New Trends in Co Activation (Studies in Surface Science and Catalysis)

Studies in Surface Science and Catalysis 64 NEW TRENDS I N CO ACTIVATION This Page Intentionally Left Blank Studie...

Author: L. Guczi

19 downloads 1265 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

DOWNLOAD PDF

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 of Fe in the zero valent state. This would imply that the Fe3" species are located at the metal-Si02 support interface. These ionic iron species could then serve as stabilizing agents for Rh, Pt or Pd and prevent sintering. The Fe promotion for oxygenate formation was attributed to enhanced migratory insertion of CO with M-H and M-alkyl. In Fe-Pd, the Mossbauer spectra showed drastic changes as a function of the FePd ratio which

251

tracked the catalytic performance in CO hydrogenation (ref. 197). At low Fe ratios, reduced Fe species appeared to be uniformly dispersed in Pd, and the catalysts had high selectivity for oxygenates. At FePd ratios greater than 0.3, FePd, Pd3Fe and a-Fe phases were formed. In the latter case hydrocarbon product distributions typical for Fe catalysts were found. It is known that oxygenated compounds can also be produced over supported Rh catalysts promoted with Mn, Zr, Ti or other additives (refs. 192,218). Fe-Pd and Fe-Ir catalysts also show good selectivity for oxygenates (refs. 219,220). In MgO promoted Fe-Pd/Si02 catalysts, iron suppressed the formation of the p-PdH phase and MgO appeared to be beneficial in terms of stabilizing the high dispersion of Pd-Fe particles (ref. 221). Addition of Fe to Pd resulted in significant enhancement of methanol formation. This increase was explained in terms of a decrease of the electron density of Pd due to alloying. A bifunctional catalytic mechanism was proposed in which Pd provides activated hydrogen, while bimetallic PdxFe particles provide the sites for CO surface species and methanol formation (ref. 222). In CoIr/AI2O3 catalysts (ref. 217) small amounts of Ir were found to enhance the formation of higher hydrocarbons. Further addition of Ir, however, caused a drastic change. Besides a decrease in rate and activation energy, methane became the main product. This behavior is characteristic for Ir dominating the catalytic function at high loading. The results were interpreted

in terms of a model where Ir would provide weakly bonded hydrogen while reactive carbon is mainly formed via CO dissociation on Co. Adding Ir to Co would then decrease the relative amount of reactive carbon with the consequence of decreased production of higher hydrocarbons. The availability of weakly bonded hydrogen could account for the decreased olefin formation. Recently, it has been reported that bimetallic catalysts containing iridium such as Ir-Ru/Si02 (ref. 222), Ir-Co/Si02 (ref. 223) and Ir-Mo/SiOz catalysts (ref. 224) do have potential for the selective production of C2-oxygenates. On Rh-Mo/AI2O3 catalysts high selectivity to oxygenates was attributed to a dual-site mechanism, where CO is activated by Rh, while Mo5+ sites activate H 2 which subsequently would migrate to the activated CO complex (ref. 225). In high pressure CO hydrogenation over SiOz supported bimetallic catalysts including FeRu, FePd, FeIr, FePt, Colr, Copt, NiIr and NiPt, methane was the main product initially. However, high methanol selectivities started to develop during the first 10-40 hours on stream (ref. 226). In the case of FeIr/Si02, an active site consisting of iron cations and metallic iridium was proposed. With time on stream, the iron cation concentration seems to increase, paralleling the onset of high methanol selectivity.

6.7 EFFECT OF SECOND METAL COMPONENT ON

CATALYST DEACTIVATION The poor activity maintenance of monometallic catalysts under syngas reaction conditions can be a problem. Carbide formation, although important for the CO hydrogenation activity of iron and nickel catalysts, is often accompanied by the growth of inactive, filamentous carbon. These carbon filaments are very stable and can lead to severe problems, such as pore plugging and structural damage to the catalyst (ref. 227). Also, graphitic deposits can block the surface sites and thereby deactivate the catalyst.

252

On Ni and Ru, carbon deactivation is relatively slow and to a large extent reversible. On Fe and Co, on the other hand, bulk carburization and graphitic deposits on the surface can lead to rapid and irreversible deactivation (ref. 228). The latter two metals, therefore, could benefit from the addition of a second metal component that is less susceptible to carbon deactivation, such as Ru. Small amounts of sulfur in the feed stream can have severe consequences for CO hydrogenation. On most metals, the methanation activity is decreased by about three orders of magnitude when sulfur is present (refs. 228,229). Sulfur poisoning causes mainly a geometric blocking of active surface sites, analogous to the effect of Ib metals. Electronic effects, if present, do not seem to make a significant contribution. Once poisoned by sulfur, it is very difficult to remove surface sulfides due to their high stability as compared to bulk sulfides. While most of the typical CO hydrogenation metals are very sensitive to even trace amounts of sulfur, Mo based catalysts show great potential as sulfur tolerant catalysts. Sulfur has to be removed anyway either upstream or downstream of the catalyst to satisfy environmental constraints, and with this in mind Dry questioned the wisdom of developing sulfur tolerant Fischer-Tropsch catalysts (ref. 18). However, the development of a sulfur tolerant catalyst may still be attractive for reasons of process reliability. A transient increase in sulfur compound concentration due to upstream failure of desulfurization would not severely impact on the function of a sulfur tolerant CO hydrogenation catalyst, while a conventional Fischer-Tropsch catalyst would need regeneration or replacement to recover from the transient exposure to sulfur. Bimetallic Mo-Fe-S or Mo-Co-S which were derived from sulfido clusters were found to be totally immune to

15 ppm H2S sulfur in the feed stream. Neither the activity nor the selectivity were affected (ref. 60). The sintering of metal particles under reaction conditions is another reason for loss of activity in CO hydrogenation. It is conceivable that the presence of a second metal component in an appropriate microstructure might influence the sintering characteristics. This aspect of stabilizing the dispersion of bimetallic CO hydrogenation catalysts, however, has not been explored in great detail. The presence of a second metal component can also modify via an ensemble effect the rate of self-poisoning reactions. Guczi and Schay (ref. 230) have suggested that the previously discussed behavior of the Ru-Au system (Fig. 3 ) (refs. 86,89-91) could be explained by such an effect. Their interpretation was based on the strikingly similar activity trends as a function of Au content for both ethane hydrogenolysis (ref. 89) and methanation (ref. 91). Although an effect of Au on the rate of self poisoning of Ru cannot be entirely excluded, closer inspection of the deactivation behavior shows that there is no systematic correlation between relative rates of deactivation and the observed activity trends (ref. 91). This points to a geometric ensemble effect of Au as the main cause for the change in specific activity of Ru for methanation and Fischer-Tropsch. It appears that the CO hydrogenation responds to gold induced changes in Ru ensemble size in a manner very similar to that in the ethane hydrogenolysis.

253

6.8 NEW REACTION PATHWAYS IN CO ACTIVATION Among the commonly accepted mechanisms for CO hydrogenation are the carbide mechanism and the CO insertion mechanism (ref. 231). The carbide mechanism assumes that CO and hydrogen adsorb dissociatively, followed by the hydrogenation of active surface carbon species. This mechanism readily accounts for the formation of methane, higher hydrocarbons and C02. It does not, however, explain the formation of oxygenated products. The CO insertion mechanism starts form associative adsorption of CO and provides a good explanation for the formation of oxygenates. This mechanism, however, does not directly account for C02 production. The latter is usually attributed to the participation of the water gas shift reaction. Addition of a second metal component to conventional CO hydrogenation catalysts may have some influence on selectivity through changes in ensemble size, relative surface coverage by hydrogen and active carbon species, etc. It is, however, not a very promising strategy for achieving significant deviations from the Anderson-Schultz-Flory product distribution. To accomplish this goal, one would have to drive the reaction into new pathways, different from the ones prevailing in typical Fischer-Tropsch catalysts. Cluster-derived catalysts do show some potential along these lines, although much more research will be needed to prove it. Examples illustrating the possibility to generate unusual selectivities are given below. In addition to the products commonly encountered in CO hydrogenation, some cluster-derived catalysts are reported to produce significant amounts of dimethyl ether as well. Over A1203-supported catalysts derived from osmium carbonyl clusters, small amounts of dimethyl ether were formed. The presence of dimethyl ether in the product spectrum was attributed to secondary reactions leading to the dehydration of methanol over acidic sites on the support (ref. 232). Much larger amounts of dimethyl ether were formed over Fe-Mn/ZrOz/A1203 catalysts (ref. 50). Again, dehydration of methanol was suggested as the pathway leading to ether formation. Such an explanation, does however, not account

for

100

the

formation of substantial amounts of dimethyl ether over MoFeS/A1203 catalysts derived from bimetallic clusters shown in 1 Fig.

20

(refs. 57,58,60). Iron based catalysts

0

(ref.3)

and

potassium promoted and MoS~

80 60

40

CH4

C2=

C2

C3+ MeOMe

Fig. 6.5 Product distribution obtained in CO hydrogenation (&/CO=3: 1, total pressure = 261 kPa over MoFeS/A1203) catalyst as a function of pretreatment immediately preceding the CO hydrogenation run (ref. 57)

254

MoS2/CoS2 catalysts (ref. 233) are known to produce oxygenates including some dimethyl ether along with alcohols when operated at high pressures. In the case of the MoFeS cluster catalysts, however, substantial amounts of dimethyl ether were produced already at low pressure, and at no time were water or methanol observed in the product. Furthermore, the differential reaction conditions used would not have been conducive for the onset of secondary dehydration reactions of methanol. The product distribution was very sensitive to the immediate pretreatment before starting a given

CO hydrogenation run (Fig. 5). Treatment in He resulted in much higher dimethyl ether selectivity than treatment in H2 (ref. 57). The Fig. 6.6 Proposed catalytic cycle containing two pathways leading to either methane or dimethyl ether as primary products (ref. 57). M denotes a catalytic site, not a metal atom.

change in selectivity was found to be comP1etely different product

and the distributions

could be restored repeatedly by

simply interrupting the CO feed stream, and exposing the catalyst.to either He or H2 at 673 K for several hours, followed by putting the catalysts back on stream in a CO-H2 mixture. A kinetic analysis narrowed the rate determining step down to one involving a species containing one C-atom and two H atoms (ref. 57). The rate law was found to be consistent with the rate-determining step being either the hydrogenation of a surface methylidene for the carbide mechanism, or the hydrogenation of a surface formyl for the CO insertion mechanism. A model was proposed that could account for the production of both methane and dimethyl ether as primary products. The key idea in this model is a catalytic cycle with a formyl intermediate as the common point of departure into two different pathways (Fig. 6.6). The first pathway results in methane formation, while the second one accounts for the production of dimethyl ether without having to rely on methanol as an intermediate. Subtle changes induced by the pretreatment determine which one of the two pathways becomes dominant. An attractive feature of this model is that the individual steps postulated are based o n proven organometallic reaction analogs.

255

The same MoFeS clusters, when deposited on MgO showed very different behavior. During an induction period lasting about 24 h, the catalyst made mainly methane. Then with continuing time on stream the activity for C2 production rapidly increased until the product stream contained about 90 mol% of C, products, with an ethane/ethylene ratio of about 2:1, followed by a gradual settling into a steady state with about 75 % C2 products (ref. 60). Close inspection of Mossbauer and EXAFS characterization data suggest that the active sites of these sulfido cluster-derived catalysts are surface 0x0-ensembles. Although it is not yet possible to establish the exact structure of these surface ensembles, the presence of the second metal component Fe or Co plays a decisive role in determining the nature of these active sites. Such examples of highly selective and stable (or at least metastable) cluster-derived bimetallic catalysts give hope that it may indeed be possible to develop molecularly designed bimetallic catalysts whose surface may act as highly selective templates for new products formed via novel catalytic pathways.

6.9 REFERENCES 1 2 3 4

5 6 7 8 9 10

11 12 13 14 15

16

J. H. Sinfelt, Bimetallic Catalysts - Discoveries, Concepts, and Applications, Wiley, New York, 1983. P.J. Flory, Principles of Polymer Chemistry, Cornell University Press, Ithaca, New York, 1967, p.318 ff. H. H. Storch, N. Golumbic, and R. B. Anderson, The Fischer- Tropsch and Related Syntheses, Wiley, New York, 1951. D. L. King, J. A. Cusumano and R. L. Garten, Catal. Rev.- Sci.Eng., 23, (1981) 233-263. Y. C. Hu, Hydrocarbon Process., 62, (1983) 88-96. G. v. d. Lee and V. Ponec, Catal. Rev.-Sci.Eng., 29, (1987) 183 A. S. C. Chan in S. Kaliaguine (Ed.), Stud. Surf. Sci. Catal., Vol. 35, “Keynotes in Energy Related Catalysis”, Elsevier, Amsterdam 1988, pp. 257- 3 18. V. Ponec, in E. G. Derouane and A.A. Lucas (Eds.), Electronic Structure and Reactivity of Metal Surfaces, Plenum, New York, 1976, pp. 537-619. W. M. H. Sachtler and R. A. van Santen, Adv. Catal., 26, (1977) 69 M. J. Kelley and V. Ponec, in Progress in Surface Science, 11(3), (1981) 139 V. Ponec, Preprints ACS Div. Pet. Chem., 29, (1984) 774 D. F. Ollis, J. Catal., 23, (1971) 131 J. H. Sinfelt and J. A. Cusumano, in J. J. Burton and R. L. Garten(Eds.), Advanced Materials in Catalysis, Academic Press, New York, 1977, pp.1 J. K. A. Clarke and A. C. M. Creaner, Ind. Eng. Chem. Product Res. Develop., 20, (1981) 574. H. H. Storch, in W. G. Frankenburg, V. I. Komarewsky, and E. K. Rideal, (Eds.), Advances in Catalysis and Related Subjects, Vol. I, Academic Press, New York, 1948, pp.115 H. Pichler, in W. G. Frankenburg, V. I. Komarewsky, and E. K. Rideal, (Eds.), Advances in Catalysis and Related Subjects, Vol. IV, Academic Press, New York, 1952, pp.271

256

17 18

R. B. Anderson, in P. H. Emmett (Ed.), Catalysis, Vol. IV, Reinhold, New York, 1956. M. E. Dry, in J. R. Anderson and M. Boudart (Eds.), Catalysis, Science and Technology, Volume 1, Springer-Verlag, Berlin, 1981, Ch.4, pp. 159

19

L. Guczi, in B.C. Gates, L. Guczi and H. Knozinger, (Eds.), Stud. Surf. Sci. Catal., Vol. 29, “Metal Clusters in Catalysis” Elsevier, Amsterdam 1986, p. 547

20 21 22 23 24

J. Zwart and R. Snel, J. Mol. Catal., 30, (1985) 305 R. Ugo and R. Psaro, J. Mol. Catal., 20, (1983) 53. Yu. I. Yermakov, J. Molec. Catal., 21, (1983) 35 J. Phillips and J. A. Dumesic, Applied Catal., 9, (1984) 1 B. C. Gates, L. Guczi and H. Knozinger (Eds.), Metal Clusters in Catalysis, Elsevier, Amsterdam 1986. J. R. Budge, J. P. Scott and B.C. Gates, J. Chem. Soc. Chem. Commun. (1983) 342. A. K. Datye, L. F. Allard, Jr., J. Schwank and B. C. Gates, in G. W. Bailey (Ed.), Proceedings of the 42nd Annual Meeting of the Electron Microscopy Society of America, San Francisco Press, San Francisco, 1984, pp. 660 L. Guczi, Z. Schay, K. Lhzir, A. Vizi and L. Marko, Surface Sci., 106, (1981) 516.

25 26

27 28

29 30

31

32 33

34 35 36 37

38 39

L. Guczi, Z. Schay, K. Matusek, I. Bogyay and G. Steffler, in T. Seiyama and K. Tanabe (Eds.), Studies in Surface Science and Catalysis, Vol. 7, Part A,, New Horizons In Catalysis, Elsevier, Amsterdam, 1981, pp. 21 1 K. La,&, Z. Schay and L. Guczi, J. Mol. Catal., 17, (1982) 205 L. Guczi, Z. Schay and I. Bogyay, in G. Poncelet, P. Grange, and P.A. Jacobs (Eds.), Studies in Surface Science and Catalysis, Vol. 16, Preparation of Catalysts 111, Elsevier, Amsterdam, 1983, pp. 451 I. BoszormCnyi, S. Dobos, L. Guczi, L. Marko, K. Lazir, W. M. Reiff, Z. Schay, L. Takacs and A. Vizi-Orosz, in Proceedings, 8th Internat. Congr. Catal., Volume 5, Verlag Chemie, Weinheim, 1984, pp. 183 2. Schay, K. LBzir, J. Mink and L. Guczi, J. Catal., 87, (1984) 179

L. Guczi, I. BoszormCnyi, S. Dobos, K. Lazar, L. Takacs and Z. Schay, in P. Barret and L.-C. Dufour (Eds.), Mater. Sci. Monogr., Vol. 28B, Reactivity of Solids, Elsevier, Amsterdam, 1985, pp. 1087 I. BoszormCnyi, S. Dobos, K. Lrizk, L. Takacs, Z. Schay and L. Guczi, Surface Sci. 156, (1985), 995 Z . Schay, I. Bogyay and L. Guczi, Acta Chim. Hung. 119, (1985) 105 I. BoszormCnyi and L. Guczi, Inorg. Chim. Acta, 112, (1986) 5 L. Guczi and S. Dobos, in Yu. Yermakov and V. Likholobov (Eds.), Homogeneous and Heterogeneous Catalysis, Proc. 5th Internat. Symp. Relations between Homogeneous and Heterogeneous Catalysis, Novosibirsk 1986, VNU Science Press, Utrecht, 1986, pp. 901 K. LBzk, W. M. Reiff, W. Morke and L. Guczi, J. Catal., 100 (1986) 118 M. Kaminsky, K. J. Yoon, G. L. Geoffroy and M.A. Vannice, J. Catal., 91, (1985) 338

257

40 41 42 43 44 45 46 47 48 49 50 51

J. R. Budge, B. F. Lucke, J. P. Scott and B. C. Gates, in Proceedings, 8th Internat. Congr. Catal., Volume 5, Verlag Chemie, Weinheim, 1984, pp. 89 A. Choplin, M. Leconte, J.-M. Basset, S. G. Shore and W.-L. Hsu, J. MoLCatal., 21, (1983) 389 J. R. Budge, B. F. Liicke, B. C. Gates and J. Toran, J. Catal., 91, (1985) 272 G. L. Ott, T. Fleisch and W. N. Delgass, J. Catal., 60, (1979) 394 A. Chen, M. A. Vannice, M. Kaminsky and G. Geoffroy, Reprints. Am. Chem. Soc.,Div. Pet. Chem., 31, (1986) 145 R. Hemmerich, W. Keim and M. Roper, J. Chem. SOC.,Chem. Commun., (1983) 428 V. L. Kuznetsov, H. F. Danilyuk, I. E. Kolosova and Y. I. Yermakov, React. Kinet. Catal. Lett., 21, (1982) 249

J. Venter, M. A. Vannice, M. Kaminsky and G. L. Geoffrey, Reprints. Am. Chem. Soc., Div. Petrol. Chem., 31, (1986) 219 L. Bruce, G. Hope and T. W. Turney, React. Kinet. Catal. Lett., 20, (1982) 175 D. Vanhove, L. Makombo and M. Blanchard, J. Chem. Research (S), 335, (1980) 4128. R. Pierantozzi, J. Catal., 106, (1987) 323 M. Ichikawa, A. Fukuoka and T. Kimura, in M. J. Phillips and M. Teman (Eds.), Proceedings of the 9th International Congress on Catalysis, V01.2, The Chemical Institute of Canada, Ottawa, 1988, pp. 569

52

J. R. Anderson, P. S. Elmes, R. F. Howe and D. E. Mainwaring, J. Catal., 50, (1977) 508

53 54

M. Ichikawa, CHEMTECH, 10, (1982) 674. J. R. Shapley, S. J. Hardwick, D. S. Foose and G. D. Stucky, Preprints. Am. Chem. Soc.,

55 56

Div. Petrol. Chem., 25, (1980) 780 M. D. Curtis, J. Schwank, L. T. Thompson, Jr., P. D. Williams and 0. Baralt, Reprints. Am. Chem. Soc.Div. Fuel Chem., 31, (1986) 44 L. T. Thompson, Jr., M. D. Curtis and J. Schwank, in Proceedings of World Congress I11

of Chemical Engineering, Tokyo, 1986, Vol. IV, pp. 426 57 58

59

60 61 62 63

L. T. Thompson, Jr., J. Schwank andM. D. Curtis, AIChE J. 35, (1989) 109 M. D. Curtis, J. Penner-Hahn, 0. Baralt, J.Schwank and L. T. Thompson, Jr., in M. J. Phillips and M. Ternan (Eds.), Proceedings of the 9th International Congress on Catalysis, V01.2, The Chemical Institute of Canada, Ottawa,1988, pp. 869 M.D. Curtis, J.Schwank, J. E. Penner-Hahn, L. T. Thompson, Jr., 0. Baralt and G . Waldo, in M. M. J. Treacy, J. M. Thomas and J. M. White (Eds.), Mater. Res. SOC.Symp. Proc. “Microstructure and Properties of Catalysts”, Vol. 111, MRS, Pittsburgh, PA, 1988, pp. 33 1 M.D. Curtis, J. E. Penner-Hahn, J. Schwank, 0. Baralt, D. J. McCabe, L. T. Thompson, Jr. and G. Waldo, Polyhedron, 7, (1988) 241 1 H. Kanai, B. J. Tan and K. J. Klabunde, Langmuir, 2 , (1986) 760 R. Bouwman and W. M. H. Sachtler, J. Catal., 19, (1970) 127 R. Bouwman and W. M. H. Sachtler, J. Catal., 26, (1972) 63

258

64 K. Christmann, G. Ertl and H. Shimizu, J. Catal., 61(1980) 397 65 66

K. Christmann and G. Ertl, J. Mol. Catal., 25, (1984) 31 H. Shimizu, K. Christmann and G. Ertl, J. Catal. 61, (1980) 412

67

J. C. Vickerman, K. Christmann and G. Ertl, J. Catal. 71, (1981) 175

68

C. H. F. Peden and D. W. Goodman,,in Proceedings, Symposium on the Surface Science of Catalysis, Philadelphia, Aug. 1984, ACS Symposium Series.

69 70 71 72 73

J. T. Yates, Jr., C. H. F. Peden and D. W. Goodman, J. Catal. 94, (1985) 321 M. Raney, Ind. Eng. Chem., 32, (1940) 1199 F. Fischer and K. Meyer, Brennstoff-Chem., 15, (1934) 84 pp. W. H. Wiser, Report 1985, DOE/ET/l4700-Tl-Vol.7,234 Kien Ru Chen, Ph.D. Thesis, Univ. Utah, 1986. Avail. Univ. Microfilms, Int., Order NO. DA8624542. D. W. McKee, Trans. Faraday SOC.,61, (1965) 2273

74 75

A. Yokoyama, H. Komiyama, H. Inoue, T. Matsumoto and H. M. Kimura, J. Catal., 68, (1981) 355

76

A. Yokoyama, H. Komiyama, H. Inoue, T. Matsumoto and H. M. Kimura, Chem. Lett., (1983) 195

77 78

M. Shibata, N. Kawata, T. Masumoto and H. M. Kimura, Chem. Lett., (1985) 1605 M. Shibata and T. Masumoto, in B.Delmon, P. Grange, P.A. Jacobs, and G . Poncelet (Eds.), Preparation of Catalysts IV, Elsevier, 1987, pp. 353 G. Kisfaludi, K. Lizar, Z. Schay, L. Guczi, Cs. Fetzer, G. Koncos and A. Lovas, Appl.

79

Surface Sci. 24, (1985) 225 80 81 82 83 84 85 86 87 88 89 90 91

G. Kisfaludi, Z. Schay, L. Guczi, G. Koncos, A. Lovas and P. Kovacs, Appl. Surface Sci. 28, (1987) 111 G. Leofanti, M. Solari, G. R. Tauszik, F. Garbassi, S. Galvagno and J. Schwank, Appl. Catal., 3, (1982) 131 A. G. Shastri, H. B. Chae, M. Bretz and J. Schwank, J. Phys. Chem., 89, (1985) 3761 T. E. Holt, A. D. Logan, S. Chakrabarty and A. K. Datye, Appl. Catal., 34, (1987) 199 A. G. Shastri, A. K. Datye and J. Schwank, Appl. Catal., 14, (1985) 119 I. W. Bassi, F. Garbassi, G. Vlaic, A. Marzi, G. R. Tauszik, G. Cocco, S. Galvagno and G . Parravano, J. Catal., 64, (1980) 405 S. Galvagno, J. Schwank, G. Parravano, F. Garbassi, A. Marzi and G. R. Tauszik, J. Catal., 69, (1981) 283 J. Schwank, G. Parravano and H. L. Gruber, J. Catal. 61, (1980) 19 G. R. Tauszik, G. Leofanti and S. Galvagno, J. Molec. Catal., 25, (1984) 357 A. K. Datye and J. Schwank, in Proceedings, 8th Internat. Congr. Catal., Volume 4, Verlag Chemie, Weinheim, 1984, pp. 587 A. K. Datye, L. F. Allard and J. Schwank, in D. B Williams and D.C. Joy (Eds.), Analytical Electron Microscopy- 1984, San Francisco Press, San Francisco, 1984, pp. 205 A. K. Datye and J. Schwank, J. Catal., 93, (1985) 256

259

92 93 94

9s 96 97 98 99 100 101

A. G. Shastri and J. Schwank, J. Catal., 95, (1985) 271 A. G. Shastri and J. Schwank, J. Catal., 95, (1985) 284 A. G. Shastri and J. Schwank, J. Catal., 98, (1986) 191 A. G. Shastri and J. Schwank, J. Catal., 100, (1986) 437 A. G. Shastri and J. Schwank, in A.D. Romig, Jr., and W. F. Chambers (Eds.), Microbeam Analysis-1986, San Francisco Press, San Francisco, 1986, pp. 623 A. G. Shastri and J. Schwank, in Proceedings of World Congress 111 of Chemical Engineering, Tokyo, 1986, Vol. 111, pp. 299 A. K. Datye, Ph.D. Thesis, Univ. Michigan, 1984. J. H. Sinfelt, J. Catal., 29, (1973) 308 A. G. Shastri, J. Schwank, and S. Galvagno, J. Catal. 100, (1986) 446 A. D. Logan and A. K. Datye, J. Catal., 112, (1987) 595

G. Blanchard, H. Charcosset, H. Dexpert, E. Freund, C. Leclercq and G. Martino, J. Catal. 70, (1981) 168 103 H. Miura and R. D. Gonzalez, J. Catal., 74, (1982) 216 104 H. Miura and R. D. Gonzalez, Ind. Eng. Chem. Prod. Res. Dev., 21, (1982) 274 105 H. Miura and R. D. Gonzalez, Reprints. Am. Chem. Soc., Div. Pet. Chem., 28, (1983) 477 106 H. Miura, T. Suzuki, Y. Ushikubo, K. Sugiyama, T. Matsuda, and R. D. Gonzalez, J. Catal., 85, (1984) 331 107 S. Alerasool, D. Boecker, B. Rejai, R. D. Gonzalez, G. del Angel, M. Azomosa and R. Gomez, Langmuir, 4, (1988) 1083 108 G. Diaz, F. Garin and G. Maire, J. Catal., 82, (1983) 13 109 J. C. M. Harberts, A. F. Bourgonje, J. J. Stephan and V. Ponec, J. Catal., 47, (1977) 92 110 K. Matusek, I. Bogyay, L. Guczi, G. Diaz, F. Garin and G. Maire, C1 Mol. Chem., 1, (1985) 335 111 D. K. Chakrabarty, K. M. Rao, N. Sundararaman and K. Chandavar, Appl. Catal., 28, (1986) 69 112 0. Okuda, T. Tatsumi, K. Fujimoto and H. Tominaga, Chem. Lett., (1983) 11.53 113 T. Tatsumi, Y. G. Shul, T. Sugiura and H. Tominaga, Appl. Catal., 21, (1986) 119 114 T. Tatsumi, Y. G. Shul, Y. Arai and H. Tominaga, in Y. Murakami, A. Iijima and J. W. Ward (Eds.), New Developments in Zeolite Science and Technology, Stud. Surf. Sci. Catal., 28, Elsevier, 1986, pp. 891 115 Y. G. Shul, T. Sugiura, T. Tatsumi and H. Tominaga, Appl. Catal., 24, (1986) 131 116 A. Ramachandran and D. K. Chakrabarty, Appl. Catal., 42, (1988) 229 117 R. D. Gonzalez, Appl. Surf. Sci., 19, (1984) 181 118 R. P. Eischens, Z. Elektrochem., 60, (1956) 782 119 Y. Soma-Noto and W. M. H. Sachtler, J. Catal., 34, (1974) 162 120 Y. Soma-Noto and W. M. H. Sachtler, J. Catal., 32, (1974) 315 . 121 Y. Soma and W. M. H. Sachtler, Japan J. Appl. Phys. Suppl.2, Pt.2 (1974) 241 102

260

122 123 124 125 126 127 128 129 130 131 132 133 134 135

136 137 138 139

140 141 142 143 144 145 146 147 148

J. C. Vickerman, K. Christmann, G. Ertl, P. Heimann, F. J. Himpsel and D. E. Eastman, Surface Sci. 134, (1983) 367 J. E. Houston, C. H. F. Peden, P. J. Feibelman and D. R. Hamman, Phys. Rev. Letters 56, (1986) 375 J. E. Houston, C. H. F. Peden, D. S. Blair and D. W. Goodman, Surface Sci., 167, (1986) 427 J. Paul and F. M. Hoffman, Surface Sci., 172, (1986) 151 M. Kiskinova, M. Tikhov and G. Bliznakov, Surface Sci.,204, (1988) 35 H. Pfniir, P. Feulner and D. Menzel, J. Chem. Phys., 79, (1983) 4613 J. Kessler and F. Thieme, Surface Sci., 67, (1977) 405 P.-L. Cao, D. E. Ellis, A. J. Freeman, Q.-Q. Zheng and S. D. Bader, Phys. Rev. B, 30, (1984) 4146 C. Harendt, K. Christmann, W. Hirschwald and J. C. Vickerman, Surface Sci., 165, (1986) 413 S. Yang, M. Yu, G. Meigs, X. H. Feng and E. Garfunkel, Surface Sci., 205, (1988) L777 and references cited therein. M. F. Brown and R. D. Gonzalez, J. Catal., 48, (1977) 292 C. M. Grill, M. L. McLaughlin, J. M. Stevenson and R. D. Gonzalez, J. Catal., 69, (1981) 454 P. J. Berlowitz, J. E. Houston, J. M. White and D. W. Goodman, Surface Sci., 205, (1988) 1 A. Matani and D. K. Chakrabarty, in M. J. Phillips and M. Teman (Eds.), Proceedings of the 9th International Congress on Catalysis, V01.2, The Chemical Institute of Canada, Ottawa, 1988, pp. 682 K. R. Christmann, in Z. Pail and P. G. Menon (Eds.), Hydrogen Effects in Catalysis, Marcel Dekker, New York, 1988, pp. 3 M. A. Barteau, J. Q. Broughton and D. Menzel, Surface Sci., 133, (1983) 433 F. Greuter and E. W. Plummer, Solid State Commun., 48, (1983) 37 J. H. Sinfelt, Y. L. Lam, J. A. Cusumano and A. E. Bamett, J. Catal., 42, (1976) 227 A. J. Rouco, G. L. Haller, J. A. Oliver and C. Kemball, J. Catal. 84, (1983) 297 G. L. Haller, D. E. Resasco and J. Wang, J. Catal. 84, (1983) 477 T. S. King, X. Wu, B. C. Gerstein, J. Am. Chem. Soc., 108, (1986) 6056 D. W. Goodman and C. H. F. Peden, J. Catal., 95, (1985) 321 C. H. Bartholomew, in 2. Pail and P. G. Menon (Eds.), Hydrogen Effects in Catalysis, Marcel Dekker, New York, 1988, pp. 139 C. H. Bartholomew, in 2. Pail and P. G. Menon (Eds.), Hydrogen Effects in Catalysis, Marcel Dekker, New York, 1988, pp. 543 D. E. Peebles, J. A. Schreifels and J. M. White, Surface Sci., 116, (1982) 117 B. E. Koel, D. E. Peebles and J. M. White, Surface Sci., 107, (1981) L367 B. E. Koel, D. E. Peebles and J. M. White, Surface Sci., 125, (1983) L87

261

149 150

151 152 153 154 155 156 157 158 159 160 161

162 163 164 165

G. E. Mitchell, J. L. Gland and J. M. White, Surface Sci., 131, (1983) 167 V. Gryaznov, J. Serov and S. Gul'yanova, in M. J. Phillips and M. Ternan (Eds.), Proceedings of the 9th International Congress on Catalysis, V01.2, The Chemical Institute of Canada, Ottawa, 1988, pp. 722 J. A. Amelse, L. H. Schwartz and J. B. Butt, J. Catal., 72, (1981) 95 Y. T. Shah and A. H. Perotta, Ind. Eng. Chem., Prod. Res. Dev., 15, (1976) 123 I. E. Wachs, D. J. Dwyer and E. Iglesia, Appl. Catal., 12, (1984) 201 T. Okuhara, T. Jin, Y. Zhou and J. M. White, J. Phys. Chem., 92, (1988) 4141 R. A. Diffenbach, R. R. Shehl, D. J. Fauth, Prw. Int. Conf. Coal Sci., (1983) 240 C. N. Satterfield and H. G. Stenger, Ind. Eng. Chem. Process Des. Dev., 23, (1984) 849 M. Araki and V. Ponec, J. Catal., 44, (1976) 439 L. J. M. Luyten, M. Van Eck, J. Van Grondelle and J. H. C. Van Hooff, J. Phys. Chem., 82, (1978) 2000 W. A. A. van Barneveld and V. Ponec, J. Catal., 51, (1978) 426 D. J. Elliott and J. H. Lunsford, J. Catal., 57, (1979) 116 H. H. Nijs, P. A. Jacobs, J. J. Verdonck and J. B. Uytterhoeven, in J. Bourdon (Ed.), Growth and Properties of Metal Clusters, Stud. Surf. Sci. Catal., Vol. 4, Elsevier, 1980, pp. 479 P. Villeger, G. Leclercq and R. Maurel, Bull. SOC.Chim. France, 9-10, Pt.1, (1979) 406 P. Villeger, J. Barrault, J. Barbier, G. Leclercq and R. Maurel, Bull. Soc.Chim. France, 9-10, Pt.1, (1979) 413 G. C. Bond and B. D. Tumham, J. Catal., 45, (1976) 128 S. Y. Lai and J. C. Vickerman, J. Catal., 90, (1984) 337

166

T. S. King, J. H. Iwaniya, R. D. Walker, W. J. Goretzke and B. C. Gerstein, AIChE 1985 Annual Meeting, Chicago, Ill., Paper 45a

I67

G. D. Zakumbaeva, L. B. Shapovalova, Zh. T. Omarov, E. N. Litvyakova and A. Sh. Kuanyshev, Neftekhimiya, 28, (1988) 507 W. A. A. Van Barneveld and V. Ponec, J. Catal., 89, (1984) 542 P. Biloen, Recl. Trav. Chim. Pays-Bas, 99, (1980) 33 J. A. Dalmon, G. A. Martin, in T. Seiyama and K. Tanabe (Eds.), Studies in Surface Science and Catalysis, Vol. 7, Part A,, New Horizons In Catalysis, Elsevier, Amsterdam, 1981, pp. 402 D. L. King, J. Catal., 51, (1978) 386

168 169 170

171 172 173 174 175 176 177

C. S. Kellner and A. T. Bell, J. Catal., 75, (1982) 251 M. Boudart and M. A. McDonald, J. Phys. Chem., 88, (1984) 2185 R. D. Kelley and W. D. Goodman, Surface Sci., 123, (1982) L743 D. W. Goodman, Surface Sci., 123, (1982) L679 M. A. Vannice and R. L. Garten, J. Mol. Catal., 1, (1975/76) 201. M. A. Vannice, Y. L. Lam and R. L. Garten, Reprints ACS Div. Pet. Chem., 23, (19781, 495

262

178 179 180

181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 20 1 202 203 204

G. L. Ott, W. N. Delgass, N. Winograd, J. Catal., 65, (1980) 253 G. L. Ott, W. N. Delgass, N. Winograd and W. E. Baitinger, J. Catal., 56, (1979) 174 L. Guczi, K. Matusek, I. Manninger, J. K i d y and M. Eszterle, in B. Delmon, P. Grange,

P. Jacobs and G. Poncelet (Eds.), Preparation of Catalysts 11, Elsevier, Amsterdam, 1979, pp. 391 Z. Schay and L. Guczi, React. Kinet. Catal. Lett., 14, (1980) 207 L. Guczi, Catal. Rev.-Sci. Eng. 23, (1981) 329 M. L. Good, M. Akbamejad, M. D. Patil and J. T. Donner, Mater. Res. Soc. Symp. Proc., 3, (1981) 271 Z. Schay and L. Guczi, J. Chem. Soc., Faraday Trans. I., 78, (1982) 1911 Z. Schay and L. Guczi, Acta Chim. Acad. Sci. Hung., 111, (1982) 607 J. W. Niemantsverdriet, D. P. Aschenbeck, F. A. Fortunato and W. N. Delgass, J. Mol. Catal., 25, (1984) 285 L. Guczi, Z. Schay, K. Matusek and I. Bogyay, Appl. Catal., 22, (1986) 289 F. J. Berry, L. Liwu, W. Chengyu, T. Renyuan, Z. Su and L. Dongbai, J. Chem. SOC., Faraday Trans. I, 81, (1985) 2293 F. Stoop and K. Van der Wiele, Appl. Catal., 23, (1986) 35 A. M. van der Kraan, R. C. H. Nonnekens and J. W. Niemantsverdriet, Hyperfine Interact., 28, (1986) 899 R. P. Elliott, Constitution of Binary Alloys: First Supplement, McGraw Hill, New York, 1966, p.43 1 M. M. Bhasin, W. J. Bartley, P. C. Ellgen and T. P. Wilson, J. Catal., 54, (1978) 120 J. W. Niemantsverdriet, A. M. van der Kraan and J. J. van Loef, J. Phys. Chem., 87, (1983) 1292 J. W. Niemantsverdriet, A. M. van der Kraan and W. N. Delgass, J. Catal., 89, (1984) 138 T. Fukushima, H. Arakawa, M. Ichikawa, J. Phys. Chem., 89, (1985) 4440 A. Fukuoka, T. Kimura and M. Ichikawa, J. Chem. Soc.,Chem. Commun.,(1988) 428 Y. Minai, T. Fukushima, S. Ogasawara, M. Ichikawa and T. Tominaga, J. Radioanal. Nucl. Chem. Letters, 106, (1986) 167 I. S. Hughes, J. 0. H. Newman and G. C. Bond, J. Mol. Catal., 25, (1984) 219 L. Guczi, K. Matusek, J. Margitfalvi, M. Eszterle and F. Till, Acta Chim. Acad. Sci. Hung., 101, (1979) 107 L. Guczi, G. KemCny, K. Matusek and J. Mink, J. Chem. Soc. Faraday Trans. I, 76, (1980) 782 G. Broden, G. Gafner and H. P. Bonzel, Appl. Phys., 13, (1977) 333 K. A. Prior, K. Schwaha and R. M. Lambert, Surf. Sci., 77, (1978) 193 R. M. Stanfield and W. N. Delgass, J. Catal., 72, (1981) 37 M. Nakamura, B. J. Wood, P. Y. Hou and H. Wise, in T. Seiyama and K. Tanabe (Eds.), Studies in Surface Science and Catalysis, Vol. 7, Part A., New Horizons In Catalysis, Elsevier, Amsterdam, 1981, pp. 432

263

205 206

R. A. Fiato, G. B. McVicker and A. A. Montagna, U.S. Patent 4,532,229. J. B. Butt and L. H. Schwartz, M. Baems and R. Malessa, Ind. Eng. Chem. Prod. Res. Dev., 23, (1984) 51

207 208 209 210 21 1 212 21 3 214

K. B. Arcuri, L. H. Schwartz, R. D. Piomowski and J. B. Butt, J. Catal., 85, (1984) 349 H. Arai, K. Mitsuishi and T. Seiyama, Chem. Letters, (1984) 1291 M. Audier,B. Baess and M. Coulon, C1 Mol Chem., 1, (1984) 33 T. A. Lin, L. H. Schwartz, and J. B. Butt, J. Catal., 97, (1986) 177 T. Ishihara, K. Eguchi andH. Arai, Chem. Letters,(l986) 1695 T. Ishihara, K. Eguchi and H. Arai, Appl. Catal., 30, (1987) 225 T. Shigematsu, K. Kodama, and Nakanishi, N., Chem. Express, 2 (1987) 317 T. Shigematsu, N. Takewa, M. Ohtsuki, K. Watanabe, Y. Bando and N. Nakanishi, Bull. Inst. Chem. Res., Kyoto Univ., 64,(1986) 277 X-2. Jiang, S. A. Stevenson and J. A. Dumesic, J. Catal., 91, (1985) 11 S . H. Chien, J. C. Chang and P. F. Chou, Bull. Inst. Chem. Acad. Sin., 34, (1987) 57 L. Guczi, K. Matusek, I. Bogyay, F. Garin, P. Esteban Puges, P. Girard and G. Maire, C1 Mol. Chem., 1, (1986) 355 M. Ichikawa and T. Fukushima, J. Chem. Soc., Chem. Commun., (1985) 321 D. C. Koningsberger, C. P. J. H. Borgmans, A. M. J. van Elderen, B. J. Kip and J. W. Niemantsverdriet, J. Chem. Soc.,Chem. Commun., (1987) 892 T. Fukushima, K. Araki and M. Ichikawa, J. Chem. SOC.,Chem. Commun., (1986) 148 G. Lietz, M. Nimz, J. Volter, K. Lizir and L. Guczi, Appl. Catal. 45, (1988) 71 H. Hamada, Y. Kuwahara, Y. Kintaichi and T. Ito, Chem. Letters, (1984) 1611 Y. Kintaichi, Y. Kuwahara, H. Hamada, T. Ito and K. Wakabayashi, Chem. Letters, (1985) 1305

215 216 217 218 219 220 22 1 222 223 224 225

226

227

228 229

Y. Kuwahara, H. Hamada and Y. Kintaichi, Chem. Letters, (1985) 205 N. A. Bhore, C. Sudhakar, K. B. Bischoff, W. H. Manogue and G. A. Mills, in M. J. Phillips and M. Ternan (Eds.), Proceedings of the 9th International Congress on Catalysis, V01.2, The Chemical Institute of Canada,Ottawa,l988, pp. 594 J. W. Niemantsverdriet, S. P. A. Louwers, J. van Grondelle, A. M. van der Kraan, F. W. H. Kampers and D. C. Koningsberger, in M. J. Phillips and M. Ternan (Eds.), Proceedings of the 9th International Congress on Catalysis, V01.2, The Chemical Institute of Canada, Ottawa, 1988, pp. 674 A. F. H. Wielers, C. M. A. M. Mesters, G. W. Koebrugge, C. J. G. van der Grift and J. W. Geus, in B. Delmon, P. Grange, P.A. Jacobs, and G. Poncelet (Eds.), Preparation of Catalysts IV, Elsevier, 1987, pp. 401 P. K Agrawal, W. D. Fitzharris and J. R. Katzer, in B. Delmon and G. F. Froment (Eds.), Catalyst Deactivation, Elsevier, Amsterdam, 1980, pp. 179 C. H. Bartholomew and J. R. Katzer, in B. Delmon and G. F. Froment (Eds.), Catalyst Deactivation, Elsevier, Amsterdam, 1980, pp. 375

264

230 231 232

233 234

L. Guczi and Z. Schay, in L. Cerveny (Ed.), Catalytic Hydrogenation, Elsevier, Amsterdam, 1986, pp. 313 A. T. Bell, Catal. Rev.-Sci. Eng., 23, (1981) 203 E. 0. Odebunmi, B. A. Matrana, A. K. Datye, L. F. AllardJr., J. Schwank, W. H. Manogue, A. Hayman, J. H. Onuferko, H. Knozinger and B. C. Gates, J. Catal., 95, (1985) 370 W. P. Dianis, Appl. Catal., 30, (1987) 99 J. M. Cowley andR. J. Plano, J. Catal., 108, (1987), 199

265

CHAPTER 7

CLASSICAL AND NON-CLASSICAL ROUTES FOR ALCOHOL SYNTHESIS Richard G. Herman Zettlemoyer Center for Surface Studies Sinclair Laboratory, No. 7 LEHIGH UNIVERSITY Bethlehem, Pennsylvania 18015 USA

266

7.1 INTRODUCTION Efficient catalytic processes for synthesizing alcohols from carbon monoxide and hydrogen have long been of practical interest because alcohols can be used directly as clean high octane fuels and as chemical intermediates to a wide range of commercially important chemicals, e.g. methanol to produce formaldehyde, ethanol to yield ethene, and higher alcohols used to synthesize alkylamines. This chapter will consider the older established routes used for the industrial synthesis of particular chemical grade alcohols from hydrogen-rich synthesis gas, e.g. H2/C0 = 2.3, as well as recently developed catalysts and processes that have commercial potential for producing mixtures of fuel-grade alcohols from synthesis gas derived from sources such as coal or biomass that yield a CO-rich synthesis gas with H2/C0 = 0.4- 1.0. Methanol synthesis is a well-developed process that occurs as one of the most active (up to 2 kg methanolkg catalyst/hr) and selective heterogeneous catalytic reactions used in an industrial process. It can be carried out as a high temperature (=4Oo0C) and high pressure (10-20 MPa) process over catalysts such as ZnO/Cr203 or as a low temperature (=25OoC) and moderature pressure (5-10 MPa) process over copper-based catalysts. Recent research has sought to develop catalysts for the synthesis of C,-C, alcohol mixtures for use as high Octane fuels and diverified chemical feedstocks. Successful approaches have included modifications of the following types of heterogeneous catalysts: HI

Modified Methanol Synthesis (Cu/ZnO-based)

+ CO

Modified Fi scher-Trop sch (Fe- or Co- based)

Modified Methanation (MoS2-based)

L Higher Alcohols

These modifications typically have involved the use of alkali, or in some cases a Group VIII metal, as a promoter to enhance the carbon-carbon bond forming reactions that take place with oxygenated intermediates over these catalysts. While this method of catalyst modification has long been known, it is noteable that the addition of alkali to these three classes of catalysts can result in: (a) (b) (c)

suppression of the formation of hydrocarbons, enhancement of the alcohol yields, and shifting of the selectivity toward C2-C4 alcohols from H 2 - p ~ 'synthesis gas, e.g. H2/CO I 1.

267

Each of these classes of catalysts will be separately discussed. Other catalyst systems are being probed for their potential in converting synthesis gas to alcohols, and these include the following: (a)

Pd-based catalysts, such as Pd/SiO,,

(b) (c)

Ni-based catalysts,

(d)

Other noble metal catalysts, such as supported Rh, Th-based isosynthesis-type catalysts, and Homogeneous phase catalytic systems that operate at high temperatures and pressures.

(e) Recent developments and prospects for these catalyst systems for the synthesis of methanol, linear

C,-C4 alcohols, and branched C,-C4 alcohols will be considered. Catalyst development and optimization can be aided by knowing the identity of the intermediates and the pathways by which these intermediates interact to transform synthesis gas into methanol and higher alcohols. Therefore, available mechanistic information will be included in the discussion where appropriate. Pathways that have been proposed for higher alcohol synthesis include: (b)

CO insertion to form a C-C bond, followed by hydrogenation, Homologation of methanol by CO via a symmetric intermediate,

(c)

Condensation of two alcohol molecules via dehydration,

(d) (e)

Aldol condensation of two aldehydic molecules, and

(a)

Methylene insertion.

More specifically, the variations of coupling two oxygenated intermediates together to fomi di-oxygenated species include the: (a)

Enolic mechanism,

(b)

Oxymethylene/Oxide mechanism, and

(c)

Formate/Alkoxide/Carboxylatemechanism, and

(d)

Direct aldol mechanism.

It will be shown that the mechanism of alcohol synthesis is catalyst specific, and cannot be generalized from one catalyst to another. A few points of fundamental mechanistic debate and importance will be addressed because they can play a role in future catalyst development and warrant additional investigation. It has long

been considered that CO is the component in the H2/CO/C02 indusnial reactant mixture that is directly hydrogenated to form methanol. However, ICI researchers, following the work of earlier Russian investigators, have presented I4C isotopic evidence that under some conditions, e.g. low conversions, high space velocities, and generally comparable concentrations of CO and CO, in the synthesis gas, CO, is the reactant in methanol synthesis rather than CO. They report that methanol is formed from CO only when very low levels of CO, are present in the reactant stream. Recently, other researchers have investigated the relative roles of CO and CO2 in methanol synthesis in

268

greater detail and have shown that indeed COz can be the predominant reactant in methanol synthesis under some experimental conditions. Another controversy in alcohol synthesis is the identity of the reactive species on the surface of the catalyst that acts as the initiator of the synthesis. Since it is known that the Cu/ZnO = 30/70 mol% catalyst is about lo3 times as active as the individual Cu and ZnO components, an intimate relationship between the catalyst components must exist. The three possibilities that have been proposed and strongly supported in the literature for the reactive species are the following: (a) Cu6+ dispersed in the ZnO matrix, metallic copper supported by the ZnO matrix, and a combination of Cu(0) and "oxidized" copper species that must be in close proximity to

(b) (c)

one another, both of which must be fairly highly dispersed. These same three possible reactive surface states have also been applied to the Pd/Si02 catalyst that synthesizes methanol, and again the real working state of the catalyst has not been definitively clarified.

7.2 METHANOL SYNTHESIS CATALYSTS Industrial alcohol production means methanol synthesis to many people since methanol synthesis is a very large scale process as it is typically camed out today. The larger plants produce between 1000 and 3000 tons (1-3 million kg) of methanol per day. On a world-wide basis, methanol

is one of the top ten organic chemicals produced with an annual production of approximately three billion kg. It is catalytically synthesized from H2/CO/C02 synthesis gas, derived from methane (natural gas) by steam reforming, by the reactions CO+2H2

4

CH30H

(7.1)

AH06mK= -100.46 kJ/mol and AG"600K = +45.36 kJ/mol

C0,+3H,

+ CH30H+H2O

MomK = -61.59 kJ/mol and

(7.2) = +61 .XO

kJ/mol.

Simultaneously occumng with methanol synthesis over catalysts such as the copper-based catalysts is the water gas shift (WGS) reaction CO+H2O

4

C02+H2

&f06m~ = -38.7 kJ/mol and

= -16.5 kJ/mol

The first catalysts used for methanol synthesis from HdCO were based on the oxides, salts, and metals studied by Patart in 1921 (ref. l), and the ZnO/Cr20j high temperature and high pressure catalysts were commercialized by Badische Anilin und Soda Fabrik (BASF) in the 1920's (refs. 2,3). It was found that pure Cr20j was a good methanol synthesis catalyst by itself at 18.0 MPa and 370-415°C (ref. 4). However, it was noted that the activity depended very much on

269

the preparation of the chromina (refs. 4 3 . As previously observed (refs. 2,3), addition of ZnO produced a large increase in the activity of the catalyst, e.g. incorporation of ZnO to the extent of 75 metal atom% with the Cr203 increased the %CO conversion from 6.3 to 14.0% at 18.0 MPa and 375°C (ref. 4). During the late 1920's, copper-based catalysts were also observed to synthesize methanol from H,/CO reactants (refs. 6-9). Pure copper oxide has been reported to be practically inactive for methanol synthesis (ref. lo), and it was observed that the small activity was dependent on the reduction temperature employed (ref. 11). While ZnO by itself exhibited catalytic activity for methanol synthesis (refs. 4,8,12,13), e.g. from H2/C0 = 2 synthesis gas at 18.0 MPa, 400-425OC, and gas hourly space velocity (GHSV) = 25,000 hr-', systematic studies of binary Cu/ZnO catalysts camed out by Frolich and co-workers (refs. 8,9,14) showed that the binary mixtures were much more active than the separate components. Although the experimental conditions utilized were severe for copper-based catalysts, i.e. temperatures >320"C and a pressure of 20.7 MPa (for methanol synthesis but 0.10 MF'a for the decomposition of methanol), a Cu/Zn ratio in the range of 20-40% was found to be optimum for the 2H2 + CO + CH30H reaction. Supports such as A1203 (ref. 15) and Cr2O3 (refs. 10,16) have been added to the Cu/ZnO catalysts, and they tended to enhance the activity of the catalysts; however, the mechanistic functioning of the supports was not investigated. Although the high activity of the copper-based catalysts was well- established during the 1920s and 1930s, the high temperature/high pressure ZnO/CrzOj catalyst remained the industrial catalyst of choice because it was relatively stable against poisons such as sulfur, chlorine, and Group VIII carbonyls, which could not be removed from the reactant synthesis gas in an economical and practical way. In the early 1960s, however, a number of efficient chemical and physical wash processes became available for the removal of the trace quantities of contaminants from synthesis gas, and this led to a renewed interest in the easily poisoned copper-based catalysts. A new generation of low pressure (
270

38 plants using the ICI process (with a total capacity of 41,126 tons/day) by the end of 1983 (ref. 24). In recent years, the trend in construction of large methanol plants has continued, as illustrated in Table 7.1 (ref. 26). Although the selective synthesis of methanol over the copper-based catalysts is now a well-developed technology, long-standing scientific issues include (i) what are the active catalytic centers, (ii) is the reactant being hydrogenated CO or CO,, and (iii) what is the overall mechanism for the synthesis of methanol. Table 7.1

Country

Methanol Plants Completed (C), Under Construction (U), Engineered (E), and Planned (P) since 1987. Source: Hydrocarbon Processing, February 1990, October 1989, February 1989, October 1988, June 1988, and October 1987. Total New or Revamped Capacity > 6.75 million tons/year. Company

India

Rama Petrochemls Ltd

Brazil

Prosint Prod Sinteticos

United States Tenneco Oil Co. Chile

Cape Punto Horn Methanol Ltd

Indonesia

Pertamina

Argentina

Petroquim Gen Mosconi

Bahrain

Gulf Petrochem. Ind. BSC

United States Ashland Chem.Co. India

Gujarat Narmada Valley Fertilizer

Iran

Natl. Petrochem. Co.

China

China Natl. Tech. Import

India

Location

Capacity

Bombay

I

Re

Rio de Janeiro

I

Ex 131

68

Process

Status

I Lurgi

C88

I ICI

C88

. Pasadena,TX Arenas

I

Re 370 748

I

Lurgi

C88

BASF

C89

I Kalinantan

330

Ensenada

25

Sitra

1

Re 438

Lurgi

C

ICI

U89

1 ICI/Uhd

U89

U89

U90

Shiraz

84

Zhibo

300

Trombay

To

Taloja

Waitara

U

55

110

I

Ex 438

Lurgi

E87

Topsoe

E90

ICI

E90

I ICI/Uhd

E90 E

Bongaigaon

AlJubail

2.7

1

To1200

Huels

1

P90

P9 1

271

Country

Company

Location

Capacity

Process

Status

-

Mt/year Argentina

Petroquim Austral SA

Rio Grande

680

ICI

P92

Mexico

Petro. Mexicanos

Cosoleacaque

825

ICI/Uhde

P96

Colombia

Empresa Colombiana Petr

Cartagena

7.2.1

P

Active State of Copper

While both pure copper and pure zinc oxide will produce small amounts of methanol from H2/CO and H2/CO/CO2 synthesis gas mixtures, higher temperatures and lower gas hourly space velocities (GHSV) than those of practical interest must be used (refs. 27-29). The active methanol synthesis catalysts contain both copper and zinc oxide to produce significant yields under the low temperature and low pressure industrial conditions. Although zinc oxide is more active than copper in the pure compounds, there is a consensus that some state of copper is the active component of the binary and ternary catalysts. In summarizing the experimental observations made before 1955, Natta stated (ref. 30), “The true catalyst of the methanol synthesis is not the metallic copper phase (cubic), but the copper oxide which is present in the incompletely reduced catalysts (for example, as copper chromite in the CuO-CrzOs catalysts).” However, there has continued to be a controversy in regard to the active state of copper and the following species have been proposed to be the catalytically significant species: (i) (ii)

metallic copper supported on ZnO (31-33),

cu6+in ZnO (27,34-40) or other oxide supports (4 1-45), and

Cu6+ on metallic copper (46-49). An observation that would tend to support proposal (i) that metallic copper plays the critical role in the methanol synthesis reaction is the high methanol synthesis activity achieved with Raney copper catalysts prepared from Cu/Zn/Al alloys (refs. 3330,s 1). Additional support for proposals (i) and (iii) is the reportedly direct proportionality of the activity of Cu/Zn/Al and Raney copper catalysts with the surface area of the metallic copper component (refs. 31,33,46,48,50,5 1). However, over the binary Cu/ZnO catalyst, no correlation of the methanol synthesis activity with the metallic copper surface area was found (refs. 28,38), nor was one observed with the bimetallic lanthanide/copper catalysts (refs. 44,4532). Support for proposal (ii) is indirectly provided by a temperature programmed reduction study of Cu/Zn/Al catalysts (ref. 53), recent EXAFS studies with binary Cu/Zn catalysts (refs. 39,40), and by analogy to homogeneous catalysts such as the (iii)

272

200c

Cu/alkali methoxide bifunctional system that

AETHANOL YIELD. a/ka catal/hr

produces methanol (ref. 54). In additon, there is a very strong electronic interaction between the copper and ZnO in the active catalysts

1500

(ref. 34).

7.2.2 Hydrogenation of CO vs CO,

1000

Industrially, a H2/CO/C02 synthesis gas mixture with CO/CO2 2 1-4 is utilized for the synthesis of methanol, and it was established early (ref. 17) that the presence of

500

0

0

5

10

15

20

25

30

CO2 CONTENT, mol%

Fig. 7.1 The dependence of methanol yield at 225°C ( ) and 250°C ( ) on the C 0 2 content in the H2/(CO+CO2) = 70/30 mol% synthesis gas at 7.6MPa and GHSV=6100l/kg catalkr. The filled data point A represents the yield obtained when the CO in the synthesis gas was replaced with argon to yield a H2/Ar/C02 = 70/24/6 mol% reactant mixture. Composed from data given in Ref. 59.

carbon dioxide greatly enhanced the durability of the catalyst in terms of carbon conversion to methanol. In the absence of C02, the catalyst had a lower initial activity, and the catalyst deactivated at a rather rapid rate, e.g. 15.4%/day for the first 2.6 days on stream and 8.2%/day

thereafter,

while

with

a

H2/CO/C02 = 75/15/10% synthesis gas the deactivation was 7.7%/day for the first 2.6

days with no deactivation after that initial period (ref. 17). Recently it has been found that the CuEnO catalyst can be stabilized and the activity doubled in the presence of CO2- free synthesis gas by doping the surface of the catalyst with a small amount of heavy alkali ion, e.g. cesium (refs. 55-57). For example, a Cs/Cu/ZnO catalyst (0.4/30/70 mol%) prepared using CsOH yielded 455.6 g methanol/kg catal/hr (22.2% CO conversion to methanol) at 25OOC and 7.6 MPa with HdCO = 2.33 synthesis gas with GHSV = 5000 hr-I (ref. 57). For an undoped Cu/ZnO = 30/70 mol% catalyst under the same reaction conditions, the methanol yield was 230.3 g/kg catal/hr (1 1.3% CO converion to methanol). Under these conditions, it is clear that CO is selectively being converted to methanol via Reaction 7.1. Engineering studies by Lurgi (ref. 58) led to the conclusion that the hydrogenation of CO to methanol proceeded much more rapidly than the reaction with CO, at the same reaction conditions. More recently, the effects of C 0 2 in H2/(C0 + COz) =70/30 vol% synthesis gas were systematically studied over Cu/ZnO = 30/70 mol% catalysts, where the H2/CO/CO2 ratios were varied between 70/30/0 and 70/0/30 vol% (ref. 59). As shown in Fig. 7.1, it was observed that methanol was produced from both H2/C0 and H2/COz nuxtures, and that reactant mixtures containing H2, CO, and C 0 2 produced significantly higher yields of methanol than either of the binary synthesis gas mixtures. Indeed, the observed yield of methanol from H2/CO/CO, = 70/28/2

273

vol% synthesis gas was 6-7 times greater than those from the binary H2/C0 and H2/C02 reactants, and it was indicated that with the listed ternary reactant, approximately 98% of the methanol produced at 250"C, 7.6 MPa, and

METHANOL YIELD, g/kg catallhr I

GHSV = 6100 vkg cat/hr was produced from CO (ref. 59). Thus, C02 was observed to be a promoter, but at concentrations higher than 2 vol% in the H2/(CO + C02) = 70/30 synthesis gas it increasingly behaved as a retardant due to competitive chemisorption. The presence of water in the synthesis gas also influences the catalytic activity of the Cu/ZnO catalysts. It was shown, see Fig. 7.2, that water in the temperature range of 215235°C was a stronger promoter at low concentrations and a stronger retardant at the higher concentrations when it replaced C 0 2 in the synthesis gas mixtures (ref. 60). Using

0

5

10

15

H 2 0 CONTENT, mol%

Fig. 7.2 The effect of water at 225°C (W) and 235°C ( 0 ) in the H2/CO = 70/30 mol% synthesis gas at 7.6MPa and GHSV = 6100 Ikg catal/hr (initial). The GHSV gradually increased because the H20 was added to the H2/CO synthesis gas. Composed from data given in ref. 60.

D20 instead of H2O resulted in a single denterium atom being incorporated into methanol to produce CH2DOH(D). This demonstrated that water was a primary reactant in methanol synthesis and that it reacted with CO to form a kinetically significant intermediate that contained hydrogen from water, e.g. formate, rather than simply forming C02, via the WGS reaction (Eq. 7.3), that would be subsequently hydrogenated to methanol via a formate species. In other studies with H2/CO/C02 mixtures as reactants, evidence obtained via isotopic labeling of CO and CO, indicated that CO, was the primary reactant (via Reaction 7.2) over the commercial Cu/Zn/Al catalysts under industrially employed reaction conditions (refs. 46,48,61-64), as well as over a Cu/ZnO = 30/70 catalyst in a batch reactor (refs. 65,66). In an early study (ref. 611, it was found that the methanol formed from a H2/CO/C02/N, = 75/1/20/4 feed with I4C carbon dioxide had the same specific radioactivity as the C02. With an ICI Cu/ZnO/AI2O3 catalyst, 12CO/14C0, and 14CO/12C02 mixtures were used for a large range of CO/CO, ratios, and quantitative analysis of 14C in the product methanol led to the conclusion that C 0 2 hydrogenation was the primary path to methanol under the reaction conditions of 250"C, 5 M a , and GHSV = 10,OOO-120,000 hr-1 when any significant quantity of C02 was present, e.g. pc02Ipco 2 0.02 (ref. 64).A recent examination of the reaction rates for methanol synthesis from CO and CO, concluded that both reactants are hydrogenated to methanol when present in H2/CO/CO, mixtures and that CO hydrogenation (Reaction 7.1) is retarded by the presence of C 0 2 but the reverse is not true (ref. 67).

214

PRODUCT YIELD,g/kg catal/hr 0

C"lZ"0

It is pertinent to point out that the C02 retardation effect is also evident over methanol synthesis catalysts that are not based on Cu/ZnO. For example, C 0 2 weakly retards methanol synthesis over supported Pd catalysts (ref. 68), but it is a strong retardant over binary lanthanide/copper catalysts (ref. 44). The C 0 2 retardation effect is due to competitive adsorption of C 0 2 on the catalytic active sites, where in some cases, e.g. with the Cu/ZnO-based catalysts, it can also undergo hydrogenation to form methanol. Thus, both

CsOOCH CONTENT, mol%

Fig. 7.3 The effect of the Cs content of the CuEnO ( 0 ) and Cu/ZnO/CrzOg (A) catalysts on the methanol yield at 250OC and 7.6MPa with CO2-free H$CO = 2.33 synthesis gas with GHSV = 5000 h r l . Due to differences in the bulk densities of the catalysts, the GHSV corresponds to 6120 I(STP)/kg catalhr for the CuEnO catalyst and 10,OOO l(STP)/kg catamr for the supported catalyst. Data are derived from Refs. 55 and 69.

reactions [7.1] and 17.21 can proceed to produce methanol, but the dominance of either one will be dictated by the reaction conditions and the catalyst.

7.2.3 Newer Methanol Synthesis Catalysts 7.2.3.1 CSiCuiZt10 Catulvsts As pointed out above, the Cu/ZnO

catalysts can be doped with heavy alkali to increase the methanol synthesis yield from C02- free synthesis gas (refs. 55-57). The catalytic activity is dependent on the alkali doping level, as shown in Fig. 7.3 (refs. 55,69,70) for Cs doped onto the binary Cu/ZnO = 30/70 mol% catalyst and the ternary Cu/Zn/Cr = 30/45/25 mol% catalyst. In both cases, the methanol selectivity was >97% for methanol synthesis reaction conditions with H&O = 2.33 at the optimum doping levels. It can be seen from the figure, that the highest methanol yields were obtained by doping the Cu/ZnO catalyst with 0.4-1.0 mol% CsOOCH. The optimum doping level in terms of both high activity and high selectivity to methanol was =0.4 mol% Cs on the Cu/ZnO (corresponding to less than a monolayer coverage ( ~ 2 0 % of ) alkali on the Cu/ZnO catalyst) because larger yields of side products were formed at higher Cs levels. The side products consist principally of ethanol and methyl formate. The optimum doping level of the Cu/ZnO/Cr203 catalyst was 2.5-3.0 mol% CsOOCH. This higher Cs level is a reflection of the higher surface area of the chromia-containing catalyst (=84 m2/g) compared with the unsupported binary catalyst (=36m2/g) and of some of the basic Cs dopant going onto the chromia to neutralize acidic sites. Indeed, correlation of the specific methanol activity (refs. 69,70) yielded a single maximum for the two catalysts at a loading level of

275

=0.032%Cs/m2/g catal. As will be discussed later, the

. L

c

high selectivity to methanol

,

,

,

I

,

1

2

3

4

5

6

7

I

,

8

9

I

,

,

m

c

m

. n

E

rn

the

0

0 I 7-

i

s!

catalysts discussed here were

4

I

+ W

prepared by different procedures. The Cu/ZnO = 30/70 mol% catalyst was prepared by aqueous coprecipitation in an acidic medium to form a single phase hydroxycarbonate The

,

-

The two optimized

precursor.

I

ul >I

can be drastically shifted towards higher oxygenates by simply changing reaction conditions.

45,

c

precursor

of

consists

H

LL

0

?

0

% H,O

IN THE H,/CO

1 0 1 1 1 2

SYNTHESIS GAS

Fig. 7.4 Comparison of the further promotion of methanol yield over Cu/ZnO = 30/70 mol% catalysts by the presence of the Cs dopant upon addition of H20 to the HdCO = 2.33 synthesis gas at 7.6 MPa and 25OOC with initial GHSV = 6120 l/kg catal/hr.

(CU0.3zn0.7)5(C03)2(OH)6

that is analogous to the naturally occurring mineral aurichalcite (ref. 7 1). In contrast, the ternary supported catalyst was prepared in basic medium by coprecipitation to form a hydrotalcite-like single phase precursor consisting of Cu2,4Zn3,6Cr2,0(OH)16C034H20(ref. 72). In the presence of damp H2/CO=2.33 synthesis gas that is initially C02-free, the Cs-promoted catalyst performs better than the undoped catalyst, as shown in Fig. 7.4 (ref. 73). The promotion in the methanol yield is significant at the lower and higher levels of HzO in the H2/CO = 2.33 synthesis gas, as indicated in Fig. 7.5. At the same time that the methanol synthesis reaction is promoted, the water

gdS

shift reaction is also promoted by the presence of Cs on the

catalyst (refs. 74,75). 7.2.3.2

ThlCu Alloy Catalvsts

The precursors for these catalysts consist of the intermetallic compounds ThZCu, ThCu2, ThCu3.6, and ThCu6 (refs. 76-78). Since activation of these catalysts was carried out in air, it was reported that the active surface consisted of metallic copper supported on thorium oxide. However, later studies by Daly (ref. 45) with the ThCu6 catalyst suggested the presence of Cu20 on the catalyst, and this was suggested to be the catalytically active species. The ThCu6 catalyst was the most active of those tested in forming methanol from C02-free synthesis gas, and it was appreciably more active than a commercial Cu/ZnO/A1203 catalyst under the same conditions (which are not close to industrial conditions utilized for this catalyst), as shown in Table 7.2

276

350,

,

,

,

,

,

,

,

,

,

,

,

,

(ref.78). It was also shown during a 480 hr test with H2/CO = 2.4 synthesis gas at 260"C, 6.1 MPa, and GHSV = 9,500 hr-I (ref. 78) that the methanol synthesis activity of the ThCu,

300

catalyst was maintained, while the methanol 250

selectivity greatly increased with time (up to 82% with nearly equal quantities of CH4 and

200

C 0 2 as the other products). It was also found

150

that the Th component could be replaced with 100

Zr, Hf, and Ce (ref. 76), and by rare earth metals (ref. 79). However, the latter catalysts were observed to be appreciably less active and selective than the Th-containing catalysts.

50 0 % H,O

IN THE H z/CO SYNTHESIS G A S

Fig. 7.5 Percent increase in the methanol yield upon doping the Cu/ZnO catalyst with Cs and adding H2O to the H2/CO synthesis. See Fig. 7.4.

7.2.3.3 ZrlCu Catalwts More recently, CuO/ZrO2 catalysts

prepared via carbonates or oxides have been prepared via precipitation techniques and studied for methanol synthesis activity at 5 MP a

and

160-300°C

with

CO-free

H2/C02 = 4 synthesis gas at GHVS = 17,100 l/kg cat/hr (ref. 80). The catalysts formed by carbonate precipitation and calcined to 360°C were observed to have surface areas between 19 and

63 m2/g, while the urea precipitated catalysts had surface areas between 132 and 191 m2/g after calcination at 360°C. The pure CuO and pure Z r 0 2 prepared by either method exhibited 0% CO conversion after reduction. For both types of catalysts, a maximum in the %CO conversion was found as function of the copper content, i.e. at =6% conversion for 40-70 wt% CuO for the carbonate precipitated catalysts and at 4 3 % conversion for 20-40 wt% CuO for the urea precipitated catalysts. The activity of the catalysts did not directly correlate with the observed surface areas (ref. 8 1). However, thermal programmed reduction (TPD) and X-ray photoelectron spectroscopy (after reduction in 10% H2) studies indicated that the active surface sites o n both types of catalysts were either well-dispersed Cu+ or Cu2+. In other studies of Cu-rich samples, it was shown that amorphous precursors yielded better catalysts than did crystalline precursors. For example, crystalline Cu2Zr (ref. 44) and Cu7&r30 (ref. 82) were poor catalysts, but amorphous Cu70Zr30 yielded a catalyst that exhibited an appreciable activity for methanol synthesis (ref. 82). In a study of low Cu content binary and ternary aerogel catalysts for methanol synthesis, it was reported that preoxidized CuO/ZrO, exhibited the best activity for the formation of methanol from CO-free H2/CO2 = 0.1 synthesis gas at 300°C and 3.2 MPa with GHSV = 30,000 hr-I (ref. 83). The catalysts were first pretreated with H2 or 0, for 2 hr, and they were then tested for 24 to 70 hr in a continuous flow reactor. The product yields are shown in Table 7.3, and the first

211

Table 7.2

Comparison of the Conversions of the ThCu, and Commercial Cu/ZnO/A1203 Catalysts Tested with CO2-free H2/C0 = 2.4 Synthesis Gas at 6.1 MPa and GHSV ~22,000hr-l.

Temperature OC

Yield of Methanol g/ml catal/hr

%CO Converted to CH30H

%CH30H in Exit Gas

ThCu

230 250 Cu/ZnO/AI2O3 240 280 300 320 340 Table 7.3

2.77 3.55

3.75 5.87 7.19 11.06 15.46

13.0 19.4 23.2 33.6 44.5

0.10 0.24 0.37 0.52 0.49

0.32 0.82 1.27 1.84 1.77

1.1 2.8 4.3 6.2 6.0

Product Yields over Zr02-based Catalysts at 300°C and 3.2 MPa with H2/CO2 = 0.1 nthesis Gas with GHSV = 30,000 hr-1. Derived from Ref. 83.

Catalyst

Surface Area

Pretreatment

m2/g

Zr02 Zr02 5% cuo/zro2 5% cuo/zro, 5% Cu0/5% Zn0/90% Zr02 5% Cu0/5% Zn0/90% ZrO,

217 201 215 219 161 1.50

3moc 300°C 300°C 300OC 300OC 300°C

Methanol Yicld gkglcat

35 54,35 243,176 406 352,300 432

Methane Yield gikglcat

12,5 250,88 13, 10 54,11 7, 3

54,7

number given is the initial activity at < 30 min test time, while a second number indicates the activity after the longer period of testing. XPS analyses of the CuO/ZnO/Zr02 catalysts indicated that the preoxidized catalysts contained twice as much copper (but less zirconia) on the surface as the prereduced catalysts. It was proposed that electron deficient copper was the major contributor to the active sites on these catalysts.

7.2.3.4 CelCu Catalysts Binary methanol synthesis lanthanide-copper intermetallics were prepared by 1CI researchers under an inert atmosphere or in vacuum (ref. 44). The CeCu alloy compositions were varied between CeCul,3 and CeCu3.2, and the phases detected prior to catalytic testing by X-ray powder diffraction (XRD) were CeCu2 and/or CeCu,. These catalysts were tested in C02-free H2/CO = 1 synthesis gas with GHSV = 6000 hr-l at 5.0 MPa and a range of temperatures. The principal products were methanol and methane, with impurity concentrations of methyl formate, methyl acetate, and ethanol also present. The catalysts are low surface area materials (<7 m2/g), and

278

they consisted of CeO2 and Cu ( 4 0 nm diameter) as detected by XRD after testing, although no crystalline Cu was initially detectable by XRD and intermediate hydride phases, both intermetallic and rare earth hydrides, were evidently present (ref. 84). The catalytic activities did not correlate with the Cu surface areas. In testing at 1.O MPa and 240°C with the H2/C0 = 1 synthesis gas, it was observed that the CeCul.7 catalyst was the most active and produced 1.95 vol% methanol and 0.05 ~ 0 1 %methane from the CO (ref. 44). It was noted that conventional Cu/ZnO/AI2O3 catalysts would exhibit no measurable activity under these conditions. However, the Ce-Cu catalysts that showed the higher activities at low temperatures had poor high temperature stability. Other rare earth-containing catalysts, e.g. with La, Pr, Nd, Gd, and Dy, were tested, but the most active catalysts were the Ce-Cu catalysts shown in Table7.4. In a study oriented toward air-stable catalysts that would yield stable methanol synthesis activities, Cu/Ce/Al = 50/50/5 alloys were prepared using a melt-spinning technique (ref. 85). Low temperature (<200°C) alcohol synthesis activities were probed using C02-free H2/C0 = 2 synthesis gas at 5 MPa and GHSV = 60,000 hr-l. It was surprisingly found that the crystalline Cu/Ce/Al catalyst was more active than the amorphous Cu/Ce/Al catalyst, e.g. the latter catalyst exhibited no activity at 130°C while the fomier catalyst exhibited appreciable activity, as shown in Fig. 7.6. This observation was directly opposite to that found with Cu/Zr catalysts, as noted previously. The amorphous Cu/Ce/Al alloy catalysts were harder to activate (higher temperature required) in H2/C0 synthesis gas than were the crystalline catalysts. It was proposed that activation proceeded via intermediate interstitial hydrides that cannot readily form from amorphous Cu/Ce/Al alloys (ref. 85). This contrasts with the easy formation of the intermediate interstitial hydrides with crystalline Cu/Ce/Al alloys (ref. 86) and the adsorption of large quantities of hydrogen by amorphous Cu/Zr alloys (ref. 87). The catalytic activity of the crystalline Cu/Ce/Al alloy at 130°C is shown in Fig. 7.6 (ref. 85). The initial sharp increase in methanol synthesis activity is reported to be due to a higher catalyst temperature caused by exothermic oxidation of the Ce component. After two days, the deactivation rate was =10%/day, which appeared to be due to sintering. In probing the effect of CO, in the synthesis gas over the Ce/Cu catalysts (ref. 44), it was shown that even low levels of 1 or 2 vol% of CO, in H2/C0 = 1 synthesis gas caused deactivation

of the catalysts. The deactivation appeared to be irreversible. The adsorption of CO, on airexposed catalysts was proposed to be the cause of the 75% lower activity observed for a sample of CeCu, that was exposure to the atmosphere for 1 week compared with a sample activated and tested immediately after preparation. Examination by XPS of samples left in air demonstrated the presence of carbonate-like species on the surface of the catalysts (ref. 44). From these studies, Owen et al. (ref. 44) proposed that the mechanism of methanol synthesis over these alloy-derived catalysts differs significantly from that over Cu/ZnO-based catalysts in terms of the role of copper in the synthesis and in the influence of C 0 2 on the activity. The catalytic activity does not seem to be associated with the large Cu crystallites, but rather with a fine dispersion of copper associated with the Ce02, as suggested by TEM analyses.

279

Table 7.4

Methanol Synthesis Activity for the Cerium-Copper Intermetallic Catalysts at 5 MPa at 240°C with H2/CO = 1 Synthesis Gas at GHSV = 72,000 hr*. Derived from Ref. 44. Methanol Yield, g/kg cat/hr Alloy Precursor

Untreated Max.

Heat Treated

Steady State

Max.

Heat Treatment

Steady State

Conditions

I

T

I

"C cecu0.39 CeCu2.0 CeCu2.81

__. _..

800

246

1424

816

1520

659

--_

___

960

502

448

310

cecu4.0

448

410

CeCu6.0

362

362

Suaaorted Pd Catalvsts Since Poutsma et al. (ref. 68) demonstrated that methanol was selectively formed over Pd/SiOz catalysts from H2/CO = 70/30 synthesis gas at 260-35OoC, 5-1 10 MPa, and GHSV = 3,300 hr-', Pd catalysts have been intensively studied (ref. 88-96) in terms of different supports and additives such as alkali. Even though some of these studies were carried out at atmospheric pressure (e.g. 88-91), methanol was observed to be the major product. In considering factors that could influence the selectivity of the Pd- based catalysts toward methanol, in light of the observation that Group VIII transition metals are usually methanation or Fischer-Tropsch catalysts, the following ideas were discussed (ref. 96):

time hr

7.2.3.5

METHANOL YIELD, g/g catal/hr

0.25

0.00

7

'

0

1

2

3

4

5

6

7

DAYS OF TESTING

Fig. 7.6 Activation and deactivation of a crystalline Cu/Ce/Al alloy catalyst at 130°C and 5 MPa using a C02-free H2/CO = 2 synthesis gas with GHSV = 60,000 hr-l. The curve was obtained by averaging the data presented in Ref. 85.

280

1.

the electronic structure of the metal is modified by the support (note that the specific methanol activity of catalysts prepared from P ~ ( K - C ~ decrease H ~ ) ~ in the order of Pd/La203 >> Pd/ZrO2 > Pd/ZnO = Pd/MgO > Pd/TiO, > Pd/AI2O3 = Pd/SiO2) and that Pd supported on different types of Si02 yields different activities and selectivities (ref. 95);

11.

...

the oxide supports change the morphology of the active metal (ref. 94);

111.

the catalyst exhibits bifunctional behavior, where the support activates CO and stabilizes

iv.

formate species and Pd activates H2 (refs. 91,97); and ionic Pd centers are the active catalytic sites that activate CO and/or stabilize oxygen-containing intermediates (ref. 89). It has been reported that the methanol synthesis activities of these catalysts do not correlate

with the Pd” surface areas but do directly correspond with the extractable Pd. concentrations (refs. 89,96). This has led to comparison (refs. 95,96) of these catalysts with the Cu/ZnO methanol synthesis catalysts (refs. 27,28,34,35). However, in contrast with the Cu/ZnO catalysts where the alkali promotion effect on activity is observed to be Cs > Rb > K > Na

=

undoped > Li

(refs. 55-57,98), the promotional/retardation effect of alkali on Pd-based catalysts was found to follow the reverse order, i.e. Li > Na >> undoped > K = Rb = Cs (refs. 89,91), where little or no methanol was produced over the (K,Rb,Cs)/Pd/support catalysts. Although the Pd catalysts are interesting because they can selective form oxygenates from H2/C0 = 1-3 synthesis gas under practical industrial reaction conditions, the yields of methanol fall in the 6- 160 g/kg or 1 catal/hr range (usually closer to the lower end of this range). 7.2.3.6 NaH-RONa-M(OAc)Z - Ctrtulvsts Suspended iiz Liuuids These catalysts are being developed (ref. 99) at the Brookhaven National Laboratory as black suspensions in organic solutions, such as tertiary amyl alcohol/tetrahydrofuran, which are active for the low temperature (1160°C) synthesis of methanol from C02-free synthesis gas. As designated above, R is a lower alkyl group containing 1-6 carbon atoms, e.g. tertiary amyl, and

M = Ni, Pd, or Co. A metal carbonyl of a Group VI metal (Mo, Cr, W), such as Mo(CO),, can also be added to the slurry. It is reported (ref. 99) that this catalyst system consists of a “hydridic” hydrogenating agent that selectively formed methanol from H2/C0 = 2 synthesis gas in a stirred autoclave at =2-5 MPa and 8O-12O0C,where the principal side-product was methyl formate. 7.2.3.7 Homopeneous Methunol Synthesis Cutulvst Initial development of a main-group solution phase oxide catalyst, CH3SiOSiCH3, has been camed out (ref. 100). The chemistry displayed by this novel oxide catalyst is distinct from homogeneous metal carbonyl catalysis and seems more relevant to metal oxide surface chemistry. The CO hydrogenation pathway occumng at 280°C and 30.4 MPa with H2/CO/C02 synthesis gas involves (a)

heterogeneous activation of hydrogen by the oxide catalyst,

281

(b) carbonylation of the resulting hydroxide to yield a formate, CH3Si02CH, (c) hydrosilation of fonna.te to yield a methyleneglycoxide, CH3SiOCH20SiCH3, and (d) elimination of formaldehyde from the methyleneglycoxide complex. Methanol is produced by hydrogenation of the resultant formaldehyde in competition with a more indirect path involving Tischenko dimerization of formaldehyde to yield methyl formate followed by its decarbonylation. Formaldehyde hydrogenation occurs via hydrosilation with CH3SiH followed by hydrolytic cleavage of the resulting CH3SiOCH3 with CH3SiOH. Isotopic labelling studies indicated that all hydrogens in methyl formate produced from CH3Si02CH arose solely from the formate hydrogen, i.e. (CH3)3Si02CDproduced only CD302CD. Further studies (ref. 101) of Step (c) in dioxane solvent at 250°C and ambient pressure found that the disproportionation of trimethylsilyl formate to methyl formate, carbon dioxide, and hexamethyldisiloxane is reversible, as represent by Eq. 7.4. 4 (CH3)3Si02CHHCO2CH3 + 2C02 + 2(CH3),Si,O HC02CH3 + C H 3 0 H+ C O

(7.4)

(7.5)

Methanol is subsequently formed by a much slower process that is shown in Eq. 7.5. The transformation of methyl formate into methanol by decarbonylation with (CH3)$i,0 in dioxane was also shown to slowly proceed as a linear function of time (ref. 101). This is apparently the first homogeneous disproportionation of formate to an organic product, and in this case the resultant methylformate is a precursor to methanol.

7.2.4 Newer Methanol Syzthesis Tcclzrzology A number of advances have recently been made in the technology of producing methanol via the strongly exothermic conversion of H2/CO/C02. The large heat release is usually a limiting factor in the conversion rate that is maintained because the equilibrium methanol yield decreases with increasing reactor temperature. Thus, advanced technology is aimed towards increasing the methanol yield while efficiently dissipating the heat that is produced during the methanol synthesis reaction. The heat is not wasted but is used to produce steam that can be utilized elsewhere in the plant. 7.2.4.I New Reactor .S,:c-ttwu fi)r Gas Plluse Methanol S~ntJ1e.si.s The base cases in methanol synthesis technology are the ICI (ref. 24) and Lurgi (ref. 25) processes that use A1203- or Cr203-supported Cu/ZnO catalysts in the pressure range of 5-10 MPa. In the ICI process, the reactor is a quench gas converter consisting of a single catalyst bed with lozenge distributors that inject cold H,/CO/CO, quench gas at designated depths of the catalyst bed. The temperature differences within the reactor can be as much as 75°C. In the Lurgi process, the reactor contains many catalyst-filled vertical tubes, e.g. 1 0 m long x 0.05 m in diameter, surrounded and cooled by pressurized water that is converted to steam. In the latter process, the temperature profile is approximately isothermal (operating temperature = 230-265°C) along the

282

Table 7.5

Comparison* of the Lurgi methanol synthesis process with the isothermal tubular and adiabatic fixed bed GSSTFR process and the RSIPR liquid adsorbent process. Lurgi

sothermal No. of Sections Total Catalyst, tons Catal. Distribution, ton sisection Reaction Tube Length, m Tube Diameter, m Recycle Energy, MW Steam Export, todton of methanol Overall CO Conversion,% Gas Recycle Ratio Cooling Water Consumption tonlton

RSIPR

GSSTFR Adiabatic

1

4

80

24

__

6,6,6,6

10

0.05 1.5 0 90 5

38

0.4 0.25 100 0

25,14 ,7,4 10 0.05 0.6 0.25 94 0.04

19

14

. .

__

0.25 100

4 50

* This data is only for comparison of the processes since the kinetics of methanol synthesis strongly depend upon the reaction conditions, e.g. pressure, feed composition, the presence of CO2, GHSV, etc. length of the catalyst bed. This has been said to allow conversions of up to 45 to 50% per pass of synthesis gas rather than perhaps 6% in the ICI reactor (ref. 102). Different engineering approaches to increase the yield of methanol and to alleviate the necessity of recycling the synthesis gas include the two newer converter systems of the Gas-Solid-Solid Trickle Flow Reactor (GSSTFR) (refs. 102-105) and the Reactor System with Interstage Product Removal (RSIPR) (ref. 105). In these converter systems, the product methanol is continuously removed from the gas phase by selective adsorption on a solid or in a liquid, which tends to drive the methanol synthesis reaction towards equilibrium. In the GSSTFR system, an adsorbent such as amorphous silica/alumina (actually a 13 wt% A1203-87 wt% SiO2 cracking catalyst (ref. 106)) trickles over the catalyst bed countercurrent to the synthesis gas flow. Either a Lurgi-type reactor can be used, or the reactor can contain a series of fixed catalyst beds with interstage cooling, via cooling coils below each bed, of the synthesis gas. In either case, the product leaves the reactor in the adsorbed state, and the adsorbent is collected in insulated tanks located below the reactor. After each tank is filled and isolated from the reactor system, desorption of the methanol is achieved by depressurizing the holding tank after the atmosphere of synthesis gas has been purged by injection of liquid methanol that vaporizes on the hot adsorbent. The desorbed methanol is condensed in a separate vessel. In the RSIPR process, a liquid rather than a solid is utilized to adsorb the product methanol at the reaction temperature. This requires a series of alternating methanol reactors and liquid adsorbent tanks. However, all of these can be operated at the same temperature and pressure so that no cooling and heating is required as in recycle processes. In addition, the methanol reactors and adsorbent tanks can be progressively smaller because of the decreasing volume of reactants and product. In this process, the liquid adsorbent is recycled, wherein the coadsorbed synthesis gas is released by expansion to reaction pressure and the methanol (and any byproduct water) is

283

subsequently released by reduction of the pressure to ~ 0 . 2MPa. The regenerated solvent is recycled to the absorbent tanks. The purged gas, consisting of synthesis gas, inerts, and C 0 2 (the conversion of which proceeds much slower to methanol than does the hydrogenation of CO) is utilized as fuel gas. A calculated comparison of these advanced processes with the Lurgi process has been made (ref. 102), and a summary is presented in Table 7.5 for a methanol plant operating at 6 MPa with a stoichiometric H2/CO ratio of 2 and having a capacity of 1,000 tons/day. 7.2.4.2 Liauid Phase Methanol Svnthesis Another approach to solving the heat transfer problem in methanol synthesis so that the CO conversion level per reactant cycle can be increased is the Liquid Phase Methanol (LPMEOH) process initially developed by Chem Systems Inc. (refs. 107-109). This process uses a supported Cu/ZnO methanol synthesis catalyst dispersed in a circulating inert hydrocarbon liquid such as a mineral oil. The liquid effectively transfers the reaction heat from the catalyst and controls the reaction temperature. In the mid-l970s, the LPMEOH research was carried out with an ebullated-bed reactor using relatively large (3-6 mm) catalyst particles fluidized by gas and liquid flow, often with C02-free synthesis gas (ref. 107). This was followed in 1979 by development of the liquid phase slurry reactor system (ref. 107). Beginning in 1981, Air Products and Chemicals, Inc. and Chem Systems Inc. jointly carried out research in scaling up the LPMEOH process to the process development unit (PDU) size. A U.S. Department of Energy owned skid-mounted pilot plant was moved from Chicago, IL to Air Products and Chemicals, Inc. facilities in LaPorte, TX. Since 1984, an extensive series of tests have been conducted with both H p i c h synthesis gas, e.g. H2/CO/C02/lnerts = 54.9/18.8/4.9/21.4 mol% with H2/(CO + CO,) = 2.3, and CO-rich synthesis gas, e.g. H2/CO/CO2/lnerts = 34.8/51.2/13.1/0.9 mol% with H2/(C0 + COz) = 0.54 (ref. 110). Typical slurries consisted of 20 wt% of the powdered catalyst in Witco 70 oil with reaction temperatures of 225,250, and 265°C (ref. 11 1). At 250°C, =5.2 m a , and GHSV = 5000 h r ' with a Hz/CO/C02/N2 = 35/5 1/13/1 mo18 synthesis gas, the yield of methanol decreased rapidly for the first 100 hr from 670 g/kg catal/hr to =450 g/kg catal/hr. A gradual deactivation was noted upon further testing for >600 hr to =400 g/kg catalhr. This demonstrated that 5-8 tons/day of fuel grade methanol could be produced from this scale of the PDU. To increase the yield of methanol at 250°C and 4 . 2 MPa with the CO- rich synthesis gas, the catalyst content of the slurry was increased to 45 wt% and the GHSV was increased to I0,OOO l/kg catal/hr and above (ref. 112). Indeed, under these conditions the yield of methanol was observed to be 960 g/kg catal/hr and 10 tons of methanouday were produced from the PDU, while the deactivation rate was 0.2%/day. A separate study using a slurry loading of 35 wt% under the same reaction conditions was camed out to determine the effect of C 0 2 concentration on the methanol yield (ref. 113). The results, shown in Fig. 7.7, indicated that the methanol productivity reached an optimum when the CO, content in the H2/C0 = 0.69 synthesis gas was 5-8 mol%. It was also shown that the addition of water to the H2/CO/C02/Inert = 51/35/13/1 mol% synthesis gas

284

METHANOL YIELD, g/kg catal/hr H2/(CO

f

COz)

* 33/66-40.W58.7

H2/CO = 0.69

"O0i

i

1 ~

with GHSV =5,000 hr-I at 25OOC and 5.2 MPa yielded a maximum in methanol yield at 2-3 molR H20 (ref. 113). At this level of water addition, the methanol yield increased from 4 1 0 g/kg catal/hr with no water addition (see Fig. 7.7) to =620 g/kg catalhr. At a higher GHSV of 10,000 hr-l, the addition of water simply produced an increasing inhibition effect on the methanol productivity. However, with a C02-free synthesis gas of H2/CO/Inerts = 58.7/40.3/1 molR at CHSV = 10,000hr-1, the methanol production rate increased from 403 g/kg catal/hr to

2oo!py 10.000

0 0

5

10

15

20

C o p CONTENT, rnol%

Fig. 7.7 The effect of C 0 2 in the CO-rich synthesis gas on methanol yield over a Cu-based slurry phase catalyst at 250°C and 5.2 MPa. Derived from Ref. 113.

77 1 g/kg catal/hr when 1% C 0 2 was added to the reactants and further to 979 g/kg catal/hr when the C 0 2 level was increased to 5%. Therefore, the presence of water can enhance the methanol productivity of the liquid phase process when the low H2/CO reactant feed contains relatively low concentrations of C 0 2

An additional approach to enhance the yield of methanol per pass is the use of staged slurry phase reactors (ref. 114), where the first reactor promotes the conversion of CO to methanol while the second reactor, following condensation of the methanol from the reactant stream, is run under conditions that enhance the conversion of C 0 2 to methanol. The flow schemes and economic evaluations for these processes, including operating performance data and plant facilities investment estimates, have been presented elsewhere for coal-generated CO-rich synthesis gas based on an integrated gasification combined cycle (IGCC) facility (ref. 1 15). A one-stage or two-stage low temperature slurry phase system for the synthesis of methanol is being developed by researchers at the Pittsburgh Energy Technology Center (refs. 116,117) based on much earlier work (ref. 1IS). The two-step process consists of carbonylation of liquid phase

methanol using sodium methoxide as a catalyst followed by hydrogenolysis of the resultant methyl formate over a slurried copper chromite or Raney copper catalyst to form two molecules of methanol, as shown below.

--

CH,OH+CO HCOOCH, + 2H2

HCOOCH,

2 CH30H

(7.6) (7.7)

The net reaction is conversion of HdCO = 2 synthesis gas into methanol under the reaction conditions of =llO"C and 3MPa. The t w o reactions can be carried out bequentially in two autoclaves, but research is underway to develop senii-batch and continuous flow proceases. It has

285

been shown that C 0 2 and H 2 0 inhibit this process and that the effect of C 0 2 is more severe than is

H2O. ‘This is due to the formation of a insoluble product (CH30C02Na) by the reaction between C02 and sodium methoxide, while the reaction between H 2 0 and sodium methoxide to form

sodium hydroxide is a reversible reaction. It seems that the hydrogenolysis step is inhibited by both the CO and methyl formate reactants. This might be caused by competitive chemisorption, where the adsorption strengths on the copper catalyst follow the order methyl formate > CO > H 2 Thus, a catalyst that dissociatively adsorbs H2 more strongly would enhance this low temperature process.

7.2.5 Deactivation of Methanol Syntlzesis Catalysts Although the current methanol synthesis catalysts are very active, selective, and rather long lived, e.g. 3 years or more, they do deactivate via poisoning and sintering processes. It is well-known that the Cu-based catalysts are poisoned by sulfur, chlorine, and iron carbonyl, and that fast reduction of the catalyst or hot spots caused by the exothermic methanol synthesis reaction can induce sintering of the catalyst components to occur. An example of the deactivation behavior of a commercial Cu/ZnO/AI2O3 catalyst is shown in Fig. 7.8 (ref. 119). It is indicated that after partial deactivation, the activity of the catalyst can be increased by increasing the reactor pressure, in this case from 7.1 MPa to 9.1 MPa. During the industrial production of methanol, a high yield of methanol is usually maintained by gradually increasing the reactor temperature as the catalyst ages and gradually loses its specific activity. However, there is a limit to how high the temperature can be increased because the sintering rate increases with temperature, especially of the metallic Cu component. In addition, the yield of side products increase as the temperature is increased. As new catalysts and processes are developed, the long term stability of the activity and selectivity is determined. The behavior of a crystalline Cu/Ce/Al alloy catalyst has already been discussed, see Fig. 7.6. The stability of a commercial Cu/ZnO/AI2O3 catalyst has also been investigated for the liquid phase slurry process being developed by Chem Systems and Air Products and Chemicals, Inc. (refs. 111,113). An example of the observed behavior is shown in Fig. 7.9, where a CO-rich synthesis gas was utilized. It was observed that a rapid deactivation of the catalyst occurred, which was then followed by a gradual uniform, but slow, deactivation of the catalyst. I n each case, the used catalyst was thoroughly characterized, and the deactivation was attributed to sintering of the Cu and ZnO components of the catalyst. A 0.4mol% Cs/Cu/znO catalyst has also been tested for its long-term stability in a microreactor system using 2.45 g of catalyst in a continuous downflow tubular reactor (refs. 69,120-122). It was found that stable activities and selectivities under methanol synthesis conditions and higher alcohol synthesis conditions were maintained for periods of time that were directly related to the purity of the inlet synthesis gas. Characterization of the catalysts showed that the deactivation rate was directly related with the quantity of iron deposited on the catalyst, evidently via iron carbonyl (refs. 120-122). Using a stainless steel reaction system with activated g/g catal/hr was calculated from charcoal traps in the inlet lines, an iron deposition rate of 1 x chemical analysis data. This corresponded to gas phase iron carbonyl concentrations of 0.46 ppm (ref. 122). Although no increase in the hydrocarbon yield was observed, the used catalysts were

286

I

Pressure Increased to 9.1 MPa

0 0.0 0

.

500

1000

5 2000

1500

TIME, hr Fig. 7.8 The deactivation profile of an industrial Cu/ZnO/A1203 catalyst under methanol synthesis conditions of 24OoC, 7.1 m a , and GHSV = 35,000 hr-l. The methanol yield increased upon increasing the pressure to 9.1 MPa. Derived from data in Ref. 119.

coated with a wax that had an average linear aliphatic chain of C,, which reduced the effective surface area of the catalyst but could be extracted by benzene. Further long term testing was carried out with a copper-lined stainless steel reactor having brass fittings and where the CO reactant was passed through a guard bed consisting of successive beds of 13X zeolite and reduced Cu/ZnO = 30/70 mol% catalyst contained in a copper-lined stainless steel reactor. The observed methanol yields are given in Table 7.6 for the first 107 hr of testing and after accelerated aging under higher alcohol synthesis conditions that were maintained

METHANOL YIELD, g/kg catal/hr

-~ ____

m m

I

r.

100

200

300

I

0 .

0

0

-.- .

400

500

600

700

TIME. hr Fig. 7.9 Deactivation behavior of a commercial Cu/ZnO/A1203 catalyst used in the sluny phase synthesis of methanol. The testing was carried out with H,/CO/C02/N2 = 35/51/13/1 mol% synthesis gas at 5.3 MPa and GHSV = 5,000 hr-*.

287

for 1000 hr. It is evident that the methanol synthesis activity had decreased after the higher alcohol synthesis testing with CO-rich synthesis gas to 60% of the original activity. Also shown in Table 7.6 is the data obtained under the higher alcohol synthesis conditions. During the first 400 hr, a steady-state activity, as well as selectivity, was maintained, as shown in Fig. 7.10. After the steady-state period, a gradual decrease in the selectivity toward the higher alcohols occurred. This was accompanied by an increase in the yield of methanol, which was due to the changing average composition of the reactant mixture over the catalyst as the selectivity changed and this affected the equilibrium conversion for methanol. After lo00 hr of testing with the HdCO = 0.70 synthesis gas at 300°C and 9.1 atm (these experimental parameters were selected by kinetic modelling of the system to produce a methanolhigher alcohol = 70/30 wt ratio high octane product, which will be described later), 96% of the total oxygenate yield was still being produced. However, the selectivity has changed from 70.7 wt% methanol to 81.4 wt% methanol.

No wax was found on the above used catalyst, and no significant sintering of the ZnO and Cu component was detected (average ZnO crystallite size from 100, 002, and 101 reflections was 14.5 nm after testing as compared with 13.0 nm after reduction but before testing, while the Cu crystallite size (1 11 reflection) was 9.5 nm after testing compared with 10.0 nm after reduction).

Table 7.6

Initial and Final Alcohol Synthesis Activities and Selectivities for the 0.4 mol% CsOOCH/Cu/ZnO Catalyst Tested With C02-free Synthesis Gas Under Methanol Synthesis and Higher Alcohol Synthesis Conditions in a Copper and Brass Reaction System. Product Yields (g/kg catalhr) ~

Products

Methanol Ethanol 1-Propanol 2-Methyl- 1 -Propano1 l-Butmol 2-Methyl- 1 -Butanol Methyl Formate Methyl Acetate Methane Ethane Water Carbon Dioxide Total Liquid Yield Wt. Ratio CO Conversion, mol%

Methanol Synthesisa start

FinishC

463.5 0.85

275.9 0.52

3.45

2.55 trace

trace 8.5 467.8 99.1/0.9 19.7

3.3 8.5 278.9 98.9/1.1 12.7

Higher Alcohol Synthesisb Finish

start 265.0 23.9 31.9 31.4 6.7 3.0 4.8 8.2 11.9 2.6 3.4 298.8 374.6 70.7/29.3 21.8

291.3 16.5 19.4 16.1 2.5 1.8 4.3 5.9 11.8 1.o 6.7 174.7 357.8 81.4/18.6 18.7 ~~~~

a Reaction conditions were H2/CO = 2.33 at 250°C and 7.6 MPa with GHSV = 6120 I(STP)/kg catalhr. Reaction conditions were H2/CO = 0.70 at 300°C and 9.1 MPa with GHSV = 3265 l(STP)/kg catalhr. Determined after the 1000 hr higher alcohol synthesis testing. Defined as methanolhigher oxygenates.

288

Methanol

t

11

4.01

0.01

I

I

I

I

I

I

I

I

I

I

I

Ethanol 1.61

1

0.41

0.0 -

I

I

I

I

I

I

I

I

I

1-Propanol

0.4 I

1

I

I

I

I

I

I

I

2-Methyl-I-Propanol

I

I 200

I

I 400

TIME.

I

I 600

I

I 800

4

I 1000

hours

Fig. 7.10 Activity profiles for a 0.4 mol% Cs/Cu/ZnO catalyst tested in a copper-lined reactor with a H$CO = 0.70 synthesis gas at 300°C and 9.1 MPa with GHSV = 3265 l(STP)/kg catal/hr.

The iron content of the catalyst had increased from 0.008 wt% in the untested catalyst to 0.010 wt% at the bottom of the catalyst bed to 0.041 wt% at the top of the catalyst bed (ref. 122).

289

7.3 HIGHER ALCOHOL SYNTHESIS CATALYSTS 7.3.1

Rationale for Higher Alcohols as Fuels

During the last two decades, there has been a growing concern that the environment on a worldwide basis has been deteriorating and that in long range planning natural resources are limited and consumed at increasing rates. This spans a wide arena for action, from lead in gasoline to acid rain to increased CO, content leading to the enhanced “greenhouse effect” to the release of carcinogenic aromatics to chlorocarbons in the atmosphere leading to “ozone holes” to land and sea pollution. Many of these problems are related to energy production and utilization. Recently, seventeen former heads of state have issued a joint statement calling on the use of alternative fuels in the industrial sector of the world economy (ref. 123). In particular, this environmental concern has been directed to the transportation market, where legislated changes to the formulation of gasoline resulted in the phasing out of lead akyls, the reduction of the high volatile components, e.g. butane, that were substituted (especially during the summer months and in the southern part of the U.S.), and the reduction of carcinogenic aromatics such as benzene and toluene. In addition, the addition of oxygenates to gasoline is now required in some areas of the United States during the winter season to reduce CO and ozone emission levels. Most liquid fuels and high octane fuel additives are currently produced from petroleum. Alternatives to petroleum-based gasoline and diesel fuel are liquid oxygenates. These are primarily ethers and alcohols, and to-date they have been utilized in North America and Europe principally as octane enhancers as lead has been (U.S.) and is being (Europe) phased out. In Europe, EEC Directive 85/536/EEC on the use of substitue fuel components in gasoline specifies that all EEC member countries have to approve the use of oxygenates in gasoline up to the limits given in Table 7.7 (ref. 124). It is predicted that in 1990 the MTBE (methyl tertiary butyl ether), methanol, t-butanol, and other oxygenate consumption in Western Europe will be 1,600,000, 1,500,000, 900,000, and

100,000 metric tons, respectively (ref. 123). However, since MTBE is dependent on the refinery

Table 7.7

EEC Directive on the Use of Oxygenates in Gasoline

______________

I Additive

1

Limit Accepted vol%

10 3 5 5 7

Methanol Ethanol Isopropanol t-Butanol Isobutanol

7 ~

II Total Oxygen Content (wtlc) ~

~~

I

2.5 ~~~

290

Table 7.8

Typical Compositions of Synthesis Gas Mixtures Produced From Brown Coal Using Oxygen Blown Gasifiers (refs. 111,129).

I ' Process Component

Texaco

S hell-Koppers

BGC-Lurgi

35.1

30.1

28.6 54.9 3.4

0.9

0.4

4.4

0.68

0.45

0.52

3.4 H2/CO Ratio

I

and steam cracker isobutene supply, the MTBE capacity will be appreciably less than the demand and the consumption, which will be met by imports of MTBE or replacement by alcohols. In the U.S., President Bush has proposed that over 1,OOO,OOO alternative-fuel vehicles be in use in nine major urban areas by 1997. During 1989, there were two bills pending in Congress that would provide for improvement in air quality through greater use of alternative fuels, and a proposal in California called for 40% of the automobiles and 70% of the trucks and buses in the state to be powered by alternative fuels by 1990 (ref. 123). Already, there are bus fleets in operation in California that are based on alcohol fuels. In these cases, alcohol fuels would not be used as octane enhancers or simply as gasoline extenders but would be the primary fuel.

In 1989, U.S. net imports of crude oil increased 9.0% over the 1988 level to about 7.2 million barreldday (ref. 125). Of course, most of this came from the Middle East and contributed a significant portion of the U.S. negative balance of trade. In contrast to petroleum, most of the world's coal is located in industrial countries, and 25% of the world's recoverable reserves are located in the U.S. (ref. 126). There are now a number of successful coal gasification technologies, and most coal gasifiers can produce a synthesis gas with a H2/C0 ratio of 0.5 to 1.1 (ref. 127). For example, in the Texaco coal gasification process (ref. 128) designed to produce H2/C0 = 0.66-0.90 synthesis gas, an output with H2/CO = 0.63 is usually achieved, with very little methane present but appreciable CO2 Additional examples are given in Table 7.8. The synthesis gas mixtures produced by these gasification processes are far below the H2/C0 or H2/(C0 + C02) ratio range of 2.3-4.0 that is obtained by reforming of natural gas and that is utilized for methanol synthesis. Thus, higher alcohol synthesis catalysts are needed that operate with a hydrogen-lean reactant mixtures and that can produce high yields of high octane alcohols.

7.3.2 Background of Higher Alcohol Synthesis Catalysts It has long been known that linear alcohols are co-products of hydrocarbon synthesis over Group VIII metals, primarily iron (ref. 130), while both linear and branched alcohols are products of CO hydrogenation over alkali-doped methanol synthesis catalysts (ref. 131). The latter catalysts, when first studied (refs. 132-136), were generally alkali-promoted oxides of zinc and manganese supported with chromia or alumina and were shown to produce 2- methyl-1-propanol in rather high

291

yields. However, the reaction conditions were severe, exemplified by the pressure range of 30-40 MF’a (300-400 atm, 4410-5880 psig) and temperatures >4OO0C (>673K, >752F). In addition to the economic penalties associated with the use of pressures in excess of 30 MPa and high temperatures that are detrimental to selectivity to alcohols, the main limitation for the industrial synthesis of higher alcohols from CO and H2 has, until recently, been the lack of selective and durable catalysts. Recently, there has been extensive research carried out to develop higher alcohol synthesis catalysts, specially those that operate at low pressures and temperatures, by the incorporation of alkali in oxide and sulfide catalysts. The systems that have been studied can be classified into four groups: (1)

Modified high

(2)

(refs. 131,137,138) or Zr/Zn/Mn/Pd (ref. 139), Modified low pressure methanol synthesis catalyst, alkali-doped

pressure

methanol

synthesis

catalysts,

alkali-doped

ZnO/CrzOg

(3)

Cu/ZnO/A1203 (refs. 69,98,140- 150), Modified Fischer-Tropsch catalysts, such as alkali-doped CuO/CoO/A1203 and

(4)

CuO/CoO/ZnO/A1203 (refs. 151-157) and alkali-promoted NiOfli02 (ref. 1581, and Alkali-doped sulfides, particularly MoSz (refs. 70,159- 169).

CuEnO and

Although some of these processes have been demonstrated on a commercial scale, the future of higher alcohol synthesis processes depends upon improvements in the reaction engineering of the higher alcohol synthesis process and in the development of highly selective catalysts. Solid state syntheses of the catalysts combined with reaction engineering investigations of the synthesis of high octane alcohols need to be performed in order to achieve these improvements. Engineering models, to be discussed later, for these processes have been developed at Lehigh University (refs. 170-172) to aid in the reaction engineering of these processes, which synthesize the higher alcohols by a number of different mechanisms that proceed by distinctly different kinetically important steps.

7.3.3 Historical Development of Higher Alcohol Synthesis The production of higher alcohols from hydrogen and carbon monoxide has been known since the beginning of this century, but as mentioned above, the processes developed lacked selectivity, overall desirable yields, and stability. In 1913, Badische Anilin und Soda Fabrik (BASF) patented a process to manufacture a mixture of alcohols, aldehydes, ketones, acids (acids are undesirable in fuel applications), and other organic compounds (ref. 173). In this process, the reaction between CO and H, occurred in the presence of alkalized oxide of cobalt or osmium and at 10 to 20MPa and 300 to 400’C. In 1923-1924, Fischer and Tropsch developed the “Synthol” process to produce higher alcohols (ref. 174). This process used an alkalized iron catalyst to convert synthesis gas to alcohols at pressures above 10 MPa and at temperatures of 400 to 450°C. After BASF established that ZnO and Cr203-based catalysts produce high yields of methanol at high pressures (2), several researchers (refs. 175-178) discovered that mixtures of methanol and higher alcohols were obtained when ZnO/Cr203 catalysts were doped with alkali salts. Between 1935 to 1945, plants using these types of catalysts were in operation in Europe. Natta et al. (ref. 131)

292

Table 7.9

Estimation of the Total Octane Demand (ref. 181)

I Estimated Gasoline Demand (million bbl/day) Average Pool Octane Total Octane Demandb

1;

1995

2010

7.12’

7.40

8.22

88.90

89.50

89.90

2.90‘

6.50

6.90

0.30 0.60

0.30 1.20

0.30 1.60

2.00

2.00 3.00

2.00 3.00

Due to:

*

* *

LeadDemise Octane Upgrade Reduced Volatility - Phase 1 - Phase 2

__

a 7.04 million bbl/day in 1987.

1 Octane Demand is equivalent to =300,000 bbl MTBE/day. Only 0.29 of this demand is expected to be fulfilled by MTBE, while 2.10 of the octane demand is expected to be fulfilled by upgrading of existing refinery units (reformer, FCC, and alkylation). Ethanol and mixed alcohols can provide the remaining 0.51 units of octane demand.

mentioned that process conditions in these plants were severe with high pressure and very short catalyst life. In the 1940s, I.G. Farbenindusme and the Ruhrchemie in Germany developed the “Synol” process (ref. 179), which was based on medium-pressure (2 MPa) Fischer-Tropsch iron catalysts with operating conditions optimized to enhance the production of alcohols. After 1945, with the growing availability of petroleum and the increasing demand for pure alcohols for chemical uses, these plants became economically unattractive and consequently were demolished. In contrast, in the 1950’s the industrial development of high pressure methanol synthesis was considered mature and zinc chromite was a very satisfactory catalyst. However, in the late 19hO’s, process and catalyst advances resulted in the development, principally by ICI (refs. 18,22,180) as described earlier, of Cu/ZnO/A1203 and Cu/ZnO/Cr203 catalysts for the low temperature, e.g. 250°C and low pressure, e.g. 5-10 m a , synthesis of methanol from natural gas-derived synthebis gas. In the early 1970s, there was renewed interest in the synthesis and use of higher alcohols as a synthetic liquid fuel, which was initiated by the Arab oil embargo that created an energy shock in the Western World. There was an intensive world-wide research effort on the production and use of synthesis gas (CO + H2) derived from coal as an alternative to crude oil. In the early 1980’s, with oil prices decreasing, research on alcohol synthesis tended to decline in importance. Currently, the main driving force for the synthesis of alcohols from synthesis gas (CO + H2) is both the anticipated increase in petroleum prices and the environmental acceptability of oxygenated fuels. Mixed alcohols (Cl-C,) are thought to be a feasible alternative (refs. 181,182), along with MTBE and other ethers, to meet the increasing demand for octane (see Table 7.9).

293

7.3.4 Newer Catalysts and Technology In this section, new catalysts that have been developed for obtaining higher alcohols from CO and H2 are summarized. Some of these catalysts have been utilized in processes that have been demonstrated on a pilot plot scale. Referring to the four types of catalysts referred to previously, the temperature and pressure ranges that are utilized can be classified as follows: (i) (ii) (iii) (iv)

Alkali/ZnO/Cr203: 300-425°C and 12.5-30 MPa, Alkali/Cu/ZnO: 275-310°C and 5-10 MPa, Alkali/Cu/Co-based catalysts: 260-340°C and 6-20 MPa, and

Alkali/MoS2: 260-350°C and 3-17.5 MPa. The range of the target process conditions for higher alcohol synthesis is in the range of those utilized for methanol synthesis, 250-290°C and 5-10 MPa, which is the lower end of the reaction conditions listed for the four types of catalysts above. Table 7.10 summarizes the different processes and their basic operating characteristics so far developed for the synthesis of higher alcohols in the presence of the catalysts mentioned above. Some

of

these

processes

have

been

discussed

in

detail

in

the

literature

(refs. 70,156,157,167-169,183-187). Recent evaluations of these systems by both industrial and academic researchers (refs. 181,186- 188) concluded, among other things, that improvements particularly in the performance of the sulfide-based catalysts are needed for more efficient conversion of synthesis gas to alcohols. The behavior of these catalyst systems will be. briefly described.

7.3.5 Higher Alcohol Synthesis over Oxide-Based Catalysts Catalysts that have been developed for higher alcohol synthesis typically contain alkali that has been surface doped onto the catalyst. The process is usually carried out with lower H2/CO ratios and at higher temperatures than is the methanol synthesis process. It can also be advantageous to utilize higher contact times (lower GHSV). The catalysts in Table 7.10 that are not based on MoS, are oxide catalysts that are typically prepared by aqueous coprecipitation techniques, followed by calcination and reduction. Some of the resultant catalysts produce linear higher alcohols, while others produce predominately branched higher alcohols. 7.3.5.1 Alkali-Promoted CulZnO Catalvsts In 1983-1985, Smith and Anderson at McMaster University and Klier et al. at Lehigh University demonstrated that the low pressure/temperature copper-based methanol synthesis catalysts could be promoted with alkali to produce the higher branched alcohols (refs. 98,141-145). For example, Smith and Anderson showed that higher alcohols were formed over a 0.S wt% K2CO3/Cu/ZnO/AI2O3 catalyst with H2/C0 = 0.5 synthesis gas at 13.2 MPa, 28S"C, and GHSV = 4000 hr-1 (ref. 143). With heavy alkali, it was found that the promotional effect was Cs > Rb > K > undoped for the higher alcohol synthesis (refs. 98,144), as well as for methanol synthesis (ref. 57). It was also observed that the optimum Cs dopant level of the Cu/Zn0=30/7Omol%

294

Table 7.10

Typical Operating Conditions for Higher Alcohol Synthesis Processes

Process (Ref.)

II

Catalyst

I

I

I

SEHTb (146,184)

KEn/Clc

Temp. "C

I

I

7x0 Conv .a

I

350-425

14 20-60

(147,185) 12-15

1

1

(151-153)

Cs/Cu/ZnO, 275-325 Cs/Cu/ZnO/Cr20

LUHAS (69,148150)

5-10

Dow ( 160,162165,169)

K/M0S2 or 255-325 WCO/MOS~

3-20

Union Carbide (161)

Alkali/MoS2

300

2-5

255-300

8.2

LUHAS-2 (167,168)

CS/MOS~

I

0 0.45-1

3,0007,000

10-20

3,000-

10-20

10,Ooo

I

i 1

0

12,000

5

2,0008,000

5 - 20

a Exclusive of

C02 SEHT = Snamprogettflnichemaldor Topsoe The catalysts usually contain Cu, e.g. K0,023C~~,0~~ZnCr0.33 Lurgi and Siid Chemie

catalyst was -0.4 mol% Cs (ref. 149) under higher alcohol synthesis conditions that utilize CO2-free CO-rich synthesis gas, e.g. see Table 7.1 1.

Table 7.11

Effect of Cesium Loading of the Binary Cu/ZnO (30/70 mol%) Catalyst on the Selectivity (S) for Higher Oxygenate Synthesis, where S is defined as C2+ Oxygenates Methanol + C2+ Oxygenates

x

100

in wt%. A H2/C0 = 0.45 Syi hesis Gas was Utilized at 310°C and 7.6 MPa wlth

b h z

~~

~~

_____

Product Yield, g/kg catal/hr Catalyst Undoped Cu/ZnO 0.25 mol% Cs/Cu/ZnO 0.34 mol% Cs/Cu/ZnO 0.43 mol% Cs/Cu/ZnO 1.50 mol% Cs/Cu/ZnO

- - _ _ _ ~

.~

157 162 217

42.8

58.4 45.8 16.7

295

In an early study (ref. 14.5) probing the influence of reaction parameters on the yield and selectivity of the higher alcohols over a 0.4 mol% Cs/Cu/ZnO catalyst from C02-free synthesis gas, it was observed that catalysts prepared with CsOH or CsOOCH gave similar catalytic conversions and selectivities. In addition, higher temperatures and contact times favored the formation of 2-methyl-1-propanol relative to methanol and that a wt ratio of 1.5 was obtained for the branched alcohol relative to methanol at 325°C and the longest contact time (4.2 sec) studied (ref. 145), as shown in Figs. 11 and 12. The alcohol selectivity was also studied as a function of the H2/CO ratio of the synthesis gas. The selectivities observed for the alcohols are given in Table 7.12 (ref. 145). It is evident from the Table 7.12 that the lower H2/C0 ratio favored carbon chain growth to the higher branched alcohols and higher selectivity to oxygenates, although the level of CO conversion was lowered. Tables 7.11 and 7.12 demonstrate that the Cs/Cu/ZnO catalyst can produce a C2-rich alcohol mixture with a rather high productivity.

Table 7.12

Composition (wt%) of the Oxygenate Product (CO2-Free Basis) formed over the CsOWCu/ZnO = 0.4/30/70 mol% Catalyst at 325°C and 7.6 MPa with Synthesis Gas at GHSV = 860 hr-l. Derived from Data in Ref. 145. ~~

~

I

I

!-

~

~~

Methanol Ethanol Propanol Butanol 2-Methyl-1 -Propano1 Pentanol 2-Methyl-1-Butanol C Z - CEsters ~ C+4 Aldehydes C4-C5Ketones Other C& Oxygenates _

.

~

~

~

1 .oo

0.75 ~

~~

. -..

~~

0.45

Products

__

~~

H2/C0 Synthesis Gas Ratio ~~

40.06 3.60 11.48 2.5 1 26.71 0.76 4.45 1.23 0.91 0.92 6.52

23.83 2.22 7.62 2.37 35.93 0.94 7.86 1.35 2.33 1.14 12.61

3.5.13 3.13 11.12 2.32 26.98 0.94 4.82 1.22 1.79 1.63 9.54

98.6

9.5.0

94.5

23.6

29.0

~~~

I 9% Oxygenates in the Product

I

~

~

I

% Carbon Conversion to Products,

exclusive of C02

~~~~~~~~~~~~

~

~~

296 100

ALCOHOL YIELD, g/kg catallhr

90

0

80 70

60 50 40

30 20

10 0 1.65

1.70

1.80

1.75

1000/K Fig. 7.11 The yield of alcohols formed over a 0.4/30/70 mol% CsOWCu/ZnO catalyst from H2/CO = 0.45 synthesis gas at 7.6 MPa and with GHSV = 1200 hr-l as a function of temperature. MeOH = methanol, EtOH = ethanol, PrOH = Propanol, 2-Me-PrOH = 2-methyl-1-propanol, and 2-Me-BuOH = 2-methyl-1-butanol.

7.3.5.2

SuDDorted Alkali-Promoted CulZnOlM@J Catalvsts A systematic study of promoting A1203- and Cr203-supported Cu/ZnO catalysts with Cs has been carried out (ref. 150), where the ternary catalysts were prepared via hydrotalcite-type precursors (ref. 72) with Cu/Zn/Cr = 30/45/25 mol%. As shown in Table 7.13 for a Cr203-containing catalyst, the %CO conversion and the selectivity toward the higher alcohols

2-Me-PrOH YIELD/MeOH YIELD RATIO 1.6 1.4 1.2

1.o 0.8 0.6

0.4 0.2

0.0 0.5

1.5

2.5

3.5

4.5

CONTACT TIME, sec Fig. 7.12 Selectivity to 2-methyl-1-propanol relative to methanol over a 0.4 mol% CsOWCu/ZnO (30/70) catalyst with Hz/CO = 0.45 at 7.6 MPa as a function of temperature and contact time.

297

exhibited a dependence on the Cs doping level, where the highest values were obtained with the 3.0 mol% doping level. Table 7.14 presents the yields of the products, and it can be observed that the 3.0 mol% Cs catalyst produced =310 g of alcohols/kg catal/hr, while the formation of dimethylether was suppressed. i.

The following observations can be made from the data given in Tables 7.13 and 7.14: Doping of the ternary Cr203-containing catalyst with Cs increased the selectivity for higher alcohols, and the maximum selectivity was produced by a Cs doping level of approximately 3 mol%,

ii.

Alcohols most affected by Cs doping were 1-propanol, 2-methyl-1-propanol, and

iii.

2-methyl- 1-butanol, while the yields of ethanol and 1-butanol were relatively unaffected, as was also observed with the Cs/Cu/ZnO catalyst, and Unlike the Cs/Cu/ZnO system, dimethylether was observed in the reaction product over the Cs/Cu/ZnO/Cr203 catalyst, but its yield decreased rapidly with increasing Cs loading of the catalyst, which is attributed to neutralization of the acidity of the support. It is evident that much higher levels of Cs doping were required with the supported catalyst

than with the unsupported CuKnO catalyst. It has been pointed out that the surface areas of the supported catalysts were appreciably higher (80-86 m2/g for the tested catalysts) than for the binary tested catalysts (32-38 m2/g). In addition, the yield of alcohols over the optimally Cs-promoted

Table 7.13

Short-term Testing Product Selectivitiesa for Cesium Formate Promoted Cu/ZnO/Cr203 Catalysts Tested under Higher Alcohol Synthesis Conditions (3 10°C, 7.6 MPa, H2/CO = 0.45, GHSV = 5330 l(STP)/kg catalhr). ~-

I

-

3 0 Conversion, mol%

~

~

~

-

p

-~

~~~

~p -

~

p

C02-FREE SELECTIVITIES. carbon atom% ~

Total

Hydrocarbons

Esters

_ _

~-

I

20.10 -~

~

-

~

p

~

--

~~

p

-

~

5.26

21.20

29.81

13.65

2 1.96

34.39

10.58

2.00

-_____

3.0 mol% Cs/Cu/Zn/Ci (20 hr)

--

p

~

0.8 mol% Cs/Cu/Zn/Ci (24 hr)

-

9.36

22.69

81.89

Dimcthy Ether

c2+

c2+

Alcohols

~ _ _ _ _ _ _ _

5.0 mol% Cs/Cu/Zn/CI (22 hr)

.

p~

30.01

16.88 1

--

. .

~~

6.96 ~~

_.

--

a The major products not included in the selectivities were the aldehydes, which constituted as much as 3.49 carbon atom% (over the undoped Cu/Zn/Cr catalyst) of the product mixture observed over these catalysts. Testing period at steady state conditions.

298

ternary catalyst was higher than that over the optimally doped binary catalyst. The Cs-promoted Cu/ZnO/A1203 catalysts were studied under the same reaction conditions, and it was observed that these catalysts gave much higher alcohol selectivites, but lower CO conversion levels, that the Cr203- supported catalyst (ref. 150), as shown in Table 7.15. The higher alcohol selectivities were mainly a reflection of the higher methanol synthesis activity of the A1203-containing catalysts, as demonstrated by a comparison of Table 7.16 with Table 7.14. The phyicochemical reasons for the difference between the Al- and Cr-containing catalysts are associated with reconstitution of the hydrotalcite-like precursor upon doping the Cu/ZnO/A1203 catalyst with Cs, as discussed elsewhere (ref. 72). The reformation of the precursor was also observed to occur with a Ga203-containing catalyst (ref. 72), and this process might result in occlusion of the Cs dopant. This would prevent the Cs from playing a catalytically active role in these catalysts. The reformation of the precursor does not occur with the Cr203- containing catalyst, which exhibits the high selectivity toward C2+ alcohols. The above catalysts were prepared in basic medium via a hydrotalcite- like precursor and were Cu-poor (Cu/Zn = 0.67). Alkali-doped Cu/Zn/Al catalysts have also been prepared from acidic medium, which were Cu-rich (e.g. Cu/Zn = 2.74) and had surface areas of =130 m2/g (ref. 147). The catalysts has the composition of CuO/ZnO/AI2O3 = 55.4/20.2/9.4 wt% and were doped via spraying the prepared calcined granulated oxide catalyst with alkali carbonate solutions. The catalysts were tested with H2-rich synthesis gas and with H2/C0 =1, as shown in Table 7.17. In both cases, the synthesis gas contained small levels of C02.

Table 7.14

Product Yields (g/kg catal/hr) over Undoped and CsOOCH-Promoted Cu/Zn/Cr = 30/45/25 mol% Catalysts under the Higher Alcohol Synthesis Conditions of 31OoC, 7.6 MPa with H2/CO = 0.45 Synthesis Gas with GHSV = 5330 I(STP)/kg catal/hr. ~~~

Product

~~

~

roduct Yields Over the Undoped and Cs Doped Catalysts i.0 mol% Cs

Undoped Methanol Ethanol 1-Propano1 2-Methyl- 1-F’ropanol 1-Butanol 2-Methyl- 1-Butanol Alkanesa Dimethylether Methylacetate Othersb ~~~~

~~

263 24.5 24.5 20.1 7.4

I

20.5 0.7

22.1 18.6 8.0 38.1 ~

219 22.7 36.1 19.2 5.5

206 25.4 34.8 30.6 6.5 5.8

15.1

I

5.6 19.7

1

1

5.8 35.4

6.3 53.6 ~

Alkanes = methane, ethane, and propane. Others = methyl esters, aldehydes, ketones, C4+ linear primary and secondary alcohols, C4+ branched primary and secondary alcohols, and methyl formate.

299

Table 7.15

Product Selectivities for Cesium Formate Promoted Cu/ZnO/A1203Catalysts Obtained after Testing for 24 hr under Higher Alcohol Synthesis Conditions (310'r 7.6 MPa, H2/C0 = 0.45, GHSV = 5330 l(STP)/kg catalhr).

I co

Catalyst

1

C02-FREE SELECTIVITIES

Conversion mol%

carbon atom%

I 1 Total

90.44 91.1 1 91.16 a

Hydro- Dimcthyl carbons Ether

C02-free Alcohols

94.30

Obtained after testing for 114 hr under methanol synthesis conditions (25OoC, 7.6 m a , H2/CO = 2.33, and GHSV = 10,000 l(STP)/kg catalhr). Obtained after testing for 141 hr under methanol synthesis conditions. Obtained after testing for 126 hr under methanol synthesis conditions.

It is observed again that the dopant promotional effect toward C2+ alcohols followed the order Cs > Rb > K and that the lower ratio H2/C0 synthesis gas produced the higher yield of the higher alcohols, although the total CO conversion was decreased by decreasing the H2 content of the synthesis gas. A distinction was not made between linear and branched products. This could be most interesting since there was a systematic trend with alkali dopant in the the CdC2 ratio, e.g. with synthesis gas A the ratio was 0.68, 1.05, and 1.27 for K, Rb, and Cs, respectively. If this were due to a change in the linear/branched selectivity among the alcohols, it might indicate a change in the dominate mechanistic pathway for the synthesis of the alcohols as the alkali dopant became more basic.

Table 7.16

Product Yields (fig catal/hr) over CsOOCH-PromotedCu/Zn/Al = 30/45/25 mo1% Catalysts under the Higher Alcohol Synthesis Conditions of 3 10°C 7.6 MPa with H2/CO = 0.45 Synthesis Gas with GHSV = 5330 l(STP)/kg catalhr. Product ~

~

~~

~

1.73 mol% C 2.5 mol% C ~~

Methanol Ethanol 1-Propano1 2-Methyl-1-Propano1 Alkanesa Dimethylether Methylformate Methylacetate ~ ~ ~ _ ~ ~ 1_ a A anes = methane and ethane.

383.9 17.58 6.84 8.55 15.42 tr.

_

6.74 5.76 _

_

_

~~~

~

405.0 12.63 6.1 1 9.37 13.94

436.0 13.87 3.12 3.58 14.03 1.22 9.46 4.09 ~

-

tr.

10.40 3.36 ~

300

Table 7.17

a

Catalytic Results of Testing the Alkali-Promoted CuO/ZnO/AI203 = 55.4/20.2/9.4 wt% Catalysts with H2/CO/C02 = 70.5/29.0/0.5% (A) and 50/49/1% (B) at 30O0C, 10.1 M a , and with GHSV = 4000 hr-'. Data Obtained from Ref. 147.

Plus 260 ppm K.

7.3.5.3

Other Oxide Catalvsts Containins Transition Metul Additives

The effect of impregnating a commercial Cu/Zn/Al catalyst with a range of transition metal cations has been studied (ref. 189) since it was reported that the addition of small quantities of Co to a conventional methanol synthesis catalyst systematically destroyed the methanol synthesis activity (ref. 190). The dopant levels were in the range of 0.25-2.0%, and the testing was camed out under typical methanol synthesis reaction conditions, i.e. at 250°C and 6.6 MPa with

Table 7.18

a

Product Selectivities Observed over Na/Cu/Ru Catalysts Supported on Activated Carbon with H$O = 2.33 Synthesis Gas at 9 MPa and GHSV = 3300 hr-*.Derived from the Table given in Ref. 191.

Catalyst 2 contained twice as much Cu as Catalyst 1 HC = Hydrocarbons Other = Other oxygenated compounds This catalyst was nimded in flowing NH3 at 400°C for 3 hr

301

Table 7.19

Conversion Levels and Oxygenate Selectivities Observed over Na/Cu/Ru/Mo Catalysts Supported on Low Surface Alumina with Synthesis Gas at 9 MPa and GHSV = 3300 h r l . Derived from the Table 7.given in Ref. 192.

I Catalyst I NdCulRulMo 1/1/1/0.03 l/I/l/O. 1 1/1/1/0.3 l/l/l/l.O

I H2/C0 I I Ratio 1

Temp.

2.33 2.33 2.33 2.33

I

Total%

("C)

ICOConv.

300 300 300 305

56.7 59.8 66.7 60.6

325 325 325

47.1 23.9 18.3

~

1/1/0.3/0.3 1/1/0.3/0.3 1/1/0.3/0.3

2.33 1.00 0.43

Product Selectivity, Wt%

42.6 45.1 39.8

I

I

i

I

28.9 32.0 32.0 57.2

i

1

H,/CO/CO, = 66.7/31.3/2.0% synthesis gas with GHVS = 7,200 hr-I. Under these methanol synthesis conditions with CO, present in the synthesis gas, it was observed that some transition metal dopants strongly inhibited (poisoned) the catalyst (Co, Re), while others exhibited little or no effect on the catalytic activity (Pd, Rh, W, V, Zr, Ti). A few produced a moderate retarding effect (Mo, Ru, Pt). At these reaction conditions, there was no change in the selectivity of the Cu/ZnO-based catalyst. The Na/Cu/Ru catalysts were further modified by changing the support from a high surface area carbon to a low surface area alumina (e.g. 4.0 m2/g) and adding Mo as a catalyst component (ref. 192). The catalyst impregnation was carried out so that the Ru loading was =3 wt%. Examples of the catalysts that were prepared and tested are shown in Table 7.19. The results indicate that increasing the Mo content of the catalyst tended to increase the CO conversion level and the selectivity towards the alcohols. It appeared that the 0.30 level of Mo was an optimum dopant level.

In addition to the alcohols, other products that were observed included paraffins, olefins, acids, and aldehydes. Elsewhere, it was disclosed (ref. 191) that using alkali-doped Cu/Ru catalysts supported on carbon (catalyst complexJactivated carbon = preferably 5/95) with C0,-free synthesis gas at higher temperatures and pressures, high selectivities to C,+ alcohols could be obtained. The activated carbon that was used was a commercial high surface area carbon (1050-1250 m2/g). For example, a catalyst of this type consisted of Na/Cu/Ru oxide that could contain up to 1% nitrogen and could contain a small amount of another metal such as Ce, Cr, Fe, or Mn (ref. 191). Examples of the catalytic conversions and selectivities are given in Table 7.18. It was clearly shown that the catalyst containing a higher level of Cu tended to produce methanol rather than higher alcohols. Nitriding the catalyst significantly decreased the activity without appreciably altering the selectivity. Table 7.19 also shows the results of changing the HdCO ratio while maintaining the other reaction conditions. Decreasing the H,/CO ratio decreased the total CO conversion. However, using the CO-rich synthesis gas increased the selectivity towards the oxygenates, especially of the alcohols.

302

Table 7.20

Conversion Levels and Oxygenate Selectivities Observed over Na/CuEh/M Catalysts with H2/C0 = 1 Synthesis Gas at 5.2 MPa and 28SOC using a Contact Time of 52 sec. Derived from Tables I1 and 111 given in Ref. 193.

Catalyst

%CO Conv.

Na/Cu/Th/M

M

y/ 1.5/1/0.15 y/1.5/1/0.05 y/1.5/1/0.30 y/1.5/1/0.15

Zn Pd Cr A1

Product Selectivity,a Wt%

to Alcohols Methanol 17.1 19.5 16.5 13.9

77.7 64.6 83.5 84.0

Iso-BuOH

Other Alcohols

7.3 15.2 6.8 6.4

15.0 20.2 9.7 9.6

aIso-BuOH = 2-methyl-1-propanol

A similar approach to catalyst development was taken with alkalized copper-based catalysts containing Th rather than Ru (ref. 193). In this case, the catalysts could be typically represented as Na,Cul.SThl.oMb, where a = 0.5-1.5 wt% of the catalyst and b = 0.05-0.15 relative to Th. The metal M could be any of a large list of additives, but it was preferred to be Cr, Zn, Al, Ti, La, V, or Pd. The catalysts were prepared by aqueous coprecipitation that was induced by the addition of a Na2C03 solution until a pH of ~ 9 . 5was achieved. Examples of these catalysts are given in Table 7.20, where the selectivities are given as wt% of the liquid product. As indicated in this table, branched products were formed over these catalysts. For the Zn- and Pd-containing catalysts given in the Table, the branched/linear C4 alcohol ratios were 5.22 and 5.85, respectively. Not discussed in detail in this chapter are supported metal catalysts that have been investigated for higher alcohol systhesis. These catalysts tend to make hydrocarbons, exhibit low catalytic activity, and might or might not contained oxidized surface cations as the active sites. However, it will be pointed out that supported Rh catalysts have been found to form C2 oxygenates from H2/CO at ambient pressure (refs. 194,195). It was observed that the products from HdCO over Rh catalysts were mainly methane with acidic supports, methanol with basic supports, and C2 oxygenates with weakly basic or neutral supports (ref. 195). Supported Rh catalysts were also investigated at 1 IWa pressure for the synthesis of ethanol using both H2ICO and H2/C02 synthesis gas mixtures (ref. 196). The observed selectivities for NbzO5- and TiO2- supported catalysts are shown in Table 7.21. It was found that C-C bond formation hardly occurred with the H2/C02 synthesis gas, and that the C 0 2 reactant favored the formation of methane. With the H2/CO reactant, appreciable selectivities to ethanol were observed, especially with the Nb205 support. A ZrO2 support gave similar product selectivities, but the catalyst was =10 times less active. In all cases, increasing the temperature decreased the alcohol selectivity and increased that of methane. Doping the catalysts with Na produced a pronounced enhancement in the ethanol selectivity. Using MgO as the support gave a catalyst that was >98.5% selective for the synthesis of methane from H2/C02

Thus, it appears that supported metal catalysts do not form active and selective catalysts for higher alcohols. However, more systematic studies like those quoted are needed to probe the effect of alkali in altering the selectivities of this class of catalysts.

303

7.3.6 Alcohol Synthesis over Alkali-Promoted MoS2 Catalysts In 1984, the Dow Chemical Company and Union Carbide Corporation separately disclosed a novel catalyst system for the production of linear alcohols from synthesis gas (CO + H2) (refs. 159-161).This new catalytic system consists of supported or unsupported alkali-doped MoS, or alkali-doped Co/MoS2, is not poisoned by sulfur like the copper-based catalysts, and produces alcohols from H2/C0 = 1 synthesis gas with selectivities of 75-90%. It has long been known that sulfide catalysts, such as MoS2, are active hydrogenation catalysts under a wide range of experimental conditions (ref. 197), and it has been demonstrated that MoS2 is a methanation catalyst at 350°C and ambient pressure (ref. 198). However, addition of KOH to the catalyst and testing at higher pressure and lower GHSV resulted in the formation of higher hydrocarbons (C,+) from synthesis gas (ref. 199). As shown in Table 7.22, Dow Chemical Co. and Union Carbide found that decreasing the temperature and increasing the GHSV shifted the selectivity from hydrocarbons toward alcohols (refs. 159-161). Additional patents have been issued for the alkali/MoS2 catalysts (refs. 162-165) and for cobalt-containing alkali/MoS2 catalysts (refs. 164,203), and literature reports (refs. 166,167,169,204) are now beginning to appear for this sulfur-resistant alcohol synthesis system.

Table 7.21

The Activity and Selectivity of 3 wt% Supported Rh Catalysts with H /CO and H2/CO Synthesis Gas Mixtures (1/1) at 1 MPa with GHSV = 2400 hr- . Alkali Impregnated Catalysts Contained Na/Rh = 1 3 1.O.

1

~

I Catalyst

Synth. Ga

1Temp. I -

Selectivity (carbon atom%) {ethane ~~

180 220

2+,HCa -

p

.-

~

18.1 20.9 23.8

30.6 40.5 68.4

86.6 94.0

3.7 5.7

-

200 -

260 280 300

5.0

72.1

13.9

1.3 0.8 0.8

34.3 44.5 47.9

55.4 52.1 49.4

- -

9.0 2.6 1.9

-

_.p~

~

260 280 300

15.8 6.4

8.1 2.7

26.5 31.7 35.3

49.6 59.2 58.9

66.2 68.7

3.6 2.8

66.3

4.9

-

a HC

= Hydrocarbon This test produced the highest observed rate of alcohol + HC formation, which was 23.7 x 10-4mol/g catalhr.

304

Table 7.22

Catalytic Activities and Principal Products Formed over MoS2 and K-doped MoS2 Catalysts from Synthesis Gas. I

I

Catalysta

(Ref.) MoS2 (198)

23.5

MoS2 (200)

63

MoS2 (120,201)

I

63

I

I

I

H2/CO

GHSV

%CO

Principal

Ratio

(hr-I)

lonversior

Productsb >97% CH4

350

0.1

3.Ic

000-6500

0.5-0.8d

255

8.3

0.92

3140

5.0

55.6% CH4 43.6% c2+nc 0.8% CH30H

8.3

0.96

7775

13.1

43.6% CH4 55.7% C2+HC 0.7% CH30H

2.05

1.97

179

52

96.3% C5+ HC

676

16.5

42.5% CH30H 32.7% C2H50H 10.6% CH4 30.7% CHjOH 33.8% C2HsOH 19.1%CH4

295

I

I

8.21

10.0

1.00

3080

17.8

10.45

1.028

3171

16.3

~

10% K2CO3/66% clay MOS~DOYC

255

53.2% CHjOH 2A.97~C2H50H

12.6% CHq

(200,201)

1 1 ,

g2%KOOCH/

-I

8.3

0.92

3140

9.5

51.7% CH30H 17.4% C2H5OH 17.1% CH4

295

8.3

0.96

7805

10.67

45.2% CH30H 26.1% C2H5OH 17.8% CH4

300

2.75

1.OO

12,000

4.4

3SS% CH30H 36.6% C2H5OH 19.0% CHd

I

--

17.4% KOH/MoS2 (161)

255

I

II

aWt%

bCarbon atom%; CO2-free basis Similar results were obtained with a H2/C0 = 1 synthesis gas After 4.5 hr of testing; initial activity was approximately 60% higher Contained 20 ppm H2S Surface area of the carbon before preparation of the catalyst g Contained SO ppm H2S

Doping the MoS2 with alkali is crucial to obtain a catalyst that will produce alcohols rather than hydrocarbons. As evident in Table 7.22, rather high doping levels of potassium were used to prepare these catalysts. The concentration effect of the alkali on MoS2 as a function of reaction temperature and flow rate on the activity and selectivity, not only for alcohols but also for the undesirable side products, mainly hydrocarbons, has been studied. The activity and selectivity of

305

the undoped, K-doped, and Cs-doped MoS, catalysts will first be compared. Cesium concentration dependence studies at different reaction temperatures, as well as the effect of total flow rate, on the product distribution will then be discussed. In a later section, the results obtained with the cobalt-containing alkali/MoS2 catalyst will be presented. The results obtained from the experiments described above allow optimization of the catalyst for high yields and selectivities of alcohols and set the stage for performing the mechanistic investigations that will also be discussed in this chapter. 7.3.6.1

Effectof Alkali Dooing of MoSz- on the Activitv and Selectivitv for Alcohol Svnthesis

The alkali-doped catalysts listed in Table 7.22 had all been promoted with K. In a study of the synthesis gas conversion as a function of the different alkali cations utilized as dopants, Kinkade (ref. 161) found that the heavier alkali were more active and selective in alcohol synthesis than were the lighter alkali, as shown in Table 7.23. It will be noted that the dopant levels were very high and the conversion and alcohol selectivity levels were quite low, although the alcohol yield over the Cs/MoS2 catalyst was appreciable. This behavior was confirmed by other researchers (refs. 120,201), who compared the selectivities and activities of MoS,, KMoS,, and CS/MO S~ catalysts, as shown in Tables 24 and 25. It is clear that practically only hydrocarbons are produced over undoped MoS2, and that the selectivity is dramatically altered toward alcohols upon promotion of the MoS2 with either potassium or cesium. From the CO conversions and product yields given in Table 7.25 it can be seen that the promotional effect of the cesium was greater than that shown by potassium. The carbon number distribution of both hydrocarbons and alcohols followed the Schulz-Flory distribution, as depicted in Fig. 7.13 for the case of the Cs-doped MoS2 catalyst. The chain growth probability (a)values obtained for each of the catalysts are tabulated in Table 7.26. I t is interesting to note that a decreased upon addition of the alkali metal to the MoS, catalyst. This negative effect of the alkali metal on the a value is different from that reported in the Fischer-Tropsch literature that established that the addition of alkali to a Fischer-Tropsch catalyst

Table 7.23

Alcohol Yields and Selectivities Obtained over Alkali-doped MoS2 Catalysts (0.616 mol alkali nitrate or acetate/mol MoS2) at 300°C and 27.6 MPa with H dCO = 1.0 Synthesis Gas with GHSV = 12,000 hr-l. Data are taken from Table I1 in ref. 161. ~~

~

~~

~

Alcohol Yield

% Selectivity to ~

Conversion

C I - C Alcohols ~

(g/l catalhr)

~~

Li (nitrate) Na (nitrate) K (nitrate) Rb (acetate) Cs (acetate)

24 53 55 66 80

0.9 1.2 2.0 2.3 4.8 ~~

~

12.8 19.2 36.8-84.8 100.8 256.0

306

Fischer-Tropsch literature that established that the addition of alkali to a Fischer-Tropsch catalyst promoted the selectivity to higher hydrocarbons, i.e. promoted an increase in the cc value (ref. 205). This difference can be understood if one accepts that two different catalytic systems are being considered. Under our operating conditions, the addition of alkali metal to MoS2 is opening a new reaction pathway by promoting the formation and/or concentration of intermediate species that give rise to the dramatic shift in selectivity from hydrocarbons to oxygenates, while under the operating conditions of a typical Fischer-Tropsch synthesis the addition of alkali metal to the Fischer-Tropsch catalyst does not change the nature of the intermediates, but only their surface concentration, and therefore the selectivity to total hydrocarbons (olefins + alkanes) is conserved. It was noted (refs. 120,201) that the distribution of the sum of hydrocarbons and alcohols fits the Schulz-Flory distribution better than those of individual products. This might be an indication that the paths of chain growth are the same for hydrocarbons and alcohols over the alkali-doped MoS2 catalyst. This will be discussed in detail in the next section.

Table 7.24

Product Selectivities Obtained over Undoped MoS2,0.22 mol Wmol MoS, (10.36 wt% = 18.03 mo18 KOOCH), and 0.22 mol Cs/mol MoS2 (20.00 wt% = 18.36 mol% CsOOCH) Catalysts. Testing was carried out at 295OC and 82 MPa with Hz/CO = 0.96 synthesis gas at GHSV = 7790 I(STP)/kg catal/hr. Selectivitya ~~

Product

~~~~

~

cs/Mos;?

MoS2

p~

Hydrocarbons C1 c2

c3 c4

c5

43.64 25.24 16.98 8.92 3.86

Alcohols

C1 c2 c3 c4

0.67

i I 1 t

17.88 3.73 1.17

17.80 3.31

p

~

38 34 26.85 8.98 1.43

45.15 26.14 5 32 tr

~ ~

p

p

I Esters Methyl Formate 1.25 Methyl Acetate 1.03 ~. Selectivity is base on carbon mole selectivity on a CO2-e

I

~__~lpp~lpp-~

0.54 1.07

~~

~

p

307

1

0

v I

,

I

-2.5

2

Carbon Number n

3

-

1

Fig. 7.13 The Anderson-Schulz-Flory production distribution obtained over the 20 wt% CsOOCWMoS2 catalyst at 295°C and 8.3 MPa with H2/CO=0.96 synthesis gas with GHSV = 7790 fig catalhr, plotted as log of the mol fraction of each type of product vs (1-carbon number, n).

Table 7.25

Product Yields Obtained over Undoped MoS2,0.22 mol K/MoS2, and 0.22 mol Cs/mol MoS, at 295"C, 82 MPa and with H2/C0 = 0.96 Synthesis Gas at GHVS = 7790 I(STP)/kg catal/hr. ~~

~~~

Product

-

co CO2 HZO CH4 C2H6 C3H8 C4H 10

CSHI, CH3OH C2"SOH C3H70H C4H90H CH300CH CH3OOCCH3

~

~

~

-~

~~

MoS~

_

.

_

wMos2 -~

~~

4050.16 23.5.8 1 135.69 89.11 48.33 3 1.79 16 51 7.11 2.75

~

~~~

aethane + ethylene bpropane + propylene

~~

~

~~

~

tr 136.95 56.99 10.08 tr 3.55 2.41

191.44 96 74 28 00 4 13 2.52 4.11

10.67

14.85

~-

13.06

~

CSfMOS2 3966.86 259.59 2 82 44.65 8.63a 2.68b

4086.92 199.89 4.65 27.00 4.64"

~

CO converted to products exclusive of C02 (molkg c a t h )

~

~~

308

Table 7.26

Chain Growth Probability (a)Values Obtained over MoS2-Based Catalysts with H2/CO = 0.96 at 295°C and 8.3 MPa and CHSV = 7787 l(STP)/kg cat/h. Chain Growth Probability (a)a

Catalyst

Alcohols

MoS~

Hydrocarbons

0.38

.__

Alcohols+ Hydrocarbons

1

0.38

(0.22 mollmol) (0.22 mol/mol) aThe chain growth probability (a)value was calculated from the Anderson-Schulz-Flory distribution yn = an-l(l - a), where yn is the mol fraction of the product with n carbon atoms and a is the chain growth probability or chain growth parameter.

7.3.6.2

Effectof Cesium Concentration on the Activitv and Selectivitv of Alcohol

Svnthesis Since the superior promotional effect of cesium has been confirmed, cesium concentration dependence studies were performed to establish the optimum Cs content for alcohol synthesis (refs. 120,167,201). The catalysts were 2oc

v 'RODUCT YIELD, glkg catallhr

15C

pretreated at 400'C for 1 hr under 2% H2/98% N2 flow (60 ml/min), before they were tested under

alcohol

synthesis

conditions

with

Hz/CO =0.96 synthesis gas. The yields of methanol, ethanol, and hydrocarbons as a function of the CsOOCH loading are given for reaction temperatures of 256°C and 295°C in Figs. 7.14 and 7.15, respectively. These Figs. show a maximum in the alcohol yield as the concentration of cesium on the catalyst increased, while the production of hydrocarbons was progressively suppressed. It

1 oc

5c

* I I

c 5

10

15

20

25

30

CSOOCH LOADING OF M o S ~ w , t%

Fig. 7.14 The effect of the CsOOCH content of the MoSz-based catalysts on the product yield ( e total alcohols; A methanol; 4 , ethanol, and m hydrocarbons) at 256°C and 8.3 MPa with Hz/CO = 0.96 synthesis with GHSV = 7750 fig catal/hr.

also appears that the maximum alcohol yield shifts toward higher dopant levels as the temperature was increased. Small quantities of olefins and methyl formate were produced over the high Cs- content catalysts. The yield of water was quite low since most of the water fomied as the by-product of hydrocarbon and

309

higher alcohol synthesis was converted to carbon dioxide through the water-gas shift reaction catalyzed by the CsOOCH/MoS, catalyst.

PRODUCT YIELD, g l k g catallhr

1

Effect ofReaction TemDerature and Pressure on the Selectivin to Alcohols at Different cs LoadinQs The selectivity to alcohols and hydrocarbons 0 5 10 15 20 25 30 as a function of temperature at different CSOOCH LOADING OF MoS2, w t % loading of cesium is shown in Fig. 7.16 Fig. 7.15 The effect of the CsOOCH content (refs. 120,201). It is evident that at a given of the MoS2-based catalysts on the product temperature, the selectivity to alcohols yield ( 0 , total alcohols; A , methanol: 4 , ethanol, and , hydrocarbons) at 295°C and increased with increasing cesium content in 8.3 MPa with H&O = 0.96 synthesis with the catalyst. Increasing the reaction GHSV = 7750 l/kg catalhr. temperature, however, led to a decrease in the selectivity for alcohols, and this effect was more noticeable over the 2.5 wt% CsOOCH catalyst in which the selectivity was dramatically altered from alcohols to hydrocarbons above 275°C. As the Cs dopant level of the catalysts was

7.3.6.3

increased, there was a less pronounced shift toward hydrocarbon formation as the temperature was

loo

-"

SELECTIVITY TO PRODUCTS, mol%

i

1

0'

230

Alcohols

I

I

I

I

I

I

1

240

250

260

270

280

290

300

TEMPERATURE. OC Fig. 7.16 The effect of the CsOOCH content (in wt%: V , 2.5; 0 , 5.3; A , 10; and m, 20) of the MoS2-based catalysts on the selectivity to alcohols (filled symbols) and hydrocarbons (open symbols) as a function of temperature. The reaction conditions are given in Figs. 14 and 15.

310

increased. At the same time, the selectivity toward C2-C, alcohols was increased by both increased dopant levels and higher temperatures. This was consistent with the higher apparent activation energies observed for the formation of ethanol and propanol over the 20 wt% Cs/MoS2 catalyst as compared with methanol formation (ref. 167). A K/Mo/Carbon = 1.5/20.0/78.5% catalyst (surface area = 330 m2/g) was studied at higher pressures with a H2/C0 = 1.21 synthesis gas at 300°C with GHSV = 3095 Vkg catal/hr (ref. 160). As the pressure was increased from 10.5 MPa to 13.9 MPa and 17.4 MPa, the total%CO conversion increased from 20.2% to 25.5% and 31.0% respectively. In each case, the major product was CO, (=38 carbon atom%). The selectivities toward the hydrocarbons showed a slight decreasing trend with increasing pressure, while that of the alcohols showed no systematic trend. The selectivity among the alcohols also showed no systematic change with pressure. 7.3.6.4

Effectof Reactant Contact Time The contact time dependence of the selectivities for alcohols and hydrocarbons formed from H&O = 0.96 synthesis gas at 8.3 MPa was studied over the 5.3 wt% CsOOCH promoted catalyst at 256°C in the range of GHSV=2500 to 75001/kg catal/hr and over a K2C03/(Co/MoS2) = 10/90 wt% catalyst at 303OC and GHSV = 1900-7600 l/kg catal/hr (refs. 120,201). The results are presented in Fig. 7.17, and it is apparent that the selectivity for alcohols was increased at short contact times, while the selectivity towards hydrocarbon synthesis was increased by higher contact times. This behavior indicates that the hydrocarbons are likely formed as secondary products, 7.3.6.5

Effectof CO?. - H2S, and Olefins in the Svnthesis Gas

SELECTIVITY TO PRODUCTS, mol% 100

40

I

I

1 Hydrocarbons

20 "

0.0

0.5

1.0

1.5

2.0

CONTACT TIME, sec Fig. 7.17 Effect of contact time on the selectivities of the 5.3 wt% CsOOCWMoS2 catalyst Nand (0 and A).

A ) and of the K ~ C O ~ / ( C O / M O=S10/90 ~ ) wt% catalyst

311

Table 7.27

The Effect of CO, in the H$CO = 1 Synthesis Gas on the Yields of the Products over a K2CO /MoS2/Carbon = 5/19/76 wt% Catalyst at 305°C and 10.1 MPa. Data are derived &om Tables XXIX and XXX in Ref. 202.

vy L

T

8

C

O

1 CH30H

166.9

1

I

Product Yield (g/kg cat/hr) C2H50H

70.4

22.5

145.7

40.6

15.0*

18.9

144.0

33.5

30.3e

12.0

81.5

25.1

C3H70H 22.9

GH,

CH4

II

64.2

I

19.9

25.1

H2O !

7.lf

21.8

ahcludes conversion to C 0 2 bTotal GHSV = 2000 hr-1 CTotal GHSV = 1600 hr-1 dTotal GHSV = 1900 hr-1 eTotal GHSV = 1800 hr-1 14.2 g/kg catal/hr

Most of the research carried out with the alkali-promoted MoS2 catalysts have utilized H2/CO = 1 synthesis gas. It was stated in the first Dow Chemical patent application (ref. 160) that the presence of larger amounts of carbon dioxide in the synthesis gas retarded the activity of the catalysts. This retardation was subsequently studied by the British Coal Research Establishment (ref. 202) with a carbon supported K/MoS2 catalyst, where the carbon initially had a surface area of 1400 m2/g. The observed effect of the COz content of the synthesis gas on the yields and selectivities of the products are shown in Tables 7.27 and 7.28. Overall, increasing the C02 content step-wise to 30.3 vol% decreased the %CO conversion while not significantly altering the alcohol/hydrocarbon ratio. At low CO, levels, e.g. comparing the 6.7 vo18 concentration with the pure H&O = 1 synthesis gas, the extent of CO conversion was hardly affected, but it is notable that the higher alcohol yield relative to methanol drastically decreased, e.g. on the basis of g/kg catal/hr, the (C,

+ C,)/C1 alcohol ratio dropped from 0.56 to 0.35 when 6.7 vol% COz was added to

the synthesis gas. It was concluded in the report that the C 0 2 additive behaved as a reversible diluent, but the influence on the selectivity among the alcohols (and hydrocarbons) was not discussed. . Researchers at Dow Chemical Co. (refs. 163,165) have studied the effect of adding small amounts of hydrogen sulfide to the H2/CO synthesis gas. An example of the influence of the H2S is shown in Table 7.29, where high pressures have been employed. In general, increasing the content of H2S tended to increase the CO conversion but decrease the selectivity towardalcohols. However, the selectivity among the alcohols was shifted to the C2-C5 linear products relative to methanol.

Thus, the presence of H2S apparently promotes the C-C bond forming reactions.

312

Table 7.27

The Effect of CO, in the H,/CO = 1 Synthesis Gas on the Yields of the Products over a K2C0 /MoS&arbon = 5/19/76 wt% Catalyst at 305OC and 10.1 MPa. Data are derived kom Tables XXIX and XXX in Ref. 202.

Vol% C0nv.a

C3H70H 166.9

30.3e

I

CH4

C2Hg

TiTrzi-

19.9

7.1

H2O

7 12.8

145.7

10.4

58.8

144.0

7.2

42.1

0.0

17.1

5.9

25.1

4.3

21.8

12.0

~

_

_

_

aIncludes conversion to C02 bTotal GHSV = 2000 hr-1 CTotalGHSV = 1600 hr-1 dTotal GHSV = 1900 hr-1 eTotal GHSV = 1800 hr-1 14.2 g/kg catalhr

Studies camed out at Union Carbide Corp. found that high selectivities toward particular linear alcohols could be obtained by feeding olefins with the H2/CO synthesis gas over alkali- or alkaline earth-promoted MoS2 catalysts (ref. 206). The presence of the C, olefin altered the selectivity from a Anderson-Schulz-Flory type of alcohol distribution (plus hydrocarbons) to one where the Cn+l alcohol predominated the product. For example, addition of ethylene to a H,/CO = 1 synthesis gas produced a high selectivity to propanol, while a H2/CO/propylene reactant mixture forms principally n-butanol.

Table 7.28

The Effect of C02 in the H2/CO = 1 Synthesis Gas on the Product Selectivitiesover a K&03/MoS~/Carbon = 5/19/76 wt% Catalyst at 305°C and 10.1 MPa. Data are derived from Tables XXIX and XXX in Ref. 202.

“Total GHSV = 2000 hr-I bTotal GHSV = 1600 hr-I CTotalGHSV = 1900 hr-l dTotal GHSV = 1800 hr-l

_

313

Table 7.29

Alcohol Yields and Selectivities Obtained over a Potassium-doped MoS2 Catalyst (10 wt% Potassium Carbonate + 66 wt% MoS2 + 20 wt% Bentonite Clay Binder + 4 wt% Carbon Lubricant) Prepared as a Mechanical Mixture. Data are taken from Table I1 in Ref. 165.

282 Pressure (MPa)

21.2

GHSV (hr’)

5220

Total CO Conv. (mol%)

Alcohol Selectivity (%) (CO2- and H2O-free)

1.10

1 1 0

11.2

I

%CO Converted to C 0 2

3390

1.18

H2KO H2S (PPW

17.7

Alcohol Distribution (%)a Cl C2 c3 c4 I

59.6 29.5 8.5 2.4 0

HzO (wt% of liquid phase)

I

1

1

I

I

60

16.0 3195 1.20

160

275 16.7 3075 1.2s 165

16.2

21.8

21.0

26.4

32.8

31.6

77.9

72.5

74.1

47.5 35.7 11.8 4.2 0.7

37.8 42.9 13.9 4.1 1.2

41.4 40.6 12.1 3.8 2.0

2.1

1.95

Time on Stream (hr)

383

339

Time with H2S (hr)

95

52

aCarbon mol% on a CO2-free basis

However, some isobutanol and butyraldehyde were also formed. In addition, the hydrogenation product, propane, was also formed, sometimes exhibiting high selectivities. Using a synthesis gas with a H2KO ratio i1 tended to decrease the selectivity toward the direct hydrogenation product while increasing the selectivity to the alcohol. Examples of the product distributions observed over two catalysts with and without the addition of ethylene to the synthesis gas are shown in Table 7.30. Under these experimental conditions, the conversion of the olefin was low (<5%), and an improvement in the conversion rate is needed.

314

Table 7.30

The product selectivities (mol% carbon, C02-free basis) obtained with H2/CO = 1/1 and H~/CO/ethylene= 4.5/4.5/1 reactants at 2.7 MPa and GHSV = 12,000 hr-1 to alcohols over alkali-promoted MoS2 catalysts.

I

I

Catalyst NaN03/MoS 26/74 mol%

Product Methanol Ethanol Propanol Othe+ Ethane Pro ane HC

E

a

I

3OOOC 47.3 26.5 6.9 0.0 2.0 0.9 16.4

I

25OOC

270°C

7.2 2.3 62.0 8.6 8.4 2.0 9.5

51.2 22.5 5.9 0.0 3.9 2.4 14.1

3.2 1.6 51.8 19.8 16.4 1.6 5.6

Other oxygenated products such as aldehydes, esters, and ketones. HC = hydrocarbons other than C2 and C, products.

7.3.6.6 Effectof Addinp Cobalt to the Alkali-Dooed MoS2- Catalyst The promotional effect of cobalt in oxide catalysts toward ethanol formation has been known since the early investigations of Taylor (ref. 135), who obtained an ethanol selectivity of 22% with respect to the condensate product using a catalyst containing CoS/CuO/MnO. Also in the 1930s, high yields of ethanol were reported to have been obtained over a CoO/ZnO catalyst (ref. 131). Later, Wender et al. (ref. 207) studied the reaction between methanol and synthesis gas

(CO + H2) in the presence of dicobalt octacarbonyl, [CO(CO)~],,and concluded that cobalt catalyzed to a great extent the homologation reaction of methanol to produce ethanol. Recently, researchers from the French Petroleum Institute (IFP) disclosed a copper/cobalt-containingoxide catalyst for the production of linear alcohols (refs. 15 1,153,155). Upon addition of cobalt to the alkali-doped MoS2 catalysts, a shift in the product selectivity in favor of ethanol has been noted (refs. 120,164,167,201,203).Thus, it is clear that cobalt promotes the C1 + C2 homologation step. For example, a K2COJ(Co/MoS2) = 10/90 wt7c catalyst was prepared and tested at 8.3 MPa and 304°C with H2/CO =0.96 synthesis gas with GHSV = 1940 Vkg cat/hr (ref. 167). The product distribution obtained over this catalyst is given in Fig. 7.18 along with that obtained over a CsOOCH/MoS2= 10/90 wt% catalyst. It is apparent that the addition of cobalt to the catalyst shifted the selectivity in favor of ethanol, as was reported previously by Dow Chemical researchers (refs. 164,203). Recently, Dianis (ref. 166) perfomied CO Temperature Program Desorption (TPD) studies on the MoS2-based catalysts, and reported the existence of two forms of CO on the cobalt-containing MoS2 catalyst. Dianis proposed that “the more strongly bound CO is less readily hydrogenated directly to methanol, and therefore, the cobalt

315

molybdenum sulfide shows better

selectivity

toward

35

higher alcohols.” but no details were given (ref. 166).

30

The most crucial step of the reaction cycle over the

25

alkali/Co/MoS2 catalysts is thought to be the oxidative

20

SELECTIVITY (exclusive of C 0 2 ) , C atom%

addition of methanol to a cobalt center by scission of the C - 0 bond. Mechanistic

15

studies have

10

shown

that

MoSz-based catalysts have the capability of breaking the

C - 0 bond of methanol to produce a methyl group, thus overcoming this crucial step. Based on this result, it is proposed that this methyl group, produced from methanol, can migrate toward the strongly bound

CO found on the cobaltcontaining MoS2 catalyst to

5

0 M e t h a n o l Propanol Ethanol Butanol

-

Methane Propane Ethane

-

Me For

Me Ac

PRODUCTS, alcohols-hydrocarbons-esters Fig. 7.18 The product distributions obtained with H2/CO = 0.96 synthesis gas at 8.3 MPa over the K,CO,/(Co/MoS) = 10/90 wt% catalyst (304OC, GHSV = 1940 l/kg catal/hr) and over the CSOOCH/MOS~ = 10/90 Wt% catalyst (29S°C, GHSV = 7790 l/kg catal/hr), where the conversions were 13.1 and 14.1 mol CO/kg catal/hr, respectively.

produce an acyl group that is subsequently

hydrogenated

to form ethanol. Analogous subsequent reactions produce the higher linear alcohols. Sakari et al. (ref. 208) performed ab initio Hartree- Fock MO calculations to show that methyl migration is an easy reaction path. Experimental evidence supporting the reaction between a methyl group and C o over the MoS2-based catalysts is discussed elsewhere in this chapter.

7.3.7 Mechanistic Implications of the Promotional Effect of Alkali The role of the alkali compound on the activity and selectivity of the MoS2-based catalysts is proposed to be two-fold. It is responsible for (i) the presence of basic centers capable of associatively activating CO and thus opening a new pathway to form alcohols, and (ii)

for the reduced availability of active hydrogen.

316

The superior promotional effect of cesium compared to that of potassium indicates that base-catalyzed reactions are being promoted in the system. The superior promotional effect of CS is by no means unique to MoS,, and it was noticed initially by Fischer and Tropsch (ref. 174) on the alkali-activated iron-base catalysts. Later systematic studies with catalysts such as Cr203/MnO (refs. 134,135), ZnO/Cr20j (ref. 131), and Cu/ZnO (ref. 98) concluded that the promotion of alcohol synthesis by alkali metals is ion specific as Cs > Rb > K > Na > Li, in the same order of their basicity. Activation of CO by alkali hydroxides is known to occur under mild conditions (ref. 209). Another indication that the basic centers created by cesium are participating actively in the system is the fact that the formation of methyl formate and the water-gas shift reaction are being promoted. These two reactions are known to be associated with the presence of basic centers. The maximum in alcohol yields as a function of cesium content observed in this study was also observed by Nunan and co-workers (ref. 5 5 ) with the Cs- promoted Cu/ZnO catalyst, although at much lower loadings of cesium, e.g. 0.45 wtlc Cs. It was reported (refs. 56,70) that the latter catalyst was bifunctional, in which the cesium component associatively activated CO and the Cu/ZnO matrix provided the hydrogenation function necessary for the synthesis of alcohols. This picture can be applied to the Cs/h40S2 catalyst. The basic component (Cs) contributes the active sites to activate CO in such a way that a new reaction pathway for the fomiation of alcohols rather than for the formation of hydrocarbons is open. The hydrogenation function necessary for the synthesis of alcohols is supplied by MoS2 It is well known that MoS, dissociatively activates H, and that its active sites are edge-like defects. The increase in the total product yield is due to the introduction of cesium to the catalyst, which increases the number of CO activation sites. The maximum activity is achieved when the CO and H2 activating components, Cs+ and MoS2, respectively, are balanced. Details of the mechanism of the formation of products over this type of catalyst will be discussed later. It should be mentioned that MoS, is also capable of activating CO, although this activation is through a reaction pathway that leads to the formation of hydrocarbons, as evidenced by the fact that undoped MoS2 produces mostly hydrocarbons.

7.3.8 Research Goals Progress in the research and development of the new alkali/MoS2 catalysts has been exciting during the last five years. Desirable goals for further development include: Higher dispersion of the alkali, Optimization of the Co component, Higher activity in terms of yield of alcohols over catalysts with and without the presence of co, Lower selectivity toward hydrocarbons, Determination of the location and chemical state of the active sites in the alkali/Co/MoS2 catalysts, and

317

Determination of the long-term stability and maintenance of the alcohol synthesis selectivity as a function of the HIS content of the synthesis gas.

7.4 MECHANISMS OF ALCOHOL SYNTHESIS 7.4.1 Mechanistic Background of Higher Alcohol Synthesis Over Oxide Cutulysts The first step in higher alcohol synthesis from H2/CO synthesis gas over metal oxide catalysts involves the formation of a carbon-carbon bond. The first hypothesis of a mechanism for this process was advanced by Fischer (ref. 210) who suggested that higher alcohols are fonned from methanol and carbon monoxide, as depicted by the following reactions. This reaction sequence

CO + 2H2 + CH30H CH,OH + CO -+ CH3COOH CH3COOH + H2 + CH3CHO + H,O CH3CHO + H2 -+ CH3CH20H would involve C 0 insertion to form the carbon-carbon bond, followed by sequential hydrogenation. A similar homologation of methanol by C0/H2 has been proposed by Natta et al. (ref. 131) and Vedage et al. (ref. 98). More recently, Mazanec proposed (ref. 21 1) that the homologation of methanol by C 0 proceeds via a symmetric intermediate, as shown by Eq. 7.10.

O-CH2+CO-O \ / M

CH2 / \ C=O

\ / M

HC = CH tf

0 0 \ / M

tf

0 0 \ / M

tf

I \ 0 0 \ /

(7.10)

M

In contract, Frolich and Cryder (ref. 132) and Morgan (ref. 212) concluded that the synthesis of higher alcohols occurs predominantly by condensation of lower alcohols. Using the Frolich and Cryder proposal that the controlling reaction in the synthesis of higher alcohols was the condensation of two methanol molecules to produce ethanol via dehydration and elimination of water (Eq. 7.1 l), Graves (ref. 133) was able to qualitatively predict the presence or absence of certain higher alcohols when simple rules for addition were involved. Because of the observations of Brown and Galloway (ref. 21 3) that appreciate yields of dimethylether were fomied during methanol synthesis over ZnO/Cr203 catalysts, a two-step dehydration mechanism involving dimethylether as an intermediate (Eqs. 12 and 13) was considered, but this was discounted because the “information was insufficient to warrant any definite conclusion” (ref. 132).

2CH30H + CH3CH20H + HZO 2CH30H +(CH3)ZO + H2O CH3CHzOH (CH3j 2 0 __f

(7.11) (7.12) (7.13)

318

Morgan (ref. 212) presented Eqs. showing how higher alcohols might be synthesized over alkali/ZnO/Cr203 high pressure catalysts through aldol condensations starting with fomialdehyde, e.g. Reaction 7.14 followed by Reaction 7.9. Since it was reported that by the aldol condensation mechanism straight chain alcohols should be formed, especially n-butanol, from two molecules of acetaldehyde (ref. 212), Graves’ experiments supported the dehydration mechanism because the “absence of n-butanol speaks against the aldol mechanism” (ref. 133). However, the aldol mechanism, Eq. 7.14, has been invoked by Fox et al. (ref. 214) for non-catalytic alcohol synthesis over alkali acetylides, where the formaldehyde molecules are derived from methanol. The latter mechanism is reminiscent of the first steps of homogeneous base- catalyzed formose chemistry (ref. 215).

2HCHO

H2

+ CH20HCHO +

CH3CHO+H20

(7.14)

To summarize, the mechanisms proposed for C-C bond formation to form c 2 - C ~alcohols can be classified as

i. ii. iii.

CO insertion into a methyl-metal bond or into the C - 0 bond of methoxide, CO homologation of methanol via a symmetric intermediate, and Coupling of aldehydic or alcoholic species. As will be described, these three mechanistic pathways were recently distinquished by

13C-NMR analysis of the products formed when small amounts of 13C-labelled reactants, e.g. were H , injected into the H2/C0 synthesis gas feed over a variety of l3CH30H or C H ~ ~ ~ C H Z O catalysts. Analogous to the various mechanisms proposed for the synthesis of ethanol and higher alcohols, a variety of mechanisms have been postulated for the synthesis of methyl formate formed as a by-product during alcohol synthesis. The possible mechanisms include i. ii.

dimerization of aldehydes (refs. 132,216), coupling of adsorbed formaldehyde with an adsorbed methoxide species (refs. 217-2191, perhaps via a hemiacetal intermediate (refs. 217,218) as shown in Eq. 7.15, and H H

\ / C

I

+

0

I

...

111.

H \ / C + CH30CHO / \ O H

CH30 CH3

I 0 I

+

I

(7.15)

I

insertion of CO into adsorbed methoxide, as reported by Tonner et al. (ref. 220) for the homogeneous carbonylation of methanol in the presence of base catalysts, represented by

Eqs. 7.16 and 7.17.

319

0 CH3O- Na+ + C O - - - +

II

CH30C-Na+

0

(7.16)

0 II

II

CH30C- Na+ + CH30H (H20)

7.4.2

+CH30CH + CH30- Na+ (NaOH)

(7.17)

Formation of C, Products Over CsICulZnO Catalysts

The mechanistic pathways by which ethanol, methyl formate, and dimethylether are formed over a 0.4 mol% Cs-promoted Cu/ZnO (30/70 niolr/c) catalyst under steady state alcohol synthesis conditions have been probed by utilizing '3C-labeled reactants, e.g. '3CH30H enriched by a factor of 22.3 over the natural abundance of 1.1 1 atom%, added to the H,/CO = 0.45 synthesis gas stream with GHSV = 3260 l(STP)/kg catal/hr at elevated temperatures and 7.6 MPa. The liquid products were collected and examined by I3C NMR, and the analytical details have been described elsewhere (refs. 56,69,148,149). The experiments were generally carried out at temperatures such as 225 and 24OOC where the methanol yields were far from equilibrium, e.g. CO conversion of 0.5 and 2.4 mol%, respectively, to ensure that only the forward synthesis reactions were being observed. However, with insight gained from the low temperature experiments, product labelling patterns obtained at higher temperature could easily be interpreted. Upon injection of methanol into the synthesis gas stream at 22S°C, the methyl formate synthesis rates were increased by more than 4-fold. However, no ethanol was present in the product mixture within the GC and NMR detection limits. Analysis of the 13C enrichnients of each carbon center of methyl formate showed that the methyl carbon NMR signal was significantly increased while the carbonyl carbon showed no enrichment. Indeed, the methyl carbon exhibited a 22.1-fold enrichment (Table 7.3 1) relative to the carbonyl carbon of methyl formate. Thus, the methyl group was derived from the injected labeled methanol while the gas phase carbon monoxide was the source for the carbonyl group, as demonstrated by Eq. 7.18. As indicated in Table 7.31, the enrichment of methanol in the collected product was less than that of the injected methanol. This was due to dilution by methanol synthesized over the catalyst from synthesis gas. Upon injection of methanol at 24OoC, GC analysis showed that the methyl formate synthesis rate was increased by a factor of >2 but that the ethanol yield was not affected by the added methanol. NMR analysis of the product mixture showed that the enrichment pattern for methyl formate was the same as that observed at 225OC. see Table 7.31. In contrast with the non-equivalency of the carbons in methyl formate, analysis showed that both carbons of ethanol were enriched, which is consistent with Eq. 7.19. The similarity of I3C enrichments in the CH3 and CH2 groups of ethanol rules out rearrangement of methyl formate as the possible pathway to ethanol formation. Likewise, the possibility of CO insertion into a C l species, as proposed earlier (refs. 98,131,210,21 l), is also ruled out. The mechanism of the carboti-carbon bond fonnation reaction indicated by this study must involve coupling of oxygenated C, species, which supports the conclusions of Morgan

320

Table 7.31

Enrichment Factors of Each Carbon of the Products Formed during Injection of Carbon-I3 Enriched Methanol into the H2/C0 = 0.45 Synthesis Gas Over a 0.4 mol% Cs/Cu/ZnO Catalyst Enrichment Factor for Each Carbon

moduct

225°C

240°C

CH30H

18.9

10.9

I

CH3 0

22.1

I

1.0

CHO

I

I

10.7 1.o

(ref. 212) and Fox et al. (ref. 214) for aldol condensation, as well as those of Frolich and Cryder (ref. 132) and Graves (ref. 133) for dehydrative coupling.

0 13CH30H + CO 2 13CH30H

4

I1

13CH30CH 13CH,13CH,0H

(7.18)

+ HzO

(7.19)

At both temperatures investigated, methyl formate was formed with incorporation of the 13C label as shown by Eq. 7.18. This eliminates the dinierization of aldehydic species (refs. 132,216) or other oxygenated species (refs. 217-219) as the synthesis pathway for methyl formate over the Cs/Cu/ZnO alcohol synthesis catalysts. In each of those pathways, including the hemiacetal route (refs. 217,218), it would be expected that the l3C label would be located in both the methyl and carbonyl groups of methyl formate, contrary to the experimental result. The NMR analysis of the product formed at 240OC showed that the carbons of dimethylether were labeled by 13C. This indicates that the likely synthesis pathway is dehydration of methanol on the residual acid sites of the catalyst, as represented by Eq. 7.12. However, there is no evidence that dimethylether is a precursor to ethanol. This study utilizing l3C labeled methanol as a probe has demonstrated that the C1 surface species that is readily formed on the Cs/Cu/ZnO catalyst from methanol is a precursor of the methyl group of methyl formate, of the CH, groups of dimethylether, and of both the CH3 and CH2 groups of ethanol. The C-C bond in ethanol is made by coupling of the oxygenated C1 surface intermediates originating from methanol and not by a CO insertion mechanism. This contrasts with the synthesis of methyl formate that occurs by CO carbonylation of methanol and not by condensation of aldehydic or methoxide species derived from methanol.

321

Methanol injection experiments have also been carried out with a non-promoted commercial Cu/ZnO/AI2O3 catalyst

(UCI

C79-2

consisting

of

the

approximate

composition

of

CuO/ZnO/AI2O3 = 43Y43313.0 wt%) at 285°C and 2 MPa with a HdCO = 1 synthesis gas (ref. 221) using mass spectroscopy (MS) to analysis the ethanol that was formed. Over the range of contact times of 0.49-9.7 sec, it was also observed that both carbons of ethanol were l3C-labelIed and that scrambling of the labels in ethanol did not occur under the reaction conditions. It was also concluded that the formation of ethanol occurred through an intermediate that was common to the synthesis of both methanol and ethanol. A subsequent study of the formation of ketones and other products from linear primary alcohols under nitrogen or CO at 285°C and 6.6 MPa was carried out over this same catalyst (ref. 222). From the distribution of the l3C isotopic labels, it was concluded that the syntheses of the 211-1 ketones, the 2n ketones, the 2n esters, and the 2n aldehydes took place via aldol-type condensation reactions of the n alcohols.

7.4.3 Formation of C, and C, Alcohols Over CslCulZnO Catalysts Isotopically labelled ethanol, CH313CH20H (24% 13C at CH2), was injected into the synthesis gas stream at the rate of 193 g/kg catal/hr to probe the mechanistic pathways leading to the higher alcohols over the 0.4 mol% Cs/Cu/ZnO catalyst (refs. 148,149). Fig. 7.19 A shows the NMR spectrum of the liquid product collected when natural abundance ethanol was injected in the H-JCO = 0.45 synthesis gas at 240°C. Methanol was the only principal product formed from the synthesis gas and detected over the catalyst at this low temperature. Increasing the reaction temperature to 300°C greatly increased the yield of the higher alcohols, as shown in Table 7.32.

Table 7.32

Yields of Products Over the 0.4 mol% Cs/Cu/ZnO Catalyst at 240°C and 300°C and 7.6 MPa with H,/CO = 0.45 Synthesis Gas with GHSV = 3260 I(STP)/kg catalhr Before (A) and After (B) the Injection of Ethanol at the Rate of 10 pl/min (193 g/kg catal/hr) into the Reactant Stream. Data are taken from Ref. 149. ~

~-

~~

~

Product Yields, g/kg catal/hr 300°C

240°C Product

A

B

B ~

Methanol Ethanol MethylAcetate 1-Propano1 2-Methyl- 1Propanol 1-Butanol 2-Butanol 2-Butanone 2-Methyl-lButanol

133 0.9 0.14 . .

__ . .

~~~

172 145 5.7

171 19 8

160 29 11

5.7

19

44

__

16.9 4.9 0.8

42.7 10.7 4.9 3.0

1.2

. .

. .

__

. .

. .

._

. .

._ ~

~~~

6.8

322 I: 0 0

A I

I

I

P

0,

0

0

N

m

I

I

*0

*",

l0 90

80

70

60

50 40 30 pprn shift f r o m TMS

20

10

0

*

*c \,C-C-OH

B

C I 0 0 I

0 I 0

c\*

I

C'

0

C-C-OH

C?

pprn shift f r o m T M S

Fig. 7.19 The I3C NMR spectra of the liquid product obtained upon injecting (A) natural abundance ethanol into the H2/C0 = 0.45 synthesis at 7.6 MPa and GHSV = 3260 l/kg catalhr over the 0.4mol% Cs/Cu/ZnO catalyst at 240T and (B) 24% enriched CH3I3CH20H at 30OOC.

Replacement of the natural abundance ethanol with 13C-labelled ethanol and increasing the reaction temperature to 300°C yielded the spectrum shown in Fig. 7.19 B when the reaction product was analyzed by NMR. It is clear that the injected ethanol was incorporated into the higher alcohols

323

Table 7.33

Enrichment Factors of Each Carbon, Relative to the Ethanol C-2 Carbon, of the Principal Products Formed during Injection of 24% Enriched CH3I3CH2OH into the H2KO = 0.45 Synthesis Gas at 7.6 MPa and GHSV = 3260 l/kg catal/hr Over a 0.4 mol% Cs/Cu/ZnO Catalyst.

I

I

I Product

I

Enrichment Factor for Each Carbon 240°C

CH30H

0.9

CH3 CH20H

1.o 10.9

CH3 CH2 CH2OH

5.5 6.4 2.3

I

I

I

300°C 0.7

I

1.o 2.4 3.1 1.4 0.9 5.0 1.8 0.6

I

and that the labelled carbon was preferentially located at particular positions in each molecule. The determined enrichment factors for each carbon of the principal products is given in Table 7.33 for the experiments involving the addition of the CH3I3CH2OH probe. The above observations and conclusions were reinforced (refs. 69,120,122,149) by injecting a 13CH20H/CH3CH20H mixture into the synthesis gas and analyzing the products formed at 260°C. The resultant l3C NMR spectrum is shown in Fig. 7.20. The 13C label was preferentially found at the C-l carbon of propanol and the C-1 carbon of 2-methyl-1-propanol. Thus, the C1 oxygenated intermediate that is added to the growing carbon chain preferentially retained its oxygen. These experiments demonstrate that over Cs/Cu/ZnO catalysts at high temperatures, Reaction 7.20 occurs selectively (ref. 149). This reaction is CH313CH20H + CO/H,

---+

13CH3CH2CH20H

(7.20)

consistent with aldol-type 0-addition with oxygen retention reversal, as shown in reaction sequence 7.21. -H2 CH3I3CH2OH

ts

H2

.H@

CH313CH0

ts

HQ

@CH2'3CH0(enolate or carbanion)

324

CH213CH0+ H2CO

+

@CH2CH213CHO) +

(7.21)

oOCH2CH213CH3 HQL

P- addition

HOCH2CH213CH3 The retention of the anionic oxygen in the I-OCH2CH213CHO]intermediate is specific to the Cs promoter that prevents the dehydration of the alcoholate oxygen and favors hydrogenation of the free 13CH0 group. Such a path constitutes a reversal of the normal aldol synthesis pattern in which CH3CH213CH20Hpropanol would be formed in the presence of hydrogen. In additional experiments over the Cu/ZnO and Cs/Cu/ZnO catalysts, injection of 1-propanol into the H2/C0 = 0.45 synthesis gas yielded dominantly 2-methyl- 1-propano1 (with 1-butanol as a minor product), and the Cs promoter enhances the rate of the P-branching (ref. 2231,

-

CHjCH2CH20H + [H,CO]

(CH3),CHCH20H (major) + CH3CH2CH2CH20H(minor).

(7.22)

The dominant P-addition to form 2-methyl-1-propanol occurs via a mechanistic path analogous to Eq. 7.21 as indicated by the 13C isotope experiments of Nunan et al. (ref. 149). This aldol path with oxygen retention reversal is further corroborated by the outcome of 2-propanol injection into the synthesis gas (ref. 223) that resulted in the dominance of 1-butanol in the C4 product.

8

I

?0

?

I

*u

?

*':=

!*O

0

I

?

*Y

*Y

9

1

1 o=v I

I

90

I

80

70

.*..

,

. . , . . . .L.. . , .

60

50

L.

40 ppm s h i f t from

I , . . . , I

30

.

*"

. . . . . . . . . . . . . . . . . . . . . . .

20

10

0

TMS

Fig. 7.20 The 13C NMR spectra of the liquid product obtained upon injecting a 13CH20H/CH3CH20H= 1/3.2 mixture (by wt) into the H2/CO = 0.45 synthesis at 7.6 MPa and GHSV = 3260 l/kg catal/hr over the 0.4 mol% Cs/Cu/ZnO catalyst at 260°C.

325

The patterns of steps C1 -+ C4 continue over the different catalysts as shown above with the exception that 2-methyl-1-propanol does not give rise to any C, products over the copper-based catalysts. This is a known feature of P-addition, in which the addition does not occur at branched carbons in aldol synthesis. The a-addition of the type represented by Eq. 7.14 between a branched C, and a C1 aldehydic intermediate also appears to be forbidden, perhaps for steric as well as for electronic reasons. The high rate of P-addition of C, to C3 and the termination of the synthesis at the branched C4 alcohol are the major factors determining the high selectivity for 2-methyl- 1-propanol.

7.4.4 Formation of Oxygenates and Hydrocarbons over AlkalilMoS2

Catalysts In contrast to the products formed over the Cs/Cu/ZnO catalyst, which tend to be branched products, the alcohols and hydrocarbons formed over the alkali/MoS, catalyst are linear. Since the mechanisms of the alcohol synthesis processes could be distinctly different over these two types of catalysts, studies were carried out in which I3C-labelled methanol was injected at the rate of 249 g methanol/kg catalkr to the reactant feed over the alkali/MoS2 catalysts under reaction conditions (refs. 167,168,201). The liquid products were collected and analyzed by NMR while the methane that was produced was analyzed by MS.

I

r

?

?

P

THF

0

THF

I

? I

Y . o

..1._ Fig. 7.21 The I3C NMR spectrum of the liquid products formed over the CsOOCH/MoS2 = 20/80 wt% catalyst from Hd C O = 0.96 synthesis gas at 295°C and 8.3 MPa with GHSV = 7465 l/kg catal/hr.

326

Table 7.34 Yields of Products Over the 20 wt% CsOOCH/MoS2Catalyst at 245°C and 295°C and 8.3 MPa with HdCO = 0.96 Synthesis Gas with GHSV = 7465 l(STP)/kg c a t a h before (A) and after (B) the Injection of l3C-enriched Methanol (24.08 atom% 13C = 21.69% enrichment factor) at the rate of 10 pl/min (249 g/kg catal/hr) into the Reactant Stream. Data are taken from Refs. 167 and 168. Product Yields, g/kg catal/hr 295°C

245°C Product Methanol Ethanol 1-Propanol Methyl Formate Methyl Acetate Methane Table 7.35

A

B

75.1 15.7 3.8 0.7

280.7 22.1 3.8 4.1

201.7 81.4 24.4 trace

280.0 91.3 27.8 0.8

4.3

3.6

5.9

7.3

43.5

65.9

trace 0.3

B

A

Enrichment Factors of Each Carbon, Relative to the TetrahydrofuranInternal Standard, of the Principal Products Formed during Injection of 24% Enriched 13CH OH into the H&O = 0.98 Synthesis Gas at 8.3 MPa and GHSV = 7465 l/kg cataldr Over a CsOOCH/MoS2 = 20/80 wt% Catalyst.

I

+

Enrichment Factor for Each Carbon 285°C

CH30H CH3 CH,OH

CH3 0

17.7

I I

11.6 1.3

12.6

14.2

I I

12.0 1.4

15.9

295°C

T T p

1

:l

:::

1.5

0.7

9.0

5.9

18.9

25.7

9.8 5.2

I

1

9.7

i.0

327

Table 7.36

Enrichment Factors of Each Carbon, Relative to the Tetrahydrofuran Internal Standard, of the Principal Products Formed during Injection of 24% Enriched 13CH30H into the H2/C0 = 0.98 Synthesis Gas at 8.3MPa and GHSV=3535l/kg catal/hr Over a K2CO,/(Co/MoS,) = 10/90 wt% Catalyst.

I

I

Enrichment Factor for Each Carbon

I

300°C

260°C

280°C

CH30H

27.5

26.9

,

22.2

CH3 CH,OH

22.8

24.0 1.2

I

21.9 1.3

Product

1.o

I

325°C

‘

,

12.5 15.9 1.8

CHzOH

Upon addition of 13CH30H to the H2/CO=0.96 synthesis gas over a 20 wt% CsOOCH/MoS2 catalyst, the rates of formation of methane, ethanol, methyl formate, and methyl acetate increased, as shown in Table 7.34. The I3C NMR spectrum of the liquid collected during the 295°C experiment is shown in Fig. 7.21, where tetrahydrofuran (THF) was added as an internal standard. Table 7.35 shows the enrichment of the carbons in the oxygenated products (NMR), as well as the enrichment observed in the methane (MS). These data show that i. ii.

preferential enrichment of the terminal carbons of the linear alcohols occurred and the methyl group of the methyl esters was preferentially enriched relative to the carbonyl carbon. Isotopically labelled methanol was also injected (rate = 239 g/kg catalhr) over the

K~CO~/(CO/MOS~) = 10/90 wt% catalyst, and the same type of 13C distribution in the products was obtained (refs. 167,168,201). However, it was surprisingly found that two isotopic species, i.e. 13CH3CH2CH20Hand CH3’3CH2CH20H, of I-propanol were formed, as shown in Table 7.36. This is distinctly different than the behavior observed for the Cs/MoS2 catalyst, see Table 7.35. The lower enrichment values of the C-2 and C-3 carbons of propanol relative to ethanol and the approximate equivalence of the two carbons suggest that adsorbed symmetric C2 species give rise to propanol. Species of the type “p-bonded ethylene” or “0-di-bonded ethylene” have been suggested in homogeneously catalyzed reactions (ref, 224), and they might exist on the surface of the K/Co/MoS2 catalyst since C2 and C3 olefins were detected in appreciable amounts among the products. It is interesting to note that Bums (ref. 225) studied the reaction of 13CH,0H with H$CO catalyzed by [Co(CO)& and observed the formation of the two isotopically labelled products

328

I4CH3CH2CH,OH and CH314CH,CH20H. Thus, the intermediate species involved in the Co-containing heterogeneous and homogeneous systems might be identical. In any case, these experimental results demonstrate that upon injection of 13CH30H into synthesis gas, the formation of ethanol and 1-propanol over the Cs/MoS2 and alkali/Co/MoS, catalysts can be represented by Eqs. 7.23-7.25.

+ +

l3CH30H+ 12C0/H2 13CH312CH20H+ "CO/H2

3CH312CH,0H 13CH'2CH,CHzOH

(7.23) (7.24)

cs/Mos, '3CH312CH,0H

13CH312CH2CH20H+ '2CH.3'3CH2CH,0H CS/(CO)MOS,

+ l2CO/H2

(7.25)

Formation of these labelled products indicate that the linear products are formed over these catalysts by a classical CO insertion process. In addition, this pathway is enhanced by the presence of cobalt in these catalysts. Further support for the CO insertion mechanism over these catalysts is provided by the observation that methane produced as a side-product was also labelled by I3C. This side-reaction, where methane is formed as a secondary product, can be represented by Eq. 7.26.

I3CH3OH

l3CH3 OH \ / M M

+

H2

--+

H '3CH4+H20+ I M

(7.26)

The higher alcohols would be formed by the continuing CO insertion reaction, e.g. the C2 -+ C3 step occurs by the same process as for the C, representing in Eqs. 7.24 and 7.25.

+ C,

step, as evidenced by the isotope reactions

7.4.5 Mechanistic Implications These studies demonstrate that the mechanism of alcohol synthesis is catalyst specific, is bifunctional, and can be significantly influenced by dopants. With the copper-based catalysts, it has been shown that: Carbon bond formation proceeds in a step-wise manner, principally involving C1 oxygenated species, Lower alcohols are incorporated into the higher alcohols, The presence of alkali enhances the C-C bond forming reactions, especially the C, --f C j step that results in a low yield of ethanol but higher yields of C3 alcohols, The presence of alkali shifts the mechanism of C-C bond formation so that the P-addition process becomes dominant and results in branched C4 and C, alcohols becoming favored products,

329

The alkali (cesium) pins the reactive dioxygenated intermediate to the surface such that the oxygen retained in the resultant alcohol is the reverse of that found in normal aldol-type synthesis products, and Branching terminates the C-C bond forming reactions so that tertiary alcohols are not formed. With the MoS2-based catalysts, it has been shown that The presence of alkali is essential to obtain selectivity and reactivity toward the synthesis of alcohols, The presence of alkali suppresses the yield of hydrocarbons formed over this hydrogenation catalyst, The alcohol synthesis activity is directly related to the basicity of the alkali dopant: thus cesium is the most active dopant, Alcohols and hydrocarbons formed over the Cs/MoS2 and K/(Co)MoS, catalysts have a common intermediate, At least a part of the hydrocarbons formed over these catalysts are formed as secondary products from the alcohols, Alcohol synthesis chain growth occurs via a classical CO insertion mechanism, and The presence of Co greatly enhances the C, dominant product.

-+ C,

step so that ethanol becomes a

7.5 KINETIC MODELS FOR THE SYNTHESIS OF ALCOHOLS 7.5.1 Introduction Kinetic modelling of methanol synthesis over both high pressure and low pressure catalysts has long been carried out so that concentration and temperature distributions in industrial reactors could be predicted (refs. 10,226). Most of the inodels (refs. 10,31,59,227-230) were based on empirical data that fit various rate Eqs. that sometimes included a term for the CO2 content of the synthesis gas (refs. 31,59,227-229). As pointed out previously, in 1983-1985 it was shown that the low pressure/temperature copper-based methanol synthesis catalysts could be promoted with heavy

alkali to produce the branched higher alcohols from H2/C0 synthesis gas mixtures (refs. 98,141-14s). At the same time, it was disclosed that alkali/MoS2 catalysts produce h e a r higher alcohols from C02-free synthesis gas (refs. 159-164). Thus, it became desirable to develop kinetic models to aid in the research, development, and scale-up of these new processes that could be used to produce mixtures of higher alcohols.

For the synthesis of branched alcohols over the K-promoted Cu/ZnO/AI2O3 catalysts, Smith and Anderson (refs. 141-143) proposed a model for carbon chain growth via a- and B-carbon addition of C, and C2 species to the growing alcohol. They calculated growth parameters that demonstrated that a addition was slow while p carbon addition was fast. It was found that the

330

Pronlotional effect was Cs > Rb > K for the higher alcohol synthesis (refs. 98,144), as well as for methanol synthesis (ref. 57). With the Cs/Cu/ZnO catalyst under the reaction conditions that were employed, Vedage et al. (ref. 98) concluded that linear chain growth was significantly slower that P-addition, in agreement with Smith and Anderson (ref. 143), while a-addition to produce secondary alcohols was insignificant. Recently, Tronconi et al. (ref. 231) developed a lumped kinetic model for the higher alcohol synthesis over a high temperature non-copper-containing Zn-Cr-K oxide catalyst. The model described the effects of process conditions such as H2/C0 feed gas ratio and space velocity on reactant conversion and total alcohol yield. However, the distribution of the individual alcohols could not be predicted by the modelling approach, and the model was developed at a fixed temperature and pressure of 400OC and 8.8 W d , respectively. These conditions are significantly more severe than those used for the alkali-promoted Cu-containing and MoS *-based catalysts.

7.5.2 Development of Kinetic Models for Higher Alcohol Synthesis As discussed earlier in this chapter, l3C-NMR studies of the mechanism of the carbon chain

growth

processes

to

produce

alcohols

and

esters

over

the

Cs/Cu/ZnO

catalysts

(refs. 56,69,122,148- 150)) and alcohols, esters, and hydrocarbons over alkali/MoS* and alkali/Co/MoS2 catalysts (refs. 167,168,201) demonstrated that higher alcohols and hydrocarbons were formed over the latter catalysts via a classical CO insertion process, viz. Reaction 7.27, while a novel aldol coupling with oxygen retention reversal process, viz. Reaction 7.20, predominantly produced the higher alcohols over the Cs/Cu/ZnO catalysts. A discussion of the coupling mechanism in terms of possible intermediates that result in the formation of alcohols and the methyl esters has been given (ref. 70).

ll

I1 C-COOC

C-C-OH

p+P,

"'

C-C-C-OH

P2

P3

I1

> C-C-C-C-OH

, .........>

C-OH I

.._.......

j

11

HCOOC

CH30H c C - O H

?

C-C-C-OH

P ' 7 -

I1 C-C-C-C-C-OH

I-e C-C-C-C-C-C-OH

C-C-C-C-OH

C-C-COOC

C-C-C-COOC

7

% C-C-C-C-C-OH

1-e

2 C-C-OH p+P 1 (lgI C-C-C-OH

- I-e a,

I

C-C-C-C-OH

I-e

C-C-C-C-COOC

2 C-C-C-C-C-OH 1 1 C-C-C-C-C-C-OH

Fig. 7.22 The kinetic reaction network for the synthesis of oxygenates over the Cs/Cu/ZnO catalyst. The reactions are linear growth (I), 1-carbon P-addition (P1,P1'), 2- and 3-carbon P-addition (p2$3), and methyl ester formation (ao,ai).

331

CH3*CH2OH + CO/H2 CH3*CH2OH + CO/H2

+ CH3*CH2CH20H ----+ *CH3CH,CH,OH

(7.27) (7.20)

Based on the mechanistic studies, kinetic models (shown schematically in Figs. 22 and 23) were developed (refs. 69,120,170-172) to describe the alcohol product (and hydrocarbon and ester side-products) distributions over the two different types of catalysts as functions of reactor operating conditions. The kinetic network in Fig. 7.22 where C-OH, HCOOC, etc. represent adsorbed surface species and where the terminating steps to form gas phase alcohols and esters are not shown, for the Cs/Cu/ZnO catalysts accounts for the experimentally observed minimum yield of ethanol and a high yield of branched 2-methyl-I -propano1 that is distinctly different from the Anderson-Schulz-Flory (A-S-F) distribution. For the alkali/MoS2 catalysts, a different kinetic model in which P-carbon addition steps are negligibly slow, shown in Fig. 7.23, accounts for the A-S-F distribution, with C, as the monomer, of alcohol and hydrocarbon products under some conditions and for a maximum in the yield of ethanol under other experimental conditions, especially with the alkali/Co/MoS2 catalyst. The kinetic model for the Cs/Cu/ZnO catalysts assumed that the reactions are first order with respect to the growing intermediate, are not reversible except for that leading to methanol, and have rate constants that are independent of chain length >3. The distribution Eqs. are derived from a steady state mass balance for each surface intermediate. The kinetic parameters were designed as follows: (i) methyl formate formation (a,) occurs via methanol carbonylation, which is at equilibrium under the reaction conditions employed (ref. 5 6 ) , whereas the other methyl esters (ad) are not at equilibrium, (ii) two P-addition rate constants are defined--one for the C2 -+ C3 step

CH30H

C2H50H

AEl

*E2

HCOOCH3

CH3COOCH3

C3H70H

‘ZHb

*E 3

C3H8

C2H5COOCH3

Fig. 7.23 The kinetic reaction network for the synthesis of alcohols, esters, and hydrocarbons over the alkali/MoS? catalysts. Indicated are rate constants, expressed relative to the rate constant k(t) for the hydrogenation and desorption of oxygenated intermediates, for the following processes: formation (K(A1))of the C, oxygenated intemiediate A , , from CO/H2, dehydration (k(H)) of the oxygenated surface intermediate to the hydrocarbon surface intermediate A,,, linear carbon chain , and growth (k(1)) by CO insertion, formation (k(e)) of the ester intermediate A E ~hydrogenation desorption (k(tH)) of the hydrocarbon intermediates, and readsorption ( k ( t ) ) of the gas phase alcohols.

332

(PI) and one for the other higher carbon

compounds (PI'), (iii) two (P2) and three (&) carbon P-additions are allowed to occur, and (iv) linear growth (1) can occur, but this latter process is

negligible with branched intermediates. These parameters are related to the particular rate constants

(kL, k,, and k ~relative ) to the rate k, of the termination reactions, e.g. (7.28) (7.29) (7.30) etc., where Acl is the surface concentration of the C , intermediate and k,, k,, and kx are the true kinetic constants for linear growth, P-addition, and ester formation, respectively. The steady state first order differential kinetics for the adsorbed surface intermediates Acj, e.g. A c ~= C-C-OH in Fig. 7.22, lead to Eqs. relating the formation and subsequent reactions of the intermediate, e.g. for the surface concentration of the C2 alcohol species (7.31)

Eqs. of this type can be derived for the surface concentration Acj of each species related to the C1-adsorbed species AC1. To calculate the gas phase concentrations from the expressions derived for the surface intermediates, it is noted that rcj = ktACj, where rcj is the molar rate of production of gas phase component Cj and k, is the termination rate constant for hydrogenation of Ac, to the gas phase product. It is assumed that the rates of the p2 and P3 processes, kB2 and k,, respectively, are the same so that b3 = P2AC3/Ac2. For differential reactor conditions, the surface concentration of the methanol precursor, A c l , is nearly constant through the reactor. Indeed, under the higher alcohol synthesis conditions employed in the experimental work, the CO conversions were generally between 25 and 35 moI% and the methanol yield was approximately equal to the equilibrium methanol yield. This leads 10 the concentrations of the surface intermediates being proportional to the gas phase concentrations C; of oxygenated products with j carbons, i.e. CJIC1= ACj/Acl. To account for integral reactor conditions in the isothermal fixed-bed plug-flow reactor, the differential equation (7.32) which defines the rate of reaction at any point in the reactor, must be solved. In this equation, ncJis the molar flow rate of component j , w = the catalyst wt, and rc, is defined above. This Eq. 7.32 allows the dependences of the yields and selectivities to be calculated as functions of parameters such as reactant flow rate, reactor pressure, and H2/C0 redctant ratio. As noted above, the methanol synthesis reaction is at or near equilibrium. Therefore, Eq. 7.32 can be rewritten for the special case of j

=

1 to include the reversibility of the methanol synthesis reaction as

333

(7.331,

where K , is the equilibrium constant of methanol from H2/C0. The presence of C02 in the higher alcohol product is a result of the water gas shift (WGS) reaction that is at equilibrium under the reaction conditions utilized. The partial pressure of CO2 was therefore calculated directly from the equilibrium constant (KWGs) of the WGS reaction. A semi-empirical expression for the surface concentration of the adsorbed methanol precursor (C-OH) was derived by considering the following sequence of reaction steps:

co + s w CO'S

(7.34)

H2 + s tf H ~ ' s

COS+ 2H2.s tf A,,

+ 2s.

(7.35) (7.36)

where s represents a surface site of the catalyst. The resu!ting expression is (7.37) where KCO and K H 2 are the equilibrium constants of the reactions represented by Eqs. 7.34 and 7.35, respectively, K is the equilibrium constant for Reaction 7.36, and K c 0 2 is the equilibrium constant for C02 adsorption. Values of Kco, K H 2 , and KC02 are taken from Klier et al. (ref. 59). The kinetic parameters I , PI, pl', p2, &, and a i are not true rate constants. By definition, they are ratios of rate constants multiplied by the surface concentration of the intemiediate being added to the growing surface species, see Eqs. 7.28-7.30. In solving the integral Eqs. and estimating the kinetic parameter values, however, the parameters were assumed to be constant and not dependent on the surface concentrations AC1,AC2, and AC3. This is a reasonable approximation since A c ~does not change dramatically through the reactor for a particular set of operating conditions while both p2 and p3 are small relative to the other terms of the growth scheme equations. Additional details

will be given elsewhere for the differential reactor model (ref. 171) and the integral reactor model (ref. 172). For the alkali-promoted MoS2 catalysts, the P-addition processes that form branched products do not occur, but hydrocarbons are observed among the products. Therefore, both adsorbed alcohol-forming surface intermediates (AAj) and hydrocarbon-forming surface intermediates (AH,) are considurd (see Fig. 7.23). In this model, the desorption of the alcohols is presumed to be reversible. However, the synthesis of the hydrocarbons and esters is considered to be irreversible. Considering the steady state surface concentration of the ethanol-fomiing intemiediate (AA2) that is formed from the preceding hydrocarbon-precursor species (A H I ) and gas phase ethanol (concentration Y c 2 0 H ) ,the following Eqs. are obtained:

334

rlNPUT.

Measured DaIa and

Initial Parameter Estimates

I Calculate the Model R e ~ ~ o n s e s

I

SUBROUTINE: GENRAT Z

I

h

e Model Distribuliao

1

I ,

M a r w a r d l ' s Algorithm

-

I I 1

Predicted Responses.

Are

They in Suitable Agreement?

I

Yes Output R B S U I I 01 ~ the

SUBROUTINE: UWHAUS

Regression Analyrm

I

I ' I I I I I

I use the R U ~ Q ~ - K U I I ~ Aloortthm to Integrate

R*actor C,/C, = G,(k)

dnl/dW = kIA,GI(hl

I 1

>I

Fig. 7.24 Schematic of the computational procedure developed for the kinetic modelling of the catalytic synthesis of alcohols and by-products over alkali-promoted Cu/ZnO- and MoS2-based catalysts.

The analogous Eqs. for the intermediate AH^) that leads to ethane give (7.40) (7.41) Eqs. such as these were obtained for all surface intermediates. For this model, the kinetic parameters are calculated relative to k(t), which is set equal to 1. Application of the derived Eqs. required an estimate of the kinetic parameters, experimental reactor data, and a non-linear regression algorithm based on Marquardt's method (ref. 232). The experimental data for the Cs/Cu/ZnO catalysts have been reported in detail elsewhere (refs. 56,69,149,171). A schematic of the overall procedure developed by Smith et al. (refs. 170-172) that is used in optimizing the parameter values is shown in Fig. 7.24. Using a given set of kinetic parameters, the yields and selectivities of the products can be predicted for a given set of reaction conditions. In this case, no refinement of the kinetic parameters is camed out, and the UWHAUS Subroutine portion of the computer program, shown schematically in Fig. 7.24, is not utilized. Therefore, the Kinetic Predictor Program is simplified. As before, the output also included the product yields (in mol/kg catal/hr) at n slices (typically n = 10) through the catalyst bed from top to bottom, which provides for extrapolation of the yields for a large range of GHSV values and contact times.

335

7.5.3 Kinetic Modelling of Alcohol Synthesis Over CslCulZnO Cataljsts The kinetic modelling has been carried out for the optimized 0.4 mol% CsOOCWWZnO (30/70) catalyst contained in a tubular, fixed bed, isothemial reactor under higher alcohol synthesis conditions, e.g. at 270-325°C and 7.6 MPa with C02-free H2/C0 = 0.45 synthesis gas with GHSV = 3265 I(STP)/kg catal/hr. An example of the products formed at 310°C and the fitted selectivities using the differential and integral model is shown in Fig. 7.25. Both linear and branched alcohols were observed, and the esters are methyl (Me) esters. The kinetic model based on growth via P-addition and linear growth 1 fit the experimental data very well, and upon decreasing the reaction temperature, the kinetic parameters decreased in value except for the higher ester formation kinetic parameter a;. It was found that the C2 + C3 growth step (PI) is the fastest reaction process in the kinetic network, while the first step, C,

+ C,,

in the network (1) is the

bottleneck in the higher alcohol synthesis. This model has been successfully tested for many process conditions and for several related catalysts, including Cs/Cu/ZnO/Crz03 catalysts. It became evident in the modelling that the

P2 process, coupling of

an oxygenated C2 species to a

growing chain, is a slow process, while the formation of the higher methyl esters becomes an equilibrium reaction at ~ 2 9 0 ° C(refs. 170,171). The kinetic model was used to predict the reaction conditions that would yield a c,-c6 oxygenate fuel over this catalyst having a composition of 70 wt% methanol and 30 wt% of c,-c6 alcohols (plus a small quantity of methyl esters would also be formed). With this ratio, the model showed that contact time and temperature were inversely related. For example, this product ratio could be maintained with a synthesis gas of H2/C0 = 0.70 at 9.1 MPa and =31OoC using a contact time of 1.0 sec, while a lower temperature of 300°C would require a longer contact time of 2.0 sec. An example of the predictions made with the model is shown in Fig. 7.26, where the predicted

Measured Differential Model

Fig. 7.25 Selectivity of the products formed over the 0.4 mol% CsICulZnO catalyst with H2/CO = 0.45 synthesis gas at 310"C, 7.6 MPa, and GHSV = 3265 l/kg catal/hr. In this fig, Me For = methyl formate, Me Ac = methyl acetate, Me Pr = methyl propanoate, Me Bu = methyl butanoate, Me Pe = methyl pentanoate, and Me IsoBu = methyl isobutanoate.

336

1200

OXYGENATE YIELD, g/kg catal/hr 3OO0C, \

1000

293OC 31OoC

~

t

800

I

600.

'

200 0.0

1 1

I

I

0.5

1.0

1.5

I

2.0

2.5

CONTACT TIME, sec Fig. 7.26 Model predictions of the total C,-C6 oxygenate yields as functions of contact time and temperature in producing a high octane alcohol fuel over the 0.4 mol% Cs/Cu/ZnO catalyst at 9.1 MPa from H-JCO = 0.70 synthesis gas. Lower temperatures and lower contact times increase the total oxygenate yield, but this is a result of producing more methanol and shifting the selectivity away from the higher alcohols, thus decreasing the C2+ oxygenate/ methanol ratio. Derived from Ref. 69.

yields of oxygenates as a function of contact time at three temperatures were detemiined. This figure shows that at 300OC with a contact time of 2.0 sec, the oxygenate yield should be 371 g/kg catalhr. Under these predicted conditions to obtain the 30/70 wt ratio of C2+ oxygenates/methanol, Table 7.6 shows that 375 g of oxygenates/kg catal/hr was observed during the f i s t 400 hr of testing (see also Fig. 7.10). Using information obtained from this kinetic model, seven Cs/Cu/ZnO catalysts were tested under these reaction conditions for more than 935 hr each (ref. 69) to demonstrate that the catalysts were intrinically stable.

7.5.4 Kinetic Modelling of Alcohol Synthesis Over AlkalilMoS2-Based Catalysts The kinetic model for the synthesis of alcohols, esters, and hydrocarbons over the alkaWMoS2-based catalysts that contain 2.5 to 30.0 wt% alkali is based on CO addition to a C, hydrocarbon precursor (see Fig. 7.23) that results in the synthesis of linear alcohols and hydrocarbons. The mechanistic foundation of this type of chain growth via CO insertion into an alkyl-metal bond was reported elsewhere (ref. 167) and discussed earlier in this chapter. Hydrogenation of the Cn+l oxygenated intermediate yields the Cn+l alcohol or hydrocarbon, and carbonylation of the intermediate produces the methyl ester.

337

A

comparison

of

SELECTIVITY (exclusive of C02), C a t o m % 1

the

experimental and calculated product selectivities over the CSOOCH/MOS~ = 20/80 wt%

Observed

@f# Calculated

catalyst at 295°C is shown in Fig. 7.27, where the activity of this catalyst corresponded to a CO conversion level, exclusive of

C02, of 15.8 mol CO/kg catalh (10.1 mol% CO conversion). It is demonstrated that the model fit the experimental data very well. The

CO conversion and product yields were predicted as a function of the H2/C0 ratio in the range of 0.67 to 1.50 (ref. 233). Maintaining all other reaction parameter constant, as well as the kinetic parameters (K(A1) = 0.038, k(t') = 7.56, k(e) = 0.026, k(H) = 0.682, k(tH) = 1.800, and k(1) = 1.083),

"

M e t h a n o l Propanol Ethanol Butanol

- Methane

Propa ne Ethane

-

Me

For

Me Ac

PRODUCTS, a l c o h o l s - h y d r o c a r b o n s - e s t e r s Fig. 7.27 The selectivity of products formed over the CsOOCH/MoS, = 20/80 wt% catalyst at 295°C and 8.3 MPa with H2/C0 =0.96 synthesis gas with GHSV = 7470 I/kg catdl/hr. SELECTIVITY (exclusive of c@), C a t o m %

resulted in the predicted yields of the alcohols and hydrocarbons to increase by =85% and the % CO conversion to increase to

1 0 8 3 0 682 7 5 6 ~~

-

17.9 mol%, exclusive of COP However, it is expected

~

that the rate of hydrogenation processes and carbon chain growth might depend on the H2/C0 partial pressure ratio as reported by Smith and Anderson (ref. 143). Therefore, it was assumed that k(1) and k(H) were directly proportional to the C o r n 2 and H2/CO feed ratios, respectively. In addition, the k(t') parameter (the reversible adsorption of the methanol product - see Fig. 7.23) was increased because of the

__

M e t h a n o l Propanol Ethanol Butanol

- Methane

~

Propane -

Ethane

__

Me

Me

For

Ac

PRODUCTS, alcohols-hydrocarbons-esters Fig 7 28 Predictions of the product yields over the CsOOCH/MoS2 = 20/80 wt% catalyst at 295°C and 8.3 MPa with H2/CO = 150 synthesis gas wlth GHSV = 7470 l/kg catdl/hr as pertinent kinetic parameter\ were altered to account for the change in the H2/CO partid pressures in the reactant feed

338

SELECTIVITY RATIOS, carbon atom%

8

:! g

,I

-

-

--;,-::I

~ -~ .

-

lr

__

-

2

- .

-1 __

__

.___

0

0

2

4

6

a

10

GHSV, Vkg catallhr (Thousands)

Fig 7.29 The product selectivity dependence observed over a K2COd(Co/MoS2) = 10/90 wt% catalyst at 302°C and 8 3 MPa with H2/C0 = 0.96 as a function of GHSV.

higher yield of methanol. Altering the model

kinetic

parameters in this way resulted in a significant increase in the yield and selectivity for methane and small increases in ethanol and methyl formate production, shown Fig. 7.28.

as in The

net result of altering the three kinetic parameters in the expected directions was an increase in the %CO conversion, exclusive of C02, to 19.3 mol%, hardly any change in the yield of Cl-C, alcohols, and a significant increase in the methane yield. Modelling of the K-doped, Co-promoted MoS2-based catalysts was also camed out. It was previously noted that the presence of Co i n these catalysts tended to increase the rate of C-C bond formation. It was also noted that higher contact times favored the formation of higher yields of hydrocarbons (see Fig. 7.17). Therefore, the selectivity dependence on GHSV of the products formed over the Co-promoted catalyst was studied using the kinetic model. Fig. 7.29 shows the influence on the higher alcohol/methanol and alcohol/hydrocarbon ratios as function of GHSV. It was observed that K(A1) and k(ej hardly varied as GHSV was decreased, but k(t'j and k(l) decreased in magnitude while k(H) and k(tH) increased. As seen in Fig. 7.29, as GHSV was decreased the Cz+OH/methanol ratio increased significantly (due to increased selectivity to ethanol and a decreasing selectivity to methanol). However, the alcohol/hydrocarbon ratio decreased drastically at the same time because of a significantly increasing selectivity to methane. In this example, ethanol is the dominant product, but the hydrocarbons make up some 27-30 wt% of the product. Increasing the GHSV stepwise increased the selectivity towards the alcohols, in particular methanol, and significantly decreased the yield of methane. The kinetic model was also used to examine the data obtained with a supported WMoSz catalyst, as reported by Dow Chemical Co. (ref. 160). An example (ref. 233) is shown in Fig. 7.30 for a K/Mo/Carbon = 1..5/20.0/78.5% catalyst tested with a H2/C0 = 1.21 synthesis gas. DOW reported the higher HC as C,+ HC and the other products as oxygenates having an average carbon number of 4.

339

The

kinetic

model,

with

SELECTIVITY (exclusive of COz), C atom%

K(A1) = 0.0073, k(t') = 23.80, k(e) = 0.037, k(H) = 1.344, k(tH) = 1.138, and k ( l ) = 1.093, predicts the distribution of the higher HC and shows most of the other products as methyl esters. Modelling the synthesis steps over a pressure range of 10.5 to 17.4 MPa at constant temperature and GHSV demonstrated that the kinetic parameters did not significantly vary. Therefore, as the product yields increased with pressure over the K/MoSz/Carbon catalyst, the selectivities were hardly affected. Melhanol Propanol Pentanol-Methane Propane Ethanol Butanot Ethane Butane

Considerations mechanistic

Olher

PRODUCTS, alcohols-hydrocarbons-other

7.5.5 Kinetic The

-

investigations

demonstrated that the pathways to higher alcohols are catalyst specific and that the

Fig. 7.30 Product selectivities over the K/MoS2/C catalyst, exclusive of COT, tested at Dow Chemical Co. at 30O0C, 10.5 MPa, and catal/hr The CO GHSV = 3095 I/kg conversion level was 7.3 mol CO/kg catal/hr

presence of alkali is essential in establishing the selectivity to higher alcohols. Linear alcohols are formed over the alkali/(Co)MoS, catalysts by CO insertion, while branched alcohols are formed over the alkali/Cu/ZnO catalyst by P-addition process involves coupling of oxygenated intermediates. Kinetic studies show that the synthesis of higher alcohols over the latter catalyst has a kinetically limiting C, + C, step, which accounts for the minimum in ethanol selectivity that is observed. However, over the alkali/MoS, catalysts, the C,

-+

C, process is a rapid process that

allows the ethanol yield to be maximized so that ethanol becomes the dominant product. The kinetic models that have been developed for these processes provide for the reaction engineering and the prediction of the influence of reaction conditions, e.g. temperature, pressure, H2/CO ratio, and GHSV (contact time), on product yields and selectivities in differential and integral reactions. It is evident that GHSV is a dominate factor in determining the yields, alcohol/hydrocarbon selectivities, and product distributions of higher alcohols formed over these catalysts. Acknowledgment The Lehigh University (LU) research discussed here has been principally carried out by G. A. Vedage, J. G. Nunan, C.-W. Young, C. E. Bogdan, J. Santiesteban, and K. J. Smith with the assistance of Donna Mitko and Roy Bastian, supervision of Kamil Klier and Gary W. Simmons, and partial sponsorship by the continuous support of the U. S . Department of Energy. Interpretation of the LU research presented here has occurred through interactive discussions among all of these people. The kinetic modelling research has been led by K. J. Smith.

340

7.6 REFERENCES 1 2

Patart, M., French Patent 540, 343 (Aug. 1921). Badische Anilin und Soda Fabrik, D. R. Patents 41.5, 686; 441, 433; and 462, 837 (1923)

3

and U.S. Patents 1,558, 559 and 1,569,775 (1923). Lormand, C., Ind. Eng. Chem., 17,430 (1925).

4 5 6 7

Molstad, M. C. and Dodge, B. F., Ind. Eng. Chem., 27, 134 (1935). Hiittig, G. F., Strial, K. S., and Kittel, H., Z. Electrochem., 39, 368 (1933). Patart, G., Ind. Eng. Chem., 17,430 (1925). Schmidt, 0. and Ufers, K., D. R. Patents 571, 355,571, 356, and 580,905 (1928).

8

Frolich, P. K., Fenske, M. R., Taylor, P. S., and Southwich, Jr., C. A., Ind. Eng. Chem., 20, 1327 (1928).

9

Frolich, P. K., Fenske, M. R., and Quiggle, D., Ind. Eng. Chem., 20, 694 (1928).

10

Natta, G., in “Catalysis”, Vol 111, ed. by P. H. Emmett, Reinhold, New York, 349 (19.55);

11

and refs. contained therein Audibert, E. and Raineau, A,, Ind. Eng. Chem., 20, 1105 (1928).

12

Cryder, D. S. and Frolich, P. K., Ind. Eng. Chem., 21,967 (1929).

13 14

Brown, R. L. and Galloway, A. E., Ind. Eng. Chem., 21, 1056 (1929). Frolich, P. K., Davidson, R. G., and Fenske, M. R., Ind. Eng Chem., 21, 109 (1928).

15

Frolich, P. K. and Lewis, W. K., Ind. Eng. Chem., 20,285 (1928).

16

Fenske, M. R. and Frolich, P. K., Ind. Eng. Chem., 21, 1052 (1929). Eguchi, T., Yamamoto, T., Yamauchi, S., Kuraishi, M., and Asakawa, K., U.S. Patent 3,

17 18

256,208 (June 14, 1966); assigned to Japan Gas-Chemical Co., Inc. Davies, P. and Snowdon, F. F., U.S. Patent 3, 326, 956 (June 20, 1967); assigned to Imperial Chemical Ind., Ltd.

19

German Patent 1,965,007 (Oct. 15, 1970); assigned to Catalysts and Chemicals, Inc.

20

Cornthwaite, D., German Patent 2,020, 194 (Sept. 10, 1970) and British Patent 1,296,212 (Nov. 15, 1972); assigned to Imperial Chemical Ind., Ltd.

21

22 23 24 2.5 26

Brocker, F. J., Marosi, L., Schrder, W., and Schwarzmann, M., German Patent 2, 056, 612 (May 31, 1972) and Brocker, F. J., German Patent 2, 116, 949 (Oct. 19, 1972); assigned to Badische Anilin- & Soda Fabrik AG. Collins, B. M., German Patent 2, 302, 658 (Aug. 2, 1973); assigned to Imperial Chemical Ind., Ltd. Stiles, A. B., German Patent 2, 320, 192 (Oct. 25, 1973); assigned to E. I. duPont de Nemours and Co. Rogerson, P. L., in “Handbook of Synfuels Technology”, ed. by R. A. Meyers, McGraw-Hill, New York, 2-45 (1984). Supp, E. and Quinkler, R. F., in “Handbook of Synfuels Technology”, ed. by R. A. Meyers, McGraw-Hill, New York, 2- 113 (1984). Compiled from Hydrocarbon Process., see Table 7.1.

341

27 28 29 30 31

Herman, R. G., Klier, K., Simmons, G. W., Finn, B. P., Bulko, J.B., and Kobylinski, T. P., J. Catal., 56, 407 (1979). Klier, K., Adv. Catal., 31, 243 (1982). Klier, K. and Herman, R. G., Published discussion of Paper No. 11-47 in Proc. 8th Intern. Congr. Catal., Berlin (1984). Ref. 10, p 405. Andrew, S. P. S., Post Congr. Symp. 7th Intern. Congr. Catal., Osaka, Japan, Paper 12 (1980).

32 33 34

Fleisch, T. H. and Mieville, R. L., J. Catal., 90, 165 (1984). Bridgewater, A. J., Wainwright, M. S., and Young, D. J., Appl. Catal., 28, 241 (1986). Bulko, J. B., Herman, R. G., Klier, K., and Simmons, G. W., J.Phys. Chem., 83, 31 18 ( 1979).

35 36

Mehta, S., Simmons, G. W., Klier, K., and Herman, R. G., J. Catal., 57, 339, (1979). Okamoto, Y., Fukino, K., Imanaka, T., and Teranishi, S., J. Phys. Chem., 87, 3730 and 3747 (1983).

37

3x 39 40

Karwacki, Jr., E. J., Anewalt, M. R., and Brown, D. M., Preprints, Div. Fuel Chem., ACS, 29(5), 210 (1984). Parris, G. E. and Klier, K., Preprints, Div. Fuel Chem., ACS, 29(5), 218 (1984) and J. Catal., 97, 374 (1986). Kau, L. S., Hodgson, K. O., and Solomon, E. I., J. Am. Chem. SOC.,111,7103 (1989). Didziulis, S. V., Butcher, K. D., Cohen, S. L., and Solomon, E. I., J. Am. Chem. Soc., 111, 71 10 (1989).

41 42 43 44 45 46 47 48 49 50 51 52

Apai, G., Monnier, J. R., and Hanrahan, M. J., J. Chem. SOC.,Chem. Commun., 212 (1984) and Appl. Surface Sci., 19, 307 (1984). Monnier, J. R., Hanrahan, M. J., and Apai, G., J. Catal., 92, 119 (1985). Apai, G., Monnier, J. R., and Preuss, D. R., J. Catal., 98, 563 (1986). Owen, G., Hawkes, C. M., Lloyd, D., Jennings, J. R., Lambert, R. M., and Nix, R. M., Appl. Catal., 33,405 (1987). Daly, F. P., J. Catal., 89, 131 (1984). Chinchen, G. C., Denny, P. J., Parker, D. G., Short, G. D., Spencer, M. S., Waugh, K. C., and Whan, D. A., Reprints, Div. Fuel Chem., ACS, 29(5), 178 (1984). Chinchen, G. C. and Waugh, K. C., J. Catal., 97, 280 (1986). Chinchen, G. C., Waugh, K. C., and Whan, D. A., Appl. Catal., 25, 101 (1986). Denise, B., Sneeden, R. P. A., Beguin, B., and Cherifi, O., Appl. Catal., 30, 353 (1987). Freidrich, J. B., Wainwright, M. S., and Young, D. J., J. Catal., 80, 1 (1983). Freidrich, J. B., Young, D. J., and Wainwright, M. S., J. Catal., 80, 14 (1983). Nix, R. M., Judd, R. W., Lambert, R. M., Jennings, J. R., and Owen, G., J. Catdl., 118, 175 (1989).

53

Matulewicz, E. R. A,, de Keijser, M. S., mol, J. C., and Kapleijn, F., Thennochim. Acta, 72, 111 (1984).

342

54

Dombek, B. D., Final Technical Report DE-AC22-84PC70022 from Union Carbide Corp. to the U.S. Department of Energy (Jan. 1987).

55

Nunan, J., Klier, K., Young, C.-W., Himelfarb, P. B., and Herman, R. G., J. Chem. Soc., Chem. Commun., 193 (1986).

56

Nunan, J. G., Bogdan, C. E., Klier, K., Smith, K. J., Young, C.-W., and Herman, R. G., J. Catal., 113,410 (1988).

57

Klier, K., Herman, R. G., and Vedage, G. A,, U.S. Patent 4, 843, 101 (June 27, 1989); assigned to Lehigh University.

58 59

Supp, E., CHEMTECH, 3,430 (1973). Klier, K., Chatikavanij, V., Herman, R. G., and Simmons, G. W., J. Catal., 74, 343 (1982).

60

Vedage, G. A., Pitchai, R., Herman, R. G., and Klier, K., Proc. 8th Intern. Congr. Catal., II,47 (1984).

61

Kagan, Y. B., Liberov, L. G., Slivinskii, E. V., Loktev, S. M., Lin, G. T., Rozovskii, A. Y., and Bashkirov, A. N., Dokl. Akad. Nauk. SSSR, 221, 1093 (1975).

62

Rozovskii, A. Y., Lin, G. T., Liberov, L. G., Sliminskii, E. V., Loktev, S. M., Kagan, Y. B., and Bashkirov, A. N., Kinet. Katal., 18,691 (1977).

63 64

Kuznetsov, V. D., Shub, F. S., and Temkin, M. J., Kinet. Katal., 23,932 (1982). Chinchen, G. C., Denny, P. J., Parker, D. G., Spencer, M. S., and Whan, D. A., Appl. Catal., 30, 333 (1987).

65

Kung, H. H., Liu, G., and Willcox, D., Preprints, Div. Fuel Chem., ACS, 29(5), 194 (1984).

66

Liu, G., Willcox, D., Garland, M., and Kung, H. H., J. Catal., 96,251 (1985).

67

Takagawa, M. and Ohsugi, M., J. Catal., 107, 161 (1987).

68

Poutsma, M. L., Elek, L. F., Ibarbia, P. A., Risch, A. P., and Rabo, J. A., J. Catal., 52, 157 ( 1978).

69

Klier, K., Herman, R. G., Simmons, G. W., Nunan, J. G., Smith, K. J., Bogdan, C. E., and Himelfarb, P. B.,“Direct Synthesis of 2-Methyl-l-Propanol/Methanol Fuels and

70

71 72 73 74

75

Feedstocks”, Final Technical Report DOE/PC/70021-F, Rev. 1,285 pp (August 1988). Klier, K., Herman, R. G., Nunan, J. G., Smith, K. J., Bogdan, C. E., Young, C.-W., and Santiesteban, J. G., in “Methane Conversion”, ed. by D. M. Bibby, C. D. Chang, R. F. Howe, and S. Yurchak, Elsevier, Amsterdam, 109 (1988). Himelfarb, P. B., Simmons, G. W., Klier, K., and Herman, R. G., J. Catal., 93,442 (1985). Nunan, J. G., Himelfarb, P. B., Heman, R. G., Klier, K., Bogdan, C. E., and Simmons, G. W., Inorg. Chem., 28,3868 (1989). Herman, R. G., Vedage, G. A,, and Klier, K., unpublished research. Bybell, D. G., Deutsch, P. P., Herman, R. G., Himelfarb, P. B., Nunan, J. G., Young, C. W., Bogdan, D. E., Simmons, G. W., and Klier, K., Preprints, Div. Pet. Chem., ACS, 31(1), 282 (1986). Klier, K., Young, C.-W., and Nunan, J. G., Ind. Eng. Chem. Fundam., 25, 36 (1986).

343

76 77 78

79 80 81 82 83 84 85 86 87 88 89

Baglin, E. B., Atkinson, G. B., and Nicks, L. J., U.S. Patent 4, 181, 630 (Jan. 1, 1980); assigned to the U.S. Department of the Interior. Baglin, E. B., Atkinson, G. B., and Nicks, L. J., Ind. Eng. Chem. Prod. Res. Dev., 20, 87 (1981). Atkinson, G. B., Baglin, E. G., Nicks, L. J., and Bauer, D. J., in "Catalytic Conversion of Synthesis Gas and Alcohols to Chemicals", ed. by R. G. Herman, Plenum Press, New York, 65 (1984). Wallace, W. E., France, J. E. and Shamsi, A,, in "Rare Earths in Modern Science and Technology", Vol. 3, Plenum Press, New York (1982). Amenomiya, Y., Ali Emish, I. T., Oliver, K. W., and Pleizier, G., Proc. 9th Intern. Cong. Catal., 2,634 (1988). Amenomiya, Y., Appl. Catal., 30,57 (1987). Shibata, M., Ohbayashi, Y., Kawata, N., Matsumoto, T., and Aoki, K., J. Catal., 96, 296 (1985). Pommier, B. and Teichner, S. J., Proc. 9th Intern. Cong. Catal., 2,610 (1988). Lambert, R. M., Nix, R. M., Jennings, J. R., and Owen, G., in "Catalysis 1987," ed. by J. W. Ward, Elsevier, Amsterdam, 131 (1988). Bryan, S. J., Jennings, J. R., Kipling, S. J., Owen, G., Lambert, R. M., and Nix, R. M., Appl. Catal., 40, 173 (1988). Nix, R. M., Rayment, T., Lambert, R. M., Jennings, J. R., and Owen, G., J. Catal., 106, 216 (1987). Dini, K. andDunlap, R. A., J. Phys., F15,273 (1985). Ichikawa, M. and Shikakura, K., Proc. 7th Intern. Congr. Catal., B17 (1980). Poels, E. K., van Broekhoven, E. H., van Barneveld, W. A. A., and Ponec, V., React. Kinet. Catal. Lett., 18,221 (1981).

90

Ryndin, Y. A., Hicks, R. F., Bell, A. T., and Yermakov, Y. I., J. Catal., 70,287 (1981).

91

Kikuzono, Y., Kagami, S., Naito, S., Onishi, T., and Tamaru, K., Faraday Discuss. Chem.

92 93 94 95

96 97 98

SOC.,72, 135 (1981). Fajula, F., Anthony, R. C., and Lunsford, J. H., J. Catal., 73,237 (1982). Driessen, J. M., Poels, E. K., Hindermann, J. P., and Ponec, V., J. Catal., 82, 26 (1983). Hicks, R. F. and Bell, A. T., J. Catal., 90, 205 (1984). Lunsford, J. H., in "Catalysis of Organic Reactions," ed. by R. L. Augustine, Marcel Dekker, New York, 95 (1985). van de Lee, G. and Ponec, V., Catal. Rev.-Sci. Eng., 29, 183 (1987). Meriaudeau, P., Dufaux, M., and Naccache, C., Proc. 8th Intern. Congr. Catal., 2, 185 (1984). Vedage, G. A., Himelfarb, P. B., Simmons, G. W., and Klier, K., ACS Symp. Ser., 279, 295 (1985).

344

99

Sapienza, R. S., Slegeir, W. A., O’Hare, T. E., and Mahajan, D., U.S. Patent Appl. 710, 879 (1986); U.S. DOE Office of Scientific and Technical Information Release PATENTS-US-A--A67 10879

100

Klingler, R. J., Bloom, I., and Rathke, J. W., Preprints, Div. Pet. Chem., ACS, 32, 61 (1986).

101

Klingler, R. J. and Rathke, J. W., Organometallics, 5,2568 (1986).

102

Westerterp, K. R., Bodewes, T. N., Vrijland, M. S. A., and Kuczynski, M., Hydrocarbon Process., 67(11), 69 (1988). Westerterp, K. R. and Kucynski, M., Hydrocarbon Process., 65(11), 80 (1986).

103 104 105 106

Westerterp, K. R. and Kucynski, M., Chem. Eng. Sci., 42, 1871 (1987). Kucynski, M., Oyevaar, M. H., Pieters, R. T., and Westerterp, K. R., Chem. Eng. Sci., 42, 1887 (1987). Kucynski, M., Van Ooteghem, A,, and Westerterp, K. R., Colloid Polymer Sci., 264, 362 (1986)

107

Sherwin, M. and Blum, D., “Liquid Phase Methanol”, Annual Report EPRI AF-202 (Project 317) to the Electric Power Research Institute, 100 pp (Aug. 1976).

108 109

110

111 112

113

114 115

Espino, R. L. and Pletzle, T. S., U.S. Patent 3, 888, 896 (June 10, 1975) and U.S. Patent 4, 031, 123 (June 21, 1977); assigned to Chem Systems Inc. Chem Systems Research and Development Group, “Development of Alcohol-Based Synthetic Transportation Fuels from Coal-Derived Synthesis Gas”, Technical Reports DOEET/14858-T1 to -T3 to the U.S. Department of Energy (1980). Studer, D. W., Holley, E. P., Hsiung, T. H., and Mednick, R. L., in “Indirect Liquefaction Contractors’ Review Meeting: Synthesis Gas Conversion to Liquid Fuels and Upgrading”, Vol. I, U.S. Department of Energy-Pittsburgh Energy Technology Center, 215 (Dec. 8-9, 1987; released in early 1988). Lewnard, J. J. and Hsiung, T. H., in “Catalysis 1987: Proc. 10th North Am. Catal. SOC. Meeting”, ed. by J. W. Ward, Elsevier, Amsterdam, 141 (1988). Frey, J. H., Studer, D. W., Henderson, J. L., and Weimer, R. F., in “Proc. Indirect Liquefaction Contractors’ Review Meeting”, U.S. Department of Energy-Pittsburgh Energy Technology Center, 5 (November 15-17, 1988; released in early 1989). Hsiung, T. H, Golden, T. C., and Underwood, R. P., in “Proc. Indirect Liquefaction Contractors’ Review Meeting”, U.S. Department of Energy-Pittsburgh Energy Technology Center, 76 (November 15-17, 1988; released in early 1989). Bonnell, L. W., Perka, A. T., and Roberts, G. W., U.S. Patent 4,766, 154 (Aug. 23, 1988); assigned to Air Products and Chemicals, Inc. Mednick, R. L. and Romeo, J. P., in “Proc. Indirect Liquefaction Contractors’ Review Meeting”, U.S. Department of Energy-Pittsburgh Energy Technology Center; 37 (November IS-17, 1988; released in early 1989).

345

116

117 118 119 120

121

122 123 124 125 126 127

128 129

130 131 132 133 134 135 136 137

Gormley, R. J., Rao, V. U. S., and Soong, Y., in “Proc. Indirect Liquefaction Contractors’ Review Meeting”, U.S. Department of Energy-Pittsburgh Energy Technology Center, 145 (November 15-17, 1988; released in early 1989). Gromley, R. J., Giusti, A. M., Rossini, S., and Rao, V. U. S., Proc. 9th Intern. Congr. Catal., 2, 553 (1988). Christiansen, J. A., U.S. Patent 1, 302,011 (1919). Boelens, C., as quoted in Bart, J. C. J. and Sneeden, R. P. A., Catal. Today, 2, 1 (1987). Klier, K., Herman, R. G., Simmons, G. W., Bogdan, C. E., Nunan, J. G., and Smith, K. J., in “Proc. 7th DOE Indirect Liquefaction Contractors’ Meeting-Vol I: Synthesis Gas Conversion to Liquid Fuels and Upgrading”, Pittsburgh, PA, December 7-9, 1987, (pp 258-379), released in early 1988. Bogdan, C. E., Nunan, J. G., Santiesteban, J. G., Herman, R. G. and Klier, K., in “Catalysis 1987: Proc. 10th North Am. Catal. SOC.Meeting”, ed. by J. W. Ward, Elsevier, Amsterdam, 745 (1988). Bogdan, C. E., Ph.D. Dissertation, Department of Chemistry, Lehigh University (1989). Hydrocarbon Process., 68(12), 63 (1989). Morandi, F., Trotta, R., Pecci, G., and Sposini, M., Energy Prog., 8(1), 1 (1988). Hydrocarbon Process., 68(12), 11 (1989). Herman, R. G., in “Catalytic Conversions of Synthesis Gas and Alcohols to Chemicals”, ed. by R. G. Herman, Plenum Press, New York, 3 (1984). Haag, W. O., Kuo, J. C., and Wender, I., in “Coal Gasification: Direct Applications and Syntheses of Chemicals and Fuels--A Research Needs Assessment”, ed. by S. S. Penner,

U.S. Department of Energy, Washington, D.C., 117 (1987). Schafer, S., Trondt, M., Langhoff, J., Konkol, W., and Hibbel, J., Energy Prog., 801, 28 ( 1988). Kuo, J. C.W., “Overview on Coal Gasification Systems”, Tutorial Lecture on Indirect Coal Liquefaction, 1984 Fall National Meeting of the American Chemical Society, Philadelphia, PA (August 26, 1984). Storch, H. H., Golumbic, N., and Anderson, R. B., “The Fischer-Tropsch and Related Syntheses”, John Wiley & Sons, New York (1951). Natta, G., Colombo, U., and Pasquon, I, In “Catalysis”, ed by P. H. Emmett, Vol. V, Chap. 3. Reinhold Publ. Co., New York (1 957). Frolich, P. K. and Cryder, D. S., Ind. Eng. Chem., 22, 1051 (1930). Graves,G. D., Ind. Eng. Chem., 23, 1381 (1931). Morgan, G. T., Hardy, D. V. N., and Procter, R. A,, J. SOC.Chem. Ind. Trans. Comniun., 51, 1T (1932). Taylor, R., J. Chem. SOC.(London), 1429 (1934); Gas World, 104,2696 (1936). Natta, G. and Rigamonti, R., Chim. e. Ind., 18. 623 (1936). Di Raffaele, P., Paggini, A., and Lagana, V., French Patent 2,482, 583 (1981); assigned to Snamprogetti.

346

138 139 140 141 142 143 144 145

Montarani, R., Aglietti, B., and Scarrosi, A., Arpel Congress on Heavy Oils, Caracas, Venezuela, Oct. 8-11, 1981. Keim, W. and Falter, W., Catal. Letters, 3,59 (1989). Hofstadt, C. E., Kochloefl, K., and Bock, O., Eur. Patent 0034338-A2, (Aug. 26, 1981); assigned to Sud Chemie A.G. Smith, K. and Anderson, R. B., Canad. J. Chem. Eng., 61,40 (1983) Smith, K. J. and Anderson, R. B., Preprints, Div. Fuel Chem., ACS, 29(5), 269 (1984). Smith, K. J. and Anderson, R. B., J. Catal., 85,428 (1984). Vedage, G. A., Himelfarb, P., Simmons, G. W., and Klier, K., Preprints, Div. Pet. Chem., ACS, 28(5), 1261 (1983). Klier, K., Herman, R. G., and Young, C. W., Preprints, Div. Fuel Chem., ACS, 29(5), 273 (1984).

Fattore, V., Notari, B., Paggini, A., and Lagana, V., U.S. Patent 4, 513, 100 (April 23, 1985); assigned to Snamprogetti S.p.A. 147 Schneider, M., Kochloefl, K., and Bock, O., U.S. Patent 4, 598, 061 (July 1, 1986); assigned to Siid-Chemie A.G. 148 Nunan, J. G., Bogdan, C. E., Herman, R. G., and Klier, K., Catal. Letters, 2,49 (1989). 149 Nunan, J. G., Bogdan, C. E., Klier, K., Smith, K. J., Young, C.-W., and Herman, R. G., J. Catal., 116, 195 (1989). 150 Nunan, J. G., Herman, R. G., and Klier, K., J. Catal., 116,222 (1989). 151 Sugier, A. and Freund, E., U.S. Patent 4, 122, 110 (Oct. 24, 1978); assigned to Institut

146

152 153 154 155

156

157 158 159

Francais du Petrole. Sugier, A. and Freund, E., U.S. Patent 4, 291, 126 (Dec. 19, 1979); assigned to Institut Francais du Petrole. Courty, P., Durand, D., Freund, E., and Sugier, A., J. Mol. Catal., 17,241 (1982). Courty, P., Durand, D., Sugier, A., and Freund, E., GB Patent 2, 118, 061 (March 25, 1983); assigned to Institut Francais du Petrole. Chaumette, P., Courty, Ph., Durand, D., Grandvaliet, P., and Travers, C., GB Patent 2, 158, 730 (May 10, 1985) and Ger. Offen. 3, 516, 928 (Nov. 14, 1985); assigned to Institut Francais du Petrole. Courty, P., Forestiere, A,, Kawata, N., Ohno, T., Raimbault, C., and Yoshimoto, M., Manuscript for the Symposium on Chemicals from Syngas and Methanol, 191st ACS National Meeting, New York, NY, April 13-18, 1986. Ohno, T., Yoshimoto, M., Asselineau, L., Courty, P., and Travers, P., Manuscript for the Spring AIChE National Meeting, New Orleans, LA, April 6-10, 1986. Uchiyama, S., Ohbayashi, Y., Hayasaka, T., and Kawata, N., Appl. Catal., 42, 143 (1988). Cochran, G. A., Conway, M. M., Murchison, C. B., Pynnonen, B. W., Quarderer, Q. H., Stevens, R. R., Stowe, R. A,, and Weihl, E. D.,Intern. Chem. Congr. Pac. Basin SOC., Honolulu, HI, Abstr. No. 03635 (Dec. 1984).

347

160

Quarderer, Q. J., and Cochran, G. A., Eur. Pat. Appl. 8, 410, 932.5 (March 16, 1984); assigned to Dow Chemical Co.

161

Kinkade, N. E., Eur. Pat. Appl. 841 16467.6 and 841 16468.4 (Dec. 28, 1984); assigned to Union Carbide Corporation. Stevens, R. R., Eur. Pat. Appl. 84116467.6 (Dec. 28, 1984); assigned to Dow Chemical co. Conway, M.M., Murchison, C. B., and Stevens, R. R., Eur. Pat. Appl. 85109215.5 (July 23, 1985); assigned to Dow Chemical Co.

162 163 164

Quarderer, G. J., Cochran, G. A., Stevens, R. R., and Murchison, C. B., Eur. Pat. Appl. 851092130.0 (July 23, 1985); assigned to Dow Chemical Co.

165

Conway, M. M., Murchison, C. B., and Stevens, R. R., U.S. Patent 4, 675, 344 (June 23,

166 167

1987); assigned to Dow Chemical Co. Dianis, W. P., Appl. Catal., 30,99 (1987). Santiesteban, J. G., Bogdan, C. E., Herman, R. G., and Klier, K., Proc. 9th Intern. Congr. Catal., Calgary, Canada, 2.561 (1988).

168

169 170 171

172

173 174 175 176 177 178 179 180 181

Klier, K., Herman, R. G., Simmons, G. W., Lyman, C. E., Santiesteban, J. G., Najbar, M, and Bastian, R., “Direct Synthesis of Alcohol Fuels over Molybdenum-Based Catalysts”, Final Technical Report DOE/PC/80014-F, 261 pp (December 1988). Murchison, C. B., Conway, M. M., Stevens, R. R., and Quarderer, G. J., Proc. 9th Intern. Congr. Catal., Calgary, Canada, 2, 626 (1988). Smith, K. J., Herman, R. G., Klier, K., Chem. Eng. Sci., 45,2639 (1990). Smith, K. J., Young, C.-W., Herman, R. G., and Klier, K., “Development of a Kinetic Model for Alcohol Synthesis over Cesium Promoted Cu/ZnO Catalysts”, Ind Eng Chem. Res., submitted. Smith, K. J., Young, C.-W., Nunan, J. G., Herman, R. G., and Klier, K., “Reaction Kinetics of the Higher Alcohol and Oxygenate Synthesis over Cesium Doped Cumno Catalysts”, to be published. Badische Anilin and Soda Fabrick, D.R. Patent 293,787 (1913); 295,202 (1914); 295,203 (1914); Span. Patent 20,488 (1913); and U.S. Patent 1, 201, 850 (1914). Fischer, F. and Tropsch, H., Brennstoff-Chem., 4, 276 (1923); 5, 201 (1924) 201; 5, 217 (1924). DuPont de Nemours, French Patent 657, 124 (1928); Brit. Patent 323,240 (1928). Fischer, F. and Tropsch, H., Ind. Eng. Chem., 21, 310 (1929). Frolich, P. K. and Lewis, W. K., Ind. Eng. Chem., 20, 354 (1928). Natta, G., Italian Patent 267,698 (1928). Wenzel, W., Angew. Chem., B20,225 (1948). Imperial Chemical Industries, Ltd., French Patent 1, 489, 682 (July 21, 1967); French Patent 2,037, 567 (Dec. 21, 1970); assigned to ICI Ltd. El Sawy, A. and Gray, D., in “Proc. 8th Indirect Liquefaction Contractors’ Review Meeting”, Pittsburgh, PA (November 15-17, 1988) pp 120-136; released in February 1989.

348

182

183 184 185 186 187

188 189 190 191 192 193 194 195 196 197 198 199

200 20 1 202

203

“Coal Liquefaction: A Research Needs Assessment-Vol I. Executive Summary”, U.S. Department of Energy Report DOE/ER-0400 prepared by Science Applications Intern. Corp. (March 1989) pp 3-21 to 3-29. Frank, M. E. and Hernandez-Robinson, A,, Manuscript for the National AIChE Meeting, New Orleans, LA, April 6-10, 1986. Paggini, A. and Sanfilippo, D., Manuscript for the National AIChE Meeting, New Orleans, LA, April 6-10, 1986. Hilsebein, W., Manuscript presented to the “1986 World Methanol Conference”, Frankfurt, West Germany, Dec. 8-10, 1986. Xiaoding, X., Doesburg, E. B. M., and Scholten, J. J. F., Catal. Today, 2, 125 (1987). Klier, K., in “Coal Liquefaction: A Research and Development Needs Assessment--Vol 11: Technical Background”, Final Report DOE/ER-0400 by Science Applications Intern. Corp. to U.S. Department of Energy, Section 5.3, pp 5-19 to 5-132 (March 1989). Mills, G. A., “Catalysts for Fuels from Syngas”, International Energy Agency Coal Research, London, Report No. IEACR/09, ISBN 92-9029-159-1, August 1988. Elliott, D. J. and Pennella, F., J. Catal., 102, 464 (1986). Lin, F. N. and Pennella, F., in “Catalytic Conversions of Synthesis Gas and Alcohols to Chemicals”, ed. by R. G. Herman, Plenum Press, New York, 53 (1984). Attig, T. G., Graham, A. M., and Pesa, F., U.S. Patent 4, 614, 750 (Sept. 30, 1986): assigned to the Standard Oil Co. Graham, A. M. and Attig, T. G., U.S. Patent 4, 72.5, 626 (Feb. 16, 1988); assigned to the Standard Oil Co. Hardman, H. F. and Beach, R. I., U.S. Patent 4, 609, 678 (Sept. 2, 1986); assigned to the Standard Oil Co. Bhasin, M. M., Barley, W. J., Ellgen, P. C., and Wilson, T. P., J. Catal., 54, 120 (1978). Ichikawa, M., Bull. Chem. SOC.Jpn., 51, 2268 (1978). Inoue, T., Iizuka, T., and Tanabe, K., Appl. Catal., 46, 1 (1989). Weisser, 0. and Landa, S., “Sulphide Catalysts, Their Properties and Applications”, Pergamon Press, New York (1973); and refs. contained therein. Saito, M. and Anderson, R. B., J. Catal., 63,438 (1980). Madon, R. J., Bucker, C. R., and Taylor, W. F., Final Report E(46-1)-8008-F to the U.S. Energy Res. Dev. Admin. (1977); as reported in R. G. Anderson, “The Fischer-Tropsch Synthesis”, Academic Press, Orlando, FL, 167 (1984). Klier, K., Herman, R. G., Simmons, G. W., and Lyman, C. E., in “Fifth DOE Indirect Liquefaction Contractors’ Meeting Proc.”, Houston, TX, Dec. 2-5, 1985, p 70. Santiesteban, J. G., Ph.D. Dissertation, Department of Chemistry, Lehigh University (1989). British Coal Research Establimshment, “Synthesis of Chemical Feedstocks and Intermediates”, Final Report ECSC No. 7220-EC/824 (March 1987). Stevens, R. R., Eur. Pat. Appl. 85109214.8 (July 23, 1985); assigned to Dow Chemical Co.

349

204

Youchang, X., Naasz, B. M., and Somorjai, G. A., Appl. Catal., 27,233 (1986).

205

Dry, M. E., in “Catalysis”, Vol. 1, ed. by J. R. Anderson and M. Boudart, Springer Verlag, 159 (1981).

206

207 208 209 210 211 212 213 214 215 216 217 218 219 220 22 1 222 223 224 225 226 227 228 229 230 23 1 232 233

Kinkade, N. E., Intern. Patent Appl. PCT/US84/02112 (Dec. 28, 1984), Intern. Publ. NO. WO 85 03, 074 (July 18, 1985) and Eur. Patent Appl. 84116468.4 (Dec. 28, 1984), Publ. No. 0 149 256 (July 24, 1985) ; assigned to Union Carbide Corp. Wender, I., Friedel, R. A,, and Orchin, M., Science, 113, 206 (195 1). Sakari, S., Kitaura, K., Morokuma, K., Okhubo, K., J. Am. Chem. Soc., 105, 2280 (1983). Thomas, G., Ann. Chim., 6 , 367 (1951). Fischer, F., Ind. Eng. Chem., 17, 576 (1925) and “Conversion of Coal into Oils”, Van Nostrand, New York, 251 (1925). Mazanec,T. J., J. Catal.,98, 115 (1986). Morgan, G. T., Proc. Roy. Soc. (London), A127,246 (1930). Brown, R. L. and Galloway, A. E., Ind. Eng. Chem., 20, 960 (1928); 21, 310 (1929); and 22, 175 (1930). Fox, J. R., Pesa, F. A,, and Curatalo, B. S., J. Catal., 90, 127 (1984). Weiss, A. H. and John, T., J. Catal., 32, 216 (1974). Cant, N. W., Tonner, S. P.. Trimm, D. L., and Wainwright, M. S., J. Catal., 91, 197 (1985). Takahaski, K., Takezawa, N., and Kobayaski, H., Chem. Lett., 1061 (1983). Denise, B. and Sneeden, R. P. A., C, Mol. Chem., I, 307 (1985). Mueller, L. L. and Griffin, G. L., J. Catal., 105, 352 (1987). Tonner, S. P., Trimm, D. L., and Wainwright, M. S., J. Mol. Catal., 18, 215 (1983). Elliott, D. J. and Pennella, F., J. Catal., 114, 90 (1988) Elliott, D. J. and Pennella, F., J. Catal., 119, 359 (1989). Young, C.-W., Herman, R. G., and Klier, K., to be published. Wender, I. and Sternberg, H. W., Adv. Catal., 9, 594 (1957). Burns, G. R., J. Amer. Chem. Soc., 77,6615 (1955). Denny, P. J. and Whan, D. A., Catal., Spec. Period. Rep., Chem. Soc. London. 2. 46 (1978). Pasquon, I. and Dente, M., J. Catal., 1, 508 (1962). Wermann, J., Lucas, K., and Gelbin, D., Z. Phys. Chem., 225, 234 (1964). Bakemeier, H. Laurer, P. R., and Schroder, W., Chem. Eng. Prog. Symp. Ser., 66(98), 1 ( 1970). Leonov, V. E., Karavaev, M. M., Tsybina, E. N., and Petrishcheva, G. S., Kinet. Katal., 14, 970 (1973); Engl. Trans., p. 848. Tronconi, E., Ferlazzo, N., Forzatti, P., and Pasquon, I., Ind. Eng. Chem. Res., 26, 2122 (1987). Marquardt, D. W., J. SOC.Ind. Appl. Math., 11,431 (1963). Herman, R. G., Smith, K. J., and Klier, K., “Kinetic Modelling of Alcohol Synthesis Over Alkali/MoSz Catalysts”, Final Report to Air Products and Chemicals, Inc. (May 1988).

350

CHAPTER 8

EFFECT OF HYDROGEN IN CONTROLLING CO HYDROGENATION

Laszlo Guczi Surface Science and Catalysis Laboratory Institute of Isotopes of the Hungarian Academy of Sciences, H-1525 Budapest, P.O. Box 77, (Hungary)

351

8.1 INTRODUCTION Hydrogen plays a vital role in determining the selectivity of a catalyst which is active for carbon monoxide hydrogenation. Generally, in most cases CO is chemisorbed by dissociative pathway and carbon species bound in different manners, are being present on the catalyst surface (refs. 1-4). The most active form of surface carbon is assigned as mobile carbon or surface carbide capable of migrating along the surface and forming CH, species by reaction with surface hydrogen.

or E’- carbide with a composition of approxiniately FeC,, 2 (ref. 5). These CH, species can be combined with each other in a surface propagation step to produce higher hydrocarbons or after being hydrogenated they can leave the surface in form of methane.

This species was also identified as

E-

Although the competition between propagation and termination reactions is an important factor, the main source of methane is the immobile carbon and inactive bulk carbide which have been identified on the surface (refs. 1,6). Immobile carbon is the main source for methane, while bulk carbide (Haeggs or X-carbide) is responsible for deactivation of the catalyst. The fomier can be considered as a multiply bound carbon being analogous to the structure found by Muttenies in Fe6C(CO)16(ref. 7). The mobile and immobile forms of carbon were distinguished by combined catalytic and i n situ Mossbauer experiments (ref. 8) under synthesis gas reaction conditions. Both on silica supported iron and iron-ruthenium (Ru/Fe=4.8) catalysts iron species with IS= 0.0 mm s-l and QS =

0.52 mm s-l was measured and assigned to an iron species with mobile carbon. Indeed, when CO was disconnected from the gas stream this doublet disappeared and Feo and Fe3+ species remainzd on the catalyst surface. Product analysis also indicated the disappearance of higher hydrocarbons

from the product stream leaving only methane as a sole reaction product. Both in the case of dissociative and associative CO chemisorption over metal catalysts hydrogen participates in the surface reaction and the rate and selectivity of CO hydrogenation is essentially controlled by the hydrogen coverage and its bonding state. On pure transition mzt:ils (single crystals, foils and films) hydrogen is easily adsorbed dissociatively in an exothermic process and rapid equilibrium is set up between the gas phase and the surface However, on supported metals or metals with promoters and poisons, hydrogen adsorption is limited by several kinetic factors. In several cases strongly bound hydrogen is foiined in an activated process and depending upon the temperature range of the CO hydrogenation reaction, these hydrogen atoms are not easily available. The second effect of hydrogen which is still under debate in the literature, is its ability to facilitate CO dissociation. The “hydrogen-assisted CO disbociation” mechanism put forhard by several authors (refs. 9,10) is refuted simply on the basis that the effect can be modellzd by assuming the acceleration of other elementary steps such as C-M + H-M = =CH-M + M and

0 - M + H-M = HlO + M (refs. 11,12). Another important phenomenon in the interaction of hydrogen with metals is the absorption or occlusion of hydrogen resulting in the formation of hydride phase. Recent investigations have demonstrated the influence of such interaction as evidenced by the observation of hydrogen induced

352

reconstruction of metal surfaces (ref. 13). Structure sensitive reactions are most affected by this process because this reconstruction occurs at or near the temperature which is normally applied in catalytic reactions. An excellent review has been recently published comprising the effect of hydrogen in catalytic reactions including the kinetics and thermodynamics of hydrogen adsorption as well a5 the participation of hydrogen in several other catalytic processes (ref. 14)

8.2 HYDROGEN ADSORPTION ON METAL SURFACE On most pure metal surfaces hydrogen adsorption occurs spontaneously, in an exothennic process. The first step is always a weak interaction in which hydrogen is adsorbed in a molecular form at temperature as low as 20 K (ref. 15). The process can be illustrated in Fig. 8.1 by a one dimensional Lennard-Jones energy diagram (ref. 16). The weak interaction is controlled by van der Waals forces and keeps the hydrogen molecule far away from the surface in a shallow energy wzll. Upon further approach of the hydrogen molecule toward the surface, the energy gained in the formation of two metal-hydrogen bonds overcompensates the energy requirement for the dissociation of a hydrogen-hydrogen bond. Thus in the chemisorbed state, the metal hydrogen atoms are in a deep energy minimum having bond energies in the range of 80-120 kJ mol-' compared to the attractive potential of 10-20 kJ mol-1 corresponding to the physisorbed btate. The intersection of the potential curves representing the physisorbed and chemisorbed states with the reference state of the hydrogen molecule couio be below, at or above the energy level of the gas phase hydrogen molecule. In the former cases "I2

hydrogen adsorption takes place without activation, while in the latter case adsorption requires an activation energy. For this process the following general equation is applied:

0

Ed=AH + Ead, where E,d, Ed and AH are the energy of activation for adsorption, for desorption and the heat of adsorption, respectively.

8.2 .I

QUANTUMCHEMI CAL APPROACH OF THE HYDROGEN BONDING

In order to obtain deeper insight into Fig. 8.1 Lennard-Jones potential diagram representing the physisorbed and chemisorbed state of H2 molecule. EDiss., EMvle-",E,d, E* and E represent the dissociation energy, the metal-{ydrogen bond strength, heat of adsorption, energy of activation and the heat of physisorption, respectively (from ref. 16).

the bonding of chemisorbed hydrogen atoms, it is worthwhile to consider the bonding state of hydrogen to Group VIII and IB metals. van Santen and associates (refs. 16,17) have pursued theoretical studies based on a simple

353

Hiickel and tight-binding theory (ref. 19) and ab initio calculations It is now generally accepted that in hydrogen adsorption on Group VIII and IB metals,

S-

electrons are involved and differences in chemisorption bonds are small and any changes in bond strength is attributed to the interaction of the hydrogen s-bands with d-electrons. The cohesive energy in IB metals is large but in Group VIII metals the d-electrons do not give significant contribution to it. Upon adsorption on transition metals, the strength of the cohesion energy among metal atoms becomes smaller. As we shall see later hydrogen adsorption is primarily affected by the sorbitals. The d-band makes only a small contribution. The width of the d-bands tends to decrease on going from left to right through a row of the Group VIII metals and increases on going from top to the bottom through the column. This is why slight changes in the heat of adsorption of hydrogen can be attributed to the position of the elements in the periodic table (ref. 20). Furthermore, as regards the heat of adsorption is concerned, there are differences between the reactivities even on the same metal which is due to difference in the face exposed and to the difference in the reactivity of the various metal atoms sitting in different position such as corner, terrace or edge. The fomier affects the adsorption site geometry, the latter influences the bond strength of a metal atom differing from each other by the coordinatively unsaturated character. Cluster model calculations assuming on-top adsorption of hydrogen (ref. 21), showed that the effect of decreasing number of neighbours of metal atoms to which hydrogen atom is adsorbed, is much smaller for a cluster embedded in an otherwise unchanged semi-infinite lattice than in an isolated cluster. Consequently, a decrease in the number of neighbours of the atoms involved in cheniisorption increases the adsorption energy of a hydrogen atom at the top-position of an s-band metal as long as the electron density of metal electrons is not higher than 1.6 electron/atom. At higher electron density a reversal effect was found. For s-metals the result of the theoretical calculations is as follows: - localization of electrons in a half-filled band increases the bond strength of hydrogen. The effect of adsorption on a completely filled band is the opposite; -

localization of electrons increases the bond strength on mono-coordinated atoms relative to that of multi-coordinated ones in a half filled band;

-

metal band filling (el/at>l) increases the bond strength of mono-coordinated relative to that of multi-coordinated atoms; - Usually multi-coordinated hydrogen is more strongly bound than mono-coordinated hydrogen. To conclude it can be established that for adsorbates with an s-type orbital such as hydrogen, multicoordination is favoured for low valence electron occupation but single-atom coordination becomes favoured at high electron occupation. Here, at low electron band occupation, the bonding is strongest for hydrogen atoms coordinated to metal atoms with the least number of metal neighbours. The effect of alloying is extremely important in hydrogen adsorption. Let us consider the electronic structure of Group VIII metals. The valence d-electron band is relatively narrow,

354

partially filled and it is overlapped by a

-t

broad, partly filled s-electron band. For IB metals the d-band is completely filled and here the Fermi level is higher than for Group VIII metals. Upon alloying, the Fermi levels should be equalized and

f

will be shifted to an average position. The d-bands become narrowed as indicated in Fig. 8.2. A study of the variation of binding energy of multi-coordinated hydrogen on alloy surfaces appears to be interesting.

The

quantumchenlical

approach was based upon a 10 atom cluster model and the binding energy of mono and tri-coordinated hydrogen was calculated as a function of the number of electrons. Using this model of a a

b

Fig. 8.2 d- and s-band structure for Group VIII and IB metals (a) and the effect of alloying (b). Solid line and dashed line are the band structures of a monometallic and an alloy, respectively (from ref. 17)

bimetallic cluster, van Santen's calculation indicated (ref. 17) that upon alloying transition metals with IB metals, the d-band width decreases and the binding energy decreases regardless of the band filling. Here

again

some

brief

conclusions can be presented regarding the effect of alloying on hydrogen bonding: -

-

-

alloying of Group VIII and IB metals results in an band filling of transition metal delectrons; band narrowing may increase the binding energy of mono-coordinated hydrogen, but that of multi-coordinated hydrogen is diminished. This decrease is the largest when the coordination of the hydrogen is decreased. Band filling tends to decrease the adsorption energy of hydrogen atoms coordinated to a transition metal atom. The binding energy for multi-coordinated hydrogen decreases more quickly than that for mono-coordinated one.

Previously it was pointed out that the number of neighbours of the metal atoms has an influence on the hydrogen adsorption, normally, through a narrowing of the d-band. This effect might be modelled through studies of changes in the hydrogen bond strength with metal dispersion. It is well known that when metal particle size is diminished the number of surface imperfections and/or the relative ratio of the number of comer, edge and terrace atoms (ref. 22) having various

355

influence from the neighbour atoms. This also gives a possibility to estimate -at least the trend - of the energetics of hydrogen bonding to differently oriented crystal faces

8.2.2 KINETICS AND ENERGETICS OF HYDROGEN ADSORPTION ON METALS As was demonstrated in the previous section, the binding state and energy of adsorbed hydrogen is one of the most significant characteristics in the interaction of hydrogen with metals. Although mainly s-electrons are involved in this process, here d- electrons also play an important role. On transition metals there are no significant trends in the change of heat of adsorption being

characteristic of the strength of hydrogen bonds (refs. 21,23) When one considers hydrogen adsorption it should first be discussed in terms of pure metals (films, foils, single crystals) and then on supported metals. As is well known in the latter case the support has a great influence on the adsorption process. On pure metals hydrogen adsorption is generally a non-activated process and depends only on the position of the metal in the periodic table as well as on the surface orientation. Here the surface can be handled, a priori, energetically homogeneous and the adsorption process takes place without activation (Ead = 0) and the heat of adsorption, AH, depends only on coverage, 0, that is, at higher coverage the repulsive forces between adsorbed hydrogen atoms tend to diminish AH. In this section a brief summary will be given on the kinetics of adsorption including activated adsorption, extent of adsorption, correlation between activation energy and the heat of adsorption, reversible and irreversible hydrogen adsorption.

8.2.2.1

Kinetics of hvdropen adsormion

On pure metals the rate of hydrogen adsorption can be described by the following equation: dn$dt = SO f(O) exp (-Ea&T)

I$

(particles m-2 s-l)

(8.1)

where so is the initial sticking coefficient at zero coverage, Ead is the energy of activation, @ in the impinging flux defined as @=dn/dt=pd(2zmkBT)(p: pressure; n: number of particle, k , Boltzman's constant). so varies over a wide range depending o n metals and the surface orientation. For instance, on Ni( 11l), Ni( 110) and Ni(100) surfaces so is 0.1, 0.96 and 0.06, respectively ((ref. 24) and more data therein). The probability of the impinging hydrogen molecule becoming dissociated depends on the shape of the potential and on the total kinetic energy of the incident H2 molecules. The kinetic energy probably exceeds the energy of activation and the heat of adsorption should be taken up by the solid, otherwise the hydrogen molecule is desorbed without dissociation. The sticking probability decreases quickly with increasing coverage, as expressed normally by a function of the form f(O)=(l-O)2 (ref. 13). However, this function is generally more complicated as both the hydrogen adsorption and desorption occur through the physisorbed state being considered as a precursor state.

356

Desorption of the dissociated hydrogen atoms also takes place according to the equation of

where O is the coverage, n is the order of desorption. For k the equation k = A,exp(-E&r) is applied where A, denotes the frequency factor. When the activation energy for hydrogen adsorption, Ead, is equal to zero, Ed is equal to AH. At equilibrium the rates of adsorption and desorption are equal and one can obtain the well known adsorption isotherms (Langmuir, Freundlich, Temkin, etc): this, however, is not the subject of this Chapter. The thermal desorption of hydrogen, one of the most powerful techniques in characterizing metal catalysts, will be discussed later (Section 8.2.2.4). The kinetics of adsorption have been carefully analysed by Aharoni and coworkers (ref. 2527). It was established that one characteristic property of the kinetics of adsorption is the S-shape curve in the reciprocal rate vs time plot. The lower, middle and upper part can be approximated by a power law, an Eiovich and Langmuir type adsorption equation, respectively. The rate can be described by the modified equation of (1):

where N(O), the number of free sites and Ead(0) and the energy of activation is function of the coverage, 0.Assuming N=NO(l-O) and E,d(@)=E,do

+ B e , the final result is

dO/dt=(klg) [ (l-O)/O)*exp(-Ead@T) where g and k are constant and Ead(O)/RT=Eoa@T

(8.4)

+ ln(gO).

For a heterogeneous surface, similar equation was derived assuming that the surface consists of an array of homogeneous patches characterized by a given value of the adsorption energy, m, heat of adsorption. This is constant for a patch varying from patch to patch. The energy of activation, Ead, depends on AH and the coverage on the patch given by:

For desorption it can be written:

The rate equation at the patch AH is: dOldt = k, (1-0) exp(-Ead/RT) - kd @ exp(-Ed/RT)

(8.7)

357

After rearrangement and assuming equilibrium

a,= l/[K-lexp(-AH)+l]

(8.8)

where K=k,Slc,. The rate equation in terms of equilibrium coverage and by neglecting the desorption is as follows: dO/dt = (l/y)[( 1-O)/O]exp(-aAH) where y=gk,. In integrated form: -ln(l-0)-0 = (t/y)exp(-aAH)

(8.10)

The 0 vs AH plot for a given k, and cx using various t/y as parameters are given in Fig. 8.3 (ref. 26). The quantity of adsorbate q taken up by the surface at time t is the sum of the quantities taken up by the patches and is given by:

4=

r n H 0 dAH

(8.1 1)

where Hg and H , are the patches with lowest and highest energy, respectively and nH is the number of adsorption sites in a patch AH. The important message here for heterogeneous surface is that for a given set of parameters the higher the AH, the higher 0 is and simultaneously the quantity taken up by the surface also increases with increasing heat of adsorption. Considering the correlation between the energy of activation of adsorption and the heat of adsorption this becomes important in the area of activated adsorption as will be discussed later.

Fig. 8.3 Plot of 0 vs AH for k,=10-7, c ~ = land various t/y values changing from lo5 to 5 lo'* in four steps denoted by 1-4.(from ref. 27).

358

Extent and stnichiornetn: o f hydrogen adsorption At equilibrium, the amount of hydrogen adsorbed should be determined by the pressure and temperature according to the adsorption isotherms. However this is generally not the case even on metal films or on foils. The adsorption of hydrogen is a fast process and large amount of hydrogen (5040%) can be chemisorbed at low pressure. When the hydrogen pressure is diminished the amount of hydrogen desorbed is smaller than that is expected according to the equilibrium pressure, that is, a part of hydrogen remains irreversibly adsorbed. If one considers the favourable case when chemisorption takes place without activation, according to the energy diagram (see Fig. 8.1) desorption requires an activation energy which equals or exceeds the heat of adsorption. Unless the temperature of desorption is increased, the onset of equilibrium takes a infinitely long time. (see later in Section 8.2.2.3) This fraction of the chemisorbed hydrogen is considered as the irreversibly adsorbed hydrogen and it has been measured to determine metallic surface area (refs. 28-32). The extent of “irreversibility” of hydrogen adsorption depends on several factors. Thus on the first place the temperature can be mentioned: at higher temperature large part of hydrogen adsorbed behaves as reversibly bound hydrogen. Furthermore, here we have to mentioned the particle size, the metal loading, metal/support interaction, promoter effect, and the degree of reduction. The irreversible fraction of chemisorbed hydrogen is used to determine surface area. On unsupported metals the extent of hydrogen adsorption is approximately 0.8- 1.0 monolayer measured at 77 K. At temperature above 200 K the extent of adsorption tends to be affected by the pressure and due to the change of the sticking coefficient higher pressure is required to achieve high coverage. On supported metal catalysts there are several problems in determining H/M=l stoichoimetry for metals. The first is migration of the hydrogen chemisorbed on the metal component to the support (spillover). Menon and associates (refs. 33,34) indicated a special care in determining the stoichoimetry on Pt/A1203 in 02-H2 titrations due to the interfemng effect of water. Guczi et a1 (ref. 35) established a stoichiometry of H/Pt = 1.3 on Pt/SiO, by comparing the O M and WM ratio. Here when the H/M=1.3 ratio is accepted the stoichoimetry for titration is exactly H/Os=3. It means that part of the hydrogen is transferred to the support after dissociation. The second problem in determining the stoichoimetry of hydrogen adsorption is the reaction of metal particles with the support. This is particularly important with supports containing OH groups and with non-noble transition metals (Fe, Co, Ni, etc) of low loading (refs. 35,361. Under these conditions small metal particles are formed which easily react with the surface OH group, for example. Although this reaction takes place only at 370 K (ref. 37), the reactivity of ultradispersed metal particles makes this reaction feasible even at room temperature. Thus, detemiining the extent of hydrogen adsorption, if any, is very uncertain and not reliable. Feo + HO = FeO + 0.5 H2 In the case of Strong Metal Support Interaction, SMSI, (excellent review in ref. 38) the decrease in the extent of hydrogen adsorption served to establish this phenomenon. Here after high 8.2.2.2

.

359

temperature reduction (as interpreted by the present view) metal particles are decorated with the moiety of the support e. g. TiO, or Lao, and thus the extent of hydrogen adsorption decreases. This effect plays an important role in the activated adsorption of hydrogen (see Section 8.2.2.3) Not much effort has been expended on experiments to determine the correlation between catalyst structure, particle size and the extent of weak - strong hydrogen adsorption. Sayari et al has studied zeolite and Si02 supported Ru particles for hydrogen chemisorption (ref. 39). Earlier it was observed that at very high dispersion all the hydrogen taken up is strongly chemisorbed (ref. 40). At room temperature and for an average particle size of about 1.6 nm, the weakly bound, fully reversible adsorption amount is 30%. Below and above this particle size the fraction of reversibly adsorbed hydrogen sharply decreases. According to their explanation there is an ensemble of about 5-10 adjacent metal atoms, presumably a B5 site, which is responsible for this fraction of weakly held hydrogen. However, there are discrepancies in this explanation because reversible adsorption was found also on particle of about 0.9 nm in diameter which is below the size of a B, site (ref. 41). Secondly, as was pointed out in Section 8.2.1 the strength of hydrogen adsorption decreases with an increase in the number of neighbour atoms and on this basis one would expect an increase of the fraction of reversibly adsorbed hydrogen with increasing particle size.

5 The phenomenon of strongly adsorbed hydrogen on metals was discovered as long as thirty years ago by Gundry (ref. 42). Later, additional evidence were obtained for this effect (refs. 43-45). A series exchange experiments between CH, and tritium showed (ref. 43) that for the exchange conducted at higher temperature there was an increase in the amount of total radioactivity which was due to the presence of the tritium strongly adsorbed on nickel and could not be pumped away at the reaction temperature. In additional experiments it was proven that tritium chemisorbed at higher temperature could be easily exchanged with hydrogen, but not with methane. The effect of strogly bound hydrogen was demonstrated in the product distribution for cyclohexane and deuterium exchange on nickel as shown in Fig. 8.4 (ref. 44). The more strongly bound hydrogen remains on the surface, the lower the cyclohexane-

D,2/D, ratio is. Since for multiple exchange adjacent nickel sites are required to be present on the surface, this strongly bound hydrogen may break up this large nickel ensembles, thus, the single exchange becomes predominant. The effect of strongly held hydrogen could be verified also in other experiments (ref. 45). Silica supported ruthenium was studied for butane hydrogenolysis and in ethane-deuterium exchange. After reduction in hydrogen at 770 K the catalyst lost its activity when the evacuation at the reduction temperature was not carried out. As the X-Ray Diffraction indicated no apparent change in the size of the metal particles, that is, no sintering occurred. When the sample was accidentally exposed to air at room temperature, the catalytic activity was restored which called our attention on the role of strongly bound hydrogen. Indeed, after reduction at 770 K, the same activity was found for butane hydrogenolysis as that measured after removing the strongly held hydrogen by evacuation at the same temperature.

360 c

2.l

- hexane -d12

Menon and coworkers (refs. 46-49) also found similar effects on alumina, silica

:-hexane -dl

and titania supported Pt catalysts, that were reduced in hydrogen in the range between

770 K and 870 K. They identified this strongly bound hydrogen through TPD measurements when the main TPD peak was shifted towards higher temperature. Similar effects were found by Nagy and 1.( associates (refs. 50,51) on unsupported platinum when the sample reduced at high temperature was also cooled down in hydrogen. They came to the similar 0 .' conclusion which was also put forward by Guczi et al (ref. 45). Accordingly, hydrogen adsorbed at high temperatures occupied a subsurface position which could clearly be 23 40 60 80 I00 I20 t lrnin) distinguished from the usual low Fig. 8.4 Variation of C6D12 to C6HI1D ratio on temperature hydrogen absorption and it Ni catalyst after different treatments. (0)after rquired higher temperature for desorption. evacuation at 670K; (A) after evacuation at 470 K; (0)after preadsorption of 50 torr D2 for 2 The proof for this was the exchange h at 570 K followed by evacuation at 570 K; between chemisorbed deuterium and ethane after preadsorption of 50 torr D2 at 670 K for 2h followed by evacuation at 570 K (from ref. 44). over Ru/Si02 sample (ref. 45). Here, after the exchange equilibrium had been achieved in the first run, the sample was mildly oxidized to remove the subsurface deuterium. In the second run the same rate was measured for exchange indicating the presence of subsurface deuterium, which had been forced to migrate to the surface by the oxygen treatment. In all these experiments the common phenomenon was the formation of strongly held hydrogen after high temperature adsorption. This is in agreement in the principle we mentioned in Section 8.2.2.2 formulated by Eq. 8.5. If high energy sites exist on the metal surface, activation endrgy is required to fill them. Now, the crucial question arisen is: what is the nature of these sites? As was shown earlier on pure metals hydrogen adsorption nonnally takes place without activation energy, while on supported metal this phenomenon is less common. Thus, the question to be answered is whether or not this is not an artifact, and if not, what are the possible explanations. The first problem with activated hydrogen adsorption whether the dissociation is promoted only by specific surface structures of the metals or not. For instance, open surfaces such as the (110) face of the fcc metals have been shown significantly change its structure upon hydrogen adsorption (refs. 52,53). This different structure can be established only at higher temperatures, thus

(a,)

361

the perturbation of the metal structure requires some activation (refs. 53-55). However, the activation energy not only involves the H-H bond rupture, but also breaking the metal-metal bond. Recently, activated hydrogen chemisorption was observed on several supported metals. Kovalinka et a1 (ref. 56) on Ni/Si02, Raupp and Dumesic (refs. 57,58) Bartholomew and associates (refs. 59-63) over supported Ni, Co and Fe catalysts, Guczi et al. on ruthenium and ruthenium-iron catalyst (ref. 64), on Fe-Re/A1203 (refs. 65,66), on PtRu/A120g (ref. 67) and on IrCo/Al203 (ref. 68) and more recently Stockwell et al. (ref. 69) studied Fe, Ni and Rh supported on various camers observed activated adsorption of hydrogen. In contrast to the conclusions reached using unsupported metals, the new binding state of hydrogen on supported catalysts is most plausibly explained by the presence of e.g. unreduced nickel (ref. 59) and aluminum oxide particles on the nickel surface or by intimate contact between small nickel crystallites. Here the surface contaminants not only unreduced particles, but others like carbon, copper or potassium (ref. 57) may also induce an activation energy for adsorption and in the case of potassium the adsorption strength is also increased. The decrease in saturation hydrogen coverage as well as the higher binding energy due to the decoration of nickel by TiO, was observed by Raupp and Dumesic (refs. 57,58). Similar effects were found on supported cobalt and iron. By way of ?.&planationit is generally assumed (ref. 32) that the presence of reduced or unreduced moieties of support decorate the metal surface. According to the quantumchemical approach given in Section 8.2.1, the strength of a multi-coordinated hydrogen atom is enhanced by decreasing the filling of the d-band or by decreasing the neighbours of the adsorption site. Here, indeed, on small metal particles the binding state have a higher energy than for a large crystal face. Furthermore, there is a linear correlation between the energy of activation and the heat of adsorption thus, binding state of higher energy requires higher activation energy. The presence of activated hydrogen adsorption is most easily observed by using temperature programmed desorption techniques. The relative amount of hydrogen in the different binding s i t e s can be measured by the amount of hydrogen desorbing at various peak temperatures. (see in detail in 8.3). There are two indications of the activated adsorption phenomenon: i) by changing the temperature of hydrogen adsorption, new desorption states appear in the TPD as was shown by several authors. ii)

there is a shift in peak position towards higher temperatures as well as an increase in the amount desorbed as the temperature of adsorption is increased. Clearly one has to distinguish the effect of a non completely reduced catalyst from the activated adsorption. In the former case the amount of hydrogen taken up by the sample increases but there is no shift in TPD peak temperature. In the latter case the peak temperature is also shifted because with the increasing adsorption temperature more strongly bound hydrogen adsorption states are available. Often a shift in the peak position instead of the appearance of a new peak is the result of poor resolution of the TPD apparatus.

362

8.3 TEMPERATURE PROGRAMMED DESORPTION OF HYDROGEN 8.3.1 DESORPTION OF HYDROGEN FROM METALS Thermal desorption of hydrogen is a rather complicated process because the thermal energy must be transferred from the solid to the adsorbed molecule. There are several energy states in the energy well characteristic of the adsorbed state which must be passed by the molecule before leaving the surface. Nevertheless, the thermal desorption of hydrogen is the most powerful method in characterizing the metal surface and in obtaining information about the binding state of hydrogen. Since the preexponential factor (A,, in equ. (2)) is independent of coverage, the position of the thermal desorption peak and the order of desorption are the most significant variables in this method. The thermal desorption of adsorbates from unsupported metals, single crystals foils, etc may have different mechanisms. Here one has to distinguish zero order (having constant number of desorption centres), first order (which is characteristic of a molecular desorption) and second order (in which desorption takes place via recombination of dissociated molecules). obviously Second order kinetics was observed for hydrogen desorption from a Ru(0001) single crystal surface (ref. 70). The TPD peak present at 450 K with 0.05 L exposure was shifted to 320 K at 50 L exposure. In general, the energy of activation can be calculated from the temperature of peak maximum. Sometimes at higher coverage a new peak appears at low temperature which can be attributed to a coverage dependent weakly adsorbed state of hydrogen (ref. 71).

8.3.2 BASIC KNOWLEDGE ABOUT TEMPERATURE PROGRAMMED DESORPTION OF HYDROGEN A comprehensive work on thermal desorption has been published by Falconer and Schwarz (ref. 72). The reader can find a full description about the different methods to determine kinetic

parameters for thermal desorption in a flow system. Here the experimental technique is rather simple. Hydrogen is blended with an inert gas (helium) carrier stream passing through the catalyst bed. At the desired adsorption temperature after satiration coverage is achieved, the flow of hydrogen is disconnected and the temperature of the catalyst is raised. The amount of hydrogen desorbing into the carrier gas stream is measured by a suitable detector downstream. The theoretical basis for this method was given by Cvetanovic and Amenomiya (ref. 73). Their results are as follows. The rate of changing the surface coverage in time as well as the mass balance for a flow system can be written as -dO/dt = k,j(@) 0" - k, C , (1 - 0 ) P

(8.12)

and

(8.13)

363

where F is the volumetric flow rate of the carrier gas, C, is the gas phase concentration of the gas to be adsorbed, Vc is the total solid volume, V , is the number of surface sites per unit solid volume. If linear temperature rise is used the equations (11) and (12) can be combined: -d@/dT = (F/p) (kd (0)@"/[F+Vc V,,, k, (l-@)P])

(8.14)

Two cases can be distingushed: when readsorption is small or readsorption occurs freely: No readsorption:

with free readsorption: -d@/dT = (FWCV , b) [A,(@) OVA, (1-0)PI exp(-AH(O)/RT)

(8.16)

The experimental techniques are based upon equations (8.15) and (8.16). Several methods are known (ref. 72) among which perhaps the most useful is the heating rate variation method. Here it is only necessary to measure the shift of the peak position towards higher temperature with increasing heating rate (refs. 73-75). Simultaneously, the peak intensity increases and reach the maximum rate within shorter time interval. It is assumed that the fractional coverage is independent of the heating rate. The necessary conditions are i) constant 0, must be applied and variable heating rate, 6. Parameters to be measured are the peak temperature Tp and peak intensity, Ip From these data the activation energy for desorption can be measured by plotting ln(p/Tp2) vs l/Tpor lnIp vs l/Tp. Hydrogen TPD has been measured for several systems. Similarly to the unsupported metals, here the strength of various binding states of hydrogen to the surface can be fingerprinted. Without going into details (it is the subject of Section 8.3) the following principles can be established. i) Dispersion effect. The degree of dispersion is one of the most important variables in determining the change in high energy states and activated hydrogen adsorption. For supported Pt (refs. 76-79) as the dispersion increased, a broadening of the TPD peak, a shift towards higher temperature was established. On Ru/A1203 not only does the amount of hydrogen recovered in TPD decrease with increasing dispersion, but the TPD peak temperature is simultaneously shifted to higher temperature as indicated in Fig. 8.5 (ref. 64). In the case of Fe-Re samples (refs. 65,66) the dispersion of iron was shown to increase with the addition of rhenium and simultaneously the high temperature peak is getting larger as compared to the low energy hydrogen TPD peak. Besides the increase of the number of coordinatively unsaturated sites, which arises due to increased dispersion one must also consider the so called "porthole" effect introduced by Rumpf et al. for CO oxidation (ref. 80). In the case of hydrogen being adsorbed in dissociative manner, it

364

I b.l 373K

400K I

-

373K

,

473K

773 K

,423 K -333K -295 K I

1

273

373

473

573 K

673

773

273

373

473

573

673

713

K

Fig. 8.5 Effect of dispersion on the TPD characteristic of Ru/A1203. (a) 10 wt% Ru/A1203, (b) 1 wt% Ru/A1203 (from ref. 64)

must find free metal sites after inpinging the surface. Let us suppose that hydrogen arrives at the surface on a support site, it then must migrate to find metal sites for dissociation. This is also the mechanism for the reverse process, because hydrogen atoms are well separated on small metal particles and thus they acquire energy for migrating to the porthole to form there a hydrogen molecule ready for desorption. Both processes are activated and proceed much slower than that on a large metal crystal where a large pool of hydrogen atoms ensures the easy desorption. ii) Effect of s u u ~ o r tand uretreatments. There are considerable differences among the generally applied supports such as alumina, silica, titania, vanadia, lanthana and magnesia. Generally silica behaves as a weakly bound support and only at very high temperatures forms silicates. In contrast, alumina easily forms spinel structures with the non noble transition metals such as nickel, cobalt and iron. Ti02 can be easily reduced to Ti3+ and the TiO, may enter electron interaction with the metal thus influencing the d-electron density on the metal (refs. 81-83). At typical reduction temperature other oxides like lanthana also tend to decorate metal surfaces (ref. 84). In all cases there is an enhanced metal-support interaction which makes the hydrogen adsorption more activated and in this sense it is very similar to that observed in highly dispersed system (ref. 85). It is difficult to separate the effect of the various support from that of calcination and reduction pretreatments carried out at higher temperature. Some supports such as lanthana (ref. 86) and titania (ref. 87) cause the same effect. For lanthana even the temperature of 670 K is sufficient to achieve a state corresponding to activated hydrogen chemisorption whereas on alumina a calcination temperature of 1170 K is necessary to transfer hydrogen into strongly bound state. The difference in supports can be well demonstrated in the Ru/AI2O3 and Ru/SiO2 systems as shown in Fig. 8.6 (ref. 64). On silica supported ruthenium the hydrogen chemisorption is not activated. With

365 348 K

(c

I 295K 77 3 K

A , (a

I -773K 673K

373 K

473 K

A

573K 473K

333K 295K I

273

373

473

573 K

3

673

773

I

273

I

373

573

473

673

I

773

K

Fig. 8.6 TPD of hydrogen from Ru/Al2O3 and Ru/SiO2 catalysts. (a) lwt%Ru/A1203, (b) 1wt'%Ru/Si02 (c) unsupported ruthenium (from ref. 64).

increasing adsorption temperature the position of the TPD peaks do not change and the amount of hydrogen desorbed remains constant. Supporting evidence for the absence of strong Ru-silica interaction is that the position of the peaks in the TPD agrees well with that measured over unsupported ruthenium. On the other hand, on Ru/A1203 the presence of a RuO,-A1203 interface causes the hydrogen adsorption to become strongly activated as indicated by the shift in the hydrogen TPD peaks. This interface has been also shown by other groups (ref. 88). Similar decoration features exist on Ni/A1203 samples reduced at higher temperature as evidenced by the discovery of a NiA120, spinel phase in addition to metallic nickel crystallites (ref. 85). The formation of Co surface phase resistant to reduction was found also to affect the activated adsorption of hydrogen (ref. 89). We have to emphasize that the H;JM = 1 stoichoimetry for hydrogen adsorption is not always fulfilled at room temperature adsorption. It is generally true for unsupported metals, single crystals, foils and films with clean surface. Whenever the morphology of the metal surface is changed due to the decoration, or formation of a metal-surface interface, the hydrogen adsorption becomes activated thus the room temperature adsorption does not give a direct measure of the adsorption stoichoimetry. To overcome this difficulty, adsorption should be performed at elevated temperatures and the amount of hydrogen recovered in the subsequent TPD should be taken as the amount of adsorbed hydrogen. iii) Effect of nromoters and poisons. Here one has to mention at the first place the classical experiments of Kovalinka and Scholten (ref. 90) in which the authors found a decrease of weakly

bound hydrogen on palladium caused by addition of Zn, Pb and Ca. Potassium was also found to increase the number of adsorption states of hydrogen over iron catalysts (ref. 91). The effect of potassium is ambiguous because while some authors found the weakening effect of hydrogen

366

adsorption (ref. 72), others (refs. 66,92) observed an increase in the number of strongly bound sites. The addition of potassium on the FeRe/SiO2 system causes on the low temperature hydrogen TPD peak to completely disappear (Fig. 8.7). Only the high temperature TPD peak remains. At the same time, XPS and in situ Mossbauer spectroscopy measurements indicates the stabilization of highly dispersed particles. Thus, it is not necessary to invoke an electronic interaction between potassium and the active metal component to explain the strongly bound state but more simply it is an effect of the high dispersion. The most common poisons for a metal catalyst are the sulphur, carbon and chlorine. Data on their effect on Fig. 8.7 Change of hydrogen TPD pattern of ReFe/SiOZ samples on the addition of potassium. (from ref. 66)

chemisorption is available mainly for CO chemisorption. However, it can be assumed that in presence of hydrogen adsorption at higher temperature these poisons can be

removed and the adsorption capacity of a metal surface regain its normal value. Only one experiment is available (ref. 93) where it was established that diminished hydrogen adsorption was attributed to the trace amount of chlorine. After hydrogenation at 970 K the chlorine could be removed and the hydrogen adsorption increased to its normal value. This effect caused a discrepancy in the judgement of the importance of hydrogen spillover over Ru/Si02 catalyst. As was stated earlier no shift in the hydrogen TPD curves was found on Ru/Si02 (ref. 64). Kakuta and White (ref. 94) found on the same system that as the exposure temperature increases there is steady growth in the additional amount of deuterium desorbed as shown in Fig. 8.8. Since the reduction temperature was only 577 K, obviously some amounts of chlorine whichever retained by the sample could be removed during the Oose t e m p e r a t u r e ( K )

Fig. 8.8 TPD peak area of deuterium desorbed vs dose temperature

subsequent increasing

deuterium temperature.

dosing Thus

at the

367

increase found could be due to this effect rather than to spillover as stated by the authors.

In conclusion, TPD is a useful technique to gain deeper insight into the binding state of hydrogen. While on unsupported metals there is no activated adsorption which is manifested by a shift of the TPD peak towards higher temperature. Many factors may affect the hydrogen adsorption on supported metals. Among them the most important are the dispersion, support and pretreatment effect and the various promoters and poisons. The impact of the activated hydrogen on the CO hydrogenation is the subject of the next Section.

8.4 EFFECT OF HYDROGEN BONDING ON THE SELECTIVITY IN CO HYDROGENATION 8.4.1 HYDROCARBON AND OLEFIN FORMATION CO chemisorption may be dissociative to form carbon and oxygen attached to the surface, or associative in which case no C - 0 bond cleavage takes place, or at the most, CO adlineation. The first process results in hydrocarbon formation whereas the latter leads to oxygenate formation. In Chapters 4 and 5 the factors determining the degree of dissociation of CO and its participation in hydrocarbon formation reactions have been discussed. The effect of bimetallic catalysts has also been highlighted in Chapter 6. Although the product is determined by the mode of CO interaction, there are several type steps in the general mechanism which could be controlled by the amount hydrogen available at the surface. Normally, chain propagation and chain termination reactions occur simultaneously and the competition is affected by hydrogen to a large extent. Furthermore, the state of the catalyst is determined by the hydrogen present as a reaction component and sometimes it prevents the catalyst deactivation. The basic reactions in CO hydrogenation are as follow:

CO + 2M -+ C-M + 0 - M H2 + 2M -+ 2H-M C-M + H-M +.... -+ CH2M CH2-M + CH3-M -+ CH3- -CH,-M CH3-M + H-M -+ CH, + M CH3- CH2-M + H-M -+ CH3- -CH3 + M CH2-M + CH3-M -+ C2H4 + H-M

(8.17) (8.18) (8.19) (8.20) (8.21) (8.22) (8.23)

As one can see reactions (8.21) and (8.22) depend on the hydrogen coverage, more precisely, on the strength of hydrogen bond to the surface. These two reactions are in competition with reaction (8.23) and when hydrogen is depleted under the reaction condition, the CO hydrogenation is shifted towards olefin formation. Frennet formulated this reaction path in the following way (ref. 95). Taking step (8.18) as the rate determining step and measuring its rate constant CO pressure, he found

368

where R is the rate of reaction e is a constant and f is function of the hydrogen pressure. As the surface coverage of hydrogen is also a function of hydrogen pressure, the rate is also influenced by the ratio of the amounts of weakly to strongly bound hydrogen. First we have to consider the effect of particle size. On Fe and Co catalysts, which are normally used in the FT reaction, a Schulz-Flory type of distribution can be found. On small particles, the chain propagation is terminated at low molecular weigth products and the main compounds are up to C, hydrocarbons with the prevailing composition of predominantly olefins. One obvious reason for this is the presence of bulk carbide (refs. 96,97). If small particles or ensembles of the active metal components could be stabilized either by selecting the proper support (ref. 98) or using carbonyl clusters (refs. 99,100) the carbide formation could be significantly suppressed. Amorphous alloys behave in a similar manner, namely, small numbers of iron atoms are surrounded by metalloids (B, P, etc) thereby preventing the formation of a-iron on the surface (ref. 101). As we have previously pointed out, on small metal particles the hydrogen available for hydrogenation is easily depleted, thus the straightforward result is the increase of olefin content among the products. Dispersion can be influenced by different methods. Synthesis gas reactions over iron combined with Re (refs. 65,66) or on nickel with rhenium in presence of CaO (ref. 102), result in increased olefin formation. As indicated in Fig. 8.9, the addition of 10 at. 7i Re to iron lead to a one and a half order of magnitude increase in the rate of reaction without effecting the olefin selectivity. The small enhancement in C,, selectivity is probably due to the slight increase in the amount of mobile carbon on the surface. When one compares the kinetic data with that of the hydrogen TPD results measured for Fe/Si02 and FegoReldSi02 samples (see Fig. 8.7), two conclusions can be drawn. First, in the presence of rhenium, small iron particles are stabilized and the formation of large carbide phases during reaction is negligible (ref. 103) as indicated by Mossbauer spectroscopy (see Fig. 8.10). On pure Fe/Si02 the TPD results do not show a low temperature TPD peak, which explanes the high olefin selectivity. On a Feg$eldSi02 sample a small TPD peak of hydrogen can be seen. However, this is not sufficient to significantly affect the olefin selectivity. The small increase in the weakly bound hydrogen undoubtedly does influence the propagation step as competition between reactions (8.23) and (8.20) is successful due to some increase in the hydrogen coverage represented by M-H. By increasing the relative coverage of weakly bound hydrogen as shown in the hydrogen TPD, the olefin content gradually decreases with the increasing atomic per cent of rhenium. The reason is again obvious: termination requires hydrogen (see steps (8.21) and (8.22)) and when the weakly bound hydrogen supply is sufficiently large, the main products are saturated hydrocarbon. C,, also drops with increasing rhenium content which is again indicative of the competition between propagation and termination steps. When hydrogen is chemisorbed in a non-activated manner such as for NiRe or for Felr (ref. 65), the low temperature hydrogen peak is predominant regardless of the adsorption temperature of hydrogen. Only very small amount of olefin is formed (selectivity of olefin is around

369 100

30% ) (ref. 102). Similar phenomena was found for RuFe/A1203

$

(ref. 64), PtRu/A1,03 (ref. 67) and CoIr/A1203 (ref. 68). Here a relationship was observed between methane selectivity and the

x + -

1

+ Y

W

proportion of weakly bound hydrogen represented by the low

W

111

temperature hydrogen TPD peak as shown in Fig. 8.1 l b for PtRu/Al,O,.

0

The shift in the hydrogen TPD peak is also an

indicator of the changing hydrogen bond strength. The higher the TPD peak maximum the less available the hydrogen for reaction. It is, therefore, expected that olefin selectivity increases with TPD peak temperature as was found for RuFe/A1203 shown in Fig. 8.1 la. Ruthenium behaves similarly as NiRe or FeIr. When ruthenium is used in the form of evaporated film (ref. 104) the only product that one can observe, is methane and the TPD data

Fig. 8.9 Rate (in mol s" gCiL1), olefin and C2+ selectivity in the CO+H, reaction over silica supported iron and ironrhenium system (from ref. 66))

indicates an activated adsorption of hydrogen (when hydrogen adsorption takes place at 673 K the peak appearing at 523K becomes larger than that desorbed at 383 K. Nevertheless it is still low temperature, so hydrogen is weakly bound to ruthenium film. When ruthenium is impregnated on silica or alumina, Ru/Si02 behaves very similar to that of unsupported ruthenium film as presented in Fig. 8.12. The main product is methane with very small quantities of olefins and higher hydrocarbons. In contrast, on Ru/A1203 where according to the hydrogen TPD measurements considerably suonger interaction exists between the metal particles and the alumina, high amount of olefin is expected and significant fraction of C2+products is formed as indicated in Fig. 8.12. As follows from the general rule of the effect of hydrogen with increasing dispersion, the peaks of the hydrogen TPD curves are shifted to higher temperature (as shown in Fig. 8.6) and consequently, there is a depletion of weakly bound hydrogen. Thus, olefin selectivity increases as shown in Fig. 8.12. Upon the addition of iron to the ruthenium catalyst, a new pattern in the H2TPD curves appears, indicating a further shift of chemisorbed hydrogen towards the stronger

H2

Fe Re

1

1

1

8

I

I

I

I

I

-6-4-2 0 2 4 Velocity (mm/s)

I

,

6

,

I

8

Fig. 8.10 In situ Mossbauer data of Fe/Si02 and FeRe/Si02 after reduction and after CO+H reaction at 670 K (from ref. 66)

370

binding states. The effect on selectivity is most dramatic for the RuFe/Si02 samples. Here we have shown (ref. 45) that due to the very poor reducibility of iron, it stays on the silica surface as a surface layer making an energy barrier for the migration of small ruthenium particles. Small metal particles can, therefor, be stabilized. In his excellent review Bartholomew pointed out (ref. 105) several other factors which influenced the selectivity values in C0+H2 reaction. The amount of activated hydrogen was followed in TPD experiments on Ni/TiO, reduced at or above 670 K and on Ni/A1203 at low metal loading. In all these samples the formation of new, highly activated hydrogen adsorption sites was accompanied by high temperature hydrogen TPD peak. For instance, on Ni/A1203, a short calcination at 770 K followed by reduction at the same temperature made the catalyst highly selective for C2+ hydrocarbon formation and the selectivity was further increased after reduction at 700°C. Cobalt supported on silica, titania and alumina behaves in different ways. The alumina supported cobalt is the least active (ref. 106).Earlier Bartholomew (ref. 107) had argued that this is due to the formation of different cobalt species on alumina, namely, CO~O,, Co2+ and cobalt aluminate which could be reduced only at successively increasing temperature. Thus, the activity is determined by the type of prevailing cobalt species at a certain reduction temperature. This was also verified by XPS studies (ref. 89). In conclusion, the influence of hydrogen on the CO+H2 reaction largely depends on the variation in reduction temperature, calcination temperature, support effects and preparation methods.

(a 1

OLefin s d e c t i v i t y vs TPD temperature

0 Hydrogen TPD peak temperature,&gree

Fig. 8.11. Selectivity of methane as a function of the proportion of weakly bound hydrogen on alumina supported PtRu (curve b). Curve a shows the olefin selectivity vs hydrogen TPD peak temperature. (from refs. 64 and 67).

371

-3 -

40

-L -

20

Po Ru/Si02

-

S

J

t

,

20

'

,

40

60

-

01%1

L 20

40

60

83 01%

Fig. 8.12 TOF and selectivity values of CO hydrogenation on Ru/SiOl and Ru/A1203 as well as RuFe bimetallic samples (from ref. 64)

8.4.2

HYDROGEN EFFECT IN ALCOHOL FORMATION

As we stated earlier (ref. 103) alcohol formation is not affected to a large extent by the bonding state of hydrogen. Rather, it is controlled by the adsorption mode of carbon monoxide. In some respects the elementary steps established for hydrocarbon formation, i. e. propagation and termination are also valid here (see in Chapter 7). In this competition hydrogen does not play such an important role as in determining hydrocarbon selectivity. Hydrogen is required as a reaction partner, consequently, in its absence alcohol formation is diminished, but there are no other products unlike the case of olefin formation in the hydrocarbon forming reactions. Here only one example is worth mentioning. On the supported palladium modified by iron and lanthanum (ref. 108-113) it was established that methanol activity on palladium is associated with PdFe bimetallic formation (refs. 108,11l,l13,I14). Addition of 16 at. % iron to Pd (2 wt%) supported on silica, increases the methanol activity over an order of magnitude as illustrated in Fig. 8 . 1 3 . According to the Mossbauer data i) formation of bimetallic PdFe particles have been proven; ii) due

to the change in the isomer shift the s-electron density at the iron nucleus decreases and it is due to the rehybridization of the d-electron orbitals which results in the decrease of d-electron density at the palladium. It would mean that further addition of iron would increase the methanol activity which is, however, not the case. Hydrogen coverage, however is also affected by iron addition to palladium. As indicated in Fig. 8.13b there is a sharp decrease in the H/Pd ratio, i. e. less and less hydrogen is available for hydrogenation of the non dissociated CO molecule. The two effects annihilate each other, thus the maximum can be explained. Namely, when large amount of

372

0

C D

n \ 2=

C

I

Oi2 2.0

0.5 2.0

1.0 2.0

2;0F&P/o 2DW/Wt%

0 26 4b ' 6 0 80 Fe/at% 0 02 0.5 1 2 5 k&t% 2

2

2

2

2 Pd/wt%

Fig. 8.13 (a) Methanol activity in CO+H2 (b) Wpd ratio as a reaction over PdFe/SiO2 catalysts function of iron on PdFe/Si02 (from ref. 111)

hydrogen is present, the amount of CO chemisorbed in appropriate form is small and as the latter increases with the amount of iron added, the hydrogen coverage decreases, so the methanol activity should pass through a maximum.

8.4.3

EFFECT OF PROMOTERS ON THE ACTIVATED HYDROGEN

A great number of investigations have been performed to study the effect of promoters on

the CO+H2 reaction because the first industrially feasible catalyst contained both thoria and potassium as promoters. A comprehensive review was written by Mross (ref. 115) on the effect of potassium and other promoters. Other studies on single crystal iron (ref. 116), on supported catalysts (refs. 117-120) and on catalysts prepared from molecular carbonyl clusters (refs. 121,122) have given excellent interpretation, however, no particular emphasize was given to the role of promoters in the activation of hydrogen. As was stated earlier the activation energy of hydrogen adsorption is changed by the presence of promoter or impurities on the surface. This, of course, affects the ratio of the coverage

of weakly to strongly held hydrogen which can alter the activity and selectivity of the catalyst.

373

Table 8.1

XPS data of Fe, FeK, FeRe and FeReK samples at impregnation and after reduction at 720 K (first and second row, respectively) (from ref. 65).

catalyst

Fe 2P312 Fe Fe FeK FeK FeRe FeRe FeReK FeReK

Peak area ratio

B. E. (eV)

711.6 710.7 711.9 711.5 710.9 710.9 711.5 711.0

Re 4fTn

Fe/Si

294 294 45.2 BG BG 41.6

293.6 294

0.13 0.03 0.09 0.07 0.09 0.03 0.06 0.06

ReISi

0.0 15 BG BG 0.02

Note: BG stands for background

Rankin and Bartholomew found a drastic change in the Fe/K/Si02 system regarding activity and selectivity (refs. 62,123). Both the catalytic activity and the activation energy for the CO+H2 reaction decreased while the selectivity for light olefin increased with increasing temperature of precalcination as well as with increasing potassium content. This was explained by the increasing energy of activation for hydrogen adsorption. On Felalumina catalysts Arakawa and Bell (ref. 124) found that the iron dispersion decreased with potassium loading and a simultaneous decrease in turnover frequency. However, this depended on the pretreatment temperature. An enhancement in activity after calcination at 373 K was reported but a decrease was noted after calcination at 473 K. It seems reasonable that activity can be influenced by promoters simply by altering the CO dissociation rate. Although this also depends on the pretreatment, there is agreement in the literature that hydrogen adsorption becomes a more activated process with the addition of promoters. This is why in the temperature range of CO+H2 reaction the coverage of hydrogen drastically decreases causing increased olefin formation. The effect of potassium was thoroughly investigated by means of Mossbauer and electron spectroscopy methods (refs. 65,66). XPS results indicated (shown in Table 8.1) that at 720 K the reduction of iron occurs to a very small extent and according to the Fe 2p3,2/Si 2p intensity ratio the sintenng which occurs is significantly prevented by potassium addition. The addition of rhenium helps somewhat in stabilizing small particles however, when both Re and K are added small particle size can be further stabilized as indicated by the Fe/Si signal ratio. The Mossbauer experiments have given more distinctive data about the state of iron. In Table 8.2 these data are presented. The interpretation of the Mossbauer data is as follows. The reduction of the pure iron catalyst at 720 K results mainly in the formation of Fe2+ ions located in different coordination types. Reduction at 870 K leads to the presence of significant amount of zero valent iron. The addition of rhenium to iron facilitates the interaction between the two metals but not necessary the formation of bimetallic particles. This is well supported by TPR measurements (ref. 65). The interaction can be suggested also on the basis of the decrease in IS and MHF values.

374

Table 8.2

Mossbauer data for Fe. FeK. FeRe and FeReK after reduction at 770 K and at 870 K (from ref. 65) samples

Component

Parameters

Fe

1

FeK

I

FeRe

FeReK

Reduction temperature: 720 K I

1

IS QS

Fe3+

0.43 1.03 19

%

1 Fe2+(L)

IS QS

1.12 0.80 28

0.87 1.51 33

IS QS

1.16 1.56 29

1.18 1.48 25

IS

1.11 2.36 32

1.16 2.08 23

%

tGq&T I

%

QS

%

Feo(singlet)

IS

15

28

i

1.08 1.00

1

1.16 1.49

2. 03

1.98 I

0.14 11 3.7

%

3.1

Reduction temperature: 870 K

' 1 0.86 24

Fe2+(H1)

1

I

'I

I

0.73 17

0.89 1.61 21

0.92 1.22 15

0.85 1.93 22

1.24 1.61 27

1.22 1.66 27

1.93 2. 05 7

I

IS

Feo(singlet)

%

21 dlc2/df

I

3.1

1.5

0.8

15

I

1.1

I

Note: IS and QS are the isomer shift and quadrupole splitting, respectively; MHF is the internal hyperfine splitting in k&;% the relative ercentage of the spectral area; dlc2/df is the goodness of the fit; in Fe2+ (L), and (H) means the F e g in low and high coordination states, respectively.

375

The effect of rhenium on the dispersion of iron particles is evident since the coalescence of particles is retarded. In order to explain the data given by the Mossbauer and XPS experiments, the surface morphology of the support should be taken into consideration. It must be assumed that coordinatively unsaturated sites are avilable at the surface which are filled up first during impregnation leaving some fraction of the impregnating ions weakly bound to the surface. This latter fraction forms metal particles during reduction which could migrate along the surface causing the formation of larger metal particles. In the presence of rhenium the CUS sites are simultaneously filled up with Fe3+ and Re7+ ions and after calcination and reduction a mixed oxide phase can be formed which obstructs free migration. Here only interfaces with superparamagnetic particles are formed in contrast to the large particles of iron formed without the rhenium. This is supported by the XPS data when one inspects the (Fe/Si)500 to (Fe/Si) impr, ratio being 0.23 for pure iron and 0.33 for FeRe sample. Addition of potassium results in the stabilization of Fe3+ and Fe2+(L) and in the hinderance

of reduction to some extent. Undoubtedly, stabilization of the initial dispersion is appearent from the XPS and Mossbauer data. From Table 8.2 one can see that the largest iron particles are present after reduction of the Fe/silica sample in hydrogen at 6 W C , whereas on the FeReK catalyst the reduction is more limited. The effect of potassium on the hydrogen TPD and on the catalytic activity has been presented in Figures 8.7 and 8.9, respectively. The most plausible explanation is as follows. First, the increase of activity on the addition of rhenium is largely due to two factors. i) the presence of Fe0,-ReOy-Si02 oxide phase helps the CO activation simply by CO adlineation, that is, binding the carbon to the metal component and the oxygen end to the reduced rhenium oxide. This mechanism is well accepted in the literature (see e.g. ref. 120). The second effect is prevention of the inactive carbide formation which is the main source of deactivation. This is attributed to the stabilization of small metal particles shown in the Mossbauer spectra (see Table 8.2). AS shown in Fig. 8.7 addition of rhenium increases the amount of weakly adsorbed hydrogen, however, it has here only of marginal importance because selectivity values do not change to a large extent. Nevertheless, addition of potassium strongly affects the amount of strongly bound hydrogen because potassium located mainly on the overlayer of the oxide component, the WSi ratio being not subjected to any changes indicated by XPS (ref. 65). It means that here the main effect is the depletion of weakly bound hydrogen desorbing in the range of 373-473 K. As a consequence, the rate of the reaction decreases in accordance with other literature data and the olefin selectivity increases as shown in Fig. 8.9

8.5 CONCLUSIONS In this Chapter the importance of weakly and strongly bound hydrogen has been established in the CO+H2 reaction. Quantum mechanical calculations allow certain predictions regarding hydrogen bonding to metal surfaces to be made. Accordingly, all factors which decreases the number of neighbour atom on a surface makes the hydrogen bonding stronger.

376

The factors influencing the activation energy of hydrogen adsorption have been reviewed. It appears that apart from some cases in which the catalyst is not properly reduced or surface contamination is present, activation of the hydrogen chemisorption is due to metal support interaction, or limited particle size. The role of strongly bound hydrogen is given for several CO hydrogenation processes, such as the Fischer-Tropsch reaction and alcohol formation. The limitations of our understanding is how the various effects can be deconvoluted into such factors as hydrogen effects, CO adlineation or hindered deactivation.

8.6 REFERENCES 1

2

K. Lazir, Z. Schay and L. Guczi, J. Mol. Catal., 17 (1982) 205 A. T. Bell and P. Winslow, Proc. 8th International Congress on Catalysis, Verlag Chemie, Weinheim, 1984 Vol. 3 p. 175

3 3a 4 5 6 7 8 9 10 11 12 13 14 15

16 17 18 19 20 21

P. Winslow and A. T. Bell J. Catal., 91 (1985) 142; A. T. Bell, Proc. 9th Int. Congress on Catalysis, Eds. M. J. Phillips and M. Teman, Chem. Inst. Canada, Ottawa, 1988, p. 134 L. Guczi, Z. Zsoldos and Z. Schay, J. Vac. Sci. Technol., A5, (1987) 1070 G. B. Raupp and W. N. Delgass, J. Catal., 58 (1979) 361 J. W. Niemantsverdriet, A. M. van der Kraan, W. L. van Dijk and J. S. van der Baan, J. Phys. Chem., 84 (1980) 3363 E.L. Muetterties and J. Stein, Chem. Rev. 79(1979)479 K. LQzQr,W. M. Reiff and L. Guczi, Hyperfine Interact. 28 (1986) 871 S. V. Ho and P. Hamott, J. Catal., 64 (1980) 262 D. Bianchi and C. 0. Bennett, J. Catal., 86 (1984) 433 A. T. Bell, Catal. Rev. Sci. Eng., 23 (1981) 203 E. L. Sughrue and C. H. Bartholomew, Appl. Catal., 2 (1982) 239 K. R. Christman, in “Hydrogen Effect in Catalysis”, Eds. Z. Pail and P. G. Menon, Marcel Decker Inc., New York, 1984, p.4 Z. Pa61 and P. G. Menon, eds “Hydrogen Effect in Catalysis, Fundamental and Practical Applications”, Marcel Dekker Inc., New York, 1988 J. E. Demuth, D. Schmeisser and Ph. Avonis, Phys. Rev. Lett., 47 (1981) 1166; ibid. 48 (1982) 199 J. E. Lennard-Jones, Trans. Faraday SOC.,28 (1932) 28 R. A. van Santen, Recl. Trav. Chim. Pays-Bas, 101 (1982) 121 and references therein R. A. van Santen, Faraday Symp. Chem. Soc. 21 (1986) paper 1 R. A. Hoffmann, M. M.-L Chen and D. L. Thorn, Inorg. Chem. 16 (1977) 503; Science 21 (1981) 995 E. L. Muetterties, T. N. Rhodin, E. Band C. F. Brucker and W. R. Pretzer, Chem. Rev. 79 (1979) 91 R. A. van Santen, Surf. Sci. 53 (1975) 35

377

22 23 24

25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49

J. R. Anderson, “Structure of Metal Catalysts”, Academic Press, New York, 1975 K. Kuchinka, J. Fusek and 0. Strouf, Collection of Czech. Chem. Comm., 46 (1981) 2328; ibid 47 (1982) 2363 M. P. Moms, M. Bowker and D. A. King, in “Sample Process at the Gas-Solid Interface”, Comprehensive Chemical Kinetics Series, Eds. C. A. Bamford, C. F. H. Tipper and R. G. Compton, Vol. 19 Elsevier, Amsterdam, 1984 C. Aharoni and Y. Suzin, J. Chem. SOC.Faraday I., 78 (1982) 2329 M. Ungarish and C. Aharoni, J. Chem. Sco. Faraday I., 79 (1983) 119 C. Aharoni, Adsorption Science and Technology, 1 (1984) 1 J. W. Geuss in Hydrogen Effect in Catalysis, Eds. Z. Paal and P. G. Menon, Marcel Decker Inc., New York, 1984 p. 85 J. H. Sinfelt and D. J. Yates, J. Catal., 8 (1967) 82 D. J. Yates and J. H. Sinfelt, J. Catal., 8 (1967) 348 P. G. Menon, in “Hydrogen Effect in Catalysis”, Eds. Z. Paal and P. G. Menon, Marcel Decker Inc., New York, 1984 p. 117 C. H. Bartholomew, in “Hydrogen Effect in Catalysis”, Eds. Z. Pail and P. G. Menon, Marcel Decker Inc., New York, 1984 p. 139 J. Prasad and P. G. Menon, J. Catal., 44 (1976) 314 J. Prasad, K. R. Murty and P. G. Menon, J. Catal., 52 (1978) 515 L. Guczi, K. Matusek and M. Eszterle, J. Catal, 60 (1979) 121

K. Lazrir, K. Matusek, J. Mink, S. Dobos, L. Guczi, L. Vizi-Orosz, L. Marko and W. M. Reiff, J. Catal., 87 (1984) 163 N. Nahon, V. Perrichon, P. Turlier and P. Busier, J. Phys. (Pans) C1 (1980) 339 R. Burch, in Hydrogen Effect in Catalysis, Eds. Z. Pail and P. G. Menon, Marcel Decker Inc., New York, 1984 p. ? M. Gundry, Proceeding of the 2nd International Congress on Catalysis, Technip, Paris, 1960p. 1095 J. G. Goodwin, J. Catal., 68 (1981) 228 R. van Hardeveld and F. Hartog, Surf. Sci. 15 (1969) 189 M. Gundry, Proceeding of the 2nd International Congress on Catalysis, Technip, Paris, 1960p. 1095 L. Guczi and P. Tetenyi, Acta Chim. Hung. 5 1 (1967) 275 L. Guczi and P. Tetenyi, Annals of the New York Acad. Sci., 213 (1973) 173 L. Guczi, K. Matusek, I. Manninger, J. Kiraly and M. Eszterle, Preparation of Catalysts II., Eds. B. Delmon, G. Poncelet and P. Jacobs, Elsevier, Amsterdam, 1979 p. 391 P. G . Menon and G. F. Froment, J. Catal., 59 (1979) 138 P. G. Menon and G. F. Froment, J. Catal., Appl. Catal., 1 (1981) 31 P. G. Menon and G. F. Froment, in “Metal-Support and Metal-Additive Effects in Catalysis”, Eds. B. Imelik et al., Elsevier, Amsterdam, 1982, p. 171 P. G. Menon and G. F. Froment, Acta Chim. Acad. Sci. Hung., 11 1 (1982) 631

378

50

F. Nagy, D. Moger, M. Hegedus, G. Mink and S. Szabo, Acta Chim. Acad. Sci. Hung., 100, (1979) 211

51 52

S. Szabo, D. Moger, M. Hegedus and F. Nagy, React. Kinet. and Catal. Lett., 6 (1976) 89 G. J. R. Jones, J. H. Onuferko, D. P. Woodruff and B. W. Holland, Surf. Sci., 147 (1984) 1 K. Christmann, F. Chehab, V. Penka and G. Ertl, Surf. Sci., 152/153 (1985) 356 K. Griffith, P. R. Norton, J. A. Davies, W. N. Unertl and T. E. Jackman, Surf. Sci.,

53 54

152/153 (1985) 374 L. Olles and A. M. Baro, Surf. Sci., 137 (1984) 607 56 J. A. Kovalinka, P. H. van Oeffelt and J. J. F. Scholten, Appl. Catal., 1 (1981) 141 57 G. B. Raupp and J. A. Dumesic, J. Catal., 95 (1985) 587 58 G. B. Raupp and J. A. Dumesic, J. Catal., 97 (1986) 85 59 G. D. Weatherbee and C. H. Bartholomew, J. Catal., 87 (1984) 55 60 J. M. Zowtiak and C. H. Bartholomew, J. Catal., 82 (1983) 107 61 J. M. Zowtiak and C. H. Bartholomew, J. Catal., 82 (1983) 230 62 J. L. Rankin and C. H. Bartholomew, J. Catal., 100 (1986) 533 63 G. D. Weatherbee, J. L. Rankin and C. H. Bartholomew, Appl. Catal., 11 (1984) 73 64 L. Guczi, Z. Schay, K. Matusek and I. Bogyai, Appl. Catal., 22 (1986) 289 65 Z. Schay, K. LBzBr, I. Bogyay and L. Guczi, Appl. Catal., 51 (1989) 33 66 Z. Schay, K. Lazar, K. Matusek, I. Bogyay and L. Guczi, Appl. Catal., 51 (1989) 49 67 K. Matusek, I. Bogyay, L. Guczi, G. Diaz, F. Garin and G. Maire, C, Mol. Chem., 1 (1985) 335 68 L. Guczi, K. Matusek, I. Bogyay, F. Garin, P. Esteban-Puges, P. Girard and G. Maire, C,

55

Mol. Chem., 1 (1986) 355 69

D. M. Stockwell, A. Bertucco, G. W. Coulston and C. 0. Bennett, J. Catal., 113 (1988) 317

70

H. Shimizu, K. Christmann and G. Ertl, J. Catal., 61 (1980) 412

71

K. Christmann, M. Ehsasi, J. H Block and W. Hirschwald, Chem. Phys. Lett. 131 (1986)

72 73 74 75 76 77 78 79 80

J. L. Falconer and J. A. Schwarz, Catal. Rev. Sci. Eng. 25 (1983) 141 R. J. Cvetanovic and Y. Amenomiya, Adv. Catal., 17 (1967) 103 J. L. Falconer and R. J. Madix, Surf. Sci. 48 (1975) 393 F. M. Ford and J. S. Kittleberger, Surf. Sci. 43 (1974) 173 J. R. Anderson, K. Foger and R. J. Breakspere, J. Catal., 57 (1979) 66 J. D. Way and J. L. Falconer, 2nd Chemical Congress of North America, Las Vegas, 1980 J. R. Katzer, Adv. in Catal. Chemistry, Snowbird, 1979 K. Foger and R. J. Anderson, Appl. Surf. Sci., 2 (1979) 335 F. Rumpf, H. Poppa and M. Boudart, Langmuir, 4, (1988) 722 J. A. Dumesic, S. A. Stevenson, R. D. Sherwood and R. T. K. Baker, J. Catal., 99 (1986) 79 J. A. Cairns, J. E. Baglin, G. J. Clark and J. F. Ziegler, J. Catal., 83 (1983) 301

192

81 82

379

83 84

85 86 87 88

89 90 91 92 93 94 95 96 91 98 99

T. Huizinga and R. Prins, in Metal-Support and Metal-Additive Effects in Catalysis, Eds B. Imelik et al., Elsevier, Amsterdam, 1982, p. 11 R. P. Underwood and A. T. Bell, J. Catal., 109 (1988) 61; ibid 11 (1988) 325 and references therein C. H. Bartholomew and R. B. Pannell, J. Catal., 65 (1980) 390 G. R. Gallaher, J. G. Goodwin and L. Guczi, submitted for publication A. D. Logan, E. J. Braunschweig and A. K. Datye, Langmuir, 4 (1988) 827 A. Bossi, F. Garbassi, A. Orlandi, G. Petrini and L. Zanderighi, Preparation of Catalysts 11, Elsevier, Amsterdam, 1979 p. 405 Z. Zsoldos, T. Hoffer and L. Guczi, J. Phys. Chem., in press J. A. Kovalinka and J. J. F. Scholten, J. Catal., 48 (1977) 374 Y. Amenomiya and G. Plazier, J. Catal., 28 (1973) 442 R. D. Gonzales and H. Miura, J. Catal., 77 (1982) 338 T. Narita, H. Miura, K. Sugiyama, T. Matsuda and R. D. Gonzales, J. Catal., 103 (1987) 492 N. Kakuta and J. M. White, J. Catal., 97 (1986) 150 A. Frennet, in “Hydrogen Effect in Catalysis”, Eds. Z. Paal and P. G. Menon, Marcel Decker Inc., New York, 1984 p. 399 G. B. Raupp and W. N. Delgass, J. Catal., 58 (1979) 337

J. A. Amelse, J. B. Butt and L. H. Schwartz, J. Phys. Chem., 82 (1978) 558 S. J. Teichner, F. Blanchard, B. Pommier and J. P. Reymond, 8th North American Catalysis Society Meeting, Philadelphia, May 1983, paper B-6 F. Hugues, P. Bussiere, J. M. Basset, D. Commereuc, Y. Chauvin, L. Bonneviot and D. Olivier, Proc. 7th Int. Congress on Catalysis, Kodansha, Tokyo, and Elsevier, Amsterdam, 1981, Part A, p. 418

100

L. Guczi, Z. Schay,K. Matusek, I. Bogyay and G. Stefler, Proc. 7th Int. Congress on Catalysis, Kodansha, Tokyo, and Elsevier, Amsterdam, 1981, Part A, p. 21 1

10 I

G. Kisfaludi, K. LazBr, Z. Schay L. Guczi, Cs. Fetzer, G. Konczos and A. Lovas, Appl. Surf. Sci., 24 (1985) 225

102

S. Engels, Eick, W. Morke, U. Maier, I. Boszornknyi, K. Matusek, Z. Schay and L. Guczi, J. Catal., 103 (1987) 105 L. Guczi, in ”Catalysis 87“ Ed. J. W. Ward, Elsevier, Amsterdam, 1987, p. 8s 2. Schay and L. Guczi, J. Chem. SOC.Faraday I. 78 (1982) 191 1 C. H. Bartholomew, in “Hydrogen Effect in Catalysis”, Eds. Z. Paal and P. G. Menon, Marcel Decker Inc., New York, 1988 p. 543 and references therein

103 104 105 106

107 108

D. G. Castner and D. S. Santilli, in ”Catalytic Materials: Relationship Between Structure and Reactivity“ Eds T. E. White et al. ACS Symposium Series 248, ACS Washington D. C., 1984, p. 39 R. C. Reuel and C. H. Bartholomew, J. Catal., 85 (1984) 78 B. M. Choudary, K. LBzk, K. Matusek and L. Guczi, Chem. Comm. (1988) 592

380

109 110

111 112 113 114 115 116

L. Guczi, G. Stefler, K. Matusek, I. Bogyay, S. Engels, H. Lausch, L. Schuster and M. Wilde, Appl. Catal., 37 (1988) 345 M. Nimz, G. Lietz, J. Volter, K. Lazar and L. Guczi, Catal. Lett., 1 (1988) 93 M. Nimz, G. Lietz, J. Volter, K. Lazar and L. Guczi, Appl. Catal., 45 (1988) 71 B. M. Choudary, K. Matusek I. Bogyay and L. Guczi, J. Catal., 122 (1990) 320 B. M. Choudary, K. LkziU, I. Bogyay and L. Guczi, J. Chem. SOC.Faraday I, (1990) K. Lazkr, M. Nimz, G. Lietz, J. Volter and L. Guczi, Hyperfine Int. 41, (1988) 657 W. D. Mross, Catal. Rev.-Sci. Eng., 25 (1983) 591 J. Benziger, R. J. Madix, Surf. Sci. 94 (1980) 119 D. A. Wesner, F. P. Boenen, H. P. Bonzel, Langmuir, 1 (1985) 478

117 118 D. A. Wesner, G. Linden, H. P. Bonzel, Appl. Surf. Sci., in press 119 K. H. Bailey, T. K. Campbell, J. L. Falconer, Appl. Catal., 54 (1989) 159 120 W. M. H. Sachtler, D. F. Shriver, W. B. Hollenberg, A. F. Lang, J. Catal., 92 (1985) 429 121 J. J. Venter, A. Chen, M. A. Vannice, J. Catal., 117 (1989) 170 122 J. J. Venter, A. A. Chen, J. Phillips, M. A. Vannice, J. Catal., 119 (1989) 451 123 J. L. Rankin. C. H. Bartholomew, J. Catal., 100 (1986) 526 124 H. Arakawa and A. T. Bell, Ind. Eng. Chem. Process Res. Dev., 22 (1983) 97

38 1

CHAPTER 9

CO ACTIVATION BY HOMOGENEOUS CATALYSTS

Michael Roper BASF-AG, D-6700 Ludwigshafen, F.R. Germany

382

9.1 INTRODUCTION Industrial application of CO activation by heterogeneous catalysis has been limited so far to few large-scale processes like methanol synthesis, Fischer-Tropsch synthesis and related reactions, or the water gas shift reaction. As a useful complement, homogeneous CO activation has been established in the chemical industry for the synthesis of a wide range of oxygenated organic products (refs. 1-6). The activation of carbon monoxide by homogeneous transition metal catalysis offers several intrinsic advantages if compared with activation by heterogeneous catalysis. They may be summarized as follows: - High activity is achieved at mild conditions since basically every transition metal atom dissolved in the liquid phase may be available as an active catalytic center. Therefore, in most cases reaction temperatures below 200 OC are applied and often temperatures of around 100 O C are sufficient. -

There are no transport limitations either of educts or products provided the concentration of CO in the liquid phase is adequate and uniform.

-

Homogeneous catalysts are often less sensitive to aging and poisoning since the catalytic centers are mononuclear and there is no surface or ensemble effect. Constant activity and selectivity can be maintained by continuous catalyst replacement by fresh or regenerated catalyst. The major problem of homogeneous catalysis remains the separation of the reaction

products from the catalyst and the recycle of the latter. Whether a homogeneously catalyzed process can be applied industrially or not depends on the successful solution of this problem. The principal routes for CO activation by homogeneous catalysis are summarized in Fig. 9.1. The most difficult is the direct hydrogenation of CO to yield alcohols, polyols, and esters. This reaction has been investigated thoroughly as an alternative source for ethylene glycol, but is hampered by low catalyst activity, severe reaction conditions, and a limited selectivity. In contrast, numerous highly successful industrial syntheses based on homogeneous CO activation were developed for the carbonylation of methanol or its derivatives and of C,, unsaturated organic substrates. The products of these processes range from base chemicals such as acetic acid or butyraldehyde/butanol to fine chemicals and specialties such as fragrances or precursors to antibiotics. While CO activation by heterogeneous catalysis has been investigated since the beginning of this century and has been applied industrially e.g. for methanol synthesis since the early twenties, its homogeneous counterpart remained dormant for the time being. The introduction of CO activation by homogeneous catalysis was the result of the pioneering work of Reppe of BASF and Roelen of Ruhrchemie who during the late 1930s discovered carbonylation and hydrocarbonylation of alkynes and of alkenes, respectively. Promoted by the increasing knowledge of the chemistry of transition metal complexes and especially of transition metal carbonyls, carbonylation and hydrocarbonylation reactions were developed into most versatile and widely industrially used tools to functionalize readily available

383

alkenes

and

alkynes.

The

most

frequently applied reaction today is olefin hydroformylation and the combined capacity of “oxo”-products accounts for about six million tons per

A lcohois PO lY0 i s

Esters

year (refs. 1,3,6,7). Carbonylationhydrocarbonylatio n of C1-products, especially of methanol and its derivatives, has been investigated

czox-

nethanol Or Derivarives

Compounds

since the early fourties. A first industrial breakthrough was the development of an acetic acid synthesis via cobalt catalyzed high pressure methanol carbonylation by BASF around 1960. About ten years later the most successful Monsanto acetic acid process based on rhodium catalyzed low pressure methanol carbonylation was introduced. Today,

Organic Substrates

Carboxyiic Acids

Fig. 9.1 Demonstrated pathways of homogeneously catalyzed CO activation.

this is the leading process for acetic acid synthesis with a still increasing capacity of about two million tons per year that is replacing the conventional oxidation routes via acetaldehyde or hydrocarbons. A related commercial process is the rhodium catalyzed carbonylation of dimethyl ether/methyl acetate to acetic anhydride. Other carbonylation processes of C, products appear viable, such as the syntheses of dimethyl oxalate, glycol aldehyde, glycol ethers, or of acetaldehyde (refs. 8-10). The direct homogeneous hydrogenation of CO in the presence of cobalt salts at 2000 - 5000 bar to yield ethylene glycol was reported first in 1948 by researchers of Du Pont. In the 1970s Pruett from Union Carbide obtained improved combined yields of ethylene glycol, I,?propane diol, and glycerol of up to 70% by use of rhodium catalysts in the presence of N-bases and ionic promoters. However, reaction conditions were still severe (> 1400 bar, > 200)c) and turnover rates were low. Side products were C,-compounds such as methanol or methyl formate (refs. 8,11,12). At milder conditions (< lo00 bar) anionic ruthenium complexes in the presence of iodide promoters have been shown to hydrogenate CO at low rates to mainly methanol and ethanol, along with small amounts of ethylene glycol (ref. 13). Despite the enormous research efforts in the late 1970s and early 1980’s the homogeneously catalyzed direct hydrogenation of CO doesn’t look very promising. If a syngasbased route to ethylene glycol is considered, indirect syntheses involving carbonylation of methanol or of formaldehyde offer far better selectivities and higher turnover rates at milder conditions (ref. 12).

384

It is the aim of this conmbution, to give - after a brief comment on mechanistic impacts of CO activation (section 9.2) - a survey of recent developments in CO hydrogenation and oxidation (sections 9.3 and 9.4), in carbonylation/hydrocarbonylation of C1 compounds (sections 9.5.1.5, 9.5.2.2, and 9.5.2.3), as well as a condensed report on the functionalization of C2+ organic substrates by CO (section 9.5).

9.2 MECHANISTIC IMPLICATIONS OF CO ACTIVATION The mechanism of CO-activation by homogeneous catalysis may be separated into four steps: Coordination of CO Activation of the reagent Conversion of coordinated CO, e.g. by migratory insertion - Product elimination and catalyst regeneration Although reactions such as CO hydrogenation, reductive carbonylation, carbonylation and oxidative carbonylation follow different pathways, there are some common principles which will be outlined with respect to the four steps mentioned above. -

-

9.2.1

COORDINATION OF CO

All transition metals of group VIII have been proven to be active homogeneous catalysts for

CO activation. With the exception of palladium and platinum, they form binary carbonyl complexes. Ligand substituted carbonyl complexes are known for all of them. Formation of carbonyl complexes e.g. by ligand exchange processes as shown in (9.1) is a common feature of transition metals: ML,+l

+

CO

ML,(CO)

+

L

(9.1)

The ease of carbonyl addition as well as the strength of the metal carbonyl bond increase with the electron density at the metal center, or which is equivalent with the metal basicity. They both are favoured by electron rich metals, low metal oxidation states, a low or even negative net charge of the complex, and by basic ligands. This is due to the weak o-donor and the strong Kacceptor properties of CO which removes charge from the metal as is illustrated by the mesomeric structuresofl: 6- 6+ 6+ 6LnM-C%

la -

c--)

LnM=C=O

lb -

-

L,MC-O 1c -

(9.2)

Increasing strength of the metal carbonyl bond corresponds to a decreasing strength of the carbon oxygen bond. Therefore, the frequency of the CO stretching bands in the infrared spectra is a measure for the strength of the metal carbonyl bond. This is shown for the following isoelectronic and isosteric complexes, where the metal carbonyl bond strength decreases in the order 2: > 3 > 4 (refs. 11,14):

385

]

[CO(CO)~ -

2, vco = 1786 cm-I

5,

Ni(C0)4

Vco = 1886 cm-l

-, 4 Vco = 2057 cm-l

Low wave numbers are also observed for bridging carbonyl ligands as in

5 (example:

Fe2(C0)9) or 6 (example: Rh4(CO),9. For neutral molecules, terminal carbonyls absorb in the range of 1850 - 2125 cm-l while bridging ones absorb at about 1700 - 1860 cm-l. Even lower wave numbers are observed for bridging carbonyls in polynuclear anions such as in [Rh5(C0)1~]-. 0

0 I

II

P

/\,/ML:C

n LnM--MLn

/"\ LnM--MLn

-5

6 -

The CO ligand can be further activated by interaction of the oxygen atom with a second metal which acts as a Lewis acid such as in (OC),Mn(Ph2PCH2PPh2)2(p-CO)Mn(C0)2 In these cases CO is a four electron donor providing two z-electrons to one metal and two o-electrons to another (ref. 15). The weakening of the carbon oxygen triple bond of CO via coordination to mono- or multinuclear metal carbonyls is the first step in CO activation. It is therefore no surprise that anionic metal carbonyls such as [Rh5(CO)15]- have been found to be most active in co hydrogenation and that anionic complexes such as [CO(CO)~]-,[Rh(CO)$&, [Ni(C0)3Il-, or [HRu(CO)11]-have been postulated as active species for carbonylation reactions (ref. 16). From a technical point of view it is noteworthy that second and third row transition metal carbonyls, especially if stabilized by further ligands such as phosphines or halides, can be used at much lower CO partial pressures than their first row analogues. This is one reason why noble metal carbonylation catalysts are used increasingly for industrial syntheses, the other being their superior activity and selectivity.

9.2.2 ACTIVATION OF THE REAGENT The following reagents which are typically used in carbonylation will be considered: hydrogen, alkynes, alkenes, alkanols, and alkyl halides. If hydrogen is present in the catalytic system, metal hydrides will form in situ by reaction with metal carbonyls or metal halides or other precursors (refs. 6,14,17):

C O ~ ( C O ) ~+ RhCI(PPh3)3

+

H2

+

H2

2 HCo(CO),,

CO

HRh(PPh3I3CO

(9.3) +

HCI

(9.4)

386

These very common reactions can be understood as oxidative additions of dihydrogen to a metal center followed (9.4) by reductive elimination of a hydrogen halide. Metal hydrides can also be formed by oxidative addition of reagents such as hydrogen halides, carboxylic acids, or alcohols. Thus, hydrogen sources of this type are often important cocatalysts in carbonylation reactions where no molecular hydrogen is present: Ni(C0I4

HX

+

HNi(CO)3X

+

(9.5)

CO

Finally, anionic metal hydrides can be generated by nucleophilic attack of e.g. hydroxide on a carbonyl to generate a hydroxycarbonyl intermediate which eliminates CO,:

Fe(CO)5

+

OH-

[HO-C(0)-Fe(C0)4]-

+

[HO-C(0)-Fe(CO)4]

+

[HFe(CO),,

1-

+

(9.6) C02

(9.7)

The attack of unsaturated substrates such as alkenes, alkadienes or alkynes usually occurs by coordination to a vacant site followed by insertion into a metal hydride bond (ref. 14):

HMLn-l(C=C-R)

+

L

----w

R-C-C-ML,

(9.9)

HML,-,(C=C-R)

+

L

-*

C-C-ML,

(9.10)

I R

As shown in (9.9) and (9.10), the insertion may ensure via a Markownikoff or an antiMarkownikoff mode. Vinyl complexes are thus generated by alkyne insertion, whereas o-ally1 complexes from insertion of 1,2- or 1,3-alkadienes. The latter may isomerize into the more stable nally1 specks:

ML,

2

T---

(9.1 1)

Saturated substrates such as alkanols, esters, ethers, or alkyl halides react usually via an oxidative addition/nucleophilic substitution path (ref. 14):

387 RX

[ML"]-

+

[R-M(x)L,]X

=

-

-b R-ML,

halide, OH, OR, 02CR,

-

[R-M(X)L,]

(9.12)

x-

+

...

(9.13)

Anionic complexes are especially susceptible to this type of reaction and typical examples are [Co(CO)&, [HFe(C0)4]-, or [Rh12(CO),]-. Since alkyl halides are far more reactive than e.g. alcohols, hydrogen halides are often used as promoters for these reagents transforming them in situ into alkyl halides.

9.2.3 COWERSION OF COORDINATED CO The most common conversion of coordinated CO is via migratory insertion yielding acyl complexes (refs. 6,14,17):

(9.14) This reaction has been investigated very thoroughly by Calderazzo for CH,Mn(CO), (ref. 18) and is believed to be the general mechanism for C-C bond formation in carbonylationhydrocarbonylation reactions. Also, this step is often the rate determining one in catalytic cycles and can be enhanced e.g. by increase of CO partial pressure. An alternative way to convert coordinated CO is by intermolecular attack of a base such as hydride, hydroxide, amide, and the like:

L,M-CSO B

=

+

6-

L,M-C-B

H, OH, OR, NH2,

...

1-

(9.15)

This type of CO activation takes place in the homogeneously catalyzed WGS reaction (attack of hydroxide, c.f. Eq. 7) and in the formation of alkyl formates (attack of alkoxide). This is also plausible for oxidative carbonylation to yield oxalates or carbonates, and may be the initial step in CO hydrogenation. Thus, a general method to synthesize formyl complexes is to react metal carbonyls with hydride reagents (ref. 19):

[ HB(OR)3 ] -

6

Fe(C0)5

-

[ H-C(0)-Fe(C0)4 ] -

+

B(OR)3

(9.16) 7 -

Reaction of 1 with protons has been demonstrated to yield methanol and traces of formaldehyde as products of stoichiornetric CO hydrogenation (ref. 14)

388

9.2.4 PRODUCT EHMINATION AND CATALYST REGENERATION In carbonylation reactions, the acyl intermediates are attacked intermolecularly by nucleophiles such as water, alcohols and the like to yield carboxylic acid derivatives along with regeneration of a metal hydride (ref. 6): R-C(O)-ML,

+

H20

R-C(O)-OH

4

H-ML,

+

(9.17)

An intermediate of type 8 may be envisioned, which generates the acid by kind of a elimination: 0-

0-

I

R-C-ML, I OH2+

8

In hydrocarbonylurion reactions the acyl intermediate is thought to be attacked via oxidative addition of dihydrogen followed by reductive elimination of the aldehyde with concomitant formation of an unsaturated metal hydride (ref. 6): R-C(O)-ML,

+

H2

R-C(O)-M(H)&,_1

R-C(O)-M(H)2Ln_1

R-C(0)-H

+

+

H-ML,-I

L

(9.18)

(9.19)

In the case of olefin hydroformylation, this mechanism has been established by means of kinetic investigations. For other reactions such as the hydrogenation of CO attack of a metal hydride followed by product elimination and formation of dinuclear carhnyl complexes may take place, as envisioned in (9.20): H-C(O)-ML,

+

H-ML,

H-C(O)-H

+

M2L2,

(9.20)

Depending on the type of catalyst and on reaction conditions, the aldehydes formed as the primary products may be attacked by metal hydrides and the alkoxy species thus formed are reduced in analogy to reactions (9.18) - (9.20) to the corresponding alcohols (ref. 6): R-C(O)-H

+

H-ML,

-*

R-CH2-O-MLn

(9.21)

In CO reduction, the intermediate formaldehyde can insert by an inverse mode to yield a hydroxymethyl complex, which can undergo subsequent CO insertion to form 9, a key intermediate for ethylene glycol formation (refs. 11,20):

389

H-C(0)-H HO-CH2-MLn

+ HO-CH2-MLn

H-ML,

+

CO

+

(9.22)

+ HO-CH2-C(0)-MLn

(9.23) 9 In oxidative curbonylution the alkoxycarbonyl species formed by reaction (9.15) may react with a nucleophile such as an alcohol followed by reductive elimination to yield carbonates: H i

[L,MC02R]-

+

HOR

b

[Ln-1M-C02R]-

I

OR

10 -

+

L

(9.24)

11 -

It remains uncertain whether intermediates such as 1are actually formed or whether the alkoxycarbonyl ligand in 14is directly attacked by e.g. an alkanol . With palladium catalysts, the reductive elimination from bis(alkoxycarbony1) species yields oxalates (9.26): (9.26)

The metal species formed in reactions (9.25) and (9.26) have to be reoxidized to close the catalytic cycle and to regenerate the active catalyst (ref. 6).

9.3 HOMOGENEOUS HYDROGENATION OF CO From an economical point of view, the direct hydrogenation of CO should yield oxygenated products, where no loss of oxygen via coproduction of valueless water or carbon dioxide has to be accepted. Thus, the desired products have the general formula C,O,+&, and among them are methanol, acetic acid, and especially ethylene glycol. Industrially, the latter product appeared to be most attractive and its apparently simple synthesis according to (9.27) has been the aim of many research efforts (refs. 11,21). 2 CO

+

3 H2

4

HO-CH2-CHz-OH

(9.27)

First reports on the homogeneously catalyzed high pressure hydrogenation of CO date back to 1948, when Gresham et al. Du Pont obtained patents on the cobalt mediated direct synthesis of ethylene glycol from syngas (ref. 22). This reaction was reinvestigated thoroughly in the early 1970's, when Pruett and coworkers from Union Carbide revealed the increased glycol selectivity and space time yields (STY) of rhodium catalysts (ref. 23). Among other group VIlI transition

390

co

I

I I

CO/H2

C2H502CH

co

CH3CH2-OH

CO/H2

HO-CH2CH2-OH

CO/H2

CO/H2

4 CH3CHCH2-OH

HO-CH2-CH-CH2-OH

I

I

OH

OH

Fig. 9.2 Products of homogeneously catalyzed CO hydrogenation

metals especially ruthenium in the presence of certain promoters was found to be active for glycol synthesis. The main products identified in homogeneous CO hydrogenation are summarized in Fig. 9.2. Among them are alcohols, diols and esters. Aldehydes such as formaldehyde, acetaldehyde, or glycolaldehyde have been determined in trace amounts in the form of acetals only. Nevertheless they can be considered as highly likely intermediates which are rapidly converted to more stable products via hydrogenation or hydrocarbonylation. The simplified and generalized catalytic cycles depicted in Figs. 9.3 and 9.4 are used to explain the formation of the observed products. Coordinated CO (c.f. Eq. 1) is attacked by a hydride to generate a formyl species 12. Whether this rate determining step occurs intra- or intermolecularly via a metal hydride formed according to Eq. (9.28), is not known. H2 + 2 ML, --+ 2 HML, (9.28)

16

It is notable that the only example known so far for the direct conversion of a metal hydride with CO into a formyl species (Eq. 29) (ref. 24) is suspected to take place via intermolecular attack of 11on a rhodium carbonyl species:

391

HRh(0EP)

CO

+

__

*

H-C(O)-Rh(OEP)

(9.29)

17 -

The formyl 12 is species to a reduced hydroxymethyl intermediate 11 which plays a key role with respect to

(OEP

=

Octaethylparphyrine)

co HOCH2CH20H

WL"(C0)

product selectivity. Undesired is the elimination

2-h ydroxyacetyl species 14. Obviously, the formation of 14 is favoured by a high co concentration

the

II

12

of

reaction is "COinsertion" into the metal hydroxymethy1 bond to yield the

which is explanation

0

H-C-ML,

an for

observed

increase

of

glycol selectivity with CO/H, pressure. Reduction of 1s.gives the 1,2-

y2"

14 -

Fig. 9.3 Simplified catalytic cycle for methanol and ethylene glycol formation from CO CH30H

CH3-CH-CH2

AH AH r 4 H

:

:

:

L

co

n

I1

0II

CH3-C-ML,

CHJ-CH-C-ML,

L

CH3-CH-MLn

CO

J z y

I

rH

OH

CH3CHz0H

Fig. 9.4 Simplified catalytic cycle for ethanol and formation

1,2-propanediol

392

dihydroxyethyl species Is,and finally ethylene glycol. The formation of higher products such as glycerol can be understood by carbonylation of Is,followed by hydrogenation. The generation of alkyl formates can be explained by secondary reactions involving nucleophilic attack of alkanols or diols on coordinated CO according to Eq. (9.15). Ethanol and 1,2propanediol are formed via hydrocarbonylation of methanol as depicted in Fig. 9.4. In this simplified scheme, methyl species 18is formed via alkylation of the metal hydride 16 by methanol. Further intermediates are the acetyl complex 19 and the 1-hydroxyethyl species B. Reduction of yields ethanol while CO insertion to 2 followed by hydrogenation gives 1,2-

a

propanediol. Again, increase of CO& pressure will favour diol formation. From this short mechanistic introduction it becomes obvious that the synthesis of ethylene glycol by direct CO hydrogenation is a very complex, multistage process, and that high selectivities will be difficult to obtain.

9.3.1 COBALT CATALYSTS AS was mentioned already, the direct synthesis of ethylene glycol by CO hydrogenation was reported first by Gresham from Du Pont (ref. 22). By use of cobalt catalysts in polar solvents such as water or acetic acid at pressures of 2000 - 5000 bar, diols and tnols along with the corresponding esters were generated. Pressures of below 1500 bar led to the preferential formation of methanol or its derivatives.

As a side reaction of the cobalt catalyzed methanol hydrocarbonylation to ethanol, Ziesecke observed traces of 2-alkoxy ethanols which were believed to be generated via the intermediate formaldehyde (ref. 25). More recent results by Keim et al. establish the suitability of unpolar solvents such as npentane for ethylene glycol synthesis (ref. 26). At 230°C and 2000 bar CO/H, (1:l) the following selectivities were obtained: 33 % ethylene glycol, 42 % methyl formate, and 19 % methanol. The yield of polyols could be improved by using higher pressures and catalyst concentrations, while at temperatures of 280°C the formation of undesired C I-products was

Table 9.1

Cobalt catalyzed high pressure CO hydrogenation (ref. 26)a

Catalyst

I

Solvent

~b

1

I

Selectivity, wt% ~

t

~

Methanol + ,Methyl formate

1400 1400 1400 1400 1700

1

I

I

61 77 69 23 39

_

C4H602

1

I

!

rate

_

[, Ethylene glycol

. .

33 4 19 21 46

a 1 mmol C02(CO)8, 10 ml of solvent

initial pressure at 25"C, corresponds to 2000 bar at 230°C

moV1.h __

-

0.84 0.61 0.27 0.27 0.88

393

favoured. The selectivity to ethanol as a methanol hydrocarbonylation product becomes significant at temperatures exceeding 250"C, as is the case upon addition of basic cocatalysts such as phosphines and amines (ref. 27). Some typical results are summarized in Table 9.1. A special advantage of unpolar solvents is the formation of biphasic product mixtures which allow catalyst recycling by simple phase separation. Feder and Rathke reported the formation of primary alcohols and of alkyl formates at pressures as low as 300 bar and temperatures of 2 W C , however at extremely low turnover rates of about 10/week. At these conditions, the rates of CO hydrogenation appear to be accelerated by use of polar solvents, and for the following solvents rates increase by a factor of 20 (refs. 28,291: heptane < benzene < 1,4-dioxan < 1,4-dioxan/water < 2,2,2-tnfluor ethanol Ethylene glycol synthesis has been reported at exceedingly mild conditions such as 130 bar and > 190°C in tetraglyme as the solvent (ref. 30). However, these results have been suspected to stem from solvent hydrolysis. According to investigations by R.B. King, active catalyst precursors for ethylene glycol synthesis must be selected from those compounds which are able to form HCo(C0)4 at reaction conditions [30].

9.3.2 RHODIUM CATALYSTS Since the early 1970s the rhodium catalyzed hydrogenation of CO to yield polyfunctional oxygenated products has been reported in numerous patent applications by Union Carbide. At pressures above 1000 bar and temperatures in the range of 210 - 250°C the favored products are

Table 9.2

Selected results of rhodium catalyzed CO hydrogenation from patents by Union Carbide (ref. 11) ~

Catalysta

T

Solvent

C4H602

rate

---I "C ~~

I

mo1P.h

bar

-~

~

230 tetraglyme

220 220

28403380 12351372 550

49

0.54

26

59

0.94

31

69

0.31

Cs-salt Cs-salt

240 1:l) 260

sulfolane ~~

1030

2.1

550

7.2

~~~

a Rh:Rh(CO);?acac;CS2[Rh,]:CS2[Rh,(CO)~~c]; 2-HP:2-hydroxy-pyridine;

T1PAB:triisopropanolammoniumborate

394

ethylene glycol, methanol, glycerol, and 1,2-propanediol. Minor products are formates, ethanol, and erythritol while the formation of methane is not observed (ref. 23). Some typical examples can be taken from Table 9.2. At optimal conditions ethylene glycol selectivities of up to 70 wt % are obtained. This is in remarkable contrast to processes like the Fischer-Tropsch synthesis, where the yield of C2-products is limited to 30 wt % according to the Schulz-Flory distribution (ref. 31). Anionic rhodium clusters have been claimed by Union Carbide to be responsible for this exceedingly high C2-selectivity. This is based on the fact that large cations such as Cs+ or (Ph3P)2Nf affect both catalytic activity and glycol selectivity. Best results are obtained, if the Rh/Cs ratio is 6:l; if this ratio is 1:1, methanol is formed almost exclusively and the mononuclear anion [Rh(CO)$- can be identified at catalytic conditions. In addition, bases like pyridine or 2hydroxypyridine are used as cocatalysts which favour anion formation by proton abstraction (ref. 23). The role of the polar solvents applied appears to be the stabilization of the anionic clusters at the severe reaction conditions. Good solvents are selected from those with a high dielectricity constant such as sulfolane, NMP, DMI (1,3-dimethyl imidazolidin-2-one), or y-butyrolactone, and from cation solvating systems such as tetraglyme-or crown ethers. Even better are mixtures of solvents from each group (ref. 32). Spectroscopic methods have been used to determine the nature of the anionic rhodium clusters at catalytic or near catalytic conditions. Thus Heaton et al. were able to demonstrate by use of high pressure NMR investigations that at a pressure of 850 bar CO the anion 22 is present almost exclusively which is a key product in rhodium carbonyl cluster chemistry (ref. 33): [Rh12(C0)30]2-

CO

+

~

*

[Rh5(C0)15]-

(9.30) 22 -

Vidal from Union Carbide used high pressure IR spectroscopy to investigate products obtained from Rh(C0)2acac in polar solvents in the presence of a base at 60 - loo0 bar COW2 and at temperatures of up to 210°C. The following cluster anions were identified, and some of them were even isolated and characterized as the cesium or ammonium salts (ref. 34): [Rh(C0)4 [Rh14(C0)25

]

1-

‘-

[Rh5(C0)15

1-

[Rh14(C0)26] 2-

[Rh13 (CO )24H2 ] 3-

[Rhl

1-

[R~G(CO)~~

( CO ) 24H3

[Rh15(C0)27

]

’-

1’-

TO obtain even more stable cluster anions, interstitial systems with a main group element in the cluster core were investigated, and especially the sulfur containing cluster anion 24 was characterized by a remarkable stability. Unfortunately, these systems proved to be less active than those formed in situ from Rh(C0)2acac; their activity decreased in the following order (refs. 3435):

395

23

24

3

Although rhodium catalysts have been demonstrated to allow remarkable selectivities for ethylene glycol, the severe reaction conditions with pressures of z lo00 bar along with the limited catalytic activity prevent so far any industrial application. From a technical point of view the recently reported use of n-alkanols such as n-hexanol as the solvent for mixed Rh/Co catalysts appears to offer some progress. Besides improved ethylene glycol yields this system facilitates product separation by use of simple extraction with water (ref. 36).

9.3.3 RUTHENIUM CATALYSTS Although reports in the literature are ambiguous, it is now well established that unpromoted ruthenium catalysts such as Ru,(CO),, or Ru(acac)j hydrogenate CO exclusively to CI-products such as methanol or methyl formate. This has been demonstrated independently by Keim et al. using toluene or NMP as the solvent at 2000 bar/230"C 1261, by Bradley using THF at 1300 bar/270"C (ref. 37), and by King et al. using 1,4-dioxan at 200 bar /180"C (ref. 30). However, it is possible to influence the catalytic properties of ruthenium to a large extent by promoters and reaction media. Dombek (refs. 38,39) and Knifton (ref. 40) reported on the ruthenium catalyzed synthesis of ethylene glycol diesters along with methyl and ethyl esters by use of aliphatic carboxylic acids as the solvent. Molten salt mixtures have also been demonstrated to be suitable as the solvent. Typical conditions are pressures in the range of 340- 430 bar and temperatures of 220°C. The formation of ethylene glycol diesters appears to be thermodynamically favoured over the direct glycol synthesis and, as an equivalent to the hydroxymethyl species 13 of Fig. 9.3, an acyloxymethyl species 26 has been proposed which undergoes "CO insertion" to give 27: LnRU-CH2-O-C(0)-R

+

CO

+

L,RU-C(O)-CH2-O-C(O)-R

(9.31) 26 27 The direct formation of ethylene glycol along with methanol and ethanol was observed by Dombek by using ruthenium catalysts in the presence of alkali metal iodides at a ratio of I:Ru = 320:l (ref. 41). Polar solvents were used at 850 bar and 230°C and the anionic species 28 and were detected in the catalytic solutions:

[H R u ~ ( C O ) ~1-~ 3

29

]

[Ru(CO)~I~-

29

Best results were obtained if the ratio of the anions 26 and 27 was 2: 1. The rate of ethylene glycol formation was in the range of 40 mol mol-1 h-1 with respect to ruthenium, and ethanol became the favoured product if HI was used instead of KI (ref. 41). The role of the various promoters used in ruthenium catalyzed CO hydrogenation was discussed in detail by Dombek as follows (ref. 16): - Stabilization of anionic carbonyl complexes by providing large cations -

Stabilization of the active center by acting as ligands

396

-

Abstraction of protons by acting as a base

-

Favouring the optimal oxidation state of the active center The initial CO activation is thought to occur via attack of the nucleophilic hydride 26 on a electrophilic CO ligand of 27 to give a formyl species in analogy to Eq. (9.15) (ref. 16). More recently, various approaches have been reported to synthesize C2+-products such as acetic acid (ref. 42) or mixtures of ethanol and propanol (ref. 43) by direct CO hydrogenation. These systems include bifunctional Co/Ru/I catalysts where the methanol produced in the first stage via rutheniumhodide catalyzed CO hydrogenation is converted in a second stage by cobalthodide catalyzed carbonylation/hydrocarbonylation. Methyl iodide is an active intermediate in these systems. However, in the light of an eventual technical realization of these processes, the presence of corrosive iodides under acidic conditions present a further obstacle.

HOMOGENEOUS OXIDATION OF CO

9.4

The base catalyzed reaction of CO with nucleophiles such as alkanols, ammonia, primary or secondary amines yields formic acid derivatives. These syntheses are applied at a large scale for the production of methyl formate and of formamides such as DMF (ref. 44): CO

+

HNU

+

H-C(O)-NU

(9.32)

Recently, transition metal catalysts such as platinurn-ethylpiperidine have been reported for the synthesis of methyl formate at mild conditions (ref. 45). At least formally, the reaction follows an oxidative addition/reductive elimination path: L,M-CEO

+

3

g + c o

CH30H

-

-W 3

+

/C(0)0CH3

(9.33)

LnM\H

fl

H-C(O)-OCH3

(9.34)

The oxidative carbonylation of methanol to yield dimethyl carbonate is catalyzed by copper(1) chloride. The reaction is carried out at > 20 bar at 90 - 1 W C either in a one pot redox system, or in two seperate reduction/oxidation steps (ref. 46):

397

2 CH30H

2 CO

+

+

1/2 02

+ (CH30)2CO

+

(9.35)

H20

A mechanism involving intermolecular methoxide transfer to carbonyl species 3 and methoxycarbonyl species 3 has been proposed (refs. 46,47): CUX2

+ CU(X)-OCH3

CH3OH

+

HX

+

(9.36)

32 (9.37)

CUX

32 -

+

34

+

HX

+

+ 2 CUX

1/11 02

+

(CH30)2CO

+ CUX2

+

(9.38)

(9.39)

1/2 H20

Dimethyl carbonate is a versatile reagent and is increasingly used to replace phosgene and dimethyl sulfate in carbonyl and methyl transfer reactions. It is produced in a 5000 t/a plant by

ENICHEM at Ravenna /Italy since 1983 (ref. 48). The synthesis of dimethyl oxalate (Eq. 40) is also of considerable commercial interest (ref. 49):

2 CH30H

+

2 CO

+

1/2 02

+ CH30C(O)C(O)OCH3

+

H20

(9.40)

The reaction occurs in the presence of palladium compounds as the catalyst, along with suitable promoters. The carbon-carbon coupling step can be envisioned by reductive elimination of two alkoxycarbonyl ligands from a Pd2+ species as outlined in chapter 2.4 (ref. 50):

2-

+

2 c1-

(9.41)

(9.42)

The catalytic cycle involves a redox step where the reduced palladium species are reoxidized with assistance of copper halide salts as is well known from the Wacker process. Typical conditions are 125°C and 70 bar of CO. By addition of orthoformates the byproduct water can be trapped and instead of oxygen quinones may be used as the oxidizing agent.

398

This reaction is applied commercially for dimethyl oxalate synthesis since 1978 by Ube IndustriedJapan by use of an indirect process. In the first stage, methanol is reacted with nitric oxide and oxygen in a distillation reactor to produce anhydrous methyl nitrite along with water which is withdrawn as a water/methanol mixture (ref. 12): 2 CH30H

+

2 NO

1/2 02

+

+

2 CH30NO

+

(9.43)

H20

Nitric acid is formed as the byproduct, and its formation is controlled by adjusting the methanol/nitric oxide/oxygen ratio and reaction conditions. The anhydrous methyl nitrite is carbonylated in the second stage over a supported PdFe catalyst to give dimethyl oxalate in 97 % yield. The coproduct nitric oxide is recycled to the methyl nitrite generator. Byproducts are dimethyl carbonate, methyl formate and methylal (refs. 12,51): 2 CH30NO

+

2 CO

CH30C(O)C(O)OCH3

~

2 NO

+

(9.44)

Dimethyl oxalate is being applied as a solvent, in agriculture (oxalamide), in the pharmaceutical industry, and in food production. With respect to its excellent selectivity, this process has been studied jointly by Ube Industries and Union Carbide for the synthesis of ethylene glycol by the GO process. The hydrogenation of dimethyl oxalate over a ruthenium catalyst gives ethylene glycol in 90 % yield. The coproduct methanol is recycled for methyl nitrite production (ref. 12): CH30C(O)C(O)OCH3

+

4 H2

+ HOCH2CH2OH

+

2 CH30H

(9.45)

The feasibility of the GO process has been demonstrated by Union Carbide in an integrated pilot plant with an ethylene glycol capacity of 5 kg/h (ref. 12). In the light of the mild conditions of the different stages of the GO process and of its remarkable overall yield, it is the most attractive alternative to the current ethylene oxide route and is clearly preferred over direct CO hydrogenation (ref. 12).

9.5 FUNCTIONALIZING REACTIONS OF CO Functionalizing reactions with carbon monoxide to produce oxygenated compounds from unsaturated hydrocarbons belong to the most important industrial applications of homogeneous catalysis. Since the pioneering discoveries of Roelen (Ruhrchemie) and Reppe (BASF) fifty years ago, hydrocarbonylation as well as carbonylation reactions have proven their usefulness in numerous examples. Today, the industrial synthesis of alcohols, aldehydes and carboxylic acids is based in many cases on reactions of carbon monoxide. An important reason for the abundant applications of carbonylation/hydrocarbonylation reactions is their high selectivity which in many cases is superior to e.g. oxidation reactions (refs. 1-7). Substrates for functionalizing reactions by carbon monoxide may be alkynes, alkenes and reactive alkyl or aryl compounds, as is summarized in Eys. (9.46) - (9.48):

399 CO

+

HY

_ j

R-CH=C-C=O

(9.46)

II R’Y R-CH=CH-R’

+

CO

+

HY

-b

(9.47)

R-CH2-CH-C=0

I I

R’ Y RX

X Y

+

= =

CO

(+

HY)

---+

OH, OR, RC02, H a l H, OH, OR, RC02, NR2,

R-C=O

I X(Y)

(+

HX)

(9.48)

Hal

All group VIII transition metals are more or less active homogeneous catalysts for carbonylationhydrocarbonylation react-ions (9.46) - (9.48), but so far only cobalt, nickel, and rhodium have found industrial application. Product selectivity is greatly affected not only by the choice of the proper catalyst metal, but also by the design of the ligand sphere and by the adjustment of reaction conditions such as solvent, pressure, and reaction temperature. Depending on the individual carbonylation/hydrocarbonylation reaction, different problems of selectivity arise. -

Thus, to achieve a high chemoselectivitv, the following side reactions have to be avoided: side reactions of educts, e.g. hydrogenation, isomerization or dehydration; secondary reactions of products, e.g. hydrogenation, aldolization, acetalization or esterization;

In addition, the control of the reeioselectivitv is in many cases an absolute necessity. The hydroformylation of a-olefins leads, for instance, to n-aldehydes or to aldehydes with a 2-methyl branch, if isomerization of the parent olefin is not considered. For many applications only the naldehydes are desired. The addition of a functional group such as a carbonyl to a prochiral olefin generates a center of chirality. If the catalyst is effective in transferring asymmetry, one of the enantiomeric carbonylationhydroformylation products will be formed in excess. Several examples of enantioselective catalvsis have been reported for the hydocarbonylation of functional olefins. The degree of selectivity control which can be achieved will be shown to depend on the substrate, on the catalyst and on the type of carbonylation/hydrocarbonylationto be carried out.

9.5.1

CARBONYLATION

With few exceptions, carbon) lation of organic substrates yields carboxylic acid derivatives. Large scale industrial applications based on carbonylation include the syntheses of acrylic acid

400

from acetylene, of acetic acid from methanol, and of acetic anhydride from methyl ethedmethyl acetate (refs. 3-6). 9.5.1.I Carbonvlation of Alkynes One of the most important applications of “Reppe reactions” is the production of acrylic acid by carbonylation of aqueous organic solutions of acetylene in the presence of catalytic amounts of nickel tetracarbonyl (ref. 3): Ni(C0)4 HCXH

+

H20

+

CO

____j

180

“c/

CH24HC02H

(9.49)

50 bar

The catalytic Reppe reaction is carried out at 160 - 200°C and 45 - 55 bar using an organic solvent such as tetrahydrofuran. A nickel halide such as NiBr2 together with a copper halide as a promoter is used as the catalyst at such low concentrations, that catalyst recovery would be unprofitable. The selectivity for acrylic acid exceeds 90 % based on acetylene and 80 % based on carbon monoxide, with propionic acid as the main byproduct. This process is carried out e.g. by BASF with a capacity of more than 100.000 tons/year of acrylic acid (refs. 1-3). Initially, the Reppe reaction was carried out in the presence of stoichiomeuic amounts of Ni(C0)4 and HC1, the former serving as the source for CO. Therefore mild conditions such as ambient pressure and temperatures as low as 40°C could be applied (ref. 52). Processes using near stoichiometric amounts of Ni(C0)4 have also been used commercially in the USA and in Japan (ref. 1).

In the catalytic process described above, the effectiveness with respect to activity increases in the following order: fluoride < chloride < bromide < iodide In the commercial operation bromide is usually preferred due to corrosion problems associated with HI (ref. 2). If nickel is added in the form of nickel(I1) halide, the corresponding hydrogen halide is generated during formation of nickel tetracarbonyl. By oxidative addition of HX a nickel(I1) hydride is formed which is believed to be the active catalytic species: (9.50)

401 Ni(C0)4

-P

HX

+

NiH(X)(C0)2

+

2 CO

(9.5 1)

Addition of acetylene to the hydride intermediate generates a vinyl complex which after insertion of CO and nucleophilic attack of water releases acrylic acid with regeneration of the initial hydride intermediate. The nucleophile water can be replaced by alcohols, amines, carboxylic acids, hydrogen halides or mercaptans yielding the corresponding acrylic esters, amides, anhydrides, halides or thioesters. If acetylene is replaced by higher alkynes, the preferred mode of insertion into the metalhydrogen bond is of the Markovnikov type. Furthermore, the addition of H-C02H to the triple bond generally takes place in a cis manner and isomerization of the alkyne is not observed. Thus, with methyl acetylene and aqueous methanol as the reaction medium the major product (>80%) is methyl methacrylate with methyl crotonate as the major byproduct (refs. 52,531: CH3'CH

CO

+

+

CH30H

*

-

CH2=C-C02CH3

I

(9.52)

CH3

With monosubstituted alkynes the regioselectivity of carboxylation is greatly affected by the nature of the substituent (ref. 53): A-C'CH

A

=

+

CO

+

H20

alkyl, a w l , CHRC02CH3,

B-CrCH

+

CO

+

H20

-

(CH2),0H,

-

A-C=CH2

I COP

n

=

(9.53)

1-3

B-C=CH

II

(9.54)

H C02H

6

=

H, CRHOH, CRZOH, C02H, C02CH3, COCH3,

CH$(CH3)20H

If disubstituted alkynes are carboxylated, fast reaction is observed if both substituents are of type A; with one substituent of type A and the other of type B rates are still reasonably. With both substituents of type B only very low yields are obtained. The main exception to this rule is acetylene which is very reactive (ref. 53). Sustituent properties also affect the regioselectivity of carboxylation of phenylarylalkynes. Para-substituents with -I or + -M properties direct the CO addition to the acetylenic carbon adjacent to phenyl. If para-substituents with +I or +M properties or ortho-substituents are present, carboxylation occurs at the acetylenic carbon next to aryl (ref. 3).

402

R-CGH4-CX-Ph

CO

+

+

H20

-L

(9.55)

Ph-C=C-CbHq-R

I 1

+ I or +M or o-R) For alkyne carbonylation most work has concentrated on nickel catalysts. However, other catalyst metals are also active in alkyne carbonylation. Cobalt is less active and tends to promote double carbonylation yielding succinic acid derivatives. Alkyl acrylate formation can be enhanced by using low acetylene and catalyst concentrations with the alcohol as the solvent at 1 1OoC/210 bar (ref. 54). A high degree of chemoselectivity can be achieved with palladium catalysts. Thus, PdBr,[P(OPh),], in the presence of perchloric acid was reported to be an active catalyst for the H C02H

(P-R:

methoxycarbonylation of acetylene to give at mild conditions methyl acrylate with a selectivity of 95 9% (ref. 55). On the other hand, palladium chloride in the presence of thiourea and controlled amounts of oxygen catalyzes the formation of dimethyl maleate with 90 % selectivity at ambient temperature and pressure (ref. 56):

H2C=CH-C02Me HCXH

+

CO

+

MeOH

(9.56) - H20

HC-C02Me

There are numerous examples of catalytic and stoichiometric carbonylations of alkynes in the literature which are often useful synthetic tools e.g. for cyclization reactions (refs. 2-4,54).

9.5.1.2 Carbonvlation ofAlkenes The hydrocarboxylation of alkenes yields mixtures of saturated acids:

R-CH2CH2-CO2H R-CH=CH2

+

CO

+

H20 R-CH-C02H

(9.57)

I

CH3

Compared to the carbonylation of alkynes, higher temperatures and pressures have to be applied, and nickel as well as cobalt, ruthenium, rhodium, palladium and platinum may be used as

403

the catalyst. Selectivities are in general lower due R-CH=CH2

to the ready double bond isomerization of the olefin

by the substrates carbonylation catalysts (refs. 3-6,57). A general mechanism of alkene carbonylation olefin formation

R-CH2-CH2-ML,

k1

I H-NL,

involving

co

coordination, of a metal

0 I/

R-CH2-CH2-C-NLn

alkyl, CO insertion, and attack of the nucleophile is given in Fig. 9.5: Among the catalysts mentioned above, nickel is the least active isomerization catalyst and is preferred,

I1

R-CH2-CH2-C-Y

Fig. 9.5 General mechanism for alkene carbonylation (ML, = CO(CO)~, Ni(CO)$, ...; Y = OH, OR, NR2, ...)

where addition of CO at the initial position of the

double bond is desired. With a-olefins, a 40:60 ratio of linear and 2-methy1 branched carboxylic acids is obtained (ref. 58). If linear or terminal acids are to be produced from internal olefins, cobalt is the catalyst of choice which regardless to the initial position of the double bond of the substrate yields the products derived from the corresponding 1-olefins. The fraction of linear/teminal acids amounts to 55 - 60 % with unmodified and to over 80 % with pyridine or picoline modified cobalt catalysts. Typical reaction conditions are 150 - 250°C and 150 - 200 bar (refs. 5839):

c,-c-c-c

cX -c=c-Cy

+ CO +

I

H20

C02H

(9.58)

co/w Cx-C-C-Cy-C02H

Noble metal carbonylation catalysts are active at temperatures below 140°C and especially rhodium and iridium allow low pressure conditions (1 bar and higher) (refs. 5,60-63).

404

The nickel catalyzed hydrocarboxylation of ethylene according to Eq. 9.57 gives propionic acid in a 95 % yield, A process operated by BASF uses nickel propionate at 200 - 240 bar and 270 - 320'C in a silver-lined reactor (refs. 5,6). With increasing degree of branching and substitution of the olefin substrate its reactivity decreases, and tetraalkylethylenes are inactive in Reppe carbonylations. The reactivity is in many cases enhanced by additives (hydrogen, water, or acids), which promote the formation of transition metal carbonyl hydrides that are generally thought to be the active species. In analogy to alkyne carbonylation the nucleophile water can be replaced by other nucleophiles such as alcohols, amines, thioalcohols or acids which yield the corresponding carboxylic acid derivatives, but are less reactive. An impressing example for control of regioselectivity by selection/mdification of the catalyst is the hydrocarboxylation of propene, which according to Eq.9.57 yields butyric and isobutyric acid (refs. 3,5,64): Table 9.3

Control of regioselectivity in propene hydrocarboxylation

I

I Catalyst

Isobutyric acid

I

Butyncacid

1

Olefins may also undergo oxidative carbonylation reactions in the presence of PdCl2/CuC12 catalysts. Thus, ethylene can be carbonylated in acetic acidacetic anhydride to give acrylic acid with 0-acetoxypropionic acid as the by-product: CHz=CHZ

+

CO

+

1/2 02

-

CH2=CH-C02H

(9.59)

P-Acetoxypropionic acid can be converted to acrylic acid by thermal cracking (refs. 3,5,65). The carbonylation chemistry of olefins by use of palladium catalysts has been reviewed comprehensively (refs. 62,63). Numerous olefins substituted by aryl or functional groups have been carbonylated and depending on reaction conditions and substrates cyclization reactions to yield e.g. lactones or lactams take place (refs. 3,557).

9.5.1.3 Carbondation of Alkadienes The carbonylation of 1,3-butadiene can be carried out in the presence of cobalt and palladium catalysts, while nickel shows inferior activity (ref. 5). The alkoxycarbonylation of butadiene can be directed with palladium catalysts in a highly chemo- and regioselective process to yield either 3-pentenoic or 3,8-nonadienoic acid esters:

405

I C=C-C-C-C-C=C-C-C02Me *

The nature of the anion determines which product is formed. In the presence of acetate butadiene dimerizes prior to alkoxycarbonylation while chloride effectively inhibits dimerization. The regioselectivity of this process is due to intermediate n-ally1 complexes, which predetermine both product linearity and the position of double bonds (ref. 66). The reaction is carried out at 100°C and 150 bar and is accelerated by addition of nitrogen bases, tetraalkylammonium, or -phosphonium salts. However, a serious drawback of these palladium systems is the loss of their initially high activity and the limited catalyst life, which has prevented so far their industrial application (ref. 5). Compared to alkoxycarbonylation, the hydrocarboxylation of butadiene according to Eq. (9.61) requires higher temperatures. C=C-C=C

+

CO

+

H20

-*

C-C=C-C-C02H

(9.61)

CobaWpyridine catalysts are used for this reaction at > 200 bar/120-16OSC and the regioselectivity improves as the amount of pyridine (picoline, isoquinoline, etc) increases. Yields of up to 95 7L have been reported. 3-Pentenoic acid is also obtained with palladium/tetraalkylammonium salt as the catalyst while rhodium yields a mixture of 3- and 4pentenoic acid, with the latter predominating (refs. 3 3 ) . The reactivity of 1,3-alkadienes in the cobalt/pyridine catalyzed hydrocarboxylation decreases with steric hindrance in the following order (ref. 3): 1,3-butadiene, isoprene > 1,3-pentadiene > 2,4-hexadiene > 2,3-dimethyl-1,3-butadiene The direct synthesis of adipic acid by double hydrocarboxylation of 1,3-butadiene has been studied repeatedly by use of cobalt and rhodium catalysts (Eq. 62). C=C-C=C

+

2 CO

+

2 H20

----+ H02C-C-C-C-C-CO2H

(9.62)

Unfortunately, the reaction requires fairly high temperatures (200 - 250"C/250 - 300 bar) which results in low selectivities (refs. 3,67,68). These problems have been circumvented by an indirect three-stage process for adipic acid from butadiene containing streams, which has been developed by BASF (ref. 5). In the first stage, butadiene is methoxycarbonylated at enhanced concentrations of both cobalt and pyridine to yield methyl-3-pentenoate:

406

C=C-C=C

120 - 140 "C

CO + MeOH

+

+

(9.63)

C-C=C-C-C02Me

300 - 1000 bar In the second stage, methoxycarbonylation is carried out at reduced pyridine concentration, increased temperatures, and lower pressures. These conditions facilitate the isomenzation of 3pentenoic ester into 4-pentenoic ester, which is the prerequisite for dimethyl adipate formation. Finally, hydrolysis of the ester gives adipic acid in an overall yield of about 70 % (ref. 5).

C-C=C-C-C02Me

+

CO

Me02-C-C-C-C-C02Me

+

MeOH

+

150

-

170 "C

150 - 200 bar

2 H20

+

.*

Me02C-C-C-C-C-C02Me

H02C-C-C-C-C-C02H

+

2 MeOH

(9.55)

Oxidative carbonylation of 1,3-butadiene has been reported to take place with palladium catalysts, and depending on conditions and catalyst composition single or double carbonylation occurs (ref. 5). Pd12/Cu12/02/enol

ether C=C-C=C-C02R

CqHg

CO

+

ROH

+

(9.66) R02C-C-C=C-C-C02R 100 "C, 125 bar

The role of Cu2+ is to reoxidize Pdo which is formed in the reaction, and enol ethers as well as ketals are added to bind the reaction water, which is detrimental to the catalyst. Despite of these precautions, catalyst life is short (ref. 5). 1,2-Alkadienes have also been carbonylated. Thus, the methoxycarbonylation of allene in the presence of ruthenium gives at high pressures methyl methylacrylate or dimethyl 1,l-dimethyl3-methyleneglutarate in moderate yields (refs. 5,69): 140 "C

C=C=C

+

CO

+

MeOH

(9.67)

200 "C Me02C-C-C-C-C02Me 800

-

1000 bar

I

C

The cobalt catalyzed carbonylation of tetraphenyl allene has been reported to give mixtures of cyclization products, such as indene, indone, and naphthalenone derivatives.

407

The carbonylation of nonconjugated alkadienes in the presence of HCo(C0)4 gives unsaturated and saturated ketones; best yields are obtained with dienes, where the double bonds are separated by one or two carbon atoms (refs. 33):

HCo(C0) 4

c=c-c-c-c=c

co

+

c-c

c-c I

I

c-c

+

\/

F-li

c-c

\/"-c

(9.68)

t

C

I

0

0

Nonconjugated alkadienes behave like monoolefins in hydrocarboxylation or alkoxycarbonylation, if palladium catalysts are applied. For example, the two double bonds of 1,5cyclooctadiene are alkoxycarbonylated independently and successively with a Pd/HCI catalyst (ref. 70).

The carbonylation of alkanes has been achieved only recently by Tanaka et al. by irradiating homogeneous solutions of Rh(PMe&CICO at 1 bar of CO and ambient temperature (ref. 71):

c-c-c-c-c

+

CO

hv

cat.

C-C-C-C-C-CHO

+

C-C-C-C-C

I

(9.69)

CHO

) 98 %

( 2 %

The same authors have also reported on the carbonylation of benzene to give benzaldehyde. 9.5.1.5 Carbonvlation of Alkanols. Esters. and Ethers From an industrial point of view, alcohols are less abundant than the corresponding olefins and are less attractive feedstocks for carbonylation reactions. Exceptions are benzyl alkohols and especially methanol, which is one of the cheapest organic bulk chemicals and which can be obtained easily from synthesis gas. It is therefor no surprise, that in alkohol carbonvlation above all methanol and its derivatives have been investigated (refs. 3,5,9,10). The synthesis of acetic acid by carbonylation of methanol is one of the most important industrial applications of homogeneous catalysis (Eq. 70). MeOH

+

CO

-

MeC02H

(9.70)

Two processes have been developed and are applied at capacities of > 2 million tons/year: The BASF process using a cobalr/iodine catalyst and the Monsanto process using a rhodiudiodine

408

catalyst. As is evident from Table 9.4, the Monsanto process operates at milder conditions and achieves a very high selectivity: In the C012 catalyzed process, HCo(C0)4 is believed to be the actual catalyst, which is methylated by methyl iodide formed from methanol and HI. Byproduct formation is due to the I/ CH3-C-I

Fig. 9.6 Simplified mechanism for methanol carbonylation (ML, = Rh(C0)212, Co(CO)4, etc.)

hydrogenating/hydrocarbony lating properties of HCO(CO)~which converts by the water gas shift reaction some of the CO into hydrogen. Thus, methane can be formed by hydrogenation of CO and acetaldehyde/ethanol by

hydrocarbonylation of methanol (refs. 72,73).

In the rhodiudiodide system anionic complexes of the type [Rh(C0)212]- have been proposed as the active species which again are methylated by methyl iodide. The iodide ligands not only stabilize rhodium complexes at low pressures of CO. They also inhibit effectively the hydrogena-ting/hydrocarbonylating pro-perties of rhodium and thus the type of side reactions occumng with cobalt (refs. 74,75). Most methanol carbonylation plants now use the Monsanto technology, although catalyst recycling is complex and small continuous losses of iodine may occur. Alternative processes based on nickel/iodide catalysts have been investigated by different companies (ref. 10).

409

A general mechanism for methanol carbonylation involving formation of methyl iodide from methanol and HI, methylation of an anionic metal carbonyl via oxidative addition, CO insertion, and reductive elimination of acetyl iodide is given in Fig. 9.6. Acetyl iodide is hydrolized to give acetic acid and HI: Closely related to methanol carbonylation is the isomerization of methyl formate (Eq. 7 1): HC02Me

-

(9.7 1)

MeC02H

This process is catalyzed by the same systems as used for methanol carbonylation and might be industrially attractive at production sites, where no CO is available (ref. 44). A number of higher alcohols have been carbonylated using mainly Rh/l or Ni/I catalysts. While primary and secondary alkyl alcohols are reported to give good yields of the corresponding carboxylic acid, tertiary alcohols tend to be dehydrated. Diols may also be carbonylated to yield at reduced selectivities linear and branched dicarboxylic acids along with esters, lactones or hydroxycarboxylic acids (refs. 5,76). Of some interest in the field of fine chemicals is the carbonylation of benzylalcohol, which produces phenylacetic acid: Rh/I2 PhCH20H

+

*

CO

PhCH2C02H

(9.72)

70 bar/175 "C

With a rhodiudiodide catalyst yields of 94 5% have been achieved at mild conditions (ref. 77). Ally1 alcohol has been carbonylated by using palladium catalysts. While the alkoxycarbonylation gives 3-butenoic acid esters (ref. 78), ally1 3-butenoate is obtained in 88 % yield in the absence of additional alcohols (ref. 79): PdC12/PPhj/SnC 12

2 C=C-C-OH

+

CO 80 "C/200 bar

- c=c-c-co~-c-c=c

(9.73)

The carbonylation of ethers to give esters and of esters to yield anhydrides can be achieved by use of e.g. cobalt, nickel, and especially rhodium catalysts in the presence of iodide and other promoters. Most interest has found the synthesis of acetic anhydride from dimethyl ether/methyl acetate (refs. 10,80).

410

Me-0-Me

-

CO

+

-

0

co

II

Me-C-0-Me

0

II

0

II

Me-C-0-C-0-Me

(9.74)

As the catalyst, a rhodium/iodide system is used together with promoters such as phosphines and early transition metal compounds like Cr(C0)6. Typical conditions are 10 - 100 bar and 150 - 200°C and, at conversions of 50 - 80 8, selectivities to acetic anhydride of up to 90 8 can be achieved. The process has been developed by Halcon and commercialized by Eastman Kodak in 1983 in a plant with a design capacity of 230 OOO tons/year (refs. 9,81). If cycloaliphatic ethers are carbonylated, lactones, hydroxyacids or dicarboxylic acids are obtained, depending on substrates, coreagents and the catalyst used, as is exemplified in Eqs. (9.75)- (9.78)(refs. 76,82,83):

-I HCO(CO)Q,

R-CH-CH2

+

CO

EtOH

+

// =

R-CH-CHzC02Et

(9.75)

1 bar, 0 "C

0

R

I2

OH

(50 % )

a lk y l

CH3-CH-CH2

CO

+

\/ 0

CO2(CO)p H20 -400 bar, 160 "C

CH3-CH=CH-C02H

(9.76)

(81 %)

CO( OAC 12

(9.77) 200 "C,

250 bar

Ni(COI4,

I(CH2I4I

CH2 /c=o \/ 0 (55 X )

CH2-CH2

I 1 CH CH \2/ 0

H02C(CH2)4C02H +

CO

+

+

*

H20 250 "C,

60 bar

(9.78)

branched isomers

(L74 %)

The carbonylation of styrene oxide can be directed either to yield hydroxyesters or, by double carbonylation under phase transfer conditions in the presence of excess of methyl iodide, to give directly 4,5-dihydro-4-phenylfuran-2,3-dione (refs. 84,881:

411

Ph-CH-CH2

//

+

0

co

CO~(CO)~/K~CO~

Ph-CH-CH20H

EtOH, 1 bar, 30 "C

C02Et

I

c02(co)8,

MeI, CgHg

I

~

(72 %)

(9.79)

Ph-C-CH2

aq. NaOH, CTAB, 1 bar, 25 "C

n 0i

HO-C

\/ C II 0

In addition, styrene oxide has been carbonylated in the presence of RhCl(CO)(PPh& as the catalyst to give a-phenyl-P-propiolactonein a yield of 67 % (ref. 86). 9.5.1.6 Carbonvlation of Orpanic Halides From an industrial point of view, organic halides are less attractive feedstocks for carbonylation if compared with the corresponding alkenes or alcohols. Their use will be resmcted to the synthesis of fine chemicals where the formation of halide salts as unavoidable coproducts either in the carbonylation step or in one of the following steps can be tolerated. The carbonylation of saturated as well as of unsaturated organic halides yields carboxylic

acid halides. If water, alcohols, or amines are present, the corresponding acids, esters, or amides are formed. In most cases at least stoichiometric amounts of a base such as NaOH are added to neutralize the hydrogen halide generated. By this the driving force of organic halide carbonylation is enhanced and very mild reaction conditions may be applied, especially under phase transfer conditions (refs. 3-5,84-85,87). Typical catalysts include nickel carbonyl, cobalt carbonyl as well as complexes of palladium and rhodium. Aryl halides can be carbonylated by use of Ni(CO), in the presence of stoichiometric amounts of bases like Ca(OH)2 to yield the corresponding salts of the aromatic carboxylic acids. The reaction takes place at mild conditions in polar aprotic solvents such as DMF or DMSO (Eq. 80) (ref. 5 ) :

ArX

+

CO

+

Ca(OHI2

Ni(COIq, DMF 100 "C, 1 bar

*

CaX [ArC02]

+

H20

(9.80)

In less polar solvents higher temperatures and pressures are required, and in THF b e n d is obtained in 80 % yield. Aryl halides are alkoxycarbonylated under mild conditions in the presence of tertiary amines by use of palladium triphenylphosphine complexes (ref. 88):

412

ArX

CO

+

+

ROH

NR'3

+

+ ArC02R

+

(9.81)

NRi3HX

Also vinvl halides can be alkoxycarbonylated in the same manner by the same catalytic system. However, with unmodified rhodium as the catalyst, alkoxycarbonylation of the olefinic double bond takes place (refs. 79,89). Rh R-CH=CHX

+

CO

+

(9.82)

R'OH R-CH=CH-C02R'

+

NR"3HX

NR"3

Allvl halides or acetates can be alkoxycarbonylated under mild conditions in the presence of nickel catalysts. This reaction may be coupled with a two carbon homologation by addition ethylene or with vinylation by addition of acetylene. Thus, allyl halide carbonylation can be directed to yield alkyl 3-butenoates (ref. 90), alkyl5-hexenoates (ref. 91), or alkyl sorbates (ref. 92):

CH2=CH-CH2Cl

+

CO

+

ROH

FL

CH2=CH-CH2-C02R CH2=CH(CH2)$02R

(9.83)

HCXH

CH3(CH=CH)2C02R

In this reaction, ethylene may be replaced by other olefins or dienes, and other catalysts such as palladium can be applied (ref. 5). If the carbonylation of allyl chloride in the presence of acetylene is carried in acetone/water, cyclopentenonacetic acid is obtained by double carbonylation (ref. 93): CH+H-CH2Cl

+

HCECH

+

2 CO

+

H20

Ni

- HCl

CH-CH2

II

I

CH CH-CH2-CO2H

\C/

(9.84)

..

II

0

This reaction principle can be varied to synthesize a variety of cyclopentane and cyclohexane derivatives (ref. 93). The carbonylation of non activated alkvl halides is generally more difficult than that of aryl halides and is usually carried out in polar solvents such as ethanol or dimethyl formamide (ref. 94). Activated alkyl halides such as chloroacetonitrile, alkyl chloroacetates or benzyl chloride can be carbonylated in high yields by use of iron, cobalt, nickel, rhodium or palladium catalysts. Thus, industrial syntheses of malonic, phenylacetic and phenylpyruvic acid esters are based on the alkoxycarbonylation of activated alkyl halides (Eq. 85, 86) (ref. 94):

413

Co2( CO)8 C1-CH2-CO2Et

CO

+

+

NaOEt

(9.85)

* Et02C-CH2-C02Et

+

NaCl

EtOH (97 %)

MeOH/NaOMe Ph-CH2-C02Me Ph-CH2-Cl

+

CO

[

+

NaCl

(95 X ) 0 II Ph-CH2-C-C02] CaCl

CO~(CO)~

(9.86)

(82 %)

In all cases C O ~ ( C Ois) ~used as the catalyst at mild conditions (50 - 60'c, < 60 bar), and apparently the type of base as well as the solvent system controls the selectivity with respect to single or double carbonylation. For the latter process, enolization of the intermediate phenylacetylcobalt carbonyl to give a vinyl cobalt complex has been proposed (ref. 84). Even secondary benzyl halides can be double carbonylated at selected conditions. Exceedingly mild conditions such as room temperature and atmospheric pressure may be applied, if these carbonylations are carried out under phase transfer conditions (ref. 87). Nonactivated alkyl halides can also undergo cobalt-catalyzed dicarbonylation, but rather drastic conditions are required (ref. 95).

9.5.2 HYDROCARBONYLATION While hydrocarbonylation of alkynes to yield e.g. hydroquinone or cyclopentenone derivatives (ref. 93) has so far found only limited interest, hydocarbonylation of alkenes is used widely in the chemical industry for large scale syntheses of aldehydes and alcohols. With a combined capacity of about 6 million tons per year the 0 x 0 synthesis is the most important application of homogeneous catalysis (refs. 1-7,96). 9.5.2. I Hvdrocarbonvlation of Alkenes The cobalt catalyzed hydrocarbonylation, hydroformylation or 0x0 reaction of alkenes to yield aldehydes has been discovered fifty years ago by Roelen of Ruhrchemie (ref. 97):

Co2( CO) 8 R-CH=CHz

+

CO

+

* R-CH2-CH2-CHO

H2

+

110-180 " C 200-300 bar

With respect to selectivity, the following problems have to be considered: - side reactions of olefins: double bond isomerization, hydrogenation

R-CH-CH3 (9.87)

CHO

414

Table 9.5

Typical reaction conditions and selectivity control of 0x0 processes (substrate: propene)

Catalyst

Co/PR3

M:olefin

LHSV I

110-180 200-300 10-3-10-2 0.5-2.0

160-200 50-100 6.10-3 0.1-0.2

aldehydes

alcohols high 88:12

I

Products CjHg yield n:iso

70:30

i

I

Rh/PR3

I

60- 120 1-50 10-4-10-3 0.1-0.25

'

Rh 100-140 100-140 10-6-10-4 0.3-0.6 aldehydes low 5050

aldehydes variable 92:8

-

regioselectivity of CO-addition (e.g. n-aldehydes or 2-methylaldehydes from 1-alkenes) secondary reactions of aldehydes: hydrogenation to alcohols, formation of high boiling residues via aldolization. Cobalt and rhodium catalysts which may be modified by phosphine ligands are used in industrial 0x0 syntheses, the active species being hydride complexes like the following (ref. 3): -

HCO (CO 1

,

HCO(CO)?BU3

HRh(C0) 4

HRh(CO),(PPh3),

A general mechanism involving olefin coordination, metal alkyl formation, CO insertion,

oxidative addition of and dihydrogen , reductive elimination of aldehyde is given in Fig. 9.7 (ref. 98): Based on these types of catalysts, Oxo-

R-CH=CH2

II

R-CH2-CH2-C-MLn

0

II

R-CH2-CH2-C-M(H)2L,

6 0 I1 R-CHyCH2-C-H

Fig. 9.7 Simplified mechanism for alkene hydrocarbonylation (ML, = Co(COI3, Co(CO)2PR3, Rh(CO)3, Rh(CO)(PR3)2, etc)

processes have been developed, which vary with respect to reaction conditions and which allow to achieve a high degree of selectivity control (ref. 3): Unmodified cobalt catalysts offer a high reactivity at a n/iso-ratio of 70:30. As they catalyze double bond isomerization of olefins, cobalt catalysts

415 produce n-aldehydes from internal olefins with almost the same n/i-ratio than from a-olefins:

cn -c=c-c,

+

co

+

(9.88)

C,-C-C-C,-CHO

H2

C,-C-C-C,-CHzOH H2 Ligand modified cobalt catalysts give even a higher yield of the n-isomer, which is transformed to the alcohol due to the high hydrogenation activity of the system, that, unfortunately, also results in an increased alkane formation. C o n catalysts are more stable than Co catalysts and are operated at lower pressures but at higher temperatures due to the reduced activity (ref. 99). Unmodified cobalt catalysts also yield alcohols as the main products, if the reaction temperature is increased sufficiently. Unmodified as well as ligand modified cobalt catalysts are preferred in syntheses of long chain (C&ls) aldehydes/alcohols, which are used to produce plasticizers and detergents. Unmodified rhodium catalysts show little activity in double bond isomerization of olefins, which is completely suppressed upon addition of ligands. Due to their very high selectivity to aldehydes and to their superior activity, unmodified rhodium catalysts are used for the hydroformylation of internal or of cyclic olefins to give branched aldehydes: (9.89)

CHO

CHO

The discovery that rhodium phosphine complexes allow considerably milder reaction conditions and thus yield mainly n-aldehydes from linear I-olefins (refs. 100-102) has led to the development of the Rh-low pressure 0 x 0 process (ref. 103): Rh/PPhj C,-C=C

+

CO

+

H2

C,-C-C-CHO

(9.90)

If the molar ratio of PPh3:Rh is in the range of 50-300, a stable and highly active catalyst system is obtained, which can be operated at the mild temperatures and pressures mentioned in Table 9.5. From linear 1-alkenes linear aldehydes are obtained in high yield along with minor amounts of 2-methyl aldehydes at a n/i-ratio of about 9, and side or secondary reactions like formation of alkanes, of alcohols, or of aldolization products are largely reduced (refs. 103,104). which This process is especially applied for the production of low boiling aldehydes (
416

in water, a somewhat higher pressure appears to be necessary than in the single phase process (refs. 105,106). Virtually all olefins can be submitted to the 0 x 0 reaction. However, both reactivity and regioselectivity are influenced by the structure of the alkene substrate (ref. 3). With $mule olefins (CnH2,,), the reactivity decreases with increasing substitution at the double bond and the following sequence of reaction rates has been reported with unmodified rhodium catalysts (ref. 3): linear a-olefins > linear internal olefins > mono-branched olefins > multi-branched olefins With the exception of ethylene and unsubstituted cyclic alkenes, the 0x0 reaction always yields mixtures of isomeric aldehydes. According to the rule of Keulemans (ref. 107), the formyl group is added preferentially to the least substituted carbon, and hardly any quartemy carbon is formed. For example, isobutene yields almost exclusively 3-methylbutanal and only about 5 % of pivalaldehyde; even branching at the carbon atom next to the double bond has a profound impact on regioselectivity (ref. 3): C I C-C=C

CO~(CO)~

CO

+

H2

+

C I C-C-C-CHO

C

I

c-c-c=c I

C

+

co

+

H2

CO~(CO)~

C

C I C-C-C-C-CHO

I

c

(9.91)

I

+

(99 %)

C-C-CHO

+

C I C-C-C-CHO

I I

cc

(9.92)

( 1 %)

Alkadienes with isolated double bonds can be hydroformylated successively to give the corresponding enals and dials, and yields improve with growing distance between the double bonds. Conkgated alkadienes are hydroformylated slowly in the presence of unmodified CO or Rh catalysts yielding exclusively saturated linear and branched monoaldehydes or -alcohols, and the low reactivity is to be atmbuted to x-ally1 intermediates. With phosphine modified Rh catalysts also dialdehydes are obtained (ratio monoakdial = 52:48), among which the branched products 2-methyl glutaraldehyde and 3-methyl glutaraldehyde prevail (ref. 108).

417

The hydroformylation of functionallv substituted olefins is a route to obtain interesting bifunctional products and has been studied in detail in recent years (ref. 109). From the many examples available, only a few, offering further insight into selectivity control, will be mentioned. The hydroformylation of allyl alcohol and allyl ethers is of interest as an alternative route for 1,4-butanediol. With phophine modified rhodium catalysts the linear isomer is obtained at a regioselectivity of about 70 % (allyl alcohol) to 80 % (allyl t-butyl ether), in analogy to nonfunctional olefins (refs. 110,111). C=C-C-OR

+

CO

+

H2

-

(9.93)

OHC-C-C-C-OR

In contrast, vinyl ethers and esters are preferentially formylated at the carbon next to oxygen: C=C-OR

+

CO

+

H2

-

(9.94)

C-C-OR

(R = alkyl, Ac, . . . I

I

CHO

Rhodium catalysts usually give better yields in these reactions, which may be of interest as alternative routes to 1,2-propanediol derivatives (refs. 112,113). Unsaturated halo compounds with limited mobility such as vinyl chloride or fluorinated olefins can be hydroformylated without particular problems under the usual conditions, and e.g. vinyl chloride gives 2-chloro propanal in 90 % yield (ref. 114). With fluorinated alkenes, high yields and almost complete control of regioselectivity can be achieved by varying the catalyst metal (Eq. 95) (refs. 115,116):

;:3 H2

co

Rf-;-C-CHO

a: 93 % b: 80 %

R p C 4

+

CO

(9.95)

+

a: 96 % Rf-C-CHO

b: 97 %

This surprising degree of selectivity control can be understood by regarding the polarity of the double bond of the olefin and considering the well known higher acidic character of cobalt carbonyl hydride compared to rhodium hydrides:

418

(9.96)

The polarity of Pt-H and Ru-H can be placed between those of rhodium and cobalt hydrides; accordingly only a low regioselectivity is obtained with Pt and Ru catalysts. There are several further examples for this type of selectivity control, of which the hydroformylation of a$-unsaturated nitriles or of a$-unsaturated carboxylic acid esters may be mentioned:

co NX-C-C-CHO

(80 % ) C=C-CGN

+

CO

+

H2

(9.97) Rh/P(OPh 1 NX-C-CHO

I

(97 %)

C

3-Cyanopropanal obtained by cobalt catalyzed hydroformylation of acryloniuile at 200300 bar and 120-150°C was used from 1963 to 1973 by the Ajinomoto Co. at up to 1000 t/mo. to produce sodium L-glutamate via Strecker synthesis (ref. 117). On the other hand, 2-cyanopropanal is formed at mild conditions in a regioselectivity of 97 %, if a rhodiudphosphite catalyst is used in methanol as the solvent (ref. 118). A remarkable control of regioselectivity has also been observed for the hydroformylation of alkyl acrylates: co OHC-C-C-C02Me C=C-CO2Me

+

Co

(85 X )

+

(9.98) OHC-C-C02Ne 110 "C, 280 bar

I

C

(98 X )

Thus, methyl-3-formylpropionateis obtained in a regioselectivity of 85 7b by use of cobalt catalysts (ref. 119), and even higher regioselectivities have been reported for the 0x0 reaction of bulky alkyl acrylates, e.g. t-butyl acrylate. In contrast, phosphine modified rhodium catalysts give almost exclusively methyl-2-formylpropionate (ref. 3).

419 Even more impressive is the control of regioselectivity on the hydroformylation of alkyl methacrylates: C

Rh/w C I C=C-C02Me

+

co

165 "C,

I

OHC-C-C-C02Me

(90 %)

270 bar

+

(9.99) C I OHC-C-Cope

80 "C,

I

54 bar

(97 %)

C

While at elevated temperatures with a rhodium/pyridine catalyst the 3-fomyl isomer prevails (ref. 3), the 2-formyl isomer is the largely preferred product at mild temperatures with a phosphine modified rhodium catalyst (ref. 120). This last example, where the fomyl group is added to the most substituted carbon, is a remarkable exception of Keulemans' rule and it should be memorized, that e.g. isobutene hardly yields any of the corresponding product, pivalaldehyde. Enantioselective hvdroformvlations of prochiral alkenes have been investigated by various authors (ref. 121). Simple olefins such as 1-butene, or cis and trans-2-butene give 2-methyl butanal in an enantiomeric excess in the range of 19-32%, if HRh(PPh3)C0 is used as the catalyst in the presence of the chiral ligand (-)-diop. Higher enantioselectivities can be achieved by use of arylalkenes such as styrene as the substrate:

(9.100)

-

cat.

=

(-)-DBP-di0p/PtCl2/SnCl2

ee

cat.

=

PhHeBuP*/

ee = 40 %

RhCltcod)

=

76

80 X

Rhodium (ref. 122) or platinudtin chloride catalysts (ref. 123) have been applied successfully at mild conditions. A rigid coordination of a chelating chiral ligand such as in the platinum system appears to be helping to achieve a high enantiomeric excess. Enantiomeric pure aldehydes are available by hydroformylation of chiral olefins such as (-)a-pinene which is an abundant natural product. If (-)-cr-pinene is hydroformylated in the presence of cobalt catalysts, rearrangement to the bornane structure occurs (ref. 124).

420

CIl*CHO

(9.101)

63 If a rhodium catalyst is applied at < 120”C/600 bar, not only the pinene structure remains intact, but also a diastereoselective synthesis takes place yielding (+)-3-pinenecarbaldehyde in a selectivity of up to 85 %. 9.5.3.2 Hvdrocarbonvlation of Alkanols Alcohols can be hydrocarbonylated in the presence of cobalt/iodide catalysts to yield the homologous aldehydes and, by their subsequent hydrogenation, alcohols: ROH

R-CHO

CO

+

+

+

H2

H2

+

4

R-CHO

+

H20

R-CHZOH

(9.102)

(9.103)

Most interest has found the hydrocarbonylation (or “homologation”) of methanol to acetaldehyde and ethanol as an alternative route to ethylene oxidation and hydration. Although both methanol and synthesis gas are cheap feedstocks, no industrial application has been reported so far (refs. 9,10,125-127). The main reason for this is the limited selectivity of reactions (9.102) and (9.103) which make product workup difficult. Major side products are methane, dimethyl ether, methyl acetate, and condensation products of acetaldehyde. 1,I-Dimethoxyethane which is formed especially at low conversions can be regarded as an intermediate that is transformed at high conversions into acetaldehyde. The selective formation of ethanol is favoured in the presence of cobalththeniumliodide catalysts at temperatures above 200°C and pressures exceeding 300 bar. Selectivities in the range of 60-80 % have been reported (ref. 125). According to more recent studies, acetaldehyde selectivity can be largely enhanced by use of cobaldiodide catalysts at pressures up to 300 bar and temperatures of below 200°C either by limiting methanol conversion or by use of special solvents (refs. 9,10,127).

421

Thus,

the

system CO~(CO)$CH~VKI (1: 10:205) was demonstrated to be highly selective for HI

acetaldehyde. Since the

\

catalyst concentration was low, yields were only in the range of

18 % acetaldehyde

with

presumably present in the form of 1,ldimethoxyethane (refs. 10,128). High yields of acetaldehyde have been reported by use of

I [HRL,I]

t

CH3CH0

[

Pt

CH3-C-RLnI

1-

L

Co12 as the catalyst and of 1,4-dioxane as

Fig. 9.8 Proposed mechanism for methanol hydrocarbonylation

the solvent (ref. 129). The catalytic activity of this system could be enhanced dramatically by use of a I/Co ratio of > 5 in the presence of ionic promoters comprising bulky cations such as alkali metal, alkali metal/crown ethers, ammonium, phosphonium, or imminium cations. With these systems, acetaldehyde yields of up to 80 % at turnover numbers of 570-1250 h-1 were obtained (refs. 127,130). Furthemore, tests at a small continuous pilot plant demonstrated the successful recycle of these catalysts with almost no loss of activity, as is essential for a potential industrial application (ref. 131). Economic evaluations have shown acetaldyehyde synthesis via reductive carbonylation of methanol to be supenor to the Wacker-Hoechst process, if reinvestment is considered (ref. 132). A proposed mechanism for methanol hydrocarbonylation by COD-catalysts is outlined in Fig. 9.8. Methyl iodide, formed from HI and methanol, alkylates the anion [Co(CO),]- to give a methyl cobalt species. Iodide assisted methyl migration yields an anionic acetyl species which after loss of CO and addition of hydrogen gives a dihydride acetyl species. Reductive elimination yields acetaldehyde, and the anion [Co(CO),]- is regenerated by loss of HI and addition of CO (ref. 127). This mechanism has been proposed considering both kinetic investigations and stoichiometric model reactions. For instance, the reaction of [Co(CO),]- with methyl iodide in THF yields in the presence of large cations such as PPN+ (PPN+ = (Ph3P)2N+) almost quantitatively the well characterized iodide substituted anionic acetyl cobaltate depicted in Eq. 104 (refs. 133,134):

422

PPN

[ Co(CO),, ]

+ PPN

CH31

+

[ CH3C(0)Co(C0)31]

(9.104)

Considering the strong effect of cations on cobaldiodine catalysts in methanol hydrocarbonylation, this anionic acetyl complex may be assumed as an intermediate in the catalytic cycle. Cobaldidide as well as rutheniudiodide catalysts have been used to hydrocarbonylate methoxy derivatives such as 1,1-dimethoxyethane, methyl acetate, methyl formate, and dimethyl ether to yield acetaldehyde (refs. 125,127). However, rates are low if compared with methanol. Methanol hydrocarbonylation has also been carried out with Run (refs. 135-137) and Fe(CO)+nine (ref. 138) catalysts, although rates are low and methane formation is rather high. With the latter system, carbon dioxide is formed as the coproduct rather than water as in eqs. (9.102)-(9.103). The methylation of a carbonyl ferrate by methyl malkyl ammonium has been postulated as the rate determining step: NR5CH3+

HFe(C0I4-

+

--b

CH3Fe(C0),,H

+

(9.105)

NR3

Highly selective two-stage processes for the synthesis of ethanol from methanol have been reported by Halcon (ref. 139), Davy McKee (ref. 140), and Humphreys and GlasgowBASF (ref. 141). The first stage consists of methanol carbonylation to acetic acidmethyl acetate, followed by heterogeneously catalyzed hydrogenation to ethanol. Among higher alcohols, the hydrocarbonylation of benzyl alcohols is worth mentioning. The resulting phenyl acetaldehydes and phenyl ethanols are of interest as fragrances. Unfortunately, benzyl alcohols tend to be hydrogenated to yield the corresponding toluenes, and iodide promoters have turned out to be not very useful in this case (ref. 125). In contrast, dialkylacetals of phenyl acetaldehydes can be obtained in high yields, if the reaction is carried out in the presence of orthoesters:

Ph-CHz-OH

+

CH(OMe)3

+

Co

+

H2

Cog (CO)B/He I PP~-CH~-CH(O~~)~

+

HCOpe

(9.106) The selectivity to the acetal is 69 % at 110"C/300 bar and the main side products are benzyl ethers (refs. 127,142). I

S

The main incentive to study formaldehyde carbonylation/hydrocarbonylation has been to develop selective syntheses for ethylene glycol. The first process of this kind has been applied by Du Pont until 1968 (ref. 1). It involved three stages: Formaldehyde was carbonylated by acid catalysis at 300-700 bar and 200-250°C to yield glycolic acid. This was estenfied with methanol and finally hydrogenated to give methanol and ethylene glycol.

423

First reports on the hydrocarbonylation of formaldehyde derivatives to yield ethylene glycol date back to patents from 1942 (ref. 143). The synthesis of 1,1,2-trimethoxyethane from methylal was demonstrated by Gresham and Brooks by use of COO as the catalyst at 160-200°C and 600900 bar (ref. 144): CH2(0CH3)2

+

CH30H

+

CO

+

H2

+

CH30CHZCH(OCH3)2

+

H20

(9.107)

According to more recent reports the selectivity of this reaction can be improved up to about 70 % (ref. 145). With Cob-catalysts, 1,1,2-trimethoxyethane is the favoured product at low iodide concentrations while at an increased I/Co ratio the methoxy moiety of methylal is hydrocarbonylated to yield acetaldehyde (ref. 146). The stoichiometric hydrocarbonylation of formaldehyde with HCo(C0)4 at atmospheric pressure and 0°C has been reported to yield glycolaldehyde quantitatively (ref. 147): CH20

+

CO

+

H2

+

HOCHZCHO

(9.108)

The catalytic hydrocarbonylation with cobalt at 200-300 bar and 110°C is far less selective (ref. 148). Researchers from Monsanto have developed a process using rhodium catalysts such as RhCI(CO)(PPh3)2. Both formaldehyde conversion and glycolaldehyde selectivity are in the range of 80-90 % (refs. 149,150). The reaction is either carried out in N,N-disubstituted amide solvents or in the presence of bases like amines or phosphines. Glycolaldehyde is easily hydrogenated in a separate step to ethylene glycol, which can be also obtained directly from formaldehyde by adjusting the hydrocarbonylation conditions. The major drawback of the Monsanto process is the need of anhydrous formaldehyde as an expensive precursor, and all attempts to use aqueous formaldehyde have so far revealed unacceptable selectivities (ref. 12). A special variant of aldehyde carbonylation is the Wakamatsu reaction, where acylamido

acids are obtained by cobalt catalyzed amidocarbonylation (ref. 151): R-CHO

+

CO

+

R’C(OINH2

+

R-CH-CO2H

I

(9.109)

NHC(0) R ’

This reaction has found increasing interest for amino acid synthesis, recently (ref. 152). For example, phenylacetaldehyde and acetamide give N-acetyl phenylalanine in 54 % yield.

9.6 CONCLUSIONS Homogeneous catalysis is a surprisingly versatile tool for carbon monoxide activation. With synthesis gas as a reliable feedstock, C1-chemistry has found especial interest in the past decade with the aim, to develop new processes for oxygenated base chemicals. Some of these are meanwhile applied commercially such as processes for acetic acid, acetic anhydride, dimethyl

424

carbonate, and dimethyl oxalate. Other processes look promising such as acetaldehyde or ethanol from methanol, especially if two stage processes are considered. Similarly, an indirect process for ethylene glycol involving oxidative carbonylation to yield dimethyl oxalate that is hydrogenated in a separate step appears to be superior to direct CO hydrogenation. Carbon monoxide has also been shown to be an extremely versatile reagent for functionalizing alkynes, alkenes, alcohols, and alkyl as well as aryl halides. Many of these reactions are highly chemo- and regioselective and can be largely influenced by subtle variations of catalysts or of reaction conditions. It is therefore no surprise that numerous

carbonylationhydrocarbonylation reactions are used industrially, and that further new applications look promising.

9.7 REFERENCES 1

K. Weissermel and H. J. Arpe, Industrielle Organische Chemie, Verlag Chemie, Weinheim, 3rd Ed., 1988.

2

C. Masters, Homogeneous Transition-metal Catalysis - a Gentleart, Science Paperbacks, London, 1981. J. Falbe (Ed.), New Syntheses with Carbon Monoxide, Springer, Berlin, 1980. I. Tkatchenko, Synthesis with Carbon Monoxide and a Petroleum Product, in G. Wilkinson, F. G. A. Stone and E. W. Abel (Eds.), Comprehensive Organometallic Chemistry, vol. 8, Pergamon Press, Oxford, 1982, p. 101.

3 4

5 6

W. Bertleff, Carbonylation, in Ullmann's Encylopedia of Industrial Chemistry, Vth Edition, Verlag Chemie, Weinheim, 1986, Vol. A 5,217. R. Ugo, Hydroformylation and Carbonylation Reactions, in:W. Keim (Ed.), Catalysis in C, Chemistry, Reidel PublishingComp., Dordrecht, 1983, p. 135.

7 8

R. A. Sheldon, Chemicals from Synthesis Gas - Catalytic Reactions of CO and H,, Reidel Publishing Comp., Dordrecht, 1983.

R. P. A. Sneeden, Organic Syntheses where Carbon Monoxide is the Unique Source of Carbon, in G. Wilkinson, F. G. A. Stone and E. W. Abel (Eds.), Comprehensive Organometallic Chemistry, Vol. 8, Pergamon Press, Oxford, 1982, p. 19. 9 M. Roper, Chemicals from Methanol and Carbon Monoxide, in A. Mortreux and F. Petit (Eds.), Industrial Applications of Homogeneous Catalysis, Reidel Publishing Comp., Dordrecht, 1988, p. 1. 10 J. Gauthier-Laffaye and R. Perron, Methanol and Carbonylation, Editions Technip, Pans, 1987. 11 W. Keim, Homogeneous Carbon Monoxide Hydrogenation, in W. Keim (Ed.), Catalysis in C1 Chemistry, Reidel Publishing Comp., Dordrecht, 1983, p. 5. 12 J. B. Saunby, Ethylene Glycol from Syngas, in Carbon One Chemical Technology - The future route to fuels and chemical feedstocks?, The Institution of Chemical Engineers, Rugby 1986, p. 153. 13 B. D. Dombek, Adv. Catal., 32 (1983) 325.

425

J. P. Collman and L. S. Hegedus, Principles and Applications of Organomansition Chemistry, University Science Books, Millvalley, 1980. 15 F. A. Cotton and J. Wilkinson, Advanced Inorganic Chemistry, John Wiley and Sons, New

14

16 17 18 19

York, 1980, p. 1049. B. D. Dombek, Comments Inorg. Chem., 4 (1985) 241. G. Henrici-Olive and S. Olive, Coordination and Catalysis, Verlag Chemie, Weinheim, 1977, p. 234. F. Calderazzo, Angew. Chem. Int. Ed. Engl., 16 (1977) 299. C. P Casey and S. M. Neumann, J. Am. Chem. SOC.,98 (1976)5395.

20

D. R. Fahey, J. Am. Chem. Soc., 103 (1981) 136.

21 22

W. Keim, Am. Chem. Soc.,Symp. Ser., 328 (1987) 1. W. F. Gresham and C. E. Schweitzer (E. I. Du Pont de Nemours and Co.) U.S. Pat.

23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

2534018 (1950); Chem. Abst., 45 (1951) 3412 h. R. L. F’ruett, Ann. N. Y. Acad. Sci., 295 (1977) 239. B. B. Wayland B. A. Woods and R. Pierce, J. Am. Chem. Soc. 104(1982) 302. K. H. Ziesecke, Brennst. Chem., 33 (1952) 385. W. Keim, M. Berger and J. Schlupp, J. Catal., 61 (1980) 359. M. Anstock, Dissertation, Rheinisch-Westfdische Technische Hochschule Aachen, 1982. J. W Rathke and H. M. Feder, J. Am. Chem. SOC.,100 (1978)3623. H. M. Feder and J. W Rathke, Ann. N. Y. Acad. Sci., 333 (1980)45. R. B. King, A. D. King, Jr. and K. Tanaka, J. Mol. Catal., lO(1981) 75. D. L. King, J. A. Cusumano and R. L. Garten, Catal. Rev. -Sci. Eng., 23 (1981) 233. L. Kaplan, Am. Chem. Soc.,Div. Petr. Chem., Prepr., 25(1980) 340. B. T. Heaton, J. Jonas, T. Eguchi, and G. A. Hoffman, J. Chem. Soc., Chem. Commun., (1981) 331. J. L. Vidal, R. C. Schoening and W. E. Walker, Am. Chem. Soc., Symp. Ser., 155 (1981) 61. J. L. Vidal, W. E. Walker and R. C. Schoening, Inorg. Chem. 20(1981) 238.

R. Mgikl, W. Bertleff, W. Harder and R. Kummer (BASF AG), Eur. Pat. Appl. 215378 (1987); Chem. Abstr., 107 (1987)9342. J. S. Bradley, in M. Tsutsui (Ed.), Fundamental Research in Homogeneous Catalysis, Plenum Press, New York, 1979, vol. 3, p. 165. B. D. Dombek, J. Am. Chem. SOC.,102 (1980) 6855. B. D. Dombek, Am. Chem. SOC.,Symp. Ser., 152 (1981) 213. J. F. Knifton, J. Am. Chem. Soc., 103 (1981) 3959. B. D. Dombek, J. Am. Chem. Soc., 103 (1981) 6508. J. F. Knifton, J. Catal., 96 (1985) 439. R. A. Head and R. Whyman, New J. Chem., 12 (1988) 675. M. Roper, Erdol, Kohle, Erdgas, Petrochem., 37 (1984) 506. R. A. Head and M. I. Tabb, J. Mol. Catal., 26 (1984) 149.

426

46

U. Romano, R. Tesei, M. M. Mauri and P. Rebora, Ind. Eng. Chem., Prod. Res. Dev., 19 (1980) 396.

47 48 49 50 51

T. Saegusa, T. Tsuda and K. Isayami, J. Org. Chem., 35 (1970)2976. M. M. Mauri, U. Romano, andF. Rivetti, Quad. Ing. Chim. Ital., 21 (1985) 1-3,6. F. Rivetti and U. Romano, Chim. Ind. (Milan) 62 (1980)(1) 7. F. Rivetti and U. Romano, J. Organomet. Chem., 154 (1978)323. Y. Shiomi, T. Matsuzaki and K. Masunaga (Ube Industries), Eur. Pat. Appl. 108359

52 53 54

(1984); Chem. Abstr., 101 (1984)72286. W. Reppe, Liebigs Ann. Chem, 582 (1953) 1. C. W. Bird, Chem. Rev., 62 (1962) 283. P. Pino and G. Braca, in I. Wender and P. Pino (Eds.), Organic Syntheses via Metal

55 56 57 58 59 60 61 62 63

Carbonyls, Vol. 2, John Wiley, New York, 1977, p. 419. 0. L. Kaliya, Doklady Acad. Nauk. SSSR, 199 (1971) 1321. B. C. Lines and R. Long, Am. Chem. Soc.,Div. Petrol. Chem. Prepr., 14 (1969) B 159. P. Pino, F. Piacenti and M. Bianchi, in I. Wender and P. Pino(Eds.), Organic Syntheses via Metal Carbonyls, Vol. 2, John Wiley, New York, 1977, p. 233. D. R. Levering and A. R. Glasebrook, J. Org. Chem., 23 (1958)1836. N. S. Imyanitov, and D. M. Rudkovskii, J. Appl. Chem. USSR, 41(1968), 157. D. Forster, A. Hershman andD. E. Moms, Catal. Rev. Sci. Eng., 23 (1981) 89. J. F. Knifton, J. Org. Chem., 41 (1976) 793. J. Tsuji: Organic Syntheses with Palladium Compounds, Springer Verlag, Berlin, 1980. J. Tsuji, Acc. Chem. Res., 2 (1969) 144.

64 S. Kunichika, S. Sakakibara, T. Okamoto and K. Takagi, Bull. Chem. SOC.Jpn., 44 (1971) 3405. 65 D. M. Fenton, K. L. Olivier and G. Biale, Am. Chem. Soc.,Div. Petrol. Chem. Prepr., 14 66 67 68 69 70 71 72 73 74 75 76

(1969) C 77. J. Tsuji, Acc. Chem. Res., 6 (1973) 5. N. S. Imyanitov, and D. M. Rudkovskii, Zh. Organ. Khim., 2(1966). N. S. Imyanitov, Zh. Prikl. Khim., 38 (1965) 2558. T. J. Kealy and R. E. Benson, J. Org. Chem., 26 (1961) 3126. J. Tsuji, S. Hosaka, J. Kiji and T. Susuki, Bull. Chem. Soc. Jpn., 39(1966) 141. Tanaka et al., J. Chem. Soc. Chem. Commun., (1987) 758. N. V. Kutepov, W. Himmele and H. Hohenschutz, Chem. Ing. Tech. 37 (1965) 383. H. Hohenschutz, N. v. Kutepov and W. Himmele, Hydrocarbon Process., 45 (1965) (11) 141. D. Forster, Adv. Organomet. Chem., 17 (1979) 255. J. F. Roth, J. H. Craddock, A. Hershman and F. E. Paulik, Chem. Tech. (1971) 600. F. Piacenti and M. Bianchi, in I. Wender and P. Pino (Eds.), Organic Syntheses via Metal Carbonyls, Vol. 2, John Wiley, NewYork, 1977, p. 1.

427

77

78 79 80 81 82

A. Masuda, H. Mitani, K. Oku and Y. Yamazaki, Nippon Kagaku Kaishi, (1982) 249; Y. Yamazaki and A. Masuda, Yuki Gosei Kagaku Koyaishi 40 (1982) 218; Chem. Abstr. 96 (1983) 199175. T. Riill, Bull. SOC.Chim. (France), (1964) 2680. J. F. Knifton, J. Organomet. Chem., 188 (1980) 223. R. V. Porcelli and V. S. Bhise (Halcon) DE 3024353 (1981);Chem. Abstr. 94 (1982) 120902. J. E. Erler and B. Juran, Hydrocarbon Proc., (1982) (2) 109. Y. Takegami, C. Yokokawa, Y. Watanabe and H. Masada, Bull. Chem. Soc. Jpn. 37 (1964) 672.

83 84 85 86 87 88 89 90 91

M. Orchin and W. Rupilius, Catal. Rev., 6 (1972) 85. M. Foa and F. Francacalanci, J. Mol. Catal., 41 (1987) 89. H. Alper, Pure Appl. Chem., 60 (1988) 35. Y.Kamiya, K. Kawoto and H. Ohta, Chem. Lett., (1980) 1549. H. Alper, Pure Appl. Chem., 52 (1980) 607. A. Schoenberg, I. Bartoletti and R. F. Heck, J. Org. Chem., 39(1974) 3318. L. Cassar and M. Foa, Chim. Ind. (Milan), 51 (1969) 673. G. P. Chiusoli, Pure Appl. Chem., 52 (1890) 635.

92 93 94 95 96

G. P. Chiusoli and L. Cassar, Angew. Chem., 79 (1967) 177. E. Amari, M. Catellani, G. P. Chiusoli and G. Salemo, Chim. Ind. (Milan), 64 (1982) 240. M. El-Chahawi and U. Range, Chem. Ztg., 102 (1978) 1. B. Fell andH. Chrobaczek, Chem. Ztg., 108 (1984) 291. P. Pino, F. Piacenti and M. Bianchi, in I. Wender and P. Pino(Eds.), Organic Syntheses via Metal Carbonyls, Vol. 2, JohnWiley, New York, 1977, p. 43. 0. Roelen and K. Ziegler, in Naturforschung und Medizin in Deutschland 1939-1946, Vol. 36, pt. I, E. Dietrich Verlag, Wiesbaden, 1948. R. F. Heck and D. S. Breslow, J. Am. Chem. Soc.,83 (1961)4023. L. H. Slaugh and R. D. Mullineaux, J. Organomet. Chem. 13(1968) 469. G. Yagupsky, C. K. Brown and G. Wilkinson, J. Chem. SOC.A., (1970) 1392. C. K. Brown and G. Wilkinson, J. Chem. Soc. A., (1970) 2753. F. H. Jardine, Polyhedron, 1 (1982) (7-8) 569. R. L. Pruett and J. A. Smith, J. Org. Chem., 34 (1969) 327. R. L. Pruett, Adv. Organornet. Chem., 17 (1979) 1. H. Bach, H. Bahrmann, W. Gick, W. Konkol and E. Wiebus, Chem.-1ng.-Tech., 59 (1987) 882. E. Kuntz, Chem. Tech., (1987) 570. A. J. M. Keulemans, A. Kwantes and T. van Bavel, Rec. Trav. Chim., 67 (1948) 298. B. Fell and W. Rupilius, Tetrahedron Lett., (1968) 3261;(1969) 2721.

97 98 99 100 101 102 103 104 105 106 107 108

G. P. Chiusoli and L. Cassar, in I. Wender and P. Pino (Eds.), Organic Syntheses via Metal Carbonyls, Vol. 2, John Wiley, NewYork, 1977, p. 297.

428

109 110

G. Botteghi, R. Ganzerla, M. Lenarda and G. Moretti, J. Mol. Catal., 40 (1987) 129. T. Kitamura, M. Matsumoto and M. Tamura (Kuraray Co.), DE3210617 (1982); Chem. Abstr. 98 (1983) 106821.

111

K. E. Smith, G. R. Chambers, R. C. Lindberg, J. N. Cawse, A. J. Dennis, G. E. Hamson and D. R. Bryant, in R. L. Augustine(Ed.), Chemical Catalysis of Organic Reactions, Marcel Dekker, New York, 1984. B. Fell and M. Barl, J. Mol. Catal. (1977) 301.

112 113

Y. Crameri, P. A. Ochsner and P. Schudel(Givaudan), EP 52775(1982); Chem. Abstr., 97 (1982) 144574.

114 115 116 117

G. Gut, M. H. Makhzangi and A. Guyer, Helv. Chim. Acta, 48(1965) 1151.

118 119 120 121

I. Ojima, Chem. Rev., 88 (1988) 1011. I. Ojima, K. Kato, M. Okabe and T. Fuchikami, J. Am. Chem. SOC.,109 (1987) 7714. T. Kaneko, Y. Izumi, I. Chibata and T. Itah, in Synthetic Production and Utilization of Amino Acids, Kodansha, Tokyo, Wiley, New York, 1974. V. P. Kurkov (Chevron Research Corp.), US 4344896 (1982);Chem. Abstr. 97 (1982) 181731. R. Iwanaga, Y. Mori T. Yoslinde (Ajinomoto Co.), JP 8177/57(1957); Chem. Abstr. 52 (1958) 14661.

c . U.Pittman Jr., W. D. Honnick and J. J. Yang, J. Org. Chem. 45 (1980) 684. H. B. Kagan, in G. Wilkinson, F. G. A. Stone and E. W. Abel(eds.), “Comprehensive Organometallic Chemistry”, vol. 8, Pergamon Press, Oxford, 1982, p. 463. 122 H. Siege1 and W. Himmele, Angew. Chem., Int. Ed. Eng., 19(1980) 178. 123 G. Consiglio, P. Pino, L. I. Flowers and C. U. Pittman Jr., J. Chem. SOC.,Chem. Commun., (1983) 612. 124 W. Himmele and H. Siegel, Tetrahedron Lett., (1976) 907. 125 G. Braca and G. Sbrana: Homologation of Alcohols, Acids and their Derivatives by CO + H2, in R. Ugo (4.):Aspects of Homogeneous Catalysis, vol. 5, Reidel Publishing Company, Dordrecht, 1984, p. 241. 126 M. Roper and H. Loevenich: “The Homologation of Methanol”, inW. Keim (ed.):“Catalysis in C1 Chemistry” Reidel Publishing Company, Dordrecht, 1983, p. 105. 127 M. Roper, Habilitationsschrift, Rheinisch-Westfalische Technische Hochschule Aachen, 1985. 128 J. Gauthier-Laffaye, R. Perron and Y. Colleuille, J. Mol. Catal. 17 (1982) 339. 129 W. E. Walker (Union Carbide Corp.), EP 37586 (1981); Chem. Abstr., 96 (1982) 68333. 130 M. Roper, K. -H. Keim, J. Korff, G. Feichtmeier and W. Keim (Union Rheinische Braunkohlen Kraftstoff AG), DE 3343519(1985); Chem. Abstr., 104 (1986) 33762. 131 J. Korff, K. H. Keim, W. Keim and M. Roper (Union Rheinische Braunkohlen Kraftstoff AG), EP 193801 (1986); Chem. Abstr. 105 (1986) 193356. 132 A. Aquilo, J. S. Alder, D. N. Freeman and R. G. H. Voorhoeve, Hydrocarbon Process., 62 (1983)(3) 57.

429

133 134 135 136

137

M. Roper, M. Schieren and B. T. Heaton, J. Organomet. Chem. 299 (1986) 131. M. Roper and C. Kruger, J. Organomet. Chem., 339 (1988) 159. G. Braca, G. Sbrana, G. Valentini, G. Andrich and G. Gregorio, J. Am. Chem. Soc., 100 (1978) 6238. G. Braca, G. Sbrana, G. Valentini, G. Andrich and G. Gregorio, Carbonylation and Homologation of Methanol, Methyl Ethers and Esters in the Presence of Ruthenium Catalysts, in M. Tsutsui (Ed.), Fundamental Research in Homogeneous Catalysis, Vol. 3, Plenum Press, 1979, p. 221.

G. Braca, L. Paladini, G. Sbrana, G. Valentini, G. Andrichand G. Gregorio, Ind. Eng. Chem. Prod. Res. Dev., 20 (1981)115.

138

M. J. Chen, H. M. Feder, J. W. Rathke, S. A. Roth and G. D. Stucky, N. J. Acad. Sci., 415 (1983) 152.

139 140

B. Juran and R. V. Porcelli, Hydrocarbon Process., (1985)(10)85. N. Harris, Process for Preparation of Ethanol via Acetate Ester, in Carbon One Chemical Technology - The future route to fuels and chemical feedstocks?, The Institution of Chemical Engineers, Rugby 1986, p. 125. C. L. Winter, Hydrocarbon Process., (1986)(4) 71. U. Bormann, Dissertation, Rheinisch-Westfalische Technische Hochschule Aachen, 1985. W. Reppe, H. Kroper, H. J. Pistor and 0. Weissbarth, Liebigs Ann. Chem., 582 (1953) 87. W. F. Gresham and R. E. Brooks (E. I. Du Pont de Nemours), US2451333(1948); Chem. Abstr., 43 (1949) 673 d. R. Markl, W. Bertleff and R. Kummer (BASF AG), DE 3627776(1988);Chem. Abstr., 109 (1988) 148881. H. Hanrath, Dissertation, Rheinisch-Westfakche Technische Hochschule Aachen, 1985. J. A. Roth and M. Orchin, J. Organomet. Chem., 172 (1979) C27. T. Yukawa, K. Kawasaki and H. Wakamatsu (Ajinomoto Co.), DE2427954 (1975); Chem. Abstr. ; 82 (1975) 124761. A. Spencer, J. Organomet. Chem., 194 (1980) 113. A. C. S. Chan, W. E. Carol1 and D. E. Willis, J. Mol. Catal., 19(1983) 377. H. Wakamatsu, J. Uda and N. Yamakami, J. Chem. SOC.Chem. Commun., (1971) 1540. J. F. Knifton (Texaco Development Corp.), EP 281707; Chem. Abstr. 110 (1989) 135714.

141 142 143 144 145 146 147 148 149 I 50 151 152

430

CHAPTER 10

INDUSTRIAL APPLICATION OF CO CHEMISTRY FOR THE PRODUCTION OF SPECIALTY CHEMICALS

Helmut Papp and Manfred Baerns Ruhr-University Bochum P.O.B. 10 21 48 D-4630 Bochum, Germany

431

10.1 INTRODUCTION Carbon monoxide being one of the feedstocks of many industrially applied processes is readily available. It can be produced from different raw materials e.g. coal, crude oil fractions and natural gas. All these sources are presently used. Abundant information has been published on these processes. In the present article the conversion of carbon monoxide to a variety of chemical products which are very often intermediates used in subsequent chemical and petrochemical processes is dealt with. The CO-consuming processes are grouped and discussed in the following. - Carbonylation of methanol to acetic acid - Synthesis of acetic anhydride by carbonylation of methyl acetate - Synthesis of acetaldehyde and ethanol

-

Synthesis of vinyl acetate Homologation of carboxylic acids and esters Oxidative carbonylation of alcohols and production of ethylene glycol Hydroformylation of olefins (0x0 process) Reppe carbonylation and related processes

-

Carbonylation of halogenated compounds

-

-

-

Koch synthesis Most of these CO conversion processes have been put into practice and play an important role in industry (e.g. acetic acid by carbonylation of methanol and hydroformylation of olefins). However, attention has been given also to such processes which are still in the development stage (e.g. synthesis of acetaldehyde or vinyl acetate from methanol). In general, when dealing with these processes the catalytic systems applied, the reaction mechanisms suggested and the commercial applications are described; furthermore, selected references are cited to give the reader an easy access to further information. -

10.2 CARBONYLATION OF METHANOL AND RELATED PROCESSES Methanol formed with high selectivity from syngas can be used as feedstock for the production of a variety of industrially important oxygen containing chemicals by carbonylation and reductive or oxidative carbonylation, thus leading to the formation e.g of acetic acid, acetic anhydride, ethanol, acetaldehyde, vinyl acetate, propionic acid, esters of acetic acid and glycol. The overall reactions of the formation of these products from methanol and CO or syngas are as follows: CH30H + CO + CH3COOH 2 CH3OH + CO + (CH,C0)20 CH30H + CO + H2 -+ CH3CHO + H20 CH30H + CO + 2 H2 -+ CH3CH2OH + H2O 2 CH30H + 2 CO + Hz + CH3COO-CHSH2 + H2O CH30H + 2 CO + H2 -+ CH3CH2COOH 2 CH30H + 2 CO + 2 H2 + CH3COOCH2CH3 + 2 HZO

(10.1) (10.2) (10.3) (10.4) (10.5) (10.6)

(10.7)

432

2 CH30H + CO + 0 2

+ CH300CCOOCH3 H2

(10.8)

+ HOCHZ-CH2OH + 2 CH30H + 2 H 20 The first two reactions, i.e. the production of acetic acid and acetic anhydride are commercially applied, whereas reactions 3, 4 and S are claimed to be ready for commercialization; economic operation of the respective processes will, however, only be possible if naphtha prices rise significantly. The rest of reactions are still in the research and development stage. All reactions are described in the following, but the main emphasis lies on the formation of acetic acid and acetic anhydride. A review dealing with selectivity problems of the carbonylation of alcohols and esters has appeared recently (ref. 41). The use of methyl formate instead of methanol as building block for the production of oxygen containing chemicals has been reviewed by Lee et al. (ref. 43). Methyl formate may be produced from methanol or directly from syngas. It can be used as versatile intermediate for the production of acetic acid, ethylene glycol, formic acid or higher molecular acids and their derivates. A commercialization of these processes has not been reported yet.

10.2.1 SYNTHESIS OF ACETIC ACID BY CARBONYLATION OF

METHANOL Carbonylation of CH30H to acetic acid has been detected by Reppe (ref. 1). He used Co, Fe or Ni carbonyls as catalysts with halide promoters at high Pco This discovery led to a commercial process of CH3COOH production by BASF, using Co catalysts with iodide promoters at PCO= 700 bar and 200°C with a selectivity to CH,COOH with respect to methanol of ca 90% (ref. 6). Monsanto researchers discovered around 1970 that Rh carbonyls with iodide promoters are much more active and selective as catalysts for the formation of acetic acid (refs.2,3); CH3COOH selectivity amounts to 99% with respect to CH30H at low CO pressures (even at 1 bar) and temperatures of ca 100°C. The commercial Monsanto process using Rh catalysts runs at 30 to 40 bar and 180°C with 10-3 mol of Rh carbonyl as catalyst and HI as promoter (refs. 4,s). It has displaced the BASF process; 90% of the production capacity installed since 1973 has been based on the Monsanto process (ref. 7). The Rh/HI catalytic system is highly corrosive calling for expensive high alloy steels for construction, the high prices for Rh and HI necessitate the complete recycle of the catalytic system as integral part of the process. Very recently Hoechst has announced a commercial process by which acetic acid and its anhydride are produced simultaneously; this process is dealt with in some detail further below. The production of acetic acid from methanol and carbon monoxide being commercially very attractive as an alternative to ethylene based processes has been dealt with in two extensive reviews (refs. 8,9). Carbonylation of higher alcohols is also possible with a similar catalytic system. From ethanol, which is obtained by homologation of methanol (see below) propionic acid may be obtained as alternative to ethylene hydroformylation and oxidation of propionic aldehyde. Similarly phenylacetic acid can be produced by carbonylation of benzylic alcohol (reaction 10.9).

433

CgHg-CH2OH + CO + C6H5-CH2-COOH

(10.9)

10.2.1.I Catalvtic Svstems Most group VIII metals can be used for the carbonylation of CH30H but with strongly differing activities. The catalyst of the Monsanto process consists of Rh carbonyl with CH3I in an aqueous HI solution. Rh is normally added as salt, e.g. Rh13 which is transformed into the active form [RhI2(C0)2]- by CO and H20 (reaction 10)

Rh13 + 3 CO + H 2 0 + [Rh12(CO)2]-+ C02 + HI + H+

(10.10)

The methyl iodide necessary as intermediate is formed via reaction (10.11): CH30H+HI

tj

CH31+H20

(10.11)

The reaction rate of acetic acid formation is first order with respect to Rh and CH3I concentration; above a certain CO pressure it is independent of

Pco (ref. 10). As in

the case of

hydroformylation heterogenization of Rh was attempted in order to overcome the problems of separating the catalytic system, but no breakthrough has been reported. Efforts to exchange the extremely corrosive iodide acidic medium by other halides have been also without success yet. A reaction mechanism based on the application of Rh catalysts has been proposed by Forster (refs. 11- 13); the respective catalytic cycle is shown in Fig. 10.1. The rate determining step is the oxidative addition of methyliodide to the active catalyst [RhI2(CO)2]- forming the unstable, coordinatively saturated complex [CH3Rh13(CO)2]-.CO insertion CH3COI CH3I leads to the acetyl species CH3CORh13(CO)2, followed by reductive elimination of CH3COI and regaining of the original [RhI2(CO),]-. The completion of the catalytic cycle is obtained by reaction of CH3COI with H2O to form CH3COOH and HI (10.12), the latter reacting with CH30H to the active CH31 (10.11) CH3COI + H20 + CH3COOH + HI (10.12) Because of the high price of Rh, research to substitute it by less expensive

0

Fig. 10.1 Catalytic cycle for the Rh catalyzed formation of acetic acid

434

metals is still continuing. Some improvements with respect to Co catalysis by adding promoters have been reported (ref. 9). But the most promising alternative seems to be a catalytic system based on Ni which has been intensively investigated. By adding special promoters like PR3, NR3 (R = Ph, Bu, Et), Sn14 and/or alkali and transition metal salts together with a high amount of CH31, a performance very similar to Rh catalysis has been reported (refs. 9,14). The catalytic action of the different promoters has been extensively discussed by Gauthier-Lafaye and Perron (ref. 9). According to Rizkalla (ref. 14) economics are already more favorable for the Ni process; an industrial application of the low pressure Ni system appears, therefore, very likely in the near future. Synthesis of acetic acid directly from syngas without the pathway via CH30H has been reported by Knifton (ref. 15). With a selectivity up to 95% acetic acid is formed on a Ru-CoVBu4PBr “melt” catalyst from syngas. The active catalytic species is assumed to be Ru(C0)3Ip

Industrial Irnuortunce ofAcetic Acid (ref.91 Acetic acid is mainly used for the production of vinyl acetate, cellulose acetate, acetic anhydride, acetyl chloride and solvent acetates. More than 4 x 106 t/a of acetic acid are produced worldwide mainly via methanol carbonylation. Apart from this route, the oxidation of acetaldehyde (reaction (10.13)) and the nonselective oxidation of C4 hydrocarbons from naphtha pyrolysis (Celanese process) are used for the production of acetic acid. 10.2.1.2

CH3CHO + 1/2 0 2 4 CH3COOH

(10.13)

METHANOL/METHYLACETATE/METHYLlODlDE/WATER 4 I

r-

I

JI

i

rt

IATER

/

CETIC

I WATER

ACETIC ACID

/

ACETIC ACID

H E A W ENDS (PROPIONIC ACID)

i ACETIC ACID

I REACTOR

t

___)

LIGHT ENDSCOLUMN

Fig. 10.2 Flow diagram of the carbonylation of methanol to acetic acid (Monsanto process)

435

A flow diagram of the Monsanto acetic acid process is shown in Fig. 10.2.

10.2.2

SYNTHESIS OF ACETIC ANHYDRIDE BY CARBONYLATION OF METHYLACETATE

Carbonylation of carboxylic acid esters leads to the formation of the corresponding acid anhydrides in the absence of water; e.g. when using methyl acetate as feed acetic anhydride is produced. This reaction was first described by Reppe (BASF) in 1951152 using high partial pressures of CO with Co, Ni or Fe carbonyl complexes as catalysts (ref. 9). After the discovery of the Rh-catalyzed CH3COOH production by Monsanto new catalysts for the low pressure carbonylation of esters have been described (refs. 16,17). The process has been commercialized by Eastman Kodak with Rh as catalyst (ref. 18).

10.2.2.1

Catalvtic Svstem

In addition to Fe, Ni and Co originally described by Reppe, group VIII metals like Rh, Pd, Ir, Pt, 0 s and Ru have been detected as suitable candidates to carbonylate methyl acetate at low pressures of CO (25 to 150 bar). As in the case of acetic acid production Rh is the most active metal (ref. 19). Necessary promoters are CH31, organic ligands (phosphines, arsines, amines or ammonium salts) and inorganic salts ( e g alkali, earth alkali, lanthanide or transition metal salts). The reaction rate is first order with respect to the concentration of Rh and of CH3I (refs. 19,201. With RhLZ(C0)I as active species (L = organic ligand) a mechanism for the carbonylation of methyl acetate has been proposed by Gauthier-Lafaye and Perron (ref. 9). Methyl iodide reacts with the active Rh complex to form CH3RhL2(CO)12 (reaction (10.14)). Subsequently, CO insertion leads to an acetyl complex (reaction (10.15)) followed by reductive elimination of CH3COI (reaction (10.16)). ( 10.14)

(10.15) (10.16) Finally, CH3COI reacts with methyl acetate to acetic anhydride CH3COI + CH3COOCH3 -+ (CH3CO),0

+ CH31

(10.17)

This reaction is very slow without promoters, but it is considerably accelerated by inorganic salts, especially Li compounds. The effect of Li is shown by reactions (10.18) and (10.19). CH3COI + Li(CH3COO) -+ (CH3CO),0 + LiI LiI + CH3COOCH3 -+ CH31 + Li(CH3COO)

(10.18) (10.19)

436

The halolysis of methyl acetate (reaction (19)) by Li iodide is the key reaction for the increase in activity due to the inorganic promoters. Sideproducts of the Rh-catalyzed process of acetic anhydride production are small amounts of acetic acid from hydrolysis of the anhydride by traces of water, of acetone and of “tars”. The latter two lead to separation problems, since acetone forms an azeotrope with CH3I and the condensation products (“tars”) have to be separated from the catalytic Rh system which is recycled as in the case of CH3COOH production. Separation processes being rather complicated are described elsewhere (ref. 9). In the recent literature less expensive alternatives than Rh have been reported, especially Ni is suggested as catalyst. Ni was first discovered as catalytic species by Reppe when carrying out the reaction at 600 to 700 bar CO and 190°C. Similar promoters as in the case of Rh lead to a high productivity at 40 to 80 bar CO at temperatures between 50 and 200°C (ref. 9). However, commercialization of this process has not been reported yet. Rhone-Poulenc has patented a Co catalyst with cocatalysts (Ru, Co, Cr,Ti, etc) and ionic iodides supposedly active at similar conditions as the Rh system (ref. 9). In contrast to Rh and Ni an inhibition of the reaction by CH31 is observed, indicating that a different reaction mechanism prevails. Acetic anhydride can also be produced by the carbonylation of methylether (reaction (10.20)) (ref. 9):

CH30CH3 + 2 CO -+ (CH3CO),O

(10.20)

The reaction is performed with similar catalytic systems (Co, Rh, Pd) and slightly more severe reaction conditions as in the ester carbonylation. It is assumed that the reaction proceeds via methylacetate as intermediate. This process has not been commercialized, because of the lower reaction rate and the readily available acetic acid, which can be easily transformed into its methylester. Attempts to heterogenize catalytic systems (Rh, Pd, Ru, Os, Ni) have also been described in the literature (ref. 39). 10.2.2.2 Industrial ImDortance of Acetic Anhvdride (ref. 91 Worldwide production of acetic anhydride amounts to about 1 x lo6 t/a. In addition to the Eastman Kodak carbonylation process acetic anhydride is produced by the oxidation of acetaldehyde and via the ketene process (reactions (10.21) and (10.22)).

CH3COOH + CH,=C=O + HZO CH,=C=O + CH3COOH + (CH,CO)20

(10.21) (10.22)

437

Acetic anhydride is used as acetylation agent for fine organic chemicals (e.g. acetyl salicylic acid) and the production of cellulose acetate (reactions (10.23) - (10.26)). cellulose + (CH3CO),O -+ cellulose-acetate + CH3COOH CH3COOH + CH30H + CH3COOCH3 + H20 CH3COOCH3 + CO -+ (CH3CO),O

(10.23) (10.24) (10.25)

This leads to an overall reaction equation of cellulose + CH30H + CO -+ cellulose acetate

(10.26)

A flow diagram of the Eastman Kodak process of acetic anhydride production is shown in Fig. 10.3. A new process patented by Hoechst is the co production of acetic acid and acetic anhydride by carbonylation of CH30H and CH3COOCH3 in one reactor. The catalytic system consists of Rh salts together with phosphonium or ammonium iodide. A flow diagram of the process to be built in the near future is drawn in Fig, 10.4. The acidfanhydride product ratio can be adapted to the needs by a change in the methanol to methyl acetate ratio in the feed. A similar process has been

Fig. 10.3 Flow diagram of the carbonylation of methyl acetate to acetic anhydride (Eastman Kodak process)

438

Table 10.1. Reaction conditions for the carbonylation of methanol and methylacetate ~~

Condition

temperature pressure catalyst complex cocatalyst

acetic

acetic

acetic acid

acid

anhydride

anhydride

Monsanto

Easunan

190-195°c 30-35 bar

50 bar

acetic acid anhydride

190°C

Rh

Rh

Fh(C0)212]-Li+ MeI/LiI

[Rh(CO),I,]-Li+ MeILiI

announced by BP. A comparison between the different processes to produce acetic acid and acetic anhydride from methanol and/or methyl acetate is given in Table 10.1. All reactions are performed in stirred tank reactors which have to be from high alloy steels or from zirkon because of the corrosiveness of the reaction mixtures. A special separation problem comes from the fact that the products have to be almost completely free of iodine ( 4 0 ppb) since most of the acetic acid is used for the production of vinyl acetate applying iodine sensitive catalysts. The production of mixtures of acetic acid, methyl acetate and acetic anhydride starting from

CO and H2 has been patented by Haldor Topsoe (ref. 42). It is performed in two reactors, in the first a mixture of methanol and methylether is produced from syngas on a methanol catalyst combined with a zeolite like ZSMS, in the second reactor the effluent of reactor one (methanol and methylether) is carbonylated in the presence of a Rh catalyst and methyl iodide to the product mixture of acetic acid, methyl acetate and acetic anhydride.

10.2.3

SYNTHESIS OF ACETALDEHYDE AND ETHANOL

The reductive carbonylation of methanol to form ethanol (see reaction (10.4)) was intensively investigated by Wender et al. (refs. 21, 22) and by BASF researchers (ref. 24). The former named the reaction homologation. Reductive carbonylation of methanol leads always to a mixture of products. Besides acetaldehyde and ethanol the latter being formed by hydrogenation of the primary product CH3CH0, acetic acid is formed by carbonylation, methane by hydrogenation and methylether by dehydration of CH30H. Additionally, products with higher molecular weight may occur. Wender et al. (refs. 21,22) used Co carbonyls as catalysts at 180 to 190°C and 250 to 350

bar syngas pressure (C0/H2 = 1) resulting in a selectivity to ethanol of 70% and low activity. Improvements where obtained by adding iodide promoters (e.g. Ag or Cu iodide, CH-jI, 12 or alkali iodides); a strong increase in activity is observed with CH31 (ref. 23). The interest in homologation of CH30H increased lately as an alternative route to ethylene starting from syngas by dehydration of the formed ethanol (reaction 11.27):

439

c

ACETIC ACID

LIGHT ENDS

4 i

4-71 I pL4

FLASH

CATALYST

I

i WATER

RES ID il E

LIGHT ENDSCOLUMN

ACETIC ACIDCOLUMN

1

RESIDUE)

ANHYDRIDECOLUMN

Fig. 10.4Flow diagram of the coproduction process of acetic acid and acetic anhydride by carbonylation of a mixture of methanol and methyl acetate (Hoechst process)

C2H50H

+ CH2=CH2 + H20

(10.27)

Newer patents describe the use of Ru compounds, and phosphine ligands (refs. 25,26) and CH3I or other iodides (ref. 23) for the homologation of CH30H. A broad range of catalyst compositions has been investigated for the conversion of methanol to predominantly ethanol (and higher alcohols). Cobalt exhibits acceptable activity for the methanol homologation at relatively mild temperatures in the range of 160 - 2 W C , however, the selectivity to either ethanol or acetaldehyde is only moderate and a lot of byproducts (esters, acids, acetals, hydrocarbons) is formed. A marked improvement of selectivity is achieved by adding promoters which are efficient for hydrogenation, e.g. Ru, Pd or Pt, thus converting the primarily formed acetaldehyde to ethanol. Further progress was achieved by ligand modification of the catalysts. At 200"C, SO0 bar syngas pressure (CO:H, = 1:3) and a catalyst comprising Co, Ru, I and a phosphine the ethanol selectivity amounted to 80% with some propanol, ethers and hydrocarbons as byproducts. Under the conditions mentioned the conversion of methanol was in the range of SO 60%. The results have been extensively reviewed (ref. 40). The formation of acetaldehyde by reductive carbonylation of CH30H has been more intensively investigated than the formation of ethanol (refs. 27-29). A pilot plant to form acetaldehyde from CH30H homologation was run by UK Wesseling for 3 years (refs. 28, 29). The results obtained showed that this process is more economic than the formation of acetaldehyde by

440

the oxidation of CH2=CH2 in the presence of a Pd-based catalyst (Wacker-process) (refs. 28, 29). The economic advantage is, however, not big enough to shut down running “Wacker-plants” and to replace them by homologation units because of the high investment costs. Additional capacity for CH3CHO is on the other hand not needed, since one of the main uses of CH3CH0, the oxidation to acetic acid has been displaced by the Monsanto process (see above). High yields of acetaldehyde (97% CH30H conversion, 80%selectivity to CH3CHO) at high rates are obtained with Co carbonyls modified with (Ph3P)~Nligands together with 12 in the presence of a polar solvent (e.g. sulfolane or dioxane) (ref. 29). The active species in this process is according to Keim (refs. 28,29) a ligand modified Co carbonyl ((Ph3P)2N[Co(CO)4]) which reacts with methyliodide to form an iodine complex (reaction (10.28)). Addition of hydrogen leads then to acetaldehyde and a hydrido carbonyl.

r

1 (10.28)

The secondary reaction of ethanol production by hydrogenating CH3CHO may be accelerated by the addition of Ru as cocatalyst to the Co system (ref. 9).

10.2.4 SYNTHESIS OF VINYL ACETATE Vinyl acetate is presently produced by the PdZ+-catalyzed reaction of acetic acid and ethylene followed by reoxidation of the reduced Pd (10.29): Pd CH3COOH + CH,=CH2

+ 1/2 0 2 +

CH2=CH-OCOCH3 + H20

(10.29)

Process schemes using reductive carbonylation of CH30H for vinyl acetate production have been recently described by Rizkalla and Goliaszewski (ref. 30) as well as Gauthier-Lafaye and Perron (ref. 9); for the overall reaction see equation (10.5). There exist different possibilities to obtain vinyl acetate, which all constitute, however, the formation of 1,l-ethylidene diacetate (EDA) as an intermediate followed by its thermal decomposition to vinyl acetate and acetic acid (reaction (10.30)):

>T CH,CH(OCOCH3)2

+

CH2=CH-OCOCH3 + CH3COOH

The intermediate EDA may be produced on different paths, e.g. 1) acidic catalyzed condensation of CH3CHO with acetic anhydride

(10.30)

44 1

CH3CHO + (CH3C0)20 + CH3CH(OCOCH3)2 2)

direct reductive carbonylation of methyl acetate with Pd or Rh catalysts promoted by CH3I and an organic base (e.g. PR3, NR3, etc)

2 CH3COOCH3 + 2 CO + H2 -+ CH3CH(OCOCH3)2 + CH3COOH 3)

(10.3 1)

(10.32)

gas phase hydrogenation of acetic anhydride heterogeneously catalyzed by Pd on carbon (ref. 9)

2 (CH,CO)20

+ H2 + CH3CH(OCOCH3)2 + CH3OH

(10.33)

The first possibility was described by Wan (ref. 31) using acetic anhydride formed by carbonylation of methyl acetate. The second process is favoured by Rizkalla and Goliaszewski (ref. 3). In addition to the noble metals Pd and Rh some transition metal catalysts such as Ni, Mo and W have been described (ref. 32). The third process is to be preferred according to GauthierLafaye and Perron (ref. 9), since the less selective direct reductive carbonylation of methyl acetate is split into two highly selective processes, the carbonylation of methyl acetate to acetic anhydride (reaction (10.2)) and the hydrogenation of acetic anhydride to form EDA (reaction (10.33)).

10.2.5 HOMOLOGATION OF CARBOXYLlC AClDS AND ESTERS Propionic acid and higher carboxylic acids may be produced by reductive carbonylation of acetic acid, this represents a homologation of carboxylic acids (see reaction (10.6)). Ruthenium catalysts promoted by HI or CH31 are described (ref. 33) for the conversion of acetic acid to mainly propionic acid and minor amounts of higher carboxylic acids (reaction (10.34)). CH3COOH + CO + 2 Hz + CH3CH2COOH + H20

(10.34)

The conditions for this reaction are rather severe, the temperature being 220°C and the syngas pressure exceeding 100 bar. The homologation of acetic acid esters used for the formation of the homologous esters (reaction (10.35) for methyl acetate) or the respective carboxylic acid 40 bar). (reaction (10.36)) (ref. 34) occurs under milder conditions (150"C, 2 CH3COOCH3 + 2 CO + 2 H2 + CH3COOGH5 + 2 CH3COOH

(10.35)

CH3COOCH3 + 2 CO + 2 Hz + CH3CH2COOH i CH3COOH

(10.36)

442

Reaction (10.35) is catalyzed by a mixture of Rh and Ru salts along with an iodide source (e.g. CH3I) and organic bases (ref. 34). In addition Ru catalysts and mixtures of Ru and Co or Ru and Mn as catalytic active metals are described in literature (ref. 9). Without Ru propionic acid (reaction (10.36)) is the main product of the reductive carbonylation of methyl acetate (ref. 34).

10.2.6 OXIDATIVE CARBONYLATION OF ALCOHOLS AND PRODUCTION OF ETHYLENE GLYCOL Formation of oxalic acid esters by oxidative carbonylation of alcohols (see equ. (10.8)) is catalyzed by Pd and Cu similar to the catalysts used in the Wacker process to produce acetaldehyde from ethylene (refs. 8,9). The oxalic acid ester formed can be hydrogenated to ethylene glycol with ruthenium or copper chromites (ref. 9) in the case of methanol as starting material. A further way to produce ethylene glycol is by reductive carbonylation of formaldehyde (reaction (10.37)) which is produced by gas phase oxidation of methanol on silver catalysts (ref. 35) or methanol dehydrogenation using a Cu/Zn/Se catalyst (refs. 36, 37).

CH20 + CO + 2 H2 + HOCHZ-CH~OH

( 10.37)

Rh catalysts are used for this reaction at 150°C and 200 bar with a glycol selectivity of approximately 50%. A further route to ethylene glycol via formaldehyde is by hydrogenating glycolic acid obtained by hydroxycarbonylation of CH20 in the presence of strong acids (reaction (10.38)).

H2 CH20 + CO + H2O + HOCH2-COOH -+ HOCH2-CH2OH

(10.38)

This process was used industrially by Dupont till 1968 before the hydrolysis of ethylene oxide became more economical A direct synthesis of ethylene glycol from syngas in the liquid phase has also been suggested. It is catalyzed by most metals forming carbonyls, but very high pressures of syngas are necessary (300 to 3.500 bar) (refs. 8,9). The different catalytic systems for this reactions have been recently reviewed (ref. 38). An industrial application of the above mentioned processes will, however, be only feasible, when the price of ethylene oxide increases considerably.

10.3 HYDROFORMYLATION OF OLEFINS ( O X 0 PROCESS) The synthesis of aldehydes from olefins and syngas (CO + H2) in the presence of metal carbonyl catalysts is denoted as hydroformylation or 0x0 process. The reaction was discovered by Roelen in 1938 (ref. 44). The basic reactions with ethylene and higher a olefins are:

443

H,C=CH, + CO + H2 + H3C-CH2-CHO 2 R-CHSH, + 2 CO + 2 H2 -+R-CH(CH3)-CHO + R-CH2-CH2-CHO

(10.39) (10.40)

The hydroformylation of a-olefins with more than two C atoms leads always to a mixture of linear and methyl branched aldehydes as indicated by equation (10.40). The rate of hydroformylation of monoolefins decreases with increasing number of C atoms and with the shift of the double bond from the a-position to an internal position in the reactant molecule (ref. 55). Branched monoolefins show a lower reaction rate of hydroformylation than linear olefins. A wide spectrum of unsaturated hydrocarbons apart from monoolefins (dienes, alcohols, esters, nitriles, amides, amines, etc.) has been investigated in the 0x0 process, showing that it can be applied as well as for the production of bulk chemicals or speciality chemicals and also for synthetic chemistry (refs. 49,501. Even the production of asymmetric compounds from prochiralic reactants is possible via the 0x0 synthesis (refs. 56,57) Because of its high commercial importance several reviews describing hydroformylation in great detail have appeared (refs. 45-50). Therefore only a summary is given here, covering the more recent developments and their technical application.

10.3.1

CATALYSTS

All metals capable of forming carbonyls are potential catalysts for hydroformylation, the relative activities, however, differ by orders of magnitude (ref. 50): catalyst Rh > Co> Ru> Mn> Fe > Cr, Mo, W, Ni 103- 104 1 10-2 10-4 10-6 0 activity/a.u. Rh and Co exhibit a superior performance in contrast to the other metals. For this reason all commercial hydroformylation processes use these metals as catalysts. The two metals, i.e. Rh and

Co, are applied either in the form of the pure carbonyls or as carbonyls modified by various ligands. In the following a description is given firstly of the pure carbonyls, secondly of the ligand modified carbonyls as catalysts, then the efforts to heterogenize the catalytic systems are mentioned and finally, the Ruhrchemiemhone-Poulenc process as the newest development is described. Unmodified carbonyls. The active forms are the hydrido-carbonyls HRh(CO), and HCo(C0)4 which are stable sufficiently only at high partial pressures of CO (>150 bar). The gain in activity by the Rh catalyst of 103 to lo4 relative to Co is commercially almost compensated by the price of Rh which is about 103 times higher than that of Co. If Rh is used a virtually quantitative recovery and recycling of the precious metal is a necessity for any commercial process. Additionally, many patents recommend the application of mixed carbonyls containing more than one metal atom, which are claimed to provide higher activity and selectivity to speciality products (ref. 49); however, no commercialization has been reported yet. Ligand modified carbonyls The substitution of one or more CO molecules by electron donating ligands has a pronounced effect on the catalytic performance of Co and Rh catalysts. The

444

stability of the catalytically active complex HMe(CO),L, (Me = Co,Rh; L = PR3, P(OR)3, AsR3, SbR3, etc; R = C6H5, C4H9,...; n + m = 4) is higher than that of HMe(C0)4, because the electron donation of the ligands L to the central atom strengthens the metal CO bond. This effect of increased complex stability allows much lower partial pressures of CO to be applied and hence, a low total pressure. The stability of the complexes implies on the other hand a strong decrease in activity, which has to be overcome by a higher catalyst concenaation, a larger reactor volume, a higher reaction temperature or a higher partial pressure of H2 The steric influence of the ligands leads to an increase in the n/iso ratio of products which is especially desirable in the case of hydroformylation of short chain olefins. The hydride character of H in HMe(CO),L, is increased by ligands, bringing about a higher hydrogenation activity. Therefore, in the case of ligand modified Co catalysts predominantly alcohols instead of aldehydes are formed. Additionally, up to 15% of the starting olefins are hydrogenated to alkanes. Co-based catalysts in the form of HCo(CO),L, have industrial importance in the SHELL process (ref. 49) for the production of higher alcohols from a-olefins. Ligand modified Rh catalysts

are used in many processes; by suspending the complex in ligand solution almost purely linear products can be obtained. The thermal stability of the Rh complex allows a destillative separation of the aldehydes formed for recycling of the catalyst. The sensitivity of the complex to poisons like S, O2 or Fe compounds requires, however, an intense purification of the feedstock (olefins and syngas). Heterogenization of Catalvsts In order to overcome the disadvantages of homogeneous catalysis, e.g. the separation and recycling of the catalyst without losses of active material, extended research efforts have been made to immobilize the active complexes. The bonding of the carbonyls to various organic and inorganic carriers has been reported in literature (e.g. refs. 49,51-54). Reduced activity because of mass transfer limitations, low stability due to the leaching of active metal from the solid matrix during reaction and increased sensitivity to poisons have prevented commercial application of heterogenized catalysts in hydroformylation till now. Phase Transfer Catalvsis Phase transfer catalysis denotes a reaction between water soluble catalyst complexes and the water-insoluble organic substrates over the watedorganic phase boundary. This procedure represents a transition between homogeneous and heterogeneous catalysis. It has commercially been applied for hydroformylation especially of propylene with ligand-modified Rh catalysts in the Ruhrchemie/Rhone Poulenc process (refs. 61,62). The Rh catalyst is kept in aqueous solution by strongly complexing ligands which are readily soluble in water. The Ruhrchemie/Rhone-Poulenc makes use of triphenylphosphine, which has been threefold substituted in the m-position by sulfonic acid groups, in the form of the trisodium salt as complexing ligand for rhodium (ref. 62a). The complex, which is shown in Fig. 10.5, is highly water soluble but virtually insoluble in the organic phase so that practically no rhodium losses occur via the organic phase.

445

The decisive advantage of this process is the ease of separation of the aqueous catalyst solution from the organic products by simple phase separation eliminating the complicated recycling procedures of catalysts in other processes. Additionally a constant high activity and selectivity of catalysts and an optimal energy use are claimed for this process (refs. 63,64). Higher olefins than propene may also be hydroformylated in the RuhrchemieRhone-Poulenc process, e.g. (ref. 62b).

10.3.2 MECHANISM The mechanism of hydroformylation proposed 1960 by Heck and Breslow (refs. 58,59) is still accepted today. A simplified version is shown in Fig. 10.6 as catalyst cycle for the Co carbonyl catalyst. The recycled catalyst component is HCO(CO)~formed from C02(C0)8 and H2 while the active species in the hydroformylation reaction is HCO(CO)~which is stable at a defined partial pressure of CO. Olefin is added to the HCO(CO)~complex (step (2)) followed by step (3) an insertion of the olefin into the Co-H-bond, step (9,an insertion of CO into the Co-alkyl-group, and step (7) hydrogenation of the Co-acyl species by H, to form aldehyde and HCo(C0)4 With RCH=CH, as reactant (step 3), the insertion of the olefin into the Co-H-bond, results in two isomeric Co-alkyl species:

qs03Na H

S03Na

Fig. 10.5 Rh-complex with water-soluble ligands (from ref. 64)

446

11 +HZ

CH3CH2CHO

Fig. 10.6 Simplified catalytic cycle for hydroformylation of ethylene

F

R-CH~-CH~-CO(CO)~ and R- H-Co(C0)3 CH3 which finally yield linear and methyl branched aldehydes. The normalhso ratio depends on steric considerations (olefin, nucleophilic ligands, central atom) and the rate of CO insertion, thus, high PCO, the presence of nucleophilic ligands and Rh as central atom increase this ratio. The hydrogenolysis of the Co-acyl intermediate (step (7)) may be different according to IR investigations (ref. 60):

CH,-CH,-COCo(CO), + HCo(C0)4 + CH3-CH2-CHO + CO2(CO)8 or CH,-CH,-COCo(CO), + HCo(C0)4 + CH3-CH2-CHO + C%(CO), C O ~ ( C O+) CO ~ + C02(CO)8

(10.41) (10.42)

447

A reaction mechanism similar to Fig. 10.1 has been proposed for Rh catalysts. The active species is supposed to be HRh(CO), or HRh(CO),L, (n+m = 4, with m = 3 in case of high ligand concentrations).

10.3.3 COMMERCIAL APPLICATIONS (ref. 60a) The existing capacity of hydroformylation plants in the world amounted in 1989 to about 5 . 7 ~ 1 0 ~ t/a with a further 1.5~106t/a in the planningkonstruction stage. Production of n-butyraldehyde from propylene having a capacity of 4 x 106 t/a is the most important process, followed by the production of long chain alcohols from a-olefins and n-propanol and propionic acid from ethylene. nButyraldehyde is either hydrogenated to n-butanol, used as solvent, or transformed to 2-ethyl hexanol (2-EH) by aldol condensation, dehydration and hydrogenation. 2-EH is used to a large extent as plasticizer for PVC in form of the 2-ethylhexyl-phthalate. The long chain alcohols are transformed to biodegradable detergents. Processes using Co as catalyst have about half the capacity as those applying Rh as catalyst for ethylene and propene hydroformylation, whereas Cobased processes still govern the majority of higher olefins hydroformylation. A comparison of the main process variants, i.e. with hydrido Co carbonyl and ligand-modified Rh carbonyls respectively is given in Table 10.2 (ref. 49). A general decision in favour of one of the processes is only possible when considering the specific situation under which the process is realized.

A simplified flow diagram of a hydroformylation process is shown in Fig. 10.7. Despite the highly different reaction conditions principally the same process steps are involved for cobalt and rhodium catalyzed hydroformylations. In the technical practice a variety of combinations is applied related to the special feature of the process (e.g. aqueous two-phase or homogeneous catalyst systems). Dependent on the catalytic system used the reactants (olefins and synthesis gas) have to be purified before entering the hydroformylation reactor. The purification stage, mostly achieved by adsorption, is more sophisticated in the case of Rh catalysts, because of the small amounts of metal used, than in the case of Co catalysts. With ligand-modified catalysts H2 may be separately added to the reactor to obtain a HdCO ratio above 1 to increase the reaction rate of hydroformylation. The olefin feed is gaseous in the case of C2H4, gaseous or liquid for C, to C4 olefins and liquid for C,, Table 10.2. Comparison of propylene hydroformylation with Co and Rh catalysts (ref. 49) Catalyst

HCo(C0)4

HRh(CO)L?

2" ("C) p (bar) Catalyst metal concentration (%) Space time (h-l) (LHSV) normal/iso ratio Formation of paraffins Poison resistance Flexibility Investment costs

110 - 180 200 - 300

60 - 120 1 - 50

0.1 - 1.0 0.5 - 2.0 4 high high high high

0.01 - 0.1 0.1 - 0.25 11 low low low low

448

VENT GAS

4

REACTANT RECYCLE

REACTOR

I

PRESSURE SEPARATOR

REGENERATION (IF NECESSARY)

, I , CATALYST (MAKE UP) Fig. 10.7 Flow diagram of a hydroformylation process (ref. 49)

olefins. The reactor design depends on the applied catalytic system. Back mixed vessels, stirredtank single or less frequently cascade reactors are used. In the case of non-corrosive catalyst systems like Co metal the reactors are made from stainless steel, with water soluble catalysts or acidic systems high alloy steels are required. The applied pressure may be fairly low (1 to 50 bar) in case of ligand modified Rh catalysts or between 200 and 300 bar for pure Co or Rh carbonyls as catalysts. The applied temperatures lie between 60 and 200°C again depending on the catalysts used. The heat of the exothermic reaction (ca 120 kJ/mol) is removed by heat exchangers cooled with water or aldehyde (ref. 63). In a high pressure separator the unconverted reactants are separated and recycled. This is followed by catalyst separation from the product stream which is then split by distillation, in case of alcohols as desired products a hydrogenation stage is additionally incorporated. The recycling of unmodified Co catalysts may be achieved as C0-l or Co2+mainly in aqueous solution or by thermal decomposition to Co metal. This is followed by a catalyst regeneration stage where a preformation of carbonyl is performed. In some process variants the aqueous solutions of Co-' or Co2+ are directly fed to the reactor.

449

In case of ligand containing Rh catalysts either a stationary technique is applied in the low pressure processes whereas more complicated separation processes are necessary in the highpressure processes with Rh carbonyl catalysts.

10.4 REPPE CARBONYLATION AND RELATED PROCESSES Carbonylation of olefins and acetylenes with CO in the presence of a nucleophilic compound with reactive H atoms (e.g. H20, HOR, HSR, HNR2, etc.) leads to the formation of carboxylic acids and the respective derivatives (e.g. esters, thioesters, amides, etc.) when group VIII metal carbonyls are used as catalysts or as reagents. This type of reactions has been invented by Reppe and investigated in detail by him and his coworkers between 1938 and 1945 (refs. 65-70). A recent review of the Reppe synthesis has been given by Sheldon (ref. 71a) and before by Mullen (ref. 71b) and Wender and Pino (ref. 71c). The capacity of production units using Reppe carbonylation is approximately 600.000 t/a (ref. 7 1b). The overall stoichiometric reaction equations are illustrated for acetylene but also for ethylene which reacts in a similar fashion (R being H or an alkyl group): (a) Acetylene HC-CH + CO + HOR -+ H,C=CHCOOR H G C H + CO + HNR, -+ H2C=CHCONR2

(10.43) (10.44)

(b) Ethylene H2C=CH2 + CO + HOR -+ CH3CH2COOR

(1 0.45)

H,C=CH2 + CO + HNR2 -+ CH3CH2CONR2 H2C=CH2 + CO + HOOCR -+ CH3CHzCOOOCR

(10.46)

+ CO + HSR -+ CH3CH2COSR H~C=CHZ H,C=CH2 + CO + HC1+ CH3CH2COC1

(10.48) (10.49)

(10.47)

If H20 is used as reactant (see eq. (10.43) and (10.45), R=H) the reaction is called hydroxv carbon ylation or hydrocarboxylation, whereby the former is the more common nomenclature. If alcohols are added instead of water an alkoxy carbonylation is performed. The carbonylation of olefins or acetylenes may be carried out in two ways: (1) when a stoichiometric amount of metal carbonyl is used at atmospheric pressure it acts simultaneously as a catalyst but also as CO supply; (2) at high CO partial pressure the synthesis occurs in the presence of small amounts of group VIII metal carbonyls as catalysts which may be produced in situ from group VIII metal compounds. The presence of mineral acids facilitates the reaction, the reasons being evident from the mechanism shown further below. In industrial processes the catalytic route is favored for economical reasons. The carbonylation of olefins needs more severe conditions with respect to CO partial pressure and temperature than that of acetylenes.

450

10.4.1

CATALYSTS

The choice of the catalytic material and of the reaction conditions affect the product distribution of the carbonylation reaction significantly. To obtain a desired product a strict adherence to the conditions is necessary which may, however, be difficult to be achieved in an industrial process. The carbonyl complexes of Ni, Co, Rh, Pd, Pt, Ru and Fe are the predominant catalysts, added either as carbonyls or generated in situ by reacting finely divided metals or metal salts with CO. The carbonyl complexes may be further modified by additional ligands, e.g. trialkyl phosphines, tertiary amines, etc. in order to guide the reactions into a desired direction. The most active catalyst for the carbonylation of acetylene is Ni(C0)4 together with mineral acid. For the carbonylation of olefins the carbonyls of Co, Rh and Ru are of similar activity as nickel carbonyl; in some cases their activity may be even higher than that of Ni(C0)4 The amount of linear or branched products formed during the reaction depends strongly on the catalyst composition (see eq. (10.50)).Starting from n-olefins mainly linear acids or derivatives are obtained when C02(CO)8 or when (R3P),PdClz together with SnC12 as a cocatalyst is used as catalyst whereas mainly branched products are obtained in the presence of Ni(CO), or PdCl, alone or (R3P)2PdCl2 without a cocatalyst. R-CH,-CH,-COOH R-CH=CH2+CO+H20

1

(10.50)

The possible variations in product distribution by changes in the catalyst system and/or in the reaction conditions are shown in the following scheme (eqs. (10.51) to (10.55)) starting from butadiene as example for the carbonylation of conjugated dienes.

+ CO + CH30H

H,COOC-CH2-CH,-CH,-CH2-COOCH3 (10.5 1) dimethyl adipate (ref. 72)

135"C, 980 bar

HOOC-CH2-CH2-CH2-CH2-COOH (10.52) adipic acid (ref. 73)

CH3=CH-CH=CH2

+ CO + ROH

CH,-CH=CH-CH2-COOR 3-pentenoic acid ester (refs. 74,751

(10.53)

451

+ CO + ROH

>

(PhaP)2 Pd (OAc)2 110 C, 48 bar

CH2=CH-CHz.-CHz-CH2-CH=CH-CH2-COOR 3,8-nonadienoic acid ester (ref. 76)

(10.54)

+ CO + ROH + 0 2

ROOC-CH2-CH=CH-CH2-COOR PdCl,,CuC12 in AcOH 2-butene dicarboxylic acid ester (ref. 77) AcOAc, - H20

(10.55)

A further variant of the carbonylation reaction is illustrated by the formation of alcohols from olefins with Fe(CO), as catalyst (ref. 71): R-CH=CH,

+ 3 CO + 2 H20 -+ R-CH2-CH2-CH2OH + 2 CO,

(10.56)

Since here high amount of CO is converted to C02, this is not an attractive alternative to the conventional hydroformylation, with the exception of n-butanol production from propylene (see further below). Formation of hydrochinone by cyclic carbonylation of acetylene having a potential for future commercial applications is finally mentioned:

+ co*

2 HCECH + 3 CO + H2O +

(10.57)

OH

Originally Fe or Co complexes have been used as catalysts for this reaction (ref. 78). More recently the use of Ru or Rh catalysts (ref. 79) have been proposed; Lonza (ref. 79a) describes that the reaction is camed out with Ru(CO), at a temperature of 100 - 300°C and a CO partial pressure of 100 - 350 bar. The main problem seems to be the recycling of the expensive noble metal catalysts, which has to be solved before any commercial application is feasible.

10.4.2 MECHANISM A mechanism for the carbonylation of olefins in the presence of catalytic amounts Of Co2(CO), has been firstly proposed by Heck and Breslow (ref. 80) and is still generally accepted. It is rather similar to the cobalt catalyzed hydroformylation mechanism proposed by the same authors. The active catalyst is assumed to be HCo(CO), which is recycled within the process. A schematic illustration of the catalytic cycle, showing generation of HCO(CO)~, addition of the olefin, insertion of the olefin into the Co-H bond and insertion of CO into the Co-alkyl bond forming finally a CO-

452

Fig. 10.8 Schematic representation of the mechanism for the formation of acids from olefins with C O ~ ( C O(from ) ~ ref. 80) acyl group is presented in Fig. 10.8. The Co-acyl bond is cleaved by the nucleophilic attack of water (or alcohols, etc.) to form an acid (or ester, etc.); hereby the catalyst is recovered. Heck (ref. 81) also gives a mechanistic explanation for the promotion of carbonylation by hydrogen halide in the presence of Ni(CO), as catalyst. He assumes the formation of an active HNi(C0)2X species by oxidative addition of HX to Ni(CO)4, which then acts as the active catalyst. Ni(CO),

+ HX + H Ni(CO),X + 2 CO

(10.58)

Similar considerations have been put forward with respect to Pd catalysts, assuming HPd(L),Cl as active species (refs. 76,81-83).

10.4.3 COMMERCIAL.APPLICATION The first commercial utilization of the Reppe carbonylation was the production of acrylic acid from acetylene using temhydrofurane as solvent: Ni(C0)A HC CH + CO + H2O + CH,=CH-COOH 40-55 bar, 180-200°C

(10.59)

453

A selectivity of 90% with respect to acrylic acid was achieved. This process, however, has been mainly replaced by the gas phase oxidation of CH2=CH-CH3. Some plants are still in operation, e.g. at BASF Ludwigshafen (appr. 1OO.OOO t/a) using a mixture of NiBr2 and CuI as catalyst precursors. Today commercial application of hydroxy carbonylation of olefins is limited because cheaper alternative processes for carboxylic acid production are now available. Only some propionic acid (appr. 60.000 t/a) is produced by this process from ethylene (BASF process) (ref. 88). CHz=CH2 + CO + HzO + CH3-CHZ-COOH

(10.60)

In this reaction Ni(CH3CH2COO), is used as catalyst dissolved in propionic acid at 200 to 400 bar and 270 to 320°C; yields of 95% are obtained. Without the addition of water the formation of propionic anhydride is achievable (ref. 84):

CH3-CH2-COOH + CO + C2H4

Ni(C0)4

+

0 H~C-CH~-C~ 0

H3C-CH2-C '

(10.61)

*O

Methoxy carbonylation of long chain olefins (I-octene or dodecene) may be a source for fatty acid esters for use in synthetic lubricants or, after hydrogenation, detergentrange alcohols (refs. 85,89).

R-CH=CH, + CO + CH30H

CO2(CO)8

+

R-CH~CHZCOOCH~

(10.62)

The Reppe reaction was formerly also used commercially for the production of butanols from CH3CH=CH2 (ref. 87).

454

The catalytic system consists of Fe(CO)5 and a tertiary amine, e.g. n-butylpyrrolidine. The formation of a carbonyl amine complex is supposed to be the active species.

The catalyst solution is prepared by mixing Fe(CO)5, n-butyl-pyrolidine, water and nbutanol as solvent at 100°C for 2 to 4 h. Propene and CO are then passed through the solution at 15 bar, where they partially react; unconverted reactants are recirculated after product separation. The product mixture separates into three phases consisting of butanol, H 2 0 and Fe(C0)5. The catalytic system is stable for up to 6 month. A mixture of 85% n- and 15% i-butanol and a overall selectivity of 90% to alcohols is obtained. The interest in Reppe carbonylation is still vivid, which may be seen by recent patent applications (e.g. (ref. 90)) A very recently developed process is the formation of acrylic acid by oxidative carbonylation of H2C=CH2 (Union Oil, selectivity to acrylic acid: approx. 85%) (ref. 86) thus replacing the formerly applied synthesis starting from acetylene (eq. (10.59)).

CH,=CH2

+ CO + 1/2 0 2

PdII, Cull

CH,=CH-COOH

( 10.65)

AcOH,ACOAc

10.4.4 CARBONYLATION OF ORGANIC HALIDES A further method to produce aldehydes, carboxylic acids and their derivatives is by insertion of CO into a carbon halogen bond (refs.71a,91-95). It proceeds at very mild conditions with transition metal complexes (Ni, Co, Fe, Rh and Pd) as catalysts. The mechanism involves oxidative addition of the organic halide to the transition metal complex M L, (reaction (10.66)) followed by CO insertion (reaction (10.67)). The acyl metal complex intermediate undergoes in situ solvolysis by water, alcohols or amines forming acids, esters or amides (reaction (10.68)). ML,

+ RX + RML,X

RML,X

+ CO + RCOML,X

(10.66) (10.67)

455

+ H70 1 ;-

RCOOH + HX + ML, RCOOR' + HX + ML,

RCOML,X

(10.68)

RCONR'2 + HX + ML, where L is CO, PPh3 etc., x may be I, Br, C1 or even RSO,, M is one of the above mentioned transition metals. Aldehydes are formed when a mixture of CO and H2 is used instead of pure CO. In this case the intermediate acyl metal complex undergoes a hydrogenolysis (ref. 71a) to aldehydes. The carbonylation of organic halides is very versatile, because it may be applied to vinylic, aromatic, heterocyclic, benzylic, aliphatic and allylic halides (ref. 9 1). A commercialization of these processes has, however, not been reported yet. Phase transfer catalysis has been successfully applied to the carbonylation of organic halides (refs. 71a,95).

10.5 KOCH REACTION Koch and coworkers (ref.96) developed a synthesis of carboxylic acids by converting olefins with CO and H 2 0 in the presence of strong acid catalysts (cf. for a recent review (ref. 102)). The hydroxy carbonylation (or hydro carboxylation) of olefins with catalysts like H2SO4, HF, H3P04 alone or together with BF3 or SbF, leads to highly branched carboxylic acids, e.g. the formation of pivalic acid from i-butene.

(10.69) H3C

CH3

The reaction conditions are relatively mild, with tempkratures ranging from 0°C to 80'C and pressures from 10 to 100 bar CO. The mechanism of the reaction is assumed to proceed via carbenium ion intermediates, i.e. by addition of a proton to the olefin, followed by isomerization, to form the most stable carbenium ion (ref. 98):

R-CHz-CHSH,

H+

+

+

R-CH2-CH-CH3

+

+

R-CH-CH,-CH,

+

/

R-C +

CH3

'CH3

(10.70)

R = alkyl After addition of CO an acylium cation is formed which reacts in a separate step with H20 or ROH to the corresponding carboxylic acid or carboxylic esters, respectively.

456

+HzO (10.7 1)

_$

-H+ CH3

H3C 0

H3C 0

One technical application of the Koch reaction is the synthesis of pivalic acid (trimethyl acetic acid, eq. (10.69)) from i-butene in a series of stirred tank reactors at 20 to 80°Cand 20 to 100 bar CO with H3POdJBF3 as catalyst. In a first stage CO is added and an acyl cation/catalyst complex is formed which is subsequently decomposed by H 2 0 in a second stage and the catalytic H3POdJBF3 mixture is recycled. Selectivity to pivalic acid with respect to CO is between 80 to 100%. Side products are carboxylic acids of dimerised butene. Isobutanol or tert. butyl alcohol may be used as raw materials too (ref. 103). Shell (refs. 99,100) and Exxon (ref. 101) apply the Koch synthesis principle commercially for the production of branched C6 to C,1 carboxylic acids (so called Versatic acids or Neo acids). Dupont uses the Koch synthesis for the synthesis of glycolic acid from CO and formaldehyde with an estimated capacity of 60.000 t/a (ref. 102). AS a consequence of the high degree of branching of the Koch acids, the carboxylic group is strongly sterically hindered. Therefore, esters of Koch acids are very stable against hydrolysis, thermal degradation and oxidation. Because of these properties Koch acids are used for the modification of polyesters, synthetic lubricants, plasticizers for PVC, and as glycid esters for the modification of alkyd resins. A further reaction connected with the name of Koch is the aryl carbonylation (KochGattermann synthesis). Aromatic aldehydes are formed from aryl compounds and CO in the presence of strong Lewis or Bronsted acid catalysts like aluminium chloride together with hydrochlorid acid. The catalyst is applied in stoichiometric amounts forming a strong complex with the aldehyde product, which has to be decomposed at the end of reaction. A low (1 bar with CuCI, as promoter) and a high (100 to 200 bar, without promoter) pressure synthesis is described (ref. 71a).

Mitsubishi reports about a modified Koch-Gattermann synthesis as a commercially viable alternative for the production of terephthalic acid from toluene (ref. 97). In this process toluene is reacted with CO in the presence of HF/BF3 to form a para-tolualdehyde/HBF4 complex, which is thermally decomposed in a continuous distillation column and the catalyst is recycled. The aldehyde is then oxidized to terephthalic acid.

$?

+ HF +BF3+

(10.72) toluene complex

457

(10.73) CHO para-tolualdehyde complex

(10.74)

CHO

CHO Decomposition

+0

2

(10.75)

c0(0Ac)~/NaBr CHO

COOH terephthalic acid

A heterogenization of the homogeneous acid catalyst system has been attempted for the Koch reaction as well. In a recent patent application Holderich et al. of BASF (ref. 104) reported about the use of zeolites as catalysts for the formation of Koch acids at temperatures from 50 to 500'C and pressures of 10 to 700 bar. A commercial application of this catalyst system has not been reported so far.

ACKNOWLEDGEMENT The authors have very much appreciated the help of some of their industrial colleagues which have contributed in numerous ways, particularly by giving their criticisms and providing US with some additional material incorporated in the text. Especially, we would like to mention Dres. D. Frohning and E. Jagers, both from Hoechst and Dr. U. Wagner from BASF.

10.6 REFERENCES

3 4

W. Reppe, Justus Liebig's Ann. Chem. 582 (1953) 1. F.E. Paulik and J.F. Roth, J. Chem. SOC.Chem. Commun. 1968, 1578. F.E. Paulik, A. Henchman, W.R. Knox and J.F. Roth, US Patent 3.769.329 (1973). J.W. Roth, J.H. Craddock, A. Hershman and F.E. Paulik, Chem. Tech. 1.(1971) 600.

5

H.D. Grove, Hydrocarbon Process. 51 (1972) 76.

1 2

458

6 7 8 9 10 11 12 13 14

H. Hohenschutz, N. von Kupetow and W. Himmele, Hydrocarbon Process. 45 (1966) 141. D. Forster and T.C. Singleton, J. Mol. Catal. 17 (1982) 299. R.A. Sheldon, “Chemicals from Synthesis Gas”, D. Reidel Publishing Company, Dordrecht, 1983, ch. 7. J. Gauthier-Lafaye and R. Perron, “Methanol and Carbonylation”, Edition Technip Paris and Rhone-Poulenc Recherches, Courbevoie, 1987, ch. 6. J. Hjortkajaer and V.W Jensen, Ind. Eng. Chem., Prod. Res. Dev. Is(1976) 46. D. Forster, J. Am. Chem. SOC. 97 (1975) 951. D. Forster, J. Am. Chem. SOC. (1976) 846. D. Forster, Adv. Organometal. Chem. (1979) 255 N. Rizkalla, in “Industrial Chemicals via C, Processes” (D.R. Fahey, ed.), ACS

z

Symposium Series 328, Washington 1987, p. 61. 15 J.F. Knifton, ibid., p. 98. 16 N. Rizkalla, German patent 2.610.036 (1976) 17 H. Kuckertz, German patent 2.450.965 (1976). 18 Chem. Week, 126(1980) 40. 19 G. Luft and M. Schrod, J. Mol. Catal. 20 (1983) 175. 20 M. Schrod and G. Luft, Ind. Eng. Chem., Prod. Res. Dev. 20 (1981) 649. 21 I. Wender, R. Levine and M. Orchin, J. Am. Chem. SOC.71(1949) 4160. 22 I. Wender, R.A. Friedel and M. Orchin, Science 113(1951)206. 23 J. Gauthier-Lafaye,R. Perron and Y. Colleuille, J. Mol. Catal. 11(1982) 339. 24 G . Wietzel, K. Eder and A. Scheurmann, German patent 867,849 (1953). 25 J.E. Bozik, T.P. Kobylinski and R.W. Pretzer, US patent 4.239.924 (1980). 26 R.A. Fiato, US patent 4.233.466 (1980). 27 R.W. Wegmann, D.C. Busby and J.B. Letts, in “Industrial Chemicals via C, Processes” @.R. Fahey, ed.)ACS Symposium Series 328, Washington 1987, p. 125. 28 29 30 31 32 33 34 35 36 37

W. Keim, ibid., p. 1. W. Keim, J. Organomet. Chem. 372 (1989) 15. N. Rizkalla and A. Goliaszewski, in “Industrial Chemicals via C, Processes” (D.R. Fahey, ed.),ACS Symposium Series 328, Washington 1987, p. 136. C.G. Wan, German patent 2.856.791 (1979). N. Rizkalla, US patent 4 335 059 (1982). J.F. Knifton, J. Mol. Catal. 11(1981) 91. E. Drent, in “Industrial Chemicals via C, Processes” (D.R. Fahey, ed.) ACS Symposium Series 328, Washington 1987, p. 155. G.A. Halbritter, W. Miilthaler, H. Sperber, H. Diem, C. Dudeck and G. Lehmann, US patent 4.072.7 17 (1968). M. Osugi and T. Uchiyama, US Patent 4.054.609 (1977). A. Meyer and A. Renken, Proc. 9th Intern. Congr. Catalysis (M.J. Phillips and M. Ternan, eds.) vol. 4, 1988, p. 1898.

459

38 39 40 41 42 43 44 45 46 47 48 49

50 51 52 53 54 55 56 57

58 59 60 60a 61 62 62a

62b

B.D. Dombek, J. Organomet. Chem. 372 (1989) 151. G. Rittner and G. Luft, Chem. Ing. Tech. 3 (1986) 668. Report T 84-054, Ministry for Science and Technology (H. Bahrmann and W. Lipps, eds., Ruhrchemie AG) 1985. G. Jenner, Appl. Catal. 2 (1989) 99. J.B. Hansen, F.H. Joensen and F.A. Haldor, DE 38 21 965 Al(1989). J.S. Lee, J.C. Kim and Y.G. Kim, Appl. Catal. 57 (1990) 1. 0. Roelen, DRP 849,548 (1938) M. Orchin and W. hpilius, Catal. Rev. 6 (1972) 85 L. Marko, in “Aspects of Homogeneous Catalysis” (R. Ugo, ed.)Reidel, Dordrecht, 1973, Vol. 2, ch. 1 F.E. Paulik, Catal. Rev. 6 (1972) 49 P. Pino, F. Piacenti and M. Bianchi, in “Organic Synthesis via Metal Carbonyls” (I. Wender and P.Pino, eds.) Vol. 2, John Wiley and Sons, New York, 1977, p. 43 B. Cornils in “New Synthesis with Carbon Monoxide” (J. Falbe, ed.), Springer Verlag, Berlin, 1980, p. 1 R.A. Sheldon “Chemicals from Synthesis Gas”, D. Reidel Publishing Company, Dordrecht, 1983, ch. 4 L.L. Murrell, in “Advanced Materials in Catalysis” (J.J. Burton and R.L. Garten, eds.) Academic Press, New York, 1977, p. 235 R.H. Grubbs, Chem. Tech. (1977) 512 A.A. Oswald and L.L. Murrell, US Patent 4 083 803 (1978) P.L. Ragg, DE 2 OOO 829 (1972) I. Wender, S. Methin, S. Ergun, H.W. Sternberg and H. Greenfield, J. Am. Chem. S 0 c . Z (1956) 5101 ibid (48) p. 136 M. Tanaka, Y. Ikeda and I. Ogata, Chem. Letters (1975) 1158 D.S. Breslow and R.F. Heck, Chem. Ind. (London) (1960) 467 R.F. Heck and D.S. Breslow, J, Am. Chem. SOC.18.(1961) 4023 N.H. Alemdarogly, J.L.M. Penninger and E. Oltay, Monatssch. Chem. 107 (1976) 1153 Lit. (49); p. 177 DE-PS 26 27 354 v. 18.06.1976 (Rhone-PoulencIndustries, E. Kuntz) DE-PS 32 34 701 v. 18.09.1982 und DE-PS 34 13 427 v. 10.04.1984 und DE-PS 35 46 123 v. 24.12.1985 (Ruhrchemie AG; B. Comils et al.) DE-PS 32 35 030 v. 22.09.1982 (Ruhrchemie AG; R. C a n e r et al.) DE-PS 32 45 883 v. 11.12.1982 (Ruhrchemie AG; R. GPtner et al.) DE-PS 34 31 634 V. 29.08.1984 (Ruhrchemie AG;L. Bexten et al.) DE-PS 34 47 030 v. 22.12.1984 (Ruhrchemie AG;B. Cornils et al.)

460

63

B. Cornils, W. Konkol and C.D. Frohning, Advances on the Ruhrchemie-Rhone-Poulenc Process based on biphasic liquid hydroformylation, Proc. 5th Symp. on Relations betw. Homog./Heterog. Catalysis, Novosibirsk 1986

64 H.W. Bach, W. Gick, W. Konkol and E. Wiebus, Proc. 9th Intern. Congr. Catalysis (M.J. 65 66 67 68 69 70 71a 71b 7lc 72 73 74 75 76 77 78 79 79a 80 81 82 83 84 85 86 87 88 89

Phillips and M. Ternan, eds.), Ontario, 1988, Vol. 1, p. 254 W. Reppe, Justus Liebigs Ann. Chem., 582 (1953) 1. W. Reppe and H. Kroper, ibid., 582 (1953) 38. W. Reppe, H. Kroper, N. v. Kupetow and H.J. Pistor, ibid., 582 (1953) 72. W. Reppe, H. Kroper, H.J. Pistor and 0. Weissbarth, ibid., 582 (1953) 87. W. Reppe et al., ibid., (1953) 116. W. Reppe and H. Vetter, ibid., 582 (1953) 133. R.A. Sheldon, “Chemicals from Synthesis Gas”, D. Reidel Publishing Company, Dordrecht, 1983, ch. 5. A. Mullen, in “New Synthesis with Carbon Monoxide” (J. Falbe, Ed.), Springer-Verlag, Berlin 1980, ch. 3. I. Wender and P. Pino, eds., “Organic Synthesis via Metal Carbonyls”, vol. 2, John Wiley and Sons, New York 1977. R. Kummer, H.W. Schneider, F.J. Weiss and 0. Lemon, German Patent, 2 837 815 (1980) BASF. Belgian Patent 770 615 (1972) to BASF. J. Tsuji, I. Kiji and S. Hosaka, Tetrahedron Letters 1964, 605. K. Bittler, N. v. Kupetow, D. Neubauer and H. Reis, Angew. Chem., 80 (1968) 352. J. Knifton, J. Catal., 60(1979) 27. D.M. Fenton and P.J. Steinwand, J. Org. Chem., 12 (1972) 2034. W. Reppe, N. v. Kupetov and A. Magin, Angew. Chem. Intern. Ed., 8 (1969) 727. P. Pino, G. Braca, G. Sbrana and A. Cuccuru, Chem. Ind. (London) 1968,1732. P.Pino, G. Braca, G. Sbrana, Swiss. Pat. 442,346 (1963), 489,450 (1968). R.F. HeckandD.S. Breslow, J. Am. Chem. Soc.,u(1963) 2013. J. Tsuji, Acc. Chem. Res., 2 (1969) 144. J.F. Knifton, J. Org. Chem., 4 (1976) 2885. D.M. Fenton and K.L. Olivier, Chem. Tech., 2 (1972) 220. R.E. Brooks et al., Ind. Eng. Chem., 49 (1957) 2004. P. Hofmann, K. Kosswig and W. Schaefer, Ind. Eng. Chem., Prod. Res. Dev., 19 (1980) 330. G.P. Chiusoli, Pure Appl. Chem., 2 (1980) 635. BASF, DE-AS 1 114 796 (1960); 1 114 797 (1960). K. Weissermel, H.J. Arpe: “Industrielle Organische Chemie”, Verlag Chemie, Weinheim, 1978. P. Hofmann, Fette, Seifen, Anstreichmittel, 85 (1983) 126.

461

90 91 92 93 94 95

E. b e n t (Shell), Eur. Pat. Appl., EP 273 489 (1987), EP 282 142 (1988), EP 291 117 (1988), EP 310 168 (1989). I. Gauthier-Lafaye and R. Perron, in “Industrial Applications of Homogeneous Catalysis” (A. Momeux and F. Petit, eds.) D. Reidel Publishing Company, Dordrecht, 1988, ch. 2. B. Fell, H. Chrobaczek and W. Kohl, Chemiker Ztg. 109 (1985) 167. M. Foa, F. Francalanci, E. Bencini and A. Gardans, J. Organomet. Chem. 285 (1985) 293. I. Pri-Bar and H. Alper, J. Org. Chem. 2 (1989) 36. H. des Abbayes, J.-C. CICment, P. Laurent, G. Tanguy and N. Thilmont, Organometallics 1 (1988) 2293.

H. Koch, Brennstoff-Chem. 36 (1955) 321. S. Fujiyama and T. Kasahara, Hydrocarbon Processing 57 (1978) 147. 98 H. Koch, Fette, Seifen, Anstrichmittel 2 (1957) 493. 99 P. Regimbean and L.A. de Boisse, FR 1 252 675 (1960). 100 J. van Dam and M.J. Waale, Chim. Ind. 90 (1963) 51 1 101 W.J. Ellis and C. Ronnig, Hydrocarbon Process. Petrol Refinger &4 (1965) 139. 102 H. Bahrmann, in “New Synthesis with Carbon Monoxide” (J. Falbe, ed.), Springer-Verlag, Berlin, 1980. ch. 5 103 Mitsubishi Gas, J.P. Kokai 80/12870 (1980); XU97245 (1981). 104 Holderich, J.G. Rewers, R. Kummer and L. Hupfer, DE 3620 581 A1 (1987). 96 97

462

CHAPTER 11

THE CATALYZED HYDROGENATION OF CARBON MONOXIDE AN OVERVIEW AND FUTURE DIRECTIONS

Gabor A. Somorjai

Department of Chemistry and Center for Advanced Materials, Lawrence Berkeley Laboratory University of California Berkeley, California 94720, (USA)

463

11.1 INTRODUCTION Studies of the catalyzed hydrogenation of carbon monoxide play a pivotal role in modem catalyst science. The reaction is exothermic whether it produces methane, methanol, high molecular weight hydrocarbons or alcohols (ref. 1). Thus, it makes fuels and chemicals from the synthetic gas feedstock, CO and H2, that can be produced readily from methane (by the steam “reforming” reaction) or by the steam gasification of coal (ref. 2). The nearly thermoneutral water gas-shift reaction (CO + H20 = C02 + H2) can be utilized to adjust the H2 to CO ratio appropriate for the formation of the desired products. Just as the early development of syn-gas based technologies was necessitated by the lack of availability of petroleum as a source of fuels and chemicals in Europe in the 1920s, ref. 2) the energy crisis in the 1970s rekindled intensive research and technology developments of syn-gas based processes. However, by the early seventies, modern surface science reached such a level of maturity that its various techniques could be employed for the CO hydrogenation studies. Low energy electron diffraction and electron microscopy could be used to explore the catalyst surface structure, electron spectroscopies (X-ray photoelectron, Auger and U.V. photoelectron-spectroscopies)were used to determine the surface composition and the oxidation states of surface atoms (ref. 1). Vibrational spectroscopies (high resolution electron energy loss and fourier transform infrared) determined the structure of chemisorbed molecules, reaction intermediates and molecular fragments. Solid state NMR and radioisotope labeling techniques were also employed extensively. Single crystal model catalysts explored the elementary reaction steps, the structure sensitivity of chemisorption, bond breaking and bond formation. The roles of promoters and catalyst supports were studied in detail (ref. 3). Surface science applied to catalyzed CO hydrogenation revealed many of the molecular details of this complex and important reaction. For most reactions that produce hydrocarbons, CO may dissociate first before the hydrogenation of its carbon fragment occurs. In other reactions leading to alcohol formation as the primary reaction product, the direct hydrogenation of molecular CO takes place (ref. 4). The primary reactions of olefin or alcohol formation is followed by secondary reactions of carbon chain growth leading to the formation of liquid or solid high molecular weight molecules (ref. 5). Promoters, alkali metal ions mostly, could be used to control the level of hydrogen saturation of the hydrocarbons (ref. 6). All these scientific investigations and the discovery of novel oxide supports, titanium oxide and zeolites permitted the development of new technologies that produce fuels or the desired chemicals with excellent selectivity. The transition metal or transition metal compound catalyst was used to control the elementary reaction, alkane, alkene or alcohol formation. The support oxide could accelerate further the metal catalyzed reaction steps or control the secondary reactions, the polymerization of the CH, fragments to produce linear or cyclic, aromatic products. By careful control of additives, the rates of CO dissociation, its insertion into an olefinic bond, hydrogenation and C-C bond formation could all be tailored to produce almost any organic molecule selectively or liquid fuels with appropriate molecular weight range and octane number.

464

The papers in this book provide an up-to-date account of the catalytic science of CO hydrogenation. In this final and concluding chapter, I shall review some of the key features of this reaction, the catalysts that carry them out and the roles of the various promoters that enhance activity or selectivity. Finally, some of the possible directions of future research will be pointed out and discussed.

11.2 THE CHEMISORPTION AND DISSOCIATION OF CARBON MONOXIDE ON CLEAN TRANSITION METALS Vibrational spectroscopy and LEED crystallography investigations reveal that CO adsorbs mostly on bridge and top sites on most transition metal surfaces at lower coverages (up to one-half monolayer) (ref. 7). In addition, chemisorption in 3-fold sites have also been reported over palladium and gem-dicarbonyl species (two CO molecules bound to one metal atom) have been reported to be present on dispersed metal particles. CO chemisorbs with higher heats of adsorption at defect sites, steps and kinks (ref. 1). It is not surprising therefore that dissociation of the C-0 bond occurs most readily at these sites. The oxygen atom formed from the dissociated molecules usually reacts with another CO molecule to produce CO, that desorbs because of weaker bonding to the metal. The net process 2 CO -+ C + CO, is exothermic and is called the Boudouard reaction (ref. 2). It can be used to titrate the amount of surface carbon formed by the quantitative detection of CO, evolution. At higher coverages above one-half monolayer, the heat of adsorption of CO declines rapidly until near one monolayer the heat of adsorption per molecule is about one-third of that of the heat of adsorption at low coverage (11-13 kcal vs 25-32 kcal) (ref. 1). This is due to repulsive CO-CO interaction in the adsorbate layer. Surface crystallography studies indicate that the adsorbed

CO species move away from top sites to new sites of low symmetry (ref. 7). Since most catalytic reactions are studied at high surface coverages, the role of these weakly adsorbed molecules in the catalytic process could be important.

11.2.1

ALKALI METAL lNDUCED CO BOND WEAKENING AND DISSOCIATION

The heat of adsorption of CO (that is an electron acceptor on most transition metals) increases by as much as 10-15 kcal/mol on coadsorbed with alkali metals (that is an electron donor) (ref. 8) on most transition metal surfaces that include Pt, Rh, Ni, and Fe. For example, CO desorbs completely from a clean Cu(ll0) surface at temperatures below 200 K whereas in the presence of coadsorbed potassium, two new binding sites are populated yielding CO desorption at 480 K and 5 5 0 K (ref. 9). This corresponds to an increase in the heat of adsorption from around 11 kcal/mol to greater than 28 kcal/mol. Not only the molecular bond energies are altered by coadsorption of alkali on transition metal surfaces but also the ordering characteristics and the structure of the adsorbed monolayer (ref. 10). LEED and HREELS studies show that benzene molecularly adsorbs at 300K in a

465

Table 11.1

Coadsorption of Adsorbates on Rh(ll1) Adsorbates

Ratio 1:l 1:l 1:l 2: 1 1:l 2: 1 1:l 1:l

CO + NO Na + C2H2 Na + IC-CH3 Na + c&

*

Coadsorbed LEED Pattern ~(4x2) ~(4x2) ~(4x2) (3x3) c(24x4)rect (3x3) c(243x4)rect ~(4x2) Disorder Disorder Disorder Mixed*

2 patterns characteristic of individual adsorbates observed suggesting phase separation into independent domains.

disordered manner on a clean Rh(ll1) surface. However, it can be readily ordered (ref. 11) by coadsorption with other molecules, such as CO and NO, that are electron acceptors. Like most organic molecules, benzene is a strong electron donor to metal surfaces. Therefore, the presence of electron acceptor-donor interactions apparently induce ordering and the formation of surface structures containing both benzene and CO molecules in the same unit cell. This is not an isolated phenomenon: Table 11.1 gives examples of several systems including those containing alkali metals where the coadsorption of an electron donor (like NO) and an acceptor (like CO) leads to the formation of ordered structures while the coadsorption of two electron donors or two electron acceptors yields disordered surface monolayers (12).

I I .2.2.

CO Dissociation

Perhaps the most frequently studied molecule whose co-adsorption with alkali metals on transition metal surfaces leads to bond dissociation is carbon monoxide. A typical CO bond dissociation yield as a function of alkali metal coverage has been studied using rhodium single crystal surfaces. CO does not dissociate at low pressures on the Rh(lll1) surface. Upon coadsorption with potassium, dissociation of the molecule occurs and at 20 times of a monolayer potassium coverage, three CO molecules dissociate per potassium atom. As the K surface coverage increases, the CO dissociation probability rapidly decreases and becomes zero over a potassium monolayer. In these studies the CO dissociation is monitored by the scrambling of doubly labelled 13Ci60 and l2ClSO isotopes (ref. 13). Increases of the CO dissociation probability by alkali coadsorption were also observed by Broden et al. (ref. 14), Benziger and Madix (ref. 15), Kiskinova (ref. 16), Kelemen (ref. 17), Berko (ref. 18), de Paola et al. (ref. 19), Luftman and White (ref. 20), Hoffman et al. (ref. 21), Weimer et

466

al. (ref. 22), Lackey et al. (ref. 9), Whitman et al. (ref. 23), and others (ref. 24,25). Usually spectroscopic studies provide convincing evidence for CO bond breaking. These include studies by both vibrational and electronic surface spectroscopies. CO does not adsorb on potassium at low pressures (
11.3 THE KINETICS OF THE CO/XREACTION Perhaps the formation of methane has been investigated in the greatest detail from the viewpoint of reaction kinetics. The reaction is positive, near first order in hydrogen pressure and negative first order with respect to CO pressure. Studies, mostly on nickel (ref. 26), reveal that the reaction is structure insensitive and that the dissociation of carbon monoxide is the rate determining reaction step. The dissociation of CO can be monitored through the exothermic Boudouart reaction: 2 CO + C + CO, that occurs by pulsing CO over supported methanation catalysts. The amount of C02 evolved is equal to the amount of carbon formed over the transition metal catalyst. When hydrogen (or even water vapor) is introduced in a subsequent pulse, methane is produced from the carbon deposit as long as the reaction temperature is below 300°C (ref. 1,5). At higher temperatures the surface carbon aggregates to form a graphitic layer that hydrogenates very slowly as compared to the atomic carbon species produced by CO dissociation at lower temperatures. Since the reaction rates and the activation energies (=24 kcal/mole) are the same when methane is produced either by a steady state flux of CO and H2 or by alternate pulses of the two reactants, there is little doubt that the dominant reaction mechanism for methane production involves CO dissociation and subsequent sequential hydrogenation of the surface carbon to C a , CH2, CH3 and finally to CH4 (ref. 27). For catalysts that hydrogenate at a slower rate as compared to nickel there is increasing probability for carbon-carbon bond formation among the CH, species leading to ethylene, propylene, ethane and propane formation. Iron, cobalt and rhodium produce a fair fraction of C2 and C, species at low conversion. Thus, a CO hydrogenation catalyst that produces methane and alkenes must be able to dissociate CO, hydrogenate the surface carbon and permit C-C bond formation (ref. 28). The hydrogenation of CO, has also been detected to yield methane (E*=16kcal/mole), C, and C3 products at low conversion. However, its kinetics has been studied to a lesser extent. It has CO + H20 is an important elementary been proposed that the water gas shift reaction, CO, + H, step to convert CO2 to CO which then dissociates to produce hydrocarbons (ref. 29). This is certainly the case for some of the transition metal catalysts. However, there is growing evidence that C02 may be hydrogenated directly and the formation of a formate intermediate is an important reaction path to produce alkanes.

11.3.1 EVIDENCE FOR SECONDARY REACTIONS Methane, ethylene and propylene are the primary CO/H, reaction products over iron catalysts at low conversion. As the conversion is increased, higher molecular weight hydrocarbons are produced. The conversion may be increased by increased catalyst loading or by longer contact

467

time between the reactants and the catalysts. It was found that even at low conversion (=1%), addition of ethylene and propylene to CO and H2 greatly increases the molecular weights of the reaction products (ref. 30). While a fraction of the added olefins hydrogenate to alkanes, the remaining fraction acts as chain initiators for the polymerization of the CH, intermediates to produce liquid and solid hydrocarbons. Ruthenium catalysts are well known for producing high molecular weight solids from CO and H2 Manipulation of the catalytic reactor configuration and reactant contact time is one way to adjust the product distribution that obeys polymerization kinetics (the so-called Schultz-Flory distribution of products) (ref. 2). Catalyst additives are often used for this purpose also. Copper addition to iron for example inhibits the secondary chain growth reaction while potassium reduces the rate of hydrogenation of the reaction intermediates thereby permitting more olefins to form and the production of higher molecular weight liquid or solid hydrocarbons through secondary reactions. Alkali catalyst promoters play a significant role in controlling the product distribution through alteration of the secondary reactions leading to chain growth. As mentioned above, the most important contribution of the alkali ions is through the reduction of the rates of hydrogenation that permits more carbon-carbon bond formation before the desorption of the hydrocarbon products. However, alkali metal ions on transition metals also increase the rates of CO dissociation markedly.

11.4 PROMOTION BY THE OXIDE-METAL INTERFACE CO hydrogenation catalysis has benefited greatly from the rediscovery of the unique catalytic behavior of oxide-metal interfaces first observed by Schwab and his coworkers (ref. 31). The effect is commonly referred to as strong metal-support interaction or SMSI. Tauster (ref. 32) reported large enhancement in the CO hydrogenation rates for transition metal catalysts when supported on high surface area titanium oxide. Subsequent studies of catalyst activation involving reduction and reoxidation using H2 and 0, respectively indicated that the catalyst is activated by optimizing the oxide-metal interface area (ref. 33) since the same catalytic behavior can be obtained by depositing the metal on the oxide support or by deposition of oxide islands on the transition metal the oxide-metal periphery area is implicated as the active site responsible for the increased reaction rates. A typical reaction rate behavior exhibits a maximum with increasing oxide coverage over a transition metal catalyst. The oxide alone is inactive while the metal is active for methane formation, for example, from CO and H2. At about SO, of a monolayer of oxide coverage, which corresponds to optimum oxide-metal interface area, the reaction rate exhibits a maximum. This large oxide-metal interface catalysis effect is observed for several transition metals including Ni, Rh, Co and Fe and for several oxides including Ti02, La203, MgO and Zr02 In addition to CO activation, other molecules that have CO bonds (acetone, alcohols, C0-J are also activated for hydrogenation. This oxide-metal interface activation phenomenon is under intense scientific scrutiny in many laboratories and several new catalyst systems have been reported and patented based on SMSI.

468

11.5 CONTROL OF SECONDARY REACTIONS DURING CO HYDROGENATION BY CONTACT TIME. BIMETALLICS AND ZEOLITES Whether CO is hydrogenated in its molecular state or after dissociation its carbon picks up hydrogen it has to form several C-H bonds before it may desorb from the catalyst surface. Hydrogenation occurs sequentially and reduction of its rate can markedly alter the product distribution. This occurs because the adsorbed intermediate can interact with other CH, fragments to form C-C bonds to produce C,

... C,

hydrocarbon products. Olefins that form may undergo

hydroformylation picking up CO and hydrogen to produce one carbon atom longer aldehydes and alcohols. Alkali metal additives reduce hydrogenation rates as mentioned above thereby leading to the formation of longer chain, higher molecular weight and less saturated organic molecules. Molybdenum sulfide catalysts promoted by alkali were found to produce methanol and longer chain alcohols while rhodium compounds in which some of the transition metal species are in their higher oxidation states produce acetaldehyde, propionaldehyde and the corresponding C2 and C, alcohols. When zeolites are utilized either as supports for transition metal catalysts or as a physical mixture with transition metal catalysts, the primary products, methanol, ethylene and propylene can undergo zeolite pore-catalyzed secondary reactions. Depending on the pore structure aromatic molecules (benzene, toluene) or straight chain hydrocarbons in the C8-C15 range can be produced, both product mix are excellent synthetic fuels or desirable chemicals. If the aim is to produce small molecules by CO hydrogenation, the inhibition of secondary reactions is necessary. This may occur at high space velocities that give rise to short contact times with the catalyst. The introduction of a second metal or alloy with low heat of adsorption for CO and/or H2 or the CH, intermediates that form could also be used to decrease the surface residence time in order to minimize the secondary reaction probability. Copper or gold have been used as metallic additives for this purpose.

11.6 FUTURE DIRECTIONS OF RESEARCH The depletion of oil reserves, which are non-renewable sources of fuels, will result in increasing oil prices. Although there may be disagreement over when synthetic gas based fuel or chemical technologies become economical, there is agreement that they will become economical in future years. Thus, research into its use to produce fuels and chemicals selectively and at high conversion must continue. There are several directions that deserve focussed research. The role and mechanisms of CO, hydrogenation should be a fruitful avenue for research. CO, hydrogenation is implicated as an important reaction path in the formation of methanol. The mechanism of the water-gas shift reaction and steam reforming both deserve major research efforts as they are important reactions for CO/H, based technologies. Heterogeneous and selective hydroformylation and carbonylation (methanol to acetic acid) should yield new technologies to produce chemicals. The production of nitrogen atom containing molecules has yet to be explored. Polymer formation from CO and H2 directly could be an interesting direction of research for the future as well. Building our future fuel and chemical technologies from syn-gas is an exciting challenge that will

469

require the joint efforts of scientists carrying out basic research and engineers who develop and build the new processes.

11.7 REFERENCES 1

Somorjai, G.A., Chemistry in Two Dimensions: Surfaces (Cornell University Press, (1981).

2

Dry, M.E. in “Catalysis. Science and Technology,” ed. J.R. Anderson and M. Boudart 1, 159, Springer Verlag (1981).

3

Bartholomew, C. in “Hydrogen Effects in Catalysis,” 543 ed 2, Paal and P.G. Menon, Dekker (1988).

4

Chinchen, G.C., Denny, P.J., Jennings, J.R., Spencer, M.S., and Waugh, K.C., Appl. Cat. 36, (1988) 1 Biloen, P. and Sachtler, W.M.H., Adv. Catal. 30, (1981) 165 Bell, A.T., Catal. Rev.-Sci. Eng. 23, (1981) 203 Koel, B.E. and Somorjai, G.A. in “Catalysis. Science and Technology,” 7, 159 ed. J.R. Anderson and M. Boudart, Springer Verlag (1985). Somorjai, G.A., Crowell, J.E., andGarfunke1, E.L., Surf. Sci. 121, 303 (1982). Lackey, D., Surman, J., Jacobs, S., Grider, D., King, D.A., Surf. Sci. 152/153 (1985). Mate, C.M., and Somojai, G.A., Surf. Sci 160 (1985) 542. Blackman, G.S., Lin, R.F., Van Hove, M.A. and Somorjai, G.A.,Acta Cryst B43, (1987) 368Van Hove, M.A., Lin, R.F., and Somorjai, G.A., J. Am. Chem. SOC.108, (1986) 2532 Somorjai, G.A., Ohtani, H., Bent, B.E., Mate, C.M., and Van Hove, M.A., Applied Surface Science 33/34, (1988) 254. Somorjai, G.A. and Bent, B., J. of Advances in Colloid and Interface Science 29, (1989) 223 Somorjai, G.A., Crowell, J.E., and Tysoe, W.T., J. Phys. Chem. 89, (1986) 1598 Broden, G., Gafner, G., and Bonzel, H.P., Surf. Sci. 84, (1979) 295 Benziger, J., and Madix, R.J., Surf. Sci. 94, (1980) 119 Kiskinova, M., Surf. Sci. 111, (1981) 584 Kelemen, S.R., Kaldor, A,, and Dwyer, D.J., Surf. Sci. 121 (1982). Berko, A. and Solymosi, F., Surf. Sci. 171, (1986) L498 dePaola, R.A., Hrbek, J., and Hoffmann, F.M., J. Chem. Phys. 82, (1985) 2484 Luftman, H.S., Sun, Y.M., and White, J.M., Surf. Sci. 141, (1983) 82 Hoffmann, F.M. and dePaola, R.A., Physical Review Letters 52, (1984) 1697 Weimer, J.J., Umbach, E., and Menzel, D., Surf. Sci. 155, (1985) 132 Whitman, L.J. andHo, W. J. Chem. Phys. 83, (1985) 4808 Solymosi, F. and Berko, A., J. Catalysis 101, (1986) 458 Solymosi, F. and Pasztor, M., J. Phys. Chem. 90, (1986) 5312 Uram, K.J., Ng, L., Folman, M., and Yates, J.T., J. Chem. Phys. 84, (1986) 2891 Kelley, R.D. and Goodman, D.W., Surf. Sci. 123, (1982) L743

5 6 7

8 9 10

11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

470

27 28 29 30 31 32 33

Goodman, D.W., Accts of Chem. Res. 17,(1984) 194 Wojciechowski, B.W., Catal. Rev.-Sci. Eng. 30, (1988) 629 Sexton, B.A. and Somorjai, G.A., J. Catal. 46, (1977) 167 Dwyer, D. and Somorjai, G.A., J. Catal. 56, (1979) 249 Schwab, G.-M. in “Catalysis. Science and Technology, Ed. J.R. Anderson and M. Boudart

2,1, Springer Verlag (1981). Tauster, S.J. andFung, S.C., J. Catal. 55, (1978) 29 Levin, M.E., Salmeron, M., Bell, A.T. and Somorjai, G.A., J. Catal. 106, (1987) 401

47 I

Subject Index acetaldehyde

250,278, 318, 383,390,408, 420424,431. 434,436, 438-440,442, 469

acetic acid

250, 382-383, 389, 392, 396, 400,404, 407-409,422423,431-441, 457,469

Acetic anhydride

383,400,404, 409,410,423, 431432,43444 1

Activated hydrogen

250-251, 360361,363-364, 367,370,372

Activation

1-3, 19,2628,32,468 barrier 4647,70-72, 76-77, 145 for CO dissociation 76-77 for carbon formation 109-111 over bimetallics 246-247 of co 87-89, 117, 144, 145-146, 226,229,233, 253,316 homogeneous 381-385,387, 396,423 energy of CO reaction 89,92-98, 101, 103, 106-107, 119, 126, 170, 173-174, 190191,238,246, 247,251, 310, 467 energy of chemisorption 46-47,77, 352,360-363, 372-373, 375376 heat of adsorption and activation 355-358 113 for methane formation

period of

Activity alloys of cobalt of chromia Cohoride C O Nsurface decrease by carbide effect of additives impurities on

136

185, 186,229 174,370-373 268 182 100-101 98

186,436,465 103-108, 112113,375 treatment on 190, 192 acidity 198 carbon formation on 203,209 315-316 alkali Fe/alumina 176 Fe/titania 178 Feborides 185 of hydrogenation 233-235,246, 247,415 of hydrofotmylation 443-444,445 formethanation 95,121,248, 252 for methanol synthesis 122, 126, 127, 128,271, 275-276,289, 297, 329, 438-439 of fused magnetite 168 nitrided catalyst 181 NiMoria 188 for oxygenate 130 of Pd/SiOZ 126 PdFe 37 1 relationship to structure 169 retarded by CO, 3 11 Ru/Au 237-241 specific, of iron 162, 170, 173, 190 steady-state 162 in surface reaction of CO 75, 77-78, 170, 172, 186, 197

472

WGS reaction

163, 172, 179

Activity pattern

118, 128, 135, 147

Additives basic effect of

159 166 162-163,464, 468-469 75-77, 89,98101, 103, 105 77,89,99, 103, 105 186,209 226-227 251,404

Electronegative Electropositive gaseous (Cl, S) modifications by promotion by Adsorption

8, 10-13, 1619,23-24,26 binding energies 48-49,52,55, 57,67,99, 163,285 co466 complex 38 of molecular CO 51.78 of co 118-120, 130134, 173, 187, 242-249,253 of CO, non activated 56 of coz 274,278,333 of c u film 97 of electron donating species 137-140 127, 166, of hydrogen 173-174, 192, 209,351-356, 365 362-370 activated of [FJtC1,]2240 of methanol 337 of oxygen 172 decrease of CO in impurities 99-101 heat of 353, 356-358, 361,465,469 inorganic precursor 236-237 probability 49,50 properties of metal cluster 175 purification by 447 rate 49,99

of Ru-0s cluster site configuration states suppression of CO strongly bound temperature Absorption spectroscopy

230-231 39,54,56,57, 60.74.75, 143-146 40-41.56, 65, 104-105, 112 123 362,373 47-49.5 1.55 41

Affinity of Ag and Cu to carbides 246 level in CO adsorption 42,64,73 metal to oxygen 130 of RUto H2 249 Alcohols

268-272,274278,281-282, 285,287-291, 293-299, 303, 308-312 as nucleophilic reagent 386-388,401, 404 branched 325,335 carbonylation of 409,41 I, 420, 422,431-432, 442-444 desorptionof 333 in hydroformylation 464,468-469 industrial synthesis of 398 by bride, nitride 183-185 production of 127-130, 180181,206, 314-318, 321322, 327-331, 335-339,382, 390,413-416, 439,447-449 on bimetallics 250 on Co/thoria 189 on K promoted MoS2 254 pathway to 146

Alcohol formation

37 1, 376,464, 451-454

Alcohol synthesis

227, 268-269, 276-280,285, 288-290.292. . ~, 295-296,307, 309

473 Aldehydes, unsaturated

147

Alkadienes

386,404-407, 416

Alkali doped methanol synthesis 288-291 iodide 438 ions 138-139,272, 316 promoters 28,77-78, 125-129,271272,274-276, 421.434.435. 464-469 mechanism 143-146 promoted Mo/carbon 176 MoS~ 184,254, 311-312, 314, 315, 325, 328, 331,333,336 cmo 247,283,296, 299,330,334 RU 395 293-294,318 Zn/Cr2Og Ru/zeolite 197- 198 Alkanals Alkanes carbonylation of formation of olefin to production of C2+

422

iron based in methanol formation from organometallics Cu/Ni RuIAu Ru-CU RuET Ru/Ni Rar sulfided surfaces of Binding energies Bonding of carbonyls of CO in transition metals CHOH complex C 0 2 to metal mode of CO Rh-Fe-0 state of hydrogen effect of

42 1 407,467-468

Alkanols

385386,392, 395396,407, 420

Alkynes

382-383,385386, 398, 400-402,413, 424

Alloys

33,132

Antibonding states

64,104

Bimetallic CO activation on CO interaction with effect of support on genesis of H2 interaction with

246 242 236 227,231 244

60, 62-63,707 1.75-76,79 444 104,111 247 465 242-244 2.50 351-355.371, 375 367

c2

alcohols

444

178-179, 182

184-185,249250,371 135-136 232 247 238-240 247 24 1 244 25 1 234 95-97

alkali dopant in C& from CH, Co for C2 growth step esters on Fe/Co on Fe, Co, Rh formation of homologation of olelin oxygenate yields for Cu

266, 319, 331-332,339 299 469 327-329 335 295 250 467 93-94, 135, 196,255 314 235 25 1 248

c2+

alcohol rich alcohol formation on bimetallic

299,301 295 186-189

474

Fdcarbide Mo/silica on Ru influenced by S oxygenate selectivity for

c3

c4

177-179 174 167 168,184 242,294,298, 299,336,338 161-163, 169171, 182, 246-248,303, 368-370

93-94, 155 addition of C1 to 325 conversion to higher hydrocarbons 208 cryogenic separation 204-206 from C02 hydrogenation 469 production of 163-164, 166 onFe/hWh/KJ 167-168 by K-promotion 180-18 1 on Co/zeolite 194 on Fe, Co, Rh 467 enhanced by alloy 185-186 alcohol 311,321 olefins 327-328 esters 295 selectivity 250,300,306, 307 growth step 335 94,208-209 alcohol 266,267,292, 298, 305, 321, 324,328,338 aldehydes 295,325 cryogenic separation 204-206 olefin 165, 168, 185, 189, 191,235 oxidation of 434 selectivity 164, 166, 169, 306,310 increased on Ni/Ce02 172 increased on Ru/Ti02 177 increased on Co/I%02 196

Carbonylation

278,280,291, 296,317-320, 331, 336, 382-389,392, 396, 398-400, 402404,406413,421-424, 429,431-442, 449-459,469

Carboxylic acid

Catalysts acid cluster derived carbonyl deactivation design development transition metal

in zeolite multimetallic

cobalt

boride iron

oxycarbide nitride boride FeReK molibdenum MoS2 promoted sulfur resistant nickel

palladium based ruthenium RuFe homogeneous Rh

388-399,404, 409,411,418, 435,441,453, 456 456,457 165, 192,432, 433,434,435 445,447, 449452 35 1 160 160, 165, 176, 187-190, 199202,210 32,38,75, 77-79,88112, 118, 122-130,209 193, 196-198, 205,210 132, 134-136, 184-186,224248 159, 162, 167, 169, 172-175, 178,204, 207-208,210 182 163, 164, 168, 170,173-174, 1 78-1 80, 190, 373 180,210 180 182 375 172, 174,176 307-317 163, 166, 180, 182, 192, 325-330 180 88-89,161, 172 267 161, 162, 177, 208 361 381,382,384 383,388,389, 393,395,409,

475

Pd co

Ru Ru/Co Ni PdICu WSn Ag carbonylation for FTS hydrogenation hydroformylation for methanol synthesis

for higher alcohols poison for single crystal slurry Catalytic activity

of metals of single crystals of Reppe reaction alkylation carbonylation center cracking cycle

performance poisoning properties

414,416-417, 448,449 389,406,407 392,405,411, 414,415, 418423,441, 448 395,418 396-398 402,408,409, 41 1,412 404 419 442 385,403,436, 438-440 159, 169, 194, 195,467 120-121,468, 469 443,444 268-281.283285, 318-323, 330-339 287-296,298, 299-300 366,369 464 203 87-88, 121, 127, 165,236, 269,273, 301, 359, 373, 375, 394-395,421 95-96 98-101, 112113 400 199 402 382 208 254-255, 387, 389-39 1, 397, 422,433,446, 45 1 25 1 105-108 141, 162, 164-165, 175,

reactions

180, 184, 190, 197 38,46,75, 77-78,238

Catalytic hydrogenation

118, 393,422, 463-464

CH30H synthesis

120-122, 125128, 145

Chain growth in Fischer-Tropsch to higher alcohols probability secondary termination of Charge donation of the complex point transfer Chemical modifications

32, 118-119, 159-161.464 295, 329-33 1, 336-337 305,308 468 193 66,105 384 137-143 59-61.73-74, 134,174 159-160

Chemically modified surfaces 98 Chemicals base by carbonylation from coal fine organic bulk Chemisorption bond strength

co of hydrogen

stepwise Coadsorbed species

198 382,423-424 43 1 210-211 409,411,443 407.437 70,73, 134, 353,464 139 242, 35 1 ,

366-367.465 189, 197,237, 244,358-359, 361,364 239 105, 132, 139, 143

476 CO molecule

42,47,52, 55-56.66, 104,131, 144 activation 387 bonding of 2-5,743, 1314, 16, 19-20, 22,24,26, 32-33, 38,40, 42-43.49-5 1, 53,56-58, 60, 62-67,70,73, 75 chemisorption of 2, 6, 16,24, 26,31,56-57, 64-66, 69,7172 electronic structure 2-3, 6-8, 1317, 19-24,28, 32, 38,42,61, 66,73 mechanism of 4344,46,49, 57 on single crystal 46 coordinated 66, 130-131, 133, 384, 387, 390,392,392 coordination of 2-4,6-8, 1314,28,30, 32-33,384385,387 desorption of 18 38-40.4451 dissociation of 26-31,242, 250,251 distance to metal 4-5, 15-18, 21,28-30 42 functionalizing of 398 hydrogenation of 38,78-79,93, 95,97-98, 106, 120-121, 146,226-229, 232-236,238240,242, 244-248, 390, 393,464,465 insertion of 395-396 interaction with bimetallics 242 methanation of 89,92-101, 103, 105, 113, 183 crystal face dependence 16, 361 modelling of 20-21.26 on single crystal metal78

metal coupling molecular orbital of non-dissociated

15.52.63-68 8, 13,28, 143 371

Commercial application

444.45 1,452, 453,458

co2

88 130 89, 108,467 45 1 454

asymmetric vibration in Boudouard reaction in carbonylation in carbonyl amine in Cs promotion CO conversion to in deactivation in higher alcohols hydrogenation in methanol in presence of in CO2 free scrubber, removal of selectivity in WGS reaction

111 161-164, 185, 189 253 333 272,274,469 127-128.268, 270,27 1 276-280,310314 293,295, 297-299.335 204 172 106, 147, 161, 464465

co

borides, nitrides bimetallics carbides in c a h n y l

in C O W layer CO hydrogenation on

hydroformylation by single crystal sulfides supported

182-183 185-186.233, 244,247,249 180,189 280,314-315, 392,404,413, 42 1 98, 175 118, 159-169, 190, 383,389, 392, 396, 402403,405409,414415, 417-423.439, 447 45 1 173 183 169-171, 175, 178,361

477 surface phase of in zeolites Contact time

365 193-194, 196, 198 270,295-296, 299,310, 335-336,339, 467-469

Conversion of acetic acid of alcohol

of Fe2+ number on surface of co of olefin single atom Coupling aldehidic species aldol carbon-carbon dipole metal12pCO oxigenated C, oxygenated C, vibrational

50 384-386 419 373-374 66,68 78-79 403,414 353-354 318 330 397 65 64-66,73,77 3 19-320 335,339 70

cs

for higher alcohols with MoS,

CsICulZnO

293-297 305-310, 327328

110-111,272, 280,298, 299, 316, 330-331, 393,394 274-276.285, 288,319-325, 334-336

cu in bimetallic

44 1 206,420,439, 440 of co 170, 172-173, 185-186, 188, 195, 197,201, 296, 309-31 1, 319,337-339, 384,431 on CufZnO 108 to methanol 270-272 of coordinated CO 387 equilibrium for methanol 287 olefin 313 V2O5 to V203 124

Coordination adsorption site for of co of chelating ligand

promotion by

CO adsorption on LEED pattern dissociation on in electron donation in methanol synthesis positively charged as promoter Fe/Cu/K2O/SiO2 FeFInEdCulK FelCulKFlgO cu/Ru

Ni/Cu FeICu CuIZnO

alkali promoted Cs promoted with alumina modelling Culiodide ZrICu CeICu for higher alcohols MoS2 based oxidative carbonylation d-orbitals on

96-98, 135136 46,53,60,64 49 63-68,70 104

120-121 146 165, 196,400, 468469 179, 189,204, 207 190 208 227,235, 238-239.243248,301 242,247 246,465 266,268,269, 271-274,280, 281,283,284, 316 296,297 275,285,286, 287,298,299, 319-325 328-331 106-112, 127, 129 438 276 277,278 291-295 339 396-397 32-33

Deactivation of

278,279,285, 35 1

Design catalyst

269 159, 160

478

FT catalyst principles reactor and process

169 184, 194, 195, 208 199-200,202, 204,207,448

Desorption activation energy of alcohol of co

352 333 53,51,12, 465 of hydrocarbon 468 of hydrogen 355-358 of methanol 282 of oxygenated intermediates 33 1 temperature programmed 39,79, 175, 314,360-364

Dispersion 106, 162 of cluster derived catalyst 234-235 effect of metal 169-174, 176, 354,363-364 of cu 121,248,278 of Fe 373,375 of PdFe 25 1-252 of m u 24 1 Dissociation

activation barrier for of co

of hydrogen metal sites for probability 70.76

76-79,90,92, 103-104, 109110 72 70-72, 165167, 183,209, 245,249,251, 367, 373, 464-469 244-245.351352,355, 358, 360 364

Early transition metals

135, 136

Electron spectroscopy

40-42,51,59, 62-63,65, 8889, 133, 185, 373

Electron density of states

353-354

density of sdiffraction donation escape of interaction microscopy microdiffraction population spectra of carbon transfer of ligands

Electronic structure Emission of ions level of ozone spectroscopies spectra Energy of activation

371,384-385 96 183,443-444, 466 137 130, 138,364 230,238-239, 241,464 240,464 144,175 93-94 99, 104-106, 142 164 209,226,228, 255,353 40-41 289 40-41 73,143

6-8, 38,45, 68, 119, 126, 170, 113-174, 190,238,247, 251, 360-364, 372-373,376 70-13, 137of adsorption 141 21,23-24,26attractive 29 134 band barrier 370 13-19,47-53, binding 75-17 92 of hydrogen 130, 133 dissociation 140,141 Fermi 144 level 43 loss 64 position potential energy diagram 108-11 1 problems in hydrogen bonding 352-358 289 production of repulsive 31-32

479 shift of sublimation transfer

61-65,76, 84 227 40

Ethylene glycol

ESD

40,57

ESDIAD

40,56,57,58

Formation of activated hydrogen of alcohol

Esters alkoxycarbonylationto carbonylation of

CO hydrogenation to c3

fatty acid formation of formylation of homologation of methyl

synthesis of oxidative addition of Ethanol hydroformylation isotope study on kinetic study on pathway for precursor for selectivity to

on Rh on Pd on Rh on Cs/Cu/ZnO synthesis of

Ethers

404 407,409, 411-412,432, 435,449 390,392 295,297 453454 314,330,401 417-418 431,441443 298,299,306, 327, 330, 331, 335,336 321,333, 382, 395 386 420 32 1 33 1 319 320,333 122-123,314315, 338-339, 383, 391-396, 422 125,302,303 I29 250 295,297,32I 308-310, 317, 327-329,431432,438-440 383, 386, 394, 406-407,409410,417, 421-422, 429, 439

382-383,388389,391-395, 398,422-424, 432,442

370-371 121-130, 136, 254,276, 310, 314,316, 321, 328,420,451 of acid anhydride 435 of acetaldehyde 439-440 of acetic acid 431-433 of alkyl acrilate 402 of aromatics 196 of bimetallic particles 227-228,236242,247,249 of C-C bonds 104,318,319, 330,338, 387-388,467, 468 of carbonyl amine 454 of carbonyl complex 179,384,404 of carboxylic acid 449 of chemisorption bond 464 of cluster anionic species 230-231 of dimethyl adipate 406 of ether 297 of ester 335,447 of ethylene glycol 391,395 of hydride phase 3.5 1 118-119, 134of higher hydrocarbons 135,243,251, 266,309.3 16, 367-368, 373, 415 of methane 90.92-94, 103, 106, 161-163, 167, 176, 184,211, 249,254,327, 394,422 of surface carbon 92-94, 104, 119, 162, 165, 168, 185, 191, 203-204,210, 246,249 234-236 of subcarbonyls of interstitial hydride 278

480

of metal alkyl bond 414 of 2-methyl-1-propano1 295 of methylformate 316 of nickel tetracarbonyl 400 of strongly bound hydrogen 360 of pivalic acid 456 F T processes

199,200,204, 206,208

FT reactors

202,203

Fuel from coal diesel, gasoline

grade gas high octane methanol for oxygenate for

159-160 167,185, 189, 199,201, 205-207,210211,292 266 283 289

Gas phase

134, 168, 189, 230,279, 291-293, 319, 331-333, 351352,363, 441442,453

Group VIII Metals alcohol formation backdonation in bimetallics

in alcohol, effect of

310-311, 313, 317 in CO hydrogenation 168 influence on deactivation 176, 183 tolerant catalyst 252

Higher alcohol accompanied by WGS function of GHSV on Cs/Cu/ZnO on Cu/Zn0/Al2O3 Homogeneous reactions

111

463,469

blocking model StlUCtUre restriction in pores

H2S

335-336

Future directions

Geometric effect

in formation of carboxylic acid 449 in homogeneous catalysis 384,399 for Reppe synthesis 435 molecular CO, state on 4849

30,33, 38, 113,244,247 252 55,62 73

193

2-3 60 134-135,236, 240,242 binding state of hydrogen on 352-354 for carbonylation 433 CO hydrogenation 169, 172, 199, 245,247-248 in FT reactions 166

Homologation of carboxylic acid by ethylene addition of methanol on cobalt

333 338 287 298,311 381-384.389390,396, 398-399,407, 413,423-425, 428429,444. 447,458,460, 462 431432,441 412 267,317-318, 420,438-439 314

Hydrocarbons c2+

branched in CO reaction

liquid

in LPG light oxidation of

184-190,370371 198 93, 103-104,

118-121, 128129, 132-135, 138, 141, 159, 161,242, 308-311,368 159, 166-167, 176-182, 199, 201-202,209, 247-248.250251,253 160 193-199.205206 434

48 1

oxygenated intermediate to 331, 398 precursor to 333,336 yields 235,285, 367-369 Hydrocarbonylation

382-384,387388,390, 392-393, 396, 398-399,408, 413-414,420424,431, 442-444,446

Hydroformylation of

399,415, 417419,439

Hydrogen activated

250-251,280, 315,359-362 adsorption of 104,106, 111-112, 119, 130, 135, 163, 166,190,210, 237,253,278, 352-359 in carbonylation 385-387 carbon removal in 90 coverage of 92-93, 120, 177 deficient carbonaceous 88,163, 165 halide 400401,411, 452 in hydroformylation 469-470 interaction of, with bimetallics 244-245,249 saturation of hydrocarbons with 464

Hydrogenation activity of acetic anhydride of alcohols of co

on single crystal

165,444-445 441 420 121, 161, 163, 169, 173-176, 178, 183-189, 193, 198, 199, 210,226-240, 247-248.253254,272,351, 368,382-383, 464,467-469 90, 92-94

homogeneous

388-392,398399,408, 413415 centers for 125 of CHO group 324 of diesel fraction 207 function 316-317 of intermediates 146-148 of surface carbon 167-168 on Cu/ZnO 274 on MoS2 303 438-439 on Ru, Pd of oxygenated intermediates 336-337 119,130 rate of rate constant for 331-332 IB metals

in bimetallics weakly bound CO on

134,242,246, 252,352-354 63,66

Image forces

137

Impurities cNoride effect of

89 234-235.237 98-10, 103, 105,372 142 125

ionized in SiO2 Industrial application

Interaction between adsorbates in bimetallics of CO with bimetallics of CO-co cluster/support electron

382,395,399, 405,420421, 430,434,436, 442,444

105 373 241-242 465-466 230-231 95, 100, 130132 of hydrogen with bimetallics 244,355 metal-support 89, 105, 123, 125, 164165, 169, 173.175, 198,358,366, 369,468

482

WNi of Fe/IB of Fe/Pt promoter/metal oxide/support Interception of intermediates

Interface metalhpport promoter/metal Intermediate acyl acyl metal complex adsorbed aldehydic dimethylether dioxygenated formate formyl hydroxymethyl interception to methane in methanol formation

methyliodide methylacetate IR

103 246 250-353 137-141, 148, 367 191 159-160, 192, 194, 196, 199, 210 124, 134 250,468 165,174,188 388-390 456 469 325 317 329 464 254,392 39 1 159-160, 168, 192,194-196 209-210 126-127, 145148 433 436 38,41,44-45, 47, 60.72, 75, 131, 134, 145, 152, 163, 177, 25 1,394,446

Ir

activity in C O B 2 119 CO dissociation on 129,135 in non activated CO adsorption 51,53,63-65, 68 inmethanol formation 120-121, 125 promoter on 141, 146 in Reppe synthesis 435 Kinetic

of adsorption Anderson-Schulz-Flory of CO hydrogenation parameter for desorption model

studies on impurities on WGS reaction

163 160-161,192194,208-209 173,467,468 362 95,287,329339 100-101, 103 107, 109-111

Koch reaction

456-458

LEED

18-19,42,48, 53,55-56,61, 77,96-98, 131,465,466

Limitations of chain growth

159-160, 193, 199-200 of I Treactor 203 of FT technology 208 technical, of I T synthesis 214

Mechanism of alcohol synthesis of alkene carbonylation

316-320 403,414, 435-436,45 1452 of C-C bond formation 387-388 of chain growth 119,160 of CO activation 145-147, 190, 384 of CO dissociation 73-78,351 of CO insertion 328-330 of C 0 2 hydrogenation 469 of crystal growth 243 of higher alcohol formation 128-129, 132 of hydroformylation 445 of methanation 92-93,467 with impurities 103-104, 107108 of methanol synthesis 270,278 of metoxide transfer 397 of methanol carbonylation 408-409,421 of site blocking 245,247-248 of WGS reaction 111

Metal additives

300-301

483

alkali alkyl acyl metal complex carbonyl cluster

clean metal surface crystallite growth electronic structure ensemble hydride hydrogen adsorption on loading, effect of orbitals oxide, promotion poisoning of single crystal support interaction

in zeolites Methanation effect of sulfur metals for Ru for Ru/Au for Rh for decreased Fe/Ni for on modified surface MoS~ poison for on single crystal Methanol to acetic acid carbonylation of

deactivation of fuel from to gasoline

316-318,421, 464-467 414 456 163, 176, 188, 192-194,401404,409-410, 441-443,454 89,93,465 181-183 280 172 384-392 352-356 169-170, 175 103 164-166,468469 98-99 95-96 88-89, 104106, 123, 169, 174 196-199 252 182-184, 188 247-248 237 24 1 249-250 98, 100-104, 125 303 168 88-90.92-97 250-25I 400-401 318, 391-396, 407-409,420422,431-432, 434,437-439, 442 285-289 178 205

non dissociative CO for formaldehyde from formation of

242 266 308-311,314315,464,469 modelling 325-327 for interception 168 homogeneous synthesis of 280 homologation 317 labelled 319-327 reaction with nitric oxide 398 supported Pd 279-280 synthesis of 88, 146-147, 194,267, 331-335,382383 synthesis catalyst 268-279 over acidic site 253-254 on Pd 371-372 promotion on 126-127 on Pt 122, 129, 135 on single crystal 111-112 technology 281-284

Methylacetate

435436.438

Mobility

235,237,245, 250.417

Modelling of cluster, calculation for FT reactor for practical catalysts for promotion kinetic of methanol formation MoS~ alcohol synthesis on alkali doped

iron based modelling of rare earth promoted

353-354 203 112-113 144 287.21 1 325-327,329331, 333-339 303-307.308316 291,293,294, 325-331, 333334 253-254 336-339 184

Modifications

175

Nature of carbonaceous residue chemical

93 170

484 metal precursor 236 product 196,197 surface segregation 228 support in genesis of catalyst 230,231,234

Oxide dispersion of ethylene in FT reaction role in C a n 0

Olefins

172, 177, 180-181, 198199,201,301, 306,368-369 carboxylic acid from 456 carbonylation of 449-453 from C 0 2 189-196 double bond isomerization 415-417 hydrofomylation 399,403-404, 417,419,431, 442445,447448,469 increase on Fe 165,170,250 light in FT 164,186,203, 310,312 oligomerization 208-210 paraffin ratio by Au 248 saturation of 205-206

Organic halides

41 1,454,456

Orientation

20,98, 130, 144.355

Overlayer alkali carbidic Cu on ZnO of Co on W(100) c1, s, P, Ni Oxidation of acetaldehyde active sites in

co homogeneous of ethylene of iron species of metal sites of methanol of propionic aldehyde of propylene stability against

iron higher alcohols on in methanol synthesis mixed Re-Fe phase nitric oxide organometallic on oxide/metal interface promoter site for CO activation styrene Zn-Cr-K 0x0 process Oxygenates C2+ promotion formation of

467 95 112,243-244 98,173 100 106 2,127 434,436 396,464,469 363,383-384 396,398 420,440 23 1 234 442 432 453 180-181

synthesis improved

106-107 442 119-120 128-129,269, 271,275 180,191,246 293,317 122-124,129 375 398 230-231,233 468 122-124,135136, 165-167 126 410,411 330 415,431, 442443 128-129,242, 294,295 123-125, 134135, 141, 195, 201,205,251, 253-254,280, 306,325,336 120-121,275, 330 146

Pd alloying chemisorption of CO in carbonylation in cluster coordination sites on d-valence electrons in ethanol synthesis in methanol synthesis role of ionic species PdZr PdIAg PdiRu Pd/Ni

133-135 105 439442,450452 232 7, 16 21, 31-33 129 111-112, 119121 125-127 236 242-245 245 245

485

FePd PdHCl in Reppe synthesis titania supported

250-251, 371372,398 407 435,436 177

Phases carbide charge transfer between Mo oxide

180-181,368 140,142 174

Perturbations electronic

105

Poisoning effect of rate of by sulfur Position of double bonds CO on atop CO on bridge of CO on surface H, in various hydrogen at subsurface of TPD Pretreatment by calcination effect of

with CO variations in

168 98-101, 103105, 109 94 183-184,226, 252 405,443,444 8, 13, 16, 1920,26-28 8, 16, 17, 3233 132 353-355 360-363 365 190-191 168, 186-187, 230,235-236, 253-254, 367, 373 178-180 209

Process adsorption aggregation

FT effect of additives in FT in high wax operation hydration of MgO liquid phase slurry to LPG multistage BASF

Celanese Easman Kodak Hoechst ICI for methanol LPMEOH Lurgi for methanol Monsanto

Ox0

Synthol Wacker Wacker-Hoechst Production of acetic acid acetic anhydride acrylic acid aromatics n-butyraldehyde carboxylic acid c2 c2+

Probability propagation

to aliphatics chain growth CO insertion

161, 164, 169, 185, 192, 207-208,268. 355,466-467, 469 355,358 237

co2 of diesel fuel ethylene glycol

formate of hydrocarbons

184-185 192-195,330 328 159-160 169 204-211 236 285 176, 181 392 291,404405, 407410,453454 434 436-437 439 269,271,281 283,284 270,281 382,383, 407-410,432433 415416,444445,447448 28 1 397-398,440, 442 42 1 434-436 436438 400,452 195 447 453,457 255 163,247 253 206 431432,442444 338 103-104, 106, 159-160, 166169, 199, 201-203, 308, 468

486

of oxygenates

linear alcohol low boiling aldehyde of methane metal catalyst methanol methylformate olefin premium grade fuel of syngas vinyl acetate Promoter Ce02

cs CuCl* effects

in FT iodide ionic hydrogen halide metal interaction oxide in oxygenates phosphines potassium and support rare earth Promotion of carbonylation chemical by Cs modelling by oxides by potassium

31, 111-112, 124-125, 127, 129, 181,291, 292,303 314 415 93,95-97, 254,467 230 268,284,285 396 135,249-251 178 88-91 440 117, 159, 198 174 111,324 457 16,88, 162163, 172, 464-465 125-126 383-384,409, 432 234-235,421422 387 137-147 165.167, 188189 128-130, 135 410 103,184, 192, 247,468 121-124, 126 178-180 452 146 110 98-99 119-120, 126, 166-167.468 163-164,280, 305

by sodium for methanol theory of

183 275,316 137-138, 140, 143

Pt cluster coadsorption on CO bond to in CO dissociation d-valence electrons with Fe/carbon with Ru/zeolite.

144 20-21 32 129,135 16,26 189 198 240-241,243 pmu 244 Pt/Pd 250 PWe 358 WA1203 Pt-H 418 in production of methanol 120-123,125 single coordination of CO 131 positively charged 146 in Reppe reaction 435 96 on W(110) titania supported 177

Quantumchemical

352,354,361

Rare earths

136, 154,303, 343

Reaction carbonylation cyclization ensemble for functionalizing heat of higher alcohol hydrocarbonylation insertion methanation methanol synthesis

pathway pulse surface

385, 387-389, 398,404,407 402,404 133-134 398 283 321-323 382-383 145, 146 24 1-242 269,271-272, 274-275,281282,284,285, 300,314-319 32,306,253254 167

487 syngas

123-126, 130, 134, 146, 184, 246,249,251, 269 promotion for 143 temperature programmed 129 WGS 2, 147, 161, 268,273,309 Reactor batch for methanol Berty bubble column continuous flow development of ebullated bed

Fr GSSTF hydroformylation ICI Lurgi Pd membrane slurry Reagent Regeneration ease of FT catalyst homogeneous catalyst stability Rh atop adsorption of CO in bimetallics carbonyl cluster in H2 adsorption in homogeneous reaction impurities on in methanol synthesis in oxygenates Rm20S

Rh/ri02 Ru/Rh catalyst

273,332 195 207 276 199-204 283 194-195 282 447-448 281 282 245 189,205,284 384-385,397, 424 200,203 252 388 166- 167, 178, 209-210 4-1, 36 16-21 135,250-251 432-435.438 9-12,23-24, 231, 385, 414-416 36 1 44 1-451,454 100-103 127-129 120,122 122-125 168,111 241,243

RNCo Rh/I Rh(ll1) single crystal modelling single crystal

393-395 407-409 466 26-31 92-96

Ru alloying of atop adsorption of CO in bimetallics carbonyl

Corn2on deactivation by S effect of K in H2 adsorption in methanol synthesis for olefin conversion preparation of promoted Ru/Ti02 Ru/MgO Rulzeolite RuICu Fe/Ru RuIAu Pt/Ru Rum RulPd Ir/Ru Ru/Si02 co/Rufi Ru-H Ru/I Ru phosphine complex single crystal Secondary reactions alkane formation of aldehydes dehydration in F T

135, 166

I , 16 184-186 395-396,45045 1 91-98, 121, 161.163, 169 101, 103, 105 165-167 359-360 124-125 208 189-198 127-128 177-179 177 197 226-227.239, 24 7 231,233,249, 35 1 237-240 240,24 1 24 1 245 25 1-252 366,369,371 396,434435, 440 418 422,442-443 439-440 173,235,242 135, 159, 176 415 4 14 253 193,467-469

488

formation of alkyl formate in 392 hydrogenolysis 248 on zeolite 195 Selectivity acetaldehyde acetal acetic acid acrylic acid to a-olefin

420 422-423 434 400,453-454 165, 167, 169, 190,250 on bimetallics 186 carbonylation 398-400,413418 on cluster derived catalyst 235-236 c2

co2 effect of impurities on effect of K effect of support on ethanol on F e b r i d e glycol to hydrocarbons limitation for liquid fuel methanol 2-methyl-1-propano1 methyl acrilate pattern for premium fuel pivalic acid shape in syngas reaction towards oxygenates

394 174 98 103,164, 165, 195,465 111 438-440 185-186 389,391-293, 442-443 135, 162, 171-172, 179, 248,338 161 199,203-206 274-276,279, 308-317, 329, 431-432 325 402 118, 119 160 457 159, 176, 192-196 117-119, 122123, 163, 367-373 125-130, 181, 250,25 1, 254, 266,285,287, 294,295,296, 298,305,306

Single crystal surfaces CO adsorption methanation over methanol synthesis on Ru(000 1) WGS reaction on Single crystals

Species adsorbed acylic alcoholic aldehydic atomic carbon carbon

carbonate in carbonylation

Cs-carbonate C”, coadsorbed Co-acyl CU”+O,

electronegative for electron transfemng formate gem-dicarbonyl HNi(C0)zX ionic isotopic oxygenated C2 OX0

from oxidic support

TiO,

18, 88-89, 466 242 95-98 111 245 106-107 95-96, 121, 146, 173,235, 244,246, 35 I , 355,362,365 244-245.331333 433 318-320 320 467 89.92-94, 234,247-25 I , 253-254 278 385-386, 388, 389,390, 391, 395397,400, 404 111 119,469 132, 138-139 445 121,268 98 105-106 273,280 465 452 126,230-231, 240-242 327-329 35 1 234 123-124, 127, 129 174-175, 191, 209 107

489 Spillover Stabilization of anionic cluster of formate dispersed metal of intermediates orbitals

244-245, 3.58, 366.367 394,395 147 366 117, 126, 130, 146, 147 14,24,28

Stoichiometry

358,365

Structure for bimetallics

150 184,228-231, 233-234,23724 1 361 32, 34,209, 226,354 97-98.467 469 169 195 92-93,95, 107,248-249, 352,465 172-173, 174, 190, 243-244, 464-465 193-195

change at Hg adsorption electronic insensitive Pore relationship to activity selectivity relationship sensitive

surface

zeolite

Sulfur charge density induced by containing cluster effect on methanation as modifier poisoning by

resistant tolerant catalyst

support acidity of basicity cluster-support interaction effect of

105 394 98-103, 186 167-169,252 109-110, 146, 168,226,229, 233-234,269, 285,303 180, 182-184, 210-211 176 159, 162, 164 297 234 23 1 169-178, 172, 177, 187-189,

227,280, 301, 367-368, 370, 464 metal-support interaction 105-107, 141, 145-146, 166167, 169, 173-174,236, 358-359.468469 modelling of 140-141 326,237, MgO 239-241 nature of 111-113,231 as promoters 121-129, 181 titania 244,241 247, 250 302 Nb05 porthole mechanism 364 ~

Surface alloy carbon chromia crystallography Cu(ll0) metallic CU

of the WCU/MOS~ double layer jellium metal

simulation modelling segregation modification Pd potential Rh(l11) Ti02 Syngas reactions

Synthesis of acetaldehyde

133,229 118-119,467 274 131 465 127-128,271, 272,275,277, 278 327 141 139 2-4,8, 88101, 121, 123-126, 131, 132, 137, 141-143.233256 6-8 13-25,27-32 133, 134,227 226-231,293 280 140 466 468 117, 118, 123, 127, 133, 148, 3.51,368 431-432.438

490 acetic acid, and acetic anhydride

407-409,421, 434-435 of adipic acid 405 carboxylic acid (Koch reaction) 456-457 91,97,102 CH4 dimethyl oxalate 397-398 ethylene glycol 392-393.395396,442-443 higher alcohols 286-300,307335,422 hydrocarbon 103,104,118, 120,135-136, 241-248 of methanol 88, 111-112, 120, 121, 123, 125-127, 145, 226,266-280, 281-285, 382383 Na-L-glutamate 418 oxygenate 120, 122, 129, 146 on MoS2 301-307 OX0 413 vinyl acetate 440

Temperature programmed

231,234,238, 361,362

Theories adsorbate perturbation 105 in alloy segregation 133 of promotion 137, 138 promotion in syngas reaction 143-146 TPD

Transfer charge, in alloys between phases of carbonyl species of electron of hydrogen of oxygen phase

98,249,274, 314,360-370, 375 134 140-142 397 106, 140, 141 364 147 410411,413, 444,456

Vinyl acetate

431,434,438, 440

Work function

3, 14-17,26, 28,33,98, 138, 140, 141, 143 44,47,59-60, 74,79

CO induced Zeolites Cu/Ru in metal particles in

458,464,469 247 226

491

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

Volume 1

Volume 2

Volume 3

Volume 4

Volume 5

Volume 6

Volume 7

Volume 8 Volume 9

Volume 10

Volume 11

Volume 12

Volume 13 Volume 14

Preparation of Catalysts I.Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the First International Symposium, Brussels, October 1417, 1975 edited by B. Delrnon, P.A. Jacobs and G. Poncelet The Control of the Reactivity of Solids. A Critical Survey of the Factors that Influence the Reactivity of Solids, with Special Emphasis on the Control of the Chemical Processes in Relation to Practical Applications by V.V. Boldyrev, M. Bulens and B. Delmon Preparation of Catalysts II. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Second InternationalSymposium, Louvain-la-Neuve, September 4-7,1978 edited by B. Delrnon, P. Grange, P. Jacobs and G. Poncelet Growth and Properties of Metal Clusters. Applications to Catalysis and the Photographic Process. Proceedings of the 32nd International Meeting of the Soci6te de Chimie Physique, Villeurbanne, September 24-28, 1979 edited by J. Bourdon Catalysis by Zeolites. Proceedings of an International Symposium, Ecully (Lyon), September 9-1 1, 1980 edited by B. Irnelik, C. Naccache, Y. Ben Taarit, J.C. Vedrine, G. Coudurier and H. Praliaud Catalyst Deactivation. Proceedings of an International Symposium, Antwerp, October 13-15,1980 edited by B. Delmon and G.F. Frornent New Horizons in Catalysis. Proceedings of the 7th International Congress on Catalysis, Tokyo, June 30-July 4, 1980. Parts A and B edited by T. Seiyama and K. Tanabe Catalysis by Supported Complexes by Yu.1. Yermakov, B.N. Kuznetsov and V.A. Zakharov Physics of Solid Surfaces. Proceedings of a Symposium, Bechyiie, September 29-October 3, 1980 edited by M. UzniEka Adsorption at the Gas-Solid and Liquid-Solid Interface. Proceedings of an International Symposium, Aix-en-Provence, September 2 1-23, 198 1 edited by J. Rouqueroland K.S.W. Sing Metal-Support and Metal-Additive Effects in Catalysis. Proceedings of an International Symposium, Ecully (Lyon), September 14-1 6, 1982 edited by B. Irnelik, C. Naccache, G. Coudurier, H. Praliaud, P. Meriaudeau, P. Gallezot, G.A. Martin and J.C. Vedrine Metal Microstructures in Zeolites. Preparation - Properties - Applications. Proceedings of a Workshop, Bremen, September 22-24, 1982 edited by P.A. Jacobs, N.I. Jaeger, P. Jird and G. Schulz-Ekloff Adsorption on Metal Surfaces. An Integrated Approach edited by J. Benard Vibrations at Surfaces. Proceedings of the Third International Conference, Asilomar, CA, September 1-4, 1982 edited by C.R. Brundle and H. Morawitz

492 Volume 15

Heterogeneous Catalytic Reactions Involving Molecular Oxygen by G. I.Golodets

Volume 16

Volume 7

Volume 8

Volume 19

Volume 20

Volume 2 1

Volume 22

Preparation of Catalysts Ill. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Third InternationalSymposium, Louvain-la-Neuve, September 6-9,1982 edited by G. Poncelet, P. Grange and P.A. Jacobs Spillover of Adsorbed Species. Proceedings of an International Symposium, LyonVilleurbanne, September 12-1 6, 1983 edited by G.M. Pajonk, S.J. Teichner and J.E. Germain Structure and Reactivity of Modified Zeolites. Proceedings of an International Conference, Prague, July 9-13, 1984 edited by P.A. Jacobs, N.I. Jaeger, P. Jirb, V.B. Kazansky and G. Schulz-Ekloff Catalysis on the Energy Scene. Proceedings of the 9th Canadian Symposium on Catalysis, Quebec, P.Q., September 30-October 3, 1984 edited by S. Kaliaguine and A. Mahay Catalysis by Acids and Bases. Proceedings of an InternationalSymposium, Villeurbanne (Lyon), September 25-27, 1984 edited by B. Imelik, C. Naccache, G. Coudurier, Y. Ben Taarit and J.C. Vedrine Adsorption and Catalysis on Oxide Surfaces. Proceedings of a Symposium, Uxbridge, June 28-29, 1984 edited by M. Che and G.C. Bond Unsteady Processes in Catalytic Reactors by Yu.Sh. Matros

Volume 23

Physics of Solid Surfaces 1984 edited by J. Koukal

Volume 24

Volume 25

Volume 26

Volume 27 Volume 28

Volume 29 Volume 3 0

Volume 3 1

Volume 32 Volume 33 Volume 34

Zeolites: Synthesis, Structure, Technology and Application. Proceedings of an International Symposium, Portoroi-Portorose, September 3-8, 1984 edited by B. Driaj, S. HoEevar and S. Pejovnik Catalytic Polymerization of Olefins. Proceedings of the InternationalSymposium on Future Aspects of Olefin Polymerization, Tokyo, July 4-6, 1985 edited by T. Keii and K. Soga Vibrations at Surfaces 1985. Proceedings of the Fourth InternationalConference, Bowness-on-Windermere, September 15-1 9, 1985 edited by D.A. King, N.V. Richardsonand S. Holloway Catalytic Hydrogenation edited by L. Cerven9 New Developments in Zeolite Science and Technology. Proceedings of the 7th InternationalZeolite Conference, Tokyo, August 17-22, 1986 edited by Y. Murakami, A. lijima and J.W. Ward Metal Clusters in Catalysis edited by B.C. Gates, L. Guczi and H. Knozinger Catalysis and Automotive Pollution Control. Proceedings of the First International Symposium, Brussels, September 8-1 1, 1986 edited by A. Crucq and A. Frennet Preparation of Catalysts IV. Scientific Bases for the Preparation of Heterogeneous Catalysts. Proceedings of the Fourth International Symposium, Louvain-la-Neuve, September 1-4,1986 edited by B. Delmon, P. Grange, P.A. Jacobs and G. Poncelet Thin Metal Films and Gas Chemisorption edited by P. Wissmann Synthesis of High-silica Aluminosilicate Zeolites by P.A. Jacobs and J.A. Martens Catalyst Deactivation 1987. Proceedings of the 4th International Symposium, Antwerp, September 29-October 1, 1987 edited by B. Delmon and G.F. Froment

493 Volume 35 Volume 36

Volume 37

Volume 38

Volume 39

Volume 40

Volume 4 1

Volume 42 Volume 43 Volume 44

Volume 45 Volume 46

Volume 47 Volume 48

Volume 49

Volume 50

Volume 5 1 Volume 52

Volume 53

Keynotes in Energy-RelatedCatalysis edited by S. Kaliaguine Methane Conversion. Proceedings of a Symposium on the Production of Fuels and Chemicals from Natural Gas, Auckland, April 27-30, 1987 edited by D.M. Bibby, C.D. Chang, R.F. Howe and S. Yurchak Innovation in Zeolite Materials Science. Proceedings of an International Symposium, Nieuwpoort, September 13-1 7,1987 edited by P.J. Grobet, W.J. Mortier, E.F. Vansant and G. Schulz-Ekloff Catalysis 1987. Proceedings of the 10th North American Meeting of the Catalysis Society, San Diego, CA, May 17-22, 1987 edited by J.W. Ward Characterization of Porous Solids. Proceedings of the IUPAC Symposium (COPS I),Bad Soden a. Ts., April 26-29, 1987 edited by K.K. Unger, J. Rouquerol, K.S.W. Sing and H. Kral Physics of Solid Surfaces 1987. Proceedings of the Fourth Symposium on Surface Physics, Bechyne Castle, September 7-1 1, 1987 edited by J. Koukal Heterogeneous Catalysis and Fine Chemicals. Proceedings of an International Symposium, Poitiers, March 15-1 7, 1988 edited by M. Guisnet, J. Barrault, C. Bouchoule, D. Duprez, C. Montassier and G. Perot Laboratory Studies of Heterogeneous Catalytic Processes by E.G. Christoffel, revised and edited by 2. Paal Catalytic Processes under Unsteady-State Conditions by Yu. Sh. Matros Successful Design of Catalysts. Future Requirements and Development. Proceedings of the Worldwide Catalysis Seminars, July, 1988, on the Occasion of the 30th Anniversary of the Catalysis Society of Japan edited by T. lnui Transition Metal Oxides. Surface Chemistry and Catalysis by H.H. Kung Zeolites as Catalysts, Sorbents and Detergent Builders. Applications and Innovations. Proceedings of an International Symposium, Wurzburg, September 48, 1988 edited by H.G. Karge and J. Weitkamp Photochemistry on Solid Surfaces edited by M. Anpo and T. Matsuura Structure and Reactivity of Surfaces. Proceedings of a European Conference, Trieste, September 13- 16, 1988 edited by C. Morterra, A. Zecchina and G. Costa Zeolites: Facts, Figures, Future. Proceedings of the 8th International Zeolite Conference, Amsterdam, July 10-14, 1989. Pans A and B edited by P.A. Jacobs and R.A. van Santen Hydrotreating Catalysts. Preparation, Characterizationand Performance. Proceedings of the Annual International AlChE Meeting, Washington, DC, November 27-December 2,1988 edited by M.L. Occelli and R.G. Anthony New Solid Acids and Bases. Their Catalytic Properties by K. Tanabe, M. Misono, Y. Ono and H. Hattori Recent Advances in Zeolite Science. Proceedings of the 1989 Meeting of the British Zeolite Association, Cambridge, April 17-1 9, 1989 edited by J. Klinowski and P.J. Barrie Catalyst in Petroleum Refining 1989. Proceedings of the First International Conference on Catalysts in Petroleum Refining, Kuwait, March 5-8, 1989 edited by D.L. Trimm, S. Akashah, M. Absi-Halabi and A. Bishara

494

Future Opportunities in Catalytic and Separation Technology edited by M. Misono, Y. Moro-oka and S. Kimura Volume 55 New Developments in Selective Oxidation. Proceedings of an International Symposium, Rimini, Italy, September 18-22, 1989 edited by G. Centi and F. Trifiro Volume 56 Olefin Polymerization Catalysts. Proceedings of the International Symposium on Recent Developments in Olefin PolymerizationCatalysts, Tokyo, October 23-25, 1989 edited by T. Keii and K. Soga Volume 57A Spectroscopic Analysis of Heterogeneous Catalysts. Part A: Methods of Surface Analysis edited by J.L.G. Fierro Volume 578 Spectroscopic Analysis of Heterogeneous Catalysts. Part B: Chemisorption of Probe Molecules edited by J.L.G. Fierro Volume 58 Introduction t o Zeolite Science and Practice edited by H. van Bekkum, E.M. Flanigen and J.C. Jansen Volume 59 Heterogeneous Catalysis and Fine Chemicals II. Proceedings of the 2nd International Symposium, Poitiers, October 2-6, 1990 edited by M. Guisnet, J. Barrault, C. Bouchoule, D. Duprez, G. Perot, R. Maurel and C. Montassier Volume 60 Chemistry of Microporous Crystals. Proceedings of the International Symposium on Chemistry of Microporous Crystals, Tokyo, June 26-29, 1990 edited by T. Inui, S. Namba and T. Tatsumi Volume 6 1 Natural Gas Conversion. Proceedings of the Symposium on Natural Gas Conversion, Oslo, August 12- 17, 1990 edited by A. Holmen, K.-J. Jens and S. Kolboe Volume 62 Characterization of Porous Solids 11. Proceedings of the IUPAC Symposium (COPS 11). Alicante, May 6-9, 1990 edited by F. Rodriguez-Reinoso,J. Rouquerol, K.S.W. Sing and K.K. Unger Volume 63 Preparation of Catalysts V. Proceedings of the Fifth International Symposium on the Scientific Bases for the Preparation of Heterogeneous Catalysts, Louvain-laNeuve, September 3-6, 1990 edited by G. Poncelet. P.A. Jacobs, P. Grange and B. Delmon Volume 6 4 New Trends in CO Activation edited by L. Guczi Volume 65 Catalysis and Adsorption by Zeolites. Proceedings of ZEOCAT 90. Leipzig, August 20-23, 1990 edited by G. Ohlmann. H. Pfeifer and R. Fricke Volume 66 Dioxygen Activation and HomogeneousCatalytic Oxidation. Proceedings of the fourth International Symposium on Dioxygen Activation and Homogeneous Catalytic Oxidation, Balatonfirred, September 10- 14, 1990 edited by L.I. Simandi Volume 67 Structure-Activity and Selectivity Relationships i n HeterogeneousCatalysis. Proceedings of the ACS Symposium on Structure-Activity Relationships in Heterogeneous Catalysis, Boston, MA, April 22-27, 1990 edited by R.K. Grasselli and A.W. Sleight Volume 68 Catalyst Deactivation 1991. Proceedings of the fifth InternationalSymposium, Evanston, IL, June 24-26, 1991 edited by C.H. Bartholomew and J.B. Butt

Volume 54

New Frontiers in Catalysis, Parts A-C (Studies in Surface Science and Catalysis)

New Developments in Selective Oxidation by Heterogeneous Catalysis: Proceedings (Studies in Surface Science and Catalysis)

Catalysts for Upgrading Heavy Petroleum Feeds, Volume 169 (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

Catalytic Polymerization of Olefins (Studies in Surface Science and Catalysis)

Past and Present in DeNOx Catalysis: From Molecular Modelling to Chemical Engineering, Volume 171 (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

Catalysis: An Integrated Approach to Homogenous, Heterogenous and Industrial Catalysis (Studies in Surface Science and Catalysis)

Recent Advances and New Horizons in Zeolite Science and Technology (Studies in Surface Science and Catalysis, Volume 102)

Fluid Catalytic Cracking VII:, Volume 166: Materials, Methods and Process Innovations (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

New Trends in Co Activation (Studies in Surface Science and Catalysis)

New Frontiers in Catalysis, Parts A-C (Studies in Surface Science and Catalysis)

New Developments in Selective Oxidation by Heterogeneous Catalysis: Proceedings (Studies in Surface Science and Catalysis)

Catalysts for Upgrading Heavy Petroleum Feeds, Volume 169 (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

Catalytic Polymerization of Olefins (Studies in Surface Science and Catalysis)

Fluid Catalytic Cracking V (Studies in Surface Science and Catalysis)

Photochemistry on Solid Surfaces (Studies in Surface Science and Catalysis)

Fluid Catalytic Cracking V (Studies in Surface Science and Catalysis)

Preparation of Catalysts VII (Studies in Surface Science and Catalysis)

Past and Present in DeNOx Catalysis: From Molecular Modelling to Chemical Engineering, Volume 171 (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

Catalysis: An Integrated Approach to Homogenous, Heterogenous and Industrial Catalysis (Studies in Surface Science and Catalysis)

Trends in Catalysis Research

New Developments in Selective Oxidation (Studies in Surface Science and Catalysis vol55)

New Trends in Ethiopian Studies

Electrochemical activation of catalysis

Catalysis and Automotive Pollution Control (Studies in Surface Science and Catalysis)

Catalysis and Automotive Pollution Control III (Studies in Surface Science and Catalysis, Volume 96)

Catalysis by Metals and Alloys (Studies in Surface Science and Catalysis)

Catalysis by Acids and Bases (Studies in Surface Science and Catalysis)

Adsorption on New and Modified Inorganic Sorbents (Studies in Surface Science and Catalysis)

Recent Advances and New Horizons in Zeolite Science and Technology (Studies in Surface Science and Catalysis, Volume 102)

Electrochemical Activation of Catalysis

Studies in Surface Science and Catalysis 110. 3rd World Congress on Oxidation Catalysis

12th International Congress on Catalysis (Studies in Surface Science and Catalysis)

Catalysis by Zeolites: International Symposium Proceedings (Studies in surface science and catalysis)

Fluid Catalytic Cracking VII:, Volume 166: Materials, Methods and Process Innovations (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

Progress in Zeolite and Microporous Materials (Studies in Surface Science and Catalysis)

Catalysts in Petroleum Refining and Petrochemical Industries 1995 (Studies in Surface Science and Catalysis)

Advanced Zeolite Science and Applications, Volume 85 (Studies in Surface Science and Catalysis)

Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis

Surface and nanomolecular catalysis

New Trends in Co Activation (Studies in Surface Science and Catalysis)

New Frontiers in Catalysis, Parts A-C (Studies in Surface Science and Catalysis)

New Developments in Selective Oxidation by Heterogeneous Catalysis: Proceedings (Studies in Surface Science and Catalysis)

Catalysts for Upgrading Heavy Petroleum Feeds, Volume 169 (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

Catalytic Polymerization of Olefins (Studies in Surface Science and Catalysis)

Fluid Catalytic Cracking V (Studies in Surface Science and Catalysis)

Photochemistry on Solid Surfaces (Studies in Surface Science and Catalysis)

Fluid Catalytic Cracking V (Studies in Surface Science and Catalysis)

Preparation of Catalysts VII (Studies in Surface Science and Catalysis)

Past and Present in DeNOx Catalysis: From Molecular Modelling to Chemical Engineering, Volume 171 (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

Catalysis: An Integrated Approach to Homogenous, Heterogenous and Industrial Catalysis (Studies in Surface Science and Catalysis)

Trends in Catalysis Research

New Developments in Selective Oxidation (Studies in Surface Science and Catalysis vol55)

New Trends in Ethiopian Studies

Electrochemical activation of catalysis

Catalysis and Automotive Pollution Control (Studies in Surface Science and Catalysis)

Catalysis and Automotive Pollution Control III (Studies in Surface Science and Catalysis, Volume 96)

Catalysis by Metals and Alloys (Studies in Surface Science and Catalysis)

Catalysis by Acids and Bases (Studies in Surface Science and Catalysis)

Adsorption on New and Modified Inorganic Sorbents (Studies in Surface Science and Catalysis)

Recent Advances and New Horizons in Zeolite Science and Technology (Studies in Surface Science and Catalysis, Volume 102)

Electrochemical Activation of Catalysis

Studies in Surface Science and Catalysis 110. 3rd World Congress on Oxidation Catalysis

12th International Congress on Catalysis (Studies in Surface Science and Catalysis)

Catalysis by Zeolites: International Symposium Proceedings (Studies in surface science and catalysis)

Fluid Catalytic Cracking VII:, Volume 166: Materials, Methods and Process Innovations (Studies in Surface Science and Catalysis) (Studies in Surface Science and Catalysis)

Progress in Zeolite and Microporous Materials (Studies in Surface Science and Catalysis)

Catalysts in Petroleum Refining and Petrochemical Industries 1995 (Studies in Surface Science and Catalysis)

Advanced Zeolite Science and Applications, Volume 85 (Studies in Surface Science and Catalysis)

Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis

Surface and nanomolecular catalysis

Recommend Documents