is weakened. The interaction of CO with a metal manifests itself by changes in the v(C0) the characteristic CO frequency (wave number). Resulting shift Av=v(CO, free)-v(C0, ads.), is built up from several contributions which has been roughly estimated for the single-coordinated CO on a metal like Pt as follows: an effect of mechanical binding to a heavy body: +30 cm-1; adsorbed dipole-image dipole interaction: -50 cm-l; effect of chemical binding usually described as direct and back donation of electrons: -60 cm-l; mutual interaction of adsorbed CO molecule: about +30 cm-l; all with regard to v=2143 cm-l, of the free CO (ref. 106). The dynamic dipoles of the adsorbed CO molecules interact with each other and this is the main component of the CO-CO interaction. This interaction leads to an increase in v(C0) with O(C0). When a l2C160 dipole layer is diluted by dipoles vibrating at a different frequency (e.g. those of 13C0 or C180) the interactions are suppressed (refs. 107,108), since they are strongest at the resonance frequency. Also alloying causes such dilution effect if CO adsorbed on individual alloy components differs significantly in v ( C 0 ) or when one component does not adsorb CO at certain conditions at all. Below more about CO adsorption by alloys. Carbon monoxide adsorbed on metals and alloys can be visualized as bound to certain “sites”. Sites consist of one or more contiguous metal atoms (ensemble of atoms). One can distinguish in the IR spectra regions of v(C0) which can be ascribe to certain ensembles, 1.e. to certain coordinations of the adsorbed CO: 2000-2130 cm-l; single coordinated (linear, terminal) double coordinated (bridged) 1880-2000 cm-l; 1700-1900 cm-l. triple (tripodic), multiply coordinated Carbon monoxide molecules interact with each other and therefore also at very low coverages they form layers of ordered structures observable by surface crystallography (LEED), or they form ordered patches clusters. Crystallographic information together with the IR spectra allowed a very detailed and well supported assignement (see e.g. (ref. 109)). The CO-CO interactions cause that CO prefers different sites at different surface coverages O(C0). This has even lead to temporary doubts about the site character of CO adsorption (ref. 1 lo), but the present generally accepted view is as described above (ref. 109).
132
There has been a discussion whether CO molecules can interact with each other through-the-metal. In principle, a through-the-metal interaction of covalently adsorbed species does exist (refs. 111-113) but estimates of its strength (ref. 114) show that this interaction is only of the size of the physical adsorption forces. Blyholder (ref. 115) and many authors after him, speculated on “competition” effects: CO molecules compete for the metal d-electrons and this competition suppresses the extent of backdonation from d into the 2a*-orbitals. However, it appeared that with some surfaces the v(e=l)-v(e=O) difference is exactly equal to that identifieable as a result of the CO-CO dynamic dipole interactions (ref. 116). However, with some other surfaces the CO-CO interactions leads to the shifts in the position of CO adsorbed and by that an additional V(e=l)-V(e=O) effect is created (ref. 117). This additional effect has been (in our opinion incorrectly) ascribed to the chemical effects, like the mentioned competition for the d-electrons. Small particles of metals, with their atomically rough surfaces, restrict shifts in the CO positions. On the other hand small particle adsorbents show a variety of different CO sites. Thus no evidence has been found yet for “competition” or any other strong through-the-metal interaction. When CO is coadsorbed with species having a dipole or forming a dipole with the metal surface, the dipole of CO is influenced (Stark effect) and to some extent also the occupation of the CO orbitals. These coadsorbed species cause then, next to a dilution effect, additional shifts in the v(C0) by this electrostatic mechanism. This is also probably a part of the v-shift caused by the fragments of hydrocarbon molecules or by benzene molecules (refs. 96,97,118-120). Finally, if the
coadsorbed species influences the position of CO on the surface, again an additional Av(C0) effect results from it (ref. 97). There are several pieces of information in the literature on the different reactivity of CO adsorbed on different sites i.e. differently coordinated. With Ni it seems to be very likely (ref. 121) that CO in multiply-coordinated form dissociates first. With other metals the information is less complete or less certain.
4.2.2 ADSORPTION OF CO ON ALLOYS With individual metals one has to consider the effects of crystal face specifity and related to it effects of particle size. With alloys or bi- or multimetallic catalysts, new aspects enter the scene, to join the just mentioned ones. When an alloy is formed from metal A and B (see Fig. 4.7) the metals dilute each other in the surface. It is then more difficult to find ensembles (2, 3 or 4 contiguous atoms) of atoms A or B in the surface than to find single atoms. At low concentrations of B in the surface of a random alloy, the probabilities are proportional to X,”, when X , is the atomic ratio of B atoms and n, the number of atoms in one ensemble. For comparable concentrations of A and B, i.e. when an atom of, say B can belong to more than one B-ensemble, the statistics are slightly more complicated. An additional complication is the deviation from the random distribution.
133
Several molecules involved in syngas reactions bind to several metal atoms
Metals
~
l
l
~
D
hydrocarbons bound by bond (e.g. ethylidene or ethylidyne like). This makes the syngas reactions sensitive to alloying as found
KIA
Surface:
(an ensemble) when possible: CO and also a carbenic or carbynic
~
surfaced--+ bulk 4 -A-
I
D(A-AI,
Q+ + + B-A-
I
I
f
f
-t
I
l
l
B - B- A -A-
I
B
I
D(B-B)
9+
B - B-6-
I
I
A Hsegr
Fig. 4.7 Alloy surface schematically.
experimentally (ref. 122). Alloying changes the composition and availability on the surface of ensembles required for the reaction. With CO, alloying of an active (in adsorption) metal (Ni, Pd) with an inactive one changes the population of the multiple and single coordinated CO (refs. 123-125). Also when the concentration of surface atoms of the required type is known, it is all but easy to calculate from it the number of available ensembles of certain size. It is namely not known if, for example, a tripod - like adsorption of a hydrocarbon species or of CO needs just three atoms in the surface or whether it requires next to it also an octahedral or tetrahedral hollow between them and atoms of a certain type underneath. (N. B. In deriving n from the comparison with the experimental data, this complication may lead to unreasanable high values of n in the XBnterms). Concentration of A and B in the surface of an AB alloy depends at given bulk concentrations (these determine the contribution of the mixing entropy to the free energy of the system) on the two factors (refs. 125,126). 1) Energy factor (heat of surface segregation). This stems from the fact that the alloys try to minimize the component in the surface with the higher dissociation energy of bonds MA -A ), D(B-B)I 2) Thermal entropy factor. With a given metal there is always a difference between the entropy of atoms in the bulk and of those in the surface. In earlier theories this was neglected but one was more and more forced to realize that this entropy difference is different for different metals and forms an additional driving force of segregation (ref. 127). The existence of a role for this terms is related to the fact that the atoms in the surface are bound to the rest of the metal by bonds of slightly different strength then atoms in the bulk of metals (ref. 128). Determination and theoretical prediction of the surface composition in multimetallic systems (with bulk metals and even more with powders or alloys on supports) is not easy. Most of the methods (electron spectroscopies) measure the signals originating from several layers under the surface, reflecting the composition along the whole penetration (actually - escape -)depth. The segregation
134
is also depth dependent and usually not limited to the outmost surface layer only. However, for these details the interested reader is referred to the reviews on this subject (e.g. ref. 126). These were the factors related to the geometrical or “ensemble size” effect of alloying. However, one can expect that the binding on a certain surface atom depends on its immediate sorrounding (see the different 1 and 2 positions in Fig. 4.7). This so called electronic or ligand effect of alloying has been a subject of vivid discussions in the literature and on many meetings. Methods of solid state physics, mainly the valence band UPS/XPS (integrated intensities) and core level XPS spectroscopies, do not supply data which would suggest a charge transfer between the alloy components. However, the data show that in number of cases (for example, in catalytic important systems Pd/Ag, Pd/Au) a redistribution takes place of electrons between the orbitals (parts of energy bands) of different symmetry. The question is whether such changes in populations are reflected by the chemisorption and catalysis behavior. There have been several claims that heats of adsorption results would show some “ligand’ effects. However, when the obvious artifacts are neglected, the rest of the data shows that alloying influences the population of different adsorption states but not the quality of the states. The most likely conclusion from that is that the availability but not the quality of atoms (i.e. of “sites”) is changed by alloying. The extended discussion on this subject has been already published (ref. 122). Adsorption of CO used as a probe and monitored by IR spectra showed that when the effects caused by dilution of the adsorbed layer are eliminated, the remaining effect which could be ascribed to the ligand effects is smaller than the size of the CO-CO interactions (refs. 106,122,129,130).
4.2.3 ALLOY BASED CATALYSTS We shall not discuss individual alloy systems, but - instead - we shall formulate some generalizations and put forward some general rules which might be useful for further studies. Let us remind the reader that the surface segregation occurring in alloys at the vacuum - metal interface can be changed and even reverted in sign, when gases are admitted to the alloy (refs. 125,131). By that, the catalytic properties are through the surface composition dependent on the composition of the gas phase. This can be a particularly complicating factor with syngas reaction: at the entrance of the reactor the gas mixture is strongly reducing, while at the end of the reactor (when CO/H, is consumed and hydrocarbons oxygenates and water are produced) the mixture can become strongly (for Fe as an example) oxidizing. An extreme of a gas induced segregation is a formation and separation of newly formed phases (oxides, carbides) of individual metals or their mixtures (mixed oxides). Alloys with easily oxidizable metals (metal-glasses, alloys with lanthanides) separate in this way and convert themselves in completely new materials under the reaction running with them. 4.2.3.1 Grow Vlll-lbmetaIs Although some of the combinations have been tested for a possible industrial use (ref. 132) the main benefit from the studies on these systems was an extension of the fundamental knowledge. It has been established by using these alloys that i) CO dissociation is a large ensemble reaction
135
(ref. 8c) ii) the presence of the inactive metal suppresses the surface concentration of active carbon and CH, intermediates and that of adsorbed hydrogen (see Fig. 4.8). Most likely, some secondary reactions like the reverse C-C bond fission are suppressed too. The combination of the effects under ii) causes, that alloying has different consequencies for the selectivity to hydrocarbon synthesis in the low and high temperature region (ref. 133) as can be seen in Fig. 4.9. As mentioned elsewhere, Ni-Cu and Co-Cu in combination with other promoters and on supports can
-2
q h
0-0-2
-3-i_. -44
u 20 60 80 100
0
LO
Cu I a t % I
produce also CH30H and C2+-oxygenates (refs. 79,80). 4.2.3.2 G r o w VIII - Grouo VIII Metals Combinations of metals like Fe-Ni, Fe-Co, Ni-Co, Fe-Ru, Rh-Ru active in FT synthesis, are offerring some advantage in comparison with individual metals, sometimes in a higher activity or selectivity in the low olefin formation (refs. 134-135). A combination of the FTS catalyzing metal with a good hydrogenation catalyst (metal) which does not dissociate CO so easily (Pt, Ir, Pd),
leads to catalysts which show synthesis of oxygenates and in particular that of methanol (ref. 136). An example of a study on these bimetallics is shown in Fig. 4.10.
VIII grow metaIs-earlv transition metals Metals of the lower groups are oxidized under the conditions of the FTS, so that the working catalyst is actually a transition Group VIII metal promoted by an oxide. The combinations Fe-Mn and Co-Mn sometimes further modified by additives as sulphur or base-metal-oxides (refs. 137,138) has some promise as a catalyst of an enhanced low olefin production. Rhodium combined with Mn (or Fe) was one of the first catalysts suggested for a C2+-oxygenate production (ref. 139).
Fig. 4.8 Activity (rate in arbitrary units, logarithmic scale) as a function of alloy composition, T=593 K, 1 atm, standard conditions; powdered alloys prepared from carbonates. 1) formation of C2 and C3 hydrocarbons; 2) formation of methane; 3) data obtained with evaporated films (T=573 K, total pressure 0.6 torr) (from ref. 47)
0 2 i
\
4.2.3.3
Fig. 4.9 Comparison of selectivities in the C2+-hydrocarbon formation over three unsupported powder catalysts (ex. carbonates) (1) pure Ni; (2) 3% Cu-Ni alloy; (3) 10% Cu-Ni alloy. Notice the different influence of alloying on the selectivity at low and high temperatures (ref. 133).
136 100 -
Methanol selectivity Copt
FePd
NlIr
%MeOH
.
U
0
10
20 30 time ( h l
LO
50
Fig. 4.10 Selectivities in CH30H formation as a function of the time on stream of various silica-supported bimetallic catalysts. (40 bar total pressure, H2/co=3, all catalysts at 545 K only FeRu at 515 K (ref. 136))
Phase separation is threedimensional and almost complete. One has to consider these systems as metals promoted by lanthanide oxides (ref. 141).
Here the same holds as said under d) (ref. 142). This is demonstrated by - IM
Fig. 4.11. Copper itself is no catalyst for CH30H formation (or such a poor one that
Hydride
the question is whether possible contaminations are not responsible for the measured “activity”), but the mentioned alloys of it are quite active after a period of ’
25
35 TWO-THETA
L5 IDEGREESI
Fig. 4.1 1 Synthesis gas activation of NdCuz precursor. Sequential diffraction patterns obtained at 35 min intervals during treatment at 8 bar and at 448 K. (a) Corresponds to virtually untransformed starting alloy. (b) Small diffraction features from residual alloy removed for clarity.
activation. During the period of activation active site are obviously created. According to what has been said earlier, these can be Cuo sites under the promoting influence of the respective oxide or Cu”+ site stabilized in that oxide. It should be mentioned on this place that under sections 4.2.3.1 to 4.2.3.5 references are quoted which should
137
jellium free electrons
Fig. 4.12 Image forces, schematically (see the text, for details)
illustrate the point of question, without an intention to make the list of papers on the indicated subject complete.
4.3 PROMOTION OF METALS. THEORIES AND THEIR VERIFICATION 4.3.1
SOME PHYSICAL PHENOMENA RELEVANT TO THE PROMOTER - METAL INTERACTION
4.3.1.I
Point charpe and dinole metal interactions: imape forces
Let us consider a situation in which an electron escapes from one of the atoms in the surface of a metal. The positive charge left behind in the metal would move to the centre of the metal, to create the maximum interaction of itself with as many metal electrons as possible (ref. 143). An electron interacts with such “mirror image” in such a way that the interaction decreases the energy
of the system by W=-e2/(41Z01)(see Fig. 4.12). Bardeen (ref. 144) analyzed later quantum-mechanically an interaction of a charge with a system of metal “free-electrons”. He showed that the charge polarizes the sea of the metal free-electrons in such a way that the interaction energy of the charge with the metal is expressed by a complicated equation which, however, can be very well approximated by W as above. Bardeen applied this theory also to the interaction of dipoles with the metal and derived in this way an expression for the adsorption energy of a molecule with a (static or fluctuating) dipole on a metallic surface. Dipole interaction energy is again formally the same as between dipole and its image in the metal. Obviously, the equation for W is only an approximation which holds for ZObetween approx. and lo4 cm. Various other equations have been suggested for smaller distances where one has to consider screening effects. When screening is involved the unreasonable limit of W for ZO-0 is
138
corrected. An example of such theories is in the papers by Henrichs and others (ref. 145). However, it is not an aim of this section to discuss the theory in all details. We shall rather turn our attention to the possible application of the mentioned ideas in the theory of promotion. It has been frequently speculated that an image charge in the metal caused by the positive charge above the metal, can be more easily donated out of the promoted metal than an electron of the promoter-free metal. However, the following has to be remarked. First, the Schottky-Bardeen theory does not speak about an increased electron density on the image distance. However, even if there were such increase in density, the electrons would not be pushed by that to be donated out of the metal. Energy W is always negative and the metal electrons would under the influence of a positive charge be less “donated” and not more.
Adsorption of strondv electrodonatinP suecies This case is in many respects similar to the one 4.3.1.2
discussed in the previous paragraph. When an atom with a very low ionisation potential (alkali atom) is adsorbed on a metal with a high work function, the adsorbed atom becomes almost fully ionized and the analogy with 4.3.1.1 is, indeed strong. Also the methods used to
Fig. 4.13 Contour maps of the induced electrostatic potential for (a) Li, (b) Na and (c) K at their equilibrium distances outside a jellium (rs=2) surface. The vertical line denotes the jellium edge and the crosses the atomic positions. The vacuum side is to the right. Contour values of +2, +1, +o, 5 , f O , 3 and +0, 1 eV are shown within a sphere centred at the atomic position and having a radius of 7 bohr. (ref. 147)
describe the system are similar (refs. 146,147) in both cases. The charge shifted towards the metal is when screening operates fairly localized in the nearest neighbourhood of the adsorbed atom and the map of the density of electrons does not suggest that there is a strong through-the-metal interaction by that. On the other hand, at horizontal distances at which other molecules (CO, hydrocarbons, etc.) can be coadsorbed, a strong electrostatic field exists which is sufficiently strong to be a potential source of promotion effects. The distribution of charge is seen in Fig. 4.13.
139
An estimate of the field at the potentially nearest site of CO adsorption reveals a size of about 0.2-0.5V/A (ref. 148). Through-the-vacuum effects of such field can
be, at least in principle, quite strong. Promoters are usually not present as atoms (it is very difficult to keep alkalis in reduced zero valent state: under conditions of FTS this is completely excluded) but as compounds: KOH, K20, K2C03, etc. To estimate the effect of the counter ion, authors (ref. 147) investigated a coadsorption of K and 0 as adsorption of two mutually non-interacting species. This is rather crude approximation but a sufficient one to demonstrate the expected effects. The results can be seen in Fig. 4.14. The strong electrostatic effect of an alkali ion is in a segment of the space preserved.
4
Fig. 4.14 Superposition of the self-consistent electrostatic potentials for isolated K and 0 cutside a jellium (rs=2) surface, illustrating the predominance of K over 0 (ref. 147).
m of oadsorbed species
The existence of such interactions was first predicted by theoretical papers, quite a long tome ago (refs. 111-114). It has been shown that the interaction is of an oscillating character, but it took some time until1 a realistic estimate could be made of the influence of the through the interactions on the coadsorption bond strength. It appeared (ref. 114) that in a monocomponent adsorption layer this type of interaction changes the chemisorption bond strength on the nearest site by about 3% (ref. 114). (The estimate has been done for H, and Re on W). The influence on the chemisorption bond strength is only a part of the effect. One has also to analyse which changes in the charge distribution and local density of state functions can be evoked by coadsorption. This has been done in several recent papers (refs. 149-151) and the conclusion is that the changes in the charge do not propagate into the metal and also upon adsorption of strongly electrodonating atoms (alkali metals) the transferred and the created charge are strongly localized. This is the consequence of a very effective screening by metal electrons. However, the changes in the local density of states usually propagate further in the metal. Feibelmann and Hamann (ref. 150) stress this point but add “it still remains to learn just what the role of the local density of states (LDOS) at the Fermi energy might be”. This caution was justified but was abounded by other authors who interpret the very small changes in LDOS as an indication of the importance of long range interactions through the metal (ref. 149). However, as far as the bonding strength is concerned this can never be a larger effect than the one discussed in the foregoing paragrapgh. The
I
heat of adsorption contains terms: N ( E ) E dE, where N ( E ) is the density of states, calculated for the system before and after the adsorption, and the effect of small deformations of the density of state function (or of its local projection-LUOS) cannot be very pronounced.
140
interaction If two metals with a different work function @ are brought into a contact with each other, the electrons flow from the metal with the lower to the metal with the higher work function. The first metal becomes positively the other negatively charged and a contact potential AV is build up between the metals which stops further transfer of electrons. Below q is the electrochemical potential or the Fermi energy of the metal electrons there are synonyms measured from the vacuum level, V the electrostatic potential of the metal with regard to “vacuum” and V+X (X is the potential of the surface double layer) the distance of the vacuum level to the Fermi energy EF in a charge free (potential free) metal (see Fig. 4.15). The following equations hold at equilibrium between the two metals. With @=-qi- eVi
The charge transferred is very small, as shown already in earlier textbooks on this subject (ref. 152) and it also cannot change @ by increasing sufficiently the highest occupied level (ref. 153). Let us now deviate slightly from the topic and discuss several consequencies of that what is stated above, for the problems of promotion. Briefly it is the following: i) Work function, CD, is defined in such a way that it is independent of the electric potential,
ii)
V, on the metal. Therefore, it is not correct to speculate that a work function of a metal can be changed by merely the electrostatic fields of a support or promoter The energy difference between the Fermi energy (highest occupied metal level at 0 K) and the vacuum level is in general not CD. Fermi energy (eletrochemical potential) is the total free energy per one mol electrons and it comprises therefore also the electrostatic energy (the eV term).
iii)
The metals at equilibrium have the same Fermi energy, E , (or q), but we do not know the position of it with respect to the vacuum. The distance from the vacuum zero level
iv)
comprises the already mentioned external elelectrostatic potential, including also the contact potential difference AV calculated above. Fermi energy can be defined with regard to the vacuum level, as above, or as solid state physists in most cases do, with regard to the bottom of the conduction band. The position is always the same (at O K the highest occupied level) only the numerical values are different in the two just mentioned cases. Further, it is important to note that in the first case the numerical values of EF do not change by adsorption or coadsorption, but in the second case they do. Ignoring these simple things has caused many errors in the literature on metals adsorption or promotion in particular in the discussions on the XPS, UPS spectra, in calculations of thermodynamic data from XPS data, etc, such error can be find in sometimes otherwise excellent papers (ref. 154).
141 Vacuum
We shall turn our attention
now
back to the metalsemiconductor interphase (metalsupport/promoter interaction) and analyze what the theory teaches us. The theory explaining the phenomena of the transfer Fig. 4.15 Graphical presentation of various definitions: work function (a), charge electrochemical potential (q) or Fermi energy (EF) (the last two are between solids, physically the same parameters) electrostatic potential, V , potential of the revolutionalized the surface double layer (W. electrotechnical industry and was crowned by a Nobel price. No wonder that people studying catalysis thought that such a theory should explain catalysis as well (refs. 155,156). Ideas of this theory have been also applied to explain the modification of the catalytic properties of metals by a support or promoter. Schwab (ref. 157) and Solymosi (ref. 158) were the first who speculated in this direction and later the idea came back in the interpretation of the SMSI phenomena. The idea was that the support or promoter, with a high laying E , (n-semiconductor) can donate electrons into the metal and change its properties considerably. It has been even speculated that by such a transfer a particle of Pt can behave more like Au or Ir, according to the support (promoter) used (ref. 159). Katzer et al. (ref. 50) thought that basic oxides can donate so much electrons to a metal particle that the particle can be changed from one producing hydrocarbons from syngas into another one, which produces oxygenates. More of such examples can be found in the literature, since the idea does not loose its popularity with years (ref. 160). However, the question still is whether the idea is physically correct. There are two problems with the above mentioned idea: Can enough electrons be transferred to change the particle chemically? ii) If i) is impossible, is it then possible that a small charge transfer would in any case change the behavior of the particle by a through-the-metal-interaction? As far as the first question is concerned a rather reliable estimates of Ne number of electrons transferred from the semiconductor onto the metal are available (ref. 161). Electrostatic potential V is related to the density of charges by the Poisson equation: i)
<
(4.3)
When electrons are transferred from the semiconductor donor levels onto the surface of the metal in contact, V inside the semiconductor decreases. The transfer of electrons continues so long
142
K- --1
AV
--
I
._I
4
I--.+*
I I I
+++
Valence band
I
Fig. 4.16 Bending of bands in the n-semiconductor as a consequence of electron transfer to a metal with a high F . The depth in which the donor levels are exhausted in the semiconductor (L) is indicated.
until a barrier is formed which prevents further transfer. At that stage the donor levels are exhausted up to the depth L in the semiconductor. If the concentration of charges created by the electron transfer (volume concentration of ionised impurities) is No, the charge transferred at equilibrium is
N&=Ne; according to (ref. 161), potential barrier AV is AV=(2d) NoL2/€
(4.4)
Thus,
Ne=d(No~A1/3/2nl
(4.5)
With €=lo, N0=lO1*/cm3, AV=2eV, which is a resonable estimate, N, transferred over onto 1 cm2 of a metal in contact is 3.5 1012/cm2.It means that about 0.003 electrons are transferred per metal atom of the interphase. As far as the second (ii) problem is concerned, there is information on it in the literature too. Calculation represented by the Figs. 4.13 and 4.14, or calculations by Feibelman and Hamann (ref. 150) show that charges transferred from adsorbates stay localized at the interphase. Of course, the same should hold for charges brought on the metal by a transfer between phases. This is the consequence of screening which the metal electrons can accomplish. The effect of screening has been explicitely analyzed by Smith et al. (ref. 162) who established that when a positive hole is located on a certain metal surface atom (e.g. upon ionisation in XPS), the electron accumulate to screen this charge on a distance of about of the distance to the nearest metal surface atom. Thus even the nearest neighbour does not feel much of the hole created by ionisation or by any other
143
process. Thus, the transfered electrons are not many and they do not influence the metal in the bulk. One can conclude that the theory does not offer a physical basis for a Schwab-Szabo-Solymosi effect.
4.3.2 MODERN THEORIES OF PROMOTION EFFECTS IN THE SYNGAS REACTION Throuplz the vacuum effectsof uromoters and of coadsorbed snecies. No effects, of course, can be totally limited to either “through the vacuum” and “through the metal” region. However in the hope that the text makes clear what the titles (i.e. our cathegorisation of effects) really mean, we have chosen for the terminology used here. It circumwents amongst others the problem of the “short range”-“long range” terminology. The analysis which we mentioned above (refs. 146,147) showed that when a positive charge (like that of alkali metal atom adsorbed on a metal with a high work function) is placed above the metal surface, the electric field created on the nearest adsorption sites is of the order which is common in electrochemical double layers (refs. 163,164) or with the Field Emission Microscopy (ref. 168). In principle, this can influence the reactivity of bonds, mainly of heteropolar ones. Several authors (refs. 146-148,163,164) predicted a role for this field, in the activation of adsorbed molecules and in catalysis. A similar conclusion has been arrived at by authors who approached the promotion problem from the side of molecular orbital theory (ref. 165). This is certainly a possibility which should be considered in the theory of promotion effects. Among the papers which appeared later, two analyzed the promotion effects by quantum mechanical methods of high accuracy. Obviously, these methods are only applicable to models, but c.leVI b the models were chosen to reflect the most relevant aspects of promotion. The models used by Bonacic-Koutecky et a1 (ref. 166) are presented in Fig. 4.17. Fig. 4.17 shows the change in the valence MO-energies (SCF-CI method) caused by the presence of Na+ in the indicated positions. Placing NaO, instead of Na+ causes a lower stabilization and Nao or Na+ placed next to Pd instead of next to CO, destabilizes the Pd-CO complex. Thus, no through the Pd atom stabilization is possible. Several, in our opinion important conclusion, can be made on basis of these calculations: i)
Stabilization is not very sensitive for the position of Na+, when Na+ is in the neighbourhood of CO,
Fig. 4.17 Molecular orbital energies of complexes indicated (ref. 166)
144
ii)
The shift in MO energy levels is rather uniform and no signs of rehydridization (predicted
by some authors (ref. 167)) are detectable The points i) and ii) indicate together that also on the very short distances of interacting particles the effect of Na+ can be best described as an effect of an electrostatic field. The changes of iii)
electron populations accompanying this effect are rather modest but could be important. A Mullikan type analysis reveals that if the number of K electrons is 4 in a free CO molecule, it is 4.0989 when CO is bound to a Pd atom. Under the influence of the Na+ ion this figure changes into 4. 1684, (model iii) and 4.2078 or 4.2192 (model ii), indicated by the geometries I11 and I1 at the bottom of Fig. 4.17, respectively. The two values for I1 are for two different distances in the complex. A disadvantage of a model which can be analyzed so accurately is that it is unable to reveal any effect related to the metallic nature of the catalysts. However, Wimmer et a1 (ref. 168) used some other methods of high accuracy and they analyzed a model consisting of a metallic two dimensional infinite slab consisting of three atomic layers, with a molecule of CO on it. The authors analyzed an effect of KO and S on CO and they came to the same conclusion as above, namely, that the main effect of the K+ created is essentially an electrostatic field effect. The authors find next to it some indication of an interaction between the alkali atom 3p electrons and the binding IT orbitals of CO. Anderson and Dowd (ref. 169) analyzed the interaction of biatomic species ZnO, FeO and Ti0 with CO on the (111) surface of a 22-atom, 2-layer thick cluster of Pt. The method used (Atom-Superposition and Electron Delocalization Molecular Orbitals Method) is not accurate enough to predict reliably dissociation energies but probably it can be trusted in questions of adsorption geometries. It is interesting to see that the authors predict the equilibrium configurations such as shown in Fig. 4.18. Next to T i 0 or FeO, the CO molecule is tilted, next to ZnO it stays perpendicularly. The strength of interaction follows the indicated order Ti > Fe > Zn, which is related by the authors to the occupancy of the d-shells. It is already known that CO placed next to KO or next to a step of the underlaying transition metal stays perpendicular to the surface (ref. 103).
However, it is nothing known about the CO-geometry in the presence of Tin+ or Vn+ ions (the best promoters). A definitive conclusion with regard to the CO geometry would be very welcome since there are theories which relate the activation of 0 0 0 the CO molecule to the 0 Ti-C 0 FeC 0 Zn C interaction of the “0’-end of it with a positive ion and 0 0 speculate that when Fe C Pt adsorbed on a metal, a CO molecule is tilted
mmm c
Fig. 4.18 Equilibrium orientation of Co on Pt clusters promoted by FeO, ZnO and T i 0 (ref. 169).
with it’s “0” end towards the ion. The
145
promoting effects of this kind are very well documented in homogeneous catalysis (ref. 170). Tomanek and Bennemann analyzed CO adsorption on a K(Ni)3 cluster (ref. 171). They came to the conclusion that CO is parallel to the surface and the barrier for dissociation is lowered by K when compared with a pure Ni cluster. The picture which emerges now from the theoretical papers and from the theoretical analysis of the spectroscopical data (AES, XPS) is that with zero-valent alkali metals there is probably an additional possible interaction of CO e.g. with KO, which can be described as a K T O complex adsorbed on a metal (refs. 167,172). Such interaction is not seen when K+ is interacting with CO. On the other hand, K+ from a K+CO- complex can generate the electrostatic field effects which have been discussed above. There is a solid experimental support for the idea of formation of (K,CO,) complexes (ref. 173) on the surface of metals weakly adsorbing CO and also a good support by IR-spectra for the existence of two kinds of K--CO interactions, one is strong and localized and second, an electrostatic interaction [ 1481. Experiments with isotopic labelled molecules show that the CO molecules most strongly influenced and localized round the promoting alkali atom do not scramble atoms easily, while scrambling occurs under the influence of a promoter atom on a distance from it (ref. 148). In the last mentioned paper a mechanism is suggested which contradicts the microscopic reversibility and it is thus not very likely. However, it is not necessary to postulate such a mechanism if one accepts that the promoter atoms (ions, actually) can promote the CO dissociation also on some distance by the electrostatic field. There are also real possibilities of a more subtle influence exterted by the promoter. Surface states of some solids can be influenced by the external fields (ref. 174) and such a field can be supplied by a promoter-ion. Also creation of the new interphase states is in principle possible and these state might play some role in chemisorption (ref. 175). It has been speculated on the possibility that promoters-transition metal oxides (oxides of metals which easily change their valency) influence CO dissociation, CO insertion or intermediate formation (formate, formyl) by either the presence of anion (oxygen) vacancies in the surface (ref. 176) or by the surface OH groups (ref. 177). Interaction of CO with the OH groups on a surface is a step which appears in many reaction schemes suggested in the literature, mainly in relation to the CH30H synthesis (ref. 177). Sshustorowich published several papers in which he showed (ref. 181) that interesting predictions can :: made on basis of semiempirical “bond order conservation” calculations. According to this i&x a promoter influences the “order” of the bonds formed between the adsorbates or reaction intermediates and the surface, by which the heat of adsorption and various activation bamers can be influenced.
4.3.3 THEORIES AND THEIR VERIFICATION Experimental data comprise already some information on all relevant topics: i)
influence of a promoter on CO adsorption,
146
ii) iii)
influence on CO activation in the direction of the insertion reaction
influence of promoters on the (proposed) intermediates of syngas reaction. The most studied item is undoubtedly the topic i). One can conclude that the downshift in the v(C0) frequencies can be very well explained by the electrostatic field created by the promoter ion(s) (refs. 146-148,178). It is not yet known if this field is also sufficient to create chemical promotion effect but a positive answer would not be surprising. It is very difficult to discern experimentally whether, for example, the improved oxygenate synthesis would be achieved by a suitable CO activation (ref. 170) or by stabilization of oxygenated intermediates,’ like carboxylates (ref. 92), or even by both of these effects. However, the first mechanism - a proper CO activation - is well documented for homogeneous reactions (ref. 170). Obviously, more studies are necessary in this direction. On several places in this review theoretical and empirical arguments are presented against the idea of a long range (longer than the nearest neighbours) through the metal interaction. Empirical data show that this is true for promoters as well as poisons (refs. 179,181). AS an exception, Goodman and coworkers (ref. 180) reported that sulfur on Ni suppressed Corn2 synthesis by a larger factor than it did with CO adsorption and this factor was much larger than it could be reasonable expected from blocking (ensemble size effect). However, it appeared later that upon CO/H2 reaction, sulfur caused a massive “carbon” deposition so that the effect observed was due to sulfur and carbon together. So far thus, convincing evidence for a longe range through the metal effect, is still missing. On the other hand an electrostatic through the vacuum effect can be a rather longe effect (V-r-l). It is not known whether it really is so, since we do not know yet whether there is a catalytic effect of this field at all. However, the possibility is there. The problems still exist not only when verifying the theories but - unfortunately - also when interpreting the data. Some authors report data showing for example, an important effect of alkali and other promoters on the dissociation of CO (ref. 181), some others report a modest or zero effect of the same promoters (ref. 182). The papers should not be suspected of the trivial experimental errors but obviously the discrepancy shows that some factors escaped the control: a distribution of the promoter over the metal and the support, interaction of the promoter with the support with possible consequencies of it, minute contamination present in the supports or promoter-precursor (e.g. carbonates) etc. As far as the future concerned there is certainly an important role here for experiments with well defined single crystals, promoted in a well defined and well controlled way. Before the theory can proceed further experimental information is required on several key-points. With regard to methanol synthesis, for example the following ones. There are already strong indications that a promoter stabilizes the positively charged forms of the active metal (cu, Pd, Pt,Ir, etc.) but the question remains - has a promoter also some other functions like: i) in activating CO for the formation of intermediates (formyls (Pd) or formates (Cu)). ii) in stabilization of the intermediates (formate or formyls, oxymethylene and methoxygroup) iii) in hydrogenation of intermediates. Analogous questions are related to the formation of C2+-oxygenates. Here, moreover one would like to know whether the pathway to alcohols is the same like that one leading to aldehydes,
147
which in their turn can most likely be formed by alkyl-CO-shift (insertion) reaction. With regard to promoters the question is - what is their role in this system of consecutive (common pathway) or parallel (split pathways) reactions?
4.4 RELATED REACTIONS 4.4.1
WATER GAS SHIFT REACTION
One of the most important related reactions is the widely applied water gas shift reaction (w.g.s.) CO + H20
-+
CO, + H2
(4.6)
In principle this reaction can be considered as intermixing of three stechiometric reaction systems:
3 H2O
=
4 Hads
+
OHads
=
2Hads
+
Oads
Obviously, reaction 2 is undesired and the additives or a choice of catalyst should suppress it. When we stay with these reactions as being also mechanistic schemes, the mechanism of the reaction would be an “oxygen transfer”: oxygen from reaction 4 would appear in the reverse reactions 2. However, it is not likely that oxygen transfer
-
if operating at all
-
is the only mechanism.
Activity of metals is like with methanol synthesis dependent on the choice of a carrier (ref. 184) and this points to a role of the carrier in the reaction and to some other intermediates. There are two acceptable suggestions in this respect: i) ii)
CO,, CO,&
+ M-OH form a formate-intermediate (refs. 185-186) + M-OH form a hydroxycarbonyl (ref. 187).
It is not known which role the additives (promoter) play in this reaction but the possibilities are in principle the same as listed in the last paragraph of the foregoing section. The at present available data point very much in the direction of the stabilization of formates (ref. 186), but the question must be considered as open.
148
4.4.2 HYDROGENATION OF UNSATURATED ALDEHYDES These compounds can be hydrogenated on two places: the C=C bond and the C=O bond. If an unmodified VIII group metal is used as a catalyst, the C=C bond is hydrogenated first. However, when promoters are used (according to the scarce information spread in widely literature (ref. 188)) - these can be alkali compounds, lanthanides, early transition metal oxides, Cd and Sn-oxides and perhaps some others (see, for example, ref. 188). It points again to a specific interaction of the promoters with the C=O group. In principle it can be activated by an influence on the “ 0 ’ end of the molecule, an interaction with a vacancy in the promoter lattice (transition metal oxides) or by formation of carboxylic intermediates. A recent fundamental study also shows, that the rate of hydrogenation of acetone can be enhanced by the same promoters as the syngas reactions (ref. 189).
4.5 ACKNOWLEDGEMENT The author expresses his most sincere thanks to his former and present coworkers for the pleasure he had in working with them, and also for the permission to reproduce their results here. The benefit from many discussions with Professors A. T. Bell (Berkeley), L. Guczi (Budapest), W. M. H. Sachtler (Evanston) and R. A. van Santen (Eindhoven) on the subject of this paper, is highly acknowledged. The financial support by SON/NWO (The Netherlands), Science Committee NATO, Science Committee EEC enabled the work and the contacts with scientists in abroad.
4.6 REFERENCES 1 2 3 4
5
6 7 8 9
10 11 12 13
P. Sabatier and J. B. Senderens, C. R. Hebd. Seances, Acad. Sci., 134 (1902) 514 E. I. Orlov, Zhur. Russ. Khim., Ob-va 40 (1908) 1588 F. Fischer and H. Tropsch, Brennst. Chem., 4 (1923) 276 H. Fischer “Herstellung Flussiger Kraftstoffe aus Kohle ”Borntrager, Berlin, 1924 H. H. Storch, H. Golumbic and R. B. Anderson “Fischer- Tropsch and Related Syntheses”, J. Wiley, Chichester, 1951 H.PIChler, Adv. Catal., 6 (1952) 272 M. A. Vannice, J. Catal., SO (1977) 228 M. A. Vannice, Catal. Revs., 14 (1976) 153 V. Ponec, in Coal Science, Vol. 3, p. 1 , M. L. Gorbaty, J. W. Larsen, I. Wender, eds. Acad. Press, N. Y., 1984 (see also ref. 11 below) V. Ponec, Catal. Revs. Sci. Eng., 18 (1978) 151 J. A. Rabo, A. P. Rich and M. L. Poutsma, 53 (1978) 295 P. R. Wentrcek, B. J. Wood, H. Wise, J. Catal., 43 (1976) 363 M. Araki and V. Ponec, J. Catal., 44 (1976) 436 E. K. Poels, E. H. van Broekhoven, W. A. A. van Barneveld and V. Ponec, React. Kin. Catal. Lett., 18 (1981) 223 M. A. Vannice, R. L. Garten, Ind. Eng. Chem. Res. Dev., 18 (1979) 186 P. Biloen, J. N. Helle and W. M. H. Sachtler, J. Catal., 58 (1979) 95 V. Ponec and W. A . A. van Barneveld, Ind. Eng. Chem. Res. Dev., 18 (1979) 268 R. C. Braddy 111, and R. Pettit, J. Amer. Chem. Soc., 103 (1981) 1287 D. Dwyer and G. A. Somorjai, J. Catal., 52 (1978) 291
149
14 15 16 17
18 19
D. W. Goodman, R. D. Kelley, T. E. Madey and J. T. Yates, J. Catal., 60 (1979) 57 D. W. Goodman, R. D. Kelley, T. E. Madey and J. M. White, J. Catal., 64 (1980) 479 R. B. Anderson, J. Feldman and H. H. Storch, Ind. Eng. Chem., 44 (1952) 2418 R. B. Anderson, Catal. Revs. Eng., 21 (1980) 53 H. H. Storch, Adv. Catal., 1 (1948) 115 H. Schafer-Stahl, Angew. Chem. Int. Ed. En& 19 (1980) 729 R. B. Anderson, L. J. Hofer, E. M. Cohen and B. J. Seligman, J. Amer. Chem. SOC.,73 (1951) 944 (a full phase diagram therein) J. W. Niemantsverdriet, A. M. van der Kraan, W. L. van Dijk and H. S. van der Baan, J. Phys. Chem., 84 (1980) 3363 P. R. Watson and G . A. Somorjai, J. Catal., 72 (1981) 347 and ibid 74 (1982) 282 K. Klier, Adv. Catal., 31 (1982) 243 G. Natta, U. Colombo and I. Pasquon in “Catalysis”, P. H. Emmett, ed. Reinhold N. Y. 1957, Vol. 5, p. 131 G. Natta in “Cata1ysis”P. H. Emmett, ed. Reinhold N. Y. 1955, Vol. 3, p. 349 E. Tranconi, C. Cristiani, N. Ferlaszo, P. Forzatti, P. L. Villa and I. Pasquon, Appl. Catal., 32 (1987) 285
20 21 22 23 24 25 26 27
28
M. L. Poutsma, L. F. Elek, P. A. Ibarbis, A. P. Rich and J. A. Rabo, J. Catal., 54 (1978) 157 M. Ichikawa, Bull. Soc. Chem. Japan, 51 (1978) 2268 J. C. J. Bart and R. P. A. Sneeden, Catalysis Today, 2(1) (1987) 1 P. J. Berlowitz and D. W. Goodman, J. Catal., 108 (1987) 364 T. Immers and A. G . T. M. Bastein, unpublished results, Leiden University G . v. d. Lee, Dissertation thesis, Leiden University, 1986 A. G. T. M. Bastein, Dissertation thesis, Leiden University, 1988 G . v. d. Lee, B. Schuller, T. L. F. Favre, H. Post and V. Ponec, J. Catal., 98 (1986) 51 1 A. G. T. M. Bastein, W. J. v. d. Boogert, G. v. d. Lee, H. Y. Luo, B. Schuller and V. Ponec, Appl. Catal., 29 (1987) 243 M. A. Vannice, C. Sudhakar, J. Phys. Chem., 88 (1984) 2429
H. Y. Luo, A. G. T. M. Bastein, A. A. J. P. Mulder and V. Ponec, Appl. Catal., 38 (1988) 24 1 29 G. v. d. Lee, A. G. T. M. Bastein, W. J. v. d. Boogert, B. Schuller, H. Y. Luo and V. Ponec, J. Chem. SOC.Faraday Trans I, 83 (1987) 2103 30 S. J. Tauster and S. C. Fung, J. Catal., 55 (1978) 29 S. J. Tauster, S. C. Fung and R. L. Garten, J. Amer. Chem. Soc., 100 (1978) 170 31 G . B. Raupp and J. A. Dumesic, J. Phys. Chem., 90 (1986) 3359 ibid: J. Catal., 95 (1985) 587 32 D. E. Resasco and G . L. Haller, J. Catal., 82 (1983) 279 33 H. R. Sadeghi and V. E. Henrich, J. Catal., 87 (1984) 279 34 S. Takatani and Y. W. Chung, J. Catal., 90 (1984) 75 ibid: Appl. Surf. Sci., 19 (1984) 34 35 S. J. Tauster, Acc. Chem. Res., 20 (1987) 389 J. A. Dumesic, S. A. Stevenson, J. J. Chludzinski, R. D. Shenvood, R. T. K. Baker, ACS Symp. Series 298 (1980) 99 C. S. KO and R. J. Gorte, 155 (1985) 296 K. Tamaru, U. Bardi and Y. Nihei, Surf. Sci., 197 (1988) L. 281 A. K. Singh, N. K. Pande and A. T. Bell, J. Catal., 94 (1985) 422
150
36
37 38 39 40 41
42
43 44 45 46 47
M. A. Vannice and R. L. Garten, J. Catal., 56 (1979) 236 M. A. Vannice and J. Vasco-Jara, in “Metal-Support and Metal-Additive Effets in Catalysis”, B. Imelik et al. ed., Elsevier, Amsterdam, 1982, p. 185 J. D. Bracey and R. Burch, J. Catal., 86 (1984) 384 R. Burch and A. R. Flambard, J. Catal., 79 (1982) 382 E. Santacessaria, S. Carra and I. Adami, Ind. Eng. Chem. Res. Dev., 16 (1977) J. Santos, J. Phillips and J, A. Dumesic, J. Catal., 81 (1983) 147 P. A. Sermon, J. Chem. SOC.Faraday Trans I, 83 (1987) 2243
J. R. Anderson, in “Structure of Metallic Catalysts” Acad. Press, N. Y., 1975; A. A. Castro, 0. A. Scelza, E. R. Benvenuto, G. T. Baronetti, S. R. Miguel and J. H. Parrera, in “Preparation of Catalysis”, G. Poncelet, P. Grange, P. A. Jacobs, eds., Elsevier, Amsterdam, 1983, Vol. II1,47 M. Ichikawa, Shokubai 21 (1979) 253 M. Ichikawa and K. Shikakura “New Horizons in Catalysis”, Proc. 7th ICC, Tokyo, 1980, T. Seiyama and K. Tanabe, eds., Elsevier, 2 (1980) 925 M. Ichikawa, K. Shikakura and M. Kawai, “Heterogeneous Catalysts in Energy Problems ”Proc. Symp. Dalian/China (1982) (A-08- 1) H. Arakawa, K. Takeuchi, T. Matsuzaki and I. Sugi, Chem. Lett. (1984) 1607 M. Kawai, M. Uda and M. Ichikawa, J. Phys. Chem., 89 (1985) 1654 A. G. T. M. Bastein, L. E. Y. Nonneman, unpublished data, Leiden University K. Takeuchi, J. R. Katzer, J. Phys. Chem., 85 (1981) 937 W. A. A. van Barneveld and V. Ponec, J. Catal. 51 (1978) 426 M. Araki and V. Ponec, J. Catal., 44 (1979) 439
48
E. K. Poels, P. J. Mangnus, J. v. Welzen and V. Ponec, Proc. 8th ICC, Berlin, 1984, Dechema, Frankfurt (1984), Vol. 11, p. 59
49 50
F. Fajula, R. C. Anthony and J. H. Lunsford, J. Catal., 73 (1982) 237 J. R. Katzer, A. W. Sleight, P. Gajarde, J. B. Michel, E. F. Gleason and S. McMillan, Faraday Disc. Chem. SOC.,72 (1981) 121 R. F. Hicks and A. T. Bell, J. Catal., 90 (1984) 205 M. A. Vannice, C. Sudhakar and M. Freeman, J. Catal., 108 (1987) 97 G. v. d. Lee and V. Ponec, Catal. Rev. Sci. Eng., 29 (1987) 183 P. Meriaudeau, M. Dufaux and C. Naccache, Proc. 8th ICC, Berlin, 1984, Dechema,
51 52 53 54
Frankfurt (1984), Vol. 11, p. 185 55 56 57 58 59
N. P. Nguyen, P. Meriaudeau and C. Naccache, Appl. Catal., 21 (1986) 337
60
J. M. Driessen, E. K. Poels, J. P. Hindermann and V. Ponec, J. Catal., 82 (1983) 26
M. Ichikawa, Chem. Tech., 6 (1982) 674 A. T. Bell, Catal. Rev. Sci. Eng., 23 (1981) 203 Yong-Bo Zhao and Yip Wah Chung, J. Catal., 106 (1987) 369 Y. Kikuzono, S. Kagami, S. Naito, T. Onishi and K. Tamaru, Faraday Disc. Chem. SOC., 72 (1981) 135
151
61 62 63 64 65 66 67
68
69 70 71 72 73 74 75a 75b 7sc 76 77 78 79 80 81
N. A. Anikin, A. A. Bagatur’yants, G. M. Zhidomirov and V. B. Kazanksi, Russ. J. Phys. Chem., 57 (1983) 393 G. Pacchioni, P. Fantucci, J. Koutecky and V. Ponec, J. Catal., 112 (1988) 34 J. P. Hindermann, A. A. Kiennemann, A. Chakor-Alami and R. Kieffer, Proc. 8th I c c , Berlin 1984, Dechema, Frankurt (1984), Vol. 11, p. 163 I J. W. Jenkins, Preprints of the 6th Canadian Symp. -Catal., Ottawa, 1979 K. Gilhooley, S. D. Jackson and S. Rigby, J. Chem. SOC. Faraday Trans I, 82 (1986) 431 C. T. Campbell and K. D a n k , J. Catal., 106 (1987) 301 P. Ien Chu, B. C. Gerstein, G. R. Sheffer and T. S. King, J. Catal., 115 (1989) 194 E. K. Poels, Dissertation Thesis, 1984, Leiden University; E. K. Poels, R. Koolstra, J. W. Geus and V. Ponec in “Metal- Support and Metal-Additive Effects in Catalysis”, B. Imelik et al. eds., Elsevier, Amsterdam, 1982, 233 Zhenh Zhong Lia, T. Okukara and M. Misono, J. Phys. Chem., 92 (1988) 723 A. Bossi, F. Garbassi and G. Pertini, Reprints 7th ICC, Tokyo, 1980, section E4 F. Solymosi and A. Erdohelyi, J. Mol. Catal., 8 (1980) 471 S. Naito, S. Kagami, H. Yoskioka, Y. Kobori, T. Onishi and K. Tamaru in “Heterogeneous Catalysts related to Energy Problems”, Proc. Symp. Dalian, China, 1982, (A-16-1) C. S. Keller and A. T. Bell, J. Catal., 71 (1981) 288 Y. Iwasawa, T. Hayasaka and S. Ogasawara, Chem. Lett. (1982) 131 J. Pritchard, Nature, 205 (1965) 1208 T. Kwan, Adv. Catal., 6 (1954) 67 J. Texter, D. H. Strome, R. G. Herman and K. Klier, J. Phys. Chem. 81 (1977) 333 R. G. Herman, K. Klier, G. W. Simmons, B. P. Finn, J. B. Bulko and T. P. Kobylinski, J. Catal., 56 (1979) 407 (see also ref. 66) K. Klier, V. Chatikavanu, R. G. Herman and G. W. Simmons, J. Catal., 74 (1982) 343 G. C. Chinchen, M. S. Spencer, K. C. Waugh and D. A. Whan, J. Chem. SOC.Faraday Trans I, 83 (1987) 2193 H. Yosioko, S. Naito and K. Tamaru, Chem. Lett, (1983) 981 G. Apai, J. R. Monnier and D. R. Preuss, J. Catal., 98 (1986) 563 M. Ichikawa, J. Chem. SOC.Chem. Commun. (1978) 566 M. Ichikawa, Bull. Chem. SOC. Japan, 51 (1978) 2268,2273 W. M. Lin, S. F. Gan, C. R. Huang, D. R. Pan and H. Noller, React. Kin. Catal. Lett., 37 (1988) 487 W. X. Pan, R. Cao and G. L. Griffin, J. Catal., 114 (1988) 447 A. A. Kiennemann, R. Breault, J. P. Hindermann and M. Laurin, J. Chem. SOC.Faraday Symp. (Bath), 21 (1988) 14 P. Coutry, D. Durnad, E. Freund, A. Sugior, J. Mol. Catal., 17 (1982) 241 S. Uchiyama, Y. Obayashi, M. Shibato, T. Uchiyama, N. Kawato and T. Konishi, J. Chem. SOC.Chem. Commun., (1985) 1671 M. M. Bhasin, W. J. Bartley, P. C. Ellgen and T. P. Wilson, J. Catal., 54 (1978) 120 S. Naito, S. Kagami, H. Yoshioka, Y. Kobori, T. Onishi and K. Tamam, in “Heterogeneous Catalysts related to Energy Problems”, Proc. symp. Dalian, China, 1982 (A-16.1) M.
152
82 83 84
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100
101 102 103
Inoue, T. Miyako, S. Yonezawa, D. Medhanavya, Y. Takegami and T. Inao, J. Mol. Catal., 45 (1988) 111 T. P. Wilson, W. J. Bartley and P. C. Ellgen in “Heterogeneous Catalysts related to Energy Problems”, Proc. symp. Dalian, China, 1982 (A-27-U) M. Arakawa, T. Hanaoka, K. Takeuchi, T. Matsuzaki and Y. Sugi, Roc. 9th ICC, Calgary, Canada, Chem. Inst. Canada (1988) Vol. 1, p. 602 H. Preliaud, B. Tardy, J. C. Bertolini and G. A. Martin, Proc. Europ. Conf. on Structure and Reactivity of Surfaces, Trieste, 1988, Elsevier, in print M. S. Spencer, J. Phys. Chem., 88 (1984) 1046 T. L. F. Favre, G. v. d. Lee and V. Ponec, J. Chem. SOC.,(1985) 230 G. N. Ilmich, V. A. Likhobolov, Y. I. Yermakov, Kinet. Kataliz., 26 (1985) 763 J. S. Rieck and A. T. Bell, J. Catal., 99 (1986) 278 A. G. T. M. Bastein and A. T. Bell, unpublished results S. Matsushita, K. Tanaka and T. Saito, J. Chem. Soc., (1985) 1209 S. D. Jackson, B. J. Brandreth and D. Winstanley, J. Catal., 106 (1987) 464 H. Orita, S. Naito and K. Tamaru, J. Catal., 90 (1984) 183 Xiao Ding Xu, E. B. M. Doesburg and J. J. F. Scholten, Catalysis Today, 2 (1987) 125 D. N. Dwyer, A. J. Hardenbergh. Appl. Surf. Sci., 19 (1984) 14 H. Orita, S. Naito and K. Tamaru, J. Phys. Chem., 89 (1985) 3066 G. B. Raupp and J. A. Dumesic, J. Phys. Chem., 90 (1986) 3359 H. Preliaud, J. A. Dalmon, C. Mirodatos and G. A. Martin, J. Catal., 97 (1986) 344 H. Preliaud, M. Primet and G. A. Martin, Appl. Surf. Sci., 17 (1983) 167 R. P. Underwood and A. T. Bell, J. Catal., 109 (1988) 61; 111 (1986) 325 see the IR spectra in ref. 58 P. A. J. M. Angevaare, H. A. C. M. Hendrickx and V. Ponec, J. Catal., 110 (1988) 11 P. A. J. M. Angevaare and V. Ponec, J. Catal., 110 (1988) 18 M. Ichikawa and T. Fukushima, J. Phys. Chem., 89 (1985) 1564 V. Ponec in “Catalysis”, G. C. Webb, G. C. Bond, eds., Specialist Periodic Reports, Chem. SOC.London, 1982, Vol. 5 , p. 48 Shen-ichi Ichi, Y. Ohno and B. Viswanathan, Surf. Sci., 161 (1985) 349 V. Ponec in NATO-AS1 Series B, Proc. NATO Workshop, Nieuwpoort, Belgium, 1988, E. G. Derouane, ed., Plenum Press, in print J. S. Somers, Th. Lindmer, M. Surman, A. M. Bradshaw, G. P. Williams, C. F. McConville and D. P. Woodruff, Surf. Sci. 183 (1987) 576 C. L. Allyn, T. Gustafsson and E. W. Plummer, Chem. Phys. Lett., 47 (1977) 127 D. Heskett, E. W. Plummer, R. A. de Paole and W. Eberhardt, Phys. Rev. B, 33 (1986) 5171 R. J. Smith, J. Anderson and G. J. Lapeyre, Phys. Rev. B, 15 (1980) 632 W. Riedel and D. Menzel, Surf. Sci., 163 (1985) 39 G. Apai, P. S. Williams, J. Stohr and D. A. Shirley, Phys. Rev. Lett., 37 (1976) 1497 G. Paolucci, M. Surman, K. C. Prince, I. L. Scorba, A. M. Bradshaw, C. F. McConville, D. P. Woodruff, Phys. Rev., 34 (1986) 1340
153
104 J. N. Allison and W. A. Goddard 111, Surf. Sci., 111 (1981) L615; 115 (1982) 553 105 106 107 108 109 110 111 112 113 114 115 116 117 118
119 120
N. D. Shinn and T. E. Madey, J. Chem. Phys., 83 (1985) 5928 N. D. Shim and T. E. M adey, Surf. Sci., 180 (1987) 615 S. D. Cameron and D. J. Dwyer, Langmuir 4 (1988) 282 F. J. C. M. Toolenaar, Dissertation thesis, Leiden University, 1983 R. M. Hammaker, S. A. Francis and R. P. Eischens, Spectrochim. Acta, 21 (1965) 1295 A. Crossley and D. King, Surf. Sci., 19 (1977) 528; 95 (1980) 131 D. Reinalda and V. Ponec, Surf. Sci., 91 (1979) 113 F. M. Hoffmann and A. M. Bradshaw, Surf. Sci., 72 (1979) 513 A. Ortega, F. M. Hoffmann and A. M. Bradshaw, Surf. Sci., 119 (1982) 79 A. M. Bradshaw and F. M. Hoffmann, Surf. Sci., 52 (1975) 449 J. Koutecky, Z. Elektrochem. (Bunsen Ges.) 60 (1956) 835; Trans Faraday SOC.,54 (1958) 1038 T. B. Grimley, Proc. Phys. SOC.,90 (1967) 751: 92 (1967) 776 T. L. Einstein and J. R. Schrieffer, Phys. Rev. B, 36 (1973) 3629 J. R. Schrieffer in “Dynamic Aspects of Surface Physics”, Proc. E. Fermi School on Physics, ed. Compositori, Bologna/Italy, 1974, p. 2540 G. Blyholder, J. Phys. Chem., 68 (1964) 2772; 79 (1976) 756 D. A. King, J. Elect. Spectrosc., 29 (1983) 11 H. A. C. M. Hendrickx, C. des Bouvrie and V. Ponec, J. Catal., 109 (1988) 120 R. Queau and R. Poiblanc, J. Catal., 27 (1972) 200 M. Primet, J. M. Basset, M. V. Mathieu and M. Prettre, J. Catal., 29 (1973) 213 M. Primet and M. V. Mathieu, J. Chim. Phys., 72 (1975) 659 A. Palazov, J. Catal., 30 (1973) 13 J. M. Basset, G. Dalmai-Imelik, M. Primet and G. A. Martin, J. Catal., 37 (1975) 22 R. W. Ittner and H. Luth, in “Proc. of 7th ICC, Tokyo, 1980, T. Seijama and K. Tanabe, eds., Kodansha, Tokyo, 1981, Vol. 1, p. 95 M. Primet, J. Catal., 88 (1984) 273 H. Ibach and G. A. Somorjai, Appl. Surf. Sci., 3 (1979) 293
121 122
W. L. van Dijk, J. A. Groenewegen and V. Ponec, J. Catal., 45 (1976) 277 V. Ponec, Adv. Catal., 32 (1983) 149 123 Y. Soma-Noto and W. M. H. Sachtler, 32 (1974) 316; 34 (1974) 162 124 J. A. Dalmon, M. Primet, G. A. Martin and B. Imelik, Surf. Sci., 50 (1975) 95 M. Primet, M. V. Mathieu and W. M. H. Sachtler, J. Catal., 44 (1976) 324
125
126 127
F. J. Kuijers, V. Ponec, J. Catal., 60 (1979) 100 V. Ponec, Mat. Sci. Eng., 42 (1980) 135 R.
A. v. Santen and W. M. H. Sachtler, J. Catal., 33 (1974) 202; 34 (1974) 13 F. C. Williams and D. Nason, Surf. Sci., 45 (1974) 377 V. S. Sundaram and P. Wynblatt, Surf. Sci., 52 (1975) 569 R. G. Donelly and T. S. King, Surf. Sci., 74 (1978) 89 P. Wynblatt and R. C. Ku, Surf. Sci., 65 (1977) 511 F. F. Abraham, N. H. Tsau and G. M. Pound, Surf. Sci., 83 (1979) 406 M. Kelley and V. Ponec, in Progress Surf. Sci., 11 (1981) 139, S. G. Davison, ed., Pergamon Press, Oxford, A. D. van Langeveld, J. W. Niemantsverdriet, J. Vac. Sci. Technol., AS(4) (1987) 558
154
128 129
L. de Temmeman, C. Cremers, H. van Hove and A. Neyens, Surf. Sci., 183 (1987) 565 F. Stoop, F. J. C. M. Toolenaar and V. Ponec, J. Catal., 73 (1982) 50 F. J. C. M. Toolenaar, F. Stoop and V. Ponec, J. Catal., 82 (1983) 1
130 131 132
H. A. C. M. Hendrikcx and V. Ponec, Surf. Sci., 192 (1987) 234 W. M. H. Sachtler and R. A. van Santen, Adv. Catal., 26 (1977) 69 Y. T. Shah and A. H. Perotta, Ind. Eng. Chem. Prod. Res. Dev., 15 (1976) 123, see table 11 on p. 127 W. A. A. van Barneveld and V. Ponec, J. Catal., 89 (1984) 542 S. Schanke, G. Sundmark and A. Holmen, Proc. 9th ICC, Calgary, 1988, Chem. Inst. Canada (1988), Vol. 2, p. 650 A. Matani and D. K. Chakrabarty, ibid, p. 682 V. Gryaznov, Juri Seroo and S. Gulyanova, ibid, p. 720 J. A. Amelse, L. H. Schwartz and J. B. Butt, J. Catal., 72 (1981) 95 T. Nakajo, H. Arakawa, K. Sano and S. Matsuhira, Chem. Lett., 4 (1987) 593 T. Nakajo, K. Sano, S. Matsuhira and H. Arakawa, J. Chem. Soc. Chem. Commun., (1987) 647 P. W. Wentrcek, J. G . McCarthy, C. M. A. Blow and H. Wise, J. Catal., 61 (1980) 232 Z. Schay and L. Guczi, React. Kinet. Catal. Lett., 14 (1980) 207 L. Guczi, G. Kemeny, K. Matusek and J. Mink, J. Chem. Soc. Faraday Trans I, 76 (1980) 782 L. Guczi, Proc. 9th Intem. Congress on Catalysis, Eds. Ternan and Phillips, Chem. Inst. Canada, 1988 Vol 5, p 114 G. L. Ott, T. Fleisch and W. N. Delgas, J. Catal., 60 (1979) 394 D. J. Elliot and J. H. Lunsford, J. Catal., 57 (1979) 11 J. W. Niemantsverdriet, S. P. A. Louwers, J. van Grondell, A. M. van der Kraan, F. W. H. Kampers and D. C. Koningsberger, Roc. 9th ICC, Calgary 1988, Chem. Inst. Canada (1988) Vol. 2, p. 674 H. Kolbel and K. D. Tilmets, Belg. Patent 837 628 (1976) B. Bussemeier, C. D. Frohning, G. Horn snd W. Kluy, Ger. Offen. 2 518 964 (1976) H. Kolbel and M. Ralek, Catal. Rev. Sci. Eng., 21 (1980) 225 J. M. Martin-Martinez, F. Rodrigues-Reinoso and M. A. Vannice, Proc. 9th ICC, Calgary, 1988, Chem. Inst. Canada, Vol. 2, p. 658 I. R. Leith, ibid, p. 658 J. J. Venker, M. Kaminsky, C. L. Geoffroy and M. A. Vannice, J. Catal., 103 (1987) 450 J. Barrault and C. Renard, N o w . J. Chim., 7 (1983) 149 T. Kuroda, M. Yoshikawa, A. Miyamoto and T. Incei, Chem. Express, l(2) (1986) 121 W. L. van Dijk, Dissertation thesis, Eindhoven Technical University, 1981 A. L. Dent and M. Lin, Adv. Chem. Ser. Am. Chem. Soc., 178 (1979) 51 L. I. Kuznetsova, A. R. Suzdorf, I. V. Usmanova, G. A. Moiseeva and V. G. Chumakov, Kinet. Katal., 29 (1988) 491 E. Kieffer, Dissertation thesis, Eindhoven Technical University, 1981 P. C. Ellgen, W. J. Bartley, M. M. Bhasin and F. P. Wilson, Adv. Chem. Ser. Am. Chem. Soc.,178 (1979) 147 L. N. Mulay, C. Lo, T. Pannaparayil and R. A. Diffenbach, J. Magn. Mater., 54 (1986) 54 V. T. Coon, T. Takeshita, W. E. Wallace and R. S. Craig, J. Phys. Chem., 80 (1976) 1878 H. Imamura and W. E. Wallace, J. Catal., 65 (1980) 127 W. E. Wallace, J. France and A. Shamsi, in “Rare Earths in Modern Science and Technology”, Plenum Press, N. Y., 1982,
133 134
135
136
137
138
139 140 141
155
142
143 144 145 146 147 148
149 150 151 152
153 154 155 156 157 158
159
Vol. 3, J. Barrault, A. Guillemont, J. C. Achard, V. Paul-Boncour, A. Percheron-Guegan, L. Hillaire and M. Coulon, Appl. Catal., 22 (1986) 273 C. A. Luengo, A. L. Cabrera, H. B. McKay and M. B. Maple, J. Catal., 47, (1977) 1 F. P. Daly, J. Catal., 89 (1984) 131 J. R. Jennings, R. M. Lambert, R. M. Nix and G. Over, Appl. Catal., in print R. M. Nix, T. Rayment, R. M. Lambert, J. R. Jennings and G. Over, J. Catal., 106 (1987) 216 M. Shibata, Y. Ohbayastri and N. Kawata, J. Catal., 96 (1985) 296 G. B. Atkinson, E. G. Baglin, L. J. Nicks and D. J. Bauer in “Catal. Conversions of Syngas and Alcohols to Chemicals”, R. G. Herrnann, ed., Plenum, N. Y., 1984 p. 65 W. Schottky, Z. Physik., 14 (1923) 63 W. Schottky, Physik. Zeitschr., 15 (1914) 872
J. Bardeen, Phys. Rev., 58 (1940) 727 J. Heinrichs, Phys. Rev. B, 4 (1973) 1346 A. Liebsch, Physics Scripta, 35 (1987) 354 s. P. Ape11 and D. R. Penn, Phys. Rev. B, 34 (1988) 6612 N. D. Lang, S. Holloway, J. K. Norskov, Surf. Sci., 150 (1985) 24 K. W. Wojciechowski, Surf. Sci., 55 (1976) 246 S. Holloway, J. K. Norskov and N. D. Lang, J. Chem. SOC.Faraday Trans I, 83 (1987) 1935 K. J. Uram, L. Ng and J. T. Yates Jr., Surf. Sci., 177 (1986) 253 L. Ng, K. J. Urarn, Z. Xu, P. L. Jones and J. T. Yates Jr., J. Chem. Phys., 86 (1987) 6523 D. Tomanek, M. Kreuzer and J. H. Block, Surf. Sci., 157 (1985) L315 M. Laren, J. P. Pendry, R. W. Joyner, Surf. Sci., 165 (1986) L86 P. J. Feibelman and D. R. Hamann, Surf. Sci., 149 (1985) 48 J. M. MacLaren, D. D. Vvedensky, J. B. Pendry and R. W. Joyner, J. Chem. SOC.Faraday Trans I, 83 (1987) 1945 Let us consider two parallel plates (metal and support, as an example) on a distance of 5 A. If the contact potential difference, V, is 1 Volt, the corresponding surface charge on the plate surface is 6=&d-l. Then, with e=8.85 10-12 C2 N-l m-2 as for vacuum 6 is 1.8 C/m2. With 10’5 atoms/cm2 the number of electrons creating a potential difference of 1 V. across the interphase is 0.01 el/atom. J. Disdier, J. M. Henmann and P.PIChat, J. Chem. SOC.Faraday Trans I, 79 (1983) 651 B. Johansson and N. Martensson, Phys. Rev. B, 21 (1980) 4427 H. P. Bonze], Surf. sci. Reports, 8 (1987) 43 R. Hauffe, Adv. Catal., 7 (1955) 213 Th. Wolkenstein, Adv. Catal., 12 (1960) 189 G. M. Schwab, J. Block, D. Schultze, Angew. Chernie, 71 (1958) 101 Z. G. Szabo and F. Solymosi, “Actes du 1lxieme Congr. Catal. ”, Paris, 1960 Ed. Technip, Paris (1961), Vol. 2, p. 1627 Z. G. Szabo, F. Solymosi and I. Batta, Z. Phys. Chem., NF, A, 125 (1958); 23 (1960) 56 F. Solymosi, Catal. Rev., 1 (1967) 232 R. A. Dalla Betta and M. Boudart in, Proc. 5th ICC, Miami Beach, 1972, Elsevier, Vol. 2 (1973) 1239 G. C. Bond and P. Sermon, Gold Bull., 6(4) (1974) 105
156
160 161 162 163
E. C. Akubuiro and X. E. Verykios, J. Catal., 103 (1987) 320; 113 (1988) 196 J. M. Henman, J. Catal., 89 (1984) 404 A. J. Dekker, Solid State Physics, McMillan, London 1958 J. R. Smith, F. J. Aelinshaus and J. G. Gay, Phys. Rev. B, 26 (1982) 1071 S. Holloway and J. K. Norskov, J. Electr. Anal. Chem., 161 (1984) 193 R. Kotz and E. Yager, J. Electr. Anal. Chem., 123 (1981) 335 K. Kunimatsu, J. Phys. Chem., 88 (1984) 2195
164 165
166 167 168 169 170
D. K. Lambert, Solid State Commun., 51 (1984) 297 J. Pancir and R. Zahradnik, Helv. Chim. Acta, 61 (1978) 59 H. Nakatsuji, T. Hayakawa and T. Yonezawa, J. Am. Chem. Soc., 103 (1981) 7426 L. Tun-Nay, I. Tvaroska and D. Tunega, COILCzechoslov. Chem. Commun., 51 (1986) 1803 V. Bonacic-Koutecky, J. Koutecky, P. Fantucci and V. Ponec, J. Catal., 111 (1988) 409 F. Umbach, Appl. Phys., A47 (1988) 25 W. Wurth, J. J. Weimer, E. Hudeczek and F. Umbach, Surf. Sci., 173 (1986) L619 E. Wimmer, A. J. Freeman, J. R. Hiskes and A. M. Karo, Phys. Rev. B, 28 (1983) 3074 A. B. Anderson and D. Q. Dowd, J. Phys. Chem., 91 (1987) 869 W. M. H. Sachtler, D. F. Shriver, W. B. Hollenberg and A. F. Lang, J. Catal., 92 (1985) 429 D. F. Shriver, Amer. Chem. SOC.Symp. Series, 152 (1981) 1 C. P. Horwitz and D. F. Shriver, Adv. Organomet. Chem., 23 (1984) 219 F. Correa, R. Nakawura, R. E. Stinson, R.
J . Bunvell and D. F. Shriver, J. Am. Chem. SOC.,102 (1980) 5112 J. P. Collman, R. G. Finke, J. N. Cawse and J. I. Brauman, J. Am. Chem. SOC.,100 (1978) 4766 171 172 173 174 175 176
177
178
D. Tomanek and K. K. Bennemann, Surf. Sci., 127 (1983) L261 D. R. Jennison, J. Vac. Sci. Technol., A5 (1987) 684 P. A. Schultz, C. H. Patterson and R. P. Messner, J. Vac. Sci. Technol., A5 (1987) 1061 D. Lackey, M. Surman, S. Jacob, D. Grider and D. A. King, Surf. Sci., 152/153 (1985) 513 D. Lackey and D. A. King, J. Chem. SOC.Faraday Trans I, 83 (1987) 2001 D. M. Kolb, J. Vac. Sci. Technol., A4(3) (1986) 1294 K. M. Ho, C. L. Fu, S. H. Liu, D. M. Kolb and G. Piazza, J. Electro. Anal. Chem., 150 (1983) 235 J. E. Houston, C. H. F. Peden, P. J. Feibelman and D. R. Hamann, Phys. Rev. Lett., 56 (1986) 375 E. K. Poels, V. Ponec, in “Catalysis” Specialist Repts, Chem Soc. London, G. C. Bond and G. Webb, eds., 1983, Vol. 6, p. 196 M. E. Levin, M. Salmeron, A. T. Bell and G. A. Somorjai, J. Catal., 106 (1987) 401 G. A. Vedage, R. Pitchai, R. G. Herman andK. Klier, Proc. 8th ICC, Berlin, 1984, Dechema, Frakfurt (1984), Vol. 11, p. 47 (and refs. therein) A. Miyamoto, M. Miura, K. Sakamoto, S. Kamitomai, Y. Kosakiand M. Murakami Ind. Eng. Chem. Prod, Res. Dev., 23(1984) 467 C. K. Rofer.De Poorter in “al. Conversions of Syngas and Alcohols to Chemicals”, R. G. Herman, ed., Plenum Press, N. Y. (1984) p. 97 D. K. Lambert, J. Chem. Phys., 89 (1988) 3847 W. Muller, P. S. Bagus, J. Electr. Spectr. Rel. Phenomena, 38 (1986) 103
157
179
J. L. Gland, R. J. Madey, R. W. McCabe and C. de Maggio, Surf. Sci., 143 (1984) 46 T. E. Madey and C. Benndorf, Surf. Sci., 164 (1985) 602 D. W. Moon, D. J. Dwyer and S. L. Bernasel, Surf. Sci., 163 (1985) 215 J. Lee, C. P. Hanrahan, J. Arias, R. M. Martin and H. Metia, Surf. Sci., 16 (1985) L543 H. P. Bonzel, J. Vac. Techno]., A2 (1984) 866
180
M. P. Kiskinova, Surf. Sci. Rept., 8 (1988) 359 D. W. Goodman, Acc. Chem. Res., 17 (1984) 194; Appl. Surf. Sci., 19 (1984) 1 M. Kiskinova and D. W. Goodman, J. Catal., 105
181 182
183
184 185 186
187 188
189
(1981) L265; 108 (1981) 64; 109 (1981) L555 E. Schustorowich, Surf. Sci., 175 (1986) 561; ibid, Surf. Sci. Rept., 6 (1986) 1 J. P. S. Badyal, A. J. Gellman and R. M. Lambert, J. Catal., 111 (1988) 383 T. Mori, A. Miyamoto, N. Takahushi, M. Fukagaya, T. Hattori and Y. Murakami, J. Phys. Chem., 90 (1986) 5197 C. H. Bartholomew and C. K. Vance, J. Catal., 91 (1985) 78 S. D. Cameron and D. J. Dwyer, Surf. Sci., 198 (1988) 315 M. Ichikawa, T. Fukushina, K. Shikakura, Proc. 8th ICC, Berlin 1984, Dechema, Frankfurt, 1984), Vol. 2, p. 69 C. T. Campbell and D. W. Goodman, Surf. Sci., 123 (1985) 413 R. P. Underwood and A. T. Bell, J. Catal., 11 (1988) 325; 109 (1988) 61 T. Matsushima, Zeit. Physik. Chem., 158 (1988) 175 M. E. Dry, T. Shingles and L. J. Boshoff, J. Catal., 2.5 (1972) 99 T. Mori. A. Miyamoto, N. Takahashi, H. Niizuma, T. Hattori and Y. Murakami, J. Catal., 102 (1986) 199 T. Mori, H. Masuda, H. Imai, A. Miyamoto, H. Niizuma, T. Hattori and Y . Murakami, J. Mol. Catal., 25 (1984) 263 S. C. Chuang, J. G. Goodwin and I. Wender, J. Catal., 95 (1985) 435 M. McLaughlin McGlory and R. D. Gonzalez, J. Catal., 89 (1984) 392 D. C. Grenoble, M. M. Estadt and D. F. Ollis, J. Catal., 67 (1981) 90 P. Mars in “The Mechanism of Heterogeneous Catalysis”, J. H. de Boer, ed., Elsevier, 1960, p. 49 (see the table on p. 60) B. W. Krupay, Y. Amenomiya, J. Catal., 67 (1981) 362 I. L. C. Freriks, P. C. de Jong-Versloot, A. G. T. G. Kortbeek and J. P. van den Berg, J. Chem. SOC. Chem. Commun., (1986) 253; see the curve B in Fig. 4.2 Y . Amenomiya and G. Plezier, J. Catal., 76 (1982) 345 Ch. K. Rofer-De Poorter in “Catalytic Conversions of Syngas and Alcohols to Chemicals”, R. G. Herman, ed., Plenum, N. Y., 1984, p. 97 Z. Poltarzewski, S. Galvagno, R. Pietropaolo and P. Staiti, J. Catal., 102 (1986) 217 s. Galvagno, Z. Poltarzewski, A. Donato, G. Nori and P. Pietropaolo, J. Mol. Catal., 35 (1986) 365 A. A. Wismeijer, A. P. G. Kieboom and H. van Bekkum, React. Kinet. Catal. Lett., 29 (1985) 311 P. N. Rylander in “Catalytic. Hydrogenation over Pt-Metals”, Acad. Press, N. Y., 1967 R. Hubaut, M. Daage and J. P. Bnelle, Appl. Catal., 22 (1986) 231 W. E. Pascol and J. F. Stenberg in 7th Conf. Catal. in Org. Synth., Chicago, W. H. Jones, ed., Acad. Press, N. Y . , 1978, D. K. Sokolskii, A. K. Zharmagambetova, N. V. Anisilmoki and A. Ualikhanova, Dokl. Akad. Nauk. SSSR, Ser. Kchim., 273 (1983) 151 M. A. Vannice and B. Sen, J. Catal., 115 (1989) 65 B. Sen and M. A. Vannice, J. Catal., 113 (1988) 52
158
CHAPTER 5
RECENT DEVELOPMENTS IN FISCHER-TROPSCH CATALYSIS
Calvin H. Bartholomew Department of Chemical Engineering, Brigham Young University, Provo, Utah 84602 (USA)
159
5.1 INTRODUCTION Fischer-Tropsch Synthesis (FTS), the production of liquid hydrocarbons from synthesis gas (CO, COz and Hz) is a promising, developing option for environmentally sound production of chemicals and fuels from coal. In view of large coal reserves and dwindling petroleum reserves it is projected to play an ever increasing role in coming decades. Catalysts for FI'S were first developed in the early 1900's. Following the discovery by Sabatier and Senderens in 1902 that CO could be hydrogenated over Co, Fe, and Ni to methane (ref. I), BASF reported the production of liquids over cobalt catalysts in 1913 (ref. 2) while Fischer and Tropsch reported production of hydrocarbons over alkalized iron in 1923 (ref. 3). Much of the early catalyst development took place in Germany during the 1930s and 40's; after World War 11, developments in Germany by Ruhrchemie and Lurgi and in the U.S. by Kellog Co. led to the commissioning of the Sasol Plant in South Africa in 1955 and which following plant expansions in the 1980s remains the only large scale commercial operation. Renewed activity in the development of FTS catalysts was stimulated in the 70's and 80's by perceived shortages of liquid petroleum. Many of the early developments of FI catalysts and processes have been reviewed by Dry (ref. 4) and Anderson (ref. 5); the details of these developments can be found in early reviews referenced in Chapter 1 of Anderson's book on the FTS (ref. 5). These early developments will not be treated here; rather this chapter will focus on catalyst and process developments which have occurred during the 70's and 80s with emphasis on examples of recent catalyst developments during the late 1980's. The discussion of catalyst developments will focus on three important areas of FT catalysis: chemical modifications (additives, promoters, supports, pretreatments, and preparation methods), interception of intemiediates (dual functionalisni and secondary reactions), and limitation of chain growth by shape selectivity. Fundamental principles of catalyst design will also be emphasized. Since other chapters in this book treat support and promoter effects (Chapter 4), bimetallics (Chapter 6), alcohol synthesis (Chapter 7), zeolites, and processing for producing chemicals (Chapter@, coverage of these topics will be brief. Several previous reviews and books have provided perspectives on catalyst and process technologies for the Fischer-Tropsch Synthesis (refs. 6-15). The most recent of these by Mills (ref. 15) provides an informative overview of present catalyst/process technology and future directions for indirect coal liquefaction; Mill's review (ref. 15) also treats the production of oxygenate fuels, the methanol to gasoline process (ref. 16), and the high temperature isosynthesis (ref. 17), subjects that are not treated in this chapter. In preparation for writing this chapter the author conducted several comprehensive searches of the FTS literature covering the past decade and discovered over 100-150 new journal articles and over 150 new patents to add to an already voluminous collection of several hundred articles. In view of the vast literature of FTS, it is not practical to review all of these new developments i n the limited space here; rather this chapter will feature some of the "most significant" developments using selected examples from journal and patent literature to illustrate principles. Present FT catalyst/process technology suffers from the following limitations: (1) limited selectivity for premium products (e.g. light olefins, gasoline, or diesel fuel), (2) catalyst
160
07
deactivation, (3) high capital cost (ref. lS), (4) heat removal, and ( 5 ) less than optimum
-
thermal efficiency (ref. IS). Selectivity limitations
06 -
are inherent in the chain growth mechanism for FT synthesis which
DEGREE OF POLYMERIZATION Fig. 5.1 Selectivity limitations on Fischer-Tropsch synthesis as determined by the Anderson-Schulz-Flory distribution function
is
governed
by
Anderson-Schulz-Flory
(ASF)
kinetics
While
(refs. 5,7).
product molecular weight can be varied by choosing process conditions and/or catalyst to achieve a given degree of
polymerization, a wide distribution of products other than methane is inherent, as shown in Fig. 5.1. For example, the maximum obtainable weight percentage of light LPG hydrocarbons (C2-C4) is 56%, of gasoline
(C5-CI1) is 47%, and of diesel fuel (C12-C17) is 40%. Attempts to circumvent the selectivity limitations of ASF kinetics have generally met with failure (ref. 7). Nevertheless, there are new approaches (ref. 7) involving (1) shape selective supports, (2) unsteady-state operation, and (3) interception of intermediates which show promise for improving selectivity beyond that predicted by ASF theory. Moreover, the design of catalysts and multistep processes which maximize production of light olefins, gasoline, or diesel fuel within the constraints of ASF kinetics has advanced considerably in the past decade. Significant progress toward the solution of deactivation, heat removal, and thermal efficiency problems has also been realized in the past two decades. With these innovations improvements in process economics of 30-40% are realizable (ref. IS). Of equal or perhaps even greater significance is the progress made during the same period in the understanding of the relationship of catalyst structure to activity and selectivity properties; this understanding provides a scientific basis for catalyst design, the principles of which will be discussed in this chapter.
5.2 NEW CATALYST DEVELOPMENTS 5.2.1 CHEMICAL MODIFICATIONS A brief review of some basic definitions and principles relating to the FT synthesis is appropriate here in providing a foundation for the ensuing discussion. The chemistry of methanation and Fischer-Tropsch (FT) synthesis processes can be described by the following set of reactions:
CO + 3 H,= CH4 + HZO
161
Table 5.1 CO hydrogenation activities and propagation probabilities of representative unpromoted Co, Fe and Ru synthesis catalysts at 480 K (H2/CO=2, latm) Catalyst
15% C0/A1203 Unsupported Fe 11% RdA1203 3% R~/A1203 a
%CO, in Prod b
%Olefin C3-Cf
ad
Ref.
1 31
54 94
0.90 0.44
18 19
88 65
0.69
20
4
0.70
21
Turnover frequency in molecules of CO converted per catalytic site per second. Mole percentage of C 0 2 in product (excluding unconverted reactants). Mole percentage of olefins in C C Propagation probability determ?ied slope of a mol % hydrocarbon versus carbon number plot.
;I%%,
CO + 2 H2 = -CH2- + H2O CO + H20 = CO2 + H2 2 co = c + coz Reaction I is the formation of methane, Reaction I1 the synthesis of hydrocarbons heavier than methane, Reaction 111 the water-gas-shift reaction, and Reaction IV the Boudouard reaction resulting in deposition of carbon. Generally, Ni catalysts are very active for Reaction I relative to Reaction I1 and hence most selective for methane relative to Co, Fe, and Ru catalysts; under typical synthesis conditions (e.g. 453-523 K, H2/CO=2) the latter three catalyst types promote Reaction I1 relative to Reaction I and are more selective for C2+ hydrocarbons. However, the selectivity for C,+ hydrocarbons is strongly influenced by reaction conditions and catalyst composition. For example, C2, hydrocarbon make decreases while methane make increases with increasing H2/CO ratio, increasing reaction temperature, and decreasing pressure. In Fischer-Tropsch synthesis hydrocarbon product selectivities are determined by the ability
of a catalyst to catalyze chain propagation versus chain termination steps. The distribution of hydrocarbon products in Fischer-Tropsch synthesis is generally described by a chain polymerization kinetics model ascribed to Anderson, Schulz, and Flory (refs. 5 7 ) henceforth referred to as the Anderson-Schulz-Flory (ASF) model. The ASF product distribution is mathematical1y represented by the following equation:
where n is the number of carbon atoms in the product, W , is the weight fraction of product containing n carbon atoms, a is the chain growth probability. Generally the value of CI is obtained by a least-squares linear regression of the logarithmic form of Eq. 5.1, the slope and intercept yielding a:
162
Values of a are influenced similar to C2+ hydrocarbon selectivity by reaction conditions and
a increase with decreasing H2/CO ratio, decreasing reaction temperature, and increasing pressure. Values of a are higher for Ru and Co catalysts catalyst composition. For example, values of
relative to Fe catalysts (see Table 5.1). Discussion of the effects of promoters, supports, additives, pretreatments and preparation is only meaningful in the context of baseline catalytic properties in the absence of these effects; i.e. referenced to unpromoted catalysts involving similar preparations and minimal support effects. Table 5.1 compares the activity and selectivity properties of representative, unpromoted Co, Fe, and Ru catalysts at 480 K, 1 atm, H2/CO=2 and low CO conversions (1.10%) such that heat/mass transport processes did not influence the measured rate; these data were obtained for catalysts of high metal loading and low dispersion so that effects of support and dispersion were minimized; both Co and 11% Ru catalysts were prepared by wet aqueous impregnation of the metal salts. The data for the Co and Fe catalysts (refs. 18,19) were obtained after 20-30 hours of reaction and hence represent steady-state activity/selectivity data, while those for Ru (ref. 20) were obtained after only 20 minutes, but the activity and selectivity data are in good agreement with those obtained under similar conditions but after 20 hours of reaction for a 3% Ru/alumina catalyst of comparable dispersion prepared by carbonyl decomposition (ref. 21) The accordance '
O
-
'
r
data with
in
Table 5.1 previous
show
in
experience
(refs. 5,14) that Co is significantly more active than Fe or Ru; Co and Ru catalysts produce heavier products consistent with the larger values of
a of 0.90 and 0.70 relative to that of
0.44 for unsupported, unpromoted iron. In line with its higher activity for the water-gas-shift
WFe Fig. 5.2 Effects of W e ratio on specific activity of iron (from ref. 24)
reaction, iron rejects a significant fraction of the oxygen in the form of C02 rather than H,O, accordingly iron catalysts can be operated at significantly lower Hz/CO ratios, e.g. 0.6-1.0 (ref. 15), compared to values of 2 for Co and Ru catalysts without significant deactivation due to carbon formation due to Reaction IV. However, if syngas is available at a H2/CO=2 ratio there is an advantage to maximizing hydrocarbon rather than CO, selectivity by using Co or Ru catalysts.
163
5.2.1.I 5.2.1.1.1
Additives and Promoters Effects of Promoters
5.2.1.1.1.1
Promotion by Potassium
While potassium promotion of iron catalysts has been known for many decades (refs. 3-51, several recent studies (refs. 22-27) provide new insights into its role in CO hydrogenation. Arakawa and Bell (ref. 24) reported that potassium promotion of Fe/alumina causes an increase in the CO conversion turnover frequency through a maximum followed by a decrease with increasing potassium content (see Fig. 5.2); potassium was also found to increase average molecular weight and olefin-to-paraffin ratio of the hydrocarbon product and to increase water-gas-shift activity. Rankin and Bartholomew (ref. 26) observed a similar increase in the average molecular weight and olefin/paraffin ratio of the product for potassium promoted Fe/silica catalysts precalcined at only 373 K (see Table 5.2). Considered as a whole previous studies (refs. 22-27) suggest that potassium affects the selectivity for CO hydrogenation by changing the kinetics and energetics of the adsorption of the reactants, hydrogen and CO, thereby affecting their relative surface coverages during reaction, i.e., leading to a lower H/C ratio favoring the formation of hydrogen-poor, olefinic products. Several recent catalyst developments (refs. 28-30) involve potassium promoters. McVicker and Vannice (refs. 28,29) prepared well-dispersed highly potassium-promoted Fe, Ru, and Ir catalysts by depositing potassium metal carbonyl clusters onto high surface area alumina and silica followed by thermal decomposition in hydrogen. The resulting catalysts were found to be more active and to exhibit higher selectivities for C2-C3 olefins as well as lower methane selectivities than the corresponding catalysts prepared by conventional means (see Table 5.3). The apparently greater promotional effect for the carbonyl-derived catalysts was hypothesized to be due to a more intimate contact of promoter and metal, since both were originally present together in the parent complex. Stowe and Murchison (ref. 30) have reported the development of a potassium-promoted Ru catalyst for synthesis of naptha from a synthesis gas of H2/CO=2, the syngas being a byproduct of the thermal cracking of petroleum crude or residuum to ethylene; the naptha product can be subsequently cracked to ethylene thus enhancing the economics of the process. The choice of catalyst was based o n the following criteria: (1) would operate at a H2/CO ratio in the range of 2-3, ( 2 ) would convert oxygen from the CO to H 2 0 rather than C02, (3) would maximize C,+ production at a high degree of polymerization while minimizing CH4 and hard wax production. Ru was the logical choice of metal since it operates at the desired H2/C0 ratio, rejects oxygen as water, and operates at a high degree of polymerization; addition of potassium was found to lower CH4 production while maximizing C2+ (see Fig. 5.3); maximum C2+ production and minimum CH, and CO, productions were obtained at a potassium loading of 0.5%. Accordingly, the optimum catalyst was found to be 1% Ru/0.5% Walumina.
164
Table 5.2 Effect of potassium level on selectivity of K-promoted Fe/SiO2 catalysts (H2/CO=2, 1 atm) (from ref. 26)
co2
T
Hydrocarbonb
OlefinC
ad
sela
15% Fe 373
I 498
17.7
I
15.9
33.6
49.4
1.0
67
0.71
a Mole percentage of converted CO appearing as CO, Wt. percentage of hydrocarbon group in hydrocarbon product. Mole % in C,-C, fraction. Propagation probability determined from the slope (of a mol%) hydrocarbon versus carbon number plot.
5.2.1.1.I .2 Promotion by Light Transition Metal Oxides (Mn, V , Ti, etc.) In combination with Fe, light transition metal oxides such as those of Mn, V, and Ti are known to promote selectivity in FTS for light olefins (refs. 6,13,15,31,32). In an extensive review of “Olefins from Syngas” Snel (ref. 32) indicates that selectivity in FTS may be influenced by several factors including basicity (ability to change the work function of the metal), electron-withdrawing ability of additive ligands, demetalization and metal support interactions; while his discussion of these factors lends some new insights, it encounters complications because
Table 5.3 Comparison of the Fischer-Tropsch catalytic behavior of bulk and supported potassium-promoted iron catalysts with a carbonyl complex derived catalysta (adapted from ref. 29) Activity -
pmol CO rnin-lg-’ Fe -.
9 15 10 14
Fe/A1203, complex derived catalyst
-- --
-~
(F) 5.5% K, 3.9% Fe/A170? 1541 convGtiGna1 catalyst
20 16 13 11
150
227
-~ -
I
0.07 I 44 56 0
(I) Bulk K/Fe a
1 19 Tr 16 4 2 26 Tr 20 4
Hz/CO=3.0, 101 kPa (1.0 atm) total pressure.
2
0
0
0
22 Tr
-
24
0
0
0
12
5
4.5
13
165
all of these factors are highly interrelated and are difficult to understand or define. Some alternative, more widely accepted and more easily defined but nevertheless closely related concepts include: (1) control of the surface H/CO ratio by electronic interaction of these metal oxide promoters with the metal surface (refs. 33,34) leading to hydrogen-poor olefinic products (ref. 27), (2) facilitation of CO dissociation at the promoter/nietal interface (so-called adlineation sites) (refs. 35-40), and (3) the formation of hard-to-reduce oxides (e.g. spinels such as Fe2Mn04) which resist formation of carbides (refs. 41-43) and hence have different adsorption/catalytic properties, e.g. lower hydrogenation activities and lower rates of carbon deposition (ref. 44), than metallic Fe catalysts which are easily carbided. A number of recent catalyst developments have involved improvements in the Fe-Mn system (refs. 43-47). Barrault et al (ref. 45) emphasized the importance of the nature of the precursor in controlling catalyst selectivity; for example, the product from a catalyst prepared from a mixed oxide, Fe,MnyO,, contained 75% C2-C4 of which 70% were olefins. Kim et al. (ref. 46) investigated the influence of Mn concentration on the selectivity of Fe finding that the percentage of olefins increased from 52 to 84% as Mn content was increased from 0 to 10.6%. Fiato and Soled (ref. 47) claimed that a slurried high surface area Mn-Fe spinels promoted with Cu and Group IA or
IIA metals provide exceptionally high catalytic activity and selectivity to a-olefins with high activity maintenance under low pressure reaction conditions; for example, an Fe-Mn spinel containing 2% K and 1% Cu was found to produce 92-94% olefins in the C2-C4 fraction and about 60wt%. C2-C20 olefins at CO conversions of 7947% while the wt% of methane was only 4.I-4.2%. Venter et al. (ref. 43) investigated the preparation and activity/selectivity properties of carbon-supported Fe/MnK cluster-derived catalysts finding them to be significantly more active than previously reported Fe-Mn catalysts and to have CzC4 olefin selectivities as high as 85-90% with the balance being methane; no other paraffins were detected; for example KFe2Mn/C reduced at 473 K produced 11-14 wt% methane, 30-31 wt% ethylene, 31-39 wt% propylene and 20-25 wt% butene. These high olefin yields remained constant during 20-40 hours of reaction. The above reports of high selectivities to light olefins are exciting developments; nevertheless the reader is
I
‘tiLLzz
15
GO2
0.0 0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2
POTASSIUM, % Fig. 5.3 Effect of potassium loading on selectivity of 1% Ru/A1203 (from ref. 30)
166
cautioned that claims such as those above (refs. 43,45) and at later points in this article of C2-C4 selectivities exceeding the upper limit from ASF theory of 56% are probably due to experimental artifacts, although their existence due to deviations from theory, particularly in the case involving a termination step to methane, cannot be ruled out altogether. The reader is referred to ref. 7 for a more detailed discussion of this issue. While oxides of V and Ti increase the light olefin selectivities of Fe, vanadium oxide when applied to the support surface of 1% Ru/Ti02, 2% V205 increases the production of liquid (Cs+) hydrocarbons from 50.4 to 70.7% and produces 14 wt% branched aliphatics (ref. 48). Moreover in the case of supported Co, promotion by Zr, Ti, or Cr oxide results in the production of a hydrocarbon mixture a large fraction of which consists of linear C20+ paraffins which can be converted through mild thermal cracking to linear Cl0-C2o olefins (ref. 49). Thus, it is not possible to generalize the effect that a given promoter will have on different Group VIII metals in FTS. 5.2.1.I .I .3 Promotion by Lanthanide and Actinide Series Oxides Promotional effects of the lanthanide and actinide series oxides are probably in some ways similar to those of the light transition metal oxides in that they are thought to (1) decorate the surfaces of metal crystallites in Group VIII metal catalysts and facilitate the dissociation of carbon monoxide and (2) lower the acidity of the catalyst support similar to K, preventing undesired side reactions such as cracking of olefins to methane and other light hydrocarbons (ref. SO). However, in addition to these effects some of these oxides, especially Tho2, promote the formation of branched hydrocarbons (refs. 17,51,52), enhance the reducibility of Co in Cokieselguhr (ref. 5 3 , and increase hydrogen adsorption during reaction (ref. 54) thereby increasing reaction rate (ref. 27).
The role of Tho, and other oxides of the actinides and lanthanides as basic additives is described in patents assigned to Gulf Research (ref. 50) in which it is claimed that basic oxides lower the support acidity which otherwise functions to crack intermediate olefins to light gaseous products (Cl-C3). Data for representative Tho2-promoted Co and Co-Ru catalysts based on this technology in Table 5.4 are indicative of low selectivities for methane and light hydrocarbons and of high selectivities for gasoline and diesel fuels. That Tho2 and similar basic oxides promote the formation of branched hydrocarbons is supported by previous studies (refs. 51,52) and by the fact that Tho2 itself is a catalyst for the isosynthesis (ref. 17). Madon et al. (ref. 51), for example, found that basic promoters dramatically increase the chain branching of 1% Ru/alumina while Schulz (ref. 52) found that by increasing the Tho2 content of Co/Aerosil from about 1.5 to 9% the degree of branching increased from about 3-4 to 30.40% at 448-453K, 17 bar, H2/CO=l. Under similar reaction conditions Sarup and Wojciechowski (ref. 55) found the degree of branching for Cokieselguhr (no promoters present) to be about 1%. It should be noted that the degree of branching is also a strong function of reaction conditions, increasing with increasing reaction temperature and decreasing reaction pressure. Recent Exxon patents (refs. 56,57) utilize lanthanide and actinide promoters for quite different purposes: (1) enhancing the selectivity of a Fe/Mn/K for light olefins (ref. 56) and (2) improving the thermal stability towards regeneration of a Co/TiO2 catalyst for synthesis of
167
Table 5.4 Hydrocarbon Selectivities of Co/Th02/A1203, Co/ThO2/A12O3+Silicdite Catalysts (from ref. 50) Catalyst
T
P
H2/CO CO ratio
K
atm
50% Co/Th02/A1203
458
1
20% co,0.5% R~/ThO,/Al203
468
1
5o%cO/Th02 /A1203+Silicalite
Co-Ru/ThO2/A1203
HC Product Distribution
conv
9% 1:l
27
1:1
22.5
and
Carbon atom% CH4 C2-C4
CyC,
C9-Czo
Czl
1
7
25
59
4
I i
4
22
57
9 1
33
52
4
I
5
11
0
premium middle distillate fuels (ref. 57). In the former case (ref. 56) the addition of Ce to a Fe/Mn/Zn/K catalyst greatly enhances activity and selectivity for C2-C3 olefins while minimizing methane formation suggesting a synergism between K and Ce. This catalyst maintains high selectivities to olefins over a range of pressures and temperatures at conversions as high as 80-90%. The second development (refs. 57a,57b) involves the addition of an oxide of Zr, Hf, Ce, or U to Co/TiO2, a catalyst developed for production of fuel hydrocarbons consisting predominantly of C ~ Oparaffins + and olefins (ref. 57c). It was found that the Co/TiO, catalyst deactivates over a period of time due to deposition of coke and requires regeneration; unfortunately the Co was found to agglomerate during the high temperature regenerative treatment in air. Fortunately, the addition of 0.1-1% of Zr, Hf, Ce, or U oxide promoter to a 11% Co/TiOz stabilizes the catalyst during treatment in air up to 773 K (after 16 hours reaction at 473 K, 280 psig, GHSV=1500 and Hz/CO=2). It was hypothesized (ref. 57) that the oxide promoters are present as highly dispersed oxides on the Ti02 support and interact strongly with C0304 during the high temperature exposure to air, forming a stable matrix which maintains the cobalt in a finely dispersed form on the support surface. Takahashi et al. (ref. 58) used pulse surface reaction rate analysis (PSRA) to study the promotion of Ru/alumina by rare earth oxides and found them to increase the rate constant for co dissociation while decreasing that for hydrogenation of surface carbon. Consistent with the PSRA results the FTS selectivity of Ru/alumina for C2+ hydrocarbons was found to increase with the addition of rare earth oxides, and this occured more effective, by than for addition of V, Nb, Mo, and W oxides. 5.2.1.I .2
Effects of Gaseous AdditivelPretreatments In principle it should be possible to add gaseous compounds to the synthesis gas feed which modify the catalyst activity/selectivity or to pretreat the catalyst with gaseous compounds that adsorb on the catalyst surface and modify its behavior. Possible modifiers include sulfur, chlorine,
and other such poisons which could under favorable circumstances and careful use selectively
168
poison undesirable reactions such as methane and carbon formation; however, in practice the application of these poisons is difficult since they may adsorb strongly and nonuniformly while poisoning the desirable synthesis reactions (refs. 59,60). The modification of selectivity by addition of olefins, methanol, and other components that intercept intermediates is yet another possibility which is discussed in a later section (5.2.3). Research from the 1930's to the early 1980's on the partial poisoning by sulfur of FT catalysts has been reviewed by Bartholomew et al. (ref. 59); this early work indicates that methane production is poisoned by sulfur to a greater extent than is C,, production. More recently Stenger and Satterfield (ref. 61) reported that addition of 1% H2S/H2 in the amount of 1 mg S/g Fe to a liquid phase fused magnetite catalyst increased its activity by about 60%; higher levels of sulfur cause a decrease in activity; methane production was only slightly lowered while the olefdparaffin ratio did not change measurably. On the other hand, Matsumoto and Satterfield (ref. 62) reported decreased methane formation by a dibenzothiophene-poisoned fused magnetite catalyst. Unfortunately, the addition of sulfur in these two studies (refs. 61,62) was not sufficiently controlled to ensure gradientless deposition. Tong and McCarty (ref. 63) studied fused Fe catalysts pretreated under carefully controlled conditions with submonolayers of chemisorbed H2S finding a three-fold decrease in methane selectivity relative to the untreated catalyst while C,C, olefin selectivity was significantly increased (see Fig. 5.4). Chlorine is a temporary poison for methanation and FTS because it is removed
-
under
HdCO ratio I1 10
Clean Fe
o)
reaction
conditions
as
HC1.
Nevertheless, the removal of adsorbed
dim-level Sulfur-treated Molefins
chlorine
qn-Paraffi
relatively slow under reaction conditions;
8-
from
supported
catalysts
is
thus, chlorine adsorbed on the catalyst during
~
2
3
4
~
6
r
a
1
2
3
~
~
Product Carbon Number
Fig. 5.4 Fischer-Tropsch synthesis at 573 K and 100 kPa on clean and medium-level sulfur-treated fused iron catalysts (from ref. 63)
6
preparation
may
cause
partial
inhibition with accompanying changes in product selectivity. For example, Hagiwara et al. (ref. 64) found that during hydrogenation of CO to light olefins over a chlorine-containing magnetite catalyst in slurry phase, the selectivity to C2-C4 olefins was increased to 48%, while that for CH, was decreased to 11%. Fujitsu et al. (ref. 65) reported that 773 K evacuation of a Rh/Ti02 catalyst to remove CI improved its r 8 CO hydrogenation activity. Dramatic effects of C1 addition with gas phase HC1 on the selectivity of Rh/silica were reported by Kim et al. (ref. 66); alcohol and
169
oxygenate selectivities were sharply reduced while the selectivity for C2+ hydrocarbons was greatly increased (from 5 to 43%); activity was reduced almost twofold by C1 addition.
In a patent assigned to Dow (ref. 66a), a process is described for adding a volatile phosphorus compound to a conventional FT catalyst in the absence of halogen to improve C2-C4 selectivity. In view of the relatively small decreases in activity and large changes in product selectivity, the controlled addition of sulfur prior to reaction at submonolayer levels and of chlorine or phosphorus during reaction appears to hold promise for specific applications such as reduction of methane and increased production of light olefins on iron catalysts. 5.2.1.2
Effectsof Suovort. Metal Loading and DisDersion
ActivitylStructure Relationships and The Roles of Dispersion and Metal-Support Interactions Numerous studies, many of which are documented elsewhere (ref. 67 in Chapter 4), provide evidence that supports can greatly influence the activity and selectivity properties of Group VIII 5.2.1.2.1
metals for CO hydrogenation. The emphasis in this section is on recent studies that illustrate principles and activity/structure relationships relating to the design and selection of F T catalysts. A number of recent studies provided quantitative evidence that specific activities of CO, I%, Fe, Ru, and Mo in FTS can vary over orders of magnitude depending upon support; hydrocarbon selectivities can also be highly support dependent. The effects of support, loading and dispersion on the CO hydrogenation activity/selectivity properties of cobalt were studied by Bartholomew et al. (refs. 68,69). Reuel and Bartholomew (ref. 68) observed variations in the initial specific activity spanning nearly three orders of magnitude for cobalt supported on alumina, silica, titania, magnesia and carbon (see Table 5.5); the order
of
decreasing
activity
from
their
data
for
3
and
10%
samples
is
Co/titania > Co/alumina = Co/silica > Co/carbon > Co/magnesia. For a given cobalthupport system both initial and steady-state activities were found to increase with increasing loading and decreasing dispersion. In the case of Co/alumina initial activity at 498 K, 1 atm, and H2/CO=2 increases from about l.10-5 (ref. 70) to 6.340-2 s-1 (ref. 68, Table 5 3 , a factor of 6,000, as metal loading is increased from 1 to 15 wt%. The average carbon number of the hydrocarbon product for these same catalysts is also support dependent but appears to correlate reasonably well with dispersion and/or metal loading, as these two properties are highly correlated (dispersion generally decreases with increasing metal loading); that is, Fig. 5.5 shows a trend of increasing average hydrocarbon carbon number with decreasing dispersion (increasing metal loading). Data in Fig. 5.6 from Fu and Bartholomew (ref. 69) show that the product distribution for Co/alumina is shifted significantly to heavier hydrocarbons as metal loading increases; indeed the polymerization probability a increases from 0.70 to 0.90 as the wt% Co increases from 3 to 15%. Two other examples of significant changes in FTS activity with support involving Fe and Mo catalysts are illustrated by data in Tables 5.6 and 5.7.
170
Table 5.5 Turnover frequencies and activation energies for conversion of CO hydrogenation on cobalt catalysts (from ref. 68)
a
Turnover frequency for CO conversion (to hydrocarbons and CO,), i.e., the number of CO molecules converted per catalytic site (based on total H2 uptake) per second at 1 atm, H2/CO=2, 498 K. These data were measured within a few minutes of initial reaction and hence correspond to initial activities. Activation energy for CO conversion based on the temperature dependence of NCO at three or four different temperatures. Extrapolated values; in most cases the extrapolation was over a small (25 to 50 K) range of temperature. Inactive up to 673 K.
The data in Table 5.6 (refs. 71,72) for relatively poorly dispersed and highly loaded Fe catalysts (effects of metal loading and dispersion are minimized) indicate a factor of 20 variation in specific activity, activity decreasing in the order Fe, Fe/carbon, Fe/silica, Fe/alumina. Olefin/paraffin molar ratios vary from a low of 0.72 for Fe/alumina to a high of 4.1 for Fe/C; in other words, Fe/C is very selective for light olefins, similar to K- and Mn-promoted Fe catalysts. Similarly, specific activities for supported Mo catalysts reduced at 773 K (ref. 73) vary over more than an order of magnitude, decreasing in the order Mo/silica, Mo/alumina, Mo/carbon, Mo/ceria (see Table 5.7) which is the same order for degree of reducibility of the Mo. Product selectivities for C2+ also vary with support and are highest for the most highly reduced catalysts, Mo/silica and Mo/carbon.
171
Another recent example of very significant effects of support on selectivity comes from the work of Barrault et al. (ref. 74a), who showed that C,, selectivities for nickel are greatly enhanced by 5
0
1
3
2
4
5
6
7
Average Carbon Number (wt. basis) Fig. 5.5 Average carbon number of hydrocarbons produced at 498 K and 1 atm for 3 and 10 wt% supported cobalt catalysts as a function of dispersion (after 3h reaction): (A) Co/sioz (impregnated); (0)Co/A1203 (impregnated); (0) Co/riO, (impregnated); (0)Co/C (evaporatively deposited) (from ref. 68)
. -2 rn
z 0
.-4
+ 0 4
LL LT
.
. -6
w 0
I
Y
-C I
. -8
-1 5
-10
CARBON NUMBER
CARBON NUMBER
Fig. 5.6 (a) Lower curves: Hydrocarbon product selectivity (wt%) for 3% C0/A1203 (prepared by impregnation and reduced at 648 K) at 473 K, 1 atm after 24 h of reaction. Upper curve: Anderson-Schulz-Flory plot for 3% Co/Al2O3 at 473 K, 1 atm after 24 h of reaction (from ref. 69) (b) Lower curves: Hydrocarbon product selectivity (wt%) for 15% Co/A1203 at 473 K, 1 atm after 24 h of reaction. Upper curve: Anderson-Schulz-Flory plot for 15% C0/A1203 at 473 K, 1 atm after 24 h of reaction (from ref. 69)
172
I1 olefin/paraffin 1 at. ratio I I1 0.61 4.1
co turnover
Catalyst
frequency ,103, s-l Fe 10% Fe/carbon 15% Fe/silica 15% Fe/ alumina
4.0 2.8 0.36 0.21
'
1.2 0.72
ref
71 72 71 71
supporting it on rare-earth oxides while methane selectivity is greatly lowered relative to that normally obtained for conventional nickel catalysts. Moreover, 90% of the C2-C4 fraction for Ni/ceria after reduction at 873 K is olefins; nevertheless, C 0 2 selectivity for the same catalyst was high (70%) indicating very significant water-gas-shift activity. In a similar study of La- and Ce-promoted Co/carbon catalysts Barrault et al. (ref. 74b) observed that these promoters give rise to a 100-fold increase in specific activity. Moreover, C2-C4 selectivity is increased from 4 to 40% and this fraction is principally olefins. What accounts for these observed, dramatic variations in activity and selectivity as a function of support and metal loading for CO hydrogenation on Group VIII metals? Several authors (refs. 20,69,75) have suggested that these changes in activity and selectivity with metal loading might result from changes in surface structure due to changes in metal crystallite size or dispersion and/or the requirement for specific ensembles of metal atoms to dissociatively adsorb CO and H2, the distribution of which varies with metal crystallite size or exposed crystallite plane. There is, after all, an undeniable trend of increasing CO hydrogenation activity with decreasing dispersion for many of the Group VIII metals (ref. 75). This is illustrated by the data in Table 5.8 and Fig. 5.7 Table 5.7 CO conversionsa, initial specific rates of CO conversion, and product selectivities for CO hydrogenation on reduced molybdenum catalysts (from ref. 73) Initial rate of CO conversion ~~
~
Product selectivity ~-
~
%C2+d
I
I --
6.7% M0/A1203 6.7% Mo/Si02 6.7% Mo/Ce02 6.7% Mo/Carbon
-~
0.63 4.2 -
6.2 ~~
-1 4.9 I
11 57 5.7 50
i1
1
1.5 6.4 0.4 1.4
5 12 5 11
~
a Reaction conditions: 623 K, space velocity=2000 h-l, 3:l H2/CO, 140 kPa.
After 20-25 hr of reaction. CO turnover frequency, the number of CO molecules converted per site per second; site densities measured by 0 2 adsorption. d Mole% of C2+ hydrocarbons in product; the product consisted of about 50 mol% C1 and C2, hydrocarbons and 50 mol% C02.
173
Table 5.8 Comparison of kinetic data of Jung et al. (ref. 76) with data from Jones et al. (ref. 72) for CO hydrogenation on iron-supported carbon catalysts
t investigator
catalyst
average crystallite diama(nm)
Jung, Walker and Vannice (ref. 76)
2.5% Fe/C-1 5.0% Fe/C- 1 5.0% Fe/V3R 4.5% Fe/V3G
0.6 1.0 9.0 54
Jones, Neubauer and Bartholomew (ref. 72)
1% Fe/C
3% Fe/C 10%Fe/C
1.5 7.9
1
I I
I
N~
103~ at?& K H2/CO=2
E b kJEo1
0.86 0.94 4.4 34
84 87 95 10.5 64
I
0.65 5.5 18
CO adsorption assuming CO/Fe,=l for data of Jung et al. Determined by H2 adsorption assuming H/Fe,=l for data of Jones et al. Activation energy for CO conversion. Turnover frequency for CO conversion, i.e., the number of CO molecules converted per catalytic site per second at 493 K, H2/CO=2, and 1 atm. Data for Jung et al. were extrapolated from 548 K and H2/CO=3 using their reported values of ECO and H, and CO partial pressure dependencies.
a Determined by
showing increases in activity with decreasing dispersion or increasing metal crystallite diameter for Fe/C (refs. 72,76) and Co/alumina (ref. 69). Nevertheless, these changes in dispersion are also correlated with changes in metal loading and extent of reduction (see Table 5.8 and Fig. 5.7) and hence may relate to metal support interactions (refs. 27,67). Furthermore, there is strong evidence from recent studies of single crystal Ni, Ru, and Co (refs. 77,78,79) and of carbonyl-derived Co/alumina (ref. 79) and Fe/alumina (ref. 80) catalysts that CO hydrogenation activities of these metals are independent of dispersion and surface structure, if in the case of the supported metals surface contamination is avoided (by preparation on well-dehydroxylated supports) and if the extent of reduction is held above 50.70%. For example, specific activity data in Fig. 5.8 obtained by Johnson et al. (ref. 79) for three sets of catalysts, cobalt overlayers on two tungsten single crystals, W(100) and W(110), having significantly different geometries, 3 and 5% Co/alumina catalysts and a CoW/alumina catalyst prepared by decomposition of carbonyls on supported aluniina dehydroxylated at 923 and 1223 K, and a highly-reduced 10% Co/alumina catalyst prepared by conventional wet impregnation, fall along the same Arrhenius plot; thus, the activities of these Co catalysts of widely varying dispersion and surface structure are the same within experimental error. Data in Fig. 5.9a obtained by Rameswaren and Bartholomew (ref. 80) show that the activities of Fe/alumina catalysts in which the extent of reduction varies with metal loading (mainly due to variations in the extent of dehydroxylation of the support during preparation) decrease with increasing dispersion, while the activities of those having approximately the same extents of reduction are invariant with dispersion (Fig. 5.9b). A similar relationship was observed for cobalt catalysts over the full range of dispersion (0.100%) (ref. 79). Furthermore, the data of Johnson et al.
174
(ref. 79) in Fig. 5.10 indicate a general trend of increasing activity with increasing extent of reduction for Co/alumina. Thus, it appears that activity correlates better with extent of reduction, a parameter relating to the degree of interaction of metal and support, rather than with dispersion, a parameter related to metal crystallite surface structure. If then, activity/selectivity variations with support and metal loading are caused by metal support interactions, what kinds of metal-support interactions are important in the data shown for Co, Fe, and Mo catalysts above? In the case of the Mo catalysts (see Table 5.7), activity and selectivity for C2+ hydrocarbons is larger for the catalysts (Mo/silica and Mo/carbon) in which Mo is reduced to the more active metallic or carbided state and smaller for the catalysts (Mo/ceria and Mo/alumina) in which strong support oxide-Mo oxide interactions maintain the less active Mo oxide phases (ref. 73). Activity/selectivity variations for moderately or poorly dispersed Co and Fe
a.
%I7
/
80
8C
catalysts on different supports (Tables 5.5 and 5.6; Figs. 5.5 and -
80
- 15
-
W
a
40
40
=”
20
20
5.6) and as a function of metal loading (Table 5.8 and Fig. 5.7)
20
10
%D
t3
are best explained by decoration of the metal with support species that act as promoters (refs. 27,38,67,74,80-87);
5
the
variation with loading can be explained by a greater extent of
0 0
5
10
15
20
lo
25
METAL LOADING (Wt.96)
decoration for catalysts of lower loading and smaller particle size (ref. 27). The promotional effect has been attributed to (1) creation
I
of
more
active
promoter/metal
sites
at
the
interface
(refs. 35-40,74) as discussed previously in the section on promoters or (2) localized charge transfer at the promoter metal interface (refs. 70,72,86-89). While decoration of metal 5
10
15
20
25
0
30
METAL LOADING (Wt.%)
Fig. 5.7 (a) Influence of metal loading on H2 adsorption uptake, dispersion and extent of reduction for Co/A1203 catalysts. (from ref. 69); (b) Dispersion, turnover frequency, activation energy and In A for CO hydrogenation on Co/Al2O3 catalysts of different loading (from ref. 69)
crystallites by reduced support species which migrate during reduction on to the surface is well-documented in metal/titania systems (refs. 27,67,83) and in metalhare earth oxide systems (refs. 67,74,8 1,82), the decoration
175
of metal crystallites by non-reducible supports such as alumina, silica, or carbon must occur
by
mechanism.
some Wheeler
other and
Bettman (ref. 84) demonstrated that the hot acidic liquid formed of during calcination nitrate-impregnated catalysts dissolves some of the alumina and dopes the metal oxide with support material during drying or calcination; subsequent reduction results in a metal surface covered with support species. Reduction of metal oxides that interact strongly with the support (e.g. Co, Fe and Ni oxides with alumina) may also lead to decoration via decomposition of the spinel (ref. 85). In highly dispersed, low
55p K 52$ K
,
~
SOY, K
475
K
459 K
a, 0 COiw(lO0)
0
5% Co(923)
0
3%Co(1223)
10%
0.0017
Co(conv)
0.0018
0.0019
0.0020 0 .0 0 2 1
0.0022 0.0023
1/T(K ') Fig. 5.8 Comparison of the Arrhenius plots for the steady-state CO turnover frequency of the 0.75 ML C O W crystals with plots for four carbonyl-derived Co/alumina catalysts. Reaction conditions were 1 atm and H*/CO=2. The Co/alumina catalysts were carbonyl-derived, except for the 10% Co(conv) catalyst, which was prepared by aqueous impregnation. The numbers in parenthesis for supported catalysts refer to the temperature of dehydroxylation of alumina support (from ref. 79)
loading catalysts the possibility of direct metal-support interactions causing modifications in the electronic, adsorption, and catalytic properties of tiny metal clusters is yet another possibility. Two kinds of evidence support this view. Recent Mossbauer data for 1-3 wt% Co/alumina and Co/carbon catalysts (refs. 70,72) indicate that small superparamagnetic clusters of metal are present having electron densities different than those of bulk Co and Fe metals; moreover the isomer shifts (a measure of the electron density at the nucleus) are negative for alumina-supported Co and Fe and positive for carbon supported Co and Fe. A recent temperature-programmed desorption study (ref. 86) provides evidence that CO does not dissociate on well-dispersed metal clusters in 1% Co/alumina, a fact which may explain the relative inactivity of this catalyst for CO hydrogenation.
176
5.2.12.2
New Catalyst Technology Involving Novel Supports and SupportlDispersion Effects
A number of recent developments in catalyst technology incorporate novel supports to enhance activity and/or selectivity. This section emphasizes relatively large pore, non-zeolitic supports (pores in the meso- and macropore range); zeolite-supported materials will be discussed in Sections. 5.2.2 and 5.2.3 dealing with shape selectivity and secondary reactions. Murchison and Murdick (refs. 90,91) developed a sulfur-tolerant alkali-promoted Mo/carbon as a synthesis catalyst for a Dow process to produce LPG and especially ethane
. 0.010
5
cracking
as
a to
feedstock ethylene.
for The
catalyst is capable of producing
70% of the hydrocarbon product as LPG accompanied by very little liquid hydrocarbons; it cokes at a very
0.001
low rate and can be operated with no apparent deactivation in the presence of 10-20 ppm H2S.
b
10.000 15% Fe/Alumina. 30% reduction.
0 1- 4 5% FeIAlumina -50% reducton
0
formation
while
disadvantages of this catalyst are (1) its relatively low activity and (2) its relatively high methane selectivity (38 at% of
n
a
p?
methane
increasing selectivities to ethane and propane. Two
1.000 rnn
Addition of K to the Mo/carbon catalysts substantially lowers
0.100
x 0
2
the carbon in the hydrocarbon 0.010
0.001
0
10
20
30
40
50
60
70
80 90 100
% Dispersion
Fig. 5.9 Activity of Fe/alumina as a function of dispersion with (a) varying extent of reduction (%R); (b) with constant percentage of reduction (from ref. 80)
product) . Titania-supported metals are generally observed to have higher specific co hydrogenation activities than the corresponding metals supported on other supports (see previous subsection)
177
.01 C o l W ( 1 2 2 3 ) 5% Co(923 5% Co(1223) n
"
/ 0
3 % Co(1223)
3% CO(923)
1% Co(l223) 1% CO(923)
20
40
60
80
100
%Reduction Fig. 5.10 Carbon monoxide turnover frequency (485 K, 1 atm, H2/CO=2) versus % reduction of the carbonyl-derived Co/alumina catalysts. The numbers in parenthesis for supported catalysts refer to the temperature of dehydroxylation of the alumina support (from ref. 79)
(refs. 67-68,89,92). Indeed, CO TOF values are higher on titania-supported Co, Ni, Rh, Pd, Pt, and Ir, comparable on Ru, but significantly lower on Fe (refs. 68,92,93). There are several recent developments of TiOz-supported FTS catalysts of apparent commercial potential (refs. 57,58,94-97). Kugler (ref. 94) studied CO hydrogenation on Ru supported on alumina, magnesia and titania. Data shown in Figs. 11 indicate that selectivity to methane is lower and selectivity to Cz-C4 olefins is higher for Ru/Ti02 and Ru/MgO catalysts relative to Ru/A1203. Moreover, the results of this study indicate that selectivity for light C2-C4 olefins is increased by decreasing the Hz/CO ratio to low values (0.5-1.0) while selectivity to ethylene and propylene can be increased by raising reaction temperature from 503-523 K to 593 K at a Hz/CO ratio of 0.5. For example, at 523 K and H2/CO=1, the Cz-C, selectivity for the Ru/TiO, catalyst was found to be 73% while that for Ru/MgO at 623 K, 1 atm, and Hz/CO=0.5 was found to be 78%. These higher than ASF-predicted values (predicted limit is 56%) were argued to be a result of low surface hydrogen concentrations which in the case of Ru/TiOz and Ru/MgO result in lower than predicted methane selectivities. A similar study of Ru on alumina, carbon, silica and titania by Vannice and Garten (ref. 95) led to similar conclusions regarding the higher yield of C2+ hydrocarbons and olefins for Ru/TiOp
178
al.
RU 'A1203
Patents filed by Mauldin et (ref. 57) assigned to
Exxon Research & Engineering
RulTi02
describe Co/titania catalysts, preferably using a titania support
a.
having a ruti1e:anatase content of at least about 2:3 upon which is dispersed either cobalt or cobalt and thoria. These catalysts possess high activity and production
of
selectivity for
premium
grade
transportation fuels from either methanol or synthesis gas with stability enabling them to be employed over long periods.
c1 c 2 c 3 CL
CARBON NUMBER
Nevertheless, these catalysts slowly lose activity as a result of coking during extended operation,
rI LO
RU /A1203
RuI MnO
1
-
must be increased to maintain acceptable conversion. Ultimately these catalyst must be regenerated at high temperatures in air, and as
I-
u
3
0 0
explained earlier rare earth promoters must be added to these
(r
a
g
and thus the operating temperature
20
-
5
c3 CL CARBON NUMBER
Fig. 5.11 Comparison of hydrocarbon product distributions for FTS on (a) Ru/A1203 and Ru/Ti02 at 503 K, 103 Wa, H2/CO=3 and (b) Ru/A1203 and Ru/MgO at 623 K, 103 kPa, H2/CO=0.5 (from ref. 94)
catalysts (ref. 57) to prevent loss of surface area during regeneration. An Fe/titania FT catalyst having higher activity than that of the previous experience (refs. 92,93) is claimed in a patent by Fiato and Kugler assigned to Exxon Research & Engineering (ref. 96). The catalyst consisting of a mixture of iron carbide and iron oxides (ilmenite) supported on titania is useful for producing substantially C2+ alkanes. This
patent discloses that a pretreatment with CO or H2 plus CO at elevated temperatures prior to use improves activity of the catalyst pretreated in H2 or He; moreover, a minimum critical iron loading
179
of 2. g of Fe203 per m2 of titania is required for useful FTS activity.
2o
Addition of a potassium promoter
I-
also improves activity and selectivity
3 0
W
for C2+ hydrocarbons. For example,
2
an Fewtitania catalyst operated at
z
0
773 K for 5 h and CO at 623 K for 5 h converts 89% of the inlet CO to 18.9 wt% CH4 and the remainder to C2+ alkanes.
0
et
al.
(ref. 97)
compared the performance of Ru supported on alumina, titania, and Y-zeolite. prepared
The by
Ru/alumina proprietary
15-
a
563 K, 2 MPa, 500 v/v/h, and 1:l H2:CO after pretreatment in H2 at
Abrevaya
I
m OI
2
10-
rK 0
>
I
8
5 -
v,
v,
4
E 0 ' 0
was
'
I
2
1
I
4
I
I
6
l
l
8
CARBON NUMBER
reverse
micelle procedure; it was determined that its activity increased, the
Fig. 5.12 Comparison of prcoduct distributions obtained with oxycarbide catalysts. (X) C-Fe, (0) BASF. (0) C-FeK, (A) C-FeCr (from ref. 99)
molecular weight of the hydrocarbon product
increased,
water-gas-shift
activity decreased, and olefin/paraffin ratio decreased with increasing metal particle size. In contrast to Ru/alumina, Ru/titania and Ru/zeolite were more stable during reaction towards metal agglomeration and loss of Ru via carbonyl formation.
5.2.1.3
Interstitial Comnounds
Iron-group metals (Co, Fe, Ni) form interstitial compounds, borides, carbides, nitrides, and sulfides as well as mixed compounds such as carbonitrides and oxycarbides; their properties are discussed in detail by Levy (ref. 98) while Anderson (ref. 5) has detailed the behavior of iron carbides, nitrides, and carbonitrides in FTS. Nitrides of Co and Ni cannot be produced by typicai catalytic preparations; otherwise it is possible to prepare high surface area fomis of these interstitial compounds for all three metals. The interstitials are discussed below in somewhat descending order of importance/application.
CarbideslOxycarbides Under synthesis conditions the active phases for Co and Ni are bulk metals while in the case of iron, carbides, oxides and/or oxycarbides are formed (refs. 4,s). According to Dry (ref. 4) upon introduction to synthesis gas of partially reduced Fe/Cu/K20/Si02 the metallic iron phase is rapidly converted to Haegg carbide (Fe5C2), following which further reduction/carbiding continues slowly for several days while activity increases. During operation of iron catalysts at high conversions 5.2.1.3.1
180
some low-activity Fe304 is also formed. The relative distribution of different carbides and oxides depends upon reaction conditions such as H2/C0 ratio, pressure, and the presence of promoters/supports (ref. 4) according to principles discussed above in the first section of this chapter; for example, one of the principal roles of K 2 0 is to maintain iron in the active carbide phase (ref. 4). As discussed earlier, high temperature treatment of Fe/Ti02 in CO or CO plus H2 is necessary to produce active carbide phases (ref. 96). Thus, it should be emphasized that a near infinite set of combinations of different carbide, oxide, and oxycarbide mixtures is possible with iron FT catalysts depending upon composition, preparation, pretreatment, and reaction conditions. This should explain why so many different combinations of activity and selectivity properties are observed and reported for iron catalysts. It
also provides almost endless opportunities for developing new catalyst formulations with new and possibly better catalytic properties. For example, Snel (ref. 99) has developed an oxycarbide catalyst, Fe203C, (x=0-1), prepared from complexes formed in a femc nitrate and citric acid concentrate followed by calcination at 673 K for 1 h followed by reduction in H2 at 433-573 K for 20 h: unpromoted, K-promoted and Cr-promoted catalysts were investigated. This catalyst was found to be about 10 times more active than a commercial BASF fused iron catalyst and to have high activity stability over several hundred hours of operation; it also features lower methane make (8 wt% vs. 19 wt%) with selectivities to CZC, olefins comparable to the BASF catalyst. The K- and Cr-promoted catalyst have unusually high selectivities for C3 hydrocarbons (see Fig. 5.12). According to Anderson (ref. 5 ) carbides of Co and Ni are inactive for FTS, although surface carbides are probably important in the working catalysts. However, Mo-carbide has been shown to be fairly active for FTS (ref. 100). With the recent development of high surface area Mo and W carbides (ref. 101), there is potential for development of active and possibly sulfur-resistant FTS catalysts.
5.2.I .3.2 Nitridesicarbonitrides Nitrides and carbonitrides of iron are reported to be more active and produce less gaseous hydrocarbons than the carbides produced from the corresponding reduced catalysts (see Table 5.9) (refs. 5,15). They are more stable than carbide catalysts against oxidation and carbon deposition and hence have longer life relative to the carbides. Moreover, they preferentially catalyze synthesis of alcohols and low boiling hydrocarbons (see Fig. 5.13). Hunimel et al. (ref. 102) investigated surface and bulk changes in unsupported iron nitride catalysts during FTS synthesis using Mossbauer spectroscopy and quantitative mass spectrometry. Upon exposure to synthesis gas, two monolayer equivalents of N was removed accompanied by carbon deposition on the surface. There was no evidence for the presence of active nitrogen on the surface after the FTS had been established; thus it was concluded that carbides rather than nitrides were the active surface phase. However, following that initial loss of nitrogen, bulk carbonitrides were apparently formed and which lost nitrogen very slowly during further reaction. These authors speculated that the greater stability of iron nitride relative to iron catalysts might be due to differences in initial carbon deposition
181
Table 5.9
Tests of reduced and nitrided catalysts at high space velocity and temperaturea (from ref. 5 )
0 1.3 2300 589
Atom ratio, N:Fe Gas composition, H&O Space velocity, h-* Temperature, K Contraction, % Yield, wt% of Total hydrocarbons CI c2
51
c3 c4
Condensed products Up to 477 K Heavy oil Wax
0.40 1 .o
2820 56 1 48
10.3 14.8 20.4 16.8
8.6 13.7 16.6 14.2
26.2 7.2 4.3
36.3 9.5 1.1
a Operating pressure, 2.53 MPa
behavior and hence long term surface carbon inventory. However, in the case of Fe/silica catalysts Delgass (ref. 103) observed that the prenitrided catalyst decayed more rapidly than that prereduced in hydrogen; it was determined that the prenitrided Fe/silica catalyst lost its nitrogen immediately
and the decomposition process caused metal crystallite growth. Thk more rapid decay of the prenitrided catalyst was attributed to larger iron carbide particles depositing carbon at a faster rate. Schulz
et
al.
(ref. 104)
investigated a nitrided commercial fused iron catalyst (BASF) using XPS. During synthesis the nitrides were apparently converted into carbonitrides at “the surface”. Thus, they concluded that carbonitrides are the active surface phases. Schulz et al. also found higher stability of the ninided catalyst against oxidation as well as increased activity and alcohol selectivity during reaction. Obviously, there is a discrepancy in the interpretation of the results from these two studies of iron nitrides (refs. 102,104) regarding the composition of the surface, i.e. whether it is a carbide or carbonitride. This discrepancy is not easily resolved, since
5 ATOM RATIO, N . F e
Fig. 5.13 Effect of initial nitrogen content on production of alcohols and olefins over iron (from ref. 5 )
182
it is difficult to show conclusively from the study of Hummel et al. (ref. 102) whether the desorbing nitrogen came from the upper two monolayers only or if any N remained on the surface. Nor is it clear how many monolayers were included in the surface analyzed by XPS in the study by Schulz et al. (ref. 104).
5.2.1.3.3 Borides Metal borides have been suggested as potential catalysts for FTS (refs. 105-107) but have received surprisingly limited investigation, despite promise as active, sulfur-resistant catalysts (refs. 105,108,109).However, several recent studies by Bartholomew et al. (refs. 106-111) provide definitive data confirming that borides of Co, Ni, and Fe are promising catalyst candidates for methanation and FTS of coal synthesis gas in view of their high activities, selectivities and sulfur resistances. Borides of cobalt show the greatest promise for FTS as indicated by data in Table 5.10 and Figs 5.14 and 5.15 from Wang and Bartholomew (ref. 110) showing them to have high activities and selectivities for C2+ alkanes. For example, COB and CoB/A1203 catalysts prepared by diborane reduction of cobalt acetate in THF are about 10 times more active than the corresponding unsupported and alumina-supported cobalt catalysts (see Table 5.10 and Fig. 5.14), while Na-promoted COB catalysts (CoB/Na and CoB/Na/A1203) prepared by NaBH4 reduction of Co acetate in diglyme have activities comparable to unpromoted cobalt catalysts (see Table 5.10). Both COB and Na-promoted COB catalysts maintain high activity during 50-70 hours of reaction. The hydrocarbon product from these catalysts at 1 atm and H2/CO=2 is very similar to that obtained over conventional cobalt catalysts (see Table 5.10 and Fig. 5.15) consisting of mainly paraffinic gasoline and diesel hydrocarbons. The observed higher activity of COB catalysts relative to Co
Table 5.10 Steady-state CO hydrogenation activity/selectivity data for cobalt and cobalt boride catalysts at 1 atm and H2/CO=2 adapted from (ref. 110).
- -~ ~-
-4-
K
S-’ ~-
COB CoB/Na
453 463 co 468 Coma 533 CoB/A1203 468 3%c0/A1203e 473 CoB/Na/Al20? 531
-
6.0 0.85 0.9 1 0.96 21 1.4 5.2
-
-
~
2.09 4.8 0.9 2.05 6.2 2.6 0.34 4.9 3.5 0.37 4.1 74 2.91 3.7 0.8 0.31 5.2 1.3 2.63 5.0 57
-
wt%
B
% ~~~~
Selectivity
‘conk Sel
~~
23 22 22 51 20 16 36
23 24 26
-
45 51 41
48
1
I8 24 52
51 54 12
~
~-~
CO turnover frequency in molecules of CO reacted per site per second. CO reaction rate in moles of CO reacted per gram catalyst per second. Mole percentage of converted CO appearing as CO,. Product distribution based on total hydrocarbons in the product. e Data from ref. 68b
a
183
catalysts is consistent with previously reported data (ref. 112) showing NIB to be initially more active than Raney Ni for
0.1000
-
0.01 0 0
Y COB-109G
7-
CO methanation and with 0 a, In the hypothesis of electron v CoiNa-100 0 0 donation from boron to z 0.001 0 the metal (ref. 113) facilitating co dissociation, if it is assumed that CO 0.0001 dissociation is the rate 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 determining step for FTS under these conditions. 1000 ( T O K ) This latter assumption is Fig. 5.14 Arrhenius plot for unsupported cobalt, cobalt bonde and reasonable for low cobalt/sodium catalysts at 1 atm and Hz/CO=2 (from ref. 110) co temperature hydrogenation (refs. 5,114). An investigation by Wang and Bartholomew (ref. 111) of iron borides provides evidence (from Mossbauer spectroscopy) that stoichiometric FeB and Fe2B/Na catalysts are obtained by diborane reduction of Fe acetatenHF and NaBH4 reduction of Fe acetate/diglyme respectively. At 21 atm, 513-558 K and H2/CO=2 the FeB catalysts are about a factor of 10 less active than Fe and Fe/K catalysts (ref. 115) while the FeB/Na catalysts have activities comparable with Fe and Fen< catalysts (see Table 5.11). Unfortunately, the boron-containing iron catalysts have high selectivities for methane relative to Fe and Fe/K (see Table 5.1 1). In view of their similarities to iron nitrides, iron borides might be expected to produce alcohols; indeed, a commercial, low surface area FeB was found to produce 10-11 wt% alcohols at relatively low reaction temperatures (498-510 K) and 21 atm (see Table 5.11, Footnote g). Unpromoted iron borides are not stable under FTS reaction conditions over long reaction times, however their stability is enhanced by sodium promotion (ref. 111). Moreover, relative to unpromoted iron these catalysts are 10 times more sulfur-resistant (last 10 times longer) in the presence of 1 ppm H2S (ref. 109). 5.2.1.3.4
Sulfides
Steady-state methanation activities of Co, Ni, and Fe in the presence of ppm levels of H,S are 3-4 orders of magnitude below those of the fresh catalysts (ref. 59). This fact indicates that the steady-state activities of the corresponding metal sulfides of Co, Ni, and Fe are essentially too low to measure under normal operating conditions. Nevertheless, it is possible that effects of sulfur poisoning could be mitigated by (1) operating under more severe operating conditions and (2) using
184
50
0
I 4 5
-1
40
-2
L 35
-3
0
C
30
-4 2
25
-5
z
-
20
-6
-
15
-7
s 10
-8
z
5
-9
L
2 1 0
S
_-.-.......
0 1
2
3
4
5
6
7
8
-1. 0-
9101112131415
Carbon Number
Fig. 5.15 Hydrocarbon product distribution and Anderson-Schulz-Flory plot for COB at 1 atm, 441 K and H2/CO=2 (from ref. 110)
promoters such as potassium that donate electrons, since the poisoning effects of sulfur may be due in part to its strong electronegativity (refs. 59,60). Moreover, it is clear that sulfur is more effective for poisoning methane formation than formation of C l + hydrocarbons (ref. 59). This latter fact coupled with the former two possibilities provide the basis for the development of sulfur-tolerant methanation, FT and alcohol-synthesis catalysts (refs. 1 16-123). Most of these previous developments involved alkali- or rare earth-promoted MoS2 catalysts: however, in a patent assigned to Battelle Develop. Corp. (Ohio) (ref. 117) a process is described for producing low boiling aliphatic (C,-C,) hydrocarbons over a sulfided CoO/ZnO/A1203 catalyst at relatively severe reaction conditions (H2/CO=1.4, 623-723 K, 600-10,000 psig and 200-6,000 h-’). Promoted Mo sulfide catalysts developed for the synthesis of alcohols (refs. 120-123) are discussed by Mills (ref. 15) and in Chapter 7. 5.2.1.4
Bimetallics
Since Chapter 6 covers bimetallic catalysts in detail, this section provides only a brief synopsis of design principles and selected, new catalyst developments. The structure and catalytic properties for CO hydrogenation and other reactions of iron-containing bimetallic catalysts has been reviewed by Guczi (ref. 124). Most of the recent catalyst developments for FTS involve Fe, Co, and Ru bimetallics--especially Co-Fe (refs. 125-127) and Co-Ru (refs. 50, 128a) catalysts, although Co-Re (ref. 128b), Co-Au (ref. 128b), and Ru-Re (ref. 128d) FTS catalysts are also reported; the principal objective of their study was to develop inore active, more selective, or more stable catalysts. A number of these investigations have met with xucccss.
185
Table 5.11 Steady-state CO hydrogenation activity/selectivity data for iron and iron boride catalysts at 21 atm and H2/CO=2 adapted from (ref. 11 1)
[ Catalyst
T
Hydrocarbon
I
Olefin
SeIectivityd
L
I
C1 C2-C4 CyCIs Alc
C?-C4
K
I FeB-Com
1
I
FeB- 1 FeB-2 FeB-Na
E$Kf
179 93 neg 173 109 105
523 547 532 513 523 523 1
1 0.50 0.26 0.30 I 0.53 0.52
I
0.67
a CO turnover frequency in molecules of CO reacted per site per second at 523 K.
CO reaction rate in moles of CO reacted per gram catalyst per second at 523 K. Mole percentage of converted CO appearing as C 0 2 . Product distribution based on total hydrocarbons in the product. Chain propagation probability. Data from ref. 115 for reduced Fez03 or K/Fe203, valid at 8 atm and H2/CO=3; rates corrected to 21 atm and H2/CO=2 using reported rate equations. Selectivity data calculated from rate distribution curves in ref. 115 6 1 1 and 10% alcohols produced at 498 and 510 K. Aldehydes rather than alcohols.
Nakamura et al. (ref. 125) studied Co-Fe alloys as a function of composition, finding a maximum in the activity for CO conversion and in selectivity for C2-C3 hydrocarbons at 50 at% Co (see Fig. 5.16). In addition these workers observed enrichment of the surface in Fe using Auger electron spectroscopy. Stanfield and Delgass (ref. 126) found that addition of cobalt to iron retarded formation of bulk carbides; at higher than 25 at% Co no carbides are formed. In a patent assigned to Exxon Research and Engineering Co. Soled and Fiato (ref. 127) describe Co-Fe bimetallic slurry catalysts (Co:Fe atomic ratio of 1:4), prepared from reduction/carbiding of high surface area spinels, having high activity (79% CO conversion) and reasonably high selectivity for conversion of C0& to cc-olefins (i.e., 16 wt% C2-C1 of which 90% is olzfins). CO-RuFTS catalysts have been reported in patents assigned to Gulf Research (ref. 50b) and Exxon Research and Engineering Co. (ref. 128a). Beuther et al. (ref. 50b) describe a Co-Ru catalyst consisting of about 20% Co, 0.5% Ru, and 25% Tho2 supported on alumina useful for convertin& synthesis gas to diesel fuel in a fluidized bed process. This catalyst technology builds on a previous patent disclosing a 50% Co/Th02/alumina catalyst (discussed in Section 5.2.1.1) (ref. 50a): the purpose of Ru addition is to reduce the cobalt loading and thereby reduce the catalyst's tendency to agghmerate and undergo attrition in the fluid bed application. Data in Table 5.4 indicate that this bimetallic catalyst has activity, and selectivity properties comparable to the 50% Co catalyst. The bimetallic catalyst is also claimed to have high attrition resistance and the proper particle size distribution for use in a fluid bed.
186
Iglesia et al. (ref. 128a) 100
80
60
40
zow 20
0
0
25
50
75
100
describe a Co-Ru/Ti02 catalyst which is claimed to have higher activity, a lower methane yield and higher C5+ yield relative to CoRiO2; moreover it is possible to regenerate this catalyst, in situ , in low temperature (473-523 K) flowing hydrogen. For example, at synthesis conditions of 473 K, H2/CO=2, and 2050 kPa, Co/”riO2 converts
0
49% CO to a product containing 7%CH4 and 85%C5+ at a space velocity of 450 h-1 while Fig. 5.16 Selectivity of Fe-Co alloys at two temperatures (A, Co-Ru/TiOz converts 61% CO 523 K B, 548 K). 0 CH4, C2fC3, C4+. Reproduced with permission of Nakamura et al. (ref. 125) to 5% CH4 and 91.4% C5+ at a space velocity of 1,200 h-l; in other words, the space time yields (conversion times space velocity) are 1.4 and 4.7 h-1 for Co and Co-Ru respectively. 0
C o g u L K , atom %
25
50
75
100
CoBULK.atom %
Moreover, TGA data in the patent confirm that Co-RuRiOz deposits less carbon in CO/H2 atmosphere and that it is removed at lower temperatures in hydrogen. The developers speculated that Ru promotes hydrogenolysis activity and an intimate association with cobalt allows carbon deposits on the catalyst to be gasified in hydrogen atmosphere. Effects of Pretreatment and Preoaration Preparation and pretreatment procedures involving a complex set of chemical and physical processes are critical in establishing from among a wide range of possible structural properties a unique set of chemical and physical properties for a given catalyst which in turn determine its unique activity and selectivity properties. For example, as pointed out previously, variations in preparation, pretreatment and reaction conditions enable almost infinite variations in the composition of iron catalysts ranging from carbides to oxides and mixtures in between. Moreover, gaseous additives such as sulfur and chlorine can greatly affect activity, selectivity, and catalyst stability during reaction. In spite of the infinite possibilities it is possible to discuss some basic principles and cite examples of recent developments illustrating these principles; the discussion of preparation (impregnation, precipitation, deposition procedures) and pretreatment (calcination, reduction, carbiding, sulfiding, etc. procedures) will be in some cases combined since the two kinds of effects are so interwoven. 5.2.1.5
187
As our knowledge and application of the chemistry of catalyst preparation is becoming more sophisticated there is a trend toward more sophisticated preparation techniques involving, for example, the use of nonaqueous impregnations, organometallic complexes, laser or microwave decompositions, high surface area spinel precursors, and micelle chemistry. Indeed, the field of catalyst preparation has grown from a well-protected art to a developing science, with international symposia now held on this subject (ref. 129). Recent developments in FTS catalyst preparation discussed below also testify to this increasing sophistication; the discussion will be divided into two parts dealing with (1) general developments in FT catalyst preparation/pretreatment and (2) preparations of FT catalysts based on metal clusters/organometallic complexes, in view of the considerable recent activity in the latter area. 5.2.15.1
General Developments in FT Catalyst PreparationlPretreatment
5.2.1S.1.I Preparation The preparation of well-dispersed, uniformly deposited supported base metals at practical metal concentrations requires special care. Classical impregnation techniques may result in nonuniform distributions through particles and pellets while ion-exchange techniques are only useful for depositing low concentrations of metals (e.g. less than 1-2 wt%). However, recently developed precipitation techniques in which pH is generally carefully controlled enable uniform deposition of the active precursor material at practical loadings. For example, Boudart et al. (ref. 130) were able to prepare well-dispersed iron on MgO by calcination/reduction of magnesium hydroxy carbonate in which part of the Mg2+ cations were exchanged with ferric ions; it was found that migration and growth of iron particles during reduction was hindered by the presence of FeO clusters interacting strongly with support. CO adsorption indicated that iron crystallite diameters ranged from 1.5 to 30 nm. Topsoe and coworkers (ref. 131) were able to improve on the dispersion as well as the uniformity of
3% Ni/Al2O3
100% Ni
d
mpregnated
I2O3
3.6% N i / S i 0 2
2.9% h
Precipitated
Precipitated
2.8’
ko,
I’rccipitatcd
Fig. 5.17 Effects of support and preparation on methane turnover frequency of nickel at 525 K: Shaded bar is proportional to the CH4 turnover frequency; unshaded bar denotes the C?+ hydrocarbon turnover frequency; total bar length is CO turnover frequency (from ref. 132)
188
Table 5.12 Effect of dehydrating promoters on FTS behavior of Ni (catalyst: 50% NiO/50% promoter) (from ref. 135)
Promoter
Tho2
1
k;;a
SNib
Nf
Conv.,%
Carbon yield
T
c, c,-c, cyc* co*
12.1 10.7
42.8
I 51.8
15.9
45.4 17.3
40 (510) 49.4
39.2
40 (514)
100 (514)
3.14
20.4
5.2
I 40 (507) I 1100 (531)
8.8 24.8
4.2 0.96
100 (543)
38.6 76.2
38.5 1.3
40 (533)
32.8
39.0
40 (542) 46.5
33.8
49.1
I
4.5
11.3
I
19.6
a Pseudo first order rate constant at 500 K in units of h-I
Nickel surface area in m2 g-1 Turnover number in molecules CO per Ni atom per second at 500 K.
impregnation and sharpness of the iron crystallite size distribution of Fe/MgO using a coprecipitation technique to form a mixed Fe/MgO hydroxide followed by calcination and reduction. Nickel is generally thought to be an active, selective catalyst for methanation - not for FTS; nevertheless, it can, through appropriate preparation and application of basic supports/promoters, be transformed into a useful FT catalyst. For example, the effects of preparation on the CO hydrogenation activity and selectivity of nickel on different supports were investigated by Bartholomew et al. (ref. 132). Their data, summarized i n Fig. 5.17, show that specific activities and C2+ selectivities are significantly greater for Ni/alumina and Ni/titania catalysts prepared by controlled pH precipitation (refs. 132-134) relative to Ni, Ni/silica, and Ni/alumina prepared by impregnation; indeed the precipitated Ni/titania has sufficiently low methane selectivity to be suitable for FTS. Hadjigeorghiou and Richardson (ref. 13.5) found in a comprehensive study of precipitated Ni catalysts supported and/or promoted by alumina, thoria, and other alkaline earth or rare earth oxides that when activated by rapid calcination (for a few seconds) to 773 K to increase the metal-support or metal-promoter interface, these materials have high selectivities for C2, hydrocarbons (see Table 5.12); in particular Ni/ThO2 has high activity (4 times that of Ni/A1203) and high selectivity for C2+ hydrocarbons (86 at% of the hydrocarbon product at 100% CO conversion, 514 K, H,/CO=2.2: 101 Wa) with low (14 at%) selectivity to methane; under these
189
conditions 16 at% of the CO reacts to C02 In other words, this catalyst is a very active, selective catalyst for FTS. A continuous precipitation method at pH ranging from 6.6 to 6.9 at 353-373 K to produce an Fe/Cu/K catalyst relatively free of nitrogen in order to facilitate its use in connection with a shape-selective zeolite such as ZSM-5 is the subject of a patent assigned to Mobil Oil Corp. (ref. 136). The low nitrogen content reduces contamination of the second-stage zeolite catalyst. Another general and relatively new approach to preparation applied to FTS catalysts involves use of nonaqueous solvents for impregnation or continuous deposition. Advantages of this approach are several fold: (1) many nonaqueous solvents of higher volatility than water are more easily evaporated from the support; ( 2 ) it enables more uniform deposition in many cases and (3) it enables hydrophobic supports to be more easily wetted with compatible solvents. For example, an evaporative, deposition technique using benzene/ethanol as a solvent and nitrogen as the evaporating agent accompanied by mechanical mixing was developed originally for the preparation of Pt-Fekarbon alloys (ref. 137) but was extended to the preparation of Co/carbon and Fe/carbon catalysts (refs. 68,72): this method enables highly-dispersed and relatively highly reduced metals to be prepared on carbon supports. Beuther et al. (ref. 50a) describe the preparation of a Co/Th02/alumina catalyst using a nonaqueous impregnation of either cobait nitrate or cobalt carbonyl; suitable solvents are ketones, lower alcohols, amides, ethers, hydrocarbons, such as pentane and hexane, and mixtures of the forgoing solvents. The preferred solvent is a mixture of ethanol and acetone. The impregnation is preferably carried out on a support previously dehydrated from 673 to 873 K in air after which the solvent is evaporated at 298-318 K, followed by heating in inert gas to 473 K and reduction in hydrogen at 523-673 K. This method produces a high surface area catalyst having an H2 chemisorption uptake of 100-300 pmoles/g. As explained previously this catalyst has high activity and selectivity for gasoline and diesel fuel hydrocarbons. Some exotic preparations have also been reported. Fiato et al. (ref. 138), for example, describe the preparation of an iron carbide catalyst for synthesis of olefins from COz and H2 by gas phase decomposition with a laser beam of a mixture of a volatile organic iron-containing compound (e.g. Fe(CO)Sj and a volatile organic first row transition metal-containing promoter compound (e .g Mn2(COjIO): the decomposition is typically carried out in the presence of diluent including an inert gas and a hydrocarbon such as ethylene to adsorb heat. A typical catalyst solid from the laser preparation (collected on a 0.5 micron filter) is mostly Fe3C and has a surface area of about 25 m2/g and converts 22-31% of the reactant CO, to hydrocarbons of which 5 6 % is methane and and 94-95% is C2+ hydrocarbons (at 543 K, 7/1 H2/C02, 3800 v/g Fe/h, 75 psig, slurry reactor); the % olefins in the Cz-C, fraction is reportedly 93.96%. Abrevaya and Targos (ref. 139) describe preparation of Ru catalysts by a reverse micelle technique in which the metal is deposited on the support in crystallites of diameter smaller than 20 nm and which do not vary more than 2 nm in size. The method involves contacting the support with a microemulsion consisting of a hydrocarbon liquid and aqueous cores containing ions of the dissolved, unreduced metal therein: separating the impregnated support; calcining to decompose the
190
metal salt; and reducing to effect metal reduction. This method was used to prepare a series of Ru/alumina FT catalysts for study of crystallite size effects (described above in Section 5.2.1.2) (ref. 97).
5.2.15.1.2 Pretreatment A number of studies indicate that activity/selectivity properties of Co and Fe FT catalysts can be significantly altered by variations in calcination and reduction conditions. For example, Reuel and Bartholomew (ref. 68) found that by increasing reduction temperature for Co/alumina and Cohitania catalyst from 650 or 675 K to 800 K, the average carbon number of the hydrocarbon product increased by 10 to 15%. Lohrengel et al. (ref. 140) studied the effect of increasing reduction temperature from 573 K to 773 K on the adsorption and catalytic properties of a precipitated Fe/Mn/Zn/Cu/K FT catalyst.
While BET surface area and pore-size distribution were little affected by the increasing reduction temperature, the heat of adsorption of hydrogen was increased 30% while that of ethylene was decreased by a factor of 2. Specific activity decreased 4-5 times and the selectivity toward olefins and short-chain hydrocarbons was significantly enhanced. Since ESCA studies revealed no significant difference in surface composition, the authors attributed the difference in behavior to formation of different catalytic surface compounds at the two different reduction temperatures; however, these effects might also be explained by more uniform spreading of Mn, Cu, Zn, and K oxides on the surface at the higher reduction temperature (ref. 27b). Dictor and Bell (ref. 115) studied the influence of reduction temperature on the physical and activity/selectivity properties of a fused iron catalyst. They observed a marked increase in activity and BET surface area with increasing reduction temperature which they attributed to the formation of additional pores upon more severe reduction. Nevertheless, the surface composition/structure of the catalyst was not independent of extent of reduction, since methane selectivity decreased and hydrogen reaction order increased with increasing reduction temperature and/or reaction time. More dramatic changes in activity and selectivity due to precalcination of a 15% Fe/3% Wsilica catalyst (reduced 36 h at 723 K) were observed by Rankin and Bartholomew (ref. 26). Their data indicate that activity decreases by a factor of 100, the activation energy decreases from 124 to 32 kJ/mol, while selectivity for light olefins increases from 60% to 99.7%. While the product distribution of the catalyst calcined at 373 K is typical of FT synthesis on F e K and consistent with the ASF model, the product of the same catalyst precalcined at 473 K consists only of methane, ethylene and propylene. The significantly lower activation energy suggests that a different mechanism may be operative on the catalyst calcined at the higher temperature. The authors attributed this dramatic change in catalytic behavior to increases in the adsorption activation energy for hydrogen due to greater silica or potassium silicate decoration of the metal surface. Pretreatments in CO/H2 are important in the case of iron catalysts, as they determine the degree of carbiding. For example, as mentioned earlier, a high temperature treatment in CO or H&O enables the creation of an active Fe/titania catalyst containing a mixture of iron carbides and oxides (ref. 96). On the other hand, Baltrus et al. (ref. 141) found that pretreatment of an Fe/K/Cu
191
catalyst in C0/H2 caused greater carbide formation than pretreatment in CO; a lower activity for the CO/H*-pretreated catalyst was explained by greater coverage of the active surface by surface carbonaceous material during pretreatment. Preparations Based on Metal Clusters/Organometallic Complexes The preparation of highly-dispersed, highly-reduced, uncontaminated supported metals, particularly base metals, is a difficult task because of the tendencies for strong oxide-support interactions and for decoration of the metal with support species or with S, C1, or N from the metal precursors. However, through decomposition of carbonyl and other organometallic complexes on carefully dehydrated supports, it is possible to produce highly-dispersed, highly-reduced, contaminant-free supported metals (refs. 21,28,29,80,142-157). Several investigators (refs. 21,28,29,80,146,147,152,153,156,157)have reported that carbonyl-derived catalysts (CDCs) are 3-50 times more active than catalysts prepared by decomposing inorganic salts onto a support. Detailed reviews of preparation, activation, and FT activity/selectivity properties of CDCs with emphasis on studies before 1985 are available elsewhere (refs. 145-148). Data from some early studies (refs. 143,146) of Fe and Co CDCs suggested that it might be possible to obtain higher selectivities for C2-C4 olefins than predicted from ASF theory; moreover the observed activities of these catalysts were unexpectedly low. However, more recent studies by Bartholomew et al. (refs. 21,80,156,157) of Co, Fe, and Ru CDCs prepared on carefully dehydroxylated supports indicate that these catalysts are generally more active than the corresponding catalysts prepared by conventional means; moreover ASF statistics are generally 0
24
-1 20
-2 -5
18
-4
12
a d
8
-7 4
d -9
0
CARBON NUMBER
Fig. 5.18 Hydrocarbon product distribution (0)and Anderson-Schulz-Flory plot (A) for carbonyl derived 3.7% Co/A1203 from ref. 156
192
Table 5.13 CO hydrogenation activities and selectivities of carbonyl-derived alumina-supported Co, Fe, and Ru at 488 K (H2/CO=2, latm) Catalyst
Nco"
co2
Hydrocarbon
Olefin
ad
Ref
Selectivityc .lo3 S-1
inProdb CH4
C2C4 CsCll C12+ C3-C7
%
%
3.7% C0/A1203~
6.2
9
23
28
3% C0/A1203~
2.8
1.3
16
24
I
I
8
80
0.78
156
54
6
38
0.70
69
42
4.5% F e / A 1 ~ 0 3 ~ 0.21
10
22
40
35
3
63
0.63
80
4.1 % Fe/A1203f
0.12
51
23
42
35
0
76
-
80
3% Ru/A1203e
1.5
4
21
25
33
21
65
0.70
21
a
_
_
Turnover frequency in molecules of CO converted per hydrogen adsorption site per second. Mole percentage of C 0 2 in product (excluding unconverted CO and H2). Wt.% hydrocarbon by carbon number grouping. Propagation probability detennined from the slope of mol % hydrocarbon versus carbon number plot. Carbonyl-derived catalyst. Catalysts prepared by aqueous impregnation.
observed (refs. 4,80,156) (see Fig. 5.18); in other words, there is no evidence to support the hypothesis of some previous authors (refs. 143,146) that small metal clusters terminate the chain-growth process causing deviations from ASF kinetics. Activity/selectivity data for representative alumina-supported CDC catalysts (refs. 2 1,80,156) listed in Table 5.13 indicate that
Co and Fe carbonyl-derived catalysts are about a factor of two more active than the catalysts of corresponding loading prepared by aqueous impregnation of metal salts. Hydrocarbon selectivities are not very much different for the two different kinds of catalysts; however, the fraction of light olefins is greater for the Co CDC relative to the conventional catalysts, while the opposite is true for Fe. However, if the catalyst activities are compared on a catalyst weight or metal weight basis, the CDC catalysts fare better because of their significantly larger metal surface areas. Contrary to earlier reports (refs. 143,147) indicating that CDC catalysts are unstable, the 3.7% Co/alumina CDC listed in Table 5.13, previously reduced at 573 K and with a dispersion of 24%, was found to be relatively stable, losing only 13% activity over a 24 hour period (ref. 156). Recent developments in metal CDC catalyst technology include (1) the preparation of highly-dispersed zeolite-supported catalysts (refs. 144,149,150,152,153), the properties of which are discussed in the next sections dealing with shape selectivity and interception of intemiediates: (2) the development of potassium-promoted catalysts in which the promoter is in more intimate contact with the promoter (refs. 28,29) as discussed in Section 5.2.1.1.1.1; and (3) of active, high surface area alumina-supported FT catalysts from the Fe and Ru groups promoted with Zr, Ti, and
~
193
Hf group metal oxides (ref. 154). The latter group of catalysts are prepared by treating an inert alumina with a nonaqueous solution of an alkoxide of the Zr, Ti or Hf group, removing the nonaqueous solvent, impregnating with a nonaqueous solution of the carbonyl from the Fe or RU group, drying and reducing. These promoted catalysts are active for converting synthesis gas to hydrocarbons in the C5-C2, range.
5.2.2
LIMITATIONS OF CHAIN GROWTH BY SHAPE SELECTIVITY
Zeolites are high surface area, crystalline alumina-silicates having well-defined microporous structures. This facilitates the application of these materials as molecular sieves, allowing only molecules of a certain minimum size to enter or leave the internal micropores of 0.4-1.1 nm due to either geometric or diffusional restrictions. Since zeolites are inorganic cation exchangers, it is a “straightforward matter” to prepare metal-containing zeolites by exchange with reducible transition-metal ions (refs. 158,159); alternatively, metals can be incorporated in the zeolite structure by decomposition of carbonyls inside the pores (refs. 144,149,150,152,153). In principle, then, the normal chain-growth process in FTS predicted by ASF kinetics might be terminated by either geometric or diffusional constraints on the product molecules in the micropores of these metal/zeolite catalysts. In addition to shape-selectivity, zeolites have well-known acid functions which facilitate secondary reactions in FTS such as cracking, hydrocracking, oligonierization, and isomerization (refs. 7,15,158,159). These latter reactions can be important in shifting selectivity in favor of premium products such as olefins or gasoline. Accordingly, there are three general approaches that might be used in the application of zeolites for FTS (ref. 7): (1) incorporating metals inside the zeolite structure to limit the size of
a.
C. t
I
9 12 15 Corbon Number
3
6
:
3
:
.
,
,
9 12 15 Carbon Number
6
i8
0
4
8
12
1 6
Carbon Number
Fig. 5.19 Semilog plot of the hydrocarbon distributions for (a) Co(C0)jNO-NaY zeolite catalq\t (a) after 6 h, (b) after 26 h, ( c ) after 47 h under synthesis gas (b) cobalt-on-kieselguhr catalyst (c) as (b) except for the presence of an equal weight of NaY zeolite pellets ( a ) after 1 h, ( h ) after 25 h (from ref. 167)
194
product molecules, ( 2 ) incorporating metal and zeolitic functions in a single bifunctional catalyst to convert intermediates to desired products, and (3) using a zeolite in a separate stage downstream from the FT reactor for product modification. The first approach is addressed in this section while the latter two approaches are addressed in the next section. There are several recent papers (refs. 149,150,153,158- 164) reporting preparation of Co inside zeolites of low acidity, for which non-ASF behavior favoring C$, hydrocarbons was apparently observed. Similar results were reported for a Ru/NaY zeolite (ref. 165) and Ru/microporous-silica (ref. 166). Several authors of these papers have attributed this behavior to shape-selectivity, since under similar conditions and run times, product distributions obtained with conventionaI cobalt catalysts were in accordance with ASF statistics. Nevertheless, the origin of these deviations from ASF kinetics has been a subject of considerable controversy (refs. 7,165,167), with possible explanations including shape-selectivity, metal dispersion effects, and experimental artifacts such as liquid product holdup in the support (refs. 7,167). Moreover, the observation of these deviations is not universal, as Zwart and Vink (ref. 152) observed ASF statistics for Fern-zeolite prepared from iron carbonyl. A recent definitive study by Ungar and Baird (ref. 167), may lay this controversy to rest.
Their results shown in Fig. 5.19 indicate that the apparent deviation from ASF kinetics on Co/NaY zeolite prepared from Co carbonyl is probably due to selective adsorption of the heavier products by the zeolite support and is only observed during the first 25-30 hours of reaction, disappearing after 50 h of reaction as sufficient product is adsorbed (see Fig. 5.19 (A)). Very similar selectivity behavior is observed after mixing an otherwise normal, ASF-observant Cokieselguhr catalyst with Nay-zeolite (Fig. 5.19 (B) and 19 (C)). Accordingly, there is no definitive evidence at present to support the notion of lasting shape-selective behavior in FTS on metal/zeolites. Moreover, the necessity of conducting the analysis of products during FTS after long periods of time (e.g.>50 h in the case of small samples of zeolites) to ensure a steady-state product distribution is emphasized.
5.2.3 INTERCEPTION OF INTERMEDIATES There are two general approaches to the interception of intermediates involving either (1) a multifunctional catalyst, typically an FT catalyst supported on a zeolite or physically mixed with a zeolite, or (2) a multistep process, typically an FTS or methanol synthesis step followed by an upgrading step involving an oxide or zeolitic catalyst. In either case the zeolite (or oxide) functions to crack or hydrocrack heavy paraffins to lighter hydrocarbons and convert a-olefins, the intermediates in chain propagation, to aromatics, branched and internal olefins (refs. 158,159,168). 5.2.3.1 Intercention i n Multifunctional Catalvsts Reported studies of multifunctional, zeolitic FT catalysts in the literature are numerous. This review will focus first on studies that illustrate selectivity/structure relationships and design principles and second on recent developments of commercial potential.
195
5.2.3.2 SelectivitvlStructure Relationshins and Desipn Princiales In a series of papers and patents (refs. 168-170) workers from Mobil Oil have described a new class of synthesis catalysts comprising a Co, Fe, or Ru FT catalyst with a ZSM-5 class zeolite to give high yields of aromatics, olefins and/or branched hydrocarbons. For example, the gasoline selectivity of fused F e K mixed with an excess volume of zeolite is over 60% of the total hydrocarbon product compared to a maximum of 40% for the Fe/K alone (ref. 168); this liquid product obtained at 593-603K, 12atm and H2/CO=2 contains negligible fractions of C13+ hydrocarbons and high fractions of branched and unsaturated hydrocarbons, e.g. isopentane and 2-methyl-2-butene in the C, fraction as well as 20-30 wt% aromatics depending upon composition. A mixture of ZSM-5 with Ru/alumina (ref. 169) produces 34 wt% aromatics at a CO conversion of 94-98%, 332 K, 51 atm, GHSV480 and H2/CO=2, while Ru/alumina produces none; addition of ZSM-5 also decreases the boiling range of 90% of the C,, overhead from 595 to 477 K. On the other hand, cobalt supported on ZSM-5 (ref. 170) produces a predominantly olefinic product at a CO conversion of 24-37%, 543 K, 13.6 atm, and H2/CO=1, of which a significant fraction is internalbranched olefins. The production of aromatics and internal olefins on these catalysts was attributed to secondary reactions on the zeolite of a-olefins, which were hypothesized to be primary intermediates in FTS (ref. 168). An interesting study of the effects of acidity on the selectivities of Fe/zeolite catalysts
Table 5.14 Berty reactor results for FTS on Fe/ZSM-5 and Fe/silicate catalysts (Process conditions: H2/CO=2, P=20.4 atm, T=553 K) (from ref. 171) Fe/Silicate (13.6% Fe) .~
39.4 16.5 Product Composition, %
co2
H2O CH, + Oxygenates
____~_~____ ~
-~ ~
CH, + Oxygenate Comp. CHn C2&
51.1 22.1 26.9
35.4 39.3 25.3 31.8
0.0, 15.3
(22%
0.0, 8.7 0.0,6.6 37.6
c3%
C3H8 C4H8, C4HI0
C5+ & Oxygenates ~
Liquid Product Comp. % Aromatics Olefins Saturates Oxygenates 8 Gasoline (BP <477 K) Research Octane No. Si02/A1203 = 29 (ZSM-5); 860 (Silicate)
26.5 2.5, 10.8, 4.8, 2.5.0
~~~
33 34 31 2 92 88.3
4 41 33 22 77 36.0
15.0
7.2 4.8
196
conducted by Rao and Gormley (ref. 171) indicates that the product selectivities of Fe on ZSM-5, a relatively high aluminum-containing, high-acidity zeolite, and on silicalite, a zeolite of relatively low acidity and aluminum content, are markedly different, although their zeolitic structures are very similar. Indeed, the data in Table5.14 show that Fe/ZSM-5 produces 33 wt% aromatics, Fe/silicalite only 4%; on the other hand the olefin and oxygenate selectivities for Fehilicalite are significantly higher, 1.e. 41 and 22% compared to 34 and 2%. The higher octane rating of the Fe/ZSM-5 product can be attributed mainly to aromatics present in the liquid, although branched products also contribute. The authors attributed the greater formation of aromatics and branched products in the case of Fe/ZSM-5 to interception of intermediates by acidic sites and the lack of products higher than C, to shape selectivity. However, as in the studies discussed in the previous section, no proof was provided that this is an effect of shape selectivity: the results could be just as well explained by cracking or hydrocracking of the C12+products. This latter supposition is supported by results showing similar selectivities obtained by simply physically mixing zeolites with FT catalysts. For example, the effect of physically mixing silicalite with Co/ThO2/alumina (ref. 50), as shown in Table 5.4, is to very substantially reduce Cg, hydrocarbon production while substantially increasing selectivities to methane and light C,-C, hydrocarbons; as observed by Rao and Gormley (ref. 171), the olefin fraction of the hydrocarbon product was increased by the addition of silicalite. One drawback to zeolite FT catalysts or catalyst mixtures is their relatively high selectivities for methane and C2 hydrocarbons, e.g. 29.33% C,-C2 for Ru/ZSM-5 (ref. 169) and 18.48% CH, in the case of Co/NaY-zeolite (ref. 150). This is undoubtedly due to cracking or hydrocracking of heavier hydrocarbons on the acidic sites of the zeolite. Four possible solutions to this problem are (1) incorporation of basic components in the bifunctional catalyst or zeolite to lower acidity (refs. 50,170,173); (2) incorporation of metals or additives such as Cu or S which poison methane production: (3) addition of olefins to the feed mixture which react to heavier products by oligomerization (refs. 5,40); and (4) changing operating conditions to favor heavier products, i.e. to operation at low H2/C0 ratios, higher pressures, and lower temperatures (ref. 169). An example of the first method is evident in the data from Gulf and Mobil patents (refs. 50,170) showing that addition of 1% Tho, lowers methane production of Co/ZSM-5 to 13.18% and Co/A120j/silicalite to 11%. Goodwin et al. (refs. 172,173) have also shown that the nature of the neutralizing cation and the Si/Al ratio in the zeolite also affect methane selectivity (see more detailed discussion below). The second principle is illustrated by the work of Rao et al. (ref. 174) showing that addition of CU to Co/Si02-ZSM-5 lowered methane and increased Cg+ yield from 72 to 76%. Melson and Zuckerman (ref. 175) have shown that methane selectivity is lower for a mixture of Ru/A1203 and ZSM-5 relative to Ru supported directly on ZSM-5 (24 vs. 44 wt%), although the production of aromatics is not significantly different; the difference in methane selectivity they attribute to the relative locations of the ruthenium sites and the zeolitic functions, their association being most intimate for the bifunctional Ru/ZSM-5 catalyst.
197
Table 5.15 Catalytic properties of Ru/Zeolites for CO hydrogenationa (adapted from ref. 1 7 3 ~ ) Catalyst
1 Si/AI
TOFb
Selectivity (wt7o'o)
iCJc
I
5.8 6.9
I
I
17
12
I
0.0
25
20 1.8 12 27
7.8 I
I
a Reaction conditions: 523 K, 1 atm, H2/CO=I, 5 min of reaction.
Turnover frequency for CO conversion based on H, chemisorption results. Isobutane in C, fraction. d M, mordenite. C
Studies by Goodwin and coworkers (refs. 172,173) provide strong evidence for the bifunctional nature of zeolite-supported FT Ru catalysts and new insights into the role of exchanged cations and acidity in determining hydrocarbon selectivity. In a study
WETAL
FUNCTIOy
lEDLlTE
CIS-
investigating the effects of
2-
the nature of the of the
~
:
neutralizing alkali cations on
co
hydrogenation
activity/selectivity Ruff-zeolites (refs. 173a,173b) it
of was
determined
that
the
concentration
and
acid
FUNCTION
; n-C-c-C-R An A l ti ;~ I
OLEFIN
0
L___-______l
TRfiNS-
2-
CO + H2
I
OLEF IN
YEAK A C I D S I T E S
strength of protonic sites significantly influenced acid-catalyzed reactions such as isomerization, oligomerization, cracking and hydrogen transfer; for instance the olefin/paraffin ratios were highest for the largest alkali ions exchanged into the zeolite, while isoparaffin selectivity was enhanced by smaller alkali ions.
I :
t HYDROGEN TRANSFER
1 Fig. 5.20 Reaction scheme for CO hydrogenation zeolite-supported metal catalysts (from ref. 17%)
on
198
A study (ref. 173c) of FTS on Ru supported on NaX, Nay, KL, and Na-mordenite zeolites covering a wide range of acidity and Si/Al ratio (1.4-5.5) reveals interesting trends in methane, olefin, and branched hydrocarbon selectivities with strength of acidity. Activity and selectivity data in Table 5.15 for fresh 3.0-3.8 % Ru/zeolite catalysts prepared by ion exchange indicate that (1) activity is roughly constant for this group of catalysts and (2) methane and isobutane selectivities increase while propylenelpropane ratio decreases with increasing Si/AI ratio, i.e. with increasing strength of acidity (the number of acid sites was constant for these catalysts). The trend of increasing methane selectivity with increasing Si/A1 ratio was also observed in studies of similar Rdzeolite catalysts of different preparations by the same group (ref. 172). It was suggested (ref. 173c) that the larger amounts of methane and branched hydrocarbons are the result of increasing cracking and isomerization activities of the more acidic zeolites, while the lower olefin/paraffin contents are a result of greater hydrogen transfer over the more acidic zeolites. It was further suggested that since the specific activity of the catalysts for CO hydrogenation was constant, i.e. not affected by the acidity of the zeolite or in any other way by the support, that the results are not due to metal support interactions as some workers had supposed but rather are due to reactions involving the acidic functions of the zeolite as shown in Fig. 5.20.
5.2.3.3 Recent Develooments in Zeolite Catalvst Technolopy In addition to the metal/ZSM-5 catalysts at Mobil (refs. 168-170), there have been a number of other recent promising developments in FT catalyst technology based on zeolites. Most of these make use of the zeolite functions to change hydrocarbon selectivities in favor of premium fuels or chemicals, although some of these developments are directed at more stable catalysts that produce less methane. For example, Miller and coworkers (Union Carbide)(ref. 176) have developed a new generation of zeolite-based multifunctional cobalt catalysts of high activity and stability with low selectivity to methane (6-7 wt%) and high selectivity to gasoline and diesel fuels (50-60%).These catalysts contain proprietary promoters other than T h o z (early formulations contained T h o 9 and proprietary zeolites; they are particularly stable at low H2/C0 ratios over a period of several hundred hours. They are capable of operating at conversions as high as 70% at a GHSV of 300 h-l and a temperature of 513 K; in other words they have activities comparable to optimized Sasol I Arge catalysts. It should be noted that the products obtained with this class of catalysts are consistent with ASF statistics for carbon numbers 1 through 40. Some other recent developments in zeolitic FT catalyst technology include: (1) a Pt-Ru/HY-zeolite catalyst with high selectivity for branched hydrocarbons (ref. 177); (2) a catalyst composite comprised of an Fe FT catalyst and a zeolite in the volume ratio of 1 to 1/20, the zeolite being a hydrogen-exchanged erionite, mordenite, faujasite or ZSM-4 which contains no more than 0.5 wt% alkali, which is selective for aromatics and internal olefins (ref. 178); ( 3 ) a catalyst composite of an FT catalyst and a silicoaluminophosphate molecular sieve catalyst which selectively produces hydrocarbons in the gasoline-diesel boiling range (ref. 179), and (4) a Co catalyst supported on a steam-stabilized, hydrophobic Y-zeolite which has enhanced stability and
199
relatively high selectivity for liquid motor fuels with relatively minor production of hydrocarbons beyond the diesel range (ref. 180). Also a method for reducing methane production in FTS by addition of olefins to the feed has been claimed (ref. 181).
3.2.3.4 Intercevtion in Two-step Processes This section focuses on catalyst developments for upgrading FT products in a second stage. Large-scale, two-stage processes will be discussed in Section 5.3.2 dealing with second generation FT processes. Most of the methods for upgrading FT products reported in the literature involve passing the FT product or fractions of the I T product over zeolite catalysts at relatively higher temperatures than used in FT synthesis. For example, a series of patents assigned to Mobil Corp. (refs. 182-184) describe passing various FT fractions over a crystalline aluminosilicate zeolite catalyst with a pore size greater than 0.5 nm, a silicdalumina ratio of at least 12 and a constraint index of 1-12 under a range of temperature/pressure conditions in order to produce various premium products at high yields including ethylene, light fuel gas, high octane gasoline (research octane number as high as 93), diesel fuel, jet fuel, and light fuel oil. In another series of patents assigned to Mobil (ref. 18%
Kuo describes the use of crystalline zeolites such as ZSMJ for producing a gasoline product of improved octane rating. Similarly Stein et al. (ref. 186) of Mobil describe upgrading FT fractions with a hydrogenation step followed by reaction on a crystalline zeolite selective for producing a gasoline product of improved octane rating. In some cases the upgraded product is further separated and fractions are passed to catalytic alkylation (ref. 185) and hydrodewaxing (ref. 186). Other recent developments in upgrading I T products include: (1) a process for hydrocracking high-boiling hydrocarbons at a low hydrogen partial pressure over a catalyst including a Group VIB or Group VIII metal supported o n a refractory cracking base to produce low boiling hydrocarbons, paraffins and isoparaffins (ref. 187), (2) a process for producing an aromatic gasoline from a light FT fraction and a light fuel oil from a heavy FT fraction by passing either over an Fe-Ga silicate (ref. 188), and (3) a hydrocracking process for upgrading a heavy FT fraction over either a IJ-zeolite composited with a hydrogenation metal or a metal- containing, large-pore zeolite having a silica/alumina ratio of 10 (e.g. ZSM-20 or dealuminized Y-zeolite to obtain a middle distillate and a lubricant base stock (ref. 189).
5.3 NEW DEVELOPMENTS IN REACTOR AND PROCESS DESIGN 5.3.1 RECENT DEVELOPMENTS IN REACTOR DESIGN 5.3.1.I Reactor TvDes and Their Characteristics It was mentioned at the beginning that FT technology suffers from limitations in regard to (1) product selectivity, (2) high capital cost, ( 3 ) heat removal, (4) thermal efficiency, and ( 5 )
200
catalyst deactivation. The extent to which a given FT process is affected by these limitations depends on catalyst, reactor, and process design. The choice of reactor depends upon a number of important attributes which include (1) size/throughput, (2) capital and operating costs,
(3) thermal efficiency, (4) heat (5) product selectivity, (6)
removal,
flexibility in terms of operating conditions
Tube bundl
and product quality, (7) maintenance of catalyst activity/ease of regeneration, and (8) reactor
ideality/stability.
Comparison
of
these reactor atmbutes with the technology limitations
mentioned
in
the
previous
paragraph leads to the conclusions that the Fig. 5.21 Fixed-bed Arge reactor at Sasol (from ref. 190))
success of FT processes is very highly dependent upon reactor design. There are basically three reactor types used for FTS: (1) fixed bed, (2) fluid bed,
and (3) slurry bed. Because of the high exothermicity of the reaction (34 kcalheacted carbon atom), all three reactor types are carefully designed for rapid heat removal using a combination of various techniques including heat exchange, recycle, fluidized and slurry beds, and staged systems (ref. 6b). The attributes, advantages and limitations of each of these reactors is different as summarized in Tables 5.16 and 5.17 and as discussed below.
In a typical fixed bed, for example, heat is removed by heat exchangers; indeed, each Sasol Arge fixed bed reactor (refs. 8,190) operates as a heat exchanger containing 2,050 tubes of 55 mm i.d. and 12 m in length eaCtDr
in which the catalyst is packed (see Fig. 5.21): heat is removed by producing steam. To facilitate temperature control while maximizing conversion and linear
mixture
Fig. 5.22 Synthol circulating fluid bed reactor at Sasol (from ref. 190)
gas velocity, a portion of the tail gas is recycled, typically with a recycle/fresh volume ratio of about 2. Since the heat exchanger design favors lower
201
Table 5.16 Comparison of operating characteristics of Sasol fixed bed, circulating fluid bed, and Rheinpreussen-Koppers Slurry bed reactors (from refs. 8,190,191)
r Characteristics
Operating Conditions Pressure, atm Temp, K H2/C0 ratio Recycle/feed ratio CO+H2 conversion Catalyst Composition Catalyst loading Nm3(CO+H2), m3cat h Performance, t/C3Jt cat day Catalyst life, (month) Reactor Catalyst vol., m3 Length, m Production, t C3Jm3 day Product quality CH,
Synthol Entrained Fluid Bed
RheinpreussenKoppers Slurry Bed (Three Phase)
23-25 49 3- 523 1.25-2.0 2-2.5 50
20-23 573-623 2.4-2.8 2-2.4 77-85
2 l(24) 533-573 0.67
100 Fe, 5 Cu, 5 K20, 25 SiO2
fused Fe/K
Fe/Cu/K
500-700 1.3.5
700 1.85
5000( l0,OOO) 5.3 (10.6)
9-12
1.3a
1.25
2.1
0.93
10 3.3 40 7 4
3 31 54 10 2
6 76
82
2
c2-c4
Gasoline, C5-Cl Diesel, CI2-kl8 Heavy oil C19-C35 L L
a
ARGE Fixed Bed
10.8 18 14 27 25 3.2 53
0 90
Mean catalyst life; catalyst is continuously added and withdrawn.
temperatures of operation (493-523 K) (ref. 190), predominantly higher hydrocarbons including gasoline, diesel fuel and waxes are produced. Typical syngas conversion is SO%, although under modified conditions conversions as high as 73% are possible (ref. 191). Fluid beds are of generally two types, fixed and circulating. The Synthol circulating, entrained (transported) fluid bed system at Sasol (Fig. 5.22) facilitates better heat removal; more isothermal operation; higher reaction temperature (exit temperature of 613 K) and hence higher selectivities for lighter products, olefins, branched products, and aromatics; and higher throughput per volume of reactor relative to the fixed bed (Table 5.16) (refs. 190,191). A combination of recycle (recycle ratio of 2) and coolers with recirculating oil (watedsteam in Sasol 11) is used for heat removal and temperature control. With on-line catalyst removal and addition process runs are much longer than for the fixed bed. Nevertheless, the Synthol fluid bed reactor is a more complex
202
system requiring an attrition-resistant catalyst and greater system maintenance. Moreover, process conditions and temperature must be adjusted to limit production of heavy hydrocarbons which would condense on the catalyst and defluidize the bed. Sasol has also investigated fixed fluidized bed reactors in which the gas is expanded but not entrained or transported finding them to give higher conversion than the circulating Synthol reactors (ref. 190). Because of their promise, Sasol conducted pilot and demonstration tests over a period of years recently commissioned a high pressure full-scale commercial unit in 1983-89 (ref. 190). Slurry bed reactors, in which a finely divided catalyst is suspended in a heavy oil (the !iT product itself) by gas bubbling up through the slurry, have been investigated fairly extensively for
FTS by Koelbel, Sasol, and Rheinpreussen-Koppers (refs. 5,9,15,192). This reactor type has the advantages of (1) capability of operation at low H2/C0 ratios without problems with carbon deposition, (2) very efficient heat transfer and uniform temperature, (3) high catalyst efficiency/performance (see Table 5.16), and (4) simple construction. Sasol tests indicate that under the same operating conditions, product selectivities obtained in the slurry reactor are very similar to those in either fluid nor fixed bed; moreover, conversions are about the same in the slurry and fixed bed reactors, although in the high-temperature, gasoline-producing mode, the fluidized bed was reported to have a higher conversion than the slurry reactor (refs. 8,190). However, the slurry reactor is probably best operated under conditions of light hydrocarbon production, since the light products can be recovered by condensation, while if predominantly heavier hydrocarbons are produced, the product must be withdrawn from the reactor, requiring separation and return of the catalyst. A modification of the slurry reactor, referred to as an ebullating bed or three phase slurry reactor, involves suspension of relatively larger catalyst particles by gas bubbles; accordingly this is a true three-phase system. The Rheinpreussen-Koppers three-phase slurry bubble column (ref. 192) and the more recently developed Chem Systems ebullating bed reactor (ref. 9) are examples of this reactor type. 5.3.1.2 ComDarison o f Attributes for Three Reactor Tvoes Based on the above reactor descriptions (Section 5.3.1.1 and Table 5.16), previous operating experience (refs. 8,9,190-192), the author's own experience in reactor design, and a UOP evaluation of advanced FT reactors (ref. 9), a comparison of the attributes, advantages, and disadvantages of three reactor types, fixed-, fluid-, and slurry-bed, is made for FTS in Table 5.17. This comparison indicates that the fluid bed leads in system throughput per unit volume of reactor while both fluid and slurry bed reactors are highly selective far high Octane gasoline products. These factors undoubtedly influenced the choice of the entrained fluidized bed reactor for the new commercial Sasol I1 and I11 plants in South Africa (refs. 8,15,190-191). The fixed bed reactor also has a high throughput, the best operational flexibility, and is the most selective for production of heavy
hydrocarbons including waxes. The slurry bed, while having significantly lower system throughput per unit reactor volume, involves significantly lower capital and operating costs, a higher thermal
203
Table 5.17 Evaluation of three common FT reactor systems based on key attributes and limitations AttributeLimitations 1.Relative thruput/ vol. reactoF
Fluid Bed
Slurry Bed
100
44
100
46
6.72
0.76
66
91
fair
good
excellent
heavy products gasoline
high octane and gasoline
light olefins
poor
fair to good
Fixed Bed
I
89
2a Relative capital costb b Operating costb (Catalyst cost in $million/year for 25,000 bbVday
3. Thermal efficiency,%b 4. Heat removal/ temp control
I
I I
5. Product selectivity favored 6. Operational flexibility a Operating conditions
b Product quality
good high T, high H2/C0 narrow velocity range good
low H$O narrow velocity range fair (light products
7a Activity maintenance b Ease of regeneration
good poor
fair very good
8a Reactor ideality b Stability i) flow ii) temp
good
fair
good fair
fair good
a
Based on comparisons of reactor production in Table 5.16. Based on UOP evaluation of advanced reactors (from ref. 9)
efficiency, and a capacity to operate at lower H2/CO ratio with less activity loss due to carbon formation. Overall, the slurry reactor appears to be the most efficient, economical system for production of light olefins and gasoline and the best match for syngas of low H2/CO ratio from advanced gasifiers.
% 5 Because of its promising operating characteristics and attributes for FTS (Tables 5.16 and 5.171, the slurry reactor has been the subject of several recent studies. These have included investigations of slurry bed performance relative to a fixed bed (ref. 193), slurry catalyst performance (refs. 194-199) including development of a carbonyl-derived Co/silica catalyst superior to that prepared from the nitrate (ref. 197) and ultrafine Fe and Fe-Co-Ni catalyst having greater activity than precipitated iron catalysts (ref. 199), heat transfer properties of the slurry
204
(refs. 200,201), effects of mass transport (refs. 202,203), hydrodynamics of three phase slurry reactor systems (refs. 204-209), effects of liquid composition on reaction rate and selectivity (ref. 210), and models of slurry reactor (refs. 200,201,205,208). These studies provide new, fundamental understanding of principles that can be used in optimizing the performance of slurry reactors. For example, the study of Kuo et al. (ref. 205) provides data on yields, product quality, reactor design, deactivation rates, wax analysis and process variables for long term, high-wax operation; it demonstrates that a slurry reactor can be operated at high reaction rates (H,+CO conversions of 70.90% at space velocities of 2.4-3.8 NL/g Fe-h) using a Fe/Cu/K catalyst in the high-wax mode on a long term (35 days) basis at low methane and ethane selectivities and with negligible catalyst loss. Unfortunately, after 35 days of operation the catalyst particles had doubled in size due to accumulation of deposits and settled terminating the run; the need for further investigation of this settling problem is clear. Other reactor types have received less attention. However, two recent investigations of fluidized beds for FTS illustrate possibilities for fresh, creative approaches to the reactor design problem. Whiting, Liu, and Squires (ref. 21 11, for example, investigated FTS using hydrogen-poor synthesis gas in a novel, two-zone dry fluidized bed designed for minimizing carbon formation and which involved unsteady-state operation with alternate exposure to hydrogen-poor synthesis gas for short periods and hydrogen-rich gas for long periods. Their results provide insights into the design and operation of such an unsteady-state reactor and optimum conditions for minimizing the carbon deposition. Silverman and coworkers (ref. 212) investigated dense-phase (fixed fluid) beds for FTS to determine the effects of catalyst properties and reactor scale-up on fluidization; their results indicate that the use of a fine powder would avoid problems previously encountered in the early,
unsuccessful use of these systems in Brownsville, Texas in the 1950's (ref. 213).
5.3.2 SECOND GENERATION FT PROCESSES 5.3.2.I Second Generation Commercial Processes The first commercial FT process based on a Co/ThO,/kieselguhr catalyst was employed in Germany during World War I1 to produce gasoline, reaching 10,000 bbl/day in 1944; these plants were discontinued after the war as less costly petroleum become available (ref. 15). The next large scale plant to be operated for a significant period of time was 8,000 bbl/day Sasol I Plant installed in 1955 in South Africa and which still operates today; this plant uses both Arge fixed bed and Synthol entrained fluid bed reactors. As a result of the unique political and economic circumstances in South Africa the Sasol plant was recently expanded; Sasol I1 and 111, each producing 50,000 bbl/day were installed in 1980 and 1983 respectively. These newer plants utilize only the large, high-capacity Synthol entrained fluid bed reactors with an alkalized iron catalyst (ref. 190). A flow schematic of the Sasol I1 plant is shown in Fig. 5.23 (ref. 190a). As in Sasol I reaction water and liquid oils are condensed out but unlike Sasol I, the product gas in Sasol I1 is next passed through a Benfield scrubber to remove CO, after which it undergoes cryogenic separation of CH4-rich, Hz-rich, C2, and C3-C, components, making possible recovery of
205
high-priced
ethane
and
PURE SYNTHESIS GAS
ethylene. As in Sasol I, the CH4-nch stream is sent to the reformer
where
CH4
is
converted to synthesis gas and recycled
to
the
reactors. The H,-rich stream undergoes purification in a pressure
swing
SYNTHOL
Synthol I
I
1 .
t
ALCOHOL
I
KETONE
I
GASIFICATION WATER
I
absorber
(PSA) to pure H2 which is used in hydrotreating units.
C3-C4
hydrocarbons
oligonierized phosphoric
over
are a
acid-kieselguhr
catalyst, the product of which is combined with the isomerized C$,
stream to
blend with gasoline as in Sasol 1. However, in Sasol 11 diesel
fuel can be maximized to a selectivity as high as 7.5% by recycling a portion gasoline and hydrocracking waxes to diesel in the hydrodewax unit. Thus, there is considerably more flexibility in the selectivities for gasoline and diesel fuel in Sasol 11; indeed, the ratio of gasoline/diesel can be varied from 1/1 to 10/1. Because the gasoline from
Sasol II is a stand-alone product rather than a blending agent, the C7-463 K fraction is hydrofined to saturate olefins and remove oxygenates following which it is reformed over a Pt/alumina catalyst. Other commercial FT-like processes include the methanol to gasoline (MTG) process developed by Mobil (refs. 15,16) and the Topsoe Integrated Gasoline Synthesis (TIGAS), both two-step processes for producing high octane gasoline at high selectivity from natural gas or syngas. In the MTG process, syngas or steam-reformed methane is converted to methanol which, in a second step, is converted nearly quantitatively over a ZSM-5 zeolite catalyst to gasoline. In the TIGAS process (ref. 21 4) syngas or steam-refoniied/autotheniial-refornled methane reacts over a multi-functional catalyst to oxygenates which are then passed through a gasoline synthesis unit, followed by separation of LPG, water and gasoline. The details of these processes are described elsewhere (refs. 15,16,214).
206
5.3.2.2 ExuerimentallConceutual Multi-Stape Processes Although the FTS cannot achieve in one step greater than 40-50% selectivity for a given premium product such as light olefins, gasoline, or diesel fuel (see Sections 5.1 and 5.2. l), it is now possible through a combination of two or three reactiodseparation stages to achieve selectivities on the order of 70-90% for almost any desired product. Accordingly, a number of highly selective "two-stage'' or multistage FT processes or combination alcohol synthesis/FTS processes have been developed in the past decade. This section will focus on examples of the former kind of process. A multistage process for producing diesel at 77% selectivity has been designed and tested at the pilot plant scale at Sam1 (ref. 190a) (see Fig. 5.24); it features low temperature FC3 in a fixed or slurry bed to produce diesel and waxes in high yield the latter of which are then catalytically hydrocracked under mild conditions to primarily diesel fuel. Production of diesel fuel is further augmented by oligomerization of C,-C, hydrocarbons. According to Dry (ref. 190a) the final diesel product has an acceptably high cetane number of 65 and meets all other diesel fuel specifications. On the other hand, if high-temperature fluidized bed reactors are substituted in this scheme the diesel fuel yield is only 57%; moreover, because 54% of this diesel fuel is a product of olefin oligomerization using the fluid bed (versus 7% using the fixed bed), it is necessary to use special zeolites rather than phosphoric acid in the GiSlFlCATlON oligomerization step to maximize the quality of the fuel by minimizing branching. Even then, the quality (cetane number of PURIFICATION 55) is still lower than with the fixed bed. Gaensslen and
I 1
. Eisenlohr
(ref. 215)
conceptualized
L
A LECTIVE
I1
CH'
i3*m
c2
12%1
c3 - C' l6%l
I
NAPHTHA
no.&i
DIESEL 177%1
Fig. 5.24 Flow scheme for high diesel fuel producing plant (from ref. 190a)
a
have
similar
process for selective diesel fuel production. It is based on the Arge fixed bed synthesis and features: (1) oligomerization of the C3-C4 alkenes to gasoline and diesel, hydrogenation of the Cg-CI8 diesel fraction, hydrocracking of the waxes to diesel fuel, and conversion of the alcohols and ketones
207
: 1 tops/naptha
/
kerosene gas-oil
0 0.75
I-
0.80
0.85
0.90
0.95
' -
probability of chain growth, a classical catalyst _new _ - _catalyst _ _ _ _ _ - _ _ - - - - - - Shelldevelopment catalyst
Fig. 5.25 Product distribution for FT synthesis as a function of alpha indicating operating region for Shell Catalyst (from ref. 217)
on a Mobil zeolite to gasoline. A two-stage slurry FTSBSM-5 upgrading process for converting hydrogen-poor synthesis gas to high quality gasoline was evaluated by Kno et al. at Mobil (refs. 205,216). Feasibility for long term (35 day) operation was demonstrated using a bench-scale unit consisting of a slurry bubble column reactor followed by a fixed-bed ZSM-5 reactor. Three Fe/Cu/K FT catalysts were evaluated some of which produced 400-800 g/g Fe, a level higher than previously reported. Gasoline yields for the 2-stage system were as high 87 wt%. A conceptual process design including a scoping cost estimate for a 27,000 bbl/day plant was conducted, yielding an 1983 capital cost of $700 million. The Shell Middle Distillate Synthesis (SMDS) process (refs. 217,218) is likewise a two-stage high-wax/crack process but uses different reactor and catalyst types to produce a heavier product, a middle distillate containing jet fuel kerosine and diesel gas oil, from a hydrogen-rich syngas produced from natural gas. The first stage is a fixed bed FTS reactor containing a silica- or alumina-supported Co catalyst promoted with Zr, Ti, or Cr oxide with a propagation probability at a Hz/CO ratio of 2 in the range of 0.8 to 0.95 and thus is highly selective for heavy liquids and waxes (see Fig. 5.25). In the second stage the heavy wax fraction is hydroisomerized and hydrocracked in a trickle bed reactor similar to that used in gas oil HDS to give a maximum yield of middle distillates containing a gas oil with a cetane number exceeding 70 and a kerosine ideal for use in jet fuels. The first commercial SMDS plant is being constructed in Sawarak, Malaysia and will come on stream in late 1992 (ref 217c)
208
Patents issued during the past 10 years describing two-stage or multistage FT processes are too numerous to describe here in detail. They include (1) two or three stage processes developed at Mobil for Fe (or otherwise)-catalyzed FTS/zeolite-upgrading to gasoline and distillates (ref. 2 19) and middle distillates (ref. 220) as well as upgrading of FT waxes by catalytic cracking followed by oligonierization to distillate range hydrocarbons (ref. 221); (2) a Shell two-stage process (ref. 222) utilizing an Fe/Cu/K/MgO type catalyst in the first stage to handle a hydrogen-poor syngas followed by a Co/Zr/Si02 type catalyst in the second stage which operates at a H2/CO=I.S, the overall system having enhanced stability relative to a single stage process; (3) a two stage process developed by Shell (ref. 223) for producing a high-viscosity lubricating oil by first FTS on a Zr-, Ti-, or Cr-promoted Co catalyst to produce heavy paraffins and second treating the paraffins at elevated temperature with an organic peroxide, (4) a multistage FT process developed at Exxon (ref. 224) for producing waxes by contacting syngas first with a promoted iron catalyst with high olefin selectivity followed by a second bed containing a Ru catalyst having high selectivity for converting olefins to heavy paraffins, and ( 5 ) a two stage process developed at Ammo (ref. 225) for converting C,-C, alkanes to higher molecular weight hydrocarbons using in the first stage a ceramic-containing autothermal reformer with nonuniform temperature gradient to facilitate the formation of syngas from light paraffins and the subsequent cooling of the syngas to FTS reaction temperature before it enters the second FTS stage.
5.4 CONCLUSIONS AND RECOMMENDATIONS 5.4.1 ASSESSMENT OF CURRENT TECHNOLOGY AND CONCLUSIONS This chapter has focused on representative Fischer-Tropsch catalyst, reactor and process developments of the past decade with an emphasis on design principles. It is evident from the work reviewed that significant progress has been made in both understanding and the application of this understanding to the development of more efficient catalyst and process technologies. Some of the important design principles discussed in this chapter include the following: 1)
2)
FT technology suffers from limitations in regard to (1) product selectivity, (2) high capital cost, (3) heat removal, (4) thermal efficiency, and ( 5 ) catalyst deactivation. The extent to which a given FT process is affected by these limitations depends on catalyst, reactor, and process design. Hydrocarbon selectivity in steady-state FT synthesis is generally governed by Anderson-Schulz-Flory (ASF) polymerization kinetics which limit the production of premium products by carbon number; accordingly the maximum obtainable weight percentages of C2-C4 hydrocarbons, gasoline, and diesel products are 56, 47 and 40% respectively. However, it is possible to obtain up to 7 0 8 waxes operating at a polymerization probability of 0.95. Attempts to circumvent ASF statistics by means of promoters, interacting supports, high metal dispersions, and shape-selective supports have
209
generally met with failure (the author is not aware of a single study in which this was
3)
conclusively demonstrated). It is possible to achieve the maximum yield of a given hydrocarbon product within the constraints of ASF kinetics, nearly 100% C2-C, olefin yields, and a large fraction of branched products by (1) careful design of the catalyst through promoters, supports, gaseous additives, interstitial compounds, and bimetallics and (2) careful choice of reaction conditions, i.e. temperature, pressure, and H2/C0 ratio. Additives such as promoters, supports, interstitial additives, and alloying metals are thought to change the chemistry of the reaction by changing the localized chemical/electronic structure of the active metals. For example, basic promoters such as K and Mn oxides favor production of olefins probably due to their presence on the metal or carbide surface where they locally promote CO dissociation while inhibiting hydrogen adsorption and thus hydrogenation to paraffins; at the same time they lower methane production by neutralizing acidic sites on the support which would otherwise crack hydrocarbon intermediates to methane.
4)
It is possible to circumvent ASF-predicted hydrocarbon yields through the use of zeolites and other promoters/supports which intercept reaction intermediates and convert them via hydrocracking, isomerization, oligomerization and/or hydrogenation to lighter, branched, heavier, or saturated products. It is also possible to cut off the heavier hydrocarbons of the predicted distribution by transient selective adsorption on the support over short periods of operation; the ASF distribution can also be sharpened by introducing olefins into the feed.
5)
Variations in pretreatment, preparation, and reaction conditions enable almost infinite variations in the composition/structure of Group VIII metal catalysts; this is especially true of iron catalysts, for which bulk compositions can range from carbides to oxides and mixtures in between. These unique compositional/structural properties, when fixed by a unique choice of these preparation/operational variables, determine in turn the catalyst's unique activity and selectivity properties.
6 ) While metal loading, extent of reduction to the metal, extent of carbiding iri the case of iron, and coverage of the surface with promoter or support species are important structural parameters which affect activity and selectivity in FTS, there is strong evidence that variations in surface metal structure have no effect on activity or selectivity; in other 7)
8)
words, CO hydrogenation on Group VIII metals is probably not structure-sensitive. In bifunctional metal/zeolite FT catalysts, there are trends of increasing methane and isobutane selectivities and decreasing olefin/paraffin ratio with increasing zeolite acidity as a result of increasing activity for hydrocracking, isomerization, and hydrogenation reactions on the zeolite. There are basically three reactor types used in FT synthesis: fixed, fluidized, and slurry
beds. The choice of reactor depends upon its (1) throughput, (2) capital and operating Costs, (3) thermal efficiency, (4) heat removal, ( 5 ) product selectivity, (6) operational flexibility, (7) catalyst activity maintenance/ease of regeneration, and (8) reactor ideality/stability. Fluidized beds currently lead in system throughput per unit volume of
210
reactor while both fluidized and slurry bed reactors are highly selective for high octane gasoline products. The fixed bed has a high throughput, the best operational flexibility, and is most selective for diesel and waxy hydrocarbons. The slurry bed, while having the lowest system throughput, involves the lowest capital and operating costs, the highest thermal efficiency and a capacity to operate at low H2/C0 ratio with minimum activity loss due to carbon formation. Overall, the slurry reactor appears to be the most efficient, economical system for production of light olefins and gasoline and the best match for syngas from advanced gasifiers. 9)
Although
FT synthesis cannot achieve in one step greater than 40.50% selectivity for a
given premium product such as light olefins, gasoline, or diesel fuel, it is now possible through a combination of two or three reaction/separation stages to achieve selectivities on the order of 70-90%for almost any desired product. Some of the key recent improvements in FT catalyst, reactor and process technology which could significantly impact the efficiency and economical production of fuels and chemicals from coal and natural gas include: 1) Catalysts Improvements in Fe/Mn catalysts prepared from mixed oxides, spinels and carbonyls and promoted with Ce, for highly-selective production of olefins and low methane production (refs. 43-47,56). Co and Ru catalysts promoted with Zr, Ti or Cr for production of waxes which in turn can be cracked to diesel fuel and olefins (ref. 49). C0/Ti02 catalysts (refs. 57,58) promoted with lanthanides having improved thermal stability towards regeneration and a Co-Ru/Ti02 catalyst (ref. 128) which can be regenerated in situ, both having high activities and selectivities for premium transportation fuels. High activity and olefin-selective Fe203C, catalysts (ref. 99). Co borides of high activity and high sulfur resistance (ref. 110). More sophisticated preparation techniques involving, for example, nonaqueous impregnations (refs. 50a,68,72,137), controlled-pH precipitations (refs. 130-136), and decomposition of metal carbonyls on dehydrated supports to produce highly-dispersed, highly-reduced, contaminant-free supported metals (refs. 21,23,29,80,142- 157). Bifunctional or mixed metal/zeolite catalysts which due to interception of intermediates produce aromatics and gasoline in yields higher than predicted by ASF theory. 2 ) Reactor Developments a) Recent investigations of fixed-fluidized-bed reactors (refs. 190,211) indicating likely improvements in performance over the early Brownsville reactor and the circulating Synthol reactor (ref. 190). b) Recent studies of slurry-bed reactors providing new, fundamental understanding of principles that can be used in optimizing their performance (refs. 193-210).
211
3 ) Process Developments a)
Conceptual and experimentally-tested two-step processes for highly-selective production of diesel fuel (refs. 190a,215).
b)
Successful pilot testing of a Mobil two-stage promoted-Fe-slurry FTS/ZSM-5 upgrading process for converting HTpoor syngas to high quality gasoline (refs. 205,216).
c)
Development and commercialization of the multi-stage Shell Middle Distillate Synthesis process to produce jet fuel kerosine and diesel gas oil from natural gas feeds (refs. 217,218).
Technology options for production of premium hydrocarbons by various combinations of these catalyst, reactor, and process technologies are summarized in Table 5.18.
5.4.2
RECOMMENDATIONS FOR FUTURE RESEARCH AND DEVELOPMENT
Based on the work reviewed in this chapter and on recommendations by Mills (ref. 15) several areas of technology are listed below which if further researched and developed could contribute significantly to more efficient, economical production of synthetic fuels and chemicals from FTS: 1) Catalyst Technology a) b)
Further improvements in catalysts which greatly reduce methane formation. Catalysts with higher resistance to deactivation by sulfur and by carbon deposits and/or coke, particularly in high-wax and slurry reactor processes and which are more easily-regenerable.
c) d) e) f)
Catalysts with higher attrition resistance for application in fixed-fluid-bed reactors. Catalysts of higher activity for slurry-bed processes. Catalysts with improved selectivity and stability for production of waxy hydrocarbons. Continued research on metal/zeolite combinations and second-stage zeolites for upgrading FT products.
2) Reactor Technology a)
b) c)
More quantitative characterization and modeling of the kinetics and mechanisms of essential reaction and transport processes in each stage of the combined slurry phase FTSESM-5 upgrading process. This should be followed by optimization studies. Investigation into the causes and prevention of catalyst settling and stability-loss in slurry reactors. Experimental/modeling studies of the performance and detailed kinetics of fixed-fluid bed FTS reactors.
3) Process Technology In view of the trend to natural gas as a feedstock for the next 3-4 decades, increased emphasis on integrated processes such as the Shell Middle Distillates Synthesis which produces premium products from natural gas via autothermal reforming to a H2-rich syngas is recommended.
212
The long-term potential for substantial improvements in scientific understanding and technology leading to a substantial, cost-effective FTS industry is very promising (ref. 15). Accordingly, it is hoped that the present shortsightedness in government and industry leading to a radical decline in FTS R&D will give way to a more enlightened, long term view and the provision of strong support for FTS and related research and development programs.
Table 5.18 Summary of some technology options for production by FTS of premium products Premium product
Catalysts
c2-c4 olefins
Fe/K, Fe/Mn, Fe/Mn/Ce Fe/K/S, Ru/Ti02,Fe/C Fe2O?CX,Mo/C
Gasoline
fused Fe/K Co/Th02/Al,0 /silicalite Fe/K/ZSM-5, ?o/ZSM-S RdZS M-5 Fe/Cu/K, ZSM-5
I
1 I
I
I
II
I
Reactors
Processes
References
Slurry, Fluid-bed
Syntho1,Koelbel RheinpreussenKoppers,Dow LPG
6,13,15,3 1,3263 90,91,94,99,190- 192
Synthol Gulf-Badger Mobil One-Stage
50,168-17 1 176-180,182186,190,191
Fluid-bed Fixed-bed Slurry/fixed-bed
I I
I
- 1
I
Mobil Two-Stage
Diesel
48-5058,
II
Co-Ru/A1203 Fe/K, Fe/Cu/K,Co/Zr, Ti or Cr/A120.+ Co/Re/A1203 Promoted Fe/Ru
1I
I
Slurry-bed (low T) Fixed-bed (low T)
1
1
Shell-Middle-Distillate Eisenlohr/Gaensslen Mobil (First Stage) Sasol-Arge Shell-Middle-Distillate (First Stage)
II I
190-192,215, 217,218 205,2 15-218, 224
214
ACKNOWLEDGEMENTS The author acknowledges financial support of his Fischer-Tropsch research and of this review from Brigham Young University; Atlantic Richfield Co., the Pittsburg Energy Technology Center, Department of Energy; and the Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences. Review of the manuscript and helpful comments by Professor G. Alex Milles (University of Delaware) and Dr. Gopalakrishnan R. Thevar (BYU) are also gratefully acknowledged.
5.5 REFERENCES 1
P. Sabatier, and J. B. Senderens, C. R. Acad. Sci., Paris, 134 (1902) 514.
2
BASF: German Pat. 293,787 (1913).
3 4
F. Fischer and H. Tropsch, Brennst.-Chem., 4 (1923) 276. M. E. Dry, in Catalysis Science and Technology, J. R. Anderson and M. Boudart, eds.,
Springer-Verlag, New York, 1981, pp. 159-256.
5 6a
R. B. Anderson, The Fischer-Tropsch Synthesis, Academic Press, New York, 1984. J. Falbe, New Syntheses with Carbon Monoxide, Berlin, FR Germany, Springer-Verlag, 1980.
6b
C. D. Frohning, H. Koelbel, M. Raleic, W. Rottig, F. Schnur, and H. Schulz, Chemical Feedstocks from Coal, J. Falbe (Ed.), Wiley, 1982, pp. 309-432.
7a 7b
D. L. King, J. A. Cusumano, and R. L. Garten, Catal. Rev.-Sci. Eng., 23 (1981) 233. Technical Limitations of Fischer-Tropsch Synthesis: Implications for New Syngas-based technologies, Catalytica Associates, Multiclient Study No. 4160, Oct., 1984.
8 9 10
11 12 13 14a 14b 14c
M. E. Dry and J. C. Hoogendoom, Catal. Rev.-Sci. Eng., 23 (1981) 265. J. Haggin, Chem. & Eng. News, (Oct. 26, 1981) 22. H. Klare and M. Pelagalli, Cheni. Techn., 33(X) (1981) 399. A. Deluzarche, R. Kieffer, J. P. Hindennann, G. Jenner, A. Kiennemann, G. Lefebvre, and A. Chauvel, L‘actualite chini., (Dec. 1982) 23. L. Koenig and J. Gaube, Chem.-1ng.-Tech., 5.5 (1983) 14. W. Keim, Am. Chem. Soc, Div. of Petr. Chem. Prepr., 31(1) (1986) 12. M. E. Dry, Am. Chem. SOC,Div. of Fuel Chern. Prepr., 31(2) (1986) 288.
16 17 18 19
W. Haag, J.C. Kou and I. Wender, in DOE Report DEER-0326, DE8800129 (1987). I. Wender and K. Klier, “Coal Liquefaction - A Research and Development Needs Assessment”, DOEER-400, Vol. 11, Ch. 5 (1985). G . A. Mills, Catalysts for Fuels from Syngas, New Directions for Research, IEACWO9, E A Coal Research, London, (1988); U.S. Nat. Tech. Info. Serv. No. IEACR 8901. C. D. Chang, Catal. Rev.-Sci. Eng., 25 (1983) 1. Y. T. Shah and A. J. Perotta, Ind. Eng. Chem. Prod. Res. & Devel., 15 (1976)123. C. H. Bartholomew and R. C. Reuel, Ind. Eng. Chem. Rod. Res. & Devel., 24 (1985) 56. C. H. Bartholomew, Ann. Tech. Rept. to DOE/PETC, DOE-PC- 90533-8, Jan. 25, 1989.
20
C. S. Kellner and A. T. Bell, J. Catal., 70 (1981) 418.
15
215
21
C. H. Bartholornew and M. D. Patil, paper in preparation, 1989.
22
J. B. Benziger and R. J. Madix, Surf. Sci., 94 (1980) 199.
23 24
H. P. Bonze1 and H. J. Krebs, Surf. Sci.,lO9 (1981) L527 . H. Arakawa and A. T. Bell, Ind. Eng. Chem. Process Des. Dev., 22 (1983) 97. D. J. Dwyer, Preprints ACS Div. Petr. Chern. (Aug. 26-31, 1985) 715. J. L. Rankin and C. H. Bartholornew, J. Catal., 100 (1986) 526-532,533-540. C. H. Bartholornew, in Hydrogen Effects in Catalysis, Z. Paal and P. G. Menon, eds., Marcel Dekker, Inc., 1985, Chapter 5 , p. 139. Ibid, Chapter 20, p. 543. G. B. McVicker and M. A.Vannice (Exxon Res. & Engr. Co.), U.S. Patent 4,191,777, Mar.
25 26 27a 27b 28
11, 1980. 29 30 31 32 33 34 35 36a 36b 37 38 39 40 41 42 43 44 45 46 47
G. B. McVicker and M. A. Vannice, J. Catal., 63 (1980) 25-34. R. A. Stowe and C. B. Murchison, Hydrocarbon Processing, (June 1984) 95-100. B. Buessemeier, C. D. Frohning and B. Cornils, Hydrocarbon Roc., 55 (1976) 105. R. Snel, Catal. Rev.-Sci. Eng., 29 (1987) 361-455. J. R. Katzer, A. W. Sleight, F. Gagardo, J. B. Michel, E. F. Gleason, and S. McMillan, Faraday Disc. Chern. SOC.,72 (1981) 121. J. M. Hemnann, J. Catal., 89 (1984) 404-412. J. D. Bracey, and R. Burch, J. Catal., 86 (1984) 384. W. M. H. Sachtler, Proc. 8th Int. Cong. Catl, Berlin, 1984, Vol V, p. 151.
W. M. H. Sachtler, D. F. Shriver, W. B. Hollenberg and A. F. Lang, J. Catal., 92 (1985) 429-43 1. M. A. Vannice and C. Sudhakar, J. Phys. Chem., 88 (1984) 151; J. Catal., 95 (1985) 227. A. G. T. M. Bastein, W. J. van der Boogert, G. van der Lee, H. Luo, 8. Schuller and V. Ponec, Appl. Catal., 29 (1984) 243- 260. M. E. Levin, M. Salmeron, A. T. Bell, and G. A. Sornorjai, J. Catal., 106 (1987) 401-409; Surf. Sci., 195 (1988) 429. A. T. Bell, Proc. 9th Int. Cong. Catal, Calgary, 1988, Vol. 5, p. 134. G. C. Maiti, R. Malessa and M. Baerns, Appl. Catal., 5 (1983) 151-170. N. K. Jaggi, L. H. Schwartz, J. B. Butt, H. Papp, and M. Baerns, Appl. Catal. 13 (1985) 347. J. Venter, M. Kaminisky, G . L. Geoffroy and M. A. Vannice, J. Catal., 103 (1987) 450-465. I. R. Leith, Proc. 9th Int. Cong. Catal, Calgary, 1988, Vol. 2, p. 658. J. Barrault, C. Renar, L. T. Yu, and J. Gal, Proc. 8th Intern. Congress Catal., Berlin, 1984, pp. 101-112. C. Kim, K. Chen, F. V. Hanson, A. G. Oblad, and Y. Tsai, Am. Chern. SOC.,Div. of Pen. Chem., Prepr. 31(1) (1986) 198-207. R. A. Fiato, and L. Soled (Exxon Res. & Engr. Co.), U. S. Patent, 4,621,102, Nov. 4, 1986.
216
48
I. E. Wachs and D. B. Yang (Exxon Res. & Engr. Co.), U. S. Patent, 4,711,871, Dec. 8, 1987.
49
S. T. Sie (Shell Canada Ltd.), Canadian Patent, 1,243,047, Oct. 11, 1988. H. Beuther, C. L. Kibby, T. P. Kobylinski, and R. B. Pannell (Gulf Res. & Dev. Co.), U. S.
50a
Patent 4,399,234, Aug. 16, 1983.
50b
Ibid, U. S. Patent 4,423,064, Nov. 1, 1983.
51
R. Madon, E. Bucker and W. Taylor, “Final Report U. S. Dept of Energy”, FE/8008-1 (1977).
52
H. Schulz, “Fischer-Tropsch CO Hydrogenation”, private communication. B. Viswanathan and R. Gopalakrishnan, J. Catal., 99 (1986) 342. R. Gopalakrishnan and B. Viswanathan, J. Chem. So., Faraday Trans. I, 82 (1986)
53 54
2635-2643.
55 56 57a 57b
57c
B. Sarup and B. W. Wojciechowski, Can. J. Chem. Eng., 62 (1984) 249. R. A. Fiato, R. Bar-Gadda, and S. Miseo (Exxon Res. & Eng. Co.), U. S. Patent 4,719,240, Jan. 12, 1988. C. H. Mauldin, S. M. Davis, and K. B. Arcuri (Exxon Res. & Eng. Co.),U. S. Patent, 4,755,536, July 5, 1988. Ibid, U. S. Patent, 4,762,959, Aug. 9, 1988. Payne and C. H. Mauldin (Exxon Res. & Eng. Co.), U. S. Patents, 4,542,122: 4,595,703; and 4,556,752.
58
N. Takahashi, T. Mori, A. Miyamoto, T. Hattori, and Y. Murakami, Appl. Catal., 38
59 60
C. H. Bartholomew, P. K. Agrawal, and J. R. Katzer, Adv. Catal., 31 (1982) 135. C. H. Bartholomew, in Catalyst Deactivation, B. Delmon and G. F. Froment (Eds.),
61
H. G. Stenger, Jr. and C. N. Satterfield, Ind. & Eng. Chem. Proc. Des. & Develop., 24
62 63
D. K. Matsumoto and C. N. Satterfield, Energy and Fuels, 1 (1987) 203-210. G. T. Tong and J. G. McCarty, Abstr. 10th North Am. Mtg. of the Catalysis So., San Diego, 1987, Paper C42. H. Hagiwara, T. Sano, H. Okado, K. Saito, K. Okabe, T. Hosoya, H. Takaya, and K. Bando, Sekiyu Gakkaishi, 29 (1986) 174-177. H. Fujitsu, N. Ikeyama, H. Ishibashi, and 1. Mochida, Proc. 9th Int. Congr. Catal., Calgary, 1988, Vol. 2, p. 689. B. J. Kip, E. G. F. Hermans and R. Prins, Proc. 9th Int. Congr. Catal., Calgary, 1988, Vol. 2, p. 821. C. B. Murchison (assigned to Dow Chemical Co.), U. S. Patent 4,539,334, Sept. 9, 1985. S. Stevenson, J. A. Dumesic, R. T. K. Baker, and E. Ruckenstein, Metal-Support Interactions in Catalysis, Sintering, and Redispersion, Van Nostrand Reinhold Catalysis Series, New York, 1987.
(1988) 61-69.
Elsevier, Amsterdam, 1987, p. 81-104. (1985) 415.
64 65 66 66a 67
217
68a 68b 69 70 71 72 73 74a 74b 75 76 77 78 79 80 81 82 83 84 85 86a 86b 87 88 89 90 91 92 93a 93b 94 95 96
R. C. Reuel and C. H. Bartholomew, J. Catal., 85 (1984) 78- 88. C. H. Bartholomew and R. C. Reuel, Ind. & Eng. Chem. Prod. Res. & Develop., 24 (1984) 56. L. Fu and C. H. Bartholomew, J. Catal., 92 (1985) 367-387. L. R. Neubauer, Ph. D. Dissertation, Brigham Young Univ., July, 1986. J. L. Rankin and C. H. Bartholomew, in preparation. V. K. Jones, L. R. Neubauer, and C. H. Bartholomew, J. Phys. Chem., 90 (1986) 4832. B. E. Concha, G. L. Bartholomew, and C. H. Bartholomew, J. Catal., 89 (1984) 536-541. J. Barrault, A. Alouche, A. Chafik, V. Paul-Boncour and S. Probst, Roc. 9th Int. Cong. Catal, Calgary, 1988, Vol. 2, p. 642. J. Barrault, A. Guilleminot, J. C. Achard, V. Paul-Boncour, and A. Percheron-Guegan, Appl. Catal., 21 (1986) 307-312. M. Boudart and M. A. McDonald, J. Phys. Chem., 88 (1984) 2185. H.-J. Jung, P. L. Walker, Jr, and M. A. Vannice, J. Catal., 75 (1982), 416. R. D. Kelley and D. W. Goodman, Surf. Sci., 123 (1982) L743. R. D. Kelley and D. W. Goodman, Amer. Chem. SOC.Div. Fuel Chem. Prepr., 25(2) (1980) 43. B. G. Johnson, C. H. Bartholomew, and D. W. Goodman, submitted for publication. M. Rameswaren and C. H. Bartholomew, J. Catal., 117 (1989) 218-236. M. A. Vannice and M. D. Mitchell, Ind. Eng. Chem. Fund., 88 (1984) 23. S. J. Rieck and A. T. Bell, J. Catal., 99 (1986) 278. G. B. Raupp and J. A. Dumesic, J. Catal., 95 (1985) 587; 96 (1985) 597: 97 (1986) 85. M. A. Wheeler and M Bettman, J. Catal., 40 (1974) 124. Y.-J. Huang, J. A. Schwarz, J. R. Diehl, and J. P. Baltrus, Appl. Catal., 37 (1988) 229. W.-H. Lee, Ph.D. Dissertation, Brigham Young University, June 1988. W.-H. Lee and C. H. Bartholomew, in preparation. L. M. Tau and C. 0. Bennett, J. Catal., 89 (1984) 285-302. M. L. Knotek, Surf. Sci., 91 (1980) L17. D. E. Resasco and G. L. Haller, J. Catal., 82 (1983) 279. C. B. Murchison and D. A. Murdick, Hydrocarbon Proc., (Jan. 1981) 156. C. B. Murchison, Fourth Int. Conf. on Chem. and Uses of Mo, Golden, CO, Aug. 9-13, 1982. M. A. Vannice, J. Catal., 74 (1982) 199-202. V. K. Jones, M. S. Thesis, Brigham Young University, April, 1985. V. K. Jones, L. R. Neubauer, and C. H. Bartholomew, paper in preparation. E. L. Kugler, Amer. Chem. SOC.Div. of Petr. Chem. Prepr., 25(3) (1980) 564-569. M. A. Vannice and R. L. Garten, J. Catal., 56 (1979) 236; 66 (1980) 242. R. A. Fiato and E. L. Kugler (Exxon Res. and Eng. Co.), U. S. Patent 4,689,313, Aug. 25, 1987.
218
97 98 99 100 l01a lOlb 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124
H. Abrevaya, W. M. Targos, H. J. Robota, and M. J. Cohn, in Catalysis, J. W. Ward (ed.), Elsevier, Amsterdam, 1987, p. 97. R. B. Levy, in Advanced Materials in Catalysis, J. J. Burton and R. L. Garten (eds.), Academic Press, 1977, p. 101. R. Snel, App. Catal., 37 (1988) 35-44.
M. Saito and R. B. Anderson, J. Catal., 63 (1980) 438; 67 (1981) 296. L. Volpe and M. Boudart, J. Solid State Chem., 59 (1985) 332,348. L. Volpe and M. Boudart, Catal. Rev.-Sci. Eng., 27(4) (1985) 515. A. A. Hummel, A. P. Wilson, and W. N. Delgass, J. Catal., 113, (1988) 236-249. W. N. Delgass, Abstracts 1l t h North Am. Meeting of the Catal. Soc.,Dearborn, May 7-1 1, 1989, CO1. R. Schulz, H.G. Feller, and M. Ralek, Proc. 9th Int. Congr. Catal., Calgary, 1988, Vol. 2, p. 800. R. C. Wade, D. G. Holah, A. N. Hughes, and B. C. Hui, Catal. Rev.-Sci. Eng., 14 (1976) 211. A. H. Uken and C. H. Bartholomew, J. Catal., 65 (1980) 402. C. H. Bartholomew, “Metal Boride Catalysts for Indirect Liquefaction”, Final report to DOE, DOE-PC-50816-12,Oct. 15, 1986. C. H. Bartholomew and A. H. Uken, Appl. Catal., 4 (1982) 19. C. H. Bartholomew and R. M. Bowman, Appl. Catal., 15 (1985) 59. J. Wang and C. H. Bartholomew, Proc. 9th Int. Congr. Catal., Calgary, 1988, Vol. 2, p. 813. C. Wang and C. H. Bartholomew, paper submitted for publication, 1989. Y. Okamoto, E. Matsunaga, T. Imanaka, and S. Teranishi, J. Catal., 74 (1982) 183. Y. Okamoto, Y. Nitta, T. Imanaka and S. Teranishi, J. Catal., 64 (1980) 397. E. L. Sughrue and C. H. Bartholomew, Appl. Catal., 2 (1982) 239. R. A. Dictor and A. T. Bell, J. Catal., 97 (1986) 121. L. H. Slaugh, U. S. Patent 3,996,256, Dec. 7, 1976. S. S. Khera (Battelle Develop. Corp.), U. S. Patent 4,077,995, March 7, 1978. C. B. Murchison and D. A. Murdick, U. S. Patent 4,15 1,190,April 24, 1979. J. Happel and M. A. Hnatow, U. S. Patent 4,151,191, April 24, 1979. C. B. Murchison, M. M. Conway, and R. R. Stevens, Abstracts 10th No. Am. Meeting of the Catalysis Soc.,San Diego, May, 1987, Paper A46. G. J. Quarderer and G. A. Cochran (Dow Chemical Co.), European Appl. No. 84102932.5, March 16, 1984. K. Klier, R. G. Herman, G. W. Simmons, and C. E. Lyman, Proc.Indirect Liquefaction Contractors Meeting, Pittsburgh, 1985, pp. 70-80. British Coal, Synthesis of Chemical Feedstocks and Intermediates,Private Communication in ref. 15. L. Guczi, Catal. Rev.-Sci. Eng., 23 (1984) 329-376.
219
M. Nakamura, B. J. Wood, P. Y. Hon, and H. Wise, 7th Int. Congr. Catal, Tokyo, June 30-July 4, 1980, Paper A29, p. 432. 126 R. M. Stanfield and W. N. Delgass, J. Catal., 72 (1981) 37- 50. 127 S. L. Soled and R. A. Fiato (Exxon Res. and Eng. Co.), U.S. Patent 4,544,671, Oct. 1, 1985. 128a E. Iglesia, S. L. Soled, and R. A. Fiato (Exxon Res. and Eng. Co.), U.S. Patent 4,738,948, Apr. 19, 1988. 128b C. H. Mauldin (Exxon Res. and Eng. Co.), U.S. Patent 4,670,475, June 2, 1987 128c P. K. Coughlin (Union Carbide Corp.) U.S. Patent 4,663,355, May 5, 1987 128d K. B. Arcuri, C. H. Mauldin, and D. H. Shaw (Exxon Res. and Eng. Co.), U.S. Patent 4,558,030, Dec. 10, 1985 129 Preparation of Catalysts IV, B. Delmon, P. Jacobs and G. Poncelet (eds.), Elsevier, 1987. (Elsevier please finish) 130a M. Boudart, A. Delbouville, J. A. Dumesic, S. Khammouma, and H. Topsoe, J. Catal., 37 (1975) 486. 130b J. A. Dumesic, H. Topse, S. Khammouma, and M. Boudart, ibid, 37 (1975) 503. 131 H. Topsoe, J. A. Dumesic, E. G. Derouane, B. S. Clausen, S. Morup, J. Villadesen, and N. Topsoe, Preparation of Catalyst, Vol. 11, Elesevier, Amsterdam, 1979, p. 365. I32 C. H. Bartholomew, R. B. Pannell, and J. L. Butler, J. Catal., 65 (1980) 335. 133 J. A. van Dillen, J. W. Geus, L. A. M. Hermans, and J. van der Meivben, Proc. of the 6th Internat. Congress on Catalsis, London, 1976. 134 J. T. Richardson and R. J. Dubus, J. Catal., 54 (1978) 207. 135 G. A. Hadjigeorghiou and J. T. Richardson, Appl. Catal., 21 (1986) 11, 37,47. 136 W. K. Bell, W. 0. Haag, G. W. Kierker, and D. J. Klocke (Mobil Oil Corp.), U. S. Patent 4,686,313, Aug. 11, 1987. 137 C. H. Bartholomew and M. Boudart, J. Catal., 25 (1972) 173. 138 R. A. Fiato, G. W. Rice, S. Miseo, and S. L. Soled (Exxon Res. and Eng. Co.), U. S. Patent 4,687,753, Aug. 18, 1987. 139 H. Abrevaya and W. M. Targos (UOP Inc.), U. S. Patent 4,714,692, Dee. 22, 1987. 140 G. Lohrengel, M. R. Dass, and M. Baerns, Preparation of Catalysts 11, Elsevier, Amsterdam, 1979. 141 J. P. Baltrus, J. R. Diehl, M. A. McDonald, and M. F. Zarochak, Appl. Catal., 48 (1989) 199-213. 142a A. Brenner and D. A. Hucul, Inorg. Chem., 18 (1979) 2836. 142b A. Brenner, J. Chem. Soc. Chem. Commun., (1979) 251. 143a D. Commereuc, Y. Chauvin, F. Hugues, J. M. Basset, J. Chem. Soc.Chem. Commun., (1980) 154. 143b F. Hugues, B. Besson and J. M. Basset, J. Chem. SOC. Chem. Commun., (1980) 719. 1 4 3 ~ F. Hugues, P. Bussiere, J. M. Basset, D. Commereuc, Y. Chauvin, L. Bonneviot, and D. Olivier, Stu. Surf. Sci. Catal., 7 (1981). 125
220
144 145
Y. W. Chen, H. T. Wang, and J. G. Goodwin, Jr., J. Catal., 83 (1983) 415. J. Phillips and J. A. Dumesic, Appl. Catal., 9 (1984) 1.
146 147 148
J. Zwart and R. Snel, J. Molec. Catal., 39 (1985) 305-352. A. Brenner, in Metal Cluster, M Moskovits (Ed.), Wiley, 1986, p. 249.
149 150 151
Metal Clusters in Catalysis, B. C. Gates, L. Guczi, and H. Knoezinger (Eds.), Elsevier, 1986, Chapter 9. R. P. Zerger, K. C. McMahon, M. D. Seltzer, R. G. Michel, and S. L. Suib, J. Catal., 99 (1986) 498-505. K. C. McMahon, S. L. Suib, B. G. Johnson, and C. H. Bartholomew, J. Catal., 106 (1987) 47-53. F. Rodriguez-Reinoso, I. Rodriguez-Ramos, A. Guerrero-Ruiz, and J. D. Lopez-Gonzalez, Appl. Catal., 21 (1986) 251-261. J. Zwart and J. Vink, Appl. Catal., 33 (1987) 383-393.
152 153 D.-K. Lee and S.-K. Ihm, J. Catal., 106 (1987) 386-393 154 P. N. Dyer, R. Pierantozzi, and H. P. Withers (Air Prod. and Chem. Inc.), U. S. Patents 4,670,472 and 4,619,910, June 2, 1987. 155 S. T. Wong and R. F. Howe, Appl. Catal., 47 (1989) 229-242. 156 B. G. Johnson, M. Rameswaren, M. D. Patil, G. Muralidhar, and C. H. Bartholomew, Catal. Today, in press, 1989. 157 M. Rameswaren, M. Patil, T. Cable, and C. H. Bartholomew, in preparation, 1989. 158 P. A. Jacobs, N. I. Jaeger, P. Jiru, and G. Schulz-Ekloff (Eds.), Metal Microstructures in Zeolites: Preparation-- Properties--Applications, Elsevier, 1982. 159 P. A. Jacobs, Metal Clusters in Catalysis, B.C. Gates, L. Guczi and H. Knoezinger (Eds.), Elsevier, 1986, Chapter 8. 160a D. Vanhove, P. Makambo, and M. Blanchard, J. Chem. Soc. Chem. Commun., (1979) 605. 160b D. Vanhove, Z. Zhugyong, L. Makambo, and M. Blanchard, Appl. Catal., 9 (1984) 327. 161 D. Fraenkel, and B. C. Gates, J. Am. Chem. SOC.,102 (1980) 2478. 162a D. Ballivet-Tkatchenko and I. Tkatchenko, J. Molec. Catal., 13 (1981) 1. 162b D. Ballevet-Tkatchenko, G. Coudurier and H. Mozzanega, Catalysis by Zeolites, B. Imelik, C. Naccache, Y. Ben Taarit, J. C. Vedrine, G. Coudurier, and H. Praliaud (Eds.), Elsevier, 1980, p. 309. 162c D. Ballivet-Tkatchenko, N. D. Chau, H. Mozzanega, M. C. Roux, and I. Tkatchenko, ACS Symposium Series, 152 (1981) 187. 163a L. F. Nazar, G. A. Ozin, F. Hugues, and J. Godber, J. Molec. Catal., 21 (1983) 313. 163b L. F. Nazar, G. A. Ozin, F. Hugues, J. Godber, and D. Rancourt, Angew. Chem. Int. Ed. Engl., 22 (1983) 624. 164 R. K. Ungar and M. C. Baird, J. Chem. Soc., Chem. Commun., (1986) 643. 165 H. H. Nijs, P. A, Jacbos and J. B. Uytterhoeven, J. Chem. SOC.,Chem. Commun., (19791, 180. 166 T. Wakui and T. Handa, J. Chem. SOC.,Chem. Commun., (1984), 994.
22 1
167 168a 168b 168c 169a 169b 170 171 172a 172b 173a 173b 173c 174 175 176a 176b 177 178 179 180 181 182 I83a 183b 183c 183d 184a
M. C. Baird, private communication, 1989. C. D. Chang, W. H. Lang, and A. J. Silvestri, J. Catal., 56 (1979) 268. P. D. Caesar, J. A. Brennan, W. E. Garwood, and J. Ciric, J. Catal., 56 (1979) 274. Mobil Oil Corp., U. S. Patent 4,086,262. T. J. Huang and W. 0. Haag, Unpublished data. Chem. Eng. News, 58(36) (1980) 44. S. A. Butter, F. V. Hanson, and H. S. Sherry (Mobil Oil Corp.), U. S. Patent 4,207,248, June 10, 1980. V. U. S. Rao and R. J. Gormley, Hydrocarbon Processing 59(11) (1980) 139.
Y . W. Chen, H. T. Wang, J. G. Goodwin, Jr, and W. K. Shiftlett, Appl. Catal., 8 (1983) 303. Y. W. Chen, H. T. Wang, and J. G. Goodwin, Jr., J. Catal., 85 (1984) 499. R. Oukaci, A. Sayari, and J. G. Goodwin, Jr., J. Catal., 102 (1986) 126. R. Oukaci, J. C. S. Wu, and J. G. Goodwin, Jr., J. Catal., 107 1987) 471. R. Oukaci, J. C. S. Wu, and J. G. Goodwin, Jr., J. Catal., 110 (1988) 47-57. V. U. S. Rao, R. J. Gormley, R. R. Schehl, R. D. H. Chi, and R. T. Obermyer, Amer. Chem. Soc., Div. of Petr. Chem. Prepr., 31(1) (1986) 169-176. G. A. Melson and E. B. Zuckerman, 9th Int. Cong. Catal, Calgary, 1988, Vol. 2, p. 807. J. G. Miller, C. L. Yang, and J. Rabo, Indirect Liquefaction Contractors Mtg., Pittsburgh, DOE Pittsburgh Energy Technology Center, 1986, pp. 1-15. ibid, 1987, pp. 1-17. H. Tominaga and T. Tatsumi, Studies on the Utilization of Coal Through Conversion, Ministry of Education, Science and Culture, Japan, SPEY-16, 1987, pp. 163-168. J. Brennan, P. D. Caesar, J. Ciric, W. E. Garwood, and R. E. Holland (Mobil Corp.), U. S. Patent 4,269,783, May 26, 1981. P. K. Coughlin and J. A. Rabo (Union Carbide Corp.), U. S. Patents 4,556,645 and 4,617,320, Oct. 14, 1986. P. K. Coughlin and J. A. Rabo (Union Carbide Corp.), U. S. Patent, 4,652,538, March 24, 1987. C. J. Kim (Exxon Res. and Eng. Co.), U. S. Patent 4,547,525, Oct. 15, 1985. J. C. Kuo, C . D. Prater, and J. J. Wise (Mobil Corp.), U. S. Patents 4,041,094 and 4,049,741, Aug. 9, 1977 and Sept 20, 1977. H. R. Ireland and T. R. Stein (Mobil Corp.), U. S. Patent 4,044,063, Aug. 23, 1977. D. Milstein and T. R. Stein (Mobil Corp.), U. S. Patent 4,044,064, Aug. 23, 1977. H. R. Ireland, A. W. Peters and T. R. Stein (Mobil Corp.), U. S. Patent 4,045,505, Au&. 30, 1977. D. Milstein and T. R. Stein (Mobil Corp.), U. S. Patent 4,071,574, Jan 31, 1978. P. D. Caeser, W. E. Garwood, W. F. Krudewig, and J. J. Wise (Mobil Corp.). U. S. Patent 4,111,792, Sept. 5 , 1978.
222
184b p. D. Caeser, W. E. Garwood, A. W. Peters, and J. J. Wise (Mobil Corp.), U. S. Patent 4,126,644, Nov. 21, 1978. 185 J. C. Kuo (Mobil Corp.), U. S. Patents 4,041,095, 4,041,096,4,046,830 and 4,046,831, Aug. 9, 1977 (2) and Sept. 6, 1977 (2). 186a H. R. Ireland and T. R. Stein (Mobil Corp.), U. S. Patent 4,041,097 , Aug. 9, 1977. 186b H. R. Ireland, A. W. Peters and T. R. Stein (Mobil Corp.), U. S. Patent 4,052,477, Oct. 4, 1977. C. J. Egan (Chevron Res. Co.), U. S. Patent 4,097,364, June 27, 1978. M. A. Boersma and S. T. Sie (Mobil Corp.), U. S. Patent 4,234,412, Nov. 18, 1980. N. Y. Chen, W. 0. Haag, T. J. Huang, and S. S. Wong (Mobil Corp.), U. S. Patent 4,500,417, Feb. 19, 1985. 190a M. E. Dry, Applied Industrial Catalysis, B.E. Leach (Ed.), Academic Press, 1983, Vol. 2, pp. 167-213. 190b M.E. Dry, Industrials Chemicals via C1 Processes, ACS Symposium Series, 328 (1987) 18-33. 190c B. Jager, M. E. Dry, T. Shingles and A. P. Steynberg, Paper 91D presented at the 1990 Spring National Meeting of the American Inst. of Chem. Engn, March 18-22, 1990, Orlando, Florida. 191 W. D. Deckwer, Oil &Gas Journal, (Nov. 10, 1980) 198. 192 H. Koelbel and M. Ralek, Catal. Rev.-Sci.Eng., 21 (1980) 225- 274. 187 188 189
193
C. N. Satterfield, G. A. Huff, Jr., H. G. Stenger, J. L. Carter, and R. J. Madon, Ind. & Eng. Chem. Fund., 24 (1985) 450.
194
C. N. Satterfield and H. G. Stenger, Ind. & Eng. Chem. Proc. Des. & Develop., 23 (1983) 849.
195
A. Lampert, J. Erickson, B. Smiley, C. Vaughan, R. Apienza, and W. Siegeir, Report to DOE, BNL 51799,1983. H. W. Pennline, R. R. Schehl, R. E. Tischer, and M. F. Zarochak, Paper 1 4 , AIChE Summer National Meeting, August 19-22, 1984, Philadelphia. W. E. Carroll, N. Cilen, and S. A. Motika, Report to DOE, DOE/PC/30021-T17, March 1985. W. E. Carroll, N. Cilen, and H. P. Withers, ACS Div. of Fuel Chem., Prepr., 31(3) (1986) 83-94. M. A. McDonald, ACS Div. of Petr. Chem., Prepr., 31(3) (1986) 17-22. E. Kikuchi, Studies On Utilization of Coal Through Conversion, Ministry of Education, Sciences, and Culture, Japan, SPEY-16, 1987, pp. 193-198. H. Itoh, H. Tanabe and E. Kikuchi, Appl. Catal., 47 (1989) Ll-L6. B. W. Brian and P. N. Dyer, Paper 14e, 1984 AIChE Summer National Meeting, Aug. 19-22, Philadelphia.
196 197a 197b 198 199a 199b 200 201a
S. Saxena and S. Shrivastava, Proc. Indirect Liquefaction Contractors Meeting, Pittsburgh, 1986, pp. 177-187.
223
201b 202a 202b
S.C. Saxena, M. Rosen, D. N. Smith, and J. A. Ruether, Chem. Eng. Commun., 40 (1986) 97-151. D. Stem, A. T. Bell, and H. Heinemann, Chem. Eng. Sci., 38 (1983) 597-605. A. T. Bell, and H. Heinemann, Proc. Univ. Coal and ARTD Coal Liquefaction Conference, June 5-7. 1984, Pittsburgh.
203
C. N. Satterfield., Proc. Univ. Coal and ARTD Coal Liquefaction Conference, June 5-7, 1984, Pittsburgh.
204
C. N. Satterfield, G. A. Huff, Jr., and H. G. Stenger, Ind. & Eng. Chem. Proc. Des. & Develop., 20 (1981) 666. K. Gupte, J. C. W. Kuo, T. M. Leib and J. Smith, 1984 AIChE Summer National Meeting, Aug. 19-22, 1984, Philadelphia. J. C. Kuo, M. Smith, K. M. Gupte, and T. M. Lieb, Proc. Indirect Liquefaction Contractors Meeting, 1985, Pittsburgh, pp. 113-144.
205a 205b 206 207 208a 208b 209 210 211a 211b 212
W. O'Dowd, D. N. Smith, and J. A. Reuther, Proc. Indirect Liquefaction Contractors Meeting, 1985, Pittsburgh, pp. 451- 478. D. B. Bukur, D. Petrov, and J. G. Daly, Proc. Indirect Liquefaction Contractors Meeting, 1985, Pittsburgh, pp. 479- 512. D. Stem, A. T. Bell, and H. Heinemann, Chem. Eng. Sci., 40 (1985) 1665-1677. Ibid, Ind. Eng. Chem. Process Des. Dev., 24 (1985) 1213-1219. K. Fujimoto, Studies On Utilization of Coal Through Conversion, Ministry of Education, Sciences, and Culture, Japan, SPEY-16, 1987, pp. 199-204. C. N. Satterfield and H. G. Stenger, Ind. & Eng. Chem. Proc. Res. & Develop., 24 (1985) 407,411. G. K. Whiting, Ph.D. Dissertation, Virginia Polytechnic Institute and State Univ., 1985. Y. A. Liu and A. M. Squires, Report to DOE, DOE/PC/50791-T3, April 30, 1984. R. W. Silverman, A. H. Thompson, A. Steynberg, Y. Yukawa and T. Shingles, Fluidization,
213 214 215 216 217a 217b 217c
A. M. Squires, Proc. Joint Mt. Chem. Eng., Chem. Industry and Engr. SOCof China, Am. Inst. of Chem. Eng., Beijing, Sept. 19-22, 1982. J. Topp-Jorgensen and H. C. Dibbern, 21st State-of-Art Symposium, ACS Div. of Industrial and Engr. Chem., Marco Island, USA, 1986. K.-H. Eisenlohr and H. Gaensslen, Erdoel Kohle, Erdgas, Petrochem. Brennst.-Chem., 35 (1982) 27. J. C. W. Kuo, Final Report to DOE, DOE/PC/30022-10, June 1983. T. Maxwell, M. J. v. d. Burgt, C. J. van Leeuwen, J. J. del'Amico and S. T. Sie, Petro Pacific 1984, Melbourne, Australia, Sept. 16-19, 1984. M. J. van der Burgt, J. van Klinken, and S. T. Sie, 5th Synfuels Worldwise Symposium, Washington, 1985, Nov. 11-13, 1985. J. Eilers, S. A. Posthuma and S.T. Sie, Paper 91A presented at the 1990 Spring National Meeting of the American Inst. of Chem. Engn, March 18-22, 1990, Orlando, Florida.
224
218a 218b 219a 219 219c 219d 219e 220 22 1 222 223 224 225
S. T. Sie (Shell Oil Co.), U. S. Patent, 4,579,986, April 1, 1986. J. K. Minderhoud and S. T. Sie (Shell Oil Co.), U. S. Patent, 4,594,468, June 10, 1986. H. R. Ireland and T. R. Stein (Mobil Corp.), U. S. Patent 4,046,829, Sept. 6, 1977. W. 0. Haag and T. J. Huang (Mobil Corp.), U. S. Patents 4,159,995 and 4,279,830, July 3, 1979 and July 21,1981. J. A. Brennan, P. D. Caeser, J. Ciric, and W. E. Garwood (Mobil Corp.), U. S. Patent 4,304,871, Dee. 8, 1981. W. 0. Haag, T. J. Huang, J. W. Kuo, and R. Shinnar (Mobil Corp.), U. S. Patent 4,252,736, Feb. 24, 1981. A. W. Chester, T.-S. Chou, Y.-F.Chu (Mobil Corp.), U. S. Patent 4,523,047, June 11, 1985. R. E. Holland and S. A. Tab& (Mobil Corp.), U. S. Patent 4,547,601, Oct. 15, 1985. W. R. Den-, Jr., W. E. Garwood, J. C. Kuo, R. M. Leib, D. M. Nace, and S. A. Tab& (Mobil Corp.), U. S.Patent 4,684,756, Aug. 4, 1987. H. Michael, M. Adriaan, L Schaper, J. Bijwaard, and M. Boersma (Shell Internationale), U. K. Patent Appl. 2,077,754A, Dee. 23, 1981. S. T. Sie (Shell Oil Co.), U. S. Patent 4,594,172, June 10, 1986. R. A. Fiato and C. J. Kim (Exxon Res. and Eng. Co.), U. S. Patent 4,624,968, Nov. 25, 1986. R. F. Blanks, J. L. Jezl, I. Puskas, and M. Stasi (Amoco Corp.), U. S. Patent 4,778,826, Oct. 18, 1988.
225
CHAPTER 6
BIMETALLIC CATALYSTS FOR CO ACTIVATION
J. Schwank Department of Chemical Engineering, The University of Michigan H. H. Dow Building, Ann Arbor, MI 48109, (USA)
226
6.1 INTRODUCTION Bimetallic catalysts play an important role in many industrial catalytic processes and represent an area of intense research interest (ref. 1). From the viewpoint of basic research, bimetallic systems give an opportunity to investigate geometrical versus electronic or ligand effects in catalysis. Studies of these effects can help to elucidate the key factors controlling the activity and
more importantly, the selectivity of metal catalysts. For the designer of catalysts, adding a second metal component allows in principle to systematically alter the size or electronic structure of catalytic surface ensembles. One could envision modifications of the adsorption characteristics, of the surface coverage with reactive intermediates, and consequently of the selectivity. The presence of a second metal Component can make its influence felt in terms of changes in reducibility of the catalyst. Furthermore, the second metal component can in certain cases alter the deactivation behavior, a fact that has proven to be beneficial in bimetallic reforming catalysts. Conventional Fischer-Tropsch catalysts are based on iron or cobalt. A major drawback of Fischer-Tropsch processes is their inherent low selectivity for desired products. Product distributions depend on the nature of the catalyst and the reaction conditions, but only in the case of methane and methanol can one achieve 100% product selectivity. The wide product spectrum typically obtained over Fischer-Tropsch catalysts is determined by the Anderson-Schultz-Flory polymerization kinetics (refs. 2,3). Several strategies have been used in attempts to break away from the Anderson-Schultz-Flory product distribution and to circumvent the selectivity limitations in CO hydrogenation (refs. 43):
1.
Limitation of chain growth by physical constraints such as encapsulation of metal particles in zeolites or in small pores in alumina;
2.
Interception of low-molecular weight surface intermediates before the polymerization reaction can propagate;
3.
Chemical modification of catalysts by additives, for example oxides of titanium, vanadium, manganese, molybdenum, and tungsten, or promoters such as potassium, or partial poisoning with sulfur;
4. Synthesis of methanol, followed by further conversion of methanol into desired products; 5. Modification of surface sites by addition of a second metal component; 6. Direct conversion of synthesis gas over organometallic clusters or cluster-derived catalysts that are not entirely subject to the Anderson-Schultz-Flory polymerization kinetics. This chapter will focus on the last two points and review methanation and Fischer-Tropsch catalysts modified by addition of a second metal component, and bimetallic catalysts derived from organometallic clusters. The purpose is to give an overview of supported bimetallic CO activation catalysts and to highlight major trends and fundamental principles. There is a certain degree of ambiguity as to the definition of a bimetallic CO hydrogenation catalyst. Many CO hydrogenation catalysts contain more than one active component, supported on more or less active carrier materials, and modified by various promoters. Especially in the patent literature one can find numerous claims for CO hydrogenation catalysts where the active metal is
227
modified by a host of metal or oxide additives. In many instances, the distinction between a second metal component and a promoter is blurred. Lee and Ponec proposed to define a promoter as “an element or compound, which itself is of negligible activity but which improves the activity, selectivity, and/or stability of the catalysts” (ref. 6). According to this definition, bimetallic
co
hydrogenation catalysts containing Ru and Cu should, in a strict sense, be called Cu promoted R u catalysts, as Cu by itself is not an active CO hydrogenation catalyst. Here, we will use the term bimetallic catalyst in a wide sense and designate any system containing two more or less reduced metal components as “bimetallic”, irrespective of the CO hydrogenation activity of the second component. However, catalysts containing typical promoter components such as potassium which are most certainly present in oxide form will not be considered as bimetallic. Similarly, Fe-Mn catalysts will not be treated here under the category of bimetallic catalysts, as the active phase of such catalysts contains a mixture of iron compounds and manganese oxide. The increasing complexity of modern catalysts and the trend towards the development of very sophisticated multicomponent catalysts are largely driven by the desire to improve selectivity. These multicomponent catalytic systems, as important as they might be for industrial purposes, are not ideally suited to illustrate fundamental principles of bimetallic catalysis, and will not be emphasized in this chapter. The review will not include homogeneous CO hydrogenation catalysts which have recently been reviewed (ref. 7) and focus mainly on heterogeneous catalyst systems. This review will also not deal explicitly with catalysts specifically developed for alcohol synthesis, such as Cu-ZnO, as there is a separate chapter devoted to this subject. The advantage of working with bimetallic heterogeneous catalysts comes with a price: the characterization of bimetallic systems is considerably more difficult than that of monometallic catalysts, and the number of experimental variables rises drastically. For example, one needs to address the question of miscibility of the metal components as a function of metal particle size. There is no guarantee that all particles have the same composition. As a further complication, concentration gradients between the surface and the center of a metal particle can exist. The effect of the support and of the preparation conditions on bimetallic particle formation needs to be explored. Under reaction conditions, one or both of the metal components could undergo changes in oxidation state. The possibility of changes in particle size and composition as a function of temperature and reactive environment needs to be taken into account. Surface enrichment of one component, and even segregation into monometallic particles need to be considered. These are just a few examples illustrating the remarkable complexity of problems to contend with in working with bimetallic supported catalysts.
6.2 THE GENESIS AND NATURE OF SURFACE SITES IN BIMETALLIC CATALYSTS In ideal bimetallic systems, it is relatively straightforward to derive the thermodynamic
equilibrium composition of the surface as a function of the bulk composition and temperature and to predict the nature and extent of surface segregation (refs. 8-10). As a general rule, the metal with the lower energy of sublimation has a tendency to segregate at the surface. Bimetallic supported
228
catalysts, however, tend to be rather non-ideal (ref. 11). A vivid illustration of this point is the formation of bimetallic aggregates or “clusters” from metals which are for all practical purposes immiscible in their bulk state. Examples are ruthenium-copper and ruthenium-gold, systems that will be discussed in more detail below. The structure and composition of such bimetallic aggregates depends on a variety of factors, and is strongly influenced by the method of catalyst preparation. The formation of small bimetallic particles on a support is in many cases kinetically, rather than thermodynamically controlled. Many of the bimetallic systems of catalytic interest contain very small metal particles o n the order of a few nanometers. As an aggregate of metal atoms gets smaller and smaller, one would expect to reach a critical threshold size below which the bulk metal properties cannot be extrapolated. For example, in a metal particle which contains just a few atoms the electrons will be localized and confined within a few A. For all practical purposes, this would qualify such a metal particle as an insulator. In a similar vein, one could start to question for very small bimetallic particles the applicability of bulk structure and composition data. A direct implication of the postulated breakdown of bulk properties could be that miscibility limits imposed by the phase diagrams may not fully apply (ref. 12). The phase diagrams describing the miscibility of metals and alloy formation are based on thermodynamic equilibria in two-component systems. In supported heterogeneous catalysts, especially when small metal particles are present, the situation is much more complicated, and one would be ill advised to uncritically apply phase diagram data to describe the composition and structure of small bimetallic particles. First of all, there is a distinct possibility that the catalyst preparation might not lead to full thermodynamic equilibration. Furthermore, the nature of the precursor compounds and their chemical interaction with the chosen support material can exert a strong influence on the sequence of metal nucleation and reduction. Finally, depending on the relative values of surface tensions at the respective metal-support interfaces, different degrees of wetting between the two metals and the support could be envisioned. This, in turn, could influence the resulting shape, structure and composition of the bimetallic particles, and in extreme cases preclude the formation of bimetallic particles altogether, at least within certain particle size ranges. Given this background, it is not surprising that there is no satisfactory theoretical treatment available that would allow a prediction of the surface composition or structure of small bimetallic particles prepared from metal components with limited bulk miscibility. Consequently, there is at present not much hope to predict a priori the geometrical or electronic structure of the surface of such bimetallic systems. There are several different ways to prepare bimetallic catalysts. Preparation methods for bimetallic catalysts have been reviewed previously (refs. 13,14). Only a brief overview shall be given here for the preparation of bimetallic CO hydrogenation catalysts. Most of the CO hydrogenation catalysts developed in the early stages of Fischer-Tropsch technology were based on iron, cobalt or nickel and prepared by precipitation techniques (refs. 15-17). Already in the early work, the addition of a second metal component was explored with the main purpose to facilitate catalyst reduction. For example, Ruhrchemie’s catalyst that found for many years application at Sasol (ref. 18) was prepared from a solution of iron and copper
229
nitrate which was precipitated by pouring it into a hot solution of sodium carbonate. After filtering and washing of the precipitate, it was reslurried with water and then impregnated with a potassium waterglass solution, filtered, dried, and reduced. Copper in precipitated iron catalysts seems to enhance the rate of reduction, so that the reduction process can be carried out under milder conditions (ref. 17). In absence of copper the degree of reduction at normal reduction temperatures is low, and the resulting activity is rather poor. Forcing the reaction to proceed at higher temperatures so that the same degree of reduction as in the copper case is obtained does still not lead to satisfactory catalyst performance. Obviously, the high temperature reduction process is detrimental to the surface area. The exact role of copper in enhancing the reduction and the resulting catalytic activity of precipitated iron catalysts is still not fully resolved, as the phenomenon has been described mainly from an empirical viewpoint. In the case of cobalt and nickel catalysts, presence of copper leads to similar enhancements of the reduction process. However, the use of copper has been discontinued as it appeared to be detrimental to the activity maintenance of the catalyst. Again, the exact reasons for this effect are not yet clear. Other supported metals that show potential as CO hydrogenation catalysts are ruthenium and osmium. Platinum, palladium and iridium have generally low activity. Rhodium tends to give oxygenated products. Molybdenum has found some interest as a possible candidate for a sulfur tolerant Fischer-Tropsch catalyst. There has been a considerable number of studies on bimetallic systems containing two of the above mentioned metals, and throughout this review we will highlight work on such bimetallic catalysts relevant to CO activation. One of the drawbacks of conventional supported bimetallic catalysts prepared by precipitation or impregnation is their inherent heterogeneity which makes it difficult to correlate the activity with the structure and composition of the surface. This state of affairs motivated a great deal of research into better defined bimetallic systems derived from well-characterized bimetallic clusters. In principle, cluster-derived catalysts offer a means to amve at well-defined catalyst stoichiomemes and surface suuctures. Recently, excellent reviews of this subject have been published (refs. 19-24). The organometallic precursor route opens the possibility to generate novel catalyst surface sites which represent metastable states with better selectivity. Such sites may not be accessible through the conventional preparation methods relying on impregnation of supports with inorganic salt solutions and reduction. Bimetallic catalysts derived form organometallic clusters should also be more uniform in terms of particle size distribution and composition, as the ratio of the two metals is fixed by the cluster composition. This makes cluster derived catalysts more amenable to spectroscopic characterization than the non-uniform conventional catalysts. Furthermore, the well defined metal framework of bimetallic molecular clusters might carry over, at least to some extent, into the spatial relationship between the two metal components in the bimetallic particles on the support. Cluster-derived supported bimetallic catalysts represent, therefore, interesting models for bimetallic catalysts, provided one is able to establish some relationship between the original structure of the organometallic cluster and the resulting aggregate of atoms on the support. In most cases, organometallic clusters when anchored onto a support decompose and often lose their
230
structural uniqueness. They still may serve as interesting precursors of bimetallic aggregates. The cluster route will then lead to highly dispersed supported metal particles similar to those encountered in conventionally prepared catalysts. The question of structural stability of the metal framework under reaction conditions remains as one of the major challenges in this field. Most catalysts derived from bimetallic organometallic clusters tend to break up under reaction conditions and agglomerate into metal particles which are not very different from those one could prepare conventionally by impregnation of a support with inorganic precursor salts. There are, however, some examples which look promising in terms of preserving the structure of the metal framework. For instance, the adsorption of H ~ R U O S ~ ( Con O y-Al203 )~~ and exposure to an equimolar mixture of H2 and CO at 373-473 K resulted in the formation of the anionic species [H3R~O~3(C0)12]' (ref. 25). This anion could be extracted from the alumina surface by ion exchange and characterized in solution, indicating that some of the original cluster framework stayed intact. The stability of the cluster framework might be related to the presence of gas phase CO. The CO being a ligand, the abundance of CO in the reaction mixture might have prevented the total decarbonylation of all the clusters. In absence of CO, the bimetallic species was very susceptible to disintegration. When heated in vacuum beyond 373 K or when pretreated in flowing H;! at 373 K (ref. 40) it suffered extensive decomposition. A similar stabilizing effect of reaction mixtures containing CO has been observed also in monometallic osmium carbonyl-derived catalysts (ref. 26). High resolution electron microscopy indicated that clusters containing 3, 4, or 6 atoms of osmium supported on A1203 did not show any significant agglomeration into larger aggregates of osmium when exposed to a 1:l mixture of CO + H2 at 548 K for up to 40 h, although the original cluster framework might have broken up. However, when the cluster-derived catalysts where exposed to flowing hydrogen at 473 K for 2 h or exposed to much milder reaction environments containing no CO (such as the hydrogenolysis of cyclopropane), a much more pronounced agglomeration into larger osmium particles was observed. It is relatively easy to anchor organometallic clusters onto oxide supports, for example through reaction with surface hydroxyl groups on the support. The ligands are then removed by decomposition at moderate temperatures. There is, however, an inherent disadvantage in utilizing supports that contain large concentrations of surface hydroxyl groups. They tend to interact with the metal framework and oxidize the metal. A high temperature reduction step is then required to activate the catalysts. This treatment can cause metal aggregation and sintering into larger metal particles. Thereby, the advantage of structural definition of the active site may be lost. The nature of the support can play an important role in the genesis of cluster-derived bimetallic particles. In the case of carbonyl clusters, the nature of the support and its pretreatment (e.g., the degree of dehydroxylation) will decisively influence the initial stages of cluster-support interaction. The bond strength between the metal atoms and the CO ligands will be affected to different degrees. In extreme cases, the interaction between the cluster compounds and the hydroxyl groups on the support can cause disintegration of the cluster framework. The decomposition of metal carbonyls during production of supported metal catalysts has been reviewed by Phillips and Dumesic (ref. 23). The following general trends have been observed
231
in many metal carbonyl systems: on hydroxylated supports, some clusters already disintegrate upon adsorption or upon heating to moderate temperatures. The decomposition tends to result in the formation of subcarbonyls. Upon increasing the temperature further, the subcarbonyl species become completely decarbonylated, and highly dispersed metal oxide species are formed. The reduction process required to bring about zero valent, metallic particles often leads to aggregation into larger metal crystallites. To illustrate the effect of the support we will discuss cluster-derived Fe-Ru catalysts, a catalyst system that has been extensively studied (refs. 27-39). Following the initial interaction between metal carbonyl clusters and surface hydroxyl groups, decomposition of the surface intermediates can occur, resulting in the disintegration of the cluster framework. The extent of cluster disintegration depends on the nature of the support, in particular the reactivity of the surface hydroxyl groups, and the structure of the original cluster. In the case of Cab-0-Sil supports, both infrared spectroscopy and Mossbauer spectroscopy indicated that bimetallic molecular clusters such as Fe,Ru(CO),, and H ~ F ~ R U ~ ( could C O )be ~ ~adsorbed onto the support without disintegration of the cluster framework (ref. 31). On A1203, on the other hand, the Fe-Ru bond was already ruptured at room temperature (ref. 31). After the carbonyl clusters had been in contact with the alumina support for 12 hours, infrared spectroscopy indicated the coexistence of various subcarbonyl species in addition to the original cluster. Upon evacuation at room temperature, evidence was obtained for decomposition of the bimetallic cluster and splitting of the Fe-Ru bond. Mossbauer spectra (isomer shift = 0.35 mm s-l, quadrupole split = 1.1 mm s-*) showed that iron in F ~ , R u ( C O ) , ~and H,F~Ru,(CO),~ had been oxidized to Fe+3 during the impregnation of the alumina support. The decomposition of the clusters led to the formation of Ru(CO), subcarbonyls, in which Ru was also in an oxidized state. After reduction in flowing H2 at temperatures up to
660 K, the isomer shift of 1.07 mm-l and the quadrupole split of 2.04 mm-l indicated the complete reduction of Fe3+ into Fez+. During the reduction process, interactions between iron and ruthenium seemed to develop (ref. 34). Based on the results of molecular isotope exchange experiments between the l 2 C 0 ligands and gaseous I3CO, it was concluded that clusters with higher iron content underwent stronger interactions with the support. This, in turn, led to greater ease of cluster disintegration and oxidation of the iron species. On Ti02, the interaction with the clusters also appeared to be rather strong (ref. 35). The amount of CO released during temperature programmed decomposition provides a good estimate of the relative strength of cluster-support interactions on various supports. Similar effects have also been observed in other supported bimetallic cluster systems. For example, in the case of H2 R ~ O~ 3 ( C 0 ) 1supported 3 on y-A1203, the bimetallic cluster framework disintegrated into aggregated Ru metal and cationic complexes of Ru and 0 s (ref. 40). Alumina supported catalysts prepared from H2RhOs3(C0)10(a~ac)and H , F ~ O S ~ ( C Oclusters )~~ also suffered from disintegration of the metal framework (refs. 41,42). The supported Rhos3 clusters fomied first triosmium clusters and mononuclear Rh complexes. When heated to higher temperatures, aggregation of Rh crystallites and the presence of mononuclear 0 s complexes were observed. Choplin et al. observed breakup of the Fe-0s metal framework already at 400 K (ref. 41).
232
Table 6.1
Bimetallic CO Hydrogenation Catalysts Prepared from Organometallic Precursors Organometallic Precursor
support
Metal Loading
Reference
wt% Carbon black Cab-0-Sil Ti02 Carbon Carbon Ti02 Carbon Cab-0-Sil y-A1203 Ti02
6.2 0.5 1 3.05 1.74
1
?/-A1203
3.07 1 1 1 2.66
Si02 Carbon Si02 Carbon
4 7.9 4
A1203
Si02 MgO Ti02 zro2 Carbon Carbon Carbon Si02 Zr02/Al2O3 Si02 Si02 Si02 Si02 Si02 Si02 Y-A1203 7/-A1203 y-A1203 y-A1203 Si02 z*2 A1203 A1203
*l2O3 A'2°3
Si02 MgO Ti02
2-4.8 2-4.8 2-4.8 1.4-2.4 49 2 2 2 2 2 2 0.72 2 2.32 1 1 0.2
1 1 1 1
39 28,30 35 39 39 35 39 31 3 1,34 35 42 41 44 45
44 46 46 46 46 46 47 47 47 48 50 51 51 51 51 51 51 40 25,40 42 52 52 53 54 54 54 55,56,57 58,59,60 60 60
233
Although the importance of the support hydroxyl group concentration is recognized, systematic studies of bimetallic cluster deposition on dehyroxylated oxide supports are still scarce. Carbon supports have the advantage that one can remove most of the surface hydroxyl or
carbox ylate groups. Therefore, there is less likelihood of cluster 6.1 Molecular structure of planar isomer of ( C P ) ~ M O ~ F ~ ~ S ( C O ) ~ decomposition due to Fig. cluster as determined by single crystal X-ray diffraction interactions with surface hydroxyl groups. Carbon supported Fe-Ru catalysts prepared from stoichiometric mixed-metal carbonyl clusters contained highly dispersed, raft-like crystallites (ref. 39). The bimetallic catalysts showed surface enrichment in iron, in agreement with observations made on unsupported Fe-Ru alloys (ref. 43). Fe-Co catalyst supported on carbon showed very high CO hydrogenation activity already after mild reduction at 473 K which led to significant cluster decarbonylation (ref. 44). Table 6.1 lists additional examples of bimetallic supported catalysts for CO activation that have been prepared via organometallic synthesis routes. In attempts to strengthen the stability of the bimetallic cluster framework, bridging or capping ligands have been A1203 used, for instance bridging Si02 no2 %!O sulfur ligands Fig. 6.2 Comparison of loss of CO ligands as cyclopentadienyl (Cp) ( C O ) ~ deposited onto various in h-(C5H4Me)2M02Fe2S2 ligands from ( C P ) ~ M O ~ F ~ ~ S clusters supports upon heating in H2 at 670 K (adapted from ref. 60)
234
(C0)g (refs. 55-60). Two isomers of this cluster can be synthesized, with the metal atoms either in a planar or in a butterfly configuration. The structure of the planar sulfided bimetallic Mo-Fe cluster
is shown in Fig. 6.1. Infrared spectroscopy and Mossbauer spectroscopy indicated that the clusters were adsorbed on the support with minimal structural perturbation (ref. 56). This conclusion was later supported by X-ray absorption near edge structure spectroscopy (XANES) where identical spectra were obtained for the pure cluster and the cluster adsorbed on A1203 (ref. 60). Upon heating, the nature of the support was found to have a drastic effect on the loss of ligands (refs. 55-60). Fig. 6.2 shows a comparison of the loss of carbon containing ligands when the cluster was deposited onto various supports and heated under H2 at 670 K. The temperature programmed decomposition behavior of the cluster was different on each support (ref. 60). A burst of CO was observed near 373 K. At higher temperatures, gradual loss of the cyclopentadienyl ligands started, along with some C02 and CH,. In addition, small amounts of H2S or (CH&S were evolved. The total amount of carbon retained was found to increase as the basicity of the support increased. Based on the overall carbon balance during TPDE under H, at temperatures up to 673 K, each
A1203 supported cluster retained on the average 5 carbon atoms. On SO,, 5.5 carbon atoms were retained. On Ti02, the number increased to 8 carbon atoms per cluster, and on MgO, 10 carbon atoms were retained (ref. 59). However, it appears that the sulfur bridging ligands did not entirely stabilize the metal framework. Mossbauer, XANES and EXAFS spectroscopies indicated that both Fe and Mo were oxidized probably through interaction with surface hydroxyl groups on the support. On A1203, these surface 0x0 species were formed at about 393 K. On MgO, on the other hand, they fomied already at room temperature. A similar formation of oxidized Fe+3 species was found by Ichikawa et al. in SiO2 supported Fe2Rh4 and Fe3Pt3 cluster-derived catalysts even after reduction in H2 at 673 K (ref. 51). Preliminary EXAFS data indicated that in the Mo-Fe system long range order reappeared as the catalyst was heated under H, atmosphere to 673 K (refs. 59,60). It could be ruled out that the new long range order was due to the formation of metallic iron particles or simple iron
oxides. It is not yet entirely clear how this newly emerging long range order relates to the original structure of the pure molecular cluster. Bimetallic catalysts prepared from cluster carbonyls can show differences in catalytic behavior as compared to conventional CO hydrogenation catalysts with the same overall metal composition. These differences can be atmbuted to several factors. First, the cluster derived catalysts could be more uniform and in a state of very high dispersion that is hard to achieve in conventional catalyst preparation. As we will see later, there are examples of conventional bimetallic catalysts where one finds vast variations in particle composition from one metal particle to the next. Secondly, as discussed above, the anchoring of clusters on hydroxyl groups of the support and the subsequent cluster decomposition may lead to partial oxidation of metal sites. These cationic surface species may serve as stabilizing centers for highly dispersed metal particles or act similar to the function of ionic promoters added to conventional catalysts. The cluster decomposition may leave behind carbonaceous species which enhance the CO hydrogenation activity. Finally, there will be different residual impurities which are specific for a given
235
preparation method. For example, one might find significant amounts of residual chlorine in conventional catalysts prepared via impregnation of the support with chloride salt solutions. On cluster derived catalysts, such chloride impurities would not be present. The presence ions such as C1- or N03- introduced during conventional catalyst impregnation may affect the acidbase Characteristics of the support. This, in turn, could have consequences for the nucleation and growth of bimetallic particles and the mobility of metal species during catalyst pretreatment. The overall performance of cluster derived catalysts could be influenced by any combination of the above effects. The verdict is not yet in which one of these effects is the most crucial one for creating different activity or selectivity of cluster-derived bimetallic catalysts. Another interesting method to prepare supported bimetallic catalysts is solvated atom metal dispersion (SMAD) (ref. 61). The synthesis is based on the use of free metal atoms in the vapor state which are solvated at low temperatures. The solution is then brought into contact with catalyst supports to deposit atoms or clusters of metal atoms on the support. A very attractive feature of this preparation method is that the formation of SMAD catalysts occurs under mild conditions, and the metal atoms are directly deposited in a reduced state. SMAD catalysts contain carbonaceous residues and represent, in a sense, pseudoorganometallic catalysts. These carbonaceous residues are remnants of carbon species formed during catalyst preparation. To some extent, the carbon bound to the metal surface might stabilize the metal dispersion and play a role in CO hydrogenation reactions. Unsupported bimetallic catalysts are from an industrial perspective less important, but can be of great help in basic research, as the effect of the second metal component can he studied undisturbed by contributions from the support or promoters. Of course, there may be situations where the presence of a support is a very crucial factor in the formation of bimetallic particles. When comparing results obtained on unsupported model catalysts with those on supported catalysts, this fundamental difference has to be kept in mind. Vapor deposition of pure metals (refs. 62,63) is ideally suited for the preparation of thin film model catalysts. A second metal component can also be deposited on the surface of single crystals, for example Cu on well defined Ru single crustal surfaces (refs. 64-69). There are several methods to prepare unsupported alloy catalysts. One approach consists of leaching one metal component from a ternary alloy to give bimetallic Raney-type materials with higher surface area (ref. 70). For example, Fischer and Meyer (ref. 71) prepared skeletal alloys of nickel and cobalt by fusing metallic nickel and cobalt with aluminum. The aluminum was then leached with an excess of hot aqueous sodium hydroxide solution. The performance of these skeletal catalysts was found to be generally inferior to precipitated catalysts. Other examples include Al-Fe-Mn alloys where the aluminum is then leached out. Such Raney Fe-Mn catalysts showed good C2C4 olefin selectivities while the overall hydrocarbon yields were similar to those on monometallic Raney-Fe catalysts (ref. 72). Since precipitated Fe-Mn catalysts showed similar improvements in selectivity, it is not likely that the special texture of the Raney catalysts was the main cause of the enhanced olefin selectivity. The effect of leaching reagents on activity and
236
selectivity in CO hydrogenation over Raney Fe-Mn catalysts has been studied in a recent dissertation (ref. 73). Another method of preparation of unsupported bimetallic catalysts starts from a mixture of appropriate metal salts which is subsequently decomposed or reduced to the metallic state (ref. 74). Nonnoble alloy catalysts are usually made from the corresponding carbonates, nitrates, or hydroxides, which are then calcined to convert them into an oxide mixture followed by reduction to generate an alloy. Bimetallics containing group VIII noble metals are often prepared from halide salts which are reduced in a suitable reducing medium, for example in flowing hydrogen gas or in hydrazine solutions. Recently, catalytic applications of amorphous alloys have been explored (refs. 75-80). A variety of techniques can be used to prepare. amorphous alloys, including vapor deposition, sputtering, electroplating, chemical plating, and rapid quenching of melts. These amorphous alloys represent non-equilibrium systems and their use as catalysts has to be restricted to temperature regimes below their crystallization temperature. Examples for the use of amorphous alloys as CO hydrogenation catalysts include Fe-Ni (ref. 75), PdZr (ref. 76), AuZr (ref. 77), and FeB (refs. 79,80). Depending on the preparation conditions, large variations in catalytic activities have been observed. These differences might be related to changes in the microstructure of the surface. It also cannot be ruled out that the surface might partially crystallize under reaction condition.
6.3 THE EFFECT OF CATALYST PREPARATION ON THE PROPERTIES OF THE SUPPORT AND THE CONSEQUENCES FOR BIMETALLIC PARTICLE FORMATION It is to be expected that the support will influence the adsorption of inorganic precursor salt solutions and the nucleation and aggregation of bimetallic particles during drying, reduction, and pretreatment. The catalytic literature contains a large number of studies concerning metal-support effects. Most of these studies focus on changes induced by metal-support interactions in the adsorption characteristics and catalytic behavior of metals. Much less attention has been paid to the other side of this issue, namely how the preparative chemistry and the nature of the metal precursors change the properties of the support. We have already recognized that the concentration of surface hydroxyl groups on the support plays a decisive role in anchoring organometallic precursor complexes and in modifying the loss of ligands during decomposition of the clusters. While it is quite obvious that the concentration of surface hydroxyl groups will increase when a support is exposed to an aqueous solution of a metal salt, one has to also consider the consequences for the morphology and surface uniformity of the support. A good example for this phenomenon is the MgO support (refs. 81-83). When MgO is exposed to an aqueous medium typically used for impregnation with metal salt solutions, bulk hydration of MgO to Mg(OH)Z occurs. When the impregnated oxide is heated to remove the water, the hydroxide is dehydrated back to MgO. However, the process of hydration-dehydration can cause massive changes in the morphology of the MgO support. As water is released from the
237
structure, the transition from hexagonal-close-packed Mg(OH), into simple cubic MgO requires a large change in specific volume. During the dehydration, the hexagonal plates of the hydroxide undergo a vertical collapse which is accompanied by a slight lateral shift of the layers to accomplish the transformation into the cubic structure. The resulting morphology of the MgO crystallites is far from the expected cube shape. Instead, a highly porous, polycrystalline material is obtained with surface areas much larger than that of the MgO initially used for impregnation. The increase in surface area is due to the breakup of the structure during the release of water. The counter-ion used in the catalyst impregnation also plays an important role. In the case of MgO, the presence of chloride ions influences the dehydration behavior of the support, enhancing the mobility of MgO for sintering and grain growth (ref. 83). A similar effect of chlorine on the morphology and surface area of the support has been found in the case of T i 0 2 (ref. 84). TO achieve surface uniformity of MgO on an atomic scale and to end up with well defined cubic particle morphologies, the MgO support has to be of extremely high purity. The presence of highly dispersed metal particles in intimate contact with MgO could be considered as impurities, and would prevent the formation of atomically uniform MgO surfaces. Other supports such as SiOz or A1203 appear to be less susceptible to massive structural changes induced by the impregnation process, but there still may be some subtle changes which may influence the formation of bimetallic particles. The shape, size, and composition of the metal particles is likely to be dependent on the surface morphology and 0 10 surface uniformity of the nolysis 518K support. Also, the pH of the metal salt solution is important, as it can change the acidbase characteristics of the support surface. This, in
4
I
m
B
f
turn, can have profound Ccl consequences for the adsorption of the metal precursors and the nucleation and metal particle aggregation lo4 process. In bimetallic + + 10 catalysts, this can be very 0 2 0 4 0 6 0 8 0 1 critical, as it will impact Atom % A u (lWAuAAu+RU)) on the sequence of Fig. 6.3 Activity of Ru-Au/SiOz and Ru-Au/MgO catalysts for nucleation of the two methanation at 518 K, H2/CO feed ratio of 3:1 and total pressure of 198 Wa. For comparison purposes, the activity for ethane metal components. It does, hydrogenolysis is also shown (adapted from ref. 86,91). The turnover therefore, not come as a frequencies are normalized with respect to Ru surface atoms as determined from hydrogen chemisorption.
ft
238
Table 6.2
EDS analysis of individual small metal particles in Ru-Au/Si02 catalyst containing 48 at% Ru and 52 at% Au (ref. 98)
Particle Size
Analysis Time
nm
s
2.0 2.0 2.0 2.0 2.0 2.0 2.5 2.5 2.5 3.O 3.0 4 .O 4.0
100 103 100 101 100 100 100 106 73 101 86 200 107
IAulIRu Backgrc
RUL
7 9 11 3 3 2 24 5 4 15 4 15 4
210 186 89 54 50 32 452 89 56 169 58 455 102
337 182 108 770 338
10 10 8 26 6
0.0429 0.0484 0.1348 0.1852 0.2000 0.2500 0.0575 0.0674 0.0893 0.0473 0.1034 0.0747 0.0784
surprise to see significant differences in the structure and behavior of bimetallic catalysts supported on different substrates. These phenomena, although in principle obvious, are not well understood. Only a very limited data base is available addressing this problem. The bimetallic systems Ru-Au/Si02 and Ru-Au/MgO serve as good illustrations for the above points (refs. 85-98). Ru and Au, similar to Ru and Cu, are elements that are for all practical purposes immiscible in the bulk state. Such bulk-immiscible systems are of great interest in catalysis, as they allow us to investigate the formation of bimetallic aggregates (sometimes also referred to as bimetallic clusters (ref. 1)) in a state of high dispersion. Two series of Ru-Au caralysts with varying amounts of Ru and Au and an overall metal loading of about 5 wt% were prepared by impregnation of S i 0 2 and MgO supports with aqueous solutions of RuC13.H20 and HAuC14.3H2O (refs. 81,85,86). The Ru-Au catalysts were subjected to a very extensive physical and chemical characterization routine and tested in various catalytic reactions, among them CO hydrogenation (refs. 89,91,98). In the SiO2 supported catalyst series, the activity dropped precipitously as the gold content of the catalyst increased. The decrease in activity was mainly atmbuted to a decrease in preexponential factor, while the apparent activation energy did not change much. Such a behavior is characteristic for a predominantly geometrical effect. Gold by itself is inactive for CO hydrogenation. Gold present on the surface of ruthenium could disrupt the Ru ensembles required for the reaction and thereby lower the specific activity of ruthenium. In the case of MgO supported Ru-Au, however, a very puzzling behavior was observed. As the Au content increased, the activity went through a maximum at intermediate Au content (Fig. 6.3). The explanation for this unexpected behavior of the Ru-Au/MgO system was provided by monitoring the genesis of bimetallic particles via temperature programmed reduction (ref. 88) and by the results of analytical electron microscopy. Careful analysis of the catalysts by transmission electron microscopy and energy dispersive X-ray spectroscopy (EDS)(refs. 89-91,98) revealed that a bimodal particle size distribution existed
239
in the Ru-Au catalyst series supported on Si02 All the ruthenium was located in small particles, less than 4 nm in diameter, along with a trace of gold. An estimation of the average surface composition of the small bimetallic particles based on a stepwise chemisorption and gas titration procedure (ref. 92) confirmed the presence of Au on the surface of these particles in agreement with Sinfelt’s model for the Ru-Cu system (ref. 99). The rest of the gold was present in large, monometallic particles having diameters between 23 to 45 nm. Table 6.2 summarizes the EDS results obtained on individual metal particles in a Ru-AulSi02 catalyst of intermediate gold content
(52 at% Au and 48 at% Ru). Exact quantification of these data cannot be performed since the X-ray count obtained from such small particle volumes is below the statistically significant threshold, and since there are no reliable calibration standards available for the bimetallic Ru-Au system. Nevertheless, the large fluctuations in characteristic X-ray signals for Ru and Au, even in particles of the same size, indicate that the composition of these particles must be very different from one particle to the next. The non-uniform distribution of the two constituent metals complicates the interpretation of global correlations of activity and gold content, such as the one shown in Fig. 6.3. In very small particles such as the ones found in the Si02 system, one could model a bimetallic particle containing about equal amounts of Au and Ru by simple surface decoration of a small Ru particle with up to a monolayer of Au or of a small Au particle with up to a monolayer of Ru. Electron microdiffraction performed on individual small bimetallic Ru-Au or Ru-Cu aggregates supports this notion, as the diffraction patterns could be ascribed to either Ru or the group Ib component without any significant perturbation of the structure of either one (refs. 95,100,234).
In the H2 reduced Ru-Au/MgO catalyst series which showed the unusual activity trend with increasing Au content, XPS indicated a surface enrichment of Ru (ref. 86). This was a surprising result in view of the thermodynamically predicted surface enrichment of the group Ib metal component. According to electron microscopy results, the MgO supported catalyst series had a much broader particle size distribution as compared to the Si02 series. Large monometallic Ru and AUparticles coexisted with large and intermediate size particles containing both metals. In contrast to the Si02 series where all the Ru was located in very small particles, the major contribution to the Ru surface area in the MgO supported catalysts was provided by particles in an intermediate size range of 4-10 nm (ref. 91). EDS proved that a large number of these particles contained significant amounts of both Ru and Au. Simple surface decoration of Au or Ru with up to a monolayer of the other element could not account for such significant amounts of both elements in one 4-10 nm particle. On the other hand, it is highly unlikely that these particles are alloys. Particles in this size range should for all practical purposes behave like bulk metal and one would not expect a violation of the thermodynamically dictated immiscibility of the two elements. Electron microdiffraction confirmed this, as it failed to show any unusual structures besides the regular Ru or Au diffraction patterns. Monometallic Ru supported on MgO showed suppressed CO hydrogenation activity compared to Ru metal or Si02 supported Ru (ref. 91). The presence of Au seemed to partially counteract this “negative support interaction” of Ru with MgO. It has been suggested that the
240
support effect of MgO is due to decoration of the Ru surface by MgO-derived species brought on by the phase transformation of Mg(OH)2 to MgO, creating electron-rich Ru sites (ref. 101). It is not entirely clear how the presence of Au interferes with the development of the MgO support effect on Ru. One might speculate that the partial restoration of the intrinsic activity of ruthenium is due to the fact that part of the Ru is supported on Au rather than on MgO. This hypothesis is supported by TPR results (ref. 88). In the case of MgO supports, Au was found to nucleate first, coinciding with the dehydroxylation of the MgO surface. After Au particles had started to deposit on MgO, the nucleation and reduction of Ru started. One could envision that some of the Ru got deposited on top of already formed Au particles. This would explain the formation of relatively large particles containing Au and Ru, with a surface enriched in Ru. Taking all the information obtained from extensive characterization and catalytic probe reactions into account, it appears that the medium-size particles on MgO are not true “bimetallic clusters” but probably phase-segregated Au particles with Ru particles deposited on their surface. However, the picture. is more complicated, as reduction of the Ru-Au/MgO system in hydrazine instead of hydrogen led to a bimodal particle size distribution reminiscent of that obtained in the Si02 series, with very small bimetallic Ru-Au particles. Hydrazine reduction is a very effective means to remove residual chloride from the MgO support, and it might very well be that differences in chloride contamination are responsible for the different end result in particle size distribution and microstructure (refs. 83,94). From TPR, we know that in the case of Ru-Au/SiOz, the nucleation of Ru and Au occurred simultaneously. The coincidence of Ru and Au nucleation in the Si02 supported catalyst series and perhaps also in the hydrazine reduced Ru-Au/MgO series might hold the key to the formation of small bimetallic particles. The lesson to be learned from the Ru-Au system is that one cannot always be certain that in bimetallic catalysts the two metal components are uniformly allocated to particles differing in size and structure. There are many other bimetallic catalyst systems where a careful analysis of this particular aspect is overdue. An extensive characterization strategy with special emphasis on analytical electron microscopy could help to develop a better understanding of many unresolved questions in the bimetallic catalysis literature. The Pt-Ru system serves as another good example for the role played by preparative variables on the surface composition of supported catalysts containing two group VIII metals. CO hydrogenation over Pt-Ru catalysts has been extensively studied by several research groups (refs. 63,102-107,110-115). Coimpregnation of Si02 or A1203 with HzPtCI, and RuC13 was found to give catalysts with major differences in particle morphology and surface composition. On SiO;! a significant surface enrichment in Pt was observed, while the ratio of Pt/Ru surface atoms appeared to be much lower when A1203 was used as support (ref. 105) (Fig. 6.4). These differences in surface composition were rationalized as follows: the 1PtCld 2- precursor anion adsorbs strongly on alumina via ligand exchange with amphoteric OH- groups. On silica such amphoteric groups are not available and the surface complex resulting from adsovtion of [PtCld2- is highly mobile. Due to the higher Brgnsted acidity of SiO, it will interact with R u + ~ cations much stronger than with [PtCl# anions (ref. 117). Thus, one could justify an initial nucleation of Ru particles whose
surface would then be partially covered by the more mobile Pt species. On alumina, both the Ru and Pt precursor complexes appeared to have comparable mobilities, and the surface composition seemed to depend in a much more intricate way on dispersion and overall composition Nominal Pt/Ru Ratio in Catalyst (refs. 106,110). Fig. 6.4 Ratio of Pt/Ru surface atoms as a function of support and nominal Based on X-ray Pt/Ru ratio (adapted from ref. 105) diffraction, conflicting results concerning the formation of bimetallic Pt-Ru aggregates have been reported. At high metal loading, Miura et al. (ref. 106) observed a new X-ray diffraction line which they attributed to the formation of bimetallic particles. In contrast to these results, Chakrabarty et al. failed to see any evidence for coclustering of Ru and Pt (ref. 111). Analytical electron microscopy confirmed the existence of bimetallic particles, but indicated non-uniform distribution of the two metal components depending on particle size and preparation conditions (refs. 102,108). Furthermore, bimodal particle size distributions were found (ref. 108), similar to the findings on the Ru-Au system (refs. 89-91). Ru-Rh/SiOZ catalysts (ref. 116) may serve as additional illustration of the effect of preparative variables on the structure of bimetallic catalysts. The drying steps and the preoxidation prior to reduction had great influence on the characteristics of bimetallic Ru-Rh catalysts. Heat drying of the impregnated SiO2, preoxidation and reduction gave catalysts with poor dispersion containing segregated metal particles. Highly dispersed bimetallic “clusters” were formed when vacuum drying at room temperature was employed followed by direct reduction. In the bimetallic “cluster” catalysts addition of more than 20 percent Rh caused a significant decrease in methanation activity, supporting the notion that Ru ensembles of a certain size are needed to effectively catalyze the methanation reaction. The picture emerging from these examples is that the surface composition of bimetallic supported catalysts is not only determined by the relative heats of sublimation of the two metal components, but even more so by the details of the preparative chemistry and the strength of interactions between support and precursor ions. As an additional factor, the influence of support
242
surface morphology should be taken into consideration, although the data base on this subject is rather limited.
6.4 THE INTERACTION OF CO WITH BIMETALLIC SURFACES Over the years, a great deal of information has accumulated on the reaction pathways and mechanisms leading to the various products obtained in CO hydrogenation. It is by now well established that the main pathway for hydrocarbon formation includes dissociative chemisorption of CO. The route leading to methanol is favored by nondissociative chemisorption of CO. The synthesis of C2+ oxygenates appears to be initiated by dissociative CO adsorption followed by CO insertion into growing alkyl chains. It is quite obvious that adding a second metal component could have a significant influence on the CO adsorption mode, and thus contribute to a change in selectivity. It has also to be kept in mind that bimetallic systems can undergo surface reconstructions and adsorption induced surface enrichment processes (ref. 109). Infrared spectroscopy proved to be a powerful tool for studying the adsorption of CO on bimetallic catalysts and to assess the relative importance of ligand versus ensemble effects due to alloying. In many cases, infrared bands appearing at wavenumbers lower than 2000 cm-1 can be attributed to CO species which are coordinated to more than one surface metal atom in a bridged bonding mode. CO linearly bonded to metal atoms tends to give bands at wavenumbers higher than 2000 cm-l. In addition, the CO band positions and intensities tend to be very sensitive to the degree of reduction of the metal. Bimetallic systems containing a group VIII and a group Ib metal offer the advantage that CO gas adsorbs strongly only on the group VIII component, while CO can easily be removed from the Ib metal sites. This makes spectral interpretation relatively straightforward. The first studies of this kind included the Ni-Cu system (refs. 118,119), the Pd-Ag system (refs. 120,121), and the Ru-Au system (ref. 87). The Pd-Ag system represents a classic example for the predominance of the ensemble effect. On monometallic Pd and on Pd-Ag alloys three bands were found at about 2060,1960 and
1920 cm-I. With increasing Ag content, the spectral features below 2000 cm-l assigned to bridged-bonded CO on Pd lost intensity, while the CO band attributed to a linear Pd-CO surface complex increased in intensity (ref. 120). The relative amounts of the bridged and the linear CO complexes depended on the geometry of the alloy surface. Ag or Cu addition appeared to decrease the number of sites containing two adjacent group VIII metal atoms needed for adsorbing CO in the bridged-bonded configuration. For linear bonding of CO, a single group VIII metal atom might have been sufficient, so that the linear-bonding mode was not disturbed as much when Ag or Cu were added. Studies of CO adsorption on model single crystal surfaces modified with group Ib elements are of great help in elucidating the effect of a second metal component on the adsorption characteristics. CO adsorption studies on Ru single crystal surfaces covered with various amounts of Cu serve as good illustrations (refs. 65,67,122-126). On Ru, CO adsorbs very strongly with an
243
initial heat of adsorption of about 160 kJ/mol (ref. 127), while the adsorption on
cu surfaces is
rather weak (45 kJ/mol) (ref. 128). These pronounced differences in adsorption energies make it possible to distinguish between strongly and weakly adsorbed CO by simply varying the adsorption temperature (ref. 65). It also makes the Cu-Ru system attractive for theoretical studies (ref. 129). In general, addition of submonolayer amounts of Cu had no drastic effect on modifying the C o uptake on Ru(0001). This is not surprising in view of the preferred terminal type of CO bonding on Ru. The deposition temperature was found to have a pronounced influence on the morphology of the Cu overlayer. Deposition at low temperature favored a Stranski-Krastanov or Volmer-Weber type island growth mechanism, while Cu deposition at higher temperatures resulted in a more uniform spreading of Cu and a layer by layer growth according to the Frank van der Menve mechanism (ref. 122). The effectiveness of Cu to block CO adsorption sites on Ru depends on the Cu location with respect to the Ru surface (ref. 126). Deposition of Au on a Ru(0001) surface followed similar trends giving island growth at lower temperatures and more pronounced spreading and lateral dispersion of Au atoms at higher Ru substrate temperatures (ref. 130). The lateral dispersion of Au following high temperature deposition suppressed the uptake of CO on Ru more effectively than in the low-T films. Given the dependence of the surface structure on the preparative conditions in single crystal studies, it does not surprise us that supported Ru-Au catalysts can exhibit massive differences in surface composition and microstructure as a consequence of changes in catalyst preparation (refs. 85-98). While Cu-Ru and Au-Ru are examples for bulk immiscible bimetallic systems, metals that are partially or completely miscible have also found much attention in the surface science literature, for instance, the Cu-Ni system (refs. 109,131).When CO is adsorbed on Ni, both bridge and top site CO stretching vibrations can be observed. CO adsorption on Cu at room temperature is negligible, so that CO represents a good probe molecule for the effect of Cu on the adsorption characteristics of Ni. Depending on the details of metal deposition and equilibration, various degrees of interdiffusion of the two metal components can lead to large differences in the surface composition. The step and defect concentration on the surface will have great influence on the intermixing of the two metal components and 2-dimensional island formation. This considerably complicates the interpretation of CO adsorption data on Cu-Ni reported by the various research groups. The situation becomes even more complex when both metal components are capable to chemisorb CO with appreciable strength. Two scenarios for the mode of CO adsorption have to be considered: both metal components could have the same mode of CO adsorption, or there could be dissociative adsorption on one metal, while the other metal would adsorb CO associatively. The interpretation of vibrational spectra of adsorbed CO on such systems is a challenging task. Examples for this type of bimetallic catalyst are Ru-Pt (ref. 132), Pt-Pd (ref. 133), Ni-Ru (ref. 1341, and Ru-Rh (ref. 135). In the simplest scenario, the infrared spectra might represent simple superimposition of the CO bands of the two constituent metals. However, in most cases shifts in band position, changes in extinction coefficients, and the appearance of new spectral features will make the interpretation a formidable challenge. One could envision differences in relative surface coverages by adsorbed CO, caused by the presence of the second metal and the
244
details of the surface geometry, changes in bonding mode from bridged to linear due to ensemble effects, dipole interactions, and so forth. Grill et al. (ref. 133) suggested that the decrease in intensity of the bridged Pd-CO bands relative to the linear Pd-CO bands in Pt-Pd may be due to ligand effects favoring the linearly adsorbed species over the bridged-bonded species. However, geometric effects and short range ordering could not be excluded. In a bimetallic Ni-Ru model system prepared by evaporating Ni onto a Ru(0001) surface (ref. 134) bonding of CO appeared to be stronger than for either of the two metal components alone indicating considerable electronic perturbation of the Ni overlayer.
6.5 THE INTERACTION OF HYDROGEN WITH BIMETALLIC SURFACES Besides being a reactant in CO hydrogenation, hydrogen is also important as a reducing agent. Interactions of hydrogen with bimetallic catalysts play a decisive role in determining the resulting microstructure and composition of bimetallic aggregates. As Christmann has pointed out in his review of hydrogen sorption on pure metals (ref. 136) there are several different processes governing the interaction of molecular hydrogen with metal surfaces: associative adsorption, dissociative adsorption and formation of atomic hydrogen on the surface, absorption of hydrogen atoms into the subsurface or bulk of the metal, and surface or bulk hydride formation. In a supported bimetallic system, additional possibilities exist: migration of hydrogen atoms from metal A to metal B which by itself would not be able to dissociate molecular hydrogen, and spillover from the metal to the support. There is strong evidence supporting the notion that hydrogen adsorption requires ensembles of several adjacent surface atoms in the proper configuration. In bimetallic systems, these surface ensembles may become disrupted by the second metal component, especially in cases where the second metal is not vely effective in hydrogen chemisorption. In general, adsorbed hydrogen is much more mobile than adsorbed carbon monoxide. In cases where atomic hydrogen can migrate to a second, inactive metal component, problems may arise with using hydrogen adsorption as a tool to determine the dispersion of the active component as the uptake of hydrogen will be much larger than warranted based on the number of surface sites of metal A. On the other hand, one could envision a scenario where the inactive component B blocks part of the surface of metal A, thereby rendering some of the surface ensembles of metal A inactive for dissociation of molecular hydrogen. This would lead to suppression of hydrogen chemisorption. Which one of the two scenarios predominates will depend on the surface structure of the bimetallic system and the relative interdispersion of the two metals. A case in point is the Ru-Cu system which has been the subject of a great number of investigations, both in supported form as well as in the form of Ru single crystals covered by Cu. Ru chemisorbs molecular hydrogen dissociatively with a chemisorption density of approximately 1 hydrogen atom per Ru surface atom (ref. 137) while pure Cu does not dissociate molecular hydrogen (ref. 138). On SiO, supported Ru-Cu catalysts, some research groups reported that CU decreased the capacity of Ru for hydrogen chemisorption (refs. 95,99,139) while others found little evidence for Ru-Cu interactions and in some cases even an increase in H2 chemisorption
245
(refs. 140,141). Evidence for spillover form Ru to Cu was also found in NMR experiments on Ru-Cu/Si02 catalysts (ref. 142). Similar discrepancies can be found in the single crystal literature. On single crystal Ru(0001) surfaces, Ertl and coworkers found that the addition of Cu significantly lowered the capacity for H2 adsorption (refs. 64,66) in contrast to Yates et al. who attributed the effect of Cu to a simple site blocking mechanism (ref. 69). There are also reports of increased uptake of hydrogen due to spillover from Ru to Cu sites (ref. 143). As in the case of CO adsorption, the discrepancies in hydrogen adsorption behavior might be due to differences in the interdispersion of Ru and Cu and differences in the size and configuration of undisturbed Ru ensembles. Recent reviews dealt with the adsorption of hydrogen and its role in CO hydrogenation over monometallic group VIII base metal catalysts (refs. 144,145). Whether or not hydrogen facilitates the dissociation of CO is still a matter of debate. It is, however, widely accepted that the hydrogen surface coverage relative to that of CO or other carbon containing surface species is a very important factor in determining activity and selectivity. To determine surface coverages on a working catalyst under reaction conditions is a very difficult proposition and becomes even more intractable for supported bimetallic catalysts.
For CO hydrogenation catalysis, the coadsorption of hydrogen and CO and the mobility of the resulting surface species needs to be considered. Such coadsorption studies on pure single crystal surfaces have given valuable insight. For example, on Ru(0001) surfaces both gases appeared to block adsorption sites for the other species, with hydrogen being less effective than CO in blocking the surface (ref. 146). On other metal surfaces such as Ni(100) (refs. 147-149), a common hydrogen and CO state was only observed when H2 was added first. Preadsorption of CO prevented the subsequent dissociative adsorption of H,. Addition of a second metal component could certainly change the coadsorption behavior of
CO and H2, especially in highly dispersed supported bimetallic systems. Unfortunately, the available data base on this subject is rather scarce. A very interesting approach to this problem has been taken by Gryaznov et al. (ref. 150) who studied the behavior of Pd-Ru, Pd-Ni and Pd-Co alloys in form of thin-walled tubes. This type of Pd membrane reactor allowed them to vary the concentration of adsorbed hydrogen species by diffusion through the palladium membrane. Studies with bimetallic model catalysts are of great value in this context. For example, Christmann and Ertl (ref. 65) found that the presence of Cu atoms decreased the uptake of strongly chemisorbed hydrogen on Ru, whereas the adsorption of CO was less affected. The effect of Cu appeared to depend o n the lateral dispersion of Cu. It is not surprising that disruption of Ru ensembles by inactive Cu would affect hydrogen adsorption more strongly as compared to CO adsorption. Dissociative hydrogen adsorption on Ru is much more demanding on ensemble size and configuration than CO adsorption. In supported Ru-Cu catalysts one would expect that the microstructure and surface composition of the individual particles will similarly affect the adsorption behavior. Even in systems which show relative uniform dispersion of metal particles, the composition can vary significantly from one particle to the next, as shown by Shastri et al. for Ru-Cu/Si02 (ref. 100). Okuhara et al. also acknowledged that such a wide variety of structures and
246
adsorption sites might explain the differences between results obtained on Cu-Ru(0001) single crystals and Cu-Ru/SiO, (ref. 154).
6.6 CO ACTIVATION OVER BIMETALLIC CATALYSTS TO assess the influence of a second metal component on the activity of a metal which has proven activity for CO hydrogenation, a basis for normalization of reaction rates in form of turnover frequencies is essential. Having gone through the above discussion of the complexity of bimetallic supported catalyst systems, we realize that it will in many cases be difficult to accurately quantify the dispersion of the two metal components and to determine the number of surface sites. This state of affairs will introduce a great deal of ambiguity into kinetic analysis. Comparisons based on apparent activation energies and preexponential factors have also to be taken with a grain
of salt, as they reflect only the average state of the catalyst. For systems containing broad or bimodal particle size distributions and non-uniform particle compositions and surface structures, global activity comparisons based on bulk composition can be outright misleading. Bimetallic CO hydrogenation catalysts can be classified into two categories: 1) systems with one active and one inactive component, and 2) systems where two metals are combined which have CO hydrogenation activity in their own right. We will first discuss case studies falling into the first category, and then move on to cover examples for the second category. We have already mentioned in our overview of catalyst preparation techniques that addition of Cu is often used to enhance the reduction of Fe catalysts (ref. 18). The milder reduction conditions preserve higher surface areas of Fe-based Fischer-Tropsch catalysts. The presence of Cu might also have a beneficial effect on the formation of iron carbides (ref. 151). Alloying of iron with Cu has been reported to suppress the activity of iron (refs. 151,152). In terms of selectivity, however, Cu does not seem to make a significant difference except for a slight increase in selectivity for C,, products (ref. 152). In a detailed characterization study of bimetallic Fe catalysts containing either Cu or Ag, it was found that the reduced and passivated iron catalysts contained a-Fe, a trace of Fej04 and some metallic Cu or Ag (ref. 153). X-ray diffraction did not reveal any evidence for alloy formation between iron and the Ib metals. XPS showed that upon reduction the Ag/Fe ratio increased while the C u F e ratio decreased. This might suggest system specific, different degrees of interaction between Fe and the Ib metals. The differences between Ag and Cu were rationalized in terms of different affinities for iron oxide: copper was partially oxidized and in a state of high dispersion on the iron oxide surface. The intimate contact and the high reducibility of copper oxide could have facilitated the reduction of iron oxide. Silver, on the other hand appeared to be present in form of poorly dispersed metallic silver which did not properly “wet” the iron oxide surface. Therefore, Ag did not significantly enhance the reduction of iron oxide. Exposure to CO + H, under reaction conditions decreased the Cu and Ag XPS signals further. These changes were attributed to the formation of iron carbides and the decreased affinity of Ag and Cu to interact with these surface carbides. The limited interaction with iron under Fischer-Tropsch reaction conditions might be the reason for the lack of selectivity differences compared to pure iron catalysts. In industrial type Fe-Cu catalysts, it is more difficult to single out the effect of Cu, as in
247
most cases alkali promoters are added which further influence activity and selectivity (refs. 155,156). Addition of a group Ib metal such as Cu to Ni or Co tends to reduce the overall CO hydrogenation activity of the group VIII metal (refs. 157-159,162,163). Ponec and coworkers interpreted their results in terms of an ensemble effect where Cu addition drastically decreased the number of Ni sites where CO can adsorb and form active surface carbon species, thereby affecting both methane and C2+ production (ref. 157). Since C2+ production decreased somewhat less than the methane production, Cu addition resulted in a slight increase in C2+ selectivity. Luyten et al. (ref. 158) and Dalmon and Martin (ref. 170), on the other hand, found that the selectivity for higher hydrocarbons decreased when Cu was added. They confimied, however, the geometric effect due to dilution of Ni ensembles by Cu. The discrepancies concerning C2+ selectivity could be due to different degrees of surface equilibration in the various catalysts. This would not be surprising in view of the fact that data on films and powders are compared with data obtained on supported catalysts. In contrast to the bulk miscible Cu-Ni system, the question of bimetallic particle formation becomes important in the bulk immiscible Ru-Cu system. The CO hydrogenation activity of the
Ru-Cu system has been investigated by several research groups (refs. 158,160,161,164-167). Bond and Turnham (ref. 164) found that the activity decreased with increasing Cu content, while the activation energy for methane formation remained more or less constant at 88 kJ/niol. The decrease in activity was attributed to a change in the preexponential factor. It was postulated that the active site on pure Ru is an ensemble of about 4 or 5 atoms, the central one bonding a CHOH complex and the surrounding Ru atoms providing sites for hydrogen atoms. Progressive dilution of Ru with Cu atoms would reduce the average number of Ru atoms surrounding the CHOH surface intermediate and decrease the number of H atoms able to participate in the rate determining step. Addition of Cu
also seemed to change the selectivity towards the formation of higher hydrocarbons. Bimetallic Ru-Cu catalysts produced initially more methane than monometallic Ru (ref. 164). However, the selectivity trends were not entirely conclusive as the low CO partial pressures might have caused some contribution of hydrogenolysis reactions. Luyten et al. (ref. 158) found a similar drop in activity when Cu was added to Ru, but pointed out that methanation and Fischer-Tropsch reactions were affected in different ways. It appeared that the synthesis of higher hydrocarbons was more sensitive to the introduction of Cu than the methanation reaction. Lai and Vickernian estimated that an ensemble of 4 to 6 Ru atoms is needed for disproportionation of CO, while hydrogenation requires a larger ensemble of about 9 to 13 adjacent Ru atoms (ref. 165), an ensemble similar in size to that estimated for Ni in Ni-Cu alloys (ref. 170). Results obtained on Cu-Ru catalysts supported on zeolites (refs. 160,161) are more difficult to interpret due to the possible contribution of the zeolite support. King et al. (ref. 166) reported that the decrease in the rate of CO hydrogenation was proportional to the decrease in the number of Ru atoms on the surface, in essence invoking a simple site blocking mechanism. A recent study of alumina supported Ru-Cu catalysts (ref. 167) reports that addition of Cu decreased the activity toward hydrogen adsorption and increased the activity
248
towards CO adsorption. Besides lowering the overall activity for CO hydrogenation, Cu also seemed to affect the selectivity, increasing the yields of CzH4 and C2H6 at the expense of CH4 The key to reconciling these different results lies in the realization that the microstructure and surface composition of these bimetallic catalysts might be very different, depending on the preparation and reaction conditions. If Cu is present in form of islands on Ru, one would expect a behavior controlled by a site blocking mechanism. In essence, there would be relative large patches of Ru left which would behave similar to pure Ru. If, on the other hand, individual Cu atoms or very small aggregates of a few Cu atoms are intimately interdispersed all over the Ru surface, then one would expect a significant breakup of Ru ensembles with drastic consequences for the activity. If one accepts the proposition that Fischer-Tropsch reactions require larger ensembles of Ru than methanation, then the presence of highly interdispersed Cu would make its influence felt also in terms of selectivity. This hypothesis, however, does not explain the recently reported surprising increase in C2 yields in the presence of Cu (ref. 167). A note by Van Barneveld and Ponec (ref. 168) might be able to shed light on this issue. They pointed out that adding Cu to Ru or Ni can cause either an increase or a decrease in Cz+ selectivity depending on the temperature. At higher temperatures, alloying increases the selectivity for higher hydrocarbons, perhaps, as suggested by Biloen, due to the Cu induced suppression of secondary chain splitting reactions (ref. 169). The lower C2+ selectivities usually encountered at lower temperatures might according to Van Barneveld and Ponec (ref. 168) be related to the lower surface concentration of CH, species on bimetallic surfaces. Clearly, more work is needed to clarify this question. The above mentioned results on Ni-Cu and Ru-Cu and the previously discussed results o n Ru-Au (refs. 89,91,98) are consistent with a model where the group Ib metal dilutes the active Ru
ensembles. Interestingly, the activity trends as a function of group Ib content in methanation are remarkably similar to those found in the structure sensitive hydrogenolysis of ethane (refs. 91,164).
To account for this similarity in activity trends, one could envision that the presence of highly dispersed Au or Cu creates an apparent dispersion effect. It is well known that the methanation activity of pure Ru declines when the dispersion of Ru gets very high (refs. 171,172). It is conceivable that Cu or Au addition breaks the Ru surface up into ensembles having a size or configuration similar to those in highly dispersed monometallic Ru particles. Ligand effects appear to be of lesser importance. Selectivity effects of Cu or Au, if noticed at all are not very strong, and the question remains if they cannot be explained by contributions from secondary reactions such as hydrogenolysis and hydrogenation, or dispersion effects. For example, by simple varying the dispersion of monometallic Ru catalysts, larger changes in the olefin/paraffin ratio could be achieved than by adding Au (ref.91). Boudart and McDonald have reviewed the question of structure sensitivity of CO hydrogenation over group VIII metals (ref. 173). For Ru, Ni and Fe alloying, particle size and specificity effects provide a strong argument for structure sensitivity of the Fischer-Tropsch synthesis. For methanation, the situation becomes somewhat more ambiguous. Over Ru and Fe, methanation appears to be sensitive to alloying and particle size effects. Ni, on the other hand, does not show significant particle size effects under selective methanation conditions, although alloying effects are still noticeable. Single crystal studies on Ru and Ni, however, suggest
249
that methanation is not a structure sensitive reaction in a strict sense (refs. 174,175). While ethane hydrogenolysis exhibited drastic changes in activity over different crystallographic planes of Ru or Ni, no significant change in methanation turnover number has been observed. Let us now consider some examples for bimetallic catalysts where both metal components are active for CO hydrogenation. Iron-based bimetallic catalysts are of special interest in view of the commercial importance of Fe as Fischer-Tropsch catalyst. It is well known that under Fischer-Tropsch reaction conditions, iron is converted into a mixture of carbides and oxides. Ru being also very active for methanation and Fischer-Tropsch reactions makes the bimetallic Fe-Ru system very attractive for CO hydrogenation. Fe is more effective in dissociating CO, while Ru has a higher affinity for hydrogen adsorption. One would therefore expect that a combination of these two metals would give very interesting CO hydrogenation behavior. The bimetallic Fe-Ru system has been extensively studied by many research groups (refs. 27,28,30-39,43,176-190). At very low Ru content (<4.5 mol%) Ru forms a substitutional solution with iron, followed by a two phase region between 4.5 and 24.5mol% Ru and a hcp structure at Ru contents higher than 24.5 mol% (ref. 191). In an XPS and SIMS study of Fe-Ru alloys in powder form, Ott et al. (ref. 43) found that the reduced catalysts showed surface enrichment of Fe. These alloys gave high propylene and ethylene yields compared to pure Fe or Ru, in agreement with the higher olefin yields observed over supported Fe-Ru catalysts (refs. 177,189). However, the catalysts were rapidly deactivated, probably due to carbon build-up. Surface enrichment was also observed in Fe-Ru catalyst prepared from mixed-metal carbonyl clusters (ref. 39). The selectivity of the bimetallic cluster catalysts, however, was different from conventionally prepared Ru-Fe catalysts. With increasing Ru content, the methane selectivity increased, and the olefin/paraffin ratio decreased. Catalysts prepared from R u ~ ( C O12 ) + Fe3(CO)12 instead of mixed-metal clusters behaved more like conventional catalysts. The presence of Ru seemed to facilitate the reduction of Fe. Guczi and coworkers (ref. 187) confirmed the enhancement of olefin production due to Fe addition in a study of S O z and A1,0,
supported Fe-Ru catalysts.
TPD experiments indicated that on the bimetallic catalysts at high temperatures an activated form of hydrogen predominated. While weakly bonded hydrogen should favor the formation of methane, one would expect that a hydrogen depleted surface, such as the one on bimetallic Fe-Ru catalysts, would give a higher olefin/paraffin ratio. Changing the surface composition and structure of a catalyst by adding a second metal component offers an opportunity to modify the population of differently bonded hydrogen and carbon species (refs. 187,189). The activity of bimetallic Ru-Fe catalysts can, therefore, exceed that of pure iron. Bimetallic Fe-Ru catalysts tend to deactivate at a much lower rate than pure Fe (ref. l89), as the presence of Ru and a larger pool of active hydrogen species prevents the formation of bulk carbide structures. Combining iron with cobalt is an interesting proposition, since both metals represent classic Fischer-Tropsch catalysts (refs. 151,203-213). Fe has a tendency to form oxides and carbides under reaction conditions, and precarbiding increases its activity. Co, on the other hand, has little affinity for oxide and carbide formation under reaction conditions. The difference in carbide formation might be related to the higher activity of Fe surfaces for CO dissociation at elevated temperatures,
250
as compared to Co (refs. 201,202). Stanfield and Delgass (ref. 203) reported interactions between Fe and Co and the formation of disordered bimetallic aggregates when the metals were supported
on Si02. The reducibility of Fe was increased and carbide formation was retarded as a function of Co loading. Under certain reaction conditions, bimetallic Fe-Co catalysts can give enhanced C, and C3 selectivities, possibly due to differences in the relative population of the surface with activated hydrogen and carbon species as compared to pure Fe or Co. According to Amelse et al. (ref. 151) alloying Fe with Co resulted in high selectivity for olefins, high shift activity, and the tendency to incorporate olefins into growing hydrocarbon chains. Similar arguments concerning surface coverage by activated hydrogen and carbon species apply for bimetallic Fe-Ni catalysts (refs. 208,212,214). Jiang et al. (ref. 215) found that a significant fraction of the iron was alloyed with nickel after reduction in H2, and the balance of the Fe was present as Fe2+ species interacting with the support. On titania, Fe was easier to reduce than on alumina. During methanation, part of the iron in the Fe-Ni alloy became oxidized to Fe2+. Under SMSI conditions, the titania supported Fe-Ni catalysts exhibited higher methanation activity and an increased selectivity towards higher hydrocarbons. Also, the deactivation characteristics appeared to be better (refs. 215,216). When Fe3(C0)12 was added to Pt/A1203, a marked increase in CO hydrogenation activity was observed along with high selectivity to low molecular weight alkenes (ref. 198). The formation of bimetallic sites was attributed to the mobility of Fe3(C0)12 under reaction conditions, resulting in a Pt surface partially covered with Fe. Evidence for interaction between Fe and Pt was also obtained by Mossbauer spectroscopy (ref. 199) and by magnetic and infrared measurements (ref. 200). Bimetallic catalysts containing Rh are very interesting in view of the ability of Rh to produce oxygenated products. Rhodium, when operated at high pressure, produces acetic acid, acetaldehyde, and ethanol along with methane. Addition of iron decreased the yields of acetic acid and acetaldehyde, and methanol and ethanol became the major products (ref. 192). Mossbauer spectroscopy indicated that after reduction in H, at 725 K, most of the Fe was still oxidized, while the rest formed an alloy with Rh (refs. 193,194). Based on a high pressure infrared study MeO, EtO-, and acyl-surface species bonded to Rh and/or Fe were proposed as reaction intermediates (ref. 195). On supported Fe-Rh, Fe-Pt and Fe-Pd catalysts derived from bimetallic carbonyl clusters, strikingly higher activities and higher selectivities for the production of Cl and C2 alcohols were observed (ref. 196). The cluster-derived Fe-Rh catalysts had a higher propensity for alcohol formation than conventional Fe-Rh/Si02 catalysts prepared by coimpregnation of Fe and Rh chlorides. EXAFS indicated the presence of highly dispersed bimetallic Fe-Rh particles having direct Rh-Fe-0 bonding. Mossbauer data suggested that Fe+3 was the most abundant Fe species even after reduction in H2 at 673 K, with the rest