Catalysis: Vol 17 (Specialist Periodical Reports)

Catalyis Volume 17 A Specialist Periodical Report Catalysis Volume 17 A Review of Recent Literature Senior Reporters...

Author: James J Spivey (Editor) | George W Roberts (Editor)

34 downloads 1180 Views 27MB 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

Catalyis Volume 17

A Specialist Periodical Report

Catalysis Volume 17 A Review of Recent Literature Senior Reporters J.J. Spivey, Louisiana State University, Baton Rouge, Louisiana, USA G.W. Roberts, NC State University, Raleigh, North Carolina, USA Reporters

0. Augustsson, Perstorp Formox, Perstorp, Sweden K. Badii, Vaxjo University, Vaxjo, Sweden M. Boutonnet, KTH, Stockholm, Sweden C.K. Costello, Northwestern University, Evanston, Illinois, USA N. Cruise, Perstorp Formox, Perstorp, Sweden K.M. Dooley, Louisiana State University, Baton Rouge, Louisiana, USA S. Eriksson, KTH, Stockholm, Sweden M. Faghihi, Vaxjii University, Vaxjo, Sweden S. Go"boIos, Hungarian Academy of Sciences, Budapest, Hungary J.G. Goodwin Jr, Clemson University, Clemson, South Carolina, USA C.S. Heneghan, Cardiff University, Cardiff, UK G.J. Hutchings, Cardiff University, Cardiff, UK S. Kim, University of Pittsburgh, Pittsburgh, Pennsylvania H.H. Kung, Northwestern University, Evanston, Illinois, USA M.C. Kung, Northwestern University, Evanston, Illinois, USA J.L. Margitfalvi, Hungarian Academy of Sciences, Budapest, Hungary C.A. Querini, INCAPE, Santa Fe, Argentina M . Rahmani, Vaxjo University, Vzxjo, Sweden W.D. Rhodes, University of Pittsburgh, Pittsburgh, Pennsylvania S. Rojas, KTH, Stockholm, Sweden M . Sanati, Vaxjo University, Vaxjo, Sweden S.H. Taylor, Cardiff University, Cardiff,UK

RS.( advancing the chemical sciences

If you buy this title on standing order, you will be given FREE access to the chapters online. Please contact [email protected] proof of purchase to arrange access to be set up.

I

Thank you.

ISBN 0-85404-229-6 ISSN 0140-0568 A catalogue record for this book is available from British Library

8 The Royal Society of Chemistry 2004 All rights reserved Apartfjom anyfair dealingfor the purposes of research or private study, or criticism or review as permitted under the terms of the U K Copyright, Designs and Patents Act, 1988, this publication may not be reproduced, stored or transmitted, in any form or by any means, without the prior permission in writing of The Royal Society of Chemistry, or in the case of reprographic reproduction only in accordance with the terms of the licences issued by the Copyright Licensing Agency in the U K , or in accordance with the terms of the licences issued by the appropriate Reproduction Rights Organization outside the U K . Enquiries concerning reproduction outside the terms stated here should be sent to The Royal Society of Chemistry at the address printed on this page. Published by The Royal Society of Chemistry, Thomas Graham House, Science Park, Milton Road, Cambridge CB4 OWF, UK Registered Charity Number 207890 For further information see our web site at www.rsc.org Typeset by Vision Typesetting, Manchester, UK Printed by Athenaeum Press Ltd, Gateshead, Tyne & Wear

Preface

It is my pleasure to welcome Prof. George Roberts of NC State University as my co-editor for Volume 17 of this book series. We have worked together to provide reviews of current topics in catalysis, and we trust that the subjects presented here are of interest. Catalysis continues to be applied to a wide range of chemical reactions. New applications of catalysis, new synthesis methods, and new research into the molecular level mechanisms have provided insight into the applied and fundamental processes occurring on the working catalyst. For example, Kerry Dooley (Louisiana State University, Baton Rouge, LA) reviews the catalysis of condensation reactions leading to ketones. There are a number of such reactions: aldol condensation and decarboxylation/condensation of acid and aldehydes, for instance. Despite a great deal of industrial and academic interest in these reactions, the mechanisms are not entirely clear. This is because of the change in mechanism (and product distribution) with temperature, and the complexity of the reaction. Prof. Dooley provides a detailed review of the most important reactions leading to ketones on a range of catalysts. Jozef Margitfalvi and Sandor G8bolos (Hungarian Academy of Sciences, Budapest, Hungary) provide a comprehensive review of the interaction of metal and metal ions in nanoscale clusters. They show that there are unique catalytic properties derived from the molecular interaction of these types of clusters. Their review summarizes the literature on five case studies that exemplify this type of interaction: Sn-Pt, supported Au, Sn-Ru, Re-Pt, and several Cu-containing catalysts. They discuss both oxidation and hydrogenation reactions on these types of catalysts, and provide detailed summaries of the literature, as well as examples from research in their own labs. S. Rojas, S. Eriksson, and M. Boutonnet (KTH, Stockholm, Sweden)focus on the use of microemulsion techniques for catalyst synthesis. They discuss this as an alternative to traditional methods such as impregnation, ion exchange, and use of organometalliccomplexes. One specific advantage of the microemulsion method is that it results in a typically narrow particle size distribution. This is true because the metal particle is formed without being influenced by the support. They describe the specific processes used to prepare catalysts with this technique. Jim Goodwin, So0 Kim, and William Rhodes (Clemson Univeristy, Clemson, SC) review the concept of turnover frequency, a widely used measure of catalytic reaction rates. They review various methods of measuring this property: chemisorption and isotopic tracing, for example. Their analysis also compares TOF values for structure-insensitive reactions like methanation and structure-

vi

Preface

sensitive reactions like ethane hydrogenolysis. Isotopic tracing is shown to be a more accurate measure of true catalytic turnover frequency. Catherine Heneghan, Stuart Taylor, and Graham Hutchings (Univ. Cardiff, Cardiff, UK) discuss the oxidation of volatile organic compounds using heterogeneous catalysts. This review supplements considerable work done by this research group in the past. Their review deals with both the more widely used noble metal catalysts as well as metal oxides. The different mechanisms on these two classes of materials are presented and analyzed. As the authors state, this may lead to the development of a catalyst with applicability to the wide range of VOCs that must be dealt with in industry. Mehri Sanati, Mohammad Rahmani, Khashayar Badii, and Mostafa Faghihi (Univ. Vaxjo, Vaxjo, Sweden), Neil Cruise and Ola Augustsson (Perstorp Formax, Perstorp, Sweden), and Jerry Spivey (Louisiana State University, Baton Rouge, LA) also review VOC oxidation catalysts, but focus on the deactivation processes that take place in industrial practice. Specifically, they focus on the deactivation mechanisms associated with silica and phosphorous poisoning. General mathematical models of the deactivation process are presented, and applied specifically to deactivation of VOC oxidation catalysts. Mayfair and Harold Kung, along with Colleen Costello (Northwestern University, Evanston, IL) review catalysts for CO oxidation over Au catalysts. This is an important reaction in the development of fuel processors to produce hydrogen for fuel cells. The authors discuss the unusual behavior of nanoparticles of Au, and point out that there is no consensus on the nature of the active site and the mechanism. Their review focuses on the preparation and effect of the support, the nature of the active site, the mechanism, and deactivation of these catalysts. Finally, Carlos Querini (INCAPE, Santiago, Argentina) reviews the literature dealing with the characterization of coke. The difficulty in identifying the chemical and physical properties of coke on the working catalyst are well known. The author describes temperature programmed methods, spectroscopy, and extraction methods as alternatives to characterize the structure of coke. He provides specific examples of these methods in a way that will helpful to those working in the field. The editors wish to thank the authors for the effort they have put into these chapters, and the Royal Society of Chemistry for their support. Comments and suggestions are welcome. James J . Spivey Department of Chemical Engineering Louisiana State University Baton Rouge, LA 70803 jjspivey/lsu.edu

George W. Roberts Department of Chemical Engineering NC State University Raleigh, NC 27695 groberts/eos.ncsu.edu

Contents

Chapter 1 Role of Metal Ion-Metal Nanocluster Ensemble Sites in Activity and Selectivity Control b y J . Margitfalvi and S. Gb'bolos 1 Introduction 1.1 Historical Background 1.2 Type of Active Sites 1.3 Mono- and Bimetallic Supported Catalysts 1.4 Promotion of Supported Metal Nanoclusters 1.5 Characterization of Supported Metal Catalysts 1.6 Subject of Contribution 2 Case Studies 2.1 Supported Sn-Pt Catalysts 2.2 CO Oxidation on Supported Gold Catalysts 2.3 Supported Sn-Ru Catalysts 2.4 Re-Pt/A1203Catalysts 2.5 Copper-Containing Catalysts 2.6 Other Types of Supported Catalysts 3 Conclusions References Chapter 2 The Destruction of Volatile Compounds by Heterogeneous Catalytic Oxidation By C.S. Heneghan, G.J. Hutchings and S.H. Taylor

1

1 1 1 2 5 5 6 8 8 47 55 67 77 91 94 96 105

1 Introduction 105 2 VOC Abatement 106 3 Operational Parameters Affecting the Catalytic Combustion of v o c s 106 3.1 Tempertature 106 3.2 System Preheating 107 3.3 Space Velocity 108 3.4 Type of VOC 108 3.5 VOC Mixtures 109 Catalysis, Volume 17 0 The Royal Society of Chemistry, 2004 vii

...

Contents

Vlll

3.6 VOC Concentration 3.7 Deactivation 4 Catalysts used for VOC Abatement 4.1 Noble Metal Catalysts 4.2 Design of Catalyst Supports 4.3 Gold as a VOC Destruction Catalyst 4.4 Metal Oxide Catalysts 4.5 Mixed Catalyst/Sorbent Systems 4.6 Comparison of Noble Metal and Oxide Catalysts 5 Conclusions References Chapter 3 CO Oxidation Over Supported Au Catalysts B y M.C. Kung, C . K . Costello and H . H . Kung

1 2 3 4 5 6

Introduction Preparation of Supported Au Catalyst Nature of Au Active Site Reaction Mechanism Catalyst Deactivation Conclusion References

Chapter 4 Coke Characterization B y C.A. Querini 1 Introduction 2 Temperature-Programmed Techniques 2.1 Temperature-Programmed Oxidation 2.2 TPO Studies of Different Catalytic Systems 2.3 Temperature-Programmed Hydrogenation 2.4 Temperature-Programmed Gasification 3 Electron Microscopy 3.1 Naphtha Reforming 3.2 Coke on Nickel Catalysts 3.3 1-Butene Isomerization 4 Electron Energy Loss Spectrocopy (EELS) 4.1 Naphtha Reforming 4.2 1-Butene Isomerization 4.3 Other Reactions on Zeolites 5 Infrared techniques (FTIR, DRIFTS) 5.1 Cracking 5.2 Isobutane Alkylation 5.3 1-Butene Skeletal Isomerization 5.4 Butene Dehydrogenation 5.5 Other Reactions on Zeolites

111 112 113 113 125 127 128 143 144 148 148 152

152 152 154 158 161 163 163

166 166 167 167 171 175 176 177 177 177 178 178 178 179 179 180 180 180 182 183 183

ix

Contents

6 Laser Raman Spectroscopy 6.1 Classic Laser Raman Spectroscopy (LRS) 6.2 UV-Raman Spectrometry (UV-RS) 7 Dissolution of Support and Solvent Extraction 7.1 Naphtha Reforming 7.2 Coke on Zeolites 7.3 Paraffins Dehydrogenation 7.4 Propene Oligomerization on Heteropoly-Acids 7.5 n-Butane Isomerization 8 Neutron Scattering and Attenuation 9 Nuclear Magnetic Resonance (NMR) 9.1 13CCP/MAS-NMR 9.2 'H NMR 9.3 '29XeNMR 9.4 129SiMAS NMR 10 Auger Electron Spectroscopy (AES) 11 X-Ray Diffraction (XRD) 12 Secondary Ion Mass Spectrometry (SIMS) 13 Sorption Capacity: Surface Area and Pore Volume 13.1 Coke on Zeolites 13.2 Residue Hydrotreating 13.3 Isobutane Dehydrogenation 14 X-Ray Photo-electron Spectroscopy (XPS) 14.1 Coke on Zeolites 14.2 Residue Hydrotreating 14.3 Isobutane Dehydrogenation 15 Ultraviolet-Visible Spectroscopy (UV-VIS) 15.1 Isobutane Alkylation 15.2 n-Butane Isomerization 16 Electron Paramagnetic Resonance (EPR) 17 Coke Formation Rate 18 Concluding Remarks References

183 183 184 186 187 187 188 188 188 189 189 190 193 193 194 194 195 196 197 197 199 199 200 200 200 200 200 20 1 201 202 203 203 206

Chapter 5 Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds 210 By M . Rahmani, K . Badii, M . Faghihi, M . Sanati, N . Cruise, 0.Augustsson and J.J. Spivey 1 Introduction 210 21 1 2 Effect of Organo-silica Compounds 2.1 Chemical Properties of Hexamethyldisiloxane (HMDS) 212 2.2 Deactivation Effect of Hexamethyldisiloxane (HMDS) on Oxidation Catalysts 212

X

Contents

2.3 Effect of Hexamethyldisiloxane (HMDS) Concentration 218 2.4 Effect of Catalysts and Supports 218 2.5 Effect of Deactivation Temperature 220 2.6 Effect of Reactor Design 22 1 2.7 Mechanism of Deactivation 223 3 Deactivation by Phosphorus Compounds 226 3.1 Introduction 226 3.2 The Influence of Phosphorus Poisoning 228 3.3 Support Effects 238 3.4 Mechanism and Kinetics 239 4 Mathematical Modeling of Deactivation by Si and P-Based Compounds 24 1 4.1 Mathematical Approaches 241 4.2 Analytical and Numerical Methods 242 4.3 Modeling of Catalyst Poisoning by Organosilicon Compounds 243 4.4 Modeling of Poisoning by Organophosphorous Compounds 243 4.5 Optimization of Active Phase Distribution For Deactivating Systems 250 4.6 Summary 252 References 254 Chapter 6 Microemulsion:An Alternative Route to Preparing Supported Catalysts B y S. Rojas, S. Eriksson and M . Boutonnet

258

1 Introduction 258 2 Formation of Nanoparticles in Microemulsions 259 2.1 What is a Microemulsion? 259 2.2 Structure of Microemulsions 260 2.3 Microemulsions as Synthesis Medium 261 2.4 Some Relevant Aspects of Microemulsions for Particle Preparation 26 1 3 Metal Oxides by Microemulsion 265 3.1 Introduction 265 3.2 Catalytic Oxide Materials 265 3.3 Oxide Materials 267 4 Metal-based Catalysts Prepared by Microemulsion 272 4.1 Introduction 272 4.2 Unsupported Catalysts 272 4.3 Supported Catalysts 275 4.4 Microemulsion vs Traditional Techniques 283 5 Concluding Remarks 288 References 289

Contents

xi

Chapter 7 Catalysis of Acid/Aldehyde/Alcohol Condensations to Ketones By K . M . Dooley 1 Introduction 2 Decarboxylative Condensation, Acids 3 Decarboxylative Condensation, Aldehydes and Alcohols 4 ‘One-step’.4:do1 Condensations to Ketones 5 Lower Temperature Condensations to Ketones 6 Catalyst Properties - Decarboxylative Condensations 7 Catalyst Properties - ‘One-step’ Aldol Condensations References

Chapter 8 Turnover Frequencies in Metal Catalysis: Meanings, Functionalitiesand Relationships By J.G. Goodwin Jr, S. Kim and W.D. Rhodes 1 2 3 4 5

Introduction Determination of TOF Based on Chemisorption Determination of TOF Based on SSITKA Relationship of TOFChem and TOFITK to Site Activity Comparison of TOFChem and TOFITK for Actual Reactions 5.1 Methanation: a Classic Structure-insensitive Reaction 5.2 Methanol Synthesis 5.3 Ethane Hydrogenolysis: a Classic Structure-sensitive Reaction 5.4 Ammonia Synthesis 6 Conclusions References

293 293 294 298 303 306 308 312 314

320 320 32 1 32 1 322 325 325 336 341 343 344 345

1 Role of ’Metal Ion-Metal Nanocluster’ Ensemble Sites in Activity and Selectivity Control BY JOZSEF L. MARGITFALVI AND SANDOR GOBOLOS

1

Introduction

1.1 Historical Background. - In heterogeneous catalysis, the entity involved in the catalytic cycle is an active site or active center located at the surface of a solid material. This idea goes back to the second half of the nineteenth century. For example, Loew suggested’ that when a molecule interacts with the catalyst the ‘sharp corners’ of the catalysts are involved in the break up of the molecule into atoms, i.e., these sites are more reactive than others are. More precise definition of the active sites was first given with respect to metal catalysts. Langmuir has described active sites as an array of sites that can chemisorb an atom or molecule in a localized mode.2 In his model Langmuir suggested that all available active sites are identical. Taylor was the first who proposed that a solid surface with catalytic properties may contain not one, but many types of active site^.^'^,^ He focused on the heterogeneity of the surface of catalysts, ascribing special activity to surface atoms whose coordination to other surface atoms is low. The other very important prediction made by Taylor is related to the ‘reaction induced’ formation of active sites. He stated ‘the amount of surface which is catalytically active is determined by the reaction catalysed’. This principle has been evidenced in several catalytic reactions. It will also be shown in this review that surface species formed in situ play an important role in the generation of a new type of active site containing ‘metal ion-metal nanocluster’ ensembles.

1.2 Type of Active Sites. - In heterogeneous catalysis the following type of actives sites can be distinguished: (i) metallic, (ii) acid-base, (iii) red-ox type, and (iv) anchored metal-complex. The catalytic sites may contain one of the above types of active sites or can include several types of sites. In case of different type of sites the catalysts are bifunctional6 or multifunctional. For instance, Pt/A12037 and Pt/mordenite* are typical bifunctional catalysts containing both metallic and acidic types of active sites. On the other hand, Pt or Pd supported on silicon carbide: nitride,1° or Pt/L-zeolite” are mono-functional catalysts. There are important industrial reactions, such as isomerization and aromatization of linear hydrocarbons, which requires bifunctional catalysts, such as chlorinated Catalysis, Volume 17 0 The Royal Society of Chemistry, 2004

1

2

Catalysis

Pt/A1203.12In these catalysts the two types of sites have to be located sufficiently close to each other so that transport between the sites would not be rate limiting in the overall process. Metal catalysed reactions are differentiated introducing the concept of facile and demanding rea~ti0ns.l~ In principle a single atom should be adequate for a facile (structure insensitive) reaction, while an ensemble of surface atoms is required to form a catalytic site adequate for demanding (structure sensitive) reactions. Consequently, there are reactions, which requires more than one species to form m~ltiplets'~ or ensemble^.'^^^^^^' In other words, some reactions depend on the surface geometry (e.9. hydrogenolysis of hydrocarbons), while other may not (e.9. hydrogenation of olefinic double bond). Red-ox type catalysts are mostly used in oxidation or related types of reactions.I8 For instance, vanadium catalysts containing ions of different valence state are used in the oxidation of benzene to maleic anhydride.19 Bismuth molybdate catalyst can be used both for the oxidation or ammoxidation of propene.20Anchored metal-complex catalysts combine the advantage of both homogeneous and heterogeneous catalysts, however in these catalysts the molecular character of the active sites is maintained.2l In the last generation of this type of catalysts, heteropolyacids are fixed first to the support and in the second step different metal-complexes are anchored to the heteropolyacid. In this way highly active and stable catalyst have been prepared for different reaction^.^^^^^

1.3 Mono- and Bimetallic Supported Catalysts. - The key factor in designing supported metal catalysts is the knowledge about the reaction mechanisms and information about the role of different types of active sites in a given step of the catalytic reaction. The performance of supported mono-functional monometallic catalysts is governed by the metal particle size, metal dispersion, overall morphology of the metal nanocluster, the character of metal-support interaction, and the electronic properties of the In bifunctional supported metal catalysts in addition to the above listed factors the metal/acid balance," and the type and strength of the acid function26play a key role in the overall performance. In case of bimetallic catalysts, other properties, such as surface composition and the potential stabilization of one of the metal components in ionic form, are the most crucial determining the performance of the catalyst. It is noteworthy that combination of modern methods enables the chemist to characterize both active sites of supported metals and the reaction intermediates formed. Additionally, quantum chemical calculations become more and more powerful tools in understanding chemical interaction controlling and governing both the catalyst structure and the catalytic perf0rmance.2~ In the last decade much attention has been paid to metal nano-clusters including supported nanoparticles as one of the promising advanced nanoscopic materials.27Elements easily forming supported metal nanoclusters are Group VIII and IB transition metals as follows: Pt, Ir, Pd, Rh, Ru, Ni, Co, and Au, Ag, Cu. It is interesting to note that the heat of formation of the oxides of these metals is low (usually below -AHf = 40 kcal/mol at 25 "C referred to one oxygen atom28).

3

1: Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

Therefore, the oxides of these metals can easily be reduced to zero valence. The reduction of metal oxides with high heat of formation (above 100 kcal/mol) (e.g. SO2,Ti02,Zr02,A1203,Ce02,Nb205,MgO, La2O3) is rather difficult, therefore they are usually applied as catalyst supports. Other transition metals, such as V, Cr, Mo, W, Mn, Re, Fe, Zn, and sp-metals such as Ga, In, Ge and Sn with an intermediate value for the heat of formation of their oxide (ca. 60-90 kcal/mol) are frequently used as promoters in supported metal catalysts. The metals with an intermediate heat of formation of oxide are usually present in the heterogeneous metal catalysts in the form of isolated ions or nano-sized oxide crystallites even under reducing or reaction c ~ n d i t i o n s . ~These ~ , ~ ~metal , ~ ~ ions behave as Lewis acid sites. These sites can be involved in the polarization of multiple bonds or electron donor groups of substrate molecules, and they can activate carbonyl compounds, nitriles, nitro-compounds, and CO molecule chemisorbed on the surface of heterogeneous ~ a t a l y s t s . ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ’ The Lewis acid strength of metal ions is said to be proportional to the generalized electronegativity of metals in their oxide form (Xi = (1 2Z)XO, where Xi and Xo are electronegativity of metal in the oxide and elemental form, respectively, and Z = valence state of metal in the oxide).36The Lewis acid strength, expressed as the electronegativity of metals in their highest oxides, is significantly higher for Cr(20.8), Mo( 16.9), W( 18.2), Mn(11.2), Re(22.5), Ge(18.0) and Sn(15.3), than for other metals. Therefore, these metals can be used as promoters in different reduction or oxidation catalysts requiring activation of the substrate molecule by Lewis acid sites. It is also known that large number of catalytic reactions, such as catalytic naphtha reforming, hydrogenation of unsaturated carbonyl compounds, oxidation of CO or methanol, require both metallic sites and Lewis acid sites for activating hydrogen or oxygen and substrate molecules, respectively. Metal nanoclusters of Group VIII or Group IB catalysts supported and stabilized on irreducible oxides and promoted by a metal ion can fulfill this requirement. In these catalysts metal ions or metal oxide species of the promoter interact with metal nanoclusters at the cluster-support interface, or can be stabilized on the top of the metal nanocluster. Due to the intimate contact between the metal ion of the promoter and the metal nanocluster bimetallic ensemble sites are formed.37.38 These types of sites can also be formed by high temperature reduction of metal catalysts supported on a slightly reducible oxide support, such as Ti02.39740 Based on a careful literature survey and recent results published by the authors of this review, it can be assumed that in a number of heterogeneous catalytic reactions ‘metal ion - metal nanocluster’ ensemble sites are operative. In these catalysts the metal ions have to be located in atomic closeness to the metal nanocluster. As far as the action of supported bimetallic catalysts is concerned, the main theories suggest either geometric and/or electronic effects to account for the improved catalytic properties. For instance, in platinum based naphtha reforming catalysts, the electronic modification of platinum particles may be induced by an interaction with an oxide layer of the promoter41or by alloy formation!2 The electronic modification results in a change in the Pt-C bond strength of adsorp-

+

4

Catalysis

tion of hydrocarbons and hence alters the activity and selectivity of the reforming type catalysts.43 Geometric or ensemble effects arise due to the dilution of the surface of the given active metal by an inactive one. For instance, this is the case for A,B, type binary alloy catalysts containing Pt, Pd or Ni as active metals and Au, Cu, Sn, etc. as diluting elemenfs.l6 The ensemble effect can induce different structure sensitivities of the reactions. For instance, the dilution of active metal (Pt) surface into smaller ensembles by addition of inactive species, such as Sn or Ge, selectively poisons demanding reactions (e.g. hydrogenolysis and coke formation) that requires relatively large clusters or ensembles of adjacent metal atoms. While structure insensitive reactions (double bond hydrogenation, aromatization or isomerization) can occur on single isolated atom^.^,^^ In bimetallic reforming-type catalysts the presence of separate oxidized promoter species, e.g. Sn(I1)and Ge(IV),46results in a change of the acidity, affecting both the activity and selectivity of the cata1yst:l Bond and co-workers have classified bimetallic or modified supported catalysts as follows: 47 i. Formation of bimetallic and/or alloy type particles from a pair of elements showing substantial or complete miscibility (for example, alloy type supported Sn-Pt, Sn-Pd catalysts, see section 2.1.2); ii. Formation of bimetallic clusters from pairs of elements showing limited solubility; (for example, supported Sn-Ru catalysts, see section 2.3); ... 111. Incorporation of a third component that cannot be reduced to the zerovalence state but coming into contact with the metal particle (PtMo03/Si02,R U - T ~ O ~ / S ~ItOwas ~ ) .suggested ~ ~ > ~ ~ that in these catalysts the character of interactions is similar to catalysts with Strong Metal Support Interaction (SM SI); iv. Addition of a third component that mainly interacts with the support. In this way either the character of metal-support interaction is altered or the electron density of supported metal nanocluster is changed (for example, alumina supported Sn-Pt catalysts prepared by conventional methods, see section 2.1.1); v. Addition of other species (for example species electronegative in character, which act as selective or non-selective temporary poisons).50 The above classification suggests that under properly chosen condition the subject of this chapter, i.e. ‘metal ion-metal nanocluster’ ensemble sites (MIMNES)” can be formed in most of the above types of catalysts. For instance, from bimetallic clusters of type (i) and (ii) MIMNES can be formed under conditions of mild oxidation. In catalysts type (iii) MIMNES should exist both under oxidative and reductive environment. In catalysts type (iv) any metalsupport interaction with the involvement of non-reducible oxide can also be considered as MIMNES. The only requirement for the formation of MIMNES is the atomic closeness of the two types of sites.

I : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

5

1.4 Promotion of Supported Metal Nanoclusters. - In the last decade a growing body of data provided evidence for the presence of specific active sites on the periphery of metal particles composed of metal site and specific sites on the support surface (adlineation sites). For example the enhanced activity of transition metal catalysts supported on/or promoted with reducible oxides TiO2, wo3, Nb2O5, etc. in carbonyl bond hydrogenation was attributed to the formation of specific s i t e ~ . ~Boffa, * , ~ ~Bell and Somorjai have proposed that the hydrogenation of carbonyl bonds on the surface of rhodium promoted by oxide species such as TiO,, ZrO,, TaO,, WO,, etc. proceeds via the activation of C=O bond through simultaneous adsorption of the carbon end to the metal site and oxygen end to the Lewis acid site on the oxide.53 A similar model was proposed by Vannice et ~ 1 to explain . ~ ~ the extremely high activity of 0.95%Pt/Ti02 reduced at 500 "C in acetophenone hydrogenation, and the enhanced selectivity toward crotyl alcohol in crotonaldehyde hydrogenat i ~ n . ~The ' model also implies the creation of special sites at the metal-support interface that can coordinate the oxygen end of the C=O bond and thereby specifically activate the carbonyl bond. The enhanced selectivity of Ru/Zr02 toward cinnamyl alcohol in cinnamaldehyde hydrogenation was also ascribed to the formation of Ru-Zr"+ sites at the periphery of the nanoparticles. The presence of mixed Ru-Zr"+ sites appeared to decrease the strength of the C=O bond, thus facilitating the hydr~genation.~~ Similar interfacial active sites created in Pt/Mo03 and Pt/W03 upon high temperature reduction were suggested to favor the isomerization of ally1 alcohol to propanal at the expense of hydrogenation to propan01.~~ Bell and S ~ m o r j aproposed i~~ the concept of the interfacial active site involving the coupling of a metal center and a Lewis acid/base site to form adjacent centers. The latter sites are formed either in the oxide support or the added promoter. It was suggested that these active sites might be crucial in the conversion of the molecules with polar functional groups (such as CN, CS and NH). Close analysis of data presented in the above references37.39.40.52~53~54.55 shows that in all cases the character of interactions strongly resembles the presence of 'metal ion-metal nanocluster' ensemble sites.

1.5 Characterizationof Supported Metal Catalysts. - Chemisorption of different probe molecules and Temperature Programmed Reduction (TPR) studies are frequently used to study the metal dispersion, surface composition and oxidation state of metals in mono- and bimetallic supported catalysts. Combined use of CO, hydrogen and oxygen chemisorption as well as oxygen-hydrogen titration can provide information about the dispersion and surface composition of metal n a n o c l ~ s t e r s TPR . ~ ~ ~studies ~ ~ of bimetallic catalysts can give information about the type, the reducibility, and the oxidation state of metal components. In addition, the position of TPR peaks can be used to characterize the type of interactions of the metal species in the c a t a l y s t ~ . ~ * ~ ~ ~ Traditionally, IR spectroscopy of adsorbed CO serves as a tool to gain knowledge about the electronic state and dispersion of supported m e t a 1 ~ . ~ ~ * ~ ~ . ~ The spectra of adsorbed CO are known to be the result of the interplay of the

Catalysis

6

interaction between metal d-orbitals and a-bonding and a-antibonding orbitals of adsorbed COeaX-ray photoelectron spectroscopy (XPS),65966 X-ray absorption fine structure (EXAFS)67and X-ray absorption near edge structure (XANES)68 are also used to determine both the electronic state and the environment of metal species in supported catalysts. There are indications in the literature suggesting the formation of electron deficient metal particles in e.g. A1203-basedand halogenated solid catalyst^.^^ However, the mechanism of this process and the nature of anchoring sites are not quite clear. Broensted acid sites, as well as strong Lewis acid sites may be considered as surface centers70 stabilizing small metal particles (Pt, Pd, Ir, Ni) and causing their positive charging.24

1.6 Subject of Contribution.- The aim of this contribution is to give an overview of catalytic systems consisting of very specific type of active sites, the so-called ‘metal ion - metal nanocluster’ ensemble sites. In this respect we shall differentiate two main types of catalysts containing ‘metal ion - metal nanoparticles’ ensemble sites as shown in Figure 1. In catalysts type I the ‘metal ion - metal nanocluster’ ensemble sites are formed at the perimeter of nanoparticles stabilized on reducible oxide supports. Typical representatives of this type of catalysts are Pt/Sn02” or R u - S ~ / A ~ ~ O ~ ~ ~ and Pt/Ti0273catalysts. In catalyst type I1 the above sites are created and stabilized at the surface of the first metal. The stabilization of highly active metal ions at the atomic closeness of metal nano-cluster can be achieved by one of the following methods:

A

B

- Pt’ensemblesites

“Snn?Pt” ensemble sites

i4

support Figure 1

Two forms of ‘metal ion-metal nanocluster’ ensemble sites. A - Type I . ‘Metal ion-metal nanocluster’ ensemble sites formed at the perimeter of nanoparticles. B - Type I I . ‘Metal ion-metal nanocluster’ ensemble sitesformed at the surface of mono- and bimetallic nanoparticles (Reproduced from ref. 107,139 with permission)

7

1: Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

(i) adsorption of metal ions at the surface of the metal n a n o ~ l u s t e r ~ ~ ~ ~ ~ (ii) selective oxidation of one of the components of supported bimetallic nanoparti~le;7~-~~ (iii) incomplete reduction of the active phase in monometallic catalysts (see supported Cu or Cu-Zn catalyst^):^.^^ (iv) incomplete reduction of one of the active phases in bimetallic catalysts (see Re-Pt/AlzOJ catalysts).” The ‘metal ion - metal nanocluster’ ensemble sites can contain either one or two metal components. In different forms of copper and gold catalysts, the metal can exist in both forms, i.e., as metal ion and the metal nanocluster. These systems will be considered as a mono-element ‘metal ion - metal nanocluster’ ensemble sites. However, as it will be demonstrated later, systems containing two elements are more common. In most of these systems the metal ions are formed from elements of well known red-ox metals, such as tin, rhenium, iron, tungsten, molybdenum, etc. while the metal nanoparticles are noble metals, such as platinum, ruthenium, etc. In this work we shall focus on preparation, stabilization, and use of catalysts containing ‘metal ion - metal nanocluster’ ensemble sites. We shall discuss different case studies summarized in Table 1. With regard to the bimetallic systems involving a platinum-group metal and a second metal, usually having lower or no catalytic activity, many studies have been carried out with the aim to investigate the chemical state of both metals. Thus, Pt, Ir, Os, Pd, Rh, and Ru catalysts among others have been studied, using Fe, Co, Ni, Ge, and Sn as second metal. Most of the studies have shown that the second metal after reduction remains as a cationic species associated with the platinum-group metal, this sites being responsible for the selectivity improvement, e.g. in the hydrogenation of unsaturated aldehydes to unsaturated alcoho1.8WJ3

Table 1

List of case studies related to the involvement of ‘metal ion - metal nanoparticle ensemble sites’ in activity or selectivity improvement

Catalysts

Reaction

Substrate

Product

Sn-Pt/SiO,

Hydrogenation Hydrogenation Low temperature CO oxidation Low temperature CO oxidation Hydrogenolysis Hydrogenolysis Hydrogenation Hydrogenation Hydrogenolysis Reductive alkylation

crotonaldehyde benzonitrile

crotylalcohol secondary amine

Au/MgO Sn-Ru/A1203 Re-Pt/A1203 Cu-chromite CuO-ZnO

*Me = CH,, Bu = C,H,.

co co

co2 co2

ethyl laurate dodecanol butylacetate butanol carbonyl compounds alcohol, amine lauric acid dodecanol ethyl laurate dodecanol butylamine MeNHBu*

8

Catalysis

In the present review the role of supported 'metal ion - metal nanocluster' ensemble sites in activity and selectivity control of different catalytic reactions will be discussed. Evaluation of literature data and interpretation of author's recent results obtained in the activation of different organic carbonyl compounds, nitriles, and the CO molecule will be given. The catalytic performance of metallic nanoclusters promoted by metal ions or reducible metal oxides will be discussed in separate chapters according to the type of metal forming the nanocluster.

2

Casestudies

2.1 Supported Sn-Pt catalysts. - 2.1.I General Approaches Used for the Preparation of Supported Sn-Pt Catalysts. Supported bimetallic Sn-Pt catalysts can be prepared in different ways. Co- or subsequent impregnation techniques are often used, followed by high-temperature decomposition of pre-adsorbed precursors of both metals in reductive or oxidative atmospheres. For example, alumina supported Sn-Pt catalyst were prepared by (i)impregnation of commercial reforming Pt catalyst (0.3 wt. YOPt and about 0.6 wt. YOCl) with an acetone solution of SnC14 x 5H20;g4(ii) co-impregnation with appropriate amount of SnC12 x 2 H 2 0 and H2PtC16in dilute HCl;g5(iii) co-impregnation using an acetone solution containing H2PtC16and snc12;86(iv) co-impregnation with nonacidic precursors Pt(NH&(OH), x XH20 and tin (11)tartrate (SnC4H406).g7 The latter method resulted in a catalyst containing no chlorine. Sn-Pt/Si02catalysts were prepared, for example, by successive impregnation with (i)an acetone solution containing H2PtC16and SnC12;(ii) by coimpregnation with appropriate amounts of H2PtC16x 5H20and SnC14 x 5H20in acetoneggor with (iii) methylene chloride solution of ci~-[PtCl~(PPh~)~] followed by impregnation with an acetone solution of SnC12.g9The main drawback of these approaches is that during high temperature treatment, the formation of bimetallic surface species takes place only by chance. The above disadvantage can be avoided by using different anchoring techniques. The major advantage of anchoring techniques is that the chemistry of the system (parent catalyst and modifier compound) controls the formation of desired surface species, an advantage not often found upon using conventional modification methods. To obtain tin-platinum supported catalysts by anchoring, the following two approaches have been applied: i) anchoring platinum complexes (e.g. PtCl, Pt(CqH7)?' into tin ions bonded to the support and ii) anchoring of individual platinum-tin complexes or clusters. In the latter approach the composition of Sn-Pt bimetallic particles can be controlled effectively by selection of the starting inorganic complexes with different Sn:Pt ratios (e.g. [PtC13SnC13l2-, [PtC12(SnCl3)2)I2-, [PtC12Sn(PPh3)2], [Pt( SnC13)5]3-, [Pt3SngC120]~-:~,~~,~~ or bimetallic ionic compound with Pt:Sn ratio 1:1 [Pt(NH3)4] [SnC16]? Anchoring of tin ions onto the surface of the support is based on the reactivity of surface OH groups of Si02 or A1203. This has been accomplished by using

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

9

either vapors of Sn(OOCCH3)2or an alcohol solution of SnC4.%Upon using this approach different organometallic and alkoxide compounds, metal halogenides, salts of organic acids, etc. were also attached to the surface of different support materials containing OH groups? It is necessary to emphasize that the surface composition of supported nanoclusters strongly depends on the method of preparation. If the formation of platinum-tin alloy phases should be avoided, tin must be introduced first onto the support by (i) exchange, (ii) coprecipitation of tin and aluminum oxides and (iii)by sol-gel synthesis of the Sn/A1203system. For example, Sn-Pt catalysts with high platinum dispersion (up to 900/) were prepared by the sol-gel method by adding tetrabutyltin to a homogeneous solution containing aluminum tri-secbutoxide (TBA), followed by impregnation of dried and calcined solids with an aqueous solution of H2PtClb.95 In solvated metal atom dispersion (SMAD) method solvated atoms prepared at very low-temperature are used as transient, highly reactive organometallic reagent for the deposition of Sn-Pt bimetallic particles onto different supports.96 In another approach chemical vapor deposition (CVD) using tin organometallic compounds was applied.97For example, the selective reaction of Sn(CH3)4vapour with Pt nanoparticles supported on Si02appears to be very promising preparation method. Methods of Controlled Surface Reactions ( C S R S ) ~ *and * ~ ~Surface Organometallic Chemistry (SOMC)'00~'0'~'02 were developed with the aim to obtain surface species with Sn-Pt interaction. In CSRs two approaches have been used: (i) electrochemical, and (ii) organometallic?' Characteristic feature of the organometallic approach is that both CSR and SOMC results in almost exclusively supported alloy type bimetallic nanoclusters. Studies on the reactivity of tin organic compounds towards hydrogen adsorbed on different transition and noble metals have revealed new aspects for the preparation of supported bimetallic catalysts. The formation of surface alloys, phase segregation at the surface, site sensitive chemisorption, changes in electronic properties, surface reconstruction, etc. have been investigated by a range of surface science methods.lo3 2.1.2 Preparation of Alloy Type Sn-Pt/SiOz Catalysts. Supported bimetallic Sn-Pt catalysts can be prepared using different methods and approaches. However, exclusive formation of alloy type nanoclusters can be achieved by using methods of surface organometallic chemistry, namely by applying Controlled Surface Reactions (CSRs) between hydrogen adsorbed on platinum and tin tetraalkyls. CSRs between Sn(C2H5)4 and hydrogen adsorbed on supported platinum (see reaction (1) below) has been first described in 1984.9' Under properly chosen experimental condition the reaction between Sn(C2H5)4and surface OH groups of the support has been completely suppressed. Consequently, reaction (1) provided direct tin-platinum interaction that was maintained upon decomposition of the primary surface complex (PSC)in a hydrogen a t m ~ s p h e r e (see ~ ~ ?reaction '~ (2) below). The net result is the formation of alloy type bimetallic surface entities

10

Catalysis

as it has been evidenced by Mossbauer spectros~opy.'~~ This approach led to the monolayer coverage of platinum by tin organometallic specieslo6as shown in Figure 2. The monolayer coverage resulted in Sn,,,,h/Pt,ratios around 0.4 (Snanchamount of tin atoms anchored, Pt, - number of surface Pt atoms). The surface chemistry involved in the formation of monolayer coverage can be written as follows:98999

Pt-Sn&,,,

+ (4 -x)/2 H2 -.Pt-Sn + (4 -x) RH

(2)

The material balance of tin anchoring indicated that under monolayer coverage of platinum by PSC the average value of x, is around 1.5. This fact pointed out that Pt nanoclusters are covered by -SnR3and -SnR2moieties formed in 1:l ratio. It has been suggested that all anchored surface organometallic species with general formula of -Snq4,)and x> 1 can be considered as Coordinatively Unsaturated Primary Surface Complex (CUPSC). The increase of the Snanch/Pts ratios required the formation of new anchoring sites for tin tetraalkyl. The CUPSC is believed to be one of these new sites, which can be used to anchor additional amount of tin tetraalkyls, provided the concentration of tin tetraalkyls is high enough. It has also been proposed that in excess hydrogen the extent of coordinative unsaturation can further be increased, i.e.,in this way higher Snanch/Ptsratios can be obtained. The validity of above suggestions has been proved and the Snanch/Pts ratio increased up to 2: 1.'06 Surface reactions involved in this new approach can be written as follow^:'^^'^^ PtHads. Pt-SnR3

A Figure 2

+

+ SnR4+Pt-SnR3 + RH x Had,. + Pt-SnR(3-,)

+

x RH

(3) (4)

B

Computer modeling of organometallic moities anchored onto the platinum nanocluster, monolayer coverage. A - anchoring of SnR, (top view); B anchoring of SnR2 (side view), R = -C2H5 (Reproduced from ref. 106 with permission)

1: Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

11

Reaction (3) and (4) describe the formation of PSC and CUPSC, respectively. The coordinatively unsaturated surface species interacts with Sn& used in large excess (reaction (5)) resulting in surface complex in the second layer (SCSL). Similar surface species have been suggested when supported rhodium was modified with tin tetrabutyl.lo1In the presence of excess hydrogen the SCSL is partially hydrogenolysed resulting in coordinatively unsaturated species in the second layer (reaction (6)),which also interacts with Sn& (reaction (7)). The net result is the formation of slab-like Multilayered Surface Complex (MLSC) on the Pt surface. Reaction (3) - Reaction (7), which take place in the presence of a solvent, are referred as tin anchoring reactions in the presence of hydrogen. This anchoring process is shown in Scheme 1. Scheme 1 shows two routes for tin anchoring. Route 1 takes place in excess hydrogen, while route 2 in excess tin tetraalkyls. Route 1 is more preferable than route 2, as it provides high tin coverages and strongly decreases the amount of tin introduced into the support either by adsorption or a side reaction with the involvement of the surface OH groups of the support. New types of tin anchoring sites were created when both PSCs and CUPSCs were mildly oxidized to Oxidized Surface Organometallic Complex (OSOC), with general formula of -Sn,Rb0,.’06The addition of trace of oxygen led to the immediate formation of ethylene, i.e. surface chemistry involved in tin anchoring was altered. In this case as shown in Scheme 2 the lone pair of electrons of the

II

IL SCSL

Scheme1

Scheme of tin anchoring in the presence of excess hydrogen or excess tin tetraalkyl (Reproduced from ref. 107 with permission)

12

Catalysis

oxygen atom in -Sn,RbO, moieties are involved in tin anchoring. As the number of anchoring sites increased further, the amount of tin anchored significantly increased.lo6 It has been proposed that in the presence of oxygen the build-up of the second and subsequent layers can be written as follows:'o6

In reaction (8) oxygen containing multilayer species (OMLSC)) are formed, which instantaneously react, forming of ethylene and MLSC (11). The approach using trace amount of oxygen during tin anchoring will be denoted as route 3. The decomposition of surface organometallic complexes formed in tin anchoring steps (see reactions (3) - (9))was accomplished as a gas-solid reaction in the temperature range between 25-300 "C. The decomposition in a hydrogen atmosphere led to the formation of alloy-type bimetallic surface entities. More details on the decomposition of different surface organometallic complexes can be found e1sewhere.lo6These Sn-Pt catalysts will be referred as (H) type catalysts. However, the decomposition of surface organometallic complexes can also be

I

support

support

Scheme 2

Tin anchoring in the presence of trace amount of oxygen (Reproducedfrom ref. 107 with permission)

I

13

I : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

carried out in the presence of oxygen.lo8These Sn-Pt catalysts will be referred as (0)type catalysts. In this case the tin organometallic species were transformed into tin-oxide like surface species on the top of supported platinum nanocluster. Upon subsequent reduction this tin-oxide type species could be partially or fully reduced resulting in alloy type bimetallic nanoparticles. In the case of a silica support, the reduction in a hydrogen atmosphere above 350°C was complete, while on alumina supported platinum part of tin-oxide migrated onto the support.lo8 Supported Sn-Pt catalysts prepared by CSRs will be designated as follows. The reaction route used will be indicated by numbers 1,2,3, corresponding to tin anchoring in the presence of excess hydrogen, excess tin tetraethyl, and trace amount of oxygen, respectively. The atmosphere (hydrogen or oxygen) used for the decomposition of surface organometallic moieties will be denoted as (H) or (0).Thus, designation, such as (H-3) indicates that the catalyst has been prepared in the presence of trace amount of oxygen and hydrogen atmosphere was used in the decomposition. 2.1.3 Characterization of Alloy Type Sn-PtlSiOz Catalysts. The alloy type SnPt/SiOz catalysts were characterized by chemisorption and Mossbauer spectroscopy. Results of CO and hydrogen chemisorption are summarized in Table 2.'08 From data presented in Table 2 the following conclusions can be drawn:

Upon increasing tin content the number of accessible Pt sites decreases; The H/Pt ratio decreases much faster than the CO/Pt ration. This is an indication that tin strongly blocks kink and corner sites involved in hydrogen activation; The hydrogen chemisorption on the supported bimetallic nanoclusters can completely be suppressed; The site blocking effect of tin is slightly higher in (0)type catalysts;

Table 2

Chemisorption properties of silica and alumina supported Sn-Pt catalysts prepared by decomposition of surface organometallic species (Reproduced from ref. 108 with permission)

Catalysts Pt/SiO; Sn-Pt/Si02 Sn-Pt/Si02 Sn-Pt/Si02 Pt/A120, Sn-Pt/A1203 Sn-Pt/A1203 Sn-Pt/A1203

Sn/Pt"

H/Ptb( A )

CO/Ptb( B )

AIB

-

0.337 n.a. (0.225) 0.162 (0.129) 0.105 (O.OO0) 0.341 0.212 (0.174) n.a. (0.122) 0.169 (0.086)

0.361 n.a. (0.278) 0.245 (0.210) 0.168 (0.157) 0.703 0.509 (0.439) n.a. (0.354) 0.486 (0.273)

1.07 n.a. (1.24)

0.6 0.8 1.7 -

1.4 2.3 4.3

1.60(00) 2.06 2.40 (2.52) n.a. (2.90) 2.88 (3.17)

in at/at, calculated from AAS analysis and the overall material balance of tin anchoring. first number catalysts type (H), in parenthesis catalysts type (0). ' Pt content: 3 wt. %. Pt content: 0.3 wt. %. a

Catalysis

14

(v) In the case of the alumina support, the decrease of the amount of chemisorbed hydrogen and C O is less pronounced. This indicates that at high Sn/Pt, > 1.4 a definite part of tin was introduced into the support and not to the p l a t i n ~ m . ~ ~ ~ ' ~ * Typical Mossbauer spectra of (H-3) type Sn-Pt/Si02 catalysts with different Sn/Pt, ratios are shown in Figure 3, the corresponding computer evaluation of the spectra is given in Table 3.1°9 These results indicate that in these (H-3) type catalysts after reduction at 300 "C there are only three tin containing species and the ratios of these species is strongly depends on the Sn/Pt, ratios. These catalysts contain two alloy phases: a platinum-rich Sn-Pt(a), which corresponds to Pt3Sn alloy phase, and a tin-rich Sn-Pt(b) phase, which can be related to Pt2Sn3or PtSn2 phases.ll0 Upon increasing the tin content the ratio of the tin rich phase increases from 14 to 54 %. Based on this finding it has been suggested that the surface of the bimetallic Pt-Sn nanocluster is enriched by tin. A characteristic feature of these catalysts is the

-6

-4

-2

0

2

4

6

v / mm s' Figure 3

Mossbauer spectra of Sn-PtlSiO, catalysts with diflerent SnlPt (atlat) ratios. SnlPt ratios: N" 1 - 0.6, N" 2 - 1 .O, N" 3 - 1.3. Catalysts type (H-3) (Reproducedfrom ref. 109 with permission)

15

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

Table 3

Results of the computer evaluation of Mossbauer spectra (Reproduced from ref. 109 with permission)

Sample No SnlPt, ~

~

1

2

Species

F WHM, mms-'

mms-l

-

-

-

-

1.56 2.31 3.33

-

1.23 0.75 0.75

70 14 16

0.97 1.03 0.87

38 47 15

1.03 1.13 0.76

32 54 14

RI,

%

~~

0.6

1.o

Sn4+ Sn-Pt(a) Sn-Pt(b) Sn2+ Sn4 Sn-Pt(a) Sn-Pt(b) Sn2 Sn4 Sn-Pt(a) Sn-Pt(b) Sn2 +

+

3

QS,

IS, mms-'

1.3

+

+

-

1.56 2.20 3.69 -

1.49 2.15 3.82

-

0.71 -

0.92 -

-

1.19

-

-

-

-

Sn-Pt/SiO, catalysts, type: (H-3). IS: isomer shift, mm s-'; QS: quadrupole splitting,mm s-'; FWHM: full-width at half-maximum, mm s-'; RI: relative spectral area (%). Errors on IS, QS and RI values are 0.03mm s-' and +/10 re]. %, respectively.

presence of Sn2+species. This behaviour has been observed only in catalysts prepared in the presence of trace amounts of oxygen. Because the relative amount of this species is almost constant (see Table 3), it has been suggested that these ionic forms of tin are stabilized at the perimeter of the Pt nanocluster. Further results of Mossbauer spectroscopy of Sn-Pt/Si02catalysts are given in Table 4, where characteristic features of (H-1) and (0-1) type catalysts are ~ompared.~' From results of Mossbauer spectroscopy presented in Table 4 the following information can be extracted: In (H) type Sn-Pt/Si02 catalyst upon its contact with air the tin is strongly oxidized; The (H) type catalyst after reduction at 300 "Ccontains two alloy phases: a platinum-rich Sn-Pt(a), which corresponds to Pt3Sn alloy phase, and a tinrich Sn-Pt(b) phase, which can be related to Pt2Sn3or PtSnz phases."' The ratio of these to alloy phases is roughly 1:l; Because no oxygen has been used in the tin anchoring step in (H) type catalyst the reduction of tin at 300 "Cis complete, i.e.,this catalyst contains no ionic forms of tin; However, the results show also that if the reduction has been performed at 200 "Ctin is not fully reduced; In the (H) type catalyst the tin rich phase oxidizes faster than the Pt rich phase (see the as received sample); The (0)type catalyst as prepared contains only tin (IV) oxide, this form of tin partially or fully covers the Pt nanocluster;

16

Catalysis

Table 4

Catalyst samples

Comparison of the Mcssbauer parameters of ( H ) and ( 0 )type of Sn-Pt/Si02catalysts (Reproduced from ref. 3 5 with permission) Species

IS, mm s-I

Sn4+ Sn-Pt(a) Sn-Pt(b)

0.09 1.25 1.98

Sn4+ Sn-Pt(a) Sn-Pt(b)

0.43 1.30 2.17

Sn-Pt(a) Sn-Pt(b) Sn4+

1.32 2.21 0.08

Sn4 Sn-Pt(a) Sn-Pt(b) Sn2+ Sn4 Sn-Pt(a) Sn-Pt(b)

0.48 1.24 2.35 2.60 0.62 1.20 2.17

+

+

QS,

FWHM,

mm s-'

Mms-'

RI, % 55 27 18 8 38 54 49 51 100

0.86 1.45 1.03 0.80 1.17 1.31 1.44

1.46 0.99 1.57 1.30 1.56

16 33 44 7 7 23 70

1-44

0.83 1.20 2.17

Catalysts with Sn/Pt (at/at) = 1.0. I S isomer shift, mm s-'; QS quadruple splitting, mm s-'; FWHM:full-width at half-maximum, mm s- *;RI: relative spectral area (%). Errors on IS, QS and RI values are 0.03 mm s- and / 10 rel. YO,respectively.

'

+

The reduction of tin(1V) oxide formed at the top of platinum nanocluster is incomplete at 200 "C; The full reduction of ionic forms of tin in (0)type catalyst requires temperatures higher than 300 "C; After reduction at 300°C the (0) type catalyst contains also alloy phases similar to the (H) type, however the ratio of these two phases is different; (0)type catalysts reduced below 300 "Ccontain also both tin (IV) and tin (11)oxides.

2.1.4 Use of Alloy Type Sn-PtlSiOl Catalysts in Selective Hydrogenation of a,/%Unsaturated Aldehydes and Ketones. 2.1.4.1 Literature Background. The hydrogenation of a,p-unsaturated aldehydes onto unsaturated alcohols is a very important reaction in the field of selective hydrogenation. In the presence of heterogeneous catalysts both the aldehyde and the olefinic double bond of unsaturated aldehydes can be hydr~genated."'*'~~.~'~ The reaction network involved in the hydrogenation of unsaturated aldehydes is given in Scheme 3.83 This scheme contains both parallel and consecutive reactions. The key issue is how to tune the selectivity of a given supported catalyst towards the formation of unsaturated alcohols. In other words the main question is how to change from

17

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

R-CH-CH-C,

Scbeme 3

qo

acrolein motonaldehyde

Reaction network invdved in the hydrogenation of a# unsaturated aldehydes

thermodynamic selectivity control to kinetic selectivity. This field has been recently reviewed by different Therefore, in this section no attempt is made for a detailed overview. We shall only refer to experimental data related to the possible formation of 'metal ion -metal nanocluster' ensemble sites and their involvement in activity and selectivity control. The first evidence related to the enhanced formation of unsaturated alcohols was mentioned by Adams et In their studies it had been demonstrated that both the activity and the selectivity of Pt black or platinum oxide catalysts to cinnamyl alcohol formation increased in the presence of iron chloride or zinc acetate. The effectiveness of the above additives had been further proved by Rylander who used carbon supported platinum ~ata1ysts.l'~Gallezot and Richard82carried out detailed kinetic studies on the hydrogenation of cinnamic aldehyde over Pt/C carbon catalysts varying the amount of FeC12 added to the reaction mixture. Figure 4 shows the influence of FeC12 both on the initial rate and the initial selectivity. These figures provided valuable information about the role of iron as a beneficial additive. The addition of iron resulted in not only pronounced increase of the selectivity for unsaturated alcohols, but considerable increase in the reaction rate. Figure 4 shows also that above the optimum iron concentration both the rate and the selectivity decreases.82 In additional experiments it has been shown that iron is interacting with platinum, i.e., it is located in atomic closeness to Pt. In the bimetallic nanocluster, due to the high electropositivity of iron, there is an electron transfer from iron to platinum. The net result is the formation of electron deficient iron species at the Pt surface. The authors suggested that these electron-deficient or low-valent iron species on the Pt surface might act as Lewis acid adsorption sites. These sites, due uZ.11491*59116

18

Catalysis A Initial selectivity (%) 50. 4

FdPt 1

.

1

1

1

0,s

Figure 4

1

1

1

1

Felpt d

1

1

1

1

03

1

1

1

.

1

1

1

InJZuenceof the FelPt ratio on the performance of Pt/C catalyst in cinnamaldehyde hydrogenation. A - initial rate of cinnamaldehyde (I 0 - 3 rnollmin gc,,J; B - initial selectivities (Reproduced from ref. 82 with permission)

to the high electron affinity of Mn+sites, can polarize the carbonyl bond, i.e., in this way they are involved in the activation of the C=O double bond as shown in Figure 5. Consequently, the key issue in this activation step is the Mn+- carbonyl interaction, taking place in the atomic closeness to the platinum sites.82 The selective reduction of the carbonyl group requires the creation of polar sites that interact with the C=O bond and thus lead to its preferential activation. This may be achieved by using bimetallic catalysts or supported noble metals on partially reducible oxides.82 Gallezot and Richard82have classified this effect into three groups: (i)Catalysts where metallic promoters are added in ionic form. This process may result in the formation of the cationic species deposited on the metal nano-particles or, due to the hydrogenation conditions, these cationic species can be reduced producing bimetallic particles. (ii) Catalysts involving bimetallic particles, where electropositive metal atoms are associated in the same particle with metal atoms of higher red-ox potential (usually, platinum-group metals). (iii) Catalysts involving oxidized metal species at the metal support interface, usually produced by migration of partially reduced support species to metal particles, resulting in decoration of metallic species. Among the transition metals, platinum and rhodium are the most studied and

1 H !

r t Figure5

Pt

P t - Pt

Scheme of C=O bond activation by electropositive iron atoms on platinum surface (Reproduced from ref. 82 with permission)

19

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

among the selectivity promoters, iron, Ge, Ga, tin etc. are widely employed in the hydrogenation of unsaturated aldehydes to unsaturated a l c ~ h o l s . ' ' ~Figure ~~'~6 and Figure 7 show the influence of different additives on the selectivity of various unsaturated alcohols over modified platinum catalysts. These results unambiguously show that tin is one of the most effective modifiers for selectivity improvement. Figure 8 shows the influence of tin in the gas phase hydrogenation of crotonaldehyde over tin-modified Rh/Si02catalysts prepared by using CSR.83$' l8 Figure 8 clearly shows the suppression of double bond hydrogenation and enhancement 100

0 E

20 0

No

Ti

V

L L d Ga

F

Gc

S

Promotor X in PtX catalyst acrolein

Figure 6

B crotonaidehydc

-

o me crotonaldehyde

Selectivity of unsaturated alcohols in hydrogenations of diflerent a,punsaturated aldehydes. Promoters used are shown. Selectivity is higher when the C = C bond bears substituents (Reproduced from ref. 8 1 with permission) 30

0

1

Na

Ti

V

Fe

Ga

Ge

' Sn

PromotorX in PtX c a t a l y s t 8 methacrolein

Figure 7

acrolein

methyl vinyl

ketone

Selectivity of unsaturated alcohols in hydrogenations of different a , p u n saturated aldehydes and ketone. Notice the zero selectivity with unsaturated ketone (Reproduced from ref. 8 1 with permission)

20

Catalysis

1

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Sn/(Rh+Sn) Figure 8

Selectioity to crotyl alcohol and n-butyraldehyde us. atomic ratio Sn/(Rh + Sn) in the gas phase hydrogenation of crotonaldehyde ( C A ) over Rh-Sn/SiO,-CSR catalysts at constant conversion of crotonaldehyde = 15 %. ( T = 40 O C , Proto[ = 2 MPa, H 2 / C A = 20) (Reproduced from ref. 83 and 11 8 with permission)

of carbonyl hydrogenation upon increasing the overall tin content. The methodology of the addition of tin into supported platinum or rhodium seems to play an important role in the behaviour of the active phase obtained. Controlled surface reactions of organometallic compounds with metal surfaces result in the formation bimetallic systems with specific properties in the hydrogenation of different unsaturated ~ o m p o ~ n d sHowever, . ~ ~ the ~ nature ~ ~ ~ ~ ~ ~ of the Sn-Pt or Sn-Rh bimetallic phase formed, and its influence on the final properties of the catalyst, have not been yet well determined and this is still a subject to be investigated. Santori et al. have studied the hydrogenation of several compounds (crotonaldehyde, cinnamaldehyde, butyraldehyde, 2-butanone, benzaldehyde and cyclohexene) over Pt/SiOz catalyst modified with tin CSRs.li9In the hydrogenation of butyraldehyde and butanone, the adsorption of the ql-(0) type appears as highly favorable, both from a geometric and electronic point of view. In the benzaldehyde hydrogenation, the increase in the catalytic activity for bimetallic catalysts associated with electronic effects, i.e. the presence of ionic tin and of phenyl group. In the case of the cyclohexene, geometric and electronic, as well as steric effects lead to a strong reduction of the hydrogenation rate of C=C bond. These results can be extrapolated to explain the behaviour of the unsaturated a,P-aldehydes. The hydrogenation of the C=O group is promoted and the adsorption modes favorable to the C=C hydrogenation are inhibited by tin. XPS analysis of silica-supported Sn-Pt catalysts shows that the binding energy of platinum shifted towards lower values of approximately 1 eV with respect to

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

21

Pt/Si02, this shift can be interpreted as an electronic transfer from tin to platinum. These modifications strengthen the hypothesis of the electronic effects induced by tin as it was previously discussed to explain changes in the H2 and CO chemisorption (see Table 2). It is important to note that the increase in the electronic density of platinum is observed not only in the systems where tin is ionic state, but also in the case of the catalyst containing tin only in zero valent state. For this reason, it was proposed that even in the case of metallic tin forming superficial alloys, polarized states of Pt*- and Sn*+ sites are generated, which would be important in the chemisorption of reagents containing double C=O and C=C bonds.’22The existence of ‘Lewis acid sites’ (electronic modifications), due to the presence of Sn*+ and ionic tin, tends to promote the attack of hydrogen on the C=O group. In further studies it has been demonstrated that a supported platinum catalyst can also be modified either by addition of cobalt, germanium, iron or tin halides. In the hydrogenation of cinnamaldehyde the modification resulted in much higher activities and selectivities for the hydrogenation of the aldehyde group than on the parent platinum catalysts as shown in Table 5.82,123 Ponec et al. has recently demonstrated that in the selective hydrogenation of crotonaldehyde over Sn-Pt catalysts the selectivity of crotylalcohol increased with time on The observed phenomenon had been ascribed to ‘reaction induced selectivity changes’, however the exact origin of these changes was not clarified. It should also be mentioned that in addition to the presence of Lewis-acid type surface entities there is another requirement for the preparation of selective catalysts used in the hydrogenation of a,P-unsaturated aldehydes. In this reaction both the adsorption of the substrate via its olefinic double bond and the re-adsorption of the formed crotylalcohol should be suppressed in order to increase the selectivity for the formation of the unsaturated alcohol. 2.1.4.2 Gas Phase Hydrogenation of Crotonaldehyde Over Alloy Type Sn-Pt/Si02 catalysts. In these studies different Sn-Pt/Si02catalysts prepared by using CSRs were tested. On the parent Pt/Si02 catalyst no crotylalcohol formation was observed. The introduction of tin into platinum resulted significant selectivity improvement and the selectivity for the formation of crotylalcohol (Sc=o) strongly depended on the Snanch/Ptsratio. A characteristic feature of this type of

Table 5

Eflect of Metal Chlorides on Cinnarnaldehyde Hydrogenation (Reproduced from ref. 82 with permission) ~~

~

k x lo3

Selectivity

Additives

(s-l g p t - 9

(%I

none COCI, FeCl, SnC1, GeCl,

0.08 0.8 2.4 10.0 17.8

10 25 64 75 94

22

Catalysis

Sn-Pt/Si02catalysts is the increase of the yield of unsaturated alcohol with time on stream as shown in Figure 9 . l l 1 * l 2 l While the yield of a saturated alcohol increases, that of saturated alcohol, and hydrocarbons decreases. The net result is the pronounced increase of the ScG selectivity. Consequently, the behaviour of alloy type Sn-Pt/Si02 catalysts prior to reaching the steady-state activity strongly resembles that observed by Ponec et ~ 2 Z . l ~ ~ However, the selectivity of Sn-Pt/Si02catalysts prepared by using CSRs, as it will be shown latter, was much higher than that prepared by conventional methods. Steady-state selectivity data obtained in gas phase hydrogenation of crotonaldehyde over different Sn-Pt/SiO, catalysts are summarized in Table 6 and Table 7.lo7 Sn-Pt/Si02 catalysts below monolayer coverage show only a slight increase in the selectivity of unsaturated alcohol, however the activity of these catalysts is much higher than that of the parent Pt/Si02catalyst. These results already show the positive effect of tin in carbonyl activation. However, due to the relatively low tin content of these catalysts, the adsorption of the substrate molecule by its olefinic double bond is still possible, consequently the formation of butiraldehyde is very pronounced. The use of Sn-Pt/SiOz catalysts with multilayer tin coverage resulted in pronounced increase in the SCe selectivity as shown in Table 7. The highest SCa selectivity was around 90 %. It should also be emphasized that the introduction of tin increased the overall activity of all alloy type Sn-Pt/SiOz bimetallic catalysts compared to the parent Pt/Si02. In this respect the behaviour of Sn-Pt/SiO, bimetallic catalysts strongly resembled platinum catalysts modified with iron (see Figure 4). 0.10

To

0 0

5

I0

15

20

Pulse number Figure 9

Time on stream pattern of the gas phase hydrogenation ofcrotonaldehyde over Sn-Pt/Si02catalyst (catalyst type (H-3), Snanch/Pts= 2.9). Reaction temperature: 80 "C,amount of catalyst: 100 mg, flow rate: 90-mllmin. Reaction products: 0 - crotylalcohol, - butyraldehyde, 0 - hydrocarbons, x - butylalcohol. (Reproduced from ref. 11 1 and 121 with permission)

23

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

Table 6

Gas phase hydrogenation of crotonaldehyde to crotylalcohol over Pt/Si02 and Sn-PtlSi02 catalysts with monolayer tin coverage (Reproduced from ref. 107 with permission) ~~

Selectivity (%)

Catalysts, SnanchlPts

Wini a

SAL

SOL

Pt/SiO,

1.68

95

0

0

5

Sn-Pt/SiO,, 0.22

6.05

82

5

13

0

Sn-Pt/Si02,0.39

4.17

71

4

20

5

UOL

HC

initial rate, w, in pmol &atp1s-' determined from the conversion - contact time dependencies extrapolating to zero conversion, measured at 5 % conversion. Temperature of re-reduction = 300 "C, reaction temperature = 60 "C, [C], = 0.64mmol/dm3, catalysts = 40-80g. Abbreviations: SAL - butiraldehyde, SOL - butanol, UOL - crotylalcohol, CH - hydrocarbons. a

Table 7

Gas phase hydrogenation of crotonaldehyde over Sn-Pt/Si02catalyst with multilayer tin coverage (Reproduced from ref. 107 with permission)

Catalysts, SnancdPts

' '

&lw?lol got - s -

Pt/SiO,

-

0.50

0

Sn-Pt/SiO,, 0.38

2.3 1

4.3 1

40

Sn-Pt/SiO,, 1.40

2.21

1.13

65

Sn-Pt/SiO,, 2.50

0.77

0.09

90

a initial rates, determined from the conversion - contact time dependencies extrapolating to zero conversion, measured at 5 % conversion. Temperature of re-reduction = 300 "C, reaction temperature = 60 "C, [C], = 0.64mmol/dm3, catalyst: 40-80g. Abbreviations see Table 6.

Figure 10 shows the dependence of the crotylalcohol selectivities of the conversion over different Sn-Pt/Si02 catalyst~.''~The higher the Sn/Pt ratio the higher the crotylalcohol selectivity. Over catalysts with high Snanch/Pts ratios the selectivity drop between 8 and 50 YOconversion was less than 10 YO.These results show that in Sn-Pt/SiOz catalysts with high Snanch/Pts ratio both requirements of the selective formation of unsaturated alcohols, i.e., the activation of the carbonyl group and the suppression of the readsorption of formed unsaturated aldehyde, are fulfilled. As shown in Figure 9 in the initial part of TOS curve the formation of hydrocarbons was very pronounced. The analysis of hydrocarbons formed in the initial part of time on stream showed marked difference between Pt/SiO, and Sn-Pt/SiO, catalyst as given in Table 8.lo9 Over the monometallic catalyst, formation of C3 hydrocarbons, while on Sn-Pt/SiO, catalyst formation of butadiene was observed. The formation of C3 hydrocarbons indicates that on the parent catalyst decarbonylation of the sub-

24

Catalysis

0.6

0.1 O

c m j w

-

0.2

0

i

0.4

0.6

Conversion Figure 10

Gas phase hydrogenation of crotonaldehyde on diferent Sn-PtlSiO, catalysts. raDependence of the crotylalcohol selectivities on the conversion. Snonch/Pts tios: 0 - 0.00, - 0.44, X - 2.1, - 2.90. Preactivation temperature: 300 "C, reaction temperature: 80 "C,[ C ] , 0.64 mmol/dm3,Catalysts type (H-3), 80 mg (Reproduced from ref. 107 with permission)

+

Table 8

Catalysts

The distribution of C& hydrocarbons formed during the pretreatment of Pt/Si02and Sn-Pt/Si02catalysts with crotonaldehyde-hydrogen mixture (Reproduced from ref. 109 with permission) Propane ~~

Pt/SiO, Sn-Pt/SiO,

Propene

1,3-butadiene

2 0

91

2-butene

butane

13 9

68 0

~

17

0

0

Snanch/Pts = 1.4,pretreatment 80 "C with crotonaldehyde- hydrogen mixture (lltorr/750 torr). Samples accumulated in the first hour of pretreatment. The overall yield of hydrocarbonsis around 4 Yo.

strate molecule takes place. Contrary to the Sn-Pt/Si02catalyst, the formation of C3hydrocarbons was not observed at all. The formation of butadiene indicates that over Sn-Pt/SiO, catalyst the abstraction of oxygen from crotonaldehyde takes place. It had been suggested that the oxygen abstracted from crotonaldehyde was involved in the oxidation of surface tin atoms. This may be the crucial step in the formation of 'metal ion - metal nanocluster' ensemble sites, i.e., in the formation of tin-containing Lewis-acid sites in atomic closeness to platinum, as shown in Figure ll.'07 Based on the above figure it has been suggested that the activation of the carbonyl group can take place on the 'Snn+ - Pt' ensemble sites as depicted in Figure 11. On the other hand the enrichment of tin at the surface of the bimetallic nanocluster strongly decreases both the adsorption of crotonaldehyde and readsorption of the formed crotyl-alcohol by their olefinic double bond. In-situ FTIR measurements provided further proof for the decarbonylation of crotonaldehyde over Pt/SiO2 catalysts.'21Under experimental conditions used in gas phase hydrogenation an intensive band of linear and a weak band of bridged

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

25

'Snm-Rensemb@site /

Figure 11

Activation of the carbonyl group in a,P unsaturated aldehydes over 'Sn"+-Pt' ensemble sites (Reproduced from ref. 107 with permission)

10.2

I

I

I

I

I

2250 2150 2050 1950 1850 1750 Wavenumber / cm-' Figure 12

Carbonyl region of FTIR spectra measured over Pt/Si02 and Sn-Pt/Si02 catalysts treated with crotonaldehydelhydrogen mixture at 80 "C.1 - PtISiO,, 2 - Sn-Pt/Si02 (Snanch/PtS = 0.5), 3 - SnPt/Si02 (Snanch/Pts = 2.3). (Reproduced from ref. 121 with permission)

26

Catalysis

carbon monoxide was observed on the parent Pt/Si02 catalysts as shown in Figure 12. Upon increasing tin content the intensity of the band of linear CO strongly diminished. This observation is in full agreement with the proposed decarbonylation of crotonaldehyde. These results can explain the low activity of the parent platinum catalyst, i.e. the low activity of Pt/Si02 catalyst can be attributed to the poisoning of Pt sites with adsorbed C0.'219125 The in situ formation of ionic forms of tin was proved by Mossbauer spectroscopy. These results are summarized in Figure 13 and Table 9.'@ As shown in data given in Figure 13 and Table 9, the freshly prepared Sn-Pt/Si02catalysts (H-3 type) have two forms of tin. The majority of tin is alloyed with platinum (isomer shift in the range of 1.40-2.31mm s-') resulting in a Pt rich (Sn-Pt(a))and a tin rich (Sn-Pt(b))phases., while the minority is in Sn2+ form (isomer shift in the range of 3.33-3.82 mm s-' Pt alloy phases). Detailed assignment of these samples has been given in section 2.1.1. In samples treated with crotonaldehyde/hydrogen mixture (samples 1-cr, 2-cr,

Table 9

~~

Results of Mossbauer spectroscopy on diflerent Sn-PtlSiOz catalysts. Comparison of catalyst freshly reduced and used in crotonaldehyde hydrogenation (Reproduced from ref. 109 with permission) ~

Samples, No

SnlPt (at/at)

Species

1

0.6

Sn4 Sn-Pt(a) Sn-Pt(b) Sn+* Sn4+ Sn-Pt(a) Sn-Pt(b) Sn+* Sn4+ Sn-Pt(a) Sn-Pt(b) Snf2 Sn+4 Sn-Pt(a) Sn-Pt(b) Sn+2 Sn+4 Sn-Pt(a) Sn-Pt(b) Sn+* Snf4 Sn-Pt(a) Sn-Pt(b) Sn+2

2

3

1-cr

1.o

1.3

0.6

2-cr

1.o

3-cr

1.3

+

IS, mms-'

QS, mms-'

F WHM, mms-'

RI,

%

-

-

1.56 2.31 3.33

1.23 0.75 0.75

70 14 16

0.97 1.03 0.87

38 47 15

1.03 1.13 0.76 0.77 1.26 1.01 0.78 1.09 1.07 1.12 1.19 0.78 0.79 1.31 0.78

32 54 14 7 62 19 12 13 31 46 10 15 21 54 10

-

-

0.71

1.56 2.20 3.69

-

-

1.49 2.15 3.86 0.69 1.56 2.3 1 3.64 0.7 1 1.42 2.28 3.57 0.85 1.54 2.23 3.64

1.19 0.96 -

1.20 0.80 -

1.18 0.44 -

0.83

-

-

Note. IS, isomer shift, mm s-I, Qs, Quadrupole splitting, mm s-I, FWHM, full width at half maximum, mm s-', RI, relative spectral areas (%). Errors on IS, QS and RI values are 0.03 mm/s, and +/- 10 re1 YO,respectively.Catalyst type (H-3).

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

-6

-4

-

2

0

v

Figure 13

2

4

6

27

8

I mm s '

Mossbauer spectra of Sn-PtlSiO, catalyst samples with diflerent SnlPt ratios. 1,2,3 - original samples, I- cr, 2-cr, 3-cr - samples treated with crotonaldehydelhydrogen mixture at 80°C (Reproduced from ref. 109 with permission)

and 3-cr, respectively) both Sn-Pt alloy phases and the Sn2+form are maintained, however in these samples due to the treatment with crotonaldehyde a new form of tin appeared with isomer shift in the range of 0.69-0.85 mm s-l. This form of tin was assigned to Sn4+.The formation of Sn4+ species in samples treated with crotonaldehyde has been considered as an indirect evidence for the 'reaction induced activation' of supported Sn-Pt nanoclusters by the substrate. This kind of activation of Sn-Pt/SiO, catalyst has been suggested in earlier studies.126 It has to be emphasized, as it emerges from data given in Table 9, that the Sn4+species are formed mostly from the platinum rich Sn-Pt(a) alloy phase. The high IS value of the Sn4+ species was attributed to the presence of delocalized diene-type ligands in the coordination sphere of Sn4+species.127Consequently, it is the first experimental evidence indicating that in the working alloy type Sn-Pt/Si02 catalyst, SnO,L, type surface organometallic species containing either butadiene or crotonaldehyde as stabilizing ligands are formed. These species should be formed in the atomic closeness of the active Pt-Sn phase. As a result, this evidence can be considered as the first proof of the formation of 'metal ion metal nanocluster' ensemble sites involved in the activation of unsaturated aldehydes as shown in Figure 11. 2.1.4.3 Selective Hydrogenation of Cinnamaldehyde. The hydrogenation of cinnamaldehyde was investigated over alloy type Sn-Pt/Si02catalysts under pressure. The corresponding experimental data are summarized in Table 10.1°7 As shown in Table 10, in this liquid phase hydrogenation reaction a similar

28

Catalysis

Table 10 Liquid phase hydrogenation of cinnamaldehyde over diflerent alloy type Sn-PtlSiOz catalysts. (Reproduced from ref. 107 with permission) Initial reaction ratea E x p . Catalysts, Pressure, No Snanch/Pts bar Overall W,,, W,,,

Selectivities,

SOL

SAL ~

1 2 3 4 5

Pt/SiO, 4 Sn-Pt,0.4 4 Sn-Pt, 1.6 4 Sn-Pt, 1.6 40 Sn-Pt, 2.4 40

2.2 17.7 1.8 3.7 2.2

1.1 13.0 1.7 3.4 1.6

1.1 4.8 0.06 0.2 0.5

40 17 5 4 9

%*

~

2 1 0 0 3

UOL

~~

58 82 95 96 88

*Selectivity measured at 10 % conversion.a in pmol GC,,-' s-' , selectivity measured at 10 Yo conversion. Reaction temperature = 60 "C,[Cl0 = 120 mmol/&,,, catalysts type (11), 50 g. Abbreviations: see Table 6.

trend has been observed as in the gas phase hydrogenation of crotonaldehyde, i.e., the introduction of tin into platinum strongly increased the reaction rate and leads to a very pronounced selectivity increase. However, the highest selectivity was obtained at relatively low Sn,,,h/Pt, ratio. It is interesting to note also that the parent Pt/Si02catalyst showed relatively high Sco selectivity in this reaction. This selectivity value was increased further by introduction of tin. The highest selectivity measured in this reaction was 96 YO at 10 % conversion. Data presented in Table 10 show also that the hydrogen pressure had a positive effect on the rates, however, it did not affect the high Sc4selectivity. In this reaction the conversion - Sc=oselectivity dependence showed also an increase, as shown in Figure 14.'07However, in this case the extent of the increase is relatively small. This type of kinetic pattern can also be attributed to the in situ formation of 'metal ion - metal nanocluster ensemble sites'. 2.1.4.4 Hydrogenation of Methyl Vinyl Ketone. The selective hydrogenation of the keto group in methyl vinyl ketone is considered a great challenge. The lack of positive results in this reaction has been attributed to (i) the steric hindrance of the alkyl groups, and (ii) the decreased reactivity of the keto carbonyl group compared to the aldehyde group in molecules with conjugated double bonds. Preliminary results obtained in the liquid phase hydrogenation of methyl vinyl ketone over a Sn-Pt/Si02 catalyst (Snanch/Pts= 1.43) are shown in Figure 15.'07 No unsaturated alcohol formation was observed on the parent catalyst. As emerges from these results, the SC+ selectivity is very low and decreases upon increasing the conversion. However, the use of catalyst treated previously with crotonaldehyde/hydrogen mixture in the gas phase resulted in a slight increase in the selectivity values. This increase was more pronounced at conversion below 10 YO, and it was maintained in the whole conversion range. This behavior was reproducible, however experimental conditions of the above treatment (i.e., duration, crotonaldehyde concentration, temperature) had no measurable influence on the selectivity improvement. Despite the modest results obtained in this reaction, the slight increase of the

29

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites 0.97

..->

c,

0.91

Y

I

-I

+

0 0)

v)

0.85

0.83 0

0.05

I

I

0.1

0.1 5

0.2

Conversion Figure 14

Selectivity-conversion dependencies in the hydrogenation of cinnamic aldehyde in batch reactor on Sn-PtlSiO, catalyst at H 2 pressure 3.5 bar - 0 and 40 bar Snonch/Pts = 1.61. Catalysts type (H-3). Temperature of the pretreatment in hydrogen = 300 " C ,reaction temperature = 40 "C, initial concentration of the substrate = 120 mmol/g,.,,, (Reproduced from ref. 107 with permission)

+

0

0.2

0.4

0.6

0.8

1

1.2

Conversion

Figure 15

Liquid phase hydrogenation of methylvinil ketone on Si02 supported Sn-Pt = 1.43. Eflect of the pretreatment with crotonaldehyde on catalyst, Snanch/Pts the selectivity. Selectivity - conversion dependencies without and with pretreatrespectively. Temperature of the pretreatment in crotonaldehyde, 0 and ment in hydrogen = 300 "C, temperature of the pretreatment with crotonaldehyde = 80 "C, concentration of crotonaldehyde = 30 mmol/dm3, reaction temperature = 40 "C, initial concentration of the substrate = 120 mmol/g,,,. (Reproduced from ref. 107 with permission)

+,

30

Catalysis

Sc4 selectivity strongly indicates on the involvement of ionic forms of tin formed during the treatment with crotonaldehyde. Consequently, in this reaction the 'metal ion-metal nanocluster' ensemble sites are involved in the activation of the keto group and the formation of unsaturated alcohol. However, these ionic forms of tin do not have enough strength to activate the keto group. 2.1.5 Use of Alloy Type Sn-Pt/Si02 Catalysts in Benzonitrile Hydrogenation. Earlier studies on liquid-phase hydrogenation of benzonitrile have shown that the addition of small amount of tin chloride to Pt/nylon catalyst increases also the rate of reduction of the -CN A similar trend has been previously reported for the hydrogenation of the carbonyl group in cinnamald e h ~ d e and , ~ ~ hydrocinnamaldehyde,12*or the -NO2 group in nitroben~ene.3~ Physico-chemical characterization of the Pt-Sn/nylon catalysts showed that tin was always present mainly as tin ions.'29 Therefore, it was suggested that the presence of tin ions enhanced the reactivity of the -CN group. The electrophillic nature of tin ions increased the polarization of the nitrile group present in the organic substrate, facilitating the attack of the chemisorbed hydrogen.33 The liquid phase hydrogenation of benzonitrile had also been investigated over alloy type Sn-Pt/Si02 catalysts prepared by CSRs.13' Prior to reaction the Sn-Pt/Si02catalysts (0.25 g) were re-reduced in H2at 300 "C. The hydrogenation of benzonitrile was carried out in ethanol at 60°C and 4 bar H2 pressure. Tin content of the catalysts ranged from 0.05 to 0.63 wt. YO,whereas Sn/Pt surface atomic ratios determined by chemisorption measurements were between 0.1 to 3.5. In the hydrogenation of benzonitrile the following reactions take place: (Ph = Phenyl group) PhCN 2 H2 PhCH2NH2 (10)

+

2 PhCH2NH2 PhCH2NH2

+

+

+ NH3 PhCH3 + NH3

(PhCH&NH

+ H2

+

(11)

(12)

The main product of the reaction was dibenzyalmine. As seen in Figure 16 the selectivity of dibenzylamine was around 75 YOand was not influenced by the level of the surface Sn/Pt atomic ratio.*30Upon increasing the surface Sn/Pt atomic ratio the selectivity of toluene formation decreased and that of the primary amine increased. The addition of tin to Pt led to an increase in the turnover frequency (TOF) by a factor of 2. TOF showed maximum at surface atomic ratio of Sn/Pt = 1 as shown in Figure 17.13' These results are in good agreement with earlier findings obtained in the liquid-phase hydrogenation of benzonitrile on catalysts prepared by the addition of tin chloride to Pt/nylon catalyst.33 Because the hydrogenation of the nitrile group takes place on the Pt sites, the rate enhancement observed upon addition of tin can only be explained by the cooperation of Pt and tin sites on the surface of alloy-type nanocluster and/or the metal support interface. The results obtained on alloy type Sn-Pt/Si02 catalysts strongly resemble results attained on silica supported Ni-Fe catalysts with 75 Yo

31

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

.

$ 80-,

.

Dibenzylamine

Benzylamine n 0.0

0.4

0.2

0.6

1.o

0.8

Sn/Pt surface atomic ratio Figure 16

Relationship between product selectivity and SnlPt surface atomic ratio in the hydrogenation of benzonitrile over Sn-PtlSiO, catalysts (Reproduced from ref. 130 with permission)

*O

1

15-

F

v,

u:

10-

?

5-

0

Figure 17

1

.

1

-

1

-

1

-

1

-

1

-

1

'

1

Correlation between turnover frequency and SnlPt surface atomic ratio in the hydrogenation of benzonitrile over Sn-PtlSiO, catalysts (Reproduced from ref. 130 with permission)

Ni and 25 % Fe ~ 0 n t e n t . In l ~ this ~ study the improved activity was attributed to the formation of either Ni-Fe"+ or Ni- Feo alloyed sites. It has been suggested that in both types of sites, due to the charge transfer form Fe, the charge on the nickel site increased. Based on the above analogy the activity increase over Sn-Pt/SiOz was explained by the formation of 'Sns+-Pt' ensemble sites on the surface of bimetallic

32

Catalysis

nanoclusters. It was suggested that highly dispersed positively charged tin species, by polarizing the triple bond, enhanced the reactivity of the -CN 2.1.6 Use of Alloy Type Sn-Pt/Si02Catalysts in Low Temperature CO Oxidation. 2.1.6.1 Literature Background. Pt/Sn02'32J33and Pd/Sn02134J35 catalysts are widely used as low temperature CO oxidation catalysts. With respect to these catalysts a synergism between the oxide and the metal phases has been suggested. Bond et al. proposed a bifunctional mechanism based on the spillover of both CO and oxygen from the noble metal to tin Sheintuch et al. also considered a spillover, but it was related only to C0.'35In another explanation the formation SnPt alloy phase has been ~uggested.'~~ Local temperature increase of Pt sites and the promotion of the reaction on adjacent SnO2 sites had also been propo~ed.'~' Recently an alternative reaction mechanism'38has been proposed suggesting that the reaction takes place at the Pt-Sn02 interface and the Lewis acid sites of the interface are involved in the activation of CO molecule as shown in Figure 18.1°7 The analysis of the above scheme and the scheme describing the activation of the carbonyl group in unsaturated aldehydes shows obvious similarities (compare Figures 5 and 11 with Figure 18). The S C e selectivity of alloy type Sn-Pt/Si02catalysts in the hydrogenation of crotylaldehyde strongly depended on the Sn/Pt (at/at) ratio and increased with the time on stream (TOS).'" This fact has been considered as direct evidence for the involvement of tin in aldehyde activation. Results obtained by in situ Mossbauer spectroscopy provided further prooflogthat in the beginning of the TOS experiment the Sn-Pt alloy phase poor in tin was oxidized by crotonaldehyde, and a correlation between the SC=O selectivity and the formation of Sn4+ sites was found. The Sco selectivity of Sn-Pt/Si02catalysts has been attributed to the in situ formed Sn4+species. For this reason it has been suggested that alloy type Sn-Pt/Si02catalysts can effectively be used in CO oxidation as the reaction atmosphere is favourable for the in situ formation of ionic forms of tin in the atomic closeness of platinum. Consequently, it has been expected that these Sn-Pt/SiO, catalysts should have higher activity than the parent Pt/Si02 and the commonly used Pt/SnO2 catalysts. 2.1.6.2 Low Temperature CO Oxidation Over Alloy Type Sn-Pt/SiO, Catalysts. Figure 19 shows selected Temperature Programmed Reaction (TPRe) results obtained in the oxidation of CO over Sn-Pt/Si02catalysts with different

I platinum I Sn02

Figure 18

Pt-Sn02 interface inuolued in the activation of CO in PtlSnO, catalysts (Accordingto ref. 135)

33

1 : Role of 'Metal Zon-Metal Nanocluster' Ensemble Sites

T

0

50

100

150

200

T PC

Figure 19

The injluence of tin content on the activity of Sn-Pt catalysts in the oxidation of CO. Temperature Programmed Reaction (TPRe) results. Catalysts: 0 Pt/SiO,; X - Sn/Pt = 0.22, (H-2) - type catalyst, A - Sn/Pt = 0.41, ( H - I ) type catalyst 0- Sn/Pt = 0.81 ( 0 - 3 ) - type catalyst (Reproduced from ref. 35 with permission)

. Figure 20

Activity of Sn-PtlSiO, catalysts in CO oxidation. Dependence of TSovalues of the Sn/Pt (atla0 ratio. (0)= 79 torr, (m) = 16 torr (Reproduced from ref. 107 with permission)

Sn/Pt The temperature (T50), at which 50 % CO conversion has been achieved was used to compare the activity of catalysts. The T5ovalues measured at two CO pressure levels are shown in Figure 20."' Results given in Figure 19 show that the parent Pt/SiOz catalyst is quite inactive. However, the introduction of tin significantly increases the activity of the catalysts, resulting in a pronounced decrease of the T50 values. Figure 20 shows that catalysts with Sn/Pt (at/at) ratio around 0.2-0.5 have high activity around room temperature at Pco = 16 torr. Further results given in Table 11 show that the TSovalues strongly depend on

34

Table 11

Catalysis

InJEuenceof the temperature of reduction on the activity of Sn-Pt/Si02 catalysts (Reproduced from ref. 35 with permission)

Catalytic runs

Catalysts, and pretreatment

7-50,

1 2 3 4*

(H) type, T H 2 = 340 "C (H) type, T H2 = 200 "C (0)type, T H=~340 "c (0)type, TH2= 200 "C (0)type, after run No4 followed regeneration at TH2= 340 "C

69 69 63 101 68

5*

"C

* TPRe experiment with temperature ramp up to 200 "C.

the temperature of reduction of the alloy type Sn-Pt/Si02catalysts in hydrogenP' There was only a minor difference between (H) and (0)type catalysts reduced at 340°C. However, the (0)type catalysts were sensitive for the temperature of reduction in hydrogen. These results are in a good agreement with results of Mossbauer spectroscopy discussed earlier (see Table 4). After reduction at 200 "C the (0)type Sn-Pt/Si02 catalysts are not fully reduced, they still contain both four- and bivalent forms of tin. Consequently, in highly active alloy type SnPt/Si02 catalysts the high activity has to be attributed to the stabilization of supported bimetallic nanoclusters and not to the formation of stabile ionic forms of tin. It should be emphasized that after TPRe run up to 350"C, all alloy-type Sn-Pt/SiOz catalysts without re-reduction had very low activity. Thus, on platinum nanoclusters covered by bulk type tin-oxide layer the number of required 'metal ion -metal nanocluster' ensemble sites is very low. The experimental data given in Table 11 strongly indicated that the activity of the alloy type Sn-Pt/Si02 catalysts was controlled by the surface composition of the bimetallic nanoclusters and the reduced form of the Sn-Pt nanoclusters is more active than a fully oxidized form. Additional experiments have proven that the activity of catalysts used in TPRe experiments can be completely restored after reduction in hydrogen at 340°C. 2.1.6.3 In situ Mossbauer Measurements. In situ Mossbauer spectra were taken during the room temperature oxidation of carbon monoxide on Sn-Pt/SiOz catalyst ((0-3) type (sample 11-3 in ref. 139) both at 27°C and -196°C. These spectra are presented in Figure 21 A and B, respectively and the corresponding data are summarized in Table 12.'39 The as received Sn-Pt/Si02catalyst contained Sn4+oxide as the dominating phase (see spectra a in Figures 21A and 21B). This finding is obvious because after preparation the sample was kept in air. The high temperature (300°C) activation in hydrogen results in the formation of bimetallic phases in overwhelming dominance (see spectra b in Figures 21A and 21B).

35

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

4

Figure 21

4

-

2

0 2 mmls

4

6

8 - 4 - 2 0 2 mmls

4

6

Mossbauer spectra of Sn-Pt/Si02catalysts ((H-3) type, SnlPt (atlat) = 0.68) under diferent experimental conditions. ( A ) spectra obtained at 27 "C,( B ) spectra obtained at -1 93 "C.Catalyst samples: (a) as received, (b) afer reduction at 300 "C, (c) in situ measurement in CO oxidation at 27 "C, (d) in situ reduction in hydrogen at 27 "C,and (f)in situ treatment in CO at 27 "C (Reproducedfrom ref. 139 with permission)

The reduced Sn-Pt/Si02catalyst had two alloy type species with isomer shifts between 1.20-1.56 mm s-'and 2.23 - 2.35 mm s-', respectively."' These species were the main components in this catalyst accounting for around 85 % of the overall amount of tin. These phases were attributed to Pt-rich (PtSn(a)) and tin-rich one (PtSn(b)) (see explanation given in section 2.1.2."' It is worth mentioning that after reduction at 300°C the Sn4+(ox)species were almost completely reduced. The Sn2+and Sn4+species found in reduced sample may have originated from organometallic species anchored at the periphery of the nano-cluster and subsequently become incorporated into the upp port.^^'^^^^ Another possible explanation for the presence of ionic forms of tin is that their full reduction requires slightly higher temperature and prolonged reduction time. Figures 21A and 21B show that the character of in situ Mossbauer spectra taken in the presence of CO + O2 at room temperature has been completely changed compared to the reduced form of the catalyst (compare spectra b and c). The oxidative atmosphere in the presence of CO resulted in (i)oxidation of both PtSn(a) and PtSn(b) alloy phases to Sn4+ in a high proportion and (ii) the

in CO at 25 "C

f

+

+

+

-

-

0.65(1)

-

0.65(2)

-

-

0.70(1)

-

-

0.65(2)

QS

0.97(3) 1.60(19) 1.40(8)

1.45(7) 0.96(7) 1.79(23) 1.53(12)

47 28 25

18 30 43 27

53 46 75 7

82 18

1.07(4) 1.53(17) 1.44(5) 1.87(7) 1.07(2) 0.90(14)

RI

FWHM 0.00(1) 1.39(2) 0.55(3) 2.76" 1.37(3) 2.4 1(6) O.Ol(1) 0.94(2) 3.4 1(2) 1.81(2) O.Ol(2) 1.24(2) 2.44(2) -0.04(2) 0.79(3) 3.31(2) 1.68(2) - 0.05(1) 0.88" 1.93(1)

IS

1.29(1) 1.66(4) 0.79( 14) 1.38(14) 1.66(16) 1.96(14) 1.22(1) 1.07(8) 0.81(6) 1.44(6) 1.03(3) 1.74(10) 1.74(5) 1.16(2) 1.03(8) 0.51(6) 1.57(4) 1.10(2) 1.46(10) 1.52(3)

0.72(1)

-

-

0.65

-

2.1 1

0.66

-

0.7 1

-

2.03

-

0.73

-

2.32"

-

FWHM

QS

77 K

3.5 4.7

77 23 4 12 41 42 62 11 4 23 31 34 35 58 10 2 29 50 21 30

4.7 (3.0) 5.1

4.9 5.9 4.6 6.6

4.1 4.8 3.1 6.1

OrfA

RI

(IS: isomer shift, relative to SnO,, mm s-'; QS: quadrupole splitting,mm s-', FWHM: line width, mm s-', RI: relative intensity %, b: fA = - d In (A3W/A,,) / dT x 10 -3) RI is a derived parameter with an error summarized from those of the components (baseline, intensity, FWHM - is estimated at ca. 10% relative YO) ("I constrained parameters

inCO O2 at 25 "C

+

in H2 at 25 "C

inCO + O2 at 25 "C

e

d

C

b

O.OO(1) Sn4+(ox) 1.37(7) PtSn(a) Sn4+(sf) Sn2 Pt Sn(a) 1.34(2) 2.24(5) PtSn(b) Sn4+(ox) O.OO(1) Sn4+(sf) 0.80'"' Sn2+ PtSn 1.58(3) Sn4+(ox) -0.01(2) Pt Sn(a) 1.23(4) 2.34(4) PtSn(b) Sn4+(ox) Sn4+ (sf) Sn2 PtSn Sn4 (ox) - 0.06( 1) 0.80" Sn 4 + ( ~ f ) PtSn 1.91(2)

As received Treated in H,at 300 "C

a

IS

Camp.

300 K

Results of in situ Mossbauer spectroscopy on (0-1)type Sn-PtlSiOz catalyst used in room temperature oxidation of CO (Reproduced from ref. 139 with permission)

Samples Treatment

Table 12

o\

w

37

1 : Role of 'Metal Ion-Metal NanocEuster' Ensemble Sites

appearance of a new alloy phase with IS value around 1.58-1.81 mm s-'(see spectra c in Figures 21A and 21B). This isomer shift is close to that of the PtSn (1:1) thus this new third alloy component has been denoted as PtSn (1:l) alloy phase. In the presence of CO + O2 at room temperature the main component is the bulk tin-oxide like Sn4+ species with IS value around zero. Furthermore, an additional new component appeared with IS between 0.80 - 0.94 mm s-' (and after a repeated CO + O2treatment at 0.79 mm s-', as well (see spectra e)). This value is beyond the lowest IS assigned for bimetallic tin alloys. Based on literature analogies'43this component has been assigned to a highly mobile and reactive surface species, Sn4+(sf),since its IS value is significantly different both from that of oxidic Sn4+and Pt rich alloy phases. The room temperature treatment of catalysts used in CO oxidation in a hydrogen atmosphere resulted in very pronounced changes in the composition of the catalyst. Interestingly, not only the mobile and highly reactive surface species, Sn4+(sf),but part of Sn4+(ox)is reversibly transformed to the original alloy phases. The reduction of the Sn4+(sf)phase is complete and the proportion of Sn4+(ox)drops from 62 to 31 YOarea in the spectrum and simultaneously the platinum-rich PtSn(a) and the tin-rich PtSn(b) components reappear in 34 and 33 YOrelative intensity, respectively. This result indicates that 82 YOof the original alloy content has been restored by room temperature hydrogen treatment (compare samples b and d). The re-reduction completely eliminated the newly formed PtSn (1:1) alloy phase, as well. Consequently, the intimate contact between Sn and Pt, both in the reduced and the working catalyst containing supported alloy-type nanoclusters, is convincingly demonstrated by the room temperature regeneration in hydrogen. This regeneration clearly proved the Sn4+ Snotransformation at room temperature both in PtSn(a) and PtSn(b) bimetallic phases. In addition, it is worth comparing the effects of reduction at 300 "C and room temperature re-activation in hydrogen: the corresponding IS values of PtSn(a) and PtSn(b) are very close and even their proportion is the same. In a subsequent room temperature CO + O2treatment partial re-oxidation of metallic tin to Sn4+(ox)and Sn4+(sf)was evidenced again with simultaneous disappearance of both PtSn(a) and PtSn(b) components in the catalyst (see spectrum e in Figure 21B). This experiment provided additional proof for the reversibility of nanocluster reconstruction. However, the treatment with pure CO (see spectrum f) resulted in only slight alteration in the ratio of the Sn4+(ox)and Sn4+(sf)phases without reconstruction of the original alloy phase. From the -d ln(A300/A77)/dT values the following information was obtained. The low values are characteristic of strong ionic bonds (for instance, for bulk-like SnOz phase a value of 1.0 x is ~btained.'~'However, in large organic complex molecules with looser bonds -d lnA/dT values close to have been obtained.l4 Strong interaction of tin with oxygen is reflected in the -d In A/dT = 3 x value estimated for Sn4+(ox)after contacting the catalyst with C O 0 2 mixture. In contrast, after room temperature re-activation in hydrogen (see

-

+

38

Catalysis

sample d), Sn4+(ox)exhibits a significantly larger -d lnA/dT value (5.87 x than in samples c or a. Further, the surface character of Sn4+(sf)is reflected in large -d In A/dT values (6.08 x after CO O2treatments. The comparison of PtSn(a) and PtSn(b) component -d In A/dT values provide further insight. For the Pt-rich PtSn(a) component, a relatively low value (4.1- 4.6 x is obtained, indicating the incorporation of tin into the core of bimetallic particles. Whereas, tin in PtSn(b) is more loosely bound as -d In A / dT = 4.8 - 6.5 x lod3indicates. Based on the -d In A/dT values the strongest Pt-Sn interaction is in the PtSn (1:1) component formed after treatment with CO (see sample f). In summary, results of Mossbauer spectroscopy studies indicate that the primary interaction with oxygen leads to a strong enrichment of tin in the surface layer in the bimetallic particles and tin, both in the surface layer and in the bulk of the nanocluster, is oxidized to Sm4+In the presence of CO O2mixture two forms of Sn4+oxide have been found, i.e., (i) a more stable one, Sn4+(ox)with isomer shift around zero, and (ii) a mobile one, Sn4+(sf)with IS = 0.79-0.94 mm s-l. During the room temperature CO oxidation no separation of the oxidized Sn4+forms was observed, as both the Sn4+(ox)and Sn4+(sf)components were easily reconverted in hydrogen at room temperature to PtSn(a) and PtSn(b) phases. During the CO + O2reaction, the surface of the nanoparticle containing of both mono and bimetallic metallic sites are probably covered mostly by Sn4+ species, while providing simultaneous access to CO on the surface of the formed PtSn (1:l) or Pt phases. In this way indirect experimental proof for the formation of 'metal ion - metal nanocluster' ensemble sites has been obtained.

+

+

2.1.6.4 In situ FTIR Measurements. Three different catalysts were investigated by in situ FTIR measurements: Pt/Si02, Sn-Pt/Si02(Sn/Pt (at/at) =0.19 sample A) and Sn-Pt/Si02(Sn/Pt (at/at) = 0.68, sample B). The characteristic frequencies of both the linear and the bridged CO bands are summarized in Table 13.'39 The addition of tin into the Pt/SiO2 catalyst decreased slightly both the frequency and the intensity of linear CO band. On the parent Pt/Si02 and the Sn-Pt/Si02, sample (A) the bridged CO band appeared around 1855 cm-'. However, on Sn-Pt/Si02, sample (B), due to the relatively high tin loading, no bridged CO band was observed. Similar results were obtained both on Pt/Si02145J46 and Sn-Pt/Si02147J48J49 catalysts. The observed behaviour of alloy type Sn-Pt/Si02 catalyst is characteristic for tin modified supported platinum catalysts and has been attributed to the dilution of the platinum surface with tin. Data in Table 13 show that in the parent platinum catalyst the addition of oxygen to CO had no detectable changes in the spectrum. However, in both Sn-Pt/Si02catalysts the addition of oxygen resulted in noticeable shift in the CO band frequencies. This shift indicates that in the presence of oxygen the surface composition of the supported tin-platinum nanocluster has been altered and the extent of these changes depends on the Sn/Pt ratio (compare samples (A) and (B)). After addition of oxygen catalyst sample (A) strongly resembles the parent platinum catalyst. On this sample the YCO frequency of linear CO band was

-

-

2083 (9.12) 2082 (9.30) -

-

2078 (8.90) 2082 (8.46)

-

1865 (0.88) 1865 (1.78) 1865 (0.65) 1860 (1.58)

a

Frequency and normalized intensity data (the latter are given in parenthesis). addition of CO first, introductionof oxygen first. Spectra measured after 40 min equilibration; n.d., not determined; n.m. not measurable. A - Sn/Pt = 0.19, B - Sn/Pt = 0.68.

4

+

In the presence of CO O2 mixtureb In the presence of CO 02mixtureC After room temperature reduction of sample No 2 in H2

2

3

Adsorption of CO

1

+

Experimental condition

N"

Catalysts

2070 (8.18) 2076 (6.66) 207 1 6.15 2083 (n.d.)

(n.m.) (n.m.) 1860 0.57 1860 (n.m.) 1860 (n.m.)

Table 13 Summary of in situ FTIR experiments: Frequency and Intensity Changes of the Linear and Bridged C O Adsorption Band under Condition of C O Oxidation at Room Temperature (Reproduced from ref. 139 with permission)

40

Catalysis

shifted towards the high frequency region (AVco= 4 cm-') and the intensity of the bridged CO band increased significantly. In sample (B) after exposure to oxygen the band position of linear CO is also shifted (Avco = 6 cm-') and there is a noticeable half width broadening. The analysis of band intensities indicates also that substantial replacement of chemisorbed CO by oxygen can only be seen in sample (B). The treatment of the catalyst used in CO oxidation for one hour at room temperature resulted in a decrease in the v C 0 band frequency from 2076 to 2071 cm-' (see Table 13 exp. No3). This new value is very close to the frequency of the v C 0 band measured on the fully reduced (B) sample (see Table 13 exp. N"1). Consequently, partial reduction of the formed oxide phase at room temperature has also been shown by in situ FTIR measurements. Results presented in Table 13 indicate also that the addition of oxygen to the Pt-Sn/Si02 (B) catalyst followed by introduction of CO has resulted in substantial differences in the FTIR spectra of chemisorbed CO (see experiment NO4 in Table 13).In this sample the low frequency v C 0 band at 2070 cm-' disappeared and the spectrum strongly resembled that of the parent platinum ((vCO)lin = 2083 cm-', (vCO)br= 1860 cm-l). This result indicates that upon contacting the supported Sn-Pt nanocluster with pure oxygen fast oxidation of the supported alloy phases takes place. The oxidation leads to rapid segregation of phases and the subsequent addition of CO cannot restore the surface composition formed in the presence of CO + O2 mixture (compare exps. No2 and No 4). This segregation leads to the reappearance of pure Pt sites ((vCO)lin = 2083 cm-', (vCO)br = 1860 cm-'). Catalytic experiments showed also that the activity of Sn-Pt/Si02 catalyst was very low when oxygen was introduced first. Difference spectra obtained on sample (B) in the presence of CO + 0 2 mixture and in pure CO are shown in Figure 22A and Difference spectra were taken at 5, 15,30 and 60 minutes after exposure of oxygen.lo7 Results obtained at 10 torr (see Figure 22A) imply that upon exposure to oxygen the intensity of the band at 2070 cm -', characteristic for the adsorption of CO on bimetallic nanoparticles, decreases, while that of the band at 2086 cm- ', characteristic for the adsorption of CO on pure platinum, increases. Parallel to the above changes, a very pronounced increase was also observed in the intensities of the (vCO)b,band at 1840 cm-'. However, the difference spectra indicate that the oxygen induced surface reconstruction is relatively fast as there is no measurable difference between these spectra. The increase of the concentration of CO up to 50 torr resulted in slightly slower, but pronounced changes induced by oxygen (see Figure 22B). The character of these changes is similar to that measured at Pco = 10 torr. However, after eight minutes only minor alteration can be seen, although after 25 minutes there are no further measurable changes in the band intensities both around 2070 cm-' and 2086 cm-'. Parallel to the above changes, a notable increase was also observed in the intensities of the (vCO)b, band at 1840 cm-'. This change appeared to be continuous in the whole time interval. These results might indicate that the bridged CO band is more sensitive to reflect minor surface reconstruction than the linear one. The comparison of results obtained at two

41

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

c

2200

I

I

2100

2000

1900

Wavenurnbers, cm.'

Figure 22

1800

1700

Wavenumbers, cm"

Diflerence FTIR spectra of adsorbed CO in the presence and absence of oxygen on Sn-Pt Si02 catalysts ( H - 1 type, SnlPt (atlat) = 0.68). A - Pco = 10 torr, Poz = 5 torr; B - Pco = 50 torr, PO2 = 25 torr, in situ room temperature spectra: (a) 5 min,, (b) 15 min, (c) 30 min and (d) 60 min of reaction (Reproducedfrom ref. 107 and 139 with permission)

CO partial pressure levels indicates a pressure dependence of surface reconstruction, i.e., the lower the CO pressure the higher the rate of oxygen induced reconstruction of supported Sn-Pt nanoclusters. The above results strongly indicate that upon introduction of the CO oxygen mixture the supported Pt-Sn alloy phases are transformed and segregated. This is a time dependent process, however the rate of these surface transformations is relatively fast. The net results of changes induced by oxygen are as follows: (i) the number of bimetallic tin-rich alloy sites around 2070 cm-' diminishes, and (ii) and number of pure Pt sites what adsorb CO around 2086 cm-' increases. The formation of pure Pt sites have been evidenced in Sample (A), as well, i.e., the site separation is characteristic both for platinum-rich and tin-rich alloy phases. These results are in a good agreement with the results of Mossbauer spectroscopy, where in the presence of oxygen the transformation of both PtSn(a) and PtSn(b) alloy phases has been shown and the formation of tin oxide like phases has been demonstrated (see Table 12). Consequently, results of in situ FTIR provided further proof for the PtSn + Sn4++ Pt conversion taking place during the low temperature CO oxidation.

+

2.1.6.5 Structure of Alloy Type Sn-Pt/Si02 Catalysts used in Low Temperature CO Oxidation. Both Mossbauer and FTIR spectroscopy provided sufficient proof of surface reconstruction during the low temperature CO oxidation. However, the above reconstruction appeared to be reversible as the reversible interconversion of PtSn c* Sn4+ + Pt was demonstrated by both spectroscopic techniques. This reversibility can only be achieved if the segregation described above is within the supported nanoparticle, i.e., when surface reactions involved in CO oxidation do not result in formation of separate Pt and tin-oxide phases on the silica support. In in situ experiments it was demonstrated first time that both metallic and ionic species can co-exist in the same supported particle (nanocluster) and both

42

Catalysis

the parent alloy phases and the newly formed ionic and metallic phases appeared to be highly mobile and reactive even at room temperature. Based on these results it had been suggested that in room temperature CO oxidation over Sn-Pt/Si02 catalysts the supported bimetallic nanocluster is oxidized and the oxidation of the alloy phases leads to strong reconstruction. The net result of these transformations is the formation of the following phases or sites: Tin oxide phase (IS = 0.0 mm s-', QS = 0.6-0.7mm s-') Appearance of a highly mobile Sn4+(sf) phase (IS = 0.79-0.94mm s-') Formation of free platinum sites (vCOlin= 2086 cm-', vCObr= 1840 cm-') Appearance of a third alloy phase, PtSn (1:l)(IS = 1.6-1.9mm s-'). The schematic view of the supported nanocluster prior and after its reconstruction in the presence of CO and O2mixture is shown in Figure 23 and 24. Based on the above results it had been suggested that the oxidation of CO takes place at the 'Sn 4+ - Pt' ensemble sites formed in situ. These sites are considered to be 'metal ion - metal nanocluster' ensemble sites. The possible

SnO, \

SiO,

Figure 23

Schematic view of silica supported Sn-Pt nanocluster after reduction in hydrogen at 300 "C.Catalyst type (H-3)

sio, fl hTI

well dispmried Pt or Ptsn (1 :l)p)use

mobite Snf++f> Sni(0x) segregated Pt

Figure 24

Schematic view of silica supported Sn-Pt nanocluster after its reconstruction in CO O2mixture at room temperature. Catalyst type (H-3) (Reproduced from ref. 139 with permission)

+

43

1 :Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

mode of the activation of the CO molecule on the above sites was also modeled and calculated. The results of these calculations are discussed in the next paragraph. 2.1.6.6 Computer Modeling of CO Activation. Results presented so far indicate that the chemical nature of heterogeneous catalytic activation of both the carbonyl group in aldehydes and the CO molecule should have a common basis. The high activity of catalysts prepared by using CSRs can be related to the activation of the carbonyl moiety by Sn4+species formed during the reaction. The ionic forms of tin stabilized on the platinum surface attract electrons either from the oxygen atom of the CO molecule or the carbonyl group. The flow of electrons from the CO (or carbonyl moiety) to Sn4+strongly weakens the carbon-oxygen triple (or double) bond. Consequently, in CO oxidation over the Sn-Pt/Si02 catalysts prepared by using CSRs the C O molecule is strongly perturbed and reacts at low temperature. The above perturbation of the CO molecule has been modeled and cal~u1ated.l~~ In the first approach ab initio Hartree-Fock calculations were made for hypothetical (CO + Sn'"))clusters, where n = 0, + 2 and +4. Results for the single CO molecule and for cases n = 0 and +4, are summarized in Table 14. These data undoubtedly demonstrate that the carbon-oxygen bond is weakened in the proximity of the charged tin atom and the weakening effect is proportional to the charge of the tin atom. The calculation of the Mulliken charges and the bond order of the contributing atoms provide further information (see data given in Table 15). These data show that the effect of the neutral tin atom in the proximity of the CO molecule on the atomic charges and bond orders is insignificant. However, when the tin atom is charged, electron transfer occurs from the CO molecule to the tin cation. The charge of the carbon atom in the C O molecule becomes more positive, and this makes the oxygen atom more negative and reduces the positive charge of the tin cation. It has been shown that the increase in the nucleophilic nature of the oxygen atom makes it possible to form a strong interaction between the C O molecule and the tin cation as shown in Table 16. As emerges from these data the Mulliken bond order increases and the distance decreases between the 0 and Sn'")atoms if the tin atom is charged. However, the increase of the positive charge on the

Table 14

Bond lengths and Mulliken bond orders of the carbon-oxygen bond in the carbon monoxide molecule and in CO + Sn'") (Reproduced from ref. 35 with permission) bond length (A)

co CO CO

+ Sno + Sn4+

Mulliken bond order

STO-3G

3-21 G ( * )

STO-3G

3-21 G( *)

1.15 1.15 1.28

1.13 1.13 1.28

2.51 2.46 1.61

2.22 2.10 1.25

44

Catalysis

Table 15

Mulliken atomic charges in the carbon monoxide molecule and in CO Mn) complexes (Reproduced from ref. 35 with permission)

+

C

co CO CO

Sn

0

STO-3G

3-21 G ( * )

STO-3G

3-21G(*)

STO-3G

3-21G(*)

0.20

0.44 0.53 1.48

- 0.20

-0.44 -0.47 -0.64

-0.03

-0.06

3.28

3.16

+ Sno 0.24 + Sn4+1.03

-0.21 -0.31

-

carbon atom reduces the electronic charge density in the proximity of the carbon nucleus, and thus reduces its ability to form a covalent bond. Hence the result of these effects is the weakening of the carbon-oxygen bond in the carbon-monoxide molecule. Consequently, the results of the ab initio Haertee-Fock calculations strongly support the involvement of the electron transfer from the CO molecule to the Sn4+site in the activation of the CO molecule. According to the results of in situ Mossbauer spectroscopy the formation and stabilization of the PtSn (1:l) alloy phase has been shown under condition of room temperature CO oxidation (see results presented in Table 12). One of the most active surfaces of the PtSn (1:l) alloy phase, the (110)phase, was chosen to model the interaction of the CO molecule with the metal surface.'39 The computer modeling and the related calculations were made on density functional level. In this model a small cluster of the (110) surface of the PtSn phase94as shown in Figure 25a, was selected in order to calculate and investigate the interaction of the CO molecule with the metal surface. Two alignments of the CO molecule relative to the metal cluster were examined, for total charges N = 0, 4. For the linear alignments, the CO molecule is perpendicular with the surface and no S n ( N ) . . 0 . . interaction is possible. In contrast, for the bent structures, the oxygen atom of the CO molecule chemisorbed on the Pt is near to a tin atom, allowing interaction between them. The structures of the above alignments are shown in Figures 25b and 25c, for neutral surface. The results of the above calculations for the CO molecule and for its interaction with N-fold charged metal clusters are presented in Table 17, for N = 0 and +4.13' These results show that with respect to the Sn(N).... 0 interaction, and the parameters of the CO bond the alignment of the CO molecule relative to the

+

Table 16

Distances and Mulliken bond orders between the oxygen and tin atoms in CO + Sn(*) (Reproduced from ref. 35 with permission) distance

CO CO

+ Sno + Sn4+

(A)

Mulliken bond order

STO-3G

3-21 G ( * )

STO-3G

3-21 G ( * )

2.66 1.93

2.72 2.0 1

0.08 0.85

0.12 0.76

45

I: Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

B

W

C

A Figure 25

Alignment of the CO molecule chemisorbed at the cluster of the (I 10) surface of the PtSn (1:l) alloy phase. ( A ) the (I 10) surface of PtSn (1:l) alloy phase (the PtSn cluster used in D F T calculation is shown in black): ( B ) linear coordination of CO on the PtSn cluster; ( C ) bent coordination of CO on the PtSn cluster. The bent structure is required for the activation of the CO molecule (Reproduced from ref. 139 with permission)

metal cluster is more important than the total charge of the system. The calculations reveal that the alignment strongly affects the CEO bond, as the bent structures have more significant effect on the bond lengths and bond orders than the linear structures. On the other hand, it has also been demonstrated that for the neutral system the linear structure is energetically more stable than the bent alignment, while the opposite relationship has been found for the charged alignment. These results support the hypothesis that on the uncharged cluster the chemisorbed CO molecule is perpendicular to the metal surface with no Sn(N).... 0 interaction. However, if the surface is charged, the alignment of the molecule is bent and the CEO bond is weakened because of the involvement of ‘Sn4+-Pt’ ensemble site in this interaction. Consequently, these calculations strongly support the relevance of our hypothesis with respect to the involvement of ‘metal ion - metal nanocluster’ ensemble sites and the C - 0 - Sn4+interaction in the increased activity of alloy type Sn-Pt/SiOz catalysts in low temperature CO oxidation.

46

Catalysis

Table 17 Bond lengths and Mayer bond orders of the C = 0 bond in the (cluster)N complexes and the carbon-monoxide molecule and in CO ienergy diflerences between the bent and linear structures (AE = Ebent- Elinear)

(Reproduced from ref. 139 with permission)

co CO CO

+ (cluster)’ + (cluster)4+

bond length (A)

Mayer bond order

linear

linear

bent

1.317 1.338 1.324

E

bent

(kcaljmol)

1.42 1.42

- 34.89

2.46 1.415 1.406

2.02 2.18

205.41

With respect to the involvement of the ‘Sn4+-Pt’ensemble sites in the activation of both the carbonyl group in unsaturated aldehydes and the CO molecule, the following difference has to be discussed. Upon using alloy type Sn-Pt/Si02 catalysts high selectivity for the formation of unsaturated alcohols was obtained at Sn/Pt (at/at) > 0.8. Contrary to that in CO oxidation, the best results were attained at Sn/Pt (at/at) = 0.4 - 0.7. The above difference was explained by a specific need for surface sites to suppress the readsorption of formed unsaturated alcohol. The suppression of the re-adsorption of unsaturated alcohol can only be achieved if platinum is strongly diluted by metallic tin, i.e., the surface concentration of the tin-rich alloy phase is relatively high. Therefore this reaction requires alloy type Sn-Pt/SiOz catalysts with high Sn/Pt (at/at) ratio. In CO oxidation there is no need for tin-rich alloy sites. The in situ formation of Pt or SnPt (1:l)sites enhances the chemisorption of both reactants. This is the reason that the segregation process (see Figure 24) is not harmful for the reaction. Contrary to that, it is very beneficial for the formation of ‘Sn4+-Pt’ensemble sites. However, at high Sn/Pt (at/at) ratio (Sn/Pt >1)the formed Pt (or SnPt (1:l))sites might be covered by the inactive Sn4+(ox) phase formed in situ. In this way the number of sites required for CO chemisorption decreases and the rate of CO oxidation diminishes, as shown in Figure 20. 2.1.7 Summary on Alloy Type Sn-Pt/Si02Catalysts. Results shown in these case studies provided unambiguous evidence that in alloy type Sn-Pt/Si02catalysts Lewis type ionic Sn”+or polarized metallic Sns+ sites can be formed. These sites due to the formation of alloy type supported nanoclusters are in atomic closeness. In this way ‘Sn4+-Pt’or MIMNES can be formed. In most of the cases these sites are formed in situ. Characteristic feature of these new types of active sites is the activation of different polarizable functional groups, such as carbonyl or nitrile, and the triple bond of carbon monoxide. Due this polarization the activity of these groups increases strongly resulting in high selectivity and high reaction rate in the hydrogenation of the carbonyl or nitril group and high rate of the oxidation of the CO molecule.

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

47

CO Oxidation on Supported Gold Catalysts. - 2.2.1 Literature Background. The first papers on the preparation and use of highly active supported gold catalysts were published in the late 1 9 7 0 ~ . ' ~ ~In3 ' ~ the ~ last 10-15 years supported gold catalysts have received great academic interest, due to their high activity in low temperature CO oxidation. Supported gold catalysts have also been used in combustion of saturated hydrocarbons, reduction of nitrogen oxides, hydrogenation of unsaturated carbonyl compounds, e t ~ . The ' ~ ~activity of supported gold catalysts appear to be a function of the gold dispersion and the concentration of interfacial sites. Supported gold catalysts have recently been reviewed by different a u f h o r ~ . ' ~ ~ ~ ' ~ ~ In studies on supported gold catalysts different gold precursor compounds and miscellaneous support materials, such as Ti02,'569'579'587'59 Ni0,16' Fe2o316071 56 c0304156"60 CuO,'@Zr02,158,161 Mg(OH)2'59and various transition metal hydroxides'62have been investigated. The support materials were classified as active (Fe304, TiOz, NiO,, COO,)and inert materials (A1203, MgO).'63 From results presented so far on low temperature CO oxidation the following main conclusions can be drawn:

2.2

(i) supported gold catalysts are highly active even at -70 "C, (ii) only nanoclusters with particle sizes in the range 1-6 nm show high activity, (iii) not only the size, but the shape of the nanoclusters has a significant influence on the activity of supported gold catalysts, (iv) supports with redox properties strongly enhance the activity of supported gold catalysts, (v) a number of different reaction mechanisms have been proposed in low temperature CO oxidation. Several models have been suggested to explain the high activity of gold catalysts. The proposed models of low temperature CO oxidation are summarized in Figure 26.'63 The first model assumes that oxygen adsorption takes place directly on the gold nanoparticles and the role of support is to stabilize very small gold particles with highly reactive gold sites or crystallite faces (reaction pathway 1 in the

Figure 26

Possible reaction schemes for the CO oxidation over supported Au catalysts (AulFe304) (Reproduced from ref. 163 with permission)

48

Catalysis

above f i g ~ r e ) . ' ~In. ' the ~ ~ second model it is suggested that the oxygen adsorption takes place on the support or at the metal-support i n t e r f a ~ e . ' ~ ~Oxygen ~'~~~'~*~ vacancies on the semiconductor type support materials, such as Fe304,T i 0 2 or ZnO may be considered as sites involved in oxygen adsorption (see reaction pathways 2a and 2b in the above figure). However, it was not discussed whether this oxygen molecule dissociates into Oads. or reacts directly with adsorbed CO (pathways 2a or 2b, respectively). In the third pathway it is suggested that oxygen after its adsorption dissociates, producing lattice oxygen, what can subsequently react at the metal-support interface with CO adsorbed on gold (see pathway 3 in the above figure).'66*'68 Reaction pathway 2 strongly resembles that of proposed for CO activation over Pt/Sn02 catalysts at the Pt-Sn02interface (see Figure 18). Costello et al. studied the deactivation and regeneration phenomenon during room temperature CO oxidation over Au/y-A1203catalyst, which was as active as the most active supported Au catalysts reported in the l i t e r a t ~ r e .The ' ~ ~ initial rapid loss of activity could be avoided if either hydrogen or water vapor was present in the reaction mixture. Thermal treatment above 100°C in a dry atmosphere also deactivated the catalyst. The original activity could be recovered by exposure of the deactivated catalyst to either hydrogen or water vapor at room temperature. These results suggested that hydroxyl group, most likely associated with a Au(1) cation located at the gold nanocluster - support interface is involved in CO activation as shown in Figure 27 and Scheme 4.17*It has been proposed that the active site is an ensemble of Au+-OH together with Au(0) atoms. As shown in Scheme 4 the CO oxidation was proposed to proceed via the insertion of CO into the Au+-OH bond to form a hydroxycarbonyl, which is oxidized to bicarbonate. Decarboxylation of bicarbonate completes the reaction cycle.171 The above results and some recent literature data related to the involvement of ionic form of gold in CO a c t i ~ a t i o n , 'encouraged ~ ~ ~ ' ~ ~ the authors of this review to demonstrate that MIMNES can also be formed in supported gold catalysts. If a non-active support, such as MgO is used in this case (AU)'+,-(AU)~, type ensemble sites can be formed provided the reduction of the gold precursor compound is controlled. In the literature there has been no investigation of the role of additives or

Au

Figure27

O H

Schematic drawing of an active site for CO oxidation over supported gold catalysts (Reproducedfrom ref. 171 with permission)

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

Scheme 4

49

Oxidation of CO over supported gold catalyst. Involvement of 'metal ion - metal nanocluster' ensemble sites in CO activation (Reproduced from ref. 171 with permission)

modifiers on the activity of supported gold catalysts in low temperature CO oxidation. The aim of this work is to demonstrate that ascorbic acid can modify Au/MgO catalysts to improve their activity in low temperature CO Oxidation. 2.2.2 Preparation and Characterization of AulMgO Catalysts. 2.2.2.1 Preparation. Three different Au/MgO catalysts were prepared using HA&& as a precursor compound:'74(i)base, (ii)re-suspended and (iii)modified with ascorbic acid. The modification of the base catalyst with ascorbic acid is associated with a slight color change. The color of the re-suspended catalyst after drying at 100 "C is beige with a light gray tone. The modified catalyst samples after modification with ascorbic acid and subsequent drying at 100°C were light bluish-violet or light pinkish-red. Ascorbic acid is considered to be a mild reducing agent, consequently it is suggested that in its presence the spontaneous auto-reduction of the gold precursor compound observed both in the impregnation and drying period and will be more controlled. It has been assumed that (i) by changing the amount of ascorbic acid added the ratio of ionic/metal gold can be altered and (ii) the controlled reduction will increase the ratio of smaller gold particles required for low temperature CO oxidation. 2.2.2.2 Characterization by U V-VIS Spectroscopy. Figures 28A and B show the diffuse reflection UV-VIS spectra of the different gold catalysts prior to and after reduction in hydrogen at 350 "C,re~pective1y.l~~ Three adsorption bands around 240,390 and 565 nm have been observed in all samples. Based on literature data these bands correspond to (i) Au+ cations, (ii) (Au),b+,and (iii) (Au),, respectiveIY.''~Figure 28A definitely shows that during the preparation of these samples spontaneous autoreduction of the gold precursor compound takes place. The autoreduction leads to the formation of both ionic and metallic forms of gold.

50

Catalysis

0.4

0.1

0.0

1

I

200

-

I

300

.

I

400

.

.

500

~

600

I

-

700

I

.

800

,

.

I

-

900

Wavelength, nm

A

0.74 0.6 0.5

1: 0.2 0.1

0.0 200300400500600700800900 Waveknght, nm

B

Figure 28

Difluse reflectance U V-VIS spectra of diflerent gold catalysts. A ) Catalysts prior to reduction. 1 - Au/Mg(OH)?P, 2 - O . ~ - A U / M ~ ( O H )3~-~2.4O~, after reduction in hydrogen at 350 C. 1 A U / M ~ ( O H ) ~B~)OCatalysts ~. Au/MgO"""",2 - 0.7-Au/MgOmod, 3 - 2.4-Au/MgOmod (Reproducedfrom ref. 174 with permission) O

The spectra reveal also that (i) autoreduction begins under conditions of the resuspension (see sample (1)in Fig. 28A, (ii) there is a certain amount of ascorbic acid which is not involved in the autoreduction (compare samples (1) and (2) in Fig. 28A), (iii) above a minimum level the addition of ascorbic acid increases the amount of both metallic gold and the positively charged gold nanoclusters, (Au), and ((AU)~'+, respectively, while amount of Au+ cations is practically constant. The reduction of samples in hydrogen at 300 "Cresulted in further increases in the intensity of bands at around 390 and 560 nm, while the band at around 240 nm completely disappeared. The disappearance of the band at 240 nm indicates

51

1: Role of 'Metal Zon-Metal Nanocluster' Ensemble Sites

the complete transformation of the gold precursor compound. After reduction in hydrogen the character of the UV-VIS spectra in the range between 300 and 600 nm is similar to that stabilized prior to the reduction. However, as shown in Figure 28B for samples 1 and 2, the band at around 560 nm is broader than in the corresponding samples shown in Figure 28A. In sample 3 the plasma-resonance peak is shiftec from 560 to 545 nm and the peak is narrower compared with the corresponding peak in the other two samples. These findings indicate that the addition of ascorbic acid enhances the stabilization of supported gold nanoparticles in the sizes below 5.0 nm required for low temperature CO oxidation. It should be noted that the reduction in hydrogen does not lead to the disappearance of (Au),"+ species. It has been suggested that the positively charged gold nanoclusters play a crucial role in the activation of CO. 2.2.2.3 Characterization by CO Chemisorption. Figure 29 shows the adsorption isotherms of CO over re-suspended Au/MgO ~ata1yst.l'~ It has to be emphasized that it was the first attempt to demonstrate that the amount of CO chemisorbed on gold could be measured. The calculated amount of chemisorbed CO shows a saturation level around 800-1000 torr of CO. Selected results of CO chemisorption are summarized in Table 18.'74These data indicate that both Au/MgO catalysts and the MgO support chemisorb CO. However, the latter has much less chemisorption capacity than the gold containing samples. The amounts of CO chemisorbed on resuspended Au/MgO catalyst (catalysts (I) and (11))roughly correspond to CO/Au = 0.10 - 0.12, i.e., the dispersion of gold in these Au/MgOcatalysts is about 10 - 12 %. Results presented in Table 18 show that the modification with ascorbic acid resulted in a substantial increase in 0.25 0.20

T

E 0.15

0.05 0.00

I

I

I

I

I

I

0

200

400

600

800

1000

I 1200

Pressure, torr

Figure 29

Adsorption isotherms of CO on Au/MgOresusp catalyst (Sample N o 3 in Table 18.). W - total volume of adsorbed CO, A - volume of physisorbed CO, 0volume of chemisorbed CO. Amount of catalyst: 0.13 g (Reproduced from ref. 174 with permission)

52

Catalysis

Table 18

CO chemisorption on MgO and diferent Au/MgOresusp catalysts (Reproduced from ref. 174 with permission) ~

VCO,

No

Catalyst

1 2 3 4 5 6 7

MgO Air and H2at 350 "C Au/MgO""usP(I)b Air and H2 at 350 "C Au/MgOesusP(II) Air and H2 at 350 "C Au/MgOesuS~(II)H2at 350 "C O . ~ - A U / M ~ O ~H2 " ~at 350 "C 0.7-Au/MgOmod Air and H2 at 350 "C 2.2-Au/MgOmod H2 at 350 "C

Treatment

mils

~~

CO/Au (atlat)

measured

Correcteda

0.095 0.554 0.547 0.569 0.540 0.525 0.813

-

-

0.642 0.632 0.675 0.630 0.600

0.113 0.111 0.120 0.110 0.103 0.176

1.005

Corrected by the amount of CO chemisorbed on MgO and for the amount of water lost due to transformation of Mg(OH), to MgO, prepared in amount of 1.3 g, (11) - reproduction of (I) in amount of 1.3g. Thermal treatment: heating in air to 350 "Cand keeping 1.5 h at this temperature followed by reduction in hydrogen for 1 h. a

the amount of CO chemisorbed and the dispersion value increased to 17.6 %. This dispersion value might correspond to a particle size below 5 nm. 2.2.2.4 FTIR Results on Adsorbed CO Molecule. C O chemisorption was performed on modified Au/MgO catalyst after two treatment procedures: (a) treatment in air followed by reduction in hydrogen, and (b) treatment in hydrogen. The corresponding spectra are presented in Figure 30A and B.'74 Over the catalyst pretreated in air and reduced in hydrogen the exposure of CO resulted in one relatively broad carbonyl band at around 2115 cm-' (see Figure 30A). Upon increasing the duration of CO exposure, the position of this peak shifted slightly to the low frequency region, to 2106 cm-'. Parallel to this shift a very broad band appeared in the low frequency region between 1800 and 2200 cm-I. The switch of the CO flow to pure helium resulted in complete removal of the CO band around 2106 cm-', but this had no influence on low

A

iIo.005

' '-D

3% CO (10 min)

-

i";"--

2

.-.

3%CO (30min)

2106

1:

/ ,

-

~

,.3,"/. CO (60min)

,

6

2125

100025

c-

g; 48 -- -J' ~'.~----.JhCO(3Ornm) 1

9

.3% CO (60 mn)

'

?

.

,

Figure 30

.

,

.

,

.

,

.

I

7

.

1

.

I

.

I

.

I

.

I

FTZR spectra of chemisorbed CO on 3.4-Au/MgOmodcatalyst, carbonyl region. A - catalyst pretreated in air followed by reduction in hydrogen; B - catalyst pretreated in hydrogen (Reproduced from ref. 174 with permission)

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

53

frequency the broad band. Complete removal of this broad band was observed only after thermal treatment in helium above 250 "C after 1 h. The sample treated only in hydrogen showed two overlapping CO bands at 2125 and 2105 cm-' (see Figure 30B). Almost similar overlapping CO bands have been obtained on Au/A1203 catalyst after its treatment with oxygen.'76The increase of the CO exposure time resulted in slight changes in the ratio of these two bands in favour of the lower frequency band. It should be mentioned that on this sample the appearance of the broad carbonyl band between 1900 and 2200 cm-I is negligible. However, the broad band between 1800 and 2200 cm-' was not observed in this sample. According to Boccuzzi et ~ 1 . the l ~ linear ~ CO band on small gold particles appears at 2106-2116 cm-'. Grunwald et a1.16' ascribed the bands around 21 112123 cm-' to CO chemisorbed on step and kink sites, while the bands around 2128-2135 cm-' were assigned to CO chemisorbed on positively polarized gold sites. Consequently, the appearance of a CO band at 2125 cm-' and above, obtained on modified Au/MgO sample pretreated only in hydrogen indicates that this catalyst might contain more ionic forms of gold than the one treated both in oxygen and hydrogen. This form of gold seemed to be quite stable as the duration of CO exposure did not result in a notable intensity change (see Fig. 30B). The asymmetric character of the CO band on a sample pretreated both in air and hydrogen indicates that the Au/MgO catalyst might also have ionic gold species, but in this sample the proportion of the ionic gold is relatively low. Based on literature data discussed above the following assignment was done for the carbonyl bans presented in Figures 30A and B: (i) the carbonyl bands at 2105 cm-': CO chemisorbed on metallic gold nanoclusters (Au)',; (ii) the carbonyl bands at 2125 cm-': CO chemisorbed on positively charged gold species (Au)'+,; (iii) the broad band between 1800 and 2200 cm-': chemisorbed CO, or spilled over to MgO. 2.2.3 Use of Au/MgO Catalysts in Low Temperature CO Oxidation, Figure 31 and Figure 32 show results in the temperature range of -30-250 "C obtained by Temperature Programmed Reaction (TPRe) technique using different types of Au/MgO cata1y~ts.l~~ These results clearly show that the base Au/MgO catalyst was less active than the re-suspended one, while the catalysts modified with ascorbic acid have the highest activity. The TPRe curves presented in Figure 31 and 32 show also unusual TPRe pattern of all Au/MgO catalysts. Upon increasing the reaction temperature from -30°C to 100 - 120°C the conversion of CO decreases. This decrease is very substantial in all catalysts. Upon further increase of the temperature, the conversion of CO increases and almost full conversion is reached around 250°C. A similar effect, i.e. a decrease in activity on increasing the temperature has recently been observed on Au/Mg(OH), catalyst under dry condition^.'^^ The authors attributed this behavior to 'negative activation energy'. It should be emphasized

54

Catalysis 1.o

0.8

--5t? 0.6 > 0

0.4 0.2 0.0

-50

0

50

100

150

200

250

300

Temperature, O C

Figure 31

Oxidation of carbon monoxide on digerent unmodijied AuIMgO catalysts using TPO techniques. Catalysts: 0 - A u / M g P s p (fresh, 0.150 9); A Au/MgOesusp(used, 0.150 g), x - Au/MgOresuSp, (fresh, 0.075 9); 0 Au/MgObase(fresh, 0.150 9); + - Au/MgOb"' (used, 0.150 9); 0Au/MgObase (fresh, 0.075 g). Catalyst pretreatment: calcination in air at 350 C for 1.5 h followed by reduction in hydrogen at 350 "Cfor 1 h (Reproduced from ref. 174 with permission) O

c

1.0

-

0.8

--

'E B

s

0

-50

0

50

100

150

Temperature, O C

Figure32

200

.-

250

300

The inJluence of the amount of ascorbic acid on the activity of Au/MgOmod catalysts. Results of TPO experiments. Catalysts: Au/MgOreSUSP, 0O.7-Au/MgOmod,0 1 . ~ - A u / M ~ O ", A" ~2.3-Au/MgOmod,x - 3.4-Au/MgOmd, + - 5.3-Au/MgOmod(the corresponding numbers indicate the amount of ascorbic acid in mg used to modify 1 g Au/Mg(OH), catalyst). Amount of catalyst: 0.075 g. Catalysts pretreated in hydrogen at 350 "Cfor 1 h (Reproduced from ref. 174 with permission)

that no other catalyst systems, such as Au/Ti02, Au/Fe203or Au/ZrO2 showed similar behavior. With respect to the activity decrease observed upon increasing the temperature, the deactivation of Au/MgO catalysts by surface species formed in situ, such

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

55

as carbonates, has been suggested. This has been confirmed by the formation of different carbonate IR bands observed under condition of CO chemisorption. The catalyst poisoning hypothesis has been further supported by TOS results obtained at -50, -30 and 0°C on different Au/MgO catalysts. In two hours on stream the modified catalysts showed no deactivation, while both the re-suspended and the base catalysts on he lowest deactivated relatively quickly. The strong deactivation of pure Au/MgO catalysts during TOS has recently also been reported by other a ~ t h o r ~ . ~ ~ ~ * ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Summing up the results of catalytic experiments the following conclusions can be drawn:

(0

the parent activity of Au/MgO catalysts increases in the following order: base <
2.3 Supported Sn-Ru Catalysts. - 2.3.1 Literature Background. In the last decade supported tin-ruthenium catalysts have been intensively studied because of their activity and high selectivity in the hydrogenation of unsaturated aldehydes, esters or acids into the corresponding General conclusion from studies devoted to Sn-Ru catalyst systems is that ionic tin species, i.e., Lewis acid sites, activates carbonyl compounds in hydrogenation reactions by polarizing the C=O bond.lgOThe mechanism of the hydrogenation of fatty acid esters into alcohol via the formation of aldehyde type intermediate is shown in Scheme 5.12

It is interesting to note that the ruthenium-tin boride system exhibited unique properties in the hydrogenation of cinnamaldehyde to unsaturated alcohol.1s1In addition, very good results have been obtained with tin promoted ruthenium catalysts in the hydrogenation of different unsaturated aldehydes.lg2 Mendes et allg3have studied the performance of different supported ruthe-

56

Catalysis

- cH30$4 *, -T -

#H

H2

CH3OH

f, RCHO

Ru---S nOx

y32R

Y Ru--%Ox Scheme5

?

H2

RCHzOH

Ru---SnOx

Mechanism of the hydrogenation of methyl oleate into oleyl alcohol via the formation of aldehyde over alumina supported Sn-Ru catalyst (Reproduced from ref. 72 with permission)

nium-tin catalysts for the liquid phase hydrogenation of oleic acid to unsaturated alcohols. Titania supported ruthenium-tin catalyst prepared by impregnation showed better performance than the alumina supported sol-gel ruthenium-tin catalyst. However, the catalyst supported on alumina is more active for the hydrogenation of C=C bond, whereas the titania supported catalyst is more active for the hydrogenation of the carboxyl group. The introduction of tin to Ru/Ti02 catalyst resulted in almost total suppression of the C=C bond hydrogenation. A possible mechanism for the hydrogenation of carboxyl bond over titania supported ruthenium catalyst with the involvement of Metal Support Interaction (MSI) is seen in Scheme 6.'83The reaction proceeds uia the formation of alkoxide surface intermediates. It has been found that titanium-oxide, when used as a support, greatly improves the activity of platinum group metals in the carbonyl-group hydrogenation in general, and the selectivity to unsaturated alcohols, in particular. The beneficial effect of Ti02 was attributed to Ti02-,patches on the metal, and Ti"+ ions that promote the hydrogenation into the desired d i r e ~ t i o n . ~ J ~ ~ The enhanced activity of transition metal catalysts supported on reducible oxides in the selective hydrogenation of the C=O bond is generally attributed to

Scheme6

A mechanism for the hydrogenation of the carboxyl group over Ru/Ti02 catalyst (Reproduced from ref. 184 with permission)

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

57

an interaction of the oxygen end of the C=O bond with exposed cations of the oxide supports. In the case of titania these sites can be Ti3+cation or 0 vacancies. These cations or 0 vacancies are created via spill-over hydrogen activated by the metal during high temperature reduction. The improved performance of metal catalysts with the addition of promoters such as tin is attributed to the presence of promoter-cations on the surface of metal, which activate the C=O bond through the interaction with the lone electron pair of the carbonyl group oxygen.183 It is known that an oxide support can strongly interact with another supported oxide, affecting its reducibility, and that the stability of supported oxides is determined by the similarity of the electronegativities of the two cations.'*' It is interesting to note, in this context, that in activated Sn-Pt/Si02and Sn-Ru/Si02 catalysts tin is mainly present as metallic Sno and as metal-tin alloys, while in activated Sn-Pt/A1203and Sn-Ru/A1203tin is otherwise stabilized as Sn2+.lS6 It is also noteworthy that, contrary to Sn-Pt system, no alloy phases exist for Sn-Ru bimetallic system. Nevertheless, the existence of two Ru-Sn intermetallic phases, Ruo.~S&.~ and R U O . ~ Shas ~ ~ .been ~, It is also worthwhile mentioning that Barbier et al. have prepared and characterized rutheniun-tin mixed oxide catalysts used in electro-oxidation of phenol. XRD of the calcined catalysts revealed the presence of Ru02, Sn02 and some trace amount of metallic ruthenium. However, there was no evidence of the formation of solid solutions of oxides.'8s Besides oxide supported Sn-Ru catalysts, carbon supported catalysts also find application in hydrogenation reaction^.'^^ '19Sn Mossbauer spectroscopy was used to investigate the tin component of ruthenium and tin supported on activated carbon catalysts containing 2 wt. % ruthenium and having Sn/(Sn+Ru) ratios between zero and 0.4. Four major components in the l19Sn Mossbauer spectra were attributed to both Sn(I1) and Sn(IV) oxides and to Ru-SnO, species formed on the surface of ruthenium metal particles. In addition to this '19Sn spectra reveal the presence of minor amounts of Ru3Sn7alloy phase.lg0 In recent years, besides other bimetallic systems (PdRe, ReSn, RhSn, CoSn) Ru-Sn/A1203catalysts have gained much attention in studying the hydrogenolysis of fatty esters to fatty It has been proposed that in the hydrogenolysis of fatty esters the active centers are metallic Ru particles in interaction with tin oxide acting as Lewis acid centers involved in the activation of the carbonyl group.180 Barrault and co-workers studied the selective hydrogenation of methyl oleate into oleyl alcohol over RuSnB/alumina catalysts.72The yield of oleyl alcohol was 75 % at 90 % methyl oleate conversion over a RuSnB/A1203catalyst for a bulk atomic ratio Sn/Ru of 4. Over such catalysts the reaction involves three steps: (1) the hydrogenation of the methyl oleate into the oleyl alcohol, (2) the transesterification between the methyl oleate and the oleyl alcohol with the formation of the heavy oleyl oleate ester, (3) the hydrogenolysis of this heavy ester into oleyl alcohol. The first and the third steps could involve mixed ruthenium-tin sites with two SnO, (x = 2) species, while the second could require tin species without

58

Catalysis

an interaction with ruthenium. It has also been found that the oleyl oleate formation decreases the alcohol yield at least at the beginning of the reaction, and the hydrogenolysis of heavy ester, giving two moles of unsaturated alcohol, is the rate-limiting step. Methyl hexadecanoate, methyl-9-octadecanoate and dimethyl succinate was hydrogenated on a-and y-alumina, silica and titania supported ruthenium-tin boride catalysts. Methyl-9-octadecanoate gave the best yield of oleyl alcohol (ca. 50 %) over the y-alumina supported catalyst at Ru/Sn =0.5 atomic ratio, 270 "C, 44 bar and 7 hours reaction time. Among Sn, Ge and Pb, tin appeared to be the best promoter. The highest activity and selectivity was attributed to Ruo sites interacting with Sn2+or Sn4+Lewis acid sites uia oxygen wherein the Lewis acid preferentially activates the C=O group of ester, facilitating hydrogen transfer from adjacent Ru-H sites. The role of boron probably lies in increasing the electronic charge density around Ru, thereby facilitating activation of H2 as a hydride.'*' Ruthenium on alumina, modified by tin compounds, has been also reported as good catalyst for selective reduction of oleic acid to the corresponding unsaturated alcohol. Study on the catalyst preparation indicated that when Ru(acach was loaded on alumina first and then tin tetra-butoxide added, the catalyst showed a better selectivity, but lower activity than that with catalysts prepared with the reversed order of impregnation. Obviously, the selectivity is highest when tin covers the Ru surface, and most of the tin remains in the unreduced state.Ig1The optimum atomic ratio of ruthenium to tin is about 1:2. Catalyst prepared by an improved sol-gel method showed higher activity and selectivity than catalysts prepared by impregnation and co-precipitation methods. Chloride was found to have a negative effect on catalytic activity. The best catalyst is prepared from chloride-free Ru and Sn raw materials. Under optimum reaction condition (250 "C and 5.6 MPa hydrogen pressure) the selectivity for 9-octadecen-1-01 and total alcohol (9-octadecen-l-o1+ stearyl alcohol) formation was 80.9 % and 97 YO,respectively, at conversion of 81.3 %.lg2 Despite the extended studies on supported Sn-Ru catalysts used in different selective hydrogenation reactions and hydrogenolysis of esters there is a need to further investigate the effect of catalyst preparation and pretreatment parameters on the performance of this bimetallic system. 2.3.2 Hydrogenolysis of Ethyl Dodecanoate on Sn-Ru/A120j Catalysts. 2.3.2.1 Catalyst Preparation. In a recent study the authors of this review have investigated the hydrogenolysis of ethyl dodecanoate to dodecanol and ethanol on different Sn-Ru/A1203catalysts. Systematic studies have been done to investigate the influence of precursor compounds, sequence of impregnation, metal loading, Sn:Ru atomic ratio, catalyst pretreatment (calcination, reduction) and reaction conditions (temperature, H2 pressure). The oxide form of catalysts was characterized by XRD and Temperature Programmed Reduction (TPR) technique~.~*.'~~ Alumina (CK-300 Ketjen y-A1203,S = 180 m2/g)supported Sn-Ru catalysts were prepared by co-impregnation or successive impregnation. Various metal precursor salts were used: RuC13xHz0, R~(acac)~, SnC12x2H20,C12H2404Sn

59

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

(dibutyltin diacetate). Catalysts with Ru- and Sn-loading ranging from 1.0 to 6.9 and 0 to 17.5 wt. %, respectively, and Sn:Ru atomic ratio ranging from 0 to 3.5 were prepared by co-impregnation. The catalyst precursors were calcined (T,) in air and then reduced (T,) in flowing hydrogen (60 ml/min) at 400 "C for 4 hours unless otherwise mentioned. The catalysts prepared by co-impregnation were designated as xM1yM2,where x and y were weight % of the corresponding metal M1 and M2. Catalysts prepared by successive impregnation were designated as e.g. xM2/yMI, in this case M1 was first impregnated onto A1203 followed by the impregnation of M2to the previously calcined monometallic sample. In Calcined catalysts were characterized by XRD and TPR the TPR experiments 0.08 g catalyst was heated up to 700°C at a rate of 10"C/min in flowing 5% H2-Ar (40 ml/min). The hydrogenolysis of ethyl dodecanoate was carried out in SS autoclave (volume = 100ml).The amount of catalyst and ester was 0.5 g (2.8 wt. %) and 20 ml(75 mmol), respectively.

2.3.2.2 X R D Results. XRD spectra obtained on calcined catalysts prepared by co- and consecutive impregnation are shown in Figure 33.'94The line broadening of the Ru02 phase was the highest for calcined catalyst prepared by co-impregnation, whereas the narrowest line was observed on catalyst prepared by successive impregnation introducing Ru first. The average crystallite size of Ru02 increased in the order: 12, 15 and 17 nm for catalysts Sn-Ru, Ru/Sn and Sn/Ru, respectively. These data suggests that co-impregnation provides the most intimate contact between crystalline Ru02and highly dispersed tin-oxide species on the alumina The effect of Sn:Ru atomic ratio on the XRD spectra of alumina supported bimetallic catalysts is shown in Figure 34.194As seen the higher the Sn:Ru ratio

3562

Figure 33

2 979

2564

2254

2015

P'lheta 1-1

1 824 I d [A]

X-ray diflractograms of oxide form of 5 %Rull .6%Sn/y-A1203catalysts. Eflect of preparation method. Sn:Ru atomic ratio: 2, Calcination: at 400 "C,Curve Method: 1- co-impregnation, 2 - and 3 - consecutive impregnation, Sn and Ru introducedjrst, respectively. RuOx - ruthenium(I V)-oxide, GamA - y-A1203 (Reproduced from ref. 194 with permission)

60

Figure 34

Catalysis

X-ray diflractograms of oxide form of 5%Ru11 .6%Sn/y-A1203catalysts. Eflect of SnlRu atomic ratio. Preparation: co-impregnation, Calcination: at 400 "C, (except sample 5 , T, = 500 " C )Curve - Sn:Ru: 1- 0 , 2 - 1,3 - 2,4 - 2.5,5 - 3.5 (Reproduced from ref. 194 with permission)

the broader the line of Ru02phase. The average crystallite size of Ru02 is 14,14, 12 and 9 nm for Samples 1-4 with Sn:Ru ratio of 0, 1,2, and 2.5, respectively. The only exception is Sample 5 with Sn:Ru ratio of 3.5 (average RuO2 crystallite size 17 nm) due to the higher calcination temperature, 500 instead of 400°C. XRD results, as expected, indicated that the higher the temperature of calcination the higher the crystallinity of Ru02. (X-ray diffractograms are not shown). These results suggest that the higher the Sn:Ru ratio the more intimate the contact between Ru02crystallites and tin-oxide species covering alumina support. Upon increasing the temperature of calcination above 400 "C the interaction between Sn and Ru phases is weakened. Formation of solid solution of tin- and ruthenium-oxide, i.e. incorporation of Sn atoms (ca. 6 YO)into Ru02 crystallite has been found for the first time in alumina supported bimetallic ~ata1ysts.l~~ As seen in Figure 35 the characteristic lines of Ru02in the bimetallic catalyst shifted to higher d(A) values compared to pure highly crystalline R ~ 0 2 .This l ~ ~finding can only be explained by replacement of tin for ruthenium in the RuOz lattice. It is interesting to note that the shift of Ru02line depended on the mode of catalyst preparation. The largest shift was observed in 11.6%Sn-5YORu/A1203catalyst prepared by co-impregnation. In addition, the higher the Sn:Ru atomic ratio for the co-impregnated catalysts the larger the shift of the RuOz line. All these results indicate that in the calcined bimetallic catalysts there is an interaction between tin and ruthenium phases. 2.3.2.3 TPR Results. Results of TPR measurements are shown in Figure 36 and Figure 37.5' The effect of Sn:Ru atomic ratio on TPR spectra of calcined at 400 "C Sn-Ru/A1203catalysts prepared by co-impregnation is shown in Figure 36. As

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites im

,

61 1

-

2.979

Figure 35

X-ray difractograms of calcined 5 %Rul1 .6%Sn/y-A1203 catalyst. Replacement of Ru by Sn in the lattice of RuO,. Preparation: co-impregnation, Sn:Ru atomic ratio: 2, Calcination: 450 "C;all lines were referred to that of Si internal standard. Curve: 1 - catalyst, 2 - simulated spectrum of RuOz with 6 %replacement of Ru by Sn, 3 - simulated spectrum of pure Ru02. RuOx ruthenium(I V)-oxide, GamA - y-A1203 (Reproduced from ref. 194 with permission)

seen in Figure 36A the monometallic Ru/A1203catalyst was completely reduced between 100 and 220 "C.The small reduction peak at 120 "Ccan be attributed to non-decomposed RuC13, whereas the shoulder at ca. 150°C can be assigned to different highly dispersed ruthenium-oxygen species. The large narrow peak observed at 190 "Ccan be attributed to the reduction of bulk Ru02 crystallite~.'~~ Figure 36A shows also the TPR spectrum of bimetallic catalyst with a Sn:Ru atomic ratio of 2. The low temperature part of the spectrum (below ca. 250 "C) can be attributed to the reduction of Ru-containing surface species, whereas the high temperature part of the spectrum can be related to the reduction of different tin-containing species. TPR spectra of Sn-Ru/A1203catalysts with Sn:Ru atomic ratios 1,2 and 3 shown in Figure 36B indicate that the higher the Sn:Ru ratio the higher the temperature of the reduction of Ru-containing species. The broadening of peaks and the complexity of the low temperature part of the TPR curve of bimetallic catalysts indicate that the reduction of RuO2 is affected by tin. This suggests a close interaction between highly dispersed tin oxide species and Ru02 crystallites. As mentioned above the presence of highly dispersed tin oxide species and replacement of about 6 % tin by ruthenium in the lattice of Ru02was evidenced by X-ray diffra~tion.'~~ The broad overlapping peaks in the high temperature part of the spectra suggest that ionic tin species have different forms and environments. Furthermore, one can assume that metallic ruthenium formed at low temperature during TPR by activating hydrogen facilitates the reduction of tin species at higher temperature^.^^^'*^ The effect of preparation method on the TPR spectra of Sn-Ru/A1203catalysts is shown in Figure 37. The temperature of the most intense peak attributed to the

62

Catalysis

A

-

i

Figure 36

A

Efect of Sn:Ru atomic ratio on TPR of Ru-Sn/A1203catalysts. ( A ) - Sn:Ru atomic ratio: (+) - 0; (X) - 2. ( B ) - Sn:Ru atomic ratio: (A) -1; (X) - 2; (0)- 3. Preparation: coimpregnation, Ru loading: 5 wt. %, calcination temperature: 400 "C (Reproducedfrom ref. 51 with permission)

reduction of RuOz crystallites increased in the order: RuSn(co-impregnation) < Ru/Sn(Sn impregnated first) < Sn/Ru(Ru impregnated first). The highest reduction temperature of the Sn/Ru type catalyst can be explained by coverage of Ru-containing species with tin oxide. Shoulders or small intensity peaks for Sn/Ru type catalyst can be attributed to the reduction of pure Ru02 particles. The high temperature part of the TPR spectra, above 260°C, assigned to the reduction of various tin species is very similar for the three catalysts prepared by different methods. Detailed discussion of TPR results obtained on Sn-Ru/A1203 catalysts are given el~ewhere.~' 2.3.2.4 Catalytic Actiuity. Prior to discussing the catalytic activity results it is interesting to mention that the mechanism of carboxylic acid hydr~genation'~ or ester hydrogenolysis shows similarities. The hydrogenation of aldehydes can be considered as a part of the mechanism of ester or carboxylic acid hydrogenation. The main and side reactions taking place in the hydrogenolysis of esters can be as follows:

+ Ha RCO, + R'OH RCOa + Ha RCHO RCHO + 2 Ha * RCHZOH

RCOOR'

+

+

(13) (14) (15)

63

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

1

0 ~~

400 a00 400 600 ebb 700 T*mpermtrrr+, OC

Figure 37

Eflect of preparation method on T P R of Ru-Sn/A1203catalysts. (A) - consecutive impregnation, Ru first (SnlRu); (A) - consecutive impregnation, Sn first (RufSn); (X) - coimpregnation (Ru + Sn). Ru loading: 5 wt. %, Sn:Ru atomic ratio: 2, calcination temperature: 400 ”C (Reproducedfrom ref. 5 1 with permission)

2RCHO RCOOCHZR RCOOR’ + RCH20H RCOOCHZR

(16)

+

+

+ RCH20H 2 RCH2OH RCH20CHZR + H20 2 R’OH R’OR’ + H20 RCH20H + R’OH RCH20R’ + H2O RCOa + Ha+RH + CO RCH20H + 2 Ha RCH3 + H20

RCOOCHZR

+ Ha

+ R‘OH

+

RCO,

+

-+

+

(17) (18) (19) (20) (21) (22)

(23) R and R‘ = alkyl, aralkyl or phenyl group; a = adsorbed species; the other compounds indicated without a subscription as stable compounds can be identified in the reaction mixture. In the first step of the reaction the ester C - 0 bond is hydrogenolysed, and an R-CO type surface species is formed (reaction 13). It is generally accepted that in the hydrogenation of carbonyl compounds the attack of the carbon atom by hydrogen atom or hydride ion is the rate-limiting step.196The R-CO type surface species is then hydrogenated to the highly reactive aldehyde (reaction 14), which is further hydrogenated to the desired alcohol (reaction 15). The so-called ‘heavy ester’ (HE), containing two longer alkyl chain, can probably be formed by coupling of two aldehyde molecule (reaction 16) or by transesterification (reaction 17). The hydrogenolysis of the HE also results in the formation of alcohol (reaction 18). The dehydration of the alcohols, as side reaction at elevated temperatures, leads to the formation of symmetric and mixed ethers (reactions 19 and 21). The decarbonylation of surface R-CO species and the hydrogenolysis of --*

64

Catalysis

the alcoholic C - 0 bond results in the formation of hydrocarbon by-products (reactions 22 and 23, respectively). Results shown in Table 19 indicate that co-impregnated Sn-Ru/A1203catalyst (row 3) was the most active and selective in producing dodecanol (ROH) from ethyl d o d e c a n ~ a t e . ~Comparable ~~'~~ conversion, but a lower selectivity to dodecanol was observed over the catalyst prepared by successive impregnation of Sn followed by Ru (row 1). In contrast, the catalyst prepared by reverse sequence of impregnation was comparatively less active with an in-between selectivity of dodecanol (row 2). These results suggest that an intimate mixing of Ru with Sn is preferable for the creation of catalytic sites. Covering of tin by ruthenium resulted in less dodecanol selectivity, while covering of Ru by Sn resulted in lower conversion. These results suggest that co-impregnation provides mixed Ru-SnO, sites to the maximum extent. Results shown in Table 19 also indicate that the combination of RuCl3xH2O and SnCI2x2H20gave the highest ethyl dodecanoate conversion and dodecanol selectivity (row 3). The combination of R ~ ( a c a cand ) ~ SnCI2x2H20resulted in a comparable conversion but considerably lower selectivity of dodecanol (row 4). Much lower conversion and selectivity of dodecanol in-between the former two catalysts was obtained over the catalyst prepared by the combination of R ~ ( a c a cand ) ~ C12H2404Sn (row 5). It is interesting to note that the selectivity of hydrocarbon (HC) and heavy ester (dodecyl dodecanoate, HE) by-products was low. Recent studies have indicated that in the hydrogenolysis of dicarboxylic acid ester chlorine poisons the activity of only reduced Sn-Ru/A1203catalysts prepared from Ru(NO)(NO~)~ and SnCI2precursors. However, the calcination at 400°C or above removed much of the chlorine and resulted in high yield of alcohol.'97 Residual amounts of chlorine promoted the hydrogenation of the carbonyl group in the hydrogenation of a,P-unsaturated aldehyde over supported Pt and Ru Therefore, it can be speculated that small amount of residual chlorine could be beneficial for a better selectivity of dodecanol in the hydrogenation of ethyl dodecanoate.

Table 19 Eflect of impregnation sequence and precursor compounds of ruthenium and tin in Sn-Ru/Alz03catalysts used in the hydrogenolysis of ethyl dodecanoate (Reproduced from ref. 51 with permission) ROH

Selectivity, % HC HE

76 79 81 74 76

4 2 3 3 2

Conversion,

%

Catalyst

Yield,% ROH ~~~

5Ru/l 1.6Sna,d 93 1 1.6Sn/5Rua," 86 5Ru 1 1.6Sna3' 94 5Ru 1 1 .6Snb.' 94 5Ru1l.6Snc.' 84

1 5 1 3 1

71 68 76 70 64

Ru precursor: RuCl,xH,O; Sn precursor: SnC12x2H,0; Ru precursor: Ru(acac),; Sn precursor: SnC12x2H20; Ru precursor: Ru(acac),; Sn precursor: C,,H,,O,Sn (Di-butyl tin di-acetate); Sn:Ru atomic ratio = 2; consecutiveimpregnation,Sn introduced first; consecutive impregnation,Ru introduced first; co-impregnation.

a

'

65

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

Results obtained in studying the effect of Ru loading on the hydrogenolysis activity of Sn-Ru/A1203catalysts prepared by co-impregnation are given in Table 20.51J93 Upon increasing the Ru content from 1 to 2 wt. % both the ethyl dodecanoate conversion and the selectivity of dodecanol significantly increased. Further increase of Ru loading up to 5 wt. % increased slightly the conversion of ethyl dodecanoate, but the selectivity of ROH remained almost unchanged, resulting in the highest yield of ROH at 5 wt. YORu. Conversion as well as selectivity of dodecanol considerably decreased with Ru loading above 5 wt. %. Selectivity of heavy ester decreased sharply with the increase of Ru loading above 2wt. %. All these results indicate that in the alumina supported catalysts both the co-operation between Sn and Ru species in the catalyst, and the metal content have an optimum value. Activity data obtained on Sn-Ru/A1203catalysts with different Sn:Ru atomic ratios, keeping the Ru loading constant at 5 wt. %, are given in Table 21.51,193 The results indicate that over the Ru/A1203catalyst (Sn:Ru =0) the conversion of ethyl dodecanoate and the selectivity of dodecanol was very low. The main products over this catalyst were various hydrocarbons. Upon incorporating tin the conversion rapidly increased as the Sn:Ru atomic ratio increased from 0 to 1, at higher Sn:Ru atomic ratio the conversion decreased slowly. The selectivity to dodecanol increased significantly with Sn:Ru atomic ratio up to a value of 1.5, above this ratio the increase in the selectivity to dodecanol was only small. Thus, the yield of dodecanol reached a maximum value of 76% at a Sn:Ru atomic ratio of 2. The selectivity of hydrocarbons and the heavy ester was small over the catalysts having a Sn:Ru atomic ratio above 1.5. These activity data suggest that an optimum Sn:Ru ratio is required to assure the co-operation of Sn and Ru sites to the maximum extent. It is worthwhile mentioning that the temperature of calcination and reduction only slightly affected the catalytic performance under these conditions. The activity, i.e. dodecanol yield of the catalysts prepared and pre-treated in different ways (5 wt. YORu; Sn:Ru atomic ratio:2) was correlated with their TPR characteristics. As seen in Figure 38 good correlation was found between the yield of dodecanol on Sn-Ru/y-A1203catalysts and the amount of hydrogen

Table 20

Eflect of Ru loading in Sn-RulAlzOj catalysts used in the hydrogenolysis of ethyl dodecanoate (Reproduced from ref. 51 with permission) Conversion,

Catalyst

%

ROH

l.ORu2.3Sn 1.3Ru3.OSn 2.ORu4.7Sn 3.ORu7.OSn 5.ORu11.6Sn 6.9Ru16.3Sn

9 44 85 88 94 87

43 55 78 76 81 70

Selectivity, % HC

11 10 6 2 3 2

HE

Yield, % ROH

31 21 3 4 1 5

4 24 66 67 76 61

Ru precursor: RuCl,xH,O; Sn precursor:SnCl2x2H,O;Sn:Ru atomic ratio Catalysts were prepared by co-impregnation.

=

2 in all cases.

66

Catalysis

Table 21 Eflect of Sn:Ru atomic ratio in Sn-Ru/A1203catalysts used in the hydrogenolysis of ethyl dodecanoate (Reproduced from ref. 5 1 with permission) Catalyst

Sn:Ru Conversion, atomic ratio %

5RuO.OSn 5Ru2.9Sn 5Ru5.8Sn 5Ru8.7Sn 5Rul1.6Sn 5Ru14.6Sn 5Ru17.5Sn

0.0 0.5 1.0 1.5 2.0 2.5 3.0

18 70 95 92 94 91 88

ROH

4 15 60 79 81 80 83

Selectivity, % HC HE

67 56 10 3 3 3 3

14 11 6 4 1 4 4

Yield, % ROH

<1 11 57 73 76 73 73

Ru precursor:RuCl,xH,O; Sn precursor:SnC12x2H,0.Catalysts were prepared by co-impregnation.

consumed above 260°C for the reduction of tin oxide species in the TPR experiment?’ It is noteworthy, that the higher the amount of hydrogen consumed in the TPR experiment above 260 “C,the higher the degree of reduction of oxidized tin species in the catalysts. We assume that the higher degree of tin-oxide reduction in the bimetallic catalysts can be explained by a stronger interaction between ruthenium- and tin-containing species. Therefore, the higher the reducibility of tin oxide species in the catalysts, the more intimate contact between rutheniumand tin-containing species, and as a consequence, the higher the catalytic activity. On the basis of this finding it was suggested that calcination and reduction of the bimetallic catalysts prepared by impregnation results in the formation of Ru-SnO, mixed sites. It is suggested that co-impregnation provided intimate contact of ruthenium with tin resulting in the formation of active ‘Sn”+-Ru’, ‘metal ion-metal nanocluster’ ensemble sites to the maximum extent. 2.3.3 Summary on Sn-RulAl203 Catalysts. - In the last decade supported tinruthenium catalysts have been extensively studied because of their activity and high selectivity in the hydrogenation of unsaturated aldehydes, esters or acids into the corresponding alcohols. The increased performance of metal catalysts with the addition of promoters such as tin is attributed to the presence of promoter-cations on the surface or at the periphery of metal nanocluster, which activate the C=O bond through the interaction with the lone electron pair of the carbonyl group oxygen. In the hydrogenolysis of fatty esters tin appeared to be the best promoter for ruthenium. The highest activity and selectivity was attributed to Ruo sites interacting with Sn2+or Sn4+Lewis acid sites via oxygen wherein the Lewis acid preferentially activates the C=O group of ester, resulting in facile hydrogen transfer from adjacent Ru-H sites. In the hydrogenolysis of ethyl dodecanoate to dodecanol on different RuSn/A1203catalysts co-impregnation provided the most active catalyst. It was

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

Figure 38

67

Correlation between the yield of dodecanol over Ru-Sn/A1203catalysts and the amount of hydrogen consumed in T P R above 260 "C (Reproduced from ref. 51 with permission)

found for the first time by XRD that a solid solution of tin- and ruthenium-oxide was formed in the catalyst precursor, i.e. Sn atoms (ca. 6 %) incorporated into Ru02 crystallite in the alumina supported bimetallic catalyst. Both TPR and activity results suggest that co-impregnation resulted in the formation of mixed Ru-SnO, sites to the maximum extent. Correlation was found between the yield of dodecanol on Ru-Sn/y-A1203catalysts and the amount of hydrogen consumed in the high temperature (above 260 "C) reduction of tin oxide species in the TPR experiment. It was suggested that the higher the degree of tin-oxide reduction in the bimetallic catalysts the stronger the interaction between ruthenium- and tin-containing species. Therefore, the higher the reducibility of tin oxide species in the catalysts, the more intimate contact between ruthenium and tin species, and as a consequence, the higher the catalytic activity. It has been also found that co-impregnation provided intimate contact of ruthenium with tin resulting in the formation of active Snn+-Ru,'metal ion-metal metal nanocluster' ensembles to the maximum extent.

2.4 Re-Pt/Alz03Catalysts. - 2.4.1. Literature Background. The first bimetallic naphtha reforming catalyst, Pt-Re/A1203was introduced in the late sixties and offered improved activity and stability compared to its monometallic Pt/A1203 Although the Pt-Re/A1203catalysts have been in use for many years, there is still considerable debate about the location and the valence state of rhenium. From the available literature on the bimetallic catalysts it is clear that the observed state of the added metal is highly dependent on a number of factors including the starting materials employed, the metal loading, the calcination and reduction temperatures and the support used. Depending on these factors, rhenium promoter may be found: (i)in a reduced form alloyed with platinum, (ii) in an oxidized form stabilized by the support, or (iii) as a combination of both of the above.200 Besides Re-Pt/A1203 reforming catalysts other rhenium-containing systems

68

Catalysis

have been also investigated, however less extensively, for their activity in the hydrogenation of different organic carbonyl compounds. Forty years ago Broadbent and co-workers studied the hydrogenation of a great variety of organic carbonyl compounds (aldehydes, ketones, esters, anhydrides, acids, carboxamides, etc.) on different reduced bulk rhenium-oxides.2°1~20z203 Rhenium heptoxide reduced by hydrogen in a solvent appeared to be highly active for the hydrogenation of carboxylic acids and carboxamides. Upon using sodium borohydride or zinc in acid solution as reducing agents the rhenium oxides were not ordinarily reduced to lower oxides or the metal during the h y d r o g e n a t i ~ n . ~It~ .is~also ' ~ noteworthy that reduced rhenium oxide catalysts showed only moderate activity toward many of the common organic functions reducible by metallic platinum or nickel catalysts. All these early results suggest that in the catalysts used in Broadbent's studies part of rhenium was in ionic form, and thus it was able to activate the C=O bond of different substrates. Nevertheless only scare data is available in the recent literature on the application of Group VIII noble metal (M) or rhenium-based mono- and Re-M bimetallic catalysts, in the hydrogenolysis of esters or hydrogenation of acids to alcohols. Recently a few publication^^^^^^^^ and patent^.^^^^^^^,^^^,^^^ have been reported on the transformation of different carbonyl compounds (saturated and unsaturated esters, acids and carboxamides) over rhenium-containing catalysts. In the bimetallic catalysts used for the hydrogenation of carbonyl compounds the rhenium was combined with Pd;O8 or Rh.*12In the case of catalysts used for the hydrogenation of unsaturated carbonyl compounds the rhenium is usually modified with tin.213 In a recent work a rhenium promoted palladium model catalyst was used to study the mechanism of the hydrogenolysis of acetic acid to Non-local density functional theory calculation (DFT) was used to examine the alternative mechanisms for the hydrogenolysis reaction. The overall surface reaction energies, at low surface coverage, were computed for a number of different possible paths by which acetic acid may be converted to ethanol over Pd(ll1). In the postulated mechanism, acetic acid dissociates to form an acetyl surface intermediate. The acetyl intermediate is then subsequently hydrogenated to ethanol via the formation of an acetaldehyde surface intermediate. Experimental observations and DFT calculations suggest that these two steps are likely to be rate determining in acetic acid hydrogenolysis. The activation barriers and overall reaction energies these same steps were also computed on Re(0001) and pseudomorphic overlayers of Pd on Re (PdML/Re(OOO1))as well. The results suggest that the C-OH bond-dissociation reaction is more favored over Re(0001) since it has a more open d band. However, bond-association reactions are favored on PdML/Re(OOO1),which has an electronic d-band structure similar to that of a noble metal. The optimal balance may require a Pd/Re alloy. Calculations performed over a Pd0.66Re0.33alloy demonstrate a nominal barrier for both C-OH bond breaking and C-H bond formation. This may be ideal for acetic acid hydrogenolysis to ethanol. Rhenium ensembles, however, should be avoided as they lead to acetic acid decarboxylation. 2.4.2 Characterization. Among different catalyst characterization methods tem-

69

I: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

perature-programmed reduction (TPR) is a useful tool for investigating the interactions and oxidation degree of metals in bimetallic catalysts.215TPR of the of Pt/A1203and Re/A1203catalysts, as well as the bimetallic catalysts containing 0.5 wt. % platinum and rhenium and 1wt. YOC1 showed distinct differences. TPR curve of the alumina supported platinum catalyst showed a peak at 280 "C, while that of the rhenium catalyst appeared at 640 "C(see Figure 39)215. Temperature of the platinum reduction peaks for the bimetallic catalysts shifted to higher temperatures due to the presence of rhenium. However, the reduction peaks of rhenium shifted in the lower temperature direction, which was attributed to hydrogen spillover from platinum to rhenium. The stronger the interaction between platinum and rhenium, the larger are the shifts in peak temperatures? Holmen and co-workerssOhave studied the influence of pre-treatment temperature on the metal function of a commercial Pt-Re/A1203reforming catalyst (EUROPT-4, CK-433 from Akzo, Holland; containing 0.3 wt. YOPt, 0.3 wt. 'YO Re and 0.95 wt. YOC1) by X-ray absorption spectroscopy. Simultaneous examination of the rhenium LIIIand platinum LIII EXAFS data allowed them to distinguish between the bimetallic interaction and the metal-support interaction from the overall spectrum. The results showed that in the catalyst dried in air at temperatures < 500°C prior to the reduction at 480"C, bimetallic species are formed. Drying at higher temperatures and in the absence of air inhibits the transport of mobile rhenium species on the surface causing no intimate contact between the two metals. Platinum LIIrEXAFS data showed that the average

(u

I

200

Figure 39

300

400

500

600

700

TPR projles of alumina supported Pt, Re and Pt-Re catalysts. The amounts of both Pt and Re were 1 wt.-% in all catalysts and the amount of Cl was 1 wt.-%. (-) Ptj, (-x-x-) PtjRej, ( ) Ptj +Re2, (-- - - -)Pt,Re2, ( o o o o o ) ReJ The catalysts were prepared by: PtJ impregnation withH,PtC1,; PtJRl co-impregnation with H2PtC1, and HReO,; Pt, + Re2 sequential impregnation with H2PtC16 and Re2(CO),,; Pt5Re2 co-impregnation with [ P t 3 ( C 0 ) J 5 [ N ( C2H5),I2and Re2(CO) Rel impregnation with H R e 0 , (Reproducedfrom ref. 21 5 with permission)

70

Catalysis

particle size of the bimetallic particles on the alumina surface was less than 1 nm. The results from the rhenium LIIIEXAFS analysis confirmed that rhenium was not completely reduced to metallic rhenium after reduction, with a significant fraction of the rhenium present in low, positive oxidation states and intimate contact with the support. The EXAFS data were consistent with a structural model of rhenium metal particles in the particle size around 1-3 nm with smaller platinum particles located within or at the boundary of the rhenium particles. Moderate heating of the catalyst in presence of air (ie., moisture) provided the best condition for transport of mobile rhenium species on the surface, and hence alloy formation. The influence of the chlorine content on the formation of bimetallic particle in Pt-Re/A1203catalysts was also studied by STEM/EDX, TPR, H2chemisorption and model reaction (hydrogenolysis of n-butane and cyclopentane)?16Indirect methods such as model reactions, hydrogen chemisorption, and TPR studies on the catalyst surface during indicate that too much chlorine (0.65-1.50 wt. YO) reduction is undesirable with respect to bimetallic particle formation. Chlorine has an inhibiting effect on the mobility of Re in the reduction process. The effect of chlorine on the bimetallic formation is, however, limited as compared to the effect of water during reduction. Addition of water enhances the formation of alloyed particles and this might be due to increased rhenium oxide mobility and stripping of chlorine of the support. EDX/STEM revealed that the particles are somewhat larger and more alloyed when water is added to the reducing gas. In summary, several investigators have attempted to use TPR to elucidate the structure and the interactions of Re and Pt in supported catalysts. The interaction between the two metals has been found to be influenced by the Re loading, the ratio of the two metals, the chlorine content, the water partial pressure in the H2 used for reduction, and especially, the temperature of drying or calcination preceding reduction.217When an alumina supported Re-Pt sample prepared from salt precursors was dried at 200"C, TPR showed that the co-reduction of the two metals occurred at a temperature intermediate between that characteristic for the two monometallic counterparts. In contrast, when the catalyst was dried or pre-oxidized at 500 "C, two separate reduction peaks were observed (see Figure 40, taken from ref. 219).216*2171218,219 The authors of this review have characterized alumina supported monometallic Pt and Re and a bimetallic catalyst containing 0.5 or 1 wt. YORe and Pt by Temperature-Programmed Reduction (TPR).220In the TPR experiments 0.08 g catalyst was heated up to 800 "C at a rate of 10 "C/min in flowing 5% H2-Ar(40 ml/min). Figure 41 shows the TPR curves of alumina supported Pt, Re and Pt-Re catalysts?20In the TPR curve of Pt catalyst the low temperature peak at 230 "C can be attributed to the reduction of PtO2, whereas the peak at ca. 350 "C can be assigned to platinum species interacting more strongly with the s ~ p p o r t . ~ * ~ ~ ~ ~ * The maximum of the TPR curve of Re/A1203 catalyst was observed at ca. 450 "C which is similar to that obtained in the case of monometallic catalyst prepared by using ammonium per~henate.~~' The amount of consumed hydrogen in TPR indicated incomplete reduction of Re in the monometallic catalyst. In the case of

71

1: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

b Figure40

I

1

I

I

1

I

I

I

I

1

I

1

I

100200300400500600 Temperature ("GI

70

The efleect of drying temperature on the T P R profiles of alumina supported Pt-Re catalyst prepared by impregnation. The amounts of both Pt and Re were 0.3 wt.-% in all catalysts and the amount of Cl was 0.6 wt. % (Reproduced from ref. 219 with permission)

0

400

401)

600

800

Temperatun, OC Figure 41

T P R of alumina supported Pt, Re and Pt-Re catalysts. Catalysts pre-reduced by the producer were calcined in air at 400°C: ( X ) -0.5%Pt-0.5%Re/A1203; ( ) -0.5%Re/A1203;(0) -0.5%Pt/A1,03 (Reproduced from ref. 220 with permission)

72

Catalysis

bimetallic catalyst it can be concluded that Pt facilitated the reduction of Re and that Pt was near the rhenium.*17The shape of the TPR curve and the characteristic temperatures of the peaks indicated simultaneous reduction of platinum and rhenium at relatively low temperature (- 300 "C) and thus the formation of bimetallic species.80The high temperature peak at ca. 450°C indicates the presence of separate oxidized rhenium species in the catalyst. Figure 42 shows the influence of pre-treatment on the TPR of Re-Pt catalyst containing 1 wt. % of both It is interesting to note that drying of the pre-calcined catalyst at 250°C in argon resulted in a single symmetric peak at about 280°C. This observation is in good agreement with previous results.80It was found that on the hydrated catalyst the reduction of rhenium is catalysed by Pt, resulting in simultaneous reduction of platinum and rhenium at low temperature (- 300 oC).80 As seen in Figure 42 the shape of the TPR curves dried in argon at 250 "Cor oxidized in air at 400 or 500 "Cis completely different. The oxidation at higher temperatures resulted in more complicated TPR spectra with broad overlapping bands. This can be an indication of the segregation of Re and Pt oxide phases in the catalyst during oxidation treatment. The high temperature part of the spectra may indicate the presence of separate oxidized Re species in the catalyst. Figure 43 shows that the higher the temperature of oxidation the higher the amount of the hydrogen consumed in high temperature part of TPR for the reduction of separate rhenium oxide species.220The low temperature peak between 200 and 300 "Cindicates the presence of the oxide precursor of bimetallic nanoclusters that can be formed during the reduction of bimetallic catalysts. All

?

6

d

J

. l i

t

Figure 42

Efect of pre-treatments on T P R of Pt-Re/Al,O, catalysts (RePt(1 .O)). Samples pre-calcined by the producer were used. Treatments prior to TPR: (0) - in Ar at 250 "C;( ) - in air 400°C; ( X ) - in air 500 "C (Reproduced from ref. 220 with permission)

+

73

I : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

1m5

4

(I

1.Q

&

at

2

-t

04

Om0 0 Figure 43

PO0 400 (100 Temperatun, O C

800

Eflect of calcination temperature on T P R of Pt-Re/Al,O, catalyst. RePt(1 .O) sample pre-reduced by the producer was calcined in air prior to TPR: (A)- at 400°C; ( X ) - at 500°C (Reproducedfrom ref. 220 with permission)

these results indicate that in the alumina supported Re-Pt catalysts pre-reduced by the producer and then oxidized prior to TPR at 400°C or higher temperatures, in addition to PtO, and the precursor of Re-Pt bimetallic species separate rhenium-oxide species are also present. The lower the temperature of drying or calcination the more intimate the contact between Re and Pt, the higher the reducibility of rhenium, the smaller the amount of consumed hydrogen in the high temperature part of TPR spectra. The degree of reduction of platinum and rhenium oxide phases determined from hydrogen uptake during TPR for the mono- and bimetallic catalysts are given in Table 22.220PtO2 and Re207phases were supposed to be present in the catalyst samples treated in argon or oxidized at 400 or 500 "Cprior to TPR. It is interesting to note that in TPR measurement carried out up to 800 "Cplatinumoxide can be fully reduced, whereas only 86 % of the rhenium-heptoxide is reduced in the monometallic catalyst. In the RePt( l.O)(e, red.) catalyst oxidized at 400"C, 96 % of the rhenium oxide phase reduced in TPR. Despite the more pronounced segregation of Pt and Re oxide phases in the RePt(l.O)(e, red.) catalyst oxidized at 500°C (see Figure 43 ) the reduction of both phases was complete in TPR. RePt( l.O)(e, calc.) catalyst oxidized at 400 and 500 "C was completely reduced in TPR. However, if the latter sample was dried in argon at 250 "Conly 79 % of the rhenium oxide phase was reduced. The most interesting result was that RePt(l.O)(e,red.) catalyst oxidized and then reduced at 400 "Ccan be further reduced in the TPR experiment at higher temperatures. The hydrogen uptake indicated that in this catalyst after reduction at 400 "Cfor 2 hours half of the rhenium was in the form of Re4+. This may suggest that the treatment

74

Table 22

Catalysis

Reduction degree of Pt and Re in the diflerent Re-Pt/A1203catalysts used in the hydrogenolysis of butyl acetate. (Reproduced from ref. 220 with permission)

Catalyst/Treatment

PtO, reduction degree, %

Pt(O.5)(b)/0400 Re(OS)(b)/400 RePt(l.O)(e,red.)/0400 RePt(l.O)(e,red.)/0500 RePt(l.O)(e,red.)/0400H400 RePt(l.O)(e,calc.)/Ar250 RePt(1.O)(e,calc.)/0400 RePt(l.O)(e,calc.)/0500

100

-

100 100 100 100 100 100

Re207reduction degree, %

86 96 100 50 (calculated for Re02) 79 100 100

Catalyst:0.5 wt. %Pt/Al,O, (Pt(0.5)),0.5 wt. YORe/Al,O, (Re(0.5))and Re-Pt/Al,O, catalysts (RePt(l.0))with metal loading 1 wt. % for both Re and Pt were customer made in the laboratory of Engelhard Co. (b = beads, e = extrudate,red. = pre-reduced, calc. = pre-calcined by the producer). Treatments in Ar, oxygen and hydrogen were used at 250,400 and 500 "C,respectively.

(oxidation in O2 and then reduction in H2)used prior to the hydrogenolysis of butyl acetate results in a catalyst still containing ionic rhenium species.220 Based on TPR results it can be concluded that intimate contact between rhenium and platinum is provided in bimetallic alloy particles on the surface of alumina supported catalysts. Therefore, one can expect that the oxidation of the catalyst followed by reduction at moderate temperature result in the formation of platinum and/or Re-Pt metallic nanoclusters and rhenium ions in atomic closeness. 2.4.3 Hydrogenolysis of Butylacetate Over Re-Pt/A1203Catalysts. In one of the authors' recent works220butyl acetate (BuOAc) was used as model compound to study ester hydrogenolysis over alumina supported Pt, Re-Pt, Sn-Pt, Re-Sn-Pt and Ru-Re-Pt catalysts. The aim of the work was to study the role and efficiency of tin and rhenium promoters in the activation of the carbonyl group of butyl acetate. It is, however, worth for mentioning that there is no data in the literature on the use of reforming type bimetallic Sn-Pt, Re-Pt or trimetallic Re-Sn-Pt catalysts for the hydrogenolysis esters to alcohols. It is known that in these types of reforming catalysts at least part of the tin221or rhenium is in ionic form.80>220 Therefore, one might expect that ionic tin or rhenium species as Lewis acid sites in the reforming type catalyst are also able to activate C=O bond in carbonyl compounds, and platinum provides hydrogen for the reaction. In previous studies, in addition to metallic rhenium, ionic rhenium species were also present in the catalysts active in the hydrogenation of carbonyl bond of different substrates.2049206 Therefore, atomic closeness of metallic and ionic species might be required to obtain bimetallic Re-Pt based catalysts for the hydrogenation of esters such as butyl acetate. The formation of 'Re"+-Pt'' ensembles may be achieved by a calcination/reduction treatment of the catalyst carried out at moderate temperatures. Such a treatment may provide high catalytic activity.

75

I: Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

Therefore, in the catalytic experiment prior to activity tests the alumina supported Re-Pt catalysts were calcined in oxygen at 400 "C and reduced at the same temperature. After such a treatment the catalyst was not completely reduced since further hydrogen consumption was observed above 400 "C in TPR.220As already shown in Table 22 this is an indication on the presence of ionic rhenium species in the catalysts reduced at 400 "Cfor 2 hours. Liquid phase hydrogenolysis of butyl acetate was carried out in a stirred (rate = 600 rpm) stainless steel autoclave. Details on the catalysts, treatments and reaction conditions are given in the corresponding Tables. The conversion, selectivity and yield as well as first order rate constant obtained in the hydrogenolysis of butylacetate on Pt, RePt and RuRePt catalysts are summarized in Table 23.220 Reaction rates listed in Table 23 indicate that, due to the absence of ionic promoter species the activity of Pt/A1203catalyst is very low. The addition of rhenium to the Pt/A1203 catalyst increased the hydrogenolysis activity at least by a factor of 20. The activity of the two different types of Re-Pt catalysts is in the same order of magnitude. Ru is often used as a component of catalysts applied for the hydrogenolysis of ester~.~~Therefore, the addition of Ru to Re-Pt catalysts further improved the hydrogenolysis activity and the selectivity to butanol. It is noteworthy that the selectivity of RePt and RuRePt catalysts is significantly higher than that of monometallic Pt catalyst. The activity data of SnPt and ReSnPt catalysts obtained in the hydrogenolysis of butyl acetate are given in Table 24.220The comparison of the data given in the first row of Table 23 and Table 24 indicate that Sn-Pt/A1203catalyst is significantly more active than monometallic Pt catalyst. Upon increasing the amount of Re added to the Sn-Pt/A1203catalyst the activity of the trimetallic catalysts monotonically increased (except for the 1.OOReSnPt catalyst). The selectivity to butanol showed up a maximum in the range of Re content of 0.5-0.75 wt. %. The results of activity tests clearly indicated that both Sn and Re significantly

Table 23

Catalyst

Hydrogenolysis of butyl acetate over alumina supported Pt, RePt and RuRePt catalysts (Reproduced from ref. 220 with permission) Conversion at 24 h

Pt(O.S)(b) 4.9 RePt(1 .O)(b) 68.4 RePt(l.O)(e) 58.3 0.3RuRePt(l.O)(e) 66.8

Selectivity BuOH, %

Yield BuOH, %

k h-'

91.8 97.6 93.1 96.8

4.5 66.3 54.3 64.7

0.0021 0.0480 0.0364 0.0460

Catalyst: 0.5 wt. %Pt/Al,O, (Pt(0.5)) and Re-Pt/Al,O, catalysts (RePt(l.0))with metal loading 1 wt. YOfor both Re and Pt were customer made in the laboratory of Engelhard Co. (b = beads, e = extrudate). Cl content was 0.8 wt. YO. 0.3 wt. % Ru was introduced by incipient wetness impregnation using RuCl, xH,O. Prior to activity test the 2 or 4 g catalyst was pre-treated in oxygen, nitrogen and hydrogen at 400 "Cfor 1,0.5 and 2 hours, respectively. = 6 MPa (at room temperature); Amount of BuOAc: 50ml(0.38 M). Reaction conditions: Phydrogen reaction temperature,T = 235 "C;reaction time, t = 24 hours; k = first order rate constant.

76

Table 24

Catalysis

Hydrogenolysis of butyl acetate on rhenium modified Sn-Pt/A1203 catalysts (Reproduced from ref. 220 with permission)

Catalyst

Conversion at 24 h

Selectivity BuOH, %

Yield BuOH, %

k h-I

SnPt 0.lOReSnPt 0.25 ReSnPt 0.5OReSnPt 0.75ReSnPt 1.00ReSnPt 1.5OReSnPt

40.5 54.5 62.2 73.2 90.7 86.0 93.6

88.8 90.0 92.1 93.9 92.3 91.5 91.4

36.2 49.0 57.3 68.7 83.8 78.7 85.6

0.0216 0.0328 0.0405 0.0548 0.0991 0.0820 0.1116

~

~~

Catalyst: Sn-Pt/Al,O, commercial reforming (SnPt), 0.3 wt. YOSn, 0.3 wt. YOPt and 1 wt. YOC1. Re content ranging from 0.1 to 1.5 wt. % was introduced by incipient wetness impregnation using NH,ReO,. Prior to activity test 4 g catalyst was pre-treated in oxygen, nitrogen and hydrogen at 400 "Cfor 1,0.5 and 2 hours, respectively. Amount of BuOAc: 50ml(0.38 M). Reaction conditions: Phydrogen = 6 MPa (at room temperature); reaction temperature, T = 235 "C;reaction time, t = 24 hours; k = first order rate constant.

improved the activity of platinum in the hydrogenolysis of butyl acetate. Therefore the activity data suggest that both Sn and Re was strongly involved in the activation of butyl acetate namely its carbonyl group. It is known that the activation of hydrogen in this reforming type alumina supported Sn-Pt and Re-Pt catalysts can be attributed to platinum or bimetallic alloy species. Furthermore literature data indicate that in reforming type Sn-Pt46and Re-Pt'O catalysts part of the promoter is in ionic state. In addition, our TPR studies (see Table 22) confirmed that in Re-Pt catalyst oxidized and then reduced at 400 "C half of the rhenium was in the form of Re4+.Therefore it is suggested that atomic closeness of positively charged Re or Sn species and platinum or bimetallic nanocluster, i.e. the formation of 'metal ion-metal nanocluster' ensembles is required to achieve high catalytic activity. 2.4.4 Summary on Re-Pt/A1203Catalysts. Besides Re-Pt/A1203reforming catalysts other rhenium-containing systems have been also investigated, however less extensively, for their activity in the hydrogenation of different organic carbonyl compounds. Early results obtained on bulk rhenium-oxides reduced in different ways suggested that in those catalysts part of rhenium was in ionic form, and thus it was able to activate the C=O bond of different substrates. Several investigators have attempted to use TPR to elucidate the structure and the interactions of Re and Pt in supported catalysts. Based on TPR results it was found that Re-Pt catalyst oxidized and then reduced at 400 "Cconsumed additional amount of hydrogen in the TPR experiment at higher temperatures. The hydrogen uptake indicated that in this catalyst half of the rhenium was in the form of Re4+.This may suggest that this treatment resulted in the formation of platinum and/or Re-Pt bimetallic nanoclusters and rhenium ions at atomic closeness, i.e. 'metal ion-metal nanocluster' ensembles sites. The introduction of Re into the Pt/A1203catalyst increased the hydrogenolysis

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

77

activity more than an order of magnitude. Sn-Pt/A1203catalyst is also significantly more active than monometallic Pt catalyst. Upon increasing the amount of Re added to the Sn-Pt/A1203both the activity and the selectivity of the catalyst to butanol very significantly improved. Both Sn and Re in its ionic form can be involved in the activation of C=O group of butyl acetate.

2.5 Copper-ContainingCatalysts. - 2.5.1 Copper Chromite Catalysts. Hydrogenation of Dodecanoic Acid. 2.5.1 .I Literature Background. Copper chromite based catalysts are frequently used for the hydrogenation of both saturated and unsaturated fatty acids (e.g oleic acid) to the corresponding alcohol. Besides copper chromites metallic (or colloidal) copper containing ionic compound as additives, showed excellent catalytic p r ~ p e r t i e s . ~ ~ ~ ? ~ ~ ~ Conventional technology of the hydrogenolysis of fatty acid methyl esters to the corresponding fatty alcohols uses copper chromite or zinc chromite based ~ a t a l y ~andt the ~ manufacturing ~ ~ ~ ~ ~ process ~ ~ ~requires ~ ~ high ~ * pressures ~ ~ ~ (200300 bar) and temperatures (250-300“C).The activity of copper chromite catalysts was significantly increased by the addition of z i n ~ . ~ ~ ~ i ~ ~ ~ Krause investigated the mechanism of a characteristic conformational hydrogenation reaction on a zinc oxide/chromic oxide catalyst, resulting in unsaturated alcohols from unsaturated The addition of Cr2O3 improved the catalytic action of ZnO in the hydrogenation of unsaturated acids by blocking its electron donating sites. It was proposed that the resultant alcohols were H-bonded to the double bond protecting the double bond from hydrogenation and the alcoholic group from dehydration. Hubaut et al.231has studied the liquid phase hydrogenation of polyunsaturated hydrocarbons and carbonyl compounds over mixed copper-chromium oxides. The selectivity of monohydrogenation was almost 100 % for conjugated dienes but much lower for a,fbunsaturated carbonyls. This was due to the adsorption competition between the unsaturated carbonyls and alcohols as primary products. It was suggested that the hydrogenation site was an octahedrally coordinated Cu+ ion with two anionic vacancies, and an attached hydride ion. The Cr3+ion in the same environment was probably the active site for side reactions (hydrodehydroxylation, nucleophilic substitution, bimolecular elimination). All these results suggest that in copper chromite type catalysts ionic copper species can be the active site for the hydrogenation of carbonyl compounds. 2.5.1.2 Hydrogenation of Dodecanoic Acid to Dodecanol. The authors of this review investigated three commercial Cu0-Zn0-A1203catalysts (CZA-1, CZA-2 and CZA-3) as well as a commercial Cu-Cr/MgO-Si02catalyst (Cu60) for their composition, bulk structure and A laboratory made Cu0-Zn0-A1203 catalyst (CZA-4)was also prepared by using a co-precipitation method. Prior to the activity test calcined samples were in situ or ex situ reduced in hydrogen for 2 hours under conditions given in the corresponding Tables. Chemical compositions of commercial and the laboratory made catalysts determined by ICP are given in Table 25.2327244

78

Catalysis

The average crystallite size of CuO and ZnO in the oxide form of catalysts, determined by XRD, are listed in Table 26.232The average crystallite sizes and estimated values in % of the distribution of copper among different species, i.e. metallic Cu, C u 2 0and of CuO in the used catalysts are also given in Table 27.232 It is noteworthy that C u 2 0and traces of CuO can only be detected in the used CZA-1 and Cu60 catalysts reduced in situ at lower temperatures (205 or 250 "C) prior to the hydrogenation of dodecanoic acid. Furthermore, the average crystallite sizes of metallic copper for the catalysts that were first reduced, and then tested, are significantly higher (especially for CZA-3 and Cu60 catalysts) than those of CuO in the catalyst's precursors. Results obtained in the hydrogenation of dodecanoic acid over different Cu0-Zn0-AI2O3catalysts and the commercial Cu-Cr/MgO-Si02 catalyst are given in Table 28. Among in situ pre-reduced Cu0-Zn0-A1203catalysts CZA-1 was the most active and selective towards dodecanol. The high selectivity of CZA-1 catalyst to dodecanol can probably be attributed to the presence of ionic copper species. Indeed, the presence of highly dispersed C u 2 0 phase in CZA-1 catalyst was also detected by XRD (see Table 27). Based on the average crystallite size data given in Table 27 the dispersion of metallic copper phase in Cu0-Zn0-A1203catalysts was calculated using the following equation: D(%) = 120/da,(nm)?33Dispersion of the metallic copper for CZA-1, CZA-2, CZA-3 (commercial samples) and CZA-4 (laboratory made, calcined at 350 "C prior to reduction) catalysts was 12,8,2.4 and 6 %, respectively. The concentration of metallic copper surface species (Cuosud)was calculated using composition (see Table 25) and dispersion data. These data for CZA-1, CZA-2, CZA-3 and CZA-4 catalysts, in situ reduced at 205"C, were 0.28, 0.27, 0.15 and 0.26 mmOl/gcatalyst (see Table 27). With respect to the activity of these catalysts in the hydrogenation of dodecanoic acid, it is interesting to note that the higher the amount of Cuosudof the CZA catalysts the higher the activity, i.e. the conversion of dodecanoic acid, of these types of catalysts (see Table 28).232 On the contrary, there was no correlation between the amount of surface metallic copper atoms and the conversion of dodecanoic acid on differently pre-reduced Cu60 catalyst. This suggests that the pre-treatment affected not only the dispersion of metallic copper, but also the surface composition (Cu/Cr atomic ratio, presence or absence of ionic copper species). Indeed, the amount of

Table 25

Composition of diflerent copper-containing catalysts (Reproduced from ref. 257 with permission) ~

~~~

Catalyst

CUO, wt. %

ZnO, wt. %

Alr03, wt. %

CZA- 1 CZA-2 CZA-3" CZA-4-350 Cu60"

37 27 51 35 80

36 26 31 26 3b

27 47 18 39 17"

a

data given by manufacturers, chromia, MgO-SiO,.

79

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

Table 26

Average crystallite sizes of CuO and ZnO in the oxide form of catalysts determined by X R D (Reproduced from ref. 257 with permission)

Catalyst ~~

Tcalcinariom

unknowna unknown unknown

CZA-1 CZA-2 CZA-3 CZA-4 CZA-4 CZA-4 Cu60b a

"C

CuO, dav.nm

ZnO, dav.nm

8 8 10 22 22 34 11

11 10 10 6 6 6

~

300 350 500

unknown

Calcination conditions for commercial samples are unknown, chromia content: 3 wt. %.

Table 27

Average crystallite size and estimated distribution of copper species in the catalysts used in the hydrogenation of dodecanoic acid (Reproduced from ref. 257 with permission)

CZA-1 CZA-2 CZA-3 CZA-4-350 Cu60 Cu60 Cu60 Cu60 Cu60

in situ insitu insitu in situ in situ in situ in situ

H2 N2/H2

205 205 205 205 205 250 270 250 250

10150 151100 501100 20/100 25/45 34/90 34/100 71/100 66/100

5/50

5/45 5/10

5/10

0.28 0.27 0.15 0.26 0.60 0.44 0.44 0.21 0.23

Prior to catalytic reaction in situ treatment was carried out in hydrogen at 9 MPa for 2 hours. Accuracy of temperature control during in situ treatment was k 5 "C.Average crystallite sizes and distribution of copper species are given in nm and %, respectively.

metallic copper surface atoms does not correlate directly, even for CZA catalysts, with the yield and selectivity to dodecanol (compare Cuosurf. and activity data in Table 27 and Table 28, respectively). Since the highest dodecanol yield was obtained on CZA-1 catalyst, it is suggested that ionic surface species should also play a role in the hydrogenation of dodecanoic acid (compare data given in Table 27 and Table 28). It is important to note that water also forms during hydrogenation of carboxylic acids. Therefore, water resistance of the catalysts under reducing condition can also influence their performance. Indeed, the presence of water in reducing atmosphere significantly increased crystallite size of the copper.234The average copper crystallite sizes given in Table 27 suggest that among CuO-ZnOA1203catalysts CZA-1 might be most resistant to water. The effect of different in situ and ex situ reduction treatments on the performance of Cu60 catalyst is also given in Table 28.232 The higher the temperature of in

80

Table 28

Catalysis

Hydrogenation of dodecanoic acid to dodecanol on diflerent copper-containing catalysts (Reproduced from ref. 257 with permission) Conversion TreatmentrC %

ROH yield

Catalyst CZA- 1 CZA-2 CZA-3 CZA-4 Cu60 Cu60 Cu60 Cu60 Cu60

in situ/205 in situ/205 in situ/205 in situ/205 in situ/205 in situ/250 in situ/270 H2/25O N/H2/250

83 19 4 19 85 36 11 19

97 93 38 82 99 98 80 90 37

%

5

Selectivity % ROH

Selectivity % R'COOR

85 21 11 23 86 37 14 21 13

13 76 87 75 13 60 83 77 85

CZA-4 catalyst was calcined at 350 "C.Prior to reaction catalysts were in situ pre-reduced in hydrogen at 9 MPa, T ( 5 "C)for 2 hours. Ex situ treatments were carried out in flowing gases at atmospheric pressure for 2 hours. Reaction conditions: T = 250 "C,PH2= 6 MPa, t = 1 1 hours. ROH = dodecanol; RCOOR = dodecyl dodecanoate, R = -C,2H25, R'= -C,,H23.

situ reduction the lower the conversion of dodecanoic acid and the selectivity to dodecanol. The highest dodecanol selectivity was obtained on Cu60 catalyst reduced in situ at 205 "C prior to the activity test. In this catalyst oxidized copper species were detected by XRD after its use in the hydrogenation of dodecanoic acid (see Table 27). This suggests again that ionic copper species may play an important role in activating dodecanoic acid. Indeed, pre-reduced copper chromites have found to be strongly deactivated in soybean oil hydrogenation due to disappearance of Cu(I1) and Cu(1) species and to decreasing Cu/Cr ratio on the catalyst s u r f a ~ e .The ~ ~ promotion ~ , ~ ~ ~ of copper with chromia results in stable catalyst structure and high copper surface area. Therefore, chromia is found to act mainly as a structural promoter.237 Ex situ reduction treatments resulted in low dispersion of metallic copper (see Table 27) and low selectivity to dodecanol (see Table 28). Low selectivity to dodecanol can probably be attributed to the low amount or absence of ionic copper species in these catalysts. The difference in the conversion of dodecanoic acid between the catalyst reduced in hydrogen and one treated first in nitrogen and then hydrogen is obvious. This difference in activity can not be explained by differences in dispersion of metallic copper, rather by differences in the surface properties of the catalysts. Indeed, Cu+ ions on the surface are required for the hydrogenation reaction, and a possible enrichment of the surface in chromium ions may enhance selectivity to side products (e.g. dodecyl d~decanoate)?~' In conclusion, all the above results suggest that in the hydrogenation of fatty acid, and more generally carboxylic acids, over chromium promoted copper catalysts the presence of Cu+ ions is required to achieve high conversion and selectivity towards alcohol. In the active catalysts both metallic copper particles and copper ions were present creating 'Cu+-Cuo' ensemble sites.

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

81

2.5.1.3 Summary on Copper Chromite Catalysts. Copper chromite based catalysts are frequently used for the hydrogenation of both saturated and unsaturated fatty acids (e.g oleic acid) to the corresponding alcohol. Copper chromite catalysts are usually promoted with zinc, however, Cu0-Zn0-A1203 catalysts are mostly used in the hydrogenolysis of fatty esters to alcohol. The authors of this review have obtained the highest reported dodecanol selectivityin the hydrogenation of dodecanoic acid on a copper chromite type commercial catalyst reduced in situ at 205°C prior to reaction. In this catalyst oxidized copper species were detected by XRD after its use in the hydrogenation of dodecanoic acid. Therefore, it was proposed that ionic copper species might play an important role in activating the surface carboxylate species in the hydrogenation of dodecanoic acid. The results strongly suggest that in the active catalysts both metallic copper particles and copper ions were present creating Cu+-Cuo ‘metal ion - metal nanocluster’ ensemble sites. 2.5.2 CuO-ZnO-A&03 Based Catalysts. 2.5.2.1 Literature Background. 2.5.2.1.I Hydrogenolysis of Esters. - The selective hydrogenolysis of methyl and ethyl acetate to ethanol over different copper-based and supported Group VIII metal (Pd, Rh, Pt, Co, Ni) catalysts has been studied in the gas phase at 175-350“C and 1-60 bar. Measurements of activity and selectivity show that the copper-based catalysts CuO/Mg0-Si02, CuO/ZnO/MnO/A1203, and CuO/ZnO/Fe203in particular exhibit very high selectivities for ethanol at nearly complete acetate conversions under moderate reaction conditions. In contrast, supported monometallic catalysts containing Pd, Rh, or Ni were found to be ineffectivein the hydrogenation of the acyl group due to multiple splitting of C - 0 and C-C bonds forming light hydrocarbons, acetic acid and carbon oxides. It is noteworthy that electropositive promoters, e.g. iron and zinc, significantly modify both the activity and the selectivity of ethanol formation in the hydrogenolysis of methyl or ethyl acetate over Cu-based catalysts.238 A series of Cu/Si02 catalysts promoted with manganese, iron, cobalt, nickel, molybdenum, magnesium and yttrium was prepared by homogeneous deposition precipitation and tested in the hydrogenolysis of methyl acetate.239For most catalysts a copper metal surface about 15 m2/&,, was measured. TPR revealed an intimate copper-promoter contact. Despite similar Cuosurface areas, the activity in ester hydrogenolysis varied considerably with the type of promoter, in the following order: Mo>Co>Zn>Mn>Fe>Y>Ni>Mg.As a function of the metaloxygen bond strength of promoter, a volcano-type catalyst activity plot was observed as shown in Figure 44.239The only exception is COO,despite the low copper surface area of the Co promoted catalyst a copper-cobalt phase is responsible for the high activity in methyl acetate hydrogenolysis. High selectivity to light alkanes on Co and Mo promoted catalysts may be due to the reduction of promoter to metallic state. There was now correlation between the conversion of methyl acetate and the generalized electronegativity. However, with respect to the alkane formation, the product of over-reduction, a correlation with generalized electronegativity was obtained, inferring that the formation of alkanes proceeds on Lewis acid sites (see Figure 45).239

82

Catalysis

:a

1"

a

coo

NO F ~ O MOO,

zno

M ~ OY,O, M ~ O ' ' CU/SLO,

Increasing promoter M-0 bond strength

Figure 44

+

Methyl acetate conversion at 197 "Cover CuJSi02catalyst doped with diflerent metal oxides (left-hand axis) and ethane selectivity (right-hand axis) at 247 "C as a function of oxygen-metal bond strength of the promoter (normalized per oxygen atom) (Reproduced from ref. 239 with permission)

In a paper published by Tanaka and O ~ a kiti was ~ ~ substantiated that a rough measure of Lewis acidity can be obtained from the generalized electronegativity of the metal ion (Xi) (Xi = (1 + 2Z)/Xo, Z = formal charge of oxide, Xo = electronegativity of the neutral atom). The acid-base properties of the promoter are important in ester hydrogenolysis as it was reported that formation of alkenes, by dehydration of alcohols, in ester hydrogenolysis proceeds on acid sites.240Under conditions of methyl acetate hydrogenolysis, however, mostly ethane is formed, likely through hydrogenation of ethene. When both activity and selectivity are taken into account, manganese and zinc are the promoters of choice.239It is important to note that both key steps of the ester hydrogenolysis reaction, namely the activation of the carbonyl group of the and the side reactions resulting in the formation of heavy ester and hydrocarbons take place on Lewis acid sites (see reaction 16 and 17, and Reaction 22 and 23, respectively in paragraph 2.3.2.3.).239 It has been proposed that the active site in copper catalysed hydrogenation reactions is the association of copper (I) in an octahedral environment with hydride ion. The ionic character of this site explains its strong reactivity toward the highly polarized C=O double bond of carbonyl However, Lewis acid sites, e.g. ionic copper or tin species can catalyse both the transesterification, i.e., formation of heavy ester in Sn-Ru/A1203catalysts72and the formation of hydrocarbon from ester via hydrogenolysis over copper-containing catalysts.239 The presence of ionic copper species in Cu0-Zn0-A1203catalysts has already been suggested and proved by several research groups.242,243,244*245 According to Fujitani the role of metal oxides is the stabilization of Cu+ species and improve-

83

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

I

o r " 5

- '

'

I

10

"

'

'

' '

I

15

'

"

0

20

-

Generalised electronegativity X,

Figure 45

Methyl acetate conversion at 197 "Cover CulSiO, catalyst doped with different metal oxides (left-hand axis) and ethane selectivity (right-hand axis) at 247 "C as a function of generalized electronegativity (see text below) (Reproduced from ref. 239 with permission)

ment of the dispersion of Cu. ZnO would control the Cu+/Cuo ratio, whereas alumina acts as a dispersion agent. It has also been found that in the hydrogenation of CO the latter is activated through interaction with oxygen containing sites of Cu -O-Cu+ type.243 +

2.5.2.1.2 Reductive Alkylation of Amines. - Mixed secondary amines, with general formula of RNHR (where R and R' are different alkyl or cycloalkyl groups) are usually prepared in the presence of a hydrogenation/dehydrogenation type catalyst using a primary or a secondary amine and an alcohol as starting materials. In the alkylation of amines with an alcohol copper-containing catalysts are recommended to avoid the transalkylation at the nitrogen atom and thus the formation of symmetrically substituted secondary or tertiary a m i n e ~ . ~ ~ ~ Cooper-containing catalysts are often used for the production of amines from alcohols, however the role of different copper surface species, i.e. metallic or ion, is still debated. Earlier studies confirmed that in the catalytic alkylation of amines with an alcohol the rate-determining step is the abstraction of a a-hydrogen from the alcohol resulting in the formation of the corresponding aldehyde type surface inte~mediate.~~' This surface intermediate can react then with an amine to produce a Schiff-base and subsequently, the secondary imine undergoes hydrogenation to a secondary amine. There are only scarce data on the catalytic mono-methylation of primary amines with methanol. The C1 methanol precursor undergoing amination was deduced to be an aldehyde type intermediate.248It has also been demonstrated

84

Catalysis

that in the methylation of fatty nitriles partial dehydrogenation of methanol is needed.249 Cu-Ni catalyst supported on Si02/A1203gave very high activity and selectivity in the dimethylamination of d o d e c a n 0 1 . ~When ~ ~ ~ the ~ ~ Cu/Ni ~ weight ratio was 4, a maximum in activity and selectivity was observed. The valence states of surface metals of catalyst were analysed by ESCA. Copper was reduced to Cuoor Cu+ in hydrogen atmosphere at 200 "C. Nickel was not reduced under the same condition. It is not surprising because the heat of formation (-AH298) of nickel oxide is higher than that of CuO, 58.4 and 38.5 kcal/mol, respectively. It has also been found that when there was much Ni2+on the catalyst surface, the selectivity for the amination reaction was low. Ni2+seemed to accelerate the disproportionation of Me2NH. These results suggested us that a cooperation of Cu+/Cuoand Ni2+ P i ored-ox pairs were responsible for the enhanced performance of bimetallic catalyst. It has been found that besides reductive amination copper catalysts are also active in the dehydrogenation of methanol to methyl f ~ r m a t e The . ~ ~ dehyd~ rogenation of methanol can be considered as a reversible reaction step of the methanol synthesis. In the methanol synthesis over Cu0-Zn0-AI2O3catalysts both ionic and metallic copper has been suggested as the active site. It has been found that the catalyst containing the maximum amount of ionic copper dissolved in the zinc oxide was the most active in the methanol synthesis.253 Among various Cu/ZnO/Zr02catalysts with the Cu/Zn ratio of 3/4, the one with 15 wt. Yo of Z r 0 2 obtained the best activity for methanol synthesis by hydrogenation of CO. The TPR, TPO and XPS analyses revealed that a new copper oxide phase is formed in the calcined Cu/ZnO/Zr02 catalysts by the dissolution of zirconium ions in copper oxide. In addition, the Cu/ZnO/Zr02 catalysts with 15 wt. YOof Z r 0 2turned out to contain the largest amount of the new copper oxide phase. When the Cu/ZnO/Zr02 catalysts were reduced, the Cu2+species present in the Z r 0 2 lattice was transformed to Cu+ species. This lead to the speculation that the addition of Z r 0 2to Cu/ZnO catalysts gave rise to the formation of Cu+ species, which was related to the methanol synthesis activity of Cu/ZnO/Zr02catalyst in addition to Cu metal particles. Consequently, the ratio of Cu+/Cuo is an important factor for the specific activity of Cu/ZnO/Zr02 catalyst for methanol synthesis.254 Okamoto et ~ 1 investigated . ~ ~ by~ XPS the surface state of co-precipitated CuO-ZnO catalysts reduced at 250°C with hydrogen. It was concluded that both the two-dimensional copper metal and Cu+ was formed from ionic copper dissolved in the ZnO lattice, while the three dimensional copper metal particles were originated mainly from crystalline copper oxide phase. The two-dimensional Cuo-Cu+species were suggested to be catalytically active for the methanol synthesis at low temperature. Copper-zinc-alumina mixed oxide catalysts has been investigated in the lowtemperature methanol synthesis. Three different copper-containing species were identified in the spent catalysts: (i) metallic copper, (ii) CuO, and (iii) copper not detectable by XRD analysis, the latter being probably related to the ZnO matrix. While no correlation existed between the catalytic activity and only one of these

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

85

species, a good fit was observed both with the sum and the product of the amount of CuO and undetected copper. Therefore, both the easily re-oxidizable form of copper and the copper related to the ZnO have been suggested as active species, even though it was not possible to determine if they have similar activities or must be present together at the same time.256 Based on XPS results it was proposed that the role of zinc oxide is to stabilize copper as Cu(1) and the latter was the active site for the methanol synthe amount of t h e ~ iMonnier ~ . ~et al.259 ~ ~have ~ found ~ ~ a~correlation ~ ~ ~ between ~ surface Cu+ species determined by XPS and methanol synthesis rate. The role of ZnO in Cu/ZnO methanol synthesis catalysts has been studied by Nakamura et a1.260using different techniques (XRD, TEM-EDX, XPS, AES). It has been found that the methanol synthesis activity increased with decreasing Cuo surface area. The role of ZnO was ascribed to the stabilization of Cu+ species, the pivotal catalytic species for the methanol synthesis. All of the above results indicate that ionic copper species are active in the methanol synthesis over Cu-Zn0-A1203catalysts. Contrary to that, in the dehydrogenation of cyclohexanol to cyclohexanone over reduced Cu-Zn0-A1203 catalysts a correlation was found between activity and reversible carbon monoxide uptake measured by IR spectroscopy, therefore, it was proposed that metallic copper species are the probable active sites in this dehydrogenation reaction.261 Literature data discussed above indicate that in the alkylation of a primary amine with an alcohol over a copper-containing catalyst the first and ratedetermining step of the reaction is the dehydrogenation of the a 1 ~ 0 h o 1 . ~ ~ ~ * ~ ~ However, the nature and the oxidation state of the active site, i.e. Cuo or Cu+ is still debated, although more and more data indicate that ionic copper species may play an important role in the dehydrogenation of methano1.252~258~259 2.5.2.2 Characterization and Catalytic Activity. 2.2.2.2.1 Hydrogenolysis of Ethyl Dodecanoate on Cu0-Zn0-A12O3 Catalysts. - The authors of this review have also investigated the liquid phase hydrogenolysis of ethyl dodecanoate over three commercial Cu0-Zn0-A1203(CZA-1, CZA-2 and CZA-3) and a laboratory made Cu0-Zn0-A1203catalyst (CZA-4) in a stirred tank reactor at 90 bar hydrogen pressure and 250 0C.232 X-ray diffraction (XRD) was used to characterize both the oxide and form of the catalysts, as well as the reduced and then used catalysts (see previous paragraph, Table 26 and Table 27).232 In situ reduced Cu0-Zn0-A12O3 catalysts. - The results of ethyl dodecanoate hydrogenolysis over in situ reduced Cu0-Zn0-A1203catalysts are summarized in Table 29.232It is noteworthy that the conversion of ethyl dodecanoate varies between 80 and 91 YO,CZA-3 and CZA-2 catalysts being most and least active, respectively. The conversion of ethyl dodecanoate seems to correlate with the copper content of the catalysts. The higher the copper concentration of the catalyst the higher the conversion (compare data given in Table 25 of the previous paragraph and Table 29 of this paragraph). The ratio copper concentration/conversion both in YOincreases in the order: CZA-2,27/80 < CZA-4,35/85 = CZA-1, 37/85 < CZA-3, 51/91. The selectivity to dodecanol only slightly varied between 89 and 93 YOon CZA-2, CZA-3 and CZA-4 catalysts. The

86

Table 29

Catalysis

Hydrogenolysis of ethyl dodecanoate to dodecanol on diflerent in situ reduced copper-containing catalysts (Reproduced from ref. 232 with permission) Conversion

ROH yield

Catalyst

%

%

ROH

Selectivity, % HC R'COOR

CZA- 1 CZA-2 CZA-3 CZA-4-300 CZA-4-350 CZA-4-400 CZA-4-500

85 80 91 84 85 81 81

52 74 84 75 76 74 73

61 93 92 89 89 91 90

19 <1 1 <1 <1 <1 <1

~

~~

6 2 1 1 <1 3 2

CZA-4 catalyst was calcined at 300,350,400 and 500 "C,respectively. Prior to reaction catalysts were in situ pre-reduced in hydrogen at 9 MPa, 205 f5 "Cfor 2 hours. Reaction conditions: T = 250 "C,P,, = 9 MPa, t = 8 hours. ROH = dodecanol; HC = hydrocarbons; R'COOR = dodecyl dodecanoate, R = -CI2Hz5, R = -C,]HZ3.

temperature of calcination only slightly affected the performance of CZA-4 catalysts. With respect to the calcination temperature the highest conversion and yield, 85 and 76 % respectively, was obtained on CZA-4 catalyst calcined at 350°C. Among in situ treated catalysts CZA-3 produced the highest 84 % dodecanol yield. Low dodecanol selectivity obtained on CZA-1 catalyst can be explained by its high reactivity to form hydrocarbon and heavy ester. This can be attributed to the presence significant amount of C u 2 0phase species provided probably by the preferred Cu/Zn/Al atomic ratio and relatively high dispersion of copper in CZA-1 catalyst (compare composition and XRD data given in Table 25 and Table 27 in the previous paragraph). Indeed, highly dispersed Cu20 was detected by XRD even in used catalysts after dodecanoic acid hydrogenation (see Table 27 in the previous paragraph). Therefore, the extent of hydrodecarbonylation of aldehyde-type reaction intermediate to undecane and CO and hydrodecarboxylation of dodecanol to dodecane (Lewis acid catalysed reactions) is more pronounced on CZA-1 catalyst than over other catalysts. Besides hydrocarbons and heavy ester, small amounts of ethers and unidentified products were also detected in the reaction mixtures. It is also noteworthy that in the hydrogenolysis of esters no water is formed in the main reactions, only minor amount of water is formed during in situ reduction of catalysts and in the side reactions. Therefore, in the hydrogenolysis of ethyl dodecanoate no information can be obtained about the water tolerance of the Cu0-Zn0-A1203catalysts. E x situ Reduced Cu0-Zn0-A1203Catalysts. - The effect of ex situ reduction temperature on the performance of CZA-3 and CZA-4 catalysts was also investigated. Results given in Table 28 indicate that over CZA-3 catalyst the reduction temperature in the range of 250-350°C had almost no effect on ethyl dodecanoate conversion, dodecanol selectivity and yield.232This suggests that treatment of CZA-3 catalyst in flowing hydrogen in the above temperature range results in similar and stable catalyst structure.

87

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

The effect of calcination temperature on the performance of CZA-4 catalyst is also shown in Table 30. The calcination of CZA-4 catalyst was performed in air at 350 "C. After calcination the catalyst was subsequently reduced under the flow of H2 at different temperatures. Results listed in Table 30 indicated that the reduction temperature hardly affected the product selectivity and the yield of dodecanol. The temperature of calcination and in situ or ex situ reduction in hydrogen only slightly affected the activity and the selectivity of both laboratory made and commercial Cu0-Zn0-AI2O3 catalysts in the hydrogenolysis of ethyl dodecanoate. In conclusion the origin, composition and treatments of Cu0-Zn0-A1203 catalysts strongly affected their phase composition and average crystallite sizes of CuO or Cu and reducibility in TPR.232 Comparable selectivity of commercial (CZA-3) and laboratory made (CZA-4) catalysts suggests that co-precipitation provides intimate contact between copper and zinc resulting in probably the formation 'Cu+-Cuo' ensembles sites that are active in the hydrogenolysis of fatty esters to alcohol. ZnO may play a role in the formation and stabilization of 'metal ion-metal nanocluster' ensemble sites. 2.5.2.3 Alkylation of Butylamine with Methanol. The authors of this review investigated the alkylation of butylamine with methanol over a reduced commercial Cu0-Zn0-AI2O3catalyst (CZA-1).244Various techniques (XRD, XPS) and titration of surface metallic copper with N 2 0 were used to estimate the amount of different type of copper species in the catalyst. The dispersion of metallic copper species was calculated from the average crystallite size determined by XRD using the following ion: D(%) = 120/d(nm).233 The surface area of metallic copper in the catalysts reduced at different temperatures was determined by titration with N20.264 Controlled surface reaction (CSR) of tin tetraethyl with hydrogen adsorbed on

Table 30

Catalyst

Hydrogenolysis of ethyl dodecanoate to dodecanol on diflerent ex situ reduced copper-containing catalysts (Reproduced from ref. 232 with permission) Treduction Conversion ROH yield % % ROH "C ~~

CZA-3 CZA-3 CZA-3 CZA-350 CZA-350 CZA-350 CZA-350

250 300 350 250 300 350 400

93 92 93 83 79 80 79

84 82 86 74 73 71 73

90 89 92 89 92 89 92

Selectivity,

% HC

R'COOR

_____

<1 <1

<1 <1 <1 <1 <1

2 1

<1 1

<1 2 1

CZA catalyst was calcined at 350 "C. Prior to reaction catalysts were ex situ pre-reduced in a hydrogen flow at atmospheric pressure for 2 hours. Reaction conditions: T = 250 "C, P,, = 9 MPa, t = 8 hours. ROH = dodecanol; HC = hydrocarbons; R'COOR = dodecyl dodecanoate, R = -CI2H2,,R'= -C, ,H,,.

88

Catalysis

metallic copper was used to characterize the reactivity of metallic copper sites on the catalyst. The CSR was carried out in benzene solvent at 50°C according to the following equation:lM Cu-H,

+ S n E t * Cu-SnEt-, + x C2Hs (x = 3-4)

(24) The methylation of butylamine with methanol was carried out in a stainlesssteel stirred autoclave. (The reaction conditions were as follows: Incatalyst = 4.5 g; pretreatment by stepwise heating in hydrogen up to Treduaion "C; heating rate: 5 "C/min, t = lh; amount of reactants: MeOH 1.3 M, BuNH2 0.34 M, molar ratio of methanol/butylamine = 3.9; H2 pressure at 20°C = 30 bar; Treactjon = 185 "C; treaction = 8 hours). The effect of the catalyst reduction temperature was studied in the temperature range of 150-500 0C?44The overall copper balance in the CuO-ZnO-Al203 catalyst can be written as follows: Cutotal = Gun+

+ CUOsufiace + C~Obulk

(25)

The temperature of reduction controls both the (ionic copper)/(metallic copper) and the (surface copper)/(bulk metallic copper) ratios. Upon increasing the reduction temperature both ratios decrease, due to partial reduction and sintering, respectively. Therefore, the reduction temperature can also alter the amount of accessible ionic copper species in the catalyst. The average crystallite size and dispersion of metallic copper and the amount of metallic and ionic copper species as well as the activity of CZA-1 catalyst reduced at different temperatures are summarized in Table 31.244 The average crystallite size data indicate that significant sintering of the copper particles starts only at reduction temperatures 300°C or above. This result is in good agreement with the findings of Gines and Ape~teguia.~~' They studied the effect of the reduction temperature on the resulting Cuocrystallite size of a Cu0-Zn0-A1203(CZA-1) water-gas-shift catalyst. Although initial copper sintering was noted at about 275 "C, severe loss of the metallic dispersion took place only above 300 "C. The amount of different copper species is also given in Table 31. The amount of surface metallic copper species was calculated from the results of N 2 0 titration. The amount of bulk metallic copper was estimated using CUOsudace and dispersion data. The total amount of ionic copper atoms was estimated by subtracting the amount of metallic surface and bulk copper atoms from the total amount of copper in the catalyst. Data given in Table 31 indicates that the alkylation activity of the catalyst decreased with the reduction temperature, whereas the amount of metallic copper increased. The amount of surface metallic copper increased rapidly up to a reduction temperature of 300 "C, while further increase in reduction temperature up to 500 "C resulted in only slight increase in surface metallic copper. It is interesting to note that upon increasing the reduction temperature form 150 to 500 "C, the amount of surface metallic copper atoms increased, while the alkylation activity of the catalyst decreased. This decrease was only slight in the temperature range of 150-250 "C, while more pronounced decrease was observed above 250 "C.

89

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

Table 31

Eflect of reduction temperature on the average crystallite size and dispersion of metallic copper and the amount of metallic and ionic copper species and the activity of Cu0-Zn0-A1203catalyst (CZA-I) (Reproduced from ref. 244 with permission)

~~

150 200 250 300 500

7.0 7.0 7.0 7.9 10.0

~

17 17 17 15 12

0.284 0.366 0.400 0.533 0.565

1.375 1.774 1.929 2.975 4.077

2.984 2.503 2.313 1.034 0

48.6 46.2 4.58 3.80 3.40

d , , - the average crystallite size was calculated from X R D line broadening of Cu(ll1) line. D - dispersion of metallic copper was calculated from the average crystallite size. CuosUrf. - the amount of metallic surface copper species was measured by N,O titration. Cuobulk - the amount of bulk copper was calculated from the amount of metallic surface copper species and the dispersion. Cu"+- the amount of ionic copper species was calculated as a difference between the total amount of copper and that of metallic copper in the catalyst.

Figure 46 shows the dependence of the rate of surface reaction (24)and the rate of alkylation as a function of the amount of surface metallic copper atoms in the catalyst?& The rate of surface reaction (24)increases with the amount of metallic copper. This behavior can be expected as the reaction of tin tetraethyl with '04 adsorbed hydrogen takes place only on metallic copper surface sites (Cuosurface). Contrary to that, as emerges from Figure 46, the alkylation activity of the catalyst decreases with increasing the amount of zero valence surface copper atoms in the catalyst. Figure 47 shows a direct correlation obtained between the sum of the amounts of metallic surface copper atoms and ionic copper atoms and the alkylation activity of the catalyst.244This linear correlation means (correlation coefficient r = 0.982)that both sites are active in the alkylation rea~tion,2'~ however the two sites should have different specific activities. Indeed, literature data indicate that in the dehydrogenation of cyclohexanol to cyclohexanone on Cu0-Zn0-A1203 catalysts, two kinds of copper active sites (monovalent copper and metallic copper) have been revealed. It has been shown that monovalent copper is significantlymore active than metallic copper. Site densities of Cu+ and Cuowere 5.2 x lo1*m-2 and 1.4 x l O I 9 m-2, and TOF 1.2 s-' and 0.086 s-'were obtained, re~pectively.~~' Therefore, it is suggested, that independent of the reduction degree of the Cu0-Zn0-A1203catalyst, the type of active sites in the ratedetermining step should be the same. Since the alkylation activity of the catalyst decreased with increasing amounts of surface metallic copper, the active sites for the partial dehydrogenation of methanol in the rate-determining step should be the ionic copper species, most probably Cu(1). XPS results also confirmed that on the surface of the reduced CuO-ZnO-Al203 (CZA-1) catalyst after treatment with methanol at room temperature for 10 minutes mainly Cu+ species could be detected.2uTherefore, one can assume that

90

Catalysis

2o 10

1

8

I

8

0,2

0

0,4

0,6

cues, mmovg catalyst Figure 46

Correlation between the amount of metallic surface copper and the activity in butylamine alkylation (A) and surface reaction (24) of tetraethyl tin (a) on CZA-1 catalyst (Reproduced from ref. 21 5 with permission)

Y

I

cues+ Figure 47

2 4 cu+', mmovg catalyst

Correlation between the sum of the amounts of ionic and surface metallic copper and the activity of CZA-1 catalyst in the alkylation of butylamine with methanol (Reproduced from ref. 244 with permission)

under reaction conditions at 185"C, in the presence of methanol and water formed in the reaction, on the surface of the catalyst the copper is in ionic state. Further evidence was obtained by Temperature Programmed Reduction (TPR). The TPR results indicate that after reduction at 200,250 and 300 "C,the extent of copper reduction was 67, 86 and 88 %, respectively.232 The non-zero intercept of the activity-copper concentration curve indicates that only a part of the ionic copper species is accessible for the reaction (see

1 : Role of ‘Metal lon-Metal Nanocluster’ Ensemble Sites

91

Figure 47). This can be explained by the fact that a certain part of the ionic copper species can be accommodated in the zinc oxide Another explanation is that the specific activity of metallic and ionic copper species is different.245In addidion, a correlation was found between the alkylation activity and the sum of surface metallic and ionic copper content of the catalyst. Based on this correlation it was suggested that in the alkylation of butylamine with methanol ionic copper species were involved in the rate-determining step of the reaction, i.e. in the dehydrogenation of alcohol into an aldehyde type surface intermediate. 2.5.2.4 Summary on Cu0-Zn0-A1203 Catalysts. - It has been proposed in the literature that the active site in copper catalysed hydrogenation reactions is the association of Cu+ in an octahedral environment with hydride ion. The ionic character of this site explains its strong reactivity toward the highly polarized C=O double bond of carbonyl compounds. The presence of ionic copper species in Cu0-Zn0-A1203catalysts has already been suggested and proved by several research groups. It has been proposed that the role of metal oxides is the stabilization of Cu+ species and improvement of the dispersion of Cu. ZnO would control the Cu+/Cuoratio, whereas alumina acts as a dispersion agent. It has also been found that in the hydrogenation of CO the latter is activated through interaction with oxygen containing sites of Cu+-0-Cu+type. The authors of this review have found that the origin, composition and treatments of Cu0-Zn0-A1203 commercial and laboratory-made catalysts strongly affected their phase composition and average crystallite sizes of CuO in the oxide form and Cu in the reduced form of catalysts, respectively. Catalytic activity and selectivity data obtained on laboratory-made catalysts suggests that co-precipitation provides intimate contact between copper and zinc resulting in the formation ‘Cu+-Cu’ensembles sites that are active in the hydrogenolysis of fatty esters to alcohol. In agreement with literature data ZnO may play a role in the formation and stabilization of ‘metal ion-metal nanoparticle’ ensemble sites. The rate-limiting step of the alkylation of butylamine with methanol on Cu0-Zn0-A1203is the dehydrogenation of the alcohol to an aldehyde type surface intermediate. Literature data indicate that in the dehydrogenation of cyclohexanol to cyclohexanone on Cu0-Zn0-A1203catalysts two kinds of copper active sites, i.e. monovalent copper and metallic copper have been revealed. It was shown that the activity of monovalent copper is 15 times higher than that of metallic copper. Upon increasing the reduction temperature the alkylation activity of the Cu0-Zn0-A1203catalyst decreased, whereas the amount of metallic copper increased. A good correlation was obtained between the sum of the amounts of metallic surface copper atoms and ionic copper atoms and the alkylation activity of the catalyst. Because the reaction rate of butylamine alkylation depends on both the amount of zero- and monovalent copper, it is suggested that the active site is an ensemble of copper ‘metal ion - metal nanocluster’. 2.6 Other Type of Supported Catalysts. - 2.6.1 ‘Ag+ - Ag’ Ensembles on Diflerent Supports. Besides copper-containing catalysts both silver and gold can

92

Catalysis

form ‘metal ion - metal nano-cluster’ type ensemble sites in heterogeneous catalysts. It has been found that the addition of both Cs(1) and Re(VI1) ions significantly increases the catalytic activity of poly(sodium acrylate) (PAA) protected Ag nanoclusters in the oxidation of ethylene to ethylene oxide performed in The promotion effect was attributed to the ensemble effect of both Cs(1) and Re(VI1) ions located near the Ag nanocluster through the interaction with PAA. Pestryakov and D a ~ y d o reported v ~ ~ ~ that silver ions are the most active silver species on the surface of different supports (pumice, aluminosilicate, a-A1203, MgO) in the oxidation of methanol to formaldehyde. Silver-containing ceramic catalyst prepared by impregnating acid washed kaolin with silver nitrate solution after calcination in air at 1200 “C appeared to be highly active in the oxidation of methanol to formaldehydeF6*In this catalyst the silver was stabilized by the presence of [A104] in ionic form. Under the reaction conditions, the silver ions were partially reduced by the atomic hydrogen released from methanol and formed metallic particles. XPS and UV-Visible diffuse reflectance spectra (UV-Vis DRS) revealed that after the reaction about one-third of silver in the catalyst was still in ionic state. Computational results showed that if silver was positively charged, Lewis acid-base interaction between methanol and silver would be strengthened, and methanol can undergo a stable chemisorbed form on silver surface, which is the key step in the oxidation of methanol to f0rmaldehyde.2~~ Also, the further oxidation of formaldehyde was inhibited and the selectivity of the partial oxidation of methanol to formaldehyde was thus enhanced over the positively charged silver catalyst.270 2.6.2 ‘Sn”+- Co’ Ensembles in Diflerent Catalysts. The hydrogenation of methyl oleate (methyl-9-octadecanoate) into oleyl alcohol (methyl-9-octadecen-1-01) was studied in the presence of a bimetallic CoSn supported over zinc oxide catalyst.271It was found that zinc oxide could activate the carbonyl group and also the hydrogen by a possible heterolytic and homolytic dissociation of dihydrogen. The selectivity to unsaturated alcohol (oleyl alcohol) is significantly enhanced by adding tin to cobalt. The activity and the selectivity to unsaturated alcohol was maximum for an atomic (Sn/Co)bulk ratio of 1. Surface composition of the catalysts was determined by XPS. For all CoSn catalysts, the selectivity to unsaturated alcohol goes through a maximum for a surface Sn/Co ratio close to 2. It has been proposed that the catalytically active species is a mixed COO-(S~O,)~ ensemble where the oxidation state of tin is slightly higher than zero. It is noteworthy that electropositive promoters, e.g. iron and zinc, exhibit significant effects modifying both the activity and the selectivity of ethanol formation in the hydrogenolysis of ethyl acetate on supported Group VIII metal catalysts, such as Pd-Zn/A1203and c ~ - R h - F e / T i o ~ . ~ ~ ~ 2.6.3 ‘M“+- Ir’ Ensembles ( M = Fe, Ge) on Diflerent Supports. Reyes et aL2I2 have recently investigated the hydrogenation of citral (3,7-dimethyl-2,6-0~tadienal) over supported iridium catalyst. The type of support (Ti02and SO2),

1 : Role of 'Metal Ion-Metal Nanocluster' Ensemble Sites

93

the effect of additives (Fe and Ge), and the reduction temperature on the catalyst performance have been examined. It was found by combined use of TEM, XRD, XPS and activity measurements that the presence of ionic species, i.e. promoters or species generated TiO, moieties upon reduction at high temperature of the reducible T i 0 2 support, which lead to catalysts active and selective in the hydrogenation of the carbonyl bond. This is explained in terms of the presence of positively charged (Fe3+,Ge4+)species in intimate contact with iridium particles, which are responsible for the polarization of the C=O bond of aldehyde. 2.6.4 M"+-RhEnsembles ( M = Sn, Ge) in Different Catalysts. The hydrogenation of citral was investigated on silica supported Sn-Rh catalysts prepared by co-impregnation, successive impregnation or an organometallic route using SnBh.273All preparation methods created active sites suitable for the formation of unsaturated alcohols. The co-impregnated catalyst was shown to be as selective as the organometallic one. This performance was attributed to an intimate contact between the metals. TPR profiles support the catalytic activity data and revealed distinct interaction degrees for each catalyst. Co-impregnation appeared as a potential method to prepare bimetallic Sn-Rh catalysts. Rh and Sn-Rh/Si02 catalysts have been prepared, characterized and used in the crotonaldehyde hydrogenation r e a ~ t i o n . 2New ~ ~ bimetallic Sn-Rh catalysts were prepared by reaction of S ~ ( I I - C ~over H ~ )a~pre-reduced Rh/Si02 precursor. It was observed that the temperature of preparation reaction also plays an important role in the nature of the Sn-Rh active sites. Thus, at lower temperatures, 90 "C,Rh(SnBQ-,), sqecies remain adsorbed on the silica surface, whereas at 500 "C, the organometallic residue decomposes leading to Sn-Rh bimetallic catalysts. Tin addition caused significant drop in hydrogen chemisorption capability but only a slight increase in metal particle size. Electron diffraction detected the presence of Rho,RhSn2 and Sn02phases for the bimetallic catalysts and Rho for the monometallic ones. XPS showed an important surface enrichment in tin suggesting that SnO, species would migrate and deposit on the metal crystals. The active sites generated upon these treatments allow the polarization of the carbonyl group and consequently an enhancement in the selectivity to crotyl alcohol is The binding energies (BE) of the Rh 3d5,2peak at ca. 307.2 eV corresponding essentially to Rho species was observed in the monometallic Rh/Si02 catalysts. The Sn 3d profile for both Sn-Rh/Si02 (sol-gel and impregnated) catalysts indicated the different species. Curve fitting of the experimental spectra indicated both reduced tin species, Sno(BE = 485.1 eV) and oxide species (BE = 488.0 eV). The discrimination between Sn2+ and Sn4+ species was however extremely difficult, because the BE of the Sn 3d512 core level is virtually the same for the two tin oxide species.275The proportion of reduced tin species was close to 50 %, higher than those usually reported for Sn-Pt and Sn-Rh catalysts prepared by conventional impregnation methods.276 2.6.5 'Mn+- Nz" ( M = Fe, Cr) Ensembles in Different Catalysts. The improved activity of silica-supported 75% Ni - 25% Fe catalyst compared to pure Ni

94

Catalysis

catalyst in the liquid-phase hydrogenation of benzonitrile was attributed to the formation of either Ni-Fe”+ sites or Ni-Feo alloyed sites, in which case there was a charge transfer from Fe to nickel. An increase in electron density of Ni due to charge transfer from Fe, should lead to increased back bonding, thereby weakening the carbon-nitrogen bond, which in turn should favor hydrogenation. Thus, the CN group n-complexed to Ni can be activated towards hydrogenation by Fe (or Fe2+)attracting electron from the orbital on nitrogen.131 It is interesting to note that chromium ions have been used as promoters in Raney nickel based catalysts for the selective acetophenone reduction.277The selectivity increase in the hydrogenation of the C=O bond of acetophenone over both chromium ion promoted Raney nickel and R u - C ~ - B / S ~ Ocatalysts ?~~ was attributed to the strong interaction of the oxygen atom of the carbonyl with Cr ions. All these results suggest the formation of ‘metal ion - metal nanocluster’ ensemble sites also in nickel-containing catalysts. 2.6.6 ‘M”+- Pd’ ( M = Fe, Sn, Ti) or ‘Pd” - Pd’ Ensembles in Digerent Catalysts. The use of PdFe heterobinuclear carbonyl complexes for the preparation of Pd-Fe/Si02catalysts lead to bimetallic particles composed of PdFe alloy and Pd-Fe3+ pairs with great intimacy.279The positive effect of iron could be attributed to the activation of the N - 0 bonds by Fen+ions in mixed site Pd-Fe3+as proposed in the hydrogenation of 2,4-dinitrotoluene on Pd-Fe/Si02catalyst.280 Underpotential deposition of different metal ions including Sn2+on Pd,99as well as CSR of tin alkyls with hydrogen adsorbed on palladium281can result in the formation of both PdSn alloy particles and Sn”+-Pdmixed sites. It is noteworthy that besides highly dispersed palladium clusters, ionic Pd was also detected by ESR in Pd/Li-A1203 catalysts prepared by anchoring PdC12 onto alumina modified with butyl lithium. These catalysts were highly active and selective in the hydrodechlorination of substituted aromatic The coexistence of Pd+ and Pdo in the Pd/Li-A1203catalysts suggests the possible formation of mixed ‘metal ion - metal nanocluster’ ensemble sites.

3

Conclusions

This contribution was aimed to review the role of ‘metal ion - metal nanocluster’ active site ensembles (MIMNES) in different heterogeneous catalytic reactions. MIMNES are stabilized at a given boundary layer. These species have significantly higher catalytic activity than supported metal nanoclusters. In this review, the formation, the characterization and the catalytic activity of different MIMNES type active sites has been described and discussed. Results obtained in our laboratory were compared with those published earlier or recently in different model reactions. It has been shown that new types of supported Au, Cu, Sn-Pt, Sn-Ru, Re-Pt, catalysts have been prepared and used for selective hydrogenation of different organic carbonyl compounds (unsaturated aldehydes, esters, carboxylic acids, carboxamides, etc.) and nitriles. Supported Sn-Pt and Au catalysts were also

1: Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

95

tested in the low temperature CO oxidation. The catalysts were characterized by chemisorption of hydrogen, oxygen and CO, hydrogen titration, temperature programmed oxidation and reduction. Different spectroscopic techniques, such as UV-VIS, FTIR, XPS, Mossbauer and methods of computational chemistry were applied to study the formation of MIMNES type active sites and the activation of different carbonyl compounds and CO molecule. SeEective hydrogenation of carbonjl compounds. The involvement of MIMNES type active site has been d.emonstrated in different earlier recent studies. Surface Organometallic Chemistry (SOC) was used to prepare alloy type supported Sn-Pt catalysts. In these supported Sn-Pt catalysts the Lewis-acid type active sites, i.e., SnO, species, or electron deficient Sn*+ entities formed in the atomic closeness of supported platinum nanoclusters were involved in the creation of ‘metal ion-metal nano-cluster’ ensemble sites. The formation of the above species has been shown by chemisorption, Mossbauer and FTIR spectroscopy. The atomic closeness of ionic and metallic species provides unusual activity and selectivity to these catalysts in variety of reactions. Sn-Pt/SiOz catalysts showed high, ‘reactant induced’ activity and unprecedented selectivity in the gas and liquid phase hydrogenation of unsaturated aldehydes into unsaturated alcohols. The in situ formation of S d 4 species in the presence of crotonaldehyde and hydrogen has been verified by in situ Mossbauer and FTIR spectroscopy. The activity of Sn-Pt/Si02catalysts showed a maximum at Sn/Pt = 1 surface atomic ratio in the liquid phase hydrogenation of benzonitrile. Alumina supported Sn-Ru and Re-Pt catalysts were found to be highly active and selective in liquid phase hydrogenolysis of ethyl dodecanoate and butyl acetate to the corresponding alcohol. The content of ionic forms of the second metals strongly affected the performance of these catalysts. The interaction between two metals in these catalysts has been shown by different experimental methods. Low temperature oxidation of CO. Sn-Pt/SiOz catalysts showed unusual high activity in the low temperature oxidation of CO by molecular oxygen. The highest activity was obtained on catalysts with a Sn/Pt atomic ratio around 0.2-0.5. Evidence for the in situ formation of mobile Sn+4species has been found. FTIR of adsorbed CO revealed the formation of Sn+4species and provided information about the in situ surface reconstruction of supported Sn-Pt bimetallic nanoclusters. DFT calculations revealed an interaction between Sn4+ and oxygen atom in the polarized CO molecule. Au/MgO catalysts containing both ionic and metallic forms of gold appeared to be also highly active in CO oxidation. Based on these results and literature analogs it seems to us extremely interesting and important to (i) design, (ii) prepare, and (iii) and test new catalytic compositions containing highly active mono- and multimetallic MIMNES type active sites. We are strongly convinced that the use of new experimental methods (catalyst preparation by combinatorial way and high throughput testing), new approaches (quantum chemical and density functional calculations) and different new in situ spectroscopic techniques shall provide us information about the formation, the stabilization, the surface composition, and the catalytic perform-

96

Catalysis

ance of highly active and selective supported multi-component catalysts containing MIMNES type active sites.

4

Acknowledgements

The authors are grateful to Mrs. I. Borbath, Dr. M. HegedGs, Dr. E. Talas, Dr. A. Tompos, Dr. N. Mahata, Dr. E. Tfirst and Dr. S. Kristyan for help with experimental work and computer modeling. S.G and J.L.M wishes to thank to the Hungarian Scientific Research Fund (OTKA Grant No T-32065 and T43570, respectively) for partial financial support.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

14. 15. 16. 17.

18. 19. 20. 21. 22. 23.

0. Loew, J . Prakt. Chem., 1875,11,372. I. Langmuir, J. Am. Chem. Soc., 1915,37, 1139. H.S. Taylor, J. Am. Chem. Soc., 1923,45,920. H.S. Taylor, Proc. Royal Soc., 1925, A108, 105. H.S. Taylor, J. Phys. Chem., 1926,30, 145. G.A. Mills, H. Heinemann, T.H. Milliken, and A.G. Oblad, Znd. Eng. Chem., 1953, 45, 134. C.G. Myers, W.H. Lang, and P.B. Weisz, Znd. Eng. Chem., 1961,53,299. G.D. Lei, and W.M.H. Sachtler, J . Catal., 1993,140,601. M.J. Ledoux, and C. Pham-Huu, Catal. Today, 1992,15,263. F.J. Cadete Santos Aries, S. Ramirez, G. G. Cervantes, E. Rogemond, and J.C. Bertolini, Appl. Catal., 2003, A238,289. V.E. Alvarez, and D.E. Resaco, J . Catal., 1996,164,467. F.G. Ciapetta, and D.N. Wallace, Catal. Rev., 1971,5,67. M. Boudart, in ‘Proceedings of the 6th International Congress on Catalysis’, ed. G.C. Bond, P.B. Wells and F.C. Tompkins, The Chemical Society, London, 1976, V O ~ .1. p. 1-9. A.A. Balandin, Ado. Catal., Rel. Subj., 1969,19, 1. N.I. Kobozev, Zhurn. Fiz. Khim (in Russian), 1939,13, 1. D.A. Dowden, in ‘Proceedings of the 6th International Congress on Catalysis’, ed. G.C. Bond, P.B. Wells and F.C. Tompkins, The Chemical Society, London, 1976, V O ~ .1, p. 621-631. V. Ponec, and W.M.H. Sachtler, in ‘Proceedings of the 6th International Congress on Catalysis’, ed. G.C. Bond, P.B. Wells, and F.C. Tompkins, The Chemical Society, London, 1976, vol. 1, p. 645-652. H. Weiner, A. Travarelli, and R.G. Finke, J . Mol. Catal., 2003, A191,253. M.Y.H. Kwan, N.W. Cant, D.L. Trimm, and M.S. Wainwright, Appl. Catal., 1987, 31,25. R.K. Grasselli, in ‘Adsorption and Catalysis on Oxide Surfaces’, ed. M. Che and G.C. Bond, Stud. Surf. Sci. Catal., Elsevier, Amsterdam, 1985, vol. 21, p. 275-283. Yu.1. Yermakov, J. Mol. Catal., 1983,21,35. R. Augustine, S. Tanielyan, S Anderson, and H. Yang, Chem., Comrn., 1999, 1527. J.A.M. Brandts, C.V. Cavenaghi, A. Gerlach, and M.J. Burk, Chimica Oggi, 2000,

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

24. 25.

26. 27. 28. 29. 30.

31. 32. 33. 34. 35. 36. 3 7. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47.

48.

49. 50. 51.

52. 53. 54.

97

18,47. A.Yu. Stakheev, and L. M. Kustov, Appl. Catal., 1999, A188, 3 (and references therein). R. Prins, in ‘Chemistry and Chemical Engineering of Catalytic Processes’, NATO Advanced Study Institute Series, Series E, Appl. Sciences, ed. R. Prins and G.C.A. Schuit, Sijthoff and Noordhoff, Alphen aan den Rijn, The Netherlands, 1980, vol. 39, p. 389-397. J.H. Sinfelt, in ‘Catalysis Science and technology, ed. J.R. Anderson and M. Boudart, Springer - Verlag, Berlin-Heidelberg, 1981, vol. 1, p. 287-300. G. Schmid, Chem. Reu., 1992,92,1709. D. Kulkarni, and I.E. Wachs, Appl. Catal., 2002, A237,121. E.M. Crabb, and R. Marshall, Appl. Catal., 2001, A217,41. G. Cocco, S. Enzo, S. Galvagno, Z. Poltarzewski, and R. Pietropaolo, J . Chem. SOC., Farad. Trans. I , 1985,81,321. S. Galvagno, H.S. Minro, R. Pietropaolo, Z. Poltarzewksi, and A. Recca, Gazz. Chim. Ital., 1986,111,265. Z. Poltarzewski, S. Galvagno, R. Pietopaolo and P. Staiti, J . Catal., 1986,102, 190. S. Galvagno, A. Donato, G. Neri, and R. Pietropaolo, J . Mol. Catal., 1990,58,215. S. Galvagno, A. Donato, G. Neri, R. Pietropaolo, and Z. Poltarzewski, J . Mol. Catal., 1987,42,379. J.L. Margitfalvi, I. Borbath, M. Hegediis, E. Tfirst, S. Gobolos, and K. Lazar, J . Catal., 2000, 196,200. K.I. Tanaka, and A. Ozaki, J . Catal., 1967,8, 1. A.B. Boffa, C. Lin, A.T. Bell, and G.A. Somorjai, Catal. Lett., 1994,27, 243. V. Ponec, J . Mol. Catal., 1998, A133,22. S.D. Lin, D.K. Sanders, and M.A.. Vannice, J . Catal., 1994,147, 370. M.A. Vannice, and B. Sen, J . Catal., 1989,115,65. R. Burch, and L.C. Garla, J . Catal., 1981,71,360. C. Betizeau, G. Leclercq, R. Maurel, C. Bolivar, H. Charcosset, and L. Tournayan, J . Catal., 1976,45, 179. F.H. Ribeiro, A.L. Bonivardi, C. Kim, and G.A. Somorjai, J . Catal., 1994,150,186. P. Biloen, F.M. Dautzenberg, and W.M.H. Sachtler, J . Catal., 1977,50,77. B. Coq, and F. Figueras, J . Catal., 1984,85, 197. J. Beltramini, and D.L. Trimm, Appl. Catal., 1987,32,71. G.C. Bond, M.R. Geltshorpe, R.R., Rajaram, and R. Yahia, in ‘Structure and Reactivity of Surfaces’, ed. C. Morterra, A. Zecchina, and G. Costa, G., Stud. Surf: Sci. Catal., Elsevier, Amsterdam, 1989, vol. 48, p. 167-177. G.C. Bond, R.R. Rajaram, and R. Burch, in ‘Proceedings ofthe 9th International Congress on Catalysis’, ed. M.J. Phillips and M. Ternan, Chem. Inst. Canada, 1988, V O ~ 3, . p. 1 130-1 137. J.P.S. Badyal, R.M. Lambert, K. Harrison, C.C.A. Riley, and J.C. Frost, J . Catal., 1991,129,486. G.C. Bond, Chem. Soc. Reu., 1991,20,441. S. Gobolos, N. Mahata, M. Hegediis, I. Borbath, and J.L. Margitfalvi, in ‘Catalysis of Organic Reactions’, ed. D.G. Morrell, Chem. Ind., Marcel Dekker, New York, 2002, V O ~ 89, . p. 565-576. S. Ito, T. Fujimori, K. Nagashima, K. Yuzaki, and K. Kunimori, Catal. Today, 2000,57,247. A.B. Boffa, A.T. Bell, and G.A. Somorjai, J . Catal., 1993,139,602. B. Coq, P.S. Kumbhar, C. Moreau, R. Moreau, and F. Figueras, J . Phys. Chem.,

98

55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80.

81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94.

Catalysis

1994,98,10180. C. Hoangvan, and 0.Zeagaoui, Appl. Catal., 1997, A164,91. R.D. Gonzalez, Appl. Surf. Sci., 1984,19, 181. V. Ponec, and G.C. Bond, ‘Catalysis by Metals and Alloys’, Stud. Surf. Sci. Catal., Elsevier, Amsterdam, 1995, vol. 95. F. Janssen, Thermochim. Acta, 1989,148, 137. S . Bhatia, J. Beltramini, and D.D. Do, Catal. Today, 1990,7,308. P.S. Braterman, ‘Metal Carbonyl Spectra’, Academic Press Inc., London, 1975. J. Ryczkowski, Catal. Today, 2001,68,263. R.P. Eischens, and W.A. Pliskin, Adv. Catal., 1958,10, 1. N. Sheppard, and C. De La Cruz, Catal. Today, 2001,70,3. G. Blyholder, J . Phys. Chem., 1964,68,2772. C. Furlani, J . Electr. Spectr. Rel. Phen., 1994,68, 569. J.J. Rehr, and A.L. Ankudinov, J . Electr. Spectr. Rel. Phen., 2001, 114-116, 1115. P. Johnston, and P.B. Wells, Rad. Phys. Chem., 1995,45,393. M.F. Garcia, Catal. Rev. Sci. Eng., 2002,44,59. Z. Karpinski, Adv. Catal., 1990,37,45. Yu. I. Yermakov, B.N. Kuznetsov, and V.A. Zakharov, in ‘Catalysis by Supported Complexes’, Stud. Surf: Sci. Catal., Elsevier, Amsterdam, 1981, vol. 8, p. 437-438. A. Boulahounsaturated aldehydesche, G. Kons, H. G., Lintz, and P. Schultz, Appl. Catal., 1992, A91, 115. Y. Pouilloux, F. Autin, C. Guimon, and J. Barrault, J . Catal., 1998,176,215. K. Liberkova, and R. Touroude, J . Mol. Chem., 2002,180,211. S. Szabo, Int. Rev. Phys. Chem., 1991,10,207. Gy. Horanyi, Catal. Today, 1994,19,285. D.I. Jerdev, and B.E. Koel, Surf: Sci., 2001,492,106. J.L. Margitfalvi, I. Borbath, M. Hegedhs, S. Gobolos, and F. Lonyi, React. Kinet. Catal. Lett., 1999,68, 133. J. Nakamura, T. Uchijima, Y. Kanai, and T. Fujitani, Catal. Today, 1996,28,23. A.N. Pestryakov, and V.V. Lunin, J . Mol. Catal., 2000, A158,325. M. Rarnning, T. Gjervan, R. Prestvik, D.G. Nicholson, and A. Holmen, J . Catal., 2001,204,292. V. Ponec, Applied Catal., 1997, A149,27. P. Gallezot, and D. Richard, Catal. Rev. Sci. Eng., 1998,40,81. P. Claus, Topics in Catal., 1998,551. R. Burch, J . Catal., 1981,71,348. N. Macleod, J.R. Fryer, and D. Stirling, G. Webb, Catal. Today, 1998,46,37. S.R. Adkins, and B.H. Davis, J. Catal., 1984,89,371. K. Balakrishnan, and J. Schwank, J . Catal., 1991,127,287. R. Srinivasan, and B.H. Davis, J . Catal., 1992,87,45. J . Llorca, N. Homs, J.L.G. Fierro, J. Sales, and P. Ramirez de la Piscina, J . Catal., 1997,166,44. Yu.1. Yermakov, B.N. Kuznetsov, and V.A. Zakharov, ‘Catalysis by Supported Complexes’, Stud. Sur$ Sci. Catal., Elsevier, Amsterdam, 1981, vol. 8, p. 351-363. V.I. Kuznetsov, A.S. Belyi, E.N. Yurchenko, M.D. Smolikov, M.T. Protasova, E.V. Zatolokina, and V.K. Duplyakin, J . Catal., 1986,99, 159. J. Llorca, P. Ramirez de la Piscina, J.L. Garcia-Fierro, J. Sales, and N. Homs, J . Catal., 1995,156, 139. R. Srinivasan, R.J. De Angelis, and B.H. Davis, J . Catal., 1987,106,449. C. Kappenstein, M. Guerin, K. Lazar, K. Matusek, and Z. Paal, J . Chem. SOC.,

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

99

Faraday Trans., 1998,94,2463. 95. R. Gbmez, V. Bertin, T. Lopez, I. Schifter, and G. Ferrat, J . Mol. Catal., 1996, A109, 55. 96. Yong-Xi Li and K.J. Klabunde, J . Catal., 1990,126, 173. 97. T. Inoue, K. Tomishige and Y. Iwasawa, J . Chem. SOC.,Faraday Trans., 1996,92, 461. 98. J. Margitfalvi, M. Hegediis, S. Giibolos, E. Kern-Talas, P. Szedlacsek, S. Szabo, and F. Nagy, in ‘Proceedings of the 8th International Congress on Catalysis’, ed. G. Ertl, Verlag Chemie, Weinheim, 1984, vol. 4, p. 903-914. 99. J. Margitfalvi, S. Szab6 and F. Nagy, in ‘Catalytic hydrogenation’ ed. L. Cerveny, , Stud. Surf: Sci. Catal., Elsevier, Amsterdam, 1986, vol. 27, p. 373-409. 100. Ch. Travers, J.P. Bournonville, and G. Martino, in: Proceedings of the 8th International Congress on Catalysis, ed. G. Ertl, Verlag Chemie, Weinheim, 1984, vol. 4, p. 89 1-902. 101. J. P. Candy, B. Didillon, E. L. Smith, T. B. Shay and J. M. Basset, J . Mol. Catal., 1994,86, 179. 102. V. D. Stytsenko, Appl. Catal., 1995,126,l. 103. J.H. Sinfelt, Surf: Sci., 2002,500, 923. 104. J.L. Margitfalvi, E. Talas, and S. Gbbolos, Catal. Today, 1989,6,73. 105. Cs. Vkrtes, E. Talas, I. Czako-Nagy, J. Ryczkovski, S. Gobolos, A. Vertes, and J. Margitfalvi, Appl. Catal., 1991,68, 149. 106. J.L. Margitfalvi, I. Borbath, E. Tfirst, and A. Tompos, Catal. Today, 1998,43,29. 107. J.L. Margitfalvi, I. Borbath, M. Hegediis, and A. Tompos, Appl. Catal., 2002, A29, 35. 108. J.L. Margitfalvi, I. Borbath, M. Hegediis, S. Gbbolos, A.Tompos, and F. Lonyi, in ‘Proceedingsof 12th International Congress on Catalysis’,ed. A.Corma, F.V. Melo, S. Mendioroz and J.L.G.Fien-o, Stud. Surf: Sci. Cat& Elsevier, Amsterdam, 2000, V O ~ .130, p. 1025-1030. 109. J.L. Margitfalvi, Gy. Vanko, I. Borbath, A. Tompos, and A. Vertes, J.Catal., 2000, 190,474. 110. V.I. Kuznetsov, A.S. Belyi, E.N. Yurchenko, M.D. Smolikov, M.T. Protasova, E.V. Zatolokina, and V.K. Duplyakin, J . Catal., 1986.99,159. 111. J.L. Margitfalvi, I. Borbath, and, A. Tompos in ‘Catalysisof Organic Reactions’, ed. F. E. Herkes, Chem. Ind., Marcel Dekker, New York, 2002, vol. 75, p. 243-254. 112. P.N. Rylander, Catalytuc Hydrogenation in Organic Synthesis, Academic Press, New York, 1979. 113. P.N. Rylander, Hydrogenation Methods, Academic Press, New York, 1985. 114. W.H. Carothers, and R. Adams, J . Am. Chem. SOC., 1924,46,1675. 115. R. Adams, and B.S. Garvey, J . Am Chem. SOC., 1926,48,477. 116. W.F. Tuley, and R. Adams, J . Am Chem. SOC. 1925,47,3061. 117. P.N. Rylander, N. Himelstein, and M. Kilroy, Engelhard Ind. Tech. Bull., 1993, 4, 49. 118. P. Claus, Chem.-1ng.-Techn., 1995,67, 1340.. 119. G.F. Santori, M.L. Casella, and O.A. Ferretti, J . Mol. Catal., 2002, A M , 431. 120. O.A. Ferretti, B. de Pauli, J.P. Candy, G. Mabilon, and J.P. Bournonville, in ‘Preparation of Catalysts IVY,ed. B. Delmon, P. Grange, P.A. Jacobs and G. Poncelet, Stud. Sur$ Sci. Catal., Elsevier, Amsterdam, 1987, vol. 31, p 713-724. 121. J.L. Margitfalvi, A. Tompos, I. Kolosova, and J. Valyon, J . Catal., 1998,174,246. 122. P. Sautet, Topics in Catal., 2000,13,213. 123. G.S. Galvagno, A. Donato, G. Neri, R. Pietropaolo, and D. Pietropaolo, J .

100

Catalysis

Mol.Catal., 1989,49,23. 124. T.B.L.W. Marinelli, and V. Ponec, J.Catal., 1995,156,51. 125. J.L. Margitfalvi, I Borbath, and A Tompos, in ‘Preparation of Catalysts VII’, ed. B. Delmon, P. A. Jacobs, R. Maggi, J.A. Martens, P. Grange, and G. Poncelet, Stud. Surf:Sci. Catal., Elsevier, Amsterdam, 1999, vol. 118, p 195-204. 126. T.B.L. Martinelli, S, Nabuurs, and V. Ponec, J . Catal., 1995,151,431. 127. P.G. Harrison, and J.J. Zuckerman, J . Am. Chem. Soc., 1969,91,6885. 128. S. Galvagno, Z. Potarzewski, A. Donato, G. Neri, and R. Pietropaolo, J . Mol. Catal., 1986,35, 365. 129. G. Cocco, S. Enzo, S . Galvagno, Z. Poltarzewski, and R. Pietropaolo, J . Chem. Soc., Farad. Trans. I , 1985,81,321. 130. S. Giibolos, N. Mahata, I. Borbath, M. Hegedhs, and J.L. Margitfalvi, React. Kinet. Catal. Lett., 2001,74, 345. 131. G.D. Yadav, and M.R. Kharkara, Appl. Catal., A, 1995,126,115. 132. J. Sarkany, M. Bartok, and R. Gonzales, J . Catal., 1983,81, 347. 133. Y.F. Yu Yao, J . Catal., 1984,87,152. 134. G.C. Bond, L.R. Molloy, and M.J. Fuller, J.C.S. Chem. Comm., 1975,796. 135. M. Sheintuch, J. Schmidt, Y. Lechtman, and G. Yahav, Appl. Catal., 1989,49,55. 136. S. Gardner, G.B. Hoflund, M.R. Davidson, and D.R. Schryer, J . Catal., 1989,115, 132. 137. N.D. Gangel, M. Gupta, and R.M. Iyer, J . Catal., 1990,126,13. 138. K. Grass, and H.G. Linz, J . Catal., 1997,172,446. 139. J.L. Margitfalvi, I.Borbath, K. Lazar, E. Tfirst, A. Szegedi, M. Hegedhs, and S. Giibolos, J . Catal., 2001,203,94. 140. M.C. Hobson, S.L. Goresh, and G.P. Khare, J . Catal., 1993,142,641. 141. R.D. Cortright, and D.A. Dumesic, J.Catal., 1994,148,771. 142. L. Stievino, F.E. Wagner, S. Calogero, S. Recchia, D. Dossi, and R. Psaro, ‘Proceedings of the 12th International Congress on Catalysis’, ed. A. Corma, F.V., Melo, s. Mendioraz, and J.L.G. Fierro, Stud. Surj Sci. Catal., 2000, vol. 130, p. 3903-3908. 143. V.I. Kuznetsov, A.S. Belyi, E.N. Yurchenko, D.M. Smolikov, M.T. Protasova, E.V. Zatolokina, and V.K. Duplyakin, J. Catal., 1986,99, 159. 144. K. Lazar, A.M. Szeleczky,N.K. Mal, and A.V. Ramaswamy, Zeolites, 1997,19,123. 145. N. Sheppard, and T.T. Nguyen, in: Advances in Infrared and Roman Spectroscopy, ed. R.G.H. Clark, R.E. Hester, (Heyden, London,) 1978, vo1.5, p. 67-148. 146. M.J. Kappers, and J.H. van der Mass, Catal. Lett., 1991, 10,365. 147. L.C. Menorval, A. Chaqroune, B. Coq, and F. Figueras, J . Chem. Soc., Faraday Trans., 1997,93, 3715. 148. A.G.T.M. Bateein, F.J.C.M. Toolenaar, and V. Ponec, J . Catal., 1984,90,88. 149. G.J. Artega, J.A. Anderson, and C.H. Rochester, J. Catal., 1999,184,268. 150. J.L. Margitfalvi, I. Borbath, M. Hegedhs, A. Szegedi, K. Lazar, S. Giibolos, and S. Kristyan, Catal. Today, 2002,73, 343. 151 H. Sakuri, A.Ueda, T. Kobayashi, and M. Haruta, J . Chem. Soc. Chem. Commun., 1977,91. 152. M. Haruta , Catal. Today, 1977,36,153. 153. G.C. Bond, and D.T. Thompson, Catal. Rev. Sci. Eng., 1999,41,319. 154. G.C. Bond, and D.T. Thompson, Gold Bull., 2000,33,41. 155. M. Haruta, and M., Date, Applied Catal., 2001, A2,427. 156. M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M.J. Genet, and B. Delmon, J . Catal., 1993,144, 175. 157. M. Haruta, Catal. Today, 1997,36,153.

1: Role of ‘Metal Ion-Metal Nunocluster’ Ensemble Sites

158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170. 171.

172. 173. 174. 175.

176. 177. 178. 179. 180. 181. 182. 183. 184. 185. 186. 187. 188. 189.

101

M. Valden, S. Pak, X. Lai, and D.W. Goodman, Catal. Lett., 1998,56,7. D.A.H., Cunningham, W. Vogel, and M. Haruta, Catal. Lett., 1999,63,43. M. Haruta, N. Yamada, T. Kobayashi, and S. J. Iijima, J . Catal., 1989,115,301. J.D. Grunwaldt, M. Maciejewski, O.S. Becker, P. Fabrizioli, and A. Baiker, J . Catal., 1999,186,458. Y. Z. Yunsaturated aldehydesn, A.P. Kozlova, K. Asakura, H. Wan, K. Tsai, and Y. Iwasawa, J.Catal., 1997.170,191. M.M. Schubert, S. Hackenberg, A.C. van Veen, M. Muhler, V. Plzak, and R.J. Behm, J . Catal., 2001,197,113. M. Haruta, Catal. Surv. Jpn., 1997,1,61. M. Valden, S. Pak, X. Lai, and D. Goodman, Catal. Lett., 1998,56,7. J.D. Grunwaldt, A. Baiker. J. Phys. Chem. B, 103 (1999) 1002. M.A. Bollinger, and M.A. Vannice, Appl. Catal., 1996, B8,417. A.K. Tripathi, V.S. Kamble, and N.M. Gupta, J . Catal., 1999,187,332. A.I. Kozlov, A.P. Kozlova, H. Liu, and Y. Iwasawa, Appl. Catal., 1999, A182,9. K. Fukushima, G.H. Takaoka, J. Matsuo, and I. Yamada, Jpn. J . Appl. Phys. 1997, 36,813. C.K. Costello, M. C. Kung, H.-S. Oh, Y. Wang, and H. H. Kung, Appl. Catal., 2002, A232,159. Z . Hao, L. An, H. Wang, T. Hu, React. Kinet. Catal. Lett., 2000,70, 153. S. Minico, S. Sciere, C. Crisafulli, A.M. Visco, S. Galvagno, Catal. Lett., 1997, 47, 273. J.L. Margitfalvi, A. Fasi, M. Hegedus, F. Lonyi, S. Gobolos, and N.Bogdanchikova, Catal. Today,2002,72, 157. A.N. Pestryakov, V.V. Lunin, A.N. Kharlanov, D.I. Kochubey, N. Bogdanchikova, and A.Yu. Stakheev, J . Mol. Struct., 2002,642, 129. R.J.H. Grisel, and B.E. Nieuwenhuys, J . Catal., 2001,199,48. F. Boccuzzi, A. Chiorino, M. Manzoli, D. Andreeva, and J. Tabakova, J . Catal., 1999,188,176. D.A.H. Cunningham, W. Vogel, H. Kageyama, S. Tsubota, and M. Haruta, J . Catal., 1998,177, 1. D.A.H. Cunningham, W. Vogel, R.M.T. Sanchez, K. Tanaka, and M. Haruta, J . Catal., 1999, 183, 183. V.M. Deshpande, K. Ramnarayan, and C.S. Narasimhan, J . Catal., 1990,121, 174. C.S. Narasimhan, V.W. Deshpande, and K.J. Ramnarayan, J . Chem. SOC., Chem. Commun., 1988,99. S . Galvagno, C. Milone, A. Donato, G.R. Neri, and R. Pietropaolo, Catal. Lett., 1993,17, 55. M.J. Mendes, O.A.A. Santos, E. JordBo, and A.M. Silva, Appl. Catal., 2001, A217, 253, and references therein. G.G. Raab, and J.A. Lercher, Catal. Lett., 1993,18,99. G.C. Bond, Appl. Catal., 1997, A157,91. Y.-X. Li, K.J. Klabunde, and B.H. Davis, J . Catal., 1991, 128, 1. L. Perring, P. Feschotte, F. Bussy, and J.C. Gachon, J . Alloys Compd., 1996, 245, 157. M. J. Chollier-Brym, M. Fontaine, L. Rou, D. Gunsaturated aldehydesy, L. Autissier, and J. Barbier, Abstract, 17th North American Catalysis Society Meeting, Toronto, Canada, June 3-8,2001, p. 219. S. Galvagno, C. Milone, A. Donato, G.R. Neri, and R. Pietropaolo, Catal. Lett., 1993, 18,349.

102

Catalysis

190. L. Stievano, S. Calogero, F.E. Wagner, S. Galvagno, and C. Milone, J. Phys. Chem. B, 1999,103,9545. 191. T.S. Tang, K.Y. Cheah, F. Mizukami, S.I. Niwa, and M. Toba, J. Amer. Oil Chem. Soc., 1994,71,501. 192. K.Y. Cheah, T.S. Tang, F. Mizukami, S.I. Niwa, M. Toba, and Y.M. Choo, J. Amer. Oil Chem. Soc., 1992,69,410. 193. S . Gbbolos, N. Mahata, and J.L. Margitfalvi, in ‘Principles and Methods for Accelerated Catalyst Design and Testing’, ed. E.G. Derouane, V. Parmon, Kluwer Academic Publishers, 2002, p. 323-332. 194. S. Gbbolos, I. Sajo, N. Mahata, and J.L. Margitfalvi, submitted to Catal. Lett. 195. P. Betancourt, A. Rives, R. Hubaut, C.E. Scott, and J. Goldwasser, Appl. Catal, 1998, A170,307. 196. G.M.R. van Druten, and V. Ponec, Appl. Catal., 2000, A191,163. 197. K. Tahara, H. Tsuji, H. Kimura, T. Okazaki, Y. Itoi, S. Nishiyama, S. Tsuruya, and M. Masai, Catal. Today, 1996,28,267. 198. B. Bachiller-Baeza, A. Guerrero-Ruiz, and I. Rodrigez-Ramos, Appl. Catal., 2000, A192,289. 199. H.E. Kluksdahl, Chevron, US Patent 3 415 737,1968. 200. N. Macleod, J.R. Fryer, D. Stirling, and G. Webb, Catal. Today 1998,46, 37, and references therein. 201. H.S. Broadbent, G.C. Campbell, W.J. Bartley, and J.H. Johnson, J. Org. Chem., 1959,24,1847. 202. H.S. Broadbent, and T.G. Selin, J. Org. Chem., 1963,28,2343. 203. H.S. Broadbent, and W.J. Bartley, J. Org. Chem., 1963,28,2345. 204. H.S. Broadbent, and J. H. Johnson, J. Org. Chem., 1962,27,4400. 205. H.S. Broadbent, and D.W. Seegmiller, J. Org. Chem., 1963,28,2347. 206. C. Hirosawa, N. Wakasa, and T. Fuchikami, Tetrahedron Lett., 1996,37,6749. 207. K. Tahara, E. Nagahara, Y. Itoi, S. Nishiyama, S. Tsuruya, and M. Masai, React. Kinet. Catal. Lett., 1996,59, 15. 208. I.D. Dobson, BP Chemicals Limited, EP 0 286 280 Al, 1988. 209. T. Fuchikami, T. Hirosawa, and N. Wakasa, Jpn. Kokai Tokkyo Koho J P 09 241 22, 1996. 210. S.W. King, Union Carbide Chemicals & Plastics Technology Corporation, USA, Jpn. Kokai Tokkyo Koho JP 08 319 261,1995. 21 1. D. Chang, and F.A. Sherrod, Dow Chemical Company, PCT Int. Appl. WO 96 38 26, 1997. 212. T. Fuchikami, T. Ga, N. Wakasa, and K. Tawara, Jpn. Kokai Tokkyo Koho, J P 09 132 541,1997 213. T.S. Tang, K.Y. Cheah, F. Mizukami, S. Niwa, M. Toba, and Y.M. Choo, J. Am. Oil Chem. Soc., 1993,70,601. 214. V. Pallassana, and M. Neurock, J. Catal., 2002,209,289. 215. Du Hongzhang, Wu Rongan, Fang Lianging, Wang Haizen, and Lin Liwu, Appl. Catal., 1991,78, 1. 216. T. Gjervan, R. Prestvik, B. T-tdal,C.E. Lyman, and A. Holmen, Catal. Today, 2001, 65, 163. 217. A.S. Fung, M.R. McDevitt, P.A. Tooley, M.J. Kelley, D.C. Koningsberger, and B.C. Gates, J. Catal., 1993,140, 190, and references therein. 218. R. Prestvik, K. Moljord, K. Grande, and A. Holmen, J. Catal., 1998,174, 119, and references therein. 219. B.H. Isaacs, and E.E. Petersen, J. Catal., 1982,77,43.

1 : Role of ‘Metal Ion-Metal Nanocluster’ Ensemble Sites

103

220. S. Gbbolos, M. Hegedb, I. Borbath, and J.L. Margitfalvi, Appl. Catal. A (submitted) 221. J.L. Margitfalvi, I. Borbath, M. Hegediis, and S. Gbbolos, Appl. Catal., 2001,219, 171. 222. C. Boelhouwer, J. van Mourik, and H. I. Waterman, Chemie und Industrie, 1960,83, 875. 223. J.D. Richter, and P.J. van den Berg, J . Amer. Oil Chem. SOC.,1969,46,158. 224. H. Adkins, and K. Folkers, J . Am. Chem. SOC.,1931,53,1095. 225. J. Sauer, and H. Adkins, J . Am. Chem. SOC.,1937,59, 1. 226. U.R. Keutzer, J . Am. Oil Chem. SOC.,1984,61,343. 227. F.J. Carduck, J.F. Falbe, T. Ceckenstein, and P. Joachim, Ger. Offen, DE 3 624 812, 1988. 228. B. Miya, B. Hoshino, and M. Matsuda, Japanese Pat. 74 29 819,1974. 229. D.S. Thakur, E. Palka, T. J. Sullivan, E. Nebesh, and B.D. Roberts, U.S. Pat. 5 134 108,1992. 230. A. Krause, Sci. Pharm., 1971,39, 179. 231. R. Hubaut, J.P. Bonnelle, and M. Daage, J . Mol. Catal., 1989,55, 170. 232. S. Gobolos, N. Mahata, I. Sajo, I. Borbath, M. Hegediis, and J.L. Margitfalvi, Appl. Catal. A (submitted). 233. H.F. Chang, M.A. Saleque, W.S. Hsu, and W.H. Lin, J . Mol. Catal., 1996,109,249. 234. S. Fujita, S. Moribe, Y. Kanamori, M. Kakudate, and N. Takezawa, Appl. Catal., 2001, A207,121. 235. C . Fragale, M. Gargano, and M. Rossi, J . Am. Oil Chem. SOC.,1982,59,465. 236. F. M. Capece, V. Di Castro, C. Furlani, G. Mattogno, C. Fragale, M. Gargano, and M. Rossi, J . Electr. Spectr. Rel. Phenomena 1982,27, 119. 237. L. Ma, B. Gong, T. Tran, and M.S. Wainwright, Catal. Today, 2000,63,499. 238. P. Claus, M. Lucas, B. Liicke, T. Berndt, and P. Birke, Appl. Catal., 1991, A79, 1. 239. D.S. Brands, E.K. Poels, and A. Bliek, Appl. Catal., 1999, A184,279. 240. F.T. van de Scheur, B. van de Linden, M. Mittelmejer-Hazeleger, J.G. Nazloomian, and L.H. Staal, Appl. Catal., 1994, A l l l , 63. 241. R. Hubaut, M. Dage, and J. P. Bonnelle, Appl. Catal., 1986,2,231. 242. T. Fujitani, M. Saito, Y. Kanai, T. Watanabe, and T. Uchijima, Chem. Lett., 1994, 1877. 243. T.M. Yurieva, I.M. Plyasova, O.V. Makarova, and T.A. Krieger, J . Mol. Catal., 1996, A113,455. 244. S. Gobolos, M. Hegedus, I. Kolosova, M. Maciejewski, and J.L. Margitfalvi, Appl. Catal., 1998, A169,201. 245. V. Z. Fridman, and A.A. Davydov, J . Catal., 2000,198,20. 246. A. Baiker, and J. Kijenski, Catal. Reu. Sci. Eng., 1985,27,653. 247. A. Baiker, W. Caprez, and W.L. Holstein, Ind. Eng. Chem. Prod. Res. Den, 1983,2, 217. 248. Y. Tsuji, R. Takeuchi, H. Ogawa, and Y. Watanabe, Chem. Lett., 1986,293. 249. J. Barrault, N. Essayem, and C. Guimon, Appl. Catal., 1993, A102,151. 250. H. Abe, K. Hashiba, Y. Yokota, and K. Okabe, Yukagaku, 1988,37,519. 251. K. Okabe, H. Abe, and Y. Yokota, in ‘Successful Design of Catalysts’ ed. T. Inui, Stud. Surf: Sci. Catal., Elsevier, Amsterdam, 1989, vol. 44, p. 299-306. 252. C.J. Jiang, D.L. Trimm, M.S. Wainwright, and N.W. Cant, Appl. Catal., 1993, A97, 145. 253. J.B. Bulko, R.G. Herman. K. Klier, and G.W. Simmons, J . Phys. Chem., 1979,83, 3118. 254. Y.W. Suh, S.H. Moon, and H.K. Rhee, Catal. Today, 2000,63,447.

104

Catalysis

255. Y. Okamoto, K. Fukino, T. Imanaka, and S. Teranishi, J. Phys. Chem., 1983, 87, 3747. 256. S. Gusi, F. Trifiro, A. Vaccari, and G. Del Piero, J. Catal., 1985,94, 120. 257. H.Y. Chen, S.P. Lau, L. Chen, J. Lin, C.H. Huan, K.L.Tan, and J.S. Pan, Appl. Suf: Sci., 1999, 152, 193. 258. K. Klier, Adv. Catal. Related Subjects, 1982,31,243. 259. J.R. Monnier, G. Apai, and M.J. Hanrahan, J. Catal., 1984,88, 523. 260. J. Nakamura, T. Uchijima, Y. Kanai, and T. Fujitani, Catal. Today, 1996,28,23. 261. S. Bailey, G. Froment, J.W. Snoeck, and K.C. Waugh, Catal. Lett., 1995,30,99. 262. K. Klier, R.G. Herman, and G.A. Vedage, Leigh Univ., Betlehem, Pa, US. Pat. 4 480 131,1984., EP 0 127 874 A2,1984. 263. L.V. Vytnova, and A.N. Bashkirov, Kinet. Katal., 1978,19,615. 264. J.W. Evans M.S. Wainwright, A.J. Bridgewater, and D.J. Young, Appl. Catal., 1983, 7,75. 265. M.J.L. Gines, and C.R. Apesteguia, Lat. Am. Appl. Res., 1995,25,215. 266. Y. Shiraishi, and N. Toshima, Colloids and Surfaces A: Physicochem. Eng. Aspects, 2000,169,59. 267. A.N. Pestryakov, and A.A. Davydov, Kinet. Catal., 1994,35,273. 268. J.-L. Li, W.-L. Dai, Y. Dong, and J.-F. Deng, Materials Lett., 2000,44,233. 269. K.N. Fan, B.R. Shen, W.N. Wang, and J.F. Deng, J. Mol. Struct. (Theochem),1998, 42, 191. 270. K.N. Fan, B.R. Shen, W.N. Wang, and J.F. Deng, J. Mol. Struct. (Theochem),1999, 469, 157. 271. K.De Oliveira, Y. Pouilloux, and J. Barrault, J. Catal., 2001,204,230. 272. P. Reyes, H. Roja, G. Pecchi, and J.L.G. Fierro, J. Mol. Catal., 2002, A179,293. 273. J.N. Coupe, E. JordBo, M.A. Fraga, and M.J. Mendes, Appl. Catal., 2000, A199,45. 274. P. Reyes, M.C. Aguirre, J.L.G. Fierro, G. Santori, and 0. Ferretti, J. Mol. Catal., 2002, A184,431. 275. S . Nishiyama, T. Hara, S. Tsuruya, and M. Masai, J. Phys. Chem. B, 1999,103,4431. 276. G. Neri, C. Milone, A. Donato, L. Mercadante, and A.M. Visco, J. Chem. Tech. Biotechnol., 1994,60, 83. 277. R.V. Malya, C.V. Rode, M. Arai, S.G. Hedge, and R.V. Chaudari, Appl. Catal., 2000, A193, 71. 278. M. Casagrande, L. Storano, A. Talon, M. Lenarda, R. Frattini, E.R. Castellon, and P.M. Torres, J. Mol. Catal., 2002, A188, 133. 279. M. Ichikawa, Polyhedron, 1988,7,2351. 280. A. Benedetti, G. Fagherazzi, F. Pinna, G. Rampazzo, , M. Selva, and G. Strukul, Catal. Lett., 1991, 10, 215. 281. H.R. Aduriz, P. Bodnariuk, B. Coq and F. Figueras, J. Catal., 1989,119,97. 282. J. Margitfalvi, S. Gbbolos, E. Talas, and M. Hegedbs, in ‘Preparation of Catalysts V’, ed. G. Poncelet, P. A. Jacobs, P. Grange and B. Delmon, Stud. Surf. Sci. Catal., vol. 63, Elsevier, Amsterdam, 1991, p. 669-678.

2 The Destruction Of Volatile Organic Compounds By Heterogeneous Catalytic Oxidation BY CATHERINE S. HENEGHAN, GRAHAM J. HUTCHINGS AND STUART H. TAYLOR

1

Introduction

The significance of abatement technologies for the removal of volatile organic compounds (VOCs) from industrial effluents has increased in importance with the introduction of legislation to control their release to the environment. Various techniques have been proposed, one of these being heterogeneous catalytic oxidation to carbon oxides and water. This has the advantage over the more common thermal oxidation process, since it requires little or no supplementary fuel and is therefore a less expensive process. Catalytic oxidation is less expensive only for relatively dilute streams of VOC-containing gases. For more concentrated gas streames, thermal oxidisers, which are less capital-intensive, may be less costly to operate if the VOC content is high enough to be thermally self-sustaining. However, relatively dilute VOC-containing effluent streams are by far the most prevalent and, consequently, catalytic oxidation is an important process. However, the characteristics of the catalyst selected for this process are of vital importance for successful process operation, and associated problems such as deactivation must be overcome if heterogeneous catalytic oxidation is to be useful commercially. Catalysts currently in use include noble metals, notably platinum and palladium, and those based on metal oxides. However, irrespective of the type of catalyst, the prime characteristics required are activity and selectivity for combustion. In this review the reactivities of a number of catalysts for the total oxidation of compounds that may be classified as VOCs are compared, and the important features of the catalyst highlighted using oxygen as oxidant. The following is not intended to be an exhaustive account of all possible catalysts, but should provide an overview of the current state of research in this area, with the aim of identifying the types of catalysts that are likely to be of use in the future, and the obstacles that must be overcome to produce a commercially viable catalyst. The development of a catalyst that may be used for the combustion of all classes of compounds under the general term VOC presents a major challenge for future research. Catalysis, Volume 17 0 The Royal Society of Chemistry, 2004 105

106

2

Catalysis

VOC Abatement

The term VOC refers to a wide-ranging class of compounds which have been given the following definition by the US Environmental Protection Agency: ' . . . Any compound of carbon, excluding carbon monoxide, carbon dioxide, carbonic acid, metal carbides or carbonates and ammonium carbonates which participates in atmospheric photochemical reactions . . . .' [11 Any organic compound with vapour pressure exceeding 0.1 mmHg under standard conditions (25°C and 760 mmHg) may therefore be regarded as a VOC. VOCs are commonly found in industrial effluents as by-products or as unused reactants, and their release to the environment has widespread environmental implications. Pollution by VOCs has been linked to the increase in photochemical smog [2], and to ozone depletion [3]. In addition, many VOCs are themselves toxic and/or carcinogenic. The US Clean Air Act (1990) [4] calls for a 90% reduction in the emissions of 189 toxic chemicals by 1998, 70% of which are classed as VOCs. Abatement technologies to control the release of VOCs to the environment are therefore of paramount importance. A variety of methods have been proposed for the removal of VOCs in industrial waste streams. Two main approaches to the problem exist, VOCs are either captured for reuse and subsequent disposal, as in adsorption, absorption and condensation processes, or they are destroyed completely, forming water, carbon oxides and, for halogenated VOCs, hydrogen halides. Carbon oxides may be released to the environment with minimal environmental implications compared to the release of the VOCs themselves. Hydrogen halides are the preferred product, as they may be more easily removed from the outlet stream, generally by aqueous scrubbers, than the halogens themselves. Catalytic oxidation and thermal oxidation are examples of the latter treatment route. The former approach enables recycling of waste reaction products and/or unused reactants, but the removal of volatile organics from the system for safe disposal will ultimately be required. In the longer term, it is possible that many organic solvents used in reactions may be replaced by water, or non-solvent processes may be developed. This may obviate the need for VOC oxidation in many cases. 3

Operational Parameters Affecting the Catalytic Combustion of VOCs

As previously established, the catalytic oxidation of volatile organics to carbon dioxide and water offers significant advantages over all other current VOC abatement methods, as it enables complete destruction at relatively low temperatures with high volumetric throughput. A variety of parameters, other than the nature of the catalyst, influence this process. These include operational temperature, pre-heating of the system, space velocity, the oxidant used for the combustion, the nature and concentration of VOC(s)in the feed gas, and the deactivation of the catalyst.

3.1

Temperature. - In general, the temperature required for the complete cata-

107

2: The Destruction Of Volatile Organic Compounds

lytic combustion of a VOC cannot be used independently as a control factor, as it varies according to the VOCs present, their concentrations, and the catalyst used. The use of high temperatures would be expected to increase the efficiency of destruction of VOCs, but can also accelerate catalyst deactivation, thus reducing activity. Industrially, relatively low temperatures would be preferred to reduce the operating costs.

3.2 System Preheating.- The thermal efficiency of a catalytic oxidation system may be enhanced by preheating the feed gas in air prior to catalytic combustion. Tichenor and Palazzolo [ 5 ] have determined the relative contribution of the pre-heater to the overall efficiency. A mixture of iso-propanol, methyl ethyl ketone, ethyl acetate, benzene and n-hexane was combusted at a space velocity of 50,000h-' in the temperature range 300-450°C over a bi-metallic Pt-Pd catalyst supported on a ceramic monolith. The results are shown in figure 1. It can be seen that the pre-heating stage contributes significantly to the efficiency of the system, particularly at higher temperatures. Pre-heated VOCs may act as a source of fuel for catalytic oxidation, reducing the overall fuel requirement necessary to sustain the combustion. In addition, the products of thermal oxidation may influence the activity of the catalyst. This has been illustrated by Zieba [ 6 ] , who indicated that thermal oxidation during preheating resulted in 520% oxidation of the effluent. The effects of pre-heating were investigated for the oxidation of ethene, methane and toluene, over two industrial combustion catalysts (0.1% Pt/alumina and copper/cobalt/manganese oxides supported on alumina) at a space velocity of 20,000 h-I, with temperatures in the range 77-477°C. The results for the system incorporating pre-heating of the VOCs were compared to those for a conventional system with no pre-heating. The system with pre-heating showed increased efficiency by 5-3070 compared to the conventional system. It was proposed that thermal oxidation was acting as a source of radicals, consisting hydrogen, oxygen, hydroxyl and organic, enhancing the production of radicals in the subsequent l00r

e 8o

1

0" , _ . . . . . . . . . . . .5.. . ....

..........

..o.. Heater

60-

300

Figure 1

system --*-

Catalyst

.*E w

0"

..B.

350 400 Inlet Temperature / OC

450

Relative contribution of heater and catalyst to system destruction efJiciencyfor bimetallic Pt-Pd catalyst supported on a ceramic monolithfor the combustion of an iso-propanol, methyl ethyl ketone, ethyl acetate, benzene and n-hexane mixture at GHSV=50,000 h-' and T = 300-450°C [ 5 ]

108

Catalysis

catalytic oxidation step, thus increasing destruction efficiency. The production of radicals in thermal oxidation has been previously reported C7-81 and their importance in gas phase oxidation reactions well established [9]. Supplementary fuel provided to sustain combustion was also found to act as a source of both thermal energy and of radicals. As the effluent gas flows through this highly active region, combustion is initiated, thus destruction begins before the VOCs reach the catalyst bed, with a resultant increase in activity. It should be noted, however, that water vapour may also be formed in the pre-heating thermal oxidation step, and may act to inhibit catalytic oxidation. The addition of 5% water vapour to the effluent gas resulted in a 5-10% reduction in activity for the catalytic toluene oxidation studied by Zieba, presumably because water was being adsorbed at the catalytically active centres and thus blocking them from participating in the oxidation. It was suggested that water produced in thermal oxidation would act similarly to depress conversion.

3.3 Space Velocity. - Space velocity has significant effects on destruction efficiency; as space velocity increases, destruction efficiency generally decreases, as may be expected for normal behaviour in heterogeneously catalysed reactions. Commercially, a catalyst capable of achieving high levels of destruction at relatively high space velocities, with no reduction in specificity towards total oxidation products, would be preferred. This will allow the catalyst to be used in simple 'end of pipe' applications, and will also reduce the amount of catalyst required to achieve complete destruction, thus reducing capital costs. The effects of increasing space velocity have been demonstrated by Vassileva and co-workers [lo]. The combustion of benzene over a 0.5 wt% Pd/300/, V205/A1203 catalyst for space velocities of 330,2,000 and 5,000 h-' at a constant oxygen to benzene molar ratio of ca. 7.5. The effects of varying space velocity are shown in figure 2. However, it should be noted that these data indicate that there may be an induction period for this catalyst, which is most marked at the lowest space velocity investigated. It can be seen that, the higher space velocities have the highest initial activities, but conversion soon reaches a constant level of less than 100%. However, the activity at a space velocity of 330h-' rapidly increases with time and becomes constant at a significantly higher conversion close to 100%. In the region in which catalytic activity is constant for each space velocity, it can be seen that activity for combustion is markedly decreased with increased space velocity. 3.4 Type of VOC. - Individual VOCs are combusted at a specific temperature according to their chemical composition, type of catalyst and the reaction conditions used. The ease of destruction of VOCs by catalytic combustion can generally be correlated to the chemical class of the compound, such that a general order for the ease of destruction may be observed. For example, Tichenor and Palazzolo [5] determined such an order for a Pt/Pd bi-metallic catalyst on a ceramic honeycomb. The inlet temperature and space velocity of the system were varied in the ranges 260-425°C and 15,OOO-80,000h-', to give 98-99% conversion of the hydrocarbons, from which the following ranking according to com-

109

2: The Destruction Of Volatile Organic Compounds

I SV = 330 h-' I SV=2000h-' II SV=5000h-' III

0

m

70

I

I

I

1

I

I

I

1

1

2

3

4

5

6

7

0.04 0.08 0.235 0.470 0.6 1.2

Figure 2

0.12 0.16 0.20 0.705 0.940 1.175 1.8 1.2.4 3.0

0.24 0.28 1.410 1.645 3.6 4.2

1

I

Timeonstream/h I

Benzenepassed

II through the III catalyst

Efects of increased space velocity on benzene combustion over 0.5 wt% Pd/30% V,O, / A1,0,; data from [ l o ]

pound class was obtained: alcohols > aldehydes > aromatics > ketones > acetates > alkanes > chlorinated hydrocarbons

All compounds were combusted with the required efficiency, with the exception of chlorinated hydrocarbons, which were seen to partially deactivate the catalyst. Chlorinated hydrocarbons are frequently difficult to destroy, with both chlorinated reagents and products acting as catalyst poisons and thus causing catalyst deactivation, resulting in a decrease in activity. A similar situation is seen for fluorinated VOCs. Comparable orders of ease of combustion may be obtained for other catalysts used in VOC abatement, activities vary according to the stability of the class of compound and the ability of the compound and/or its oxidised products to act as catalyst poisons. Specific compounds within these general classes may have higher or lower destruction efficiencies depending on their exact nature and on the composition of the catalyst used for the combustion. VOC Mixtures. - Industrial gas effluents rarely consist of a single VOC, hence it is essential to determine the effects caused by any interaction between components of a VOC mixture. The effects of using different mixture compositions are rarely predictable, although they are often significant. If there is no interaction between the components of a mixture, it would be expected that the mixture would be combusted with a similar efficiency to the pure components. This is rarely seen. For example, over a Pt/Pd catalyst at 305"C,90% conversion of hexane is observed, whereas only 75% conversion is seen under the same 3.5

110

Catalysis

conditions when hexane is present in a mixture with iso-propanol, methyl ethyl ketone, ethyl acetate and benzene [S]. The decrease in conversion of hexane when present in a mixture is illustrated in figure 3. However, these effects are not universal, as the same study determined that the destruction of ethyl acetate is greater when present in a mixture of hydrocarbons than when combusted singly, and there was no observable difference in the destruction efficiency of benzene when combusted alone or in mixtures. Gangwal et al. [ll] have also studied the oxidation of n-hexane and benzene mixtures. A number of general statements may be made about the interactions between classes of compound and the subsequent effects on combustion. Aliphatic hydrocarbons are usually combusted with greater efficiency alone, with significantly lower activity observed for these compounds in mixtures with aromatics. In contrast, esters are frequently destroyed with greater efficiency when present in mixtures, although this is probably due to homogeneous gas phase reaction effects [12]. A decrease in conversion of a VOC when present in a mixture as compared to the activity of the pure compound is generally attributable to the existence of competitive mechanisms for the destruction of the various components of the mixture. This has been observed by Papenmeier and Rossin [13] between chloroform and dichloromethane oxidised by a 3% Pt/alumina catalyst in the temperature range 300-4OO0C,with each chloro-organic depressing the reactivity of the other compared to the pure compound. The combustion of both compounds occur by similar mechanisms, involving adsorption of the chloro-organic onto an oxygen covered platinum surface and subsequent decomposition, and both are also inhibited by the formation of HCl. As both chloro-organics have similar adsorption equilibrium constants, competitive adsorption effects are expected in the two component mixture. Increasing the concentration of one chloro-organic relative to the other causes that compound to occupy a greater 100

90 -

r\

0 o -

[ : :

60 50

.

*

*

*

'

*

.

*

.

'

.

.

.

"

~

.

.

.

EfSects of addition of other VOCs on hexane combustion over bimetallic Pt-Pd catalyst at GHSV= 50,000 h-' in the range 250-450°C (total VOC concentration = 1,200 ppm, VOC mixture of hexane, benzene, ethyl acetate, methyl ethyl ketone and iso-propanol); data adapted from [ 5 ] : p hexane; v hexane in mixture

2: The Destruction Of Volatile Organic Compounds

111

proportion of the adsorption sites, with the result that fewer sites are available for the adsorption and oxidation of the other component of the mixture, thus decreasing its combustion. Papenmeier and Rossin state that competitive adsorption effects occurring in mixed feed streams may result in the reversal of the order in which individual species are combusted relative to the order observed for pure compounds. This is illustrated by a study of the combustion of lean mixtures (200-2000 ppm) of aromatic hydrocarbons over a Pt/alumina catalyst, in the temperature range 100-350°C at an overall space velocity of 134,000 h-' [141. Under these conditions, maximum conversions of 80% were observed. Relative activity for combustion both alone and in 2, 3 and 4 component mixtures were determined, and strengths of adsorption of each compound on the catalyst surface calculated. The reactivity for the pure compounds decreased in the order: benzene > toluene > ethylbenzene > o-xylene > styrene In mixtures, this order was reversed, as the relative strengths of adsorption of the aromatic compounds dictated the extent of surface coverage and hence reactivity. Competition for adsorption sites between compounds is the reason given for the reversal. Strongly adsorbed compounds will block catalytic sites, and thus reduce activity. For example, styrene, the most strongly adsorbed of these compounds, will decrease the adsorption of the other components of a mixture onto the catalyst surface, and thus inhibits their oxidation. If these results are applied to all VOCs, it can be proposed that inhibition is due to competition for adsorption sites between VOCs. Compounds that have low activity for combustion tend to be strongly adsorbed on the catalyst and thus will inhibit the oxidation of more reactive and hence weakly adsorbed compounds. Therefore, in mixtures of VOCs of differing reactivities, it can be expected that the more reactive VOCs will not be oxidised to the same extent as the less active compounds. In order to ensure that all components of a mixture of VOCs are completely oxidised it is necessary to increase the reaction temperature from that which would be required to combust the components of the mixture separately. Commercially, increasing reaction temperature increases the cost requirements of the system, so it would be beneficial to develop catalytic systems in which interaction between VOCs and competition for sites on the catalyst surface are minimised. 3.6 VOC Concentration. - Catalytic oxidation is ideally suited to the destruction of low concentrations of VOCs, this is a major advantage for this abatement technique, since it allows low levels of VOC to be combusted which is essential if industry is to comply with current air pollution legislation. The applicability of various abatement techniques to differing VOC concentrations is shown in table 1 [l]. Commercially, a catalyst capable of efficiently destroying a wide range of concentrations of VOCs would be preferred, so that legislation concerning their release could be complied with regardless of the actual concentration. However, relatively high concentrations have been used in

112

Table 1

Catalysis

Applicability of VOC abatement technologies to VOC concentration; data adapted from [ l ]

Abatement Method

Thermal Oxidation (no heat recovery) Thermal Oxidation (with heat recovery) Catalytic Oxidation Adsorption Absorption Condensation Biofiltration Membrane Technology UV Oxidation

Minimum Concentration /ppm

Maximum Concentration /ppm

20

1000

lo00

3

50 20 lo00 6000 500 1 1

10000 20000 20000 10000 2000 1000 30000

m

a number of studies, since VOCs can thus be used as fuel, thereby reducing the need to supply additional fuel and decreasing the cost requirements of the process. The specific effects of varying VOC concentration will therefore depend on their chemical composition and heating characteristics. The effects of increasing VOC concentration has been considered by Tichenor and Palazzolo [5] for a bi-metallic platinum-palladium catalyst in the combustion of a mixture of hydrocarbons at total concentrations of 1,200 and 6,000 vppm. The higher concentration resulted in a higher destruction efficiency, a result which was more evident at lower temperatures (305°C)than at higher temperatures (400°C).However, this effect may well be due to the heat liberated during the reaction. As the reactor is essentially adiabatic, this would explain the higher conversions experienced for the higher VOC concentrations.

3.7 Deactivation. - The effects of catalyst deactivation in VOC abatement have been reviewed extensively by Spivey and Butt [151, and will not, therefore, be discussed in depth here. The causes of deactivation include high temperatures and the presence of chlorinated, fluorinated and/or sulfated substances, including the reactant VOCs and their partially oxidised products, water or other catalyst poisons, such as metals, in the feed gas. It should be noted, however, that the presence of water vapour in the effluent stream may be beneficial for some catalysts, particularly in terms of their specificity towards completely oxidised products; this is discussed later in this review. Noble metal catalysts are highly susceptible to poisoning, particularly by chlorinated compounds and it has been suggested that metal oxide catalysts have greater poison tolerance than noble metal catalysts, hence these potentially may find widespread use for the oxidation of chlorinated organics. The deactivation of zeolite catalysts is a major problem associated with their use as total oxidation catalysts, and is frequently attributed to the deposition of organic and/or halogenated material in the zeolite pores, preventing the entry of VOCs. This deactivation is generally reversible, with the catalyst requiring regular

2: The Destruction Of Volatile Organic Compounds

113

reactivation by heating in air. The time and cost requirements for catalyst reactivation are major factors which must be considered if it is to be of use commercially. A catalyst that spends a greater proportion of its lifetime being reactivated than catalysing VOC destruction will obviously be of less use than one with long-term stable activity.

4

Catalysts Used for VOC Abatement

Many different catalytic systems have been used for oxidation reaction and this is reflected in the diversity of catalysts used for VOC combustion. However, the types of catalyst used can be catorgorised generally as noble metal and metal oxide based, examples of both systems are discussed in the following sections.

4.1 Noble Metal Catalysts. - Supported noble metals are the most commonly used catalysts for the combustion of volatile organic compounds, accounting for 75% of all such catalysts currently employed in commercial applications [16]. Numerous reviews concerning their catalytic activity, mechanism of destruction and deactivation have been published previously [l5, 17, 181. Noble metals catalysts have high activity for combustion at relatively low temperatures, and show high selectivity for the formation of carbon dioxide and water, with minimal partial oxidation products. However, the susceptibility of supported noble metals to deactivation by poisoning, particularly by halogenated compounds, and the high cost of the noble metal component has initiated research into other possible catalysts for VOC abatement. The mechanism of hydrocarbon destruction over platinum catalysts has been extensively described by Golodets [171 and Volter [191. The mechanisms may also be applied to oxidation over palladium catalysts, since Patterson and Kemball [20] have previously determined that there is no fundamental difference between the mechanism of oxidation over these catalysts. Golodets [17] states that the complete oxidation of hydrocarbons occurs via a surface redox cycle known as the Mars-van Krevelan mechanism, which can be described.

In the first step, molecular oxygen is adsorbed onto a catalyst surface site, (), forming 02(ads).The molecular oxygen dissociates to produce two adsorbed oxygen species, which may themselves adsorb gas-phase hydrocarbon molecules to form CnH,-0 on the catalyst surface, as in step 3. The adsorbed hydrocarbonoxygen species may further react with adsorbed oxygen to produce, via partially oxidised intermediates, carbon dioxide and water. These are able to dissociate from the catalyst surface into the gas phase. The precise reaction mechanism for complete oxidation over noble metal

114

Catalysis

catalysts is not yet clear. The reaction may follow either a Langmuir-Hinshelwood type mechanism (i.e. reaction between adsorbed oxygen and adsorbed hydrocarbon) or an Eley-Rideal type mechanism (i.e.reaction between adsorbed oxygen and gas phase hydrocarbon). Investigations by Volter et al. [19] have determined that during the catalytic cycle, for complete oxidation over a supported platinum catalyst both active reduced metallic (Pt) sites and inactive oxidised (Pt") sites exist. The catalyst studied consisted of highly dispersed platinum supported on alumina. At 5OO0C,the platinum was transformed into an oxidised Pt4+surface complex. Increasing the temperature caused this complex to decompose, forming poorly dispersed metallic platinum. Both forms of platinum were found to catalyse the oxidation of n-hexane, but the metallic form had significantly higher activity. Combining this observation with the Mars-van Krevelan mechanism, it was proposed that the reduced sites are generated by the hydrocarbon, i.e.:

--

C,H, + oxidised catalyst O2 + reduced catalyst

reduced catalyst oxidised catalyst

+ product

The kinetics of the combustion of hydrocarbons is a complex issue, the activity of a catalyst being dependant on many different factors. In determining the kinetics of combustion, therefore, each catalytic system must be considered on its own merits. The noble metals most commonly used as components of combustion catalysts are platinum and palladium, either singly or as a bi-metallic catalyst. Other noble metals can be active, but tend to have lower activity for combustion [20], and undergo sintering more easily, are more susceptible to loose metal component as volatile material, and can be irreversibly oxidised at high temperatures [21]. The varying activity of different noble metals can be attributed to differences in the heat of adsorption of oxygen on the surface. The activity of platinum, palladium, rhodium and gold for the oxidation of ethylene in oxygen has been investigated by Patterson and Kemball [20]. Thin films of noble metal were prepared by evaporation for this reaction. The oxidation of ethylene on platinum occurred at a higher rate than on palladium, under identical reaction conditions, Arrhenius plots of the initial rates of reaction indicated that this was due to a decrease in activation energy from palladium (59 kJ mol-') to platinum (47 kJ mol-'). In addition to the lower rate, partial oxidation of ethylene, primarily to acetic acid, occurred significantly for the palladium catalyst. Kinetic analysis implied that acetic acid was formed over platinum inhibiting the combustion, although experimentally, none was detected. Taking into account the activity of gold and rhodium, the order of combustion reactivity was determined to be: Pt > Pd > Rh >> Au This can be correlated reasonably well to the heats of adsorption of oxygen on evaporated metal films, as determined by Brennan and co-workers; these were 275 kJ mol-' for platinum and palladium and 312 kJ mol-', for rhodium [22].

2: The Destruction Of Volatile Organic Compounds

115

The discrepancy between activity and heat of adsorption for platinum and palladium is not discussed. However, the authors assert that the differences in the activities of platinum and palladium are not due to a fundamental difference in mechanism, but rather are caused by a greater strength of adsorption of ethylene in the presence of oxygen on platinum than on palladium, as indicated by a zero order dependence of ethylene on platinum compared to a first order dependence of palladium. The highest catalytic activity is observed for metals which have weakly bound oxygen species. The low activity of gold cannot be explained by strongly bound oxygen species, but is probably due to the fact that the chemisorption of oxygen does not occur readily on gold, although gold/oxide interfaces do provide sites for oxidant adsorption. From the early work by Patterson and Kemball[20], it is clear that platinum tended to have the highest activity for complete oxidation of noble metals, and had a relatively low tendency to catalyse partial oxidation. For these reasons, platinum is the most commonly used and widely-studied of the noble metals. The destruction of chloroform has also been studied by Lou and Lee using a Pt/A1203 alloy catalyst [23]. The authors have concentrated on the fact that many catalysts do produce undesirable products such as CC14, C2Ch and CO and they have considered the nature of the adsorbed species by interpreting kinetic data. The catalyst was prepared using a wash coating method to produce a catalysts with a bulk density of 0.6-0.7 g ml-' with a BET surface area of 1.267 m'g-', an active metal surface area of 65-75% and porosity of 80.03%. Catalysts were tested at atmospheric pressure and temperatures ranging from 200 to 475"C, with space velocities of 22,500 h-' and 57,000 h-'. The study shows that at a constant temperature, reaction rate increases almost linearly with CHC13 content. The effect of an oxygen rich environment on the reaction was studied, with the conclusion that there is no noticeable difference for temperatures between 290 and 350°C, as the rate was zero order with respect to O2concentration. The main products of this reaction were C02, Cl2 and HCl, with trace amounts of CO and CC14produced across the temperature range. Between 275 and 400°C a linear relationship between the destruction of CHC13 and the formation of C02, C12and HCl was reported. Above this temperature, conversion of CHCl3 did not continue to increase at the same rate with temperature and neither did the production of HCl. In contrast, relative C12 concentration increased and this observation is attributed to the Deacon reaction. 2HCl

+

'/202-----*

H20

+ Cl2

The paper concludes that the Pt catalysts produced by the wash coat method are highly active for the oxidation of CHC13,and produces no significant amounts of products that contain C-Cl bonds. HCl and C12are the only chlorine containing compounds formed at temperatures above 450°C. It must be noted that all experiments were conducted using dry air. They have suggested that water vapour may have an effect and indeed the authors highlight that it requires further study. Rhodium and palladium may also be effective catalysts, as illustrated by

116

Catalysis

Nagata and co-workers [24]. In the oxidation of a mixture of CFC115 (CF3CF2Cl)and n-butane in air at 500°C and a gas hourly space velocity of 15,366 h-' rhodium and palladium catalysts supported on alumina showed activity for combustion comparable to that of platinum at the same temperature. 100% conversion of n-butane and 40% conversion of CFC115 was obtained over all three catalysts under these conditions. This observation is not unconditionally observed, since the activity of the noble metal may be adversely affected by certain reaction conditions. One cause of this is the differing susceptibilities of noble metals to deactivation by VOCs and the products of complete and incomplete oxidation. For example, palladium supported on alumina catalysts are unsuitable for the combustion of CFC113 (C2C13F3),as the noble metal component sacrificially reacts with product fluorine and chlorine species and is thus lost from the catalyst by volatilisation [25]. In contrast, platinum, rhodium and iridium show high activity for combustion under the same reaction conditions, although all are greatly affected by poisoning, causing deactivation. Platinum catalysts are not always appropriate for the combustion of volatile organics. At high temperatures platinum catalysts are susceptible to sintering, causing deactivation [26]. However, it has been stated that temperatures in excess of 300°C are essential in the combustion of chlorinated hydrocarbons in order to prevent strong interaction of HCl with the catalyst surface, leading to extensive deactivation of the catalyst [27]. This gives a limited range of temperatures in which the noble metal catalyst may be used, which is not ideal given the widely varying susceptibilities of different VOCs to catalytic combustion. Structural changes, similar to those seen for platinum, are observed for palladium supported catalysts at temperatures in excess of 450°C in the presence of oxygen [28]. In general, however, noble metals other than platinum and palladium have few commercial applications in catalytic combustion, due to their relative instability and lower activity. The use of rhodium as the component of automotive exhaust catalysts responsible for hydrocarbon oxidation is widely reported, though this has been extensively reviewed by Shelef and Graham [29] and will not be dealt with here. An interesting recent development in the use of gold supported catalysts in combustion has been reported by Haruta and co-workers [30]. The activity of gold for the low-temperature oxidation of carbon monoxide has previously been widely reported [31]. In this study, Haruta determined that a catalyst consisting of 1 wt% gold supported on a-Fe203is active for the combustion of methanol and its decomposition derivatives (HCOOH, HCHO) at temperatures below lOO"C, at a gas hourly space velocity of 2,OOOh-'. Comparison of T50, the temperature at which the catalyst combusts 50% of the initial amount of hydrocarbon, for Au/Fe203and the conventional combustion catalysts Pd/A1203 and Pt/A1203 gave the order of reactivity:

which is in contrast to the order determined by Patterson and Kemball [20], in which gold was found to have low activity. In all cases, the only carbon-contain-

2: The Destruction Of Volatile Organic Compounds

117

ing product observed was COZ. In addition to the obvious advantages of a catalyst with comparable activity to platinum and palladium for low temperature oxidation, it was proposed that the Au/Fe203catalyst may be beneficial as its activity was not suppressed but enhanced by the presence of moisture. The reason for this is not clear, although FT-IR studies have indicated that the adsorption of C O on the catalyst surface is enhanced by the presence of moisture ~321. One class of possible alternative catalysts that show high activity and selectivity for combustion are bimetallic noble metal catalysts, which may have significant advantages over the single component catalyst. Activity is not always enhanced, as illustrated by a study into the combustion of chlorobenzene and xylene over platinum catalyst both in the pure form and with palladium or manganese added as active ingredients [33]. The catalysts considered were 0.15% Pt, 0.1% Pt,O.l% Pt/0.02% Pd,O.l% Pt/O.l% Mn, 0.05% Pt and0.05% Pt/0.02% Pd, all supported on Ba-modifed y-alumina. At a constant space velocity of 10,000 h-I, temperature was varied in the range 177-467"C, to determine the temperature required to give 50% and 90% conversion of a variety of concentrations of xylene and chlorobenzene, both separately and in a two-component mixture. Pure chlorobenzene could not be oxidised with 90% efficiency at any temperature over any catalyst other than 0.15% Pt/alumina , which gave 90% conversion at 440°C. This catalyst was also found to be more active than all others for combustion of chlorobenzene-xylene mixtures, achieving 90% conversion of a 1 mg dm-3 xylene and 2 mg dm-3 chlorobenzene feed at 300°C. Addition of palladium to the platinum catalyst was not seen to increase activity, with 90% conversion of the above mixture at 327°C for the 0.1% Pt/0.02% Pd catalyst. Addition of manganese decreased the temperature at which 90% conversion of the mixture was obtained to 290°C. This catalyst underwent significant deactivation as a result of poisoning by chlorinated compounds, with xylene conversion falling by 47% following exposure to the chlorinated compound. The catalyst could not be completely reactivated by heating in air at 347"C, and the deactivation observed is similar that seen for the pure platinum catalyst. Due to their low combustion activities, platinum-manganese and platinum-palladium were not found to be acceptable alternatives to the single component catalyst. Platinum-palladium bi-metallic catalysts, however, have shown high activity in the combustion of certain volatile organics. This was illustrated by Skoglundh et al. [34], who proposed the use of a platinum-palladium bi-metallic catalyst, with a Pd/Pt ratio of 4/1, supported on hydrothermally pre-treated alumina, for the oxidation of xylene. The hydrothermal support was prepared by treating alumina at 814°C for 2 hours in 100% steam; its merits and preparation are discussed later in this review. This catalyst had a lower light off temperature (235°C)for the oxidation of 220ppm xylene at a space velocity of 144,000h-' and a total noble metal concentration of 5 pg mol-', than the corresponding pure platinum catalyst. Conversely, a 1/4 ratio of platinum to palladium does not promote the combustion activity in the same manner. The behaviour of bi-metallic catalysts may differ markedly from those of the

118

Catalysis

pure noble metal catalyst. A bi-metallic Pt-Pd catalyst supported on hydrophobic fluorinated carbon studied by Sharma and co-workers [161 demonstrates this, The bi-metallic catalyst showed higher combustion activity in the range 200-4OO0C, at a space velocity of 3,000-15,000 h-1, for the combustion of methylene chloride, with a maximum 60% conversion observed at 400"C, significantly higher than for the pure metal supported on the same material. Cordonna et al. made similar observations for the oxidation of a number of hydrocarbons over bi-metallic platinum-palladium catalysts. Catalytic activity was found to increase markedly with increased platinum content, the most active catalyst had a Pt/Pd ratio of 4/1, in accordance with previous work by Skoglundh [34]. Gonzalez-Velasco et al. have further studied the use of platinum-palladium catalysts [35] and have addressed the problems associated with the destruction of hydrogen deficient chlorinated VOCs in an attempt to minimise the production of COC4 and C12. This has been approached by the addition of toluene, water and hexane to the reactant and feed. Many studies have previously added water to the feed, and although studies have investigated mixtures of the VOCs, not many have investigated the importance of hydrogen transfer from other reactants during VOC destruction. The catalysts employed were 0.5 wt% Pt and Pd on y-A1203.They were calcined in air at 550°C for 4 h before being reduced at the same temperature for a further 2 h. Characterisation of the catalysts showed loadings of 0.44 wt% Pt and 0.42 wt% Pd with dispersions from hydrogen hemisorption of 53% for Pt and 43% for Pd. Trichloroethylene was fed into the reactor after mixing with various concentrations of water (1000 ppm, 7500 ppm, 15000 ppm) and it was reported that over Pd/A1203,adding water had no effect on the reaction. Over Pt/A1203 the catalyst activity was enhanced by water up to 400"C, but was inhibited above this temperature. Increasing the concentration of the water caused this enhancement or inhibition to be more pronounced. The activity of both catalysts was effected by the addition of hexane of toluene. Over Pd/A1203 the light off temperature for the reaction was reduced from 400 to 325°C in the presence of both toluene and hexane. Over Pt/A1203the decrease in light-off temperature was from 425 to 325°C for toluene and hexane. The authors suggest that this enhancement in activity is due to the exothermic oxidation of the added hydrocarbons increasing the surface temperature of the catalyst. Without added water the amount of C2Cl4 formed reached a maximmum at 450°C for both Pt/A1203 and Pd/A1203, with approximately twice the yield formed over Pd/A1203compared to Pt/A1203.With the addition of water (1000 ppm) C2C14 was still produced but the concentration was reduced considerably, and by the addition of 1500 ppm water it was reduced further. As expected similar improvements in HCl selectivity over C12were observed when water was added. The addition of water also decreased CO selectivity, over Pd/A1203 the amount of CO decreased by a factor of 4 between 350 and 400°C. With higher H20 concentrations (7500,15000 ppm) no CO was observed at all. The authors suggest that the formation of OH- species on the surface of the metal decreases

2: The Destruction Of Volatile Organic Compounds

119

selectivity towards CO by promoting the water gas shift reaction and the observed reactivity is consistent with previous work [36]. The addition of hexane and toluene also had a beneficial effect. Over Pd/A1203 the amount of C2Ch formed was 40 times lower in the presence of hexane whilst it was 20 times lower with toluene added. C2C14formation was also inhibited and it was apparent that the production of C12was suppressed. It was also noted that the chlorine balance was often low, suggesting that chlorine was retained by the catalyst and this eventually could lead to deactivation. These results indicate that in the bi-metallic catalyst the active surface is not a simple mechanical mixture of platinum and palladium, and a new structure responsible for the high activity is formed on the surface, which has not as yet been satisfactorily determined, although alloy effects are thought to play an important role. Platinum catalysts promoted with base metals have been studied by Bo-Hyuk Jang et aE. [37]. The catalysts contained vanadium, chromium, manganese, cobalt, copper and barium and were prepared by impregnation with the various promoters added to a slurry of Pt/A1203. Three groups of catalyst were prepared containing different base metal loadings of 2, 6 and 18 wt%. All catalysts were dried for 20 min at 120"C, and then calcined at 500°C for 1 h. Catalysts were tested in a fixed bed microreactor with a gas stream consisting of 10% C02,4% O2and 1000 ppm of chlorinated hydrocarbon with a gas hourly space velocity of 20,000 h-'. The system temperature was then raised from ambient to 450°C at a ramp rate of 5" min-'. The chromium-promoted catalyst was the most effective for destroying the chlorinated hydrocarbons, with 85% conversion at 450°C. It was evident that the chromium catalysts were most active, with vanadium the next active showing a maximum conversion of 53%. The other catalysts did not achieve 50% conversion. Further studies with the most active chromium and vanadium promoters showed that in general the activity increased as the loading of chromium and vanadium increased. Accelerated ageing of the catalysts on line was performed by heating to 800°C for 120 h. This treatment caused the catalyst activities to drastically decrease. The best catalyst was still the chromium-promoted system but conversion decreased to 40% at 450°C. The performance of the vanadium catalyst after ageing was poor and the authors attribute this to the melting point of vanadium (690"C), which caused loss of the promoter from the catalyst. It is also evident that the BET surface area of the vanadium aged catalyst decreased by 84% from the initial value. However, these studies indicate that the incorporation of other metals with platinum may have a beneficial effect for complete oxidation activity. A number of catalysts consisting of mixed noble metal - metal oxide have been proposed for combustion processes and these are discussed later in this review. Most noble metal catalysts are supported, the most common support material being y-alumina. This role of the support is to increase the dispersion and thus enhance activity, economically this also results in a decrease of the quantity of the costly noble metal component required for the catalyst. The dispersion of the noble metal on the catalyst support is therefore of prime importance. In a highly

120

Catalysis

dispersed catalyst, the interactions between the noble metal and the support can have considerable effects on catalytic behaviour. A poorly dispersed catalyst may have less interaction with the support, resulting in properties similar to that of the bulk metal. As metal concentration is increased particle size may increase, so support interactions will decrease. Variations in particle size are also associated with changes in the density of active sites, with associated effects on the catalysis. The concentration of the noble metal can affect both catalytic activity and resistance to deactivation, hence the amount of noble metal on a supported catalyst is of prime importance. The former is illustrated by Skoglundh [34], who noted that the overall noble metal content of a Pt-Pd catalyst (ratio 4/1) is of vital importance in determining the light off temperature for xylene oxidation, as the light off temperature markedly decreased as noble metal content increased in the range 5-20 pg mol-'. These results are shown in figure 4. Regarding susceptibility to deactivation, Mendyka [33] determined that a

I

Figure 4

I

I

I

I

Efect of noble metal content on T,, for the combustion of 220 ppm xylene by Pt-Pd bimetallic catalyst supported on a hydrothermally treated washcoats at GHSV= 144,000 h-' [34]: v 5 pmol per gram, o 1 Opmol per gram, h 20 pmol per gram

2: The Destruction Of Volatile Organic Compounds

121

catalyst with high noble metal content (0.6-1.5% Pt) is relatively poison-resistant in the oxidation of xylene and chlororbenzene, under the conditions previously described. However, the use of additional noble metal will significantly increase the cost of the catalyst, a highly undesirable situation in commercial operations, and will not completely prevent poisoning. As such a completely different type of catalyst with higher poison resistance might be preferable for effluents in which the levels of chlorinated compounds and other poisons are such that deactivation of a noble metal catalyst is likely. The effects of metal concentration on activity are discussed in greater depth by Spivey and Butt [l5]. Their review deals extensively with catalyst deactivation, hence the causes of deactivation are not considered in this review. However, the effects of water and chlorinated compounds in total oxidation have led to the development of a variety of novel combustion catalysts, which will be discussed briefly. The presence of moisture in the feed gas is a common cause of catalyst deactivation. Water vapour may be present as a contaminant in the effluent stream, and is also produced by combustion. This poses a particular problem for alumina supported catalysts. Adsorption of water vapour on the surface of the catalyst is unavoidable at the low temperatures necessary for catalytic oxidation to be economically viable. It has been suggested that the use of a catalyst which is hydrophobic in nature and will thus prevent adsorption of water on its surface will have major advantages over the conventional hydrophilic catalysts. In addition, when the hydrophobicity of a catalyst to water increases, hydrocarbons are often more readily adsorbed on the catalyst surface where they can subsequently react. This will result in an increase in the number of active reduced sites on the catalyst surface, which have previously been shown to be essential in the mechanism of combustion reactions over noble metal catalysts. Hydrophobic systems have been extensively studied by Chuang and co-workers [38], and have comparable activity to conventional hydrophilic catalysts, such as Pt/alumina, at significantly reduced temperatures. The high catalytic activity appears to be due to the hydrophobicity of the catalyst to water and to the high ratio of the surface reduction rate constant (k,) over surface re-oxidation rate constant (k3), obtained from the Mars-van Krevelan mechanism previously detailed. The hydrophobic catalyst employed consisted of 6 mm ceramic Raschig rings coated with a mixture of hydrophobic fluorinated carbon (containing 60% fluorine) and Teflon, impregnated to a platinum loading of 0.2 wt.%. The contact angle of the catalyst with water was 109",indicating its hydrophobic surface properties. The catalyst was tested for the combustion of a feed stream consisting of 250 vppm of benzene, toluene or xylene, or a bi- or tri-component mixture of these, diluted in air to 10-60 vppm at a gas hourly space velocity of 3,200 h-' in the temperature range 90-150"'. The experimental procedure consisted of measuring the reactor inlet and outlet concentrations at a series of different feed compositions and constant temperatures. The oxidation of benzene, toluene and xylene, as the sole component of the effluent, was complete in the temperature range 9O-15O0C,with no partial oxidation products or CO detected. At 130°C and inlet concentration of 45 vppm over 90% of the VOC stream was converted. This was significantly higher activity than that observed for a conventional hydrophilic catalyst (250-

122

Catalysis

300°C required for comparable conversion). It can be seen that there is a significant decrease in the temperature required for combustion to occur with the hydrophobic catalysts. A detailed kinetic study of the combustion process indicated that the surface reduction rate constant, kl, in the Mars-van Krevelan mechanism increases with temperature faster than the surface re-oxidation rate constant, k3. That is, the surface concentration of adsorbed oxygen (02) is higher than that of (C,H,-0) at higher temperatures. This occurs in spite of the fact that kl is more sensitive to variations in temperature than k3.The ratio kl/k3 for a Pt/hydrophobic catalyst in the combustion of benzene, toluene and xylene at 90-130°C was calculated to be in the range 466-3880, which can be compared to 24-2490 for a conventional hydrophilic catalyst (Pt-Ni/A1203)in the oxidation of benzene and n-hexane at 140-221°C C341The higher values of this ratio for the hydrophobic catalyst suggested that the hydrophobicity of the catalyst may accelerate the desorption of water from the catalyst surface and thus accelerate the forward reaction (step 4 in the Mars-van Krevelan mechanism). This may result in a reduced catalyst surface, which has previously been determined to favour high activity. A similar situation exists in the oxidation of CO on pure platinum metal [18], such that the surface concentration of CO decreases as the reaction temperature increases. Because the carbon atom of CO is known to adsorb to the platinum surface, a similar adsorption mechanism may exist for structurally similar hydrocarbons. This postulate is consistent with the mechanism employed. The addition of water to the feed gas does not affect the conversion of benzene, toluene and xylene on the hydrophobic catalyst, as there was no significant loss in conversion as the proportion of water vapour increased. A further advantage of the hydrophobic catalyst is that the combustion of VOC mixtures containing two and three components did not depress conversion when compared to the single components. This was in clear contrast to results previously described, in which interactions between the components of mixtures often resulted in the inhibition of oxidation of the mixture’s weakly adsorbed components, with the result that increased temperature was required to achieve high levels of destruction compared to the components alone. A further benefit of using a hydrophobic catalyst is that there is lower sensitivity of the reaction rate to the oxygen to hydrocarbon ratio. Chuang demonstrated this by comparing the combustion of formaldehyde over Pt/hydrophobic support to that of conventional hydrophilic Pt/alumina [39]. It is generally accepted [17] that the activation energy for oxygen desorption is much lower than that for combustion, and this was confirmed by detailed kinetic studies of the hydrophobic catalyst. The rate limiting step for both catalysts was the reaction between formaldehyde and surface oxygen and maintaining a high oxygen to formaldehyde ratio was of vital importance in achieving high C 0 2 selectivity. For the Pt/alumina hydrophilic catalyst, C 0 2 selectivity falls as the ratio decreased, particularly at low temperature. However, for the hydrophobic catalyst selectivity only decreased if the ratio was below 500, a situation which is rarely encountered in practical applications. Because the selectivity to C 0 2 was

2: The Destruction Of Volatile Organic Compounds

123

low and the dependence on the ratio was high for the hydrophilic catalyst, the hydrophobic catalyst might therefore offer significant advantages. It was observed that whilst conversion of formaldehyde took place at 63"C, activity was low and selectivity to C 0 2 negligible, with partial oxidation to HCOOH occurring at a significant rate. Selectivity to C02increases with increasing temperature, reading 100% at 125°C and 100% conversion was observed at temperatures in excess of 150°C. This suggests two reaction pathways between formaldehyde and adsorbed oxygen, one producing complete oxidation products at high temperatures, the other forming the partial oxidation product HCOOH at lower temperatures. The authors suggest that this indicates that at least two types of active sites are involved in the reaction. With increased temperature selectivity to C02 increased, indicating that sites active for complete oxidation become more active. Detailed kinetic studies showed that the rate of consecutive oxidation of HCOOH becomes increasingly greater than the rate of desorption of surface HCOOH. As a result, less of the unwanted partial oxidation product HCOOH was produced and selectivity to C02 increased. The presence of water vapour in the feed gas may have a beneficial effect on the combustion of chlorinated VOCs. For ease of removal from the effluent, HCl is the preferred chlorinated product of combustion. As HCl generally inhibits the rate of combustion, it might have been expected that any increase in its formation would result in a decrease in reaction rate. However, this does not occur, since the inhibition effects of C12are similar to those of HCl. It has been widely reported that water in the feed gas acts to enhance the production of HCl during the oxidation of a chlorinated VOC. This is sometimes attributed to the use of water as a source of hydrogen atoms [40]. However, other papers have proposed the existence of alternative reaction mechanisms in the presence of water, which give increased selectivity to the production of HCl and may also increase the conversion of the organic material. One such mechanism has been detailed by Rossin and co-workers [41], who state that the presence of water appears to play an important role in the overall reaction sequence, minimising the production of Clz in favour of HCl without significantly affecting the catalytic activity. During the oxidation of chloroform over 2% Pt/alumina, HCl was the major chlorinated reaction product. The oxidation of chloroform in humid air might be expected to proceed as follows: CHC4

+ 0 2 e C O 2 + HCl + C12

This reaction is not consistent with the observed results, as negligible C12 was produced, suggesting that the majority of C12was converted to HCl. As there is water in the effluent, it might be supposed that the Deacon reaction: 4HCl

+ + 2C12 + H20 0 2

is responsible for the production of HCl. Studies of the kinetics of the process indicate that this is not the case. Equilibrium concentrations of HCl and Cl2, calculated using the values determined in this study, indicate that the Deacon reaction, if it occurred, would be shifted towards the formation of C12,hence this cannot be responsible for the preference to HCl.

124

Catalysis

Bond and Sadeghi [42] previously suggested that the hydrogen required to make HCl may be produced by the combustion of the hydrocarbon. Lester further speculated that the formation of HCl can be favoured by adding hydrogen, in the form of water or organics, to the feed gas [43]. It was further proposed that the conversion of halogenated organics in humid air streams involves hydrolysis, and the presence of water is essential to limit the formation of Cl2. Thus, a reaction mechanism consistent with Rossin and Farris' products is:

+

+

CHC13 [ O l d [COC12] HCl C02 + 3HC1 [COCl2] + H20The first step involved an interaction between chloroform and adsorbed oxygen, whereby HCl is abstracted to yield a phosgene intermediate, which is consistent with the results obtained form a kinetic study of the reaction. Phosgene then undergoes rapid hydrolysis to yield C 0 2and HCl. A slight inhibition of the rate is observed in the presence of water. This is attributed to the rate limiting step for the reaction being the reaction between chloroform and oxygen to form phosgene. If the reaction between phosgene and water were the rate limiting step, the addition of water would be expected to increase reaction rate. This mechanism was found to be consistent with the products and kinetics of the reaction. Further evidence for the existence of a hydrolysis mechanism is provided by Papenmeier and Rossin [12] and by Agarwal and co-workers [40]. Agarwal detailed a hydrolysis reaction mechanism which is said to predominate in the presence of water vapour, providing both increased activity and increased selectivity to HCl compared to the oxidation mechanism operating in 'dry' conditions. This mechanism was consistent with the observed reaction products and with the kinetics of the oxidation. The initial observation was that, in the oxidation of cyanogen chloride in air over a 2.15% Pt/alumina catalyst, 78,000 ppm water vapour in the inlet gas acts to enhance conversion, giving 98% conversion at 375°C and a space velocity of 170,000 cm3h-'g-' compared to a maximum 20% conversion for the corresponding dry feed stream at 440°C and 46,000 cm3h-'g-'.. The conversion in the presence of water was mainly to HCl and C02, with decreased selectivity to partial oxidation products and C12.The dry feed stream produced negligible HC1 and C02, with CO the only carboncontaining product detected. It was proposed that the enhancement in conversion was due to the existence of an alternative hydrolysis pathway in the presence of water. Such a mechanism has been reported previously by Lester and Marinangeli [44] for the oxidation of cyanogen chloride over a platinum supported on titania catalyst. Agarwal [40] found that the presence of water vapour acted to reduce the apparent activation energy in their system, from 96 kJ mol-' at 390-440°C for the dry feed gas to 54 kJ mol-' at 185-215°C for the humid feed gas. Experimental conditions were such that diffusion resistance was negligible, and therefore, the lower activation energy for the humid feed gas suggests the presence of an alternative reaction mechanism. This was confirmed by further experiments in which the catalyst particle size was reduced to approximately half its original size, consequently no change in apparent activation energy was

2: The Destruction Of Volatile Organic Compounds

125

detected. By varying the oxygen concentration of the feed stream in the range 0-21%, it was determined that cyanogen chloride conversion is independent of the concentration of oxygen, hence the rate of its destruction is zero order over this range. This suggests that hydrolysis was the major pathway for cyanogen chloride destruction. Two reaction mechanisms are proposed, one for the dry feed stream: 2CNCl

+

02-4

Cl2

+ N2 + 2/X C02 (where 1<X<2)

and an alternative hydrolysis reaction for the humid feed stream: CNCl

+ H20-

NH3

+ HCl + C02

It was suggested that in the dry feed stream, only oxidation may occur, as shown in the former mechanism, whereas, in the humid feed stream the latter hydrolysis mechanism is predominant for destruction, resulting in increased destruction of highly stable cyanogen chloride, and also increased selectivity towards the HCl and C02.Thus, the enhanced selectivity to HCl in the presence of water may be due to the existence of an alternative reaction pathway, and not merely to the presence of an additional supply of hydrogen atoms, as has often been stated. Whilst cyanogen chloride is not itself classified as a VOC by the US EPA [11, it is a suitable model for other chlorinated organics which are VOCs. Its destruction is also of great interest in military applications for the removal of chemical warfare agents from air.

4.2 Design of Catalyst Supports. - A significant proportion of all VOCs are halogenated, and growing awareness of ozone depletion has resulted in increased legislation against the emission of halogenated compounds. An industrial catalyst is desired which is capable of combusting halogenated and other highly stable VOCs efficiently at relatively low temperatures with minimal catalyst deactivation. However, halogenated compounds have been implicated as a major cause of deactivation of noble metal catalysts. This is often due to the reaction between alumina, the most commonly used support material and halogen species, forming aluminium halides and thus degrading the support and blocking catalytic sites [16,25,45]. Therefore, a current area of much research is into the development of alternative support materials for noble metal combustion catalysts. Supports considered include stainless steel (a mechanical rather than chemical support) [46], silica [47], titania [47], zirconia[48] and carbon [38]. Pre-treatments of the support material have also been considered, for example sulfation [49] and hydrothermal pre-treatment [34]. In the oxidation of methane over palladium catalysts, as studied by Muto et al. [SO], the activity of the supported catalysts decreased in the order: Pd/Si02 > Pd/A1203> Pd/A1203-Si02 However, the reverse order was observed for platinum catalysts also for the oxidation of methane [51]. The acidic properties of the support are of prime importance, with solid superacids showing high activity at relatively low temperatures [49]. Ishikawa and co-workers have found that in the oxidation of

126

Catalysis

propane by supported platinum catalyst, the activity of the supported catalysts follows the order:

-

-

Pt/Si02 > Pt/A1203 Pt/Ti02 Pt/Ce02 > Pt/ Zr02 > Pt/La203 With the exception of silica, this correlates well to the order of acid strength of the catalyst: Ti02 > Si02 > Z r 0 2> La203 The type of support influences the metal dispersion also, which may account for the high activity of silica, and the very low activity of lanthana. As acidic properties appear to influence activity, the doping of the catalyst with SO:' has been proposed as a means of increasing the acidity of the catalyst, and thus increasing its activity [46,48]. Zirconia has been proposed by Hubbard et al. [47] to be a possible superior support material to alumina. For highly dispersed platinum, the oxidation of propane occurs at a rate as high as two orders of magnitude higher over Pt/zirconia than over Pt/alumina. At higher platinum concentrations, the influence of support is negligible and the rate constants for both catalysts are the same. The change in activity was proposed to be due to a difference in interaction between noble metal and support. The differences in activity between supports might be expected due to the great differences in structure for alumina (poorly crystalline, high BET surface area) and zirconia (highly crystalline, low surface area). Analysis of infrared CO adsorption data indicates that there is little interaction between platinum and zirconia [52]. It can therefore be proposed that highly dispersed platinum is active for propane oxidation provided that it is supported on an inert material, such as zirconia. It is also stated that interaction with alumina deactivates platinum, with only one third of sites active for propane oxidation compared to the zirconia supported catalyst [53]. Hence, a zirconia support may have significant advantages over an alumina support in combustion of noble metals. The effects of hydrothermal pre-treatment of an alumina support have been investigated by Skoglundh [34]. Alumina was treated either thermally at 500°C in air for 2 hours, or hydrothermally by treating at 814°C for 2 hours in 100% steam. The noble metal component of both catalysts consisted of Pt-Pd in the ratio 80:20. Relative to the thermally treated catalyst, the hydrothermally treated catalyst had a lower light-off temperature in the combustion of xylene than both the thermally treated catalyst and the pure, unsupported platinum particles. It has been determined that hydrothermal pre-treatment decreased surface area and shifted the pore size distribution of the catalyst to larger pores [54]. It is suggested that larger molecules diffuse more rapidly in a hydrothermally treated support and can thus reach catalytically active metal more rapidly, hence the increased activity. Hydrothermal pre-treatment may also alter the metal-support interaction. Measurements of platinum particle size on this support have indicated that the noble metal is present as larger crystallites than in thermally treated catalysts, hence there is less interaction with the support material. It is

2: The Destruction Of Volatile Organic Compounds

127

clear that these thermal treatments affected both the structure of the support and the morphology of the platinum crystallites. Other possible pre-treatment have been considered, for example, the exposure of Pt/zirconia and Pt/alumina to hydrogen at 500°C increase the oxidation rate of both catalysts [53]. However, pre-treatment of Pt/zirconia in oxygen at 500°C has negligible effect on its activity [48]. This is said to be due to a lack of interaction between noble metal and support, preventing re-dispersion of Pt at low concentrations.

4.3 Gold as a VOC Destruction Catalyst. - Continued research into the use of noble metal catalysts for complete oxidation reactions is required to determine the composition of catalysts most active for the process and the mechanism by which these operate. In spite of considerable research into alternative supports, varied noble metal loadings, etc., the susceptibility to deactivation of these catalysts remains a problem, particularly in the oxidation of chlorinated compounds. For this reason, alternative classes of catalysts active for VOC combustion are required. Recently, there has been considerable interest in gold as a heterogeneous oxidation catalyst. Gold based catalysts on a variety of supports have been employed by Baoshu Chen et al. for the complete oxidation of dichloromethane [55]. The use of alumina has been shown to degrade at high temperatures with HCl. Supporting the metal on Ti02/V205or Cr203/A1203has lessened this tendency. It has also been shown [56] that a small amount of gold on a cobalt oxide can enhance chlorinated hydrocarbon destruction. The gold catalyst, Au/C0304 was prepared by co-precipitation, by ageing the catalyst at 60-70°C for an hour before it was washed with distilled water. After drying, the catalyst was prepared by calcination for 8 h at 350°C. Water was co-fed into the reaction feed stream to suppress the Deacon reaction that forms chlorine gas. The effect of gold loading on activity was investigated, and the results compared with conventional noble metal catalysts. Reaction conditions were 500 ppm dichloromethane and 0.5 wt% water in air, giving a gas hourly space velocity of 15,000 h-'. Catalysts with a gold loading of 0.2,1,5 and 10Wt% were prepared and all demonstrated an increase in activity compared to the Co304 support alone. For example the temperature for 50% conversion was 3 10°C without gold but decreased to 210°C when it was present. The reaction rate was also increased by 25 times at 300°C when in gold was added to the support. It was also reported that the activity was independent of gold loading. In some cases CHC13 and CCl, were produced as by-products at temperatures below 250"C, however, above this temperature only HCl and C02 were produced. Comparison with other noble metal catalysts showed that the gold catalysts were active at lower temperatures. A conversion of 50% dichloromethane was achieved at 210°C for Au/Co304 compared to Cr203/A1203(260"C), 0.5% Pt/A1203 (345°C) and 0.5% Pd/A1203 (395°C). These observations were explained in part by poisoning of the noble metal catalysts by chlorine at low temperatures. Whilst it has been reported that that over a test run of 130 h the 5% Au/Co304catalyst produced a consistent conversion of 95%. Conversely the

128

Catalysis

0.5% Pt/A1203catalyst was stable for around 70 h with a lower conversion of 40% whilst the 0.5% Pd/A1203catalyst showed a decrease in conversion from 25% conversion to zero after 25 h. The catalytic combustion of n-hexane, benzene and 2-propanol was investigated using Au/Ce02/A1203and Au/A1203by Centeno et aZ. [57]. The catalysts were prepared by the method of deposition-precipitation. They showed that ceria enhances the fixation and final dispersion of the Au nanoparticles and stabilises them at lower crystallite sizes. The addition of ceria therefore improves the activity of the gold particles for the oxidation of VOCs. They propose that the enhanced activity may be caused by an increase in the mobility of lattice oxygen and controlling and maintaining the required oxidation state of the active Au nanoparticles. Minico et.aZ. [58] have studied the oxidation of alcohols, acetone and toluene using Au/Fe203prepared by coprecipitation in the presence of excess oxygen. The high activity of the catalysts was ascribed to the increase in the mobility of lattice oxygen.

4.4 Metal Oxide Catalysts. - The use of metal oxide catalysts for oxidation reactions has been well documented, and general reviews concerning catalytic activity [171 and mechanistic principles [59] have been published. Metal oxide based catalysts have been specifically applied to the combustion of VOCs [l5] although their use is not as widespread as catalysts based on noble metal systems have been most extensively studied. It is generally accepted that oxide catalysts show greater resistance to poisons when compared to noble metal catalysts, they also have the advantage that the catalyst tends to be less expensive. However, metal oxides frequently show lower catalytic activities which may require the use of lower space velocities and higher temperatures to give comparable performance. A class of oxide catalysts which have been employed for combustion reactions, particularly hydrocarbon combustion are oxides with the perovskite structure, possessing the general formula AB03[60]. The activities of several unsubstituted component B oxides (B03)have been compared with perovskite oxides for the catalytic oxidation of propylene [61], this is shown in figure 5. Catalytic activity is expressed in terms of the temperature at which the combustion of propylene occurred at a given rate. Catalysts below the straight line showed enhanced activity over the component oxide due to the formation of a perovskite structure. Conversely for catalysts above the line activity was reduced. Although there is a degree of scatter most catalyst are generally distributed close to the line indicating that the activity of the unsubstituted perovskite oxides are primarily determined by the nature of the B component oxide. The most active catalysts being based on the oxides of Co and Mn. The catalytic combustion of methane over perovskite type catalysts has been investigated by Arai et al. [62]. Methane is one of the most stable alkanes and is relatively difficult to combust by virtue of the high strength of the C-H bond which must be activated. Studies were performed using relatively high space velocities in the range 45,000-50,000 h-' with a 2% methane feed in air. The catalytic activity, expressed as the temperature required for 50% conversion, is

2: The Destruction Of Volatile Organic Compounds SrCe0,-CeO 650

-

o2

129 SrTi0,-Ti0

o2

CaTi0,-TiO, 0

Figure 5

Efect of perovskite substitution on the activity for propylene oxidation 1611; T is the temperature at which propylene oxidation rate = 1 O-' mol m-2 s-I

shown in table 2 for a series of unsubstituted perovskite type oxides. Carbon dioxide was the sole reaction product over all the catalysts tested. Comparison of the activity was made with a Pt/alumina catalysts, and although the perovskite oxides were less active, this was only marginal in the cases of LaCo03, LaMn03 and LaFe03which showed much lower surface areas. The activity of the lanthanum perovskite oxides were enhanced by substitution of Sr2+,this was particularly prevalent for the lanthanum manganate variable, La&Sro.4Mn03,which showed a T50%of 482"C, 36°C lower than Pt/alumina. The substitution of Sr2+ for La3+ leads to the formation of positive holes and/or oxygen vacancies, and it is the formation of these defective structures which are considered to impart the high catalytic activities. It has been reported that on cooling in oxygen and heating in vacuum these oxides absorb or liberate large amounts of oxygen, this phenomenon is clearly important in catalytic reactions and the behaviour of these oxides and the nature of the oxygen species have been previously reviewed [60]. A more detailed study of the combustion of methane over a series of Mg doped LaCr03 perovskites has been reported by Saracco et al. [63]. The catalysts prepared were LaCrl.xMgx03 , where 0 < X< 0.5, and were synthesised by the method denoted as the 'citrate method', which briefly consisted dissolving the constituent metal nitrates in citric acid solution. After heating the solution the catalyst precursor was obtained and subsequently calcined at 1100°Cto produce the final catalyst. Catalysts were screened for activity using 1.5% methane, 18%

130

Table 2

Catalysis

Activity of perovskite type oxides for methane combustion expressed as temperature required for 50% conversion [62]

Catalyst

G O / OC

Surface area /m2g-' ~

~~~

LaCo0,

3.0

535

LaMn0,

4.0

579

LaFe0,

571

LaCuO,

3.1 0.6

LaNi0,

4.8

702

LaCr0,

1.9

780

146.5

518 834

Pt/alumina thermal oxidation

-

672

oxygen with the balance helium. Substitution of Mg into the perovskite structure enhanced the combustion activity, as LaCr03 showed a TSo%= 692°C which was reduced to 553°C for LaCro.sMgo.s03. Catalyst activity was improved further by supporting the perovskites on MgO, although more importantly the support stabilised small perovskite crystallites reducing sintering. Kinetic analysis of the supported catalyst showed that the reaction was first order with respect to methane, and that dissociative adsorption of oxygen exerted a significant role on the reaction mechanism. The experimental observations were consistent with the operation of an Eley-Rideal type mechanism for methane combustion over these cat a1y st s. The combustion of other VOCs by perovskites, besides alkanes and alkenes, has also been investigated. Ling et al. [64] have studied LaNi03catalysts for the combustion of ethanol and acetaldehyde, comparing activity of that for methane combustion. Oxidation of 1 vol.% VOC in air (total flow-100 ml min-', 0.lg catalyst) followed the order for ease of combustion; ethanol z acetaldehyde >> methane The temperature required to attain 90% methane conversion was ca 600°C whilst equivalent conversion of ethanol and acetaldehyde were reached around 400°C The preparation of monolith perovskites by extrusion of plastic pastes, comprising perovskite powder, binder, acid peptizers and some surfactant, suitable for use in high temperature incineration processes have been reported by Isupova et al. [65]. Preliminary catalytic activity results showed that the monoliths were active for the combustion of butane, gasoline, methane and chloroform in air at GHSV = 12,000 h-', although the required reaction temperature was high, in the region of 1000°C.It was reported that catalysts were used continually for over one month at 900°C without loss of monolith integrity, mechanical strength and catalytic activity. The preparation of such active monoliths is an interesting concept and one which may be applied to other oxide catalysts,

2: The Destruction Of Volatile Organic Compounds

131

especially for use in VOC destruction which requires high volumetric throughput with a low pressure drop. The destruction of halogenated VOCs, particularly those of short chain length, are of great industrial and environmental importance, and a considerably number of studies using oxide catalyst have investigated this area. The oxidation of 0.74 vol.% dichloroethylene by a wide range of oxides has been studied by Imamura at 3,600 h-' space velocity [66]. Catalytic activity was defined in terms of C02 yield and at 650°C the activity was ranked in the order: Cr203 > Mn203 > Co304 > CuO > La203> Ce02 > NiO > MgO-CaO > MgO > CaO > ZnO > Si02-Al203 > V205 > Si02-Ti02 Although transitions metal oxides demonstrated the highest activity they all showed high yields of C12 whilst acidic catalysts such as Ti02-Si02and Si02A1203 produced HCl almost exclusively. C12 is highly reactive and can readily react with carbon containing substrates forming harmful by-products. The product distribution over the metal oxides changed quickly with time as CO selectivity increased at the expense of C02, indicating that the combustion activity decreased with time on line. This behaviour was not observed over the acidic catalysts which were concluded to be more suitable for dichloroethylene destruction. A range of solid acid catalysts, including Ti02/Si02,zeolite Y, various mordenites and A1203/Si02,were tested for the destruction of 1% 1,2-dichloroethane in air at 3,600 h-' space velocity [67]. The most active catalyst was TiOz/SiOz which showed 100% conversion at 400°C, CO was a major product showing a CO/CO2 ratio of ca. 2.5. Zeolite Y also showed a high conversion of 98.5% at the same temperature, however, the carbon balance was poor (33.5%0),highlighting that the zeolite Y catalyst tended to have a short lifetime, being rapidly deactivated by deposition of carbonaceous organic residue in the zeolite pore structure and the catalyst surface. All of the zeolite-based catalysts were rapidly deactivated and had inferior activity to TiO2/SiO2, which showed stable activity, these other zeolites were also deactivated by carbon deposition. Imamura et al. [68] have also investigated the destruction of dichlorodifluoromethane (CC12F2)over a similar range of solid acid catalysts used previously [67] and single and mixed oxides. Again Ti02/Si02exhibited the best performance, initially showing 95.7% conversion at 5,900 h-' space velocity, other acidic catalysts such as mordenite and A1203/Si02showed higher conversions, however, these catalysts deactivated quickly. The TiOz/SiO2 catalyst also deactivated, conversion decreased markedly after 150 minutes reaching a steady state of ca. 10% after approximately 300 minutes use. This deactivation was due to the attack of corrosive fluorine on the silicon component of the catalyst, as during use silicon was removed from the catalysts and deposited at the reactor outlet. The addition of CaO prolonged catalyst lifetime by reacting with fluorine, although the catalyst still showed significant deactivation as conversion decreased to 50% after 600 minutes time on line. Comparison with transition metal oxides (Ti02,Cr203,Mn203,Co304and Fe203),ZrP207,Zr02/Mo03,Ti02/Zr02 and CaO showed that all, with the exception of Cr2O3 and CaO, gave appreci-

132

Catalysis

ably CC12F2conversion although activity was generally lower than Ti02/Si02 and deactivation was a very rapid process. These studies highlight problems associated with the corrosive nature fluorine from the destruction of fluorinated VOCs which often leads to catalyst deactivation of both metal oxides and noble metals [25]. This is illustrated by the work of Imamura and co-workers in comparing the destruction of 1,2 trichloroethylene and dichlorodifluoromethane over the same group of catalysts. The oxidation of CC12F2by a Ti02 (anatase) catalyst has also been investigated by Karmakar and Greene [69]. Catalytic activity was determined using 1500-2000 ppm CC12F2in air with 10,500 h-' space velocity in the presence and absence of 5000-6000 ppm co-fed water. The addition of water had no significant effect on initial catalytic activity, but dramatically increased the HCl/Cl ratio and increased the selectivity to C02, which was 100% at all reaction temperatures. Time on line studies at 300°C with co-fed water showed that less than 5% decrease in conversion was observed after 4 days continuous operation. Over the same time period the Ti02surface area was reduced from 170 m2g-' to 40 m2g-', 50% of the reduction took place in the first 1-1.5 hours, simultaneously the catalyst activity increased. This increase was due to the increase in catalyst acidity which was a consequence of surface fluorination. It is interesting to note the differences in Ti02 deactivation characteristics between the two studies investigating CCl2F2 decomposition [67,68]. The differences can be related to the concentration of the fluorinated VOC and fluorine containing products and residence time within the catalyst bed. The former catalysts [68], which deactivated quickly, used a CC12F2concentration 3-4 times greater and a space velocity approximately 0.5 that of the latter [69]. The influence of water in the feed also had an important effect as the extent of deactivation was significantly reduced by the presence of water. The decomposition of HCFC-22 (CHClF2) by a series of acidic single and dual component metal oxides has been studied by Li et al. [70]. Initial studies over single metal oxides showed the order of reactivity; Ti02> Z r 0 2> Cr203 > W03 Ti02 was the most active, in agreement with other catalyst systems for the decomposition of fluorinated VOCs [67,68], producing 97.3YOconversion at 400°C with 84.1YOselectivity towards COX. ZrOz was marginally less active but only showed 36.2% COX selectivity at the same temperature. The range of dual component oxides tested included Zr02/Ti02, Cr203/Ti02, Cr203/Zr02, Co304/Zr02,V205/Co304,Zr02/V205, Zr02/MnO2, V205/Mn02,Co304/Mn02 and Zr02/W03.The highest activity was shown by Zr02/Ti02,Cr203/Ti02and Cr203/Zr02producing at least 85% CHClF2 conversion at 350°C and >99% at 400°C. The addition of water vapour increased the conversion and selectivity to COX whilst decreasing the selectivity to CHF3. It was proposed that water promoted the removal of fluoride ions from the catalyst surface which was then more active for the decomposition reaction. Time on line studies of Cr203/Zr02 in the presence of water indicated that conversion decreased marginally over 50 hours operation, whilst initially C 0 2selectivity decreased but levelled at ca. 70%

2: The Destruction Of Volatile Organic Compounds

133

after 20 hours. In the absence of water, CHClF2 conversion also decreased slightly with time, however, a dramatic increase in C 0 2selectivity accompanied by an increase in CHF3 selectivity was observed, such that after 50 hours operation CHF3was the predominant product with ca, 60% selectivity. Powder X-ray diffraction studies of the Cr203/Zr02fresh catalyst identified a Cr203 phase, whilst after use in the absence of water the diffraction intensity of Cr2O3 was reduced and intense peaks from ZrF4 were identified. For the catalyst tested in the presence of co-fed water Cr2O3 was unaffected and only relatively small ZrF4diffraction peaks were present showing that water in the feed suppressed the transformation of the oxide to a fluoride phase maintaining catalyst activity. Nagata et al. [24] have studied the decomposition of a range of CFCs including CFC-113 (CF2C1CFCl2), CFC-114 (CF2ClCF2Cl) and CFC-115 (CF3CF2Cl)in the presence of hydrocarbons, which were also oxidised, by a series of acidic metal oxides and supported metal oxides. The relative order for CFCs destruction was; CFC-113 > CFC-114 > CFC-115 and was related to the number of chlorine atoms in the substrate as destruction was successively more difficult as the number decreased. The most active catalyst was y - A l 2 0 3 showing higher conversions than zeolites Y, L, mordenite, ferrierite and ZSM-5 and the mixed oxide Si02/A1203.100% conversion of CFC-113 and CFC-114 over y-A1203 in the presence of n-butane were reported at 450°C and 500°C respectively, whilst CFC-115 conversion was 45.1% at 600°C. The only carbon products were CO and C02. CFC-115 conversion over y-A1203increased as the partial pressure of ethane, propane and n-butane increased but no increase was observed when the partial pressure of methane was altered. Supporting chromium, cobalt, zinc, molybdenum, cerium, vanadium and tungsten oxides on y-Al203 enhanced the activity for CFC-115 destruction, vanadium and tungsten were the most effective but only increased conversion to 59.0% and 60.0% respectively at 600°C. The y-A1203 catalysts did show some deactivation during the first two hours on stream ,however, after 4 hours they were reported to show stable activity which is promising but a considerably quantity of CFC-115 was not destroyed and activity must therefore be improved further. The catalytic oxidation of dichloromethane was investigated by Van den Brink et al. [71], using y-A1203, which is commonly employed as a support of noble metals for catalytic oxidation. Studies used a combination of flow and infrared spectroscopy experiments over a range of reaction temperatures. This paper is interesting as it provides a comparison with many of the studies using alumina supported catalysts, and it demonstrates that alumina is not a passive component in many chlorinated the VOC oxidation reactions. Studies were performed using Teflon coated tubing to prevent reactions between the substrate and vessel walls. 0.3 g of catalyst was used with a gas flow of 50 ml min-' with a composition of 1000 ppm CH2C12,89% He, 10% 0 2 and 1% H20. from a baseline temperature of 200°C a heating rate of 5" min-' was applied up to the desired reaction temperature and the system was allowed to stabilise for 20 min before effluent gases were analysed.

134

Catalysis

Initially dichloromethane was introduced to the reactor at 300°C and after 10 min on line no dichlormethane was eluted from the catalyst bed and CO was the only reaction product detected. CO remained the only product for a further 25 min on line. After this time the CO MS signal was reduced and CH3Cl was also detected in the reactor effluent along with unreacted CH2C12. The conversion of CH2C12was 92%. HCl was only detected after several hours on stream once the y-A1203surface was saturated with chorine. Above 350°C C02 was produced, but CO still remained the dominant product. The carbon balance for these experiments decreased from 95% at 325°C to 60% at 500°C over a period of 30 h. A Darkening of the catalyst was also observed and was attributed to the deposition of carbonaceous material on the surface. Water was also important in the reaction, with 1YOin the feed the temperature required for 50% conversion of CH2C12was increased from 270 to 320°C when compared with no co-fed water. Through FTIR studies it was suggested that there is a strong interaction between CH2C12 and adsorbed O H groups on the catalyst surface, and dichloromethane formed hydrogen bonds with the OH groups. The authors identified two types of species present on the surface when CH2C12 reacted with the y-A1203 at 250°C. They propose a reaction scheme (figure 6) in which the first step was the displacement of a chlorine atom from dichloromethane at the alumina surface to form a chloromethoxy species (1). Further interaction resulted in the loss of another chlorine to form species (2) which is in equilibrium with the chemisorbed formaldehyde analogue (3).

c1

I

A1

/ \

Figure 6

c1

/A
A1 \

Proposed scheme for dichloromethane destruction on a y-AI2O3catalyst [71]

2: The Destruction Of Volatile Organic Compounds

135

It is suggested that the chloride species displaced can combine with a proton to form HCl which subsequently reacts with the alumina causing partial chlorination of the surface. The species (2) and (3) can undergo further reaction leading to destruction products. The authors also suggest that water regenerates surface hydroxyl groups and prevents chlorination of the surface. The addition of water to the feed stream increased the conversion of dichloromethane, thus supporting the mechanism. Interestingly no HCl was observed for several hour after the start of the experiment suggesting that the regeneration was not a facile process. The reaction of chlorinated surface with O2to produce an A1-0 surface and C12 has been previously reported [72,73], but has been discounted due to the different conditions under which the experiments were performed. Finally the authors noted that the alumina catalysts tended to favour the formation of CO rather than C 0 2and that deactivation did occur. A study detailing the use of first row transition metal oxide supported on carbonaceous materials for the oxidation of hexane, butane and toluene has been reported by Drago et al. [74]. Catalysts were prepared by incipient wetness impregnation of a series of carbon supports by solutions of the metal nitrate salts. Carbonaceous supports Kureha (Kureha Chemicals Japan) and AX21 (Anderson Development Co.) were unsuitable as supports as both produced catalysts which deactivated and were reduced in mass during use. This was particularly prevalent for the MnO/AX21 catalyst which lost 44% mass on use. The best supports were Ambersorb0563 and Ambersorb0572 (Rohm and Haas Co.) as these were thermally stable during the reaction period. The best catalysts were specifically MnO/A572 and CoO/A572, which showed conversions of 96.5% and 98.9% respectively at 250°C for 2000 ppm n-hexane after 70 hours time on line, although it must be noted that a relatively low space velocity of 180 cm3g-'h-' was used. Ambersorb0572 catalyst containing ZnO were considerably less active after time on line, although the initial activity was comparable the manganese and cobalt based catalysts, this deactivation was due to the deposition of coke. Prolonged use of the CoO/A572 catalyst indicated that after 80 hours deactivation commenced, regeneration in air at 200°C for 20 hours restored activity, which again decayed as the combustion process proceeded. Further regeneration in air increased the activity although the initial activity of 99 YOconversion was not reproduced, it is also important to note that continuous cycling through the regeneration process did not reduce the catalyst mass via loss of carbon. The authors highlight the importance of metal oxide support interactions and suggest that the support stabilises the cobalt oxide in an oxidation state which is highly efficient for oxidation of the hydrocarbon before it is re-oxidised to the initial active species. Evidence for this hypothesis has been drawn from earlier studies of styrene production from ethyl benzene over the same catalysts [75], as it has been shown that the support can act as an electron source or sink facilitating oxidation and reduction of the metal oxide. A possible reaction mechanism based on magnetic susceptibility and XPS studies has been proposed for the oxidation of hydrocarbons over CoO/A572 (figure 7). The initial step involves oxygen absorption onto the cobalt oxide phase forming either (0$2species (A), or two metal 0x0 species (B), which then require

+

136

Catalysis

A572

A

B

Figure 7

Proposed reaction mechanism for the oxidation of hydrocarbons by CoOIAmbersorb85 72 [741

two electrons either from the carbon support or other Co ions in the cluster. Oxidation of the hydrocarbon then takes place with reduction of the Co'*' species to Co". The stabilisation effect of the support towards Co" thus promotes the oxidation cycle. The partially oxidised hydrocarbons produced by the reaction scheme are activated towards further oxidation via a sequence of similar reactions eventually leading to the production of carbon dioxide. It must be noted however, that if such a sequential mechanism was in operation it may be expected that some partially oxygenated products may be produced especially considering the relatively low reaction temperatures employed. We have investigated the oxidative destruction of a range of chemically diverse VOCs by uranium oxide based catalysts. The destruction of 1% benzene and butane in air was studied at a high space velocity of 70,000 h-' [76]. The oxide u308 was found to be a particularly active combustion catalyst for benzene destruction, showing 99.9% conversion at 400°C. u308 was also active for butane combustion, although activity was lower with a conversion of ca. 80% at 600°C. A silica supported uranium oxide catalyst was synthesised by impregnation of fumed silica by uranyl nitrate solution, followed by drying and calcination ultimately at 800°C. Powder X-ray diffraction studies identified U308as the supported phase and in-situ diffraction studies showed that the phase was stable up to 600°C in humidified air. The supported catalyst demonstrated equivalent activity for benzene combustion as the unsupported variant, however, a significant hysteresis in conversion was observed. The U308/Si02catalyst only exhibited trace conversion at 350°C and 99.9 at 400°C when the temperature was steadily increased from lOO"C, whilst decreasing the furnace temperature below

2: The Destruction Of Volatile Organic Compounds

137

400°C did not results in loss of activity, on the contrary a conversion >92% was maintained at 200°C. The supported catalyst was also significantly more active for butane destruction than U308,as 99.9% conversion was achieved at 500°C. The catalyst performance for butane combustion was further improved by the addition of 1 mol% copper and chromium oxides, the effect was most marked for the chromium oxide doped catalyst which converted 93% butane to carbon oxides at 400°C. Doping the catalyst with copper oxide also had a beneficial effect by increasing the C 0 2 selectivity to greater than 99%. CO oxidation studies revealed that the CuO/U308/Si02catalyst was the most active, showing rates at least one order of magnitude greater than other materials. The addition of the copper component increased C 0 2yield by sequential oxidation of CO. Catalysts based on uranium oxide are also particularly active for the destruction of the chlorinated VOCs chlorobenzene and chlorobutane [77]. Both were destroyed by u308 at 350°C and 70,000 h-' space velocity, showing 99.7% and >99.5% conversions respectively. Time-on-line studies for the destruction of 0.12% chlorobenzene at 450°C showed that the catalyst was not deactivated as 99.9 YO conversion was maintained during 400 hours continuous operation. These catalysts were also active for the oxidative abatement of other VOCs and it has been demonstrated that toluene, butylacetate and cyclohexanone can also be destroyed at relatively low temperatures. Considering the high space velocities employed in these studies, uranium based catalysts are amongst some of the most active oxide catalysts investigated for VOC destruction. The combustion of acetaldehyde and trimethylamine, which are common VOC odour pollutants, have been investigated over a series of mainly metal oxide based catalysts [78]. Catalysts were supported on y-A1203 and were all prepared by impregnation of the nitrate solution with the exception of the Pt system which was prepared using H2PtCls.Catalyst activity was screened using a GHSV = 30,000 h-' with 50 ppm VOC the balance being air. The order of activity for acetaldehyde combustion was:

-

Ag > Mn203 CuO > PdO > Pt > Fe2O3 > NiO > Co304 The activity of Ag was significantly better than all other catalysts showing 90% conversion at 200"C, whilst the best oxides were Mn2O3 and CuO both showing 87% conversion at 300°C. Trimethylamine was oxidised at slightly lower temperatures and followed the order of activity:

-

Ag = Mn2O3 = Pt > PdO CuO > Fe2O3 > Co304> NiO Based on these results a series of Ag/Mn203/y-A1203catalysts with varying Ag content were prepared by co-impregnation to a total loading of 2 wt.%. The addition of Ag improved the activity of both supported and bulk Mn2O3, although after 2 hours calcination Ag/y-A1203 was still more active than Ag/Mn203/y-A1203.However, when the catalysts were calcined for 22 hours the activity was approximately equal and after 42 hours calcination Ag/Mn203/yA1203was more active. Surface area determination showed that these effects were not dependent on the catalyst surface area. The combustion activity of Ag/Mn203/y-A1203 was not significantly effected by the presence of water vapour

138

Catalysis

in the feed. Oxygen TPD studies indicated that surface oxygen species were ca. 2.9 greater on the Ag/Mn203based system compared to those on Mn2O3, and it was these surface oxygen species which were important for the enhancement of combustion activity. The incorporation of metals with oxide based catalysts have also been investigated, Vassileva and co-workers have studies the addition of Pd [lo] and Ag [79] to 30 wt. % V2O5 supported on y-A1203. Both catalysts were tested for the destruction of benzene at 300 h-' space velocity, the addition of Pd and Ag both promoted the catalyst activity. The promotional effect was due to the activation of oxygen by the metal component and modification of the vanadium redox properties. Results from ESR spectroscopy and X-ray diffraction studies were consistent with the proposal that V4+ species within the V2O5 lattice were responsible for the catalyst activity. The most active catalysts were those with high metal dispersion, particularly in the case of Ag, as the activity was reduced when Ag was chemically bound to vanadium oxide and metallic Ag phases were present. Additional studies of the Pd/V205/y-A1203 [SO] system have identified the phases during use and it has been proposed that the efficient delocalisation of electrons in V4+ion clusters in these phases facilitated the redox cycle important for oxidation. The addition of metals to oxide catalysts, which are already active for combustion processes, is an interesting approach which attempts to combine the beneficial aspects of both types of catalyst system. The incorporation of the metal component generally increases the activity, however, the destruction of halogenated VOCs have not been investigated and problems associated with deactivation may take place. The oxidative destruction of methylene chloride has been studies by Jiang et al. [S 13 over sulfated oxide catalysts, consisting Ti02/S04, ZrOz/S04 and Ce02/S04,prepared by sulfuric acid impregnation of the oxides. Experimental conditions for catalyst testing used a reactant stream of 959 ppm methylene chloride in air at GHSV = 2,210 h-'. Ti02/S04was the most active catalysts showing complete CH2C12conversion at 275"C, at equivalent temperature ZrOz/S04and Ce02/S04showed conversions of 90.6% and 82.9% respectively. HCl was the sole chlorine containing product, although CO was a major reaction product showing selectivities in the range 86-89%, the balance was C02. Determination of the oxygen adsorption capacity and acidity, by desorption of NH3, showed a direct relationship with catalyst activity suggesting that both factors may be important for catalytic activity. The addition of 2 vol.% water to the reaction feed suppressed catalyst activity markedly from 100% conversion to 50.1YOfor Ti02/S04at 275°C. In an attempt to improve the low C 0 2selectivity shown by the sulfated materials a bifunctional catalysts consisting Ti02/S04 with 5% CuO was developed. The addition of CuO did produce a beneficial effect, increasing the C02 selectivity from 12% to 60% at 275°C. The authors consider that one possible mechanism for the increase in C 0 2 selectivity was sequential oxidation of CO by the CuO component which is a similar conclusion shared by the present authors over uranium oxide catalysts for the oxidation of butane and benzene [76].

2: The Destruction Of Volatile Organic Compounds

139

The destruction of benzene by Cu/Cr and Co/Cr mixes oxide systems supported on y-A1203 and y-A1203/Si02have also studies by Vass and Georgescu [82]. The catalyst precursors were prepared by co-precipitation of 1:1 molar ratios of the metal nitrates in aqueous solutions containing tartaric acid. The supported catalyst was prepared by two methods, the first involved solubilisation of the precursor followed by impregnation of the support with the precursor solution. The second involved formation of the precursor on the support by successiveimpregnation with tartaric acid and then metal nitrate solutions. After drying the precursors were calcined at 700°C for 6 hours prior to use. The catalysts were tested for the combustion of 5% benzene at 4,000 h-' space velocity and 1% benzene at 10,000 h-'. Under both conditions the catalysts prepared by successive impregnations were more active, and irrespective of the preparation method the y-Al203 catalyst was more active than that supported on y-A1203/Si02.The Cu/Cr system was more active than Co/Cr, the best performance was shown by the Cu/Cr/y-A1203catalysts which showed 100% conversion of 1% benzene at 320°C and 10,000h-'. The authors report that in the region of 70-80% conversion traces of maleic anhydride were detected and concluded that it was sequentially combusted as at higher conversions C 0 2was the sole reaction product. Kang and Wan [83] have also studied the combustion activity of y-A1203 supported chromium and cobalt oxide catalysts, with particular reference to the effects of acid and base additives. These catalysts were found to be active for the combustion of ethane and activity was increased by the preparation of a mixed chromium/cobalt supported oxide phase. This enhancement was attributed to the release of lattice oxygen from the binary oxide being a more facile process than the singular oxides. The addition of a base additive, potassium, to Cr/Co/yA1203 reduced the ethane conversion whilst increasing the C 0 2 selectivity to 100%. Conversely the addition of Si02,an acidic component, enhanced ethane conversion and reduced C 0 2selectivity. It was therefore considered that initial ethane activation was via C-H bond breaking on Br-nsted acid sites of the Si02/A1203.Whilst the base additive enhanced the adsorption of oxygen which is involved in the formation of COz, the reverse effect on C02 production is observed for the acid additive. Agarwal et al. [SS] have investigated the long term activity of a commercial chromia-alumina catalyst (ARI. Technologies Ltd.) for the destruction of a mixture of C1-C2chlorinated VOCs and a mixture of Cs-C9hydrocarbons with 50 ppm trichloroethylene. Catalyst deactivation was studied at constant conversion, increasing the temperature when necessary to compensate for loss of intrinsic catalyst activity. Over 153 days in a fixed bed reactor at 23,970 h-' space velocity no increase in temperature was required for the oxidation of the chlorinated feed stream although CO selectivity gradually increased indicating that deactivating to place to some extent. Activity in a fluidised bed showed no deactivation under the same conditions, it was concluded that physical attrition and loss of chromium, via the oxychloride, were beneficial by continually exposing fresh chromium catalysts. Although beneficial to catalyst activity the loss of chromium by volatile product formation has further environmental conse-

Catalysis

140

quences and release to the environment must be avoided. Destruction of the mixed chlorinated/hydrocarbon stream indicated that benzene and trichloroethylene were the most difficult to destroy and the temperature had to be increased from 385°C to 418°C over 210 days to maintain >99YOconversion. This decrease in the activity of the catalyst was due to a decrease in the catalyst surface area which decreased by approximately 20% for the mixed stream whilst remaining unchanged for the solely chlorinated stream. Manganese oxide catalysts promoted with transition metal oxides, in particular CuO, have long been established as effective oxidation catalysts for carbon monoxide [85]. The same type of catalyst has also been applied to the oxidation of VOCs. A detailed kinetic study of the combustion of acetone over a supported hopcalite (CuMn204)catalyst has been reported by Linz and Wittstock [86]. Studies monitored the reaction products along the length of a fixed catalyst bed showed that partially oxidised products, particularly acetaldehyde, were significant products within the bed, although at the bed exit only ca. 2% of the carbon products were partially oxidated the balance was COZ. The same type of behaviour was exhibited during the oxidation of isopropyl alcohol and butyl acetate and the oxidation process of the VOCs can be described by a simple parallelconsecutive reaction network (figure 8). A series of rate equations for the reaction network were expressed and rate constant calculated from curve fitting the experimental data. The derived rate constants obeyed Arrhenius behaviour and by using the kinetic data the reactor can be designed specifically so that the outlet concentration of the pollutant VOC and partially oxidised by-products can be reduced below legislative limits. Klissurski et al. [87] have examined the combustion of acetone, toluene and styrene by zinc-cobalt spinel oxides supported on alumina. Catalysts were prepared by co-precipitation with sodium carbonate from a mixed zinc/cobalt nitrate solution at pH 9. The supported catalyst was prepared by deposition of the precursor on y-A1203 from a suspension in dimethylformamide and water. The supported precursor was dried at 150°C and calcined at 300°C to produce the catalyst. The bulk and supported catalysts both showed the formation of zinc cobaltite spinel structures which were thermally stable. Microreactor studies at 15,000 h-' space velocity showed that the components of a mix of acetone, toluene and styrene were destroyed at 225"C, 280°C and 350°C respectively. The VOC concentrations were not specifically expressed but it is assumed that they

acetaldehyde acetone

'14

partially oxidied products (PO) Figure 8

Simple parallel-consecutive reaction network describing the oxidation of VOCs by a supported hopcalite catalyst [86]

2: The Destruction Of Volatile Organic Compounds

141

are within the range 0.5-3.0 gm-3. The catalyst showed no deactivation or partial oxidation products and remained active when the space velocity was varied in the range 7,500-30,000 h-'. Activity determination in a semi-large scale pilot plant installation at 10,OOO h-' showed similar trends, showing 99.9% acetone conversion at 180°C and conversions of 99.8% at 280°C for toluene and 320°C for styrene. A mix of VOCs containing acetone, ethyl acetate, benzene, toluene and styrene were also efficientlycombusted at 320°C in the pilot plant tests and it is evident from these studies that oxide catalysts based on cobalt spinels demonstrate high combustion activity. An interesting development in the destruction of VOCs, in particular chlorinated compounds, has been the development of catalysts based on copper and chlorine [88,89]. Catalysts were prepared by incipient wetness impregnation of a silica support by metal chloride and KCl solutions. A wide range of metal chlorides, including CuCl, ZnC12, FeC12, MnC12, CoC12, SnC12, NiC12, SbC13, MoClS, and CdC12 were prepared and tested for the destruction of methylene chloride (10.000 ppm) at 300 h-' space velocity. CuC1/KCl/SiO2 was the most active of the chloride catalysts, showing 100% conversion at 350"C, this compared to 38.0% for MnCl2/KC1/SiO2which was the best performance after the CuCl based catalyst. Comparison with the oxides Cr2O3, CuO, Co304and M n 0 2 supported on Si02showed that CuC1/KCl/SiO2 was more active and although CrzOJSi02 activity was only slightly lower the selectivity to COX from CuC1/KCl/SiO2was significantly better than the oxide catalysts which produced appreciable quantities of non-combustion products. After use the Cr203/Si02 catalyst showed an 18% decrease in the Cr content, probably due to the formation of volatile Cr02C12,whilst analysis of CuC1/KCl/SiO2 showed no decrease in Cu content or loss of activity. Kinetic investigations of methylene chloride oxidation indicated that the reaction was first order with respect to CH2C12and zero order with respect to O2 above 0.15 atm partial pressure, pulsing experiments with CHzC12and O2 indicated that significant amounts of active oxygen were stored by the catalyst during 0 2 pulses. It has also been shown that the catalyst can function in a redox cycle mode and evidence from TPR studies suggests that the copper (11)oxychloride species, CuO. CuC12, was the active component for CH2C12destruction. The oxidation of deuterated methylene chloride showed a kinetic isotope effect, kH/kD = 1.48, suggesting that C-H bond breaking was an important procedure in the rate determining step. The CuC1/KC1/SiO2and CuC1/Si02 catalysts have also been used to destroy a range of other VOCs including 1,2 dichlorobenzene, carbon tetrachloride and ethylene oxide which were all completely destroyed within the temperature range 300-500°C.These catalysts based on copper and chlorine show high activity for the destruction of a range of VOCs, the results presented here were all obtained at a relatively low gas hourly space velocity of 300 h-', it would therefore be interesting to investigate these catalyst at higher flow rates which would more applicable for industrial applications. The use of zeolite catalysts in the total oxidation of VOCs has been previously highlighted [58-591, zeolite based catalysts have particularly been used for the destruction of a range of chlorinated organics. Early studies by Chatterjee and

142

Catalysis

Greene [90] investigated the oxidative destruction of methylene chloride by zeolite Y in the H form and ion exchanged with cerium and chromium. The zeolite was supported on a low surface area cordierite honeycomb monolith by application as a wash coat. Below 425°C the catalytic activity decreased in the order; Cr-Y > H-Y > Ce-Y however, above 425°C the activity was similar, with all catalysts showing >90% conversion, major reaction products were CO and HCl. The addition of 2.7 vol.% water into the reaction stream suppressed the conversion, particularly below 400"C, and was most marked with the H-Y and Ce-Y catalysts, this effect was reversible and it was considered due to active site blocking by the water molecule. The oxygen uptake capacity and acidity for the zeolites correlated directly with activity and it was considered that the exchanged metal cation determined these parameters and ultimately the activity. A dual site oxidation mechanism involving the adsorption of methylene chloride at Br-nstedacid sites and oxygen adsorption on the metal cation was feasibly proposed. The use of metal exchanged zeolites has been investigated further by Chatterjee et al. [91] to include other transition metals such as cobalt and manganese in addition to chromium. The catalysts were prepared as previously described [90] and were tested for the destruction of methylene chloride, trichloroethylene and carbon tetrachloride in the presence of 1.3 vol.% water. Co-Y showed superior performance producing complete conversion of methylene chloride and carbon tetrachloride at 350°C and 200°C respectively, whilst conversion of trichloroethylene was not observed below 400°C. Impregnation of the transition metal exchanged zeolites with chromium solution resulted in the formation of Cr2O3 in the matrix, significantly improving the destruction of trichloroethylene to >90% at 325°C. Partially oxygenated products were not detected and the major reaction products were CO, C 0 2 and HCl, 100% C02 selectivity was observed for carbon tetrachloride oxidation over all catalysts. However, for the oxidation of methylene chloride and trichloroethylene, CO selectivities for the ion exchanged catalysts were high, CO2/CO ratios were in the range 0.02-0.05, the added Cr2O3 improved C 0 2 selectivity but the C 0 2 / C 0 ratio was still only 0.2-0.3. A more detailed study investigating the effects of catalyst composition on cobalt exchanged chromium impregnated zeolite Y has been reported [92]. Increasing the degree of cobalt exchange increased the acidity and oxygen uptake of the catalysts without effecting the surface area. Increasing the Cr2O3 impregnation caused a decrease in acidity by dealumination and higher loadings reduced the surface area by structural loss and pore/channel blocking. The destruction of methylene chloride was improved by increased Co exchange, whilst increasing Cr203loading was detrimental, on the other hand trichloro-ethylene destruction increased with increasing Cr203and to some extent Co exchange. The authors concluded that methylene chloride destruction was essentially a function of the acid sites, forming a carbonium ion which was subsequently oxidised by oxygen adsorbed at the cobalt cation sites. The presence of Cr203was detrimental as a consequence of the reduced surface area and acidity. Conversely, trichloroethy-

2: The Destruction Of Volatile Organic Compounds

143

H\c/H

CI

' v I 'Cl

0-

I

H+

co2+

Al /o\Si

Al /*\Si

I t

HCI

Figure 9

co3+

CO

H

I

Al /'\Si

"7'

c---

Reaction mechanism proposed for methylene chloride oxidation by Cr203/Co-Y [931

lene conversion was strongly dependent on the chromium sites although the acidic and cationic sites were also influential in the destructive sequence. Surface reaction mechanisms based on evidence from in situ transmission FTIR studies, have been proposed for the destruction of methylene chloride and carbon tetrachloride over Co-Y loaded with Cr203[93]. The proposed reaction mechanism for methylene chloride destruction is shown in figure 9. A similar type of mechanism was also proposed for carbon tetrachloride oxidation, important steps in both mechanisms were the adsorption of the chlorinated hydrocarbon on the Br-nsted acid sites and dissociative adsorption of molecular oxygen on Co cation sites. The formation of CO and C02 products were via parallel reaction pathways, with little or no COz formed by sequential CO oxidation. The formation of phosgene as a reaction intermediate has been identified during CCl, oxidation at space velocities above 15,000 h-' in a wet feed stream and as low as 2,400 h-' in a dry stream. The selectivity of phosgene also increased with time on line as the catalysts deactivated from 99% conversion to ca. 10% conversion after 22.5 h use. The use of zeolite-based catalysts for VOC oxidation, in particular chlorinated VOCs, has been demonstrated at low temperatures, however, they deactivate relatively quickly and will therefore require regular regeneration, which may prove impractical for commercial operation. The very low selectivity towards C 0 2and the formation of potentially hazard by-products such as phosgene also needs to be addressed. 4.5 Mixed Catalyst/Sorbent Systems. - The process efficiency of catalytic oxidation may be enhanced by incorporating a sorbent component in the system. In conventional catalytic oxidation acid products such as HCl and C12 produced

144

Catalysis

during chlorinated hydrocarbon destruction must be scrubbed from the effluent stream before emission to the atmosphere. The use of a catalyst-sorbent system simultaneously destroys the VOC substrate and captures the halogenated oxidation products negating the need for downstream scrubbers. Stenger et al. [94] have described the use of such catalysts for the destruction of trichloroethylene, trichlorofluoromethane and toluene in the concentration range 30-350 ppm at 6,700 h- space velocity. The catalyst comprised copper and manganese oxides, which were the catalytic components, supported on the sorbent component sodium carbonate. All the VOCs were completely destroyed in the temperature range 250-400°C. Powder X-ray diffraction showed that sodium carbonate captured the chlorine products and was converted to sodium chloride with subsequent release of COz,this was confirmed by X-ray photoelectron studies. A chlorine balance determined during trichloroethylene destruction indicated that 98% of the chlorine was retained by the sorbent. Catalysts lifetime testing showed that there was no trichloroethylene in the effluent, however, dichloromethane was a by-product and increased slightly in concentration from 1 ppm to 5 ppm over 5 days operation. Theoretically the lifetimes of such catalyst-sorbent systems are limited and lifetime will be a function of the flow rate, VOC type and concentration. A different type of catalyst-sorbent system based on transition metal exchanged zeolites for the destruction of chlorinated VOCs has been investigated by Greene et al. [95]. Chromium exchanged Y and ZSM-5 based zeolites were used to physisorb VOCs from a humid air stream at ambient temperatures, the upper portion of the bed was then heated to catalytically active temperatures, typically 300"C, whilst the lower portion was slowly temperature ramped to desorb the VOCs which were catalytically destroyed. A gas hourly space velocity of 2,400 h-' was used and Cr-ZSM-5 was found to have the highest activity for simple VOC oxidation, 9 5 YO conversion of trichloroethane and dichloromethane at 300"C, and also had a relatively high sorption capacity. A process which alternatively trapped and catalytically destroyed 110 ppm trichloroethylene using a Cr-ZSM-5 sorbent/catalyst has been described. The cycle time was determined to be ca. 24 hours and comparison of the relative energy use showed a 93% reduction compared to a conventional reactor as the sorbentcatalyst process only required heating for 7% of the cycle time. 4.6 Comparison of Noble Metal and Oxide Catalysts. - A few studies have directly compared the activity of noble metal and oxide based catalysts. The combustion of a range of Cs-C9hydrocarbon VOCs in humidified air by 0.1% Pt/3 YONi/A1203and ceria promoted hopcalite commercial catalysts has been compared [96]. The Pt based catalyst showed no deactivation during 253 continuous operation, whilst over 297 days the temperature of hopcalite catalyst required an 85°C increase to maintain > 99% conversion. However, the final operating temperature of the hopcalite catalyst was 4OO0C, 30°C lower than the isothermal operating temperature of the Pt system. A first order concentration deactivation model was developed and predicted a 362 day lifetime for the

GHSV/h-'

70000 70000 70000 15000 10500 10000 10000 9500 6700 5900 5900 5900 5900 5900 5000 5000 3600 3600 3600 3600 3200 2400 2400 2400 2400 2361

Temp. / "C

400 350 450 400 400 325 390 430 270 500 500 500 500 500 500 500 400 650 400 400 130 350 350 350 350 325

voc(s)

benzene chloro benzene butane mixture of Cs-C9hydrocarbons CFzClz benzene benzene mixture of C5-C9hydrocarbons trichloroet h ylene dichlorofluoro methane dichlorofluoro methane dichlorofluoro methane dichlorofluoro methane dichlorofluoro methane C,CI,F, C2C13F3 1,2- dichloroethane trichloroethane 1,2- dichloroethane 1,2- dichloroethane benzene, toluene and xylene

methylene chloride methylene chloride trichloroethylene trichloroethylene trichloroethane

Comparison catalyst activties for combustion of VOCs

Cu/U/Si02 Ce/Cu/Mn02 Ti02 Cu-Cr/y-A120, Co-Cr/y-A1203 0.1%Pt/3 %Ni/A1203 Cu/MnO/Na2C03 mordenite A Si02/A1203 Zr02/Mo03 Ti02-SiOz ZrP207 Pt/Zr02 Pt/ZrOz-(PO,) Ti02/Si02 Cr203 zeolite Y mordenite B Pt/ hydrophobic fluorinated carbon (60% fluorine) co-Y H-Y Cr-ZSM-5 Cr-Y co-Y

U308

U308

Catalyst

Table 3

1406 996 1098 1023 1500

10000 10000 10000 500 2000 10000 10000 500 31 6000 6000 6000 6000 6000 11000 11000 10000 10000 10000 10000 45

Inlet

99.9 99.7 95.0 99 98.0 100 100 99 99 100 99.2 95.7 95.7 95.3 99 98 100 99.7 98.5 95.3 95 100 100 99.4 98 100

<10 31 500 5 40 0 0 5 1 0 5952 258 258 282 110 220 0 30 150 470 3 0 0 7 21 0

Conversion

/%

Outlet

Concentration / ppm

93 93 95 95 91

77 77 77 96 69 82 82 96 97 68 68 68 68 68 98 98 66 67 66 66 38

Re$

350 350 250 250 300 530 600 500 500 400 400 400 400 350 275 275

methylene chloride methylene chloride benzene benzene dimethylmethyl phosphonate

chlorobenzene CF3CFZCI dichlorofluoro methane dichlorofluoro methane 1,2clichloroethane 1,2- dichloroethane 1,2- dichloroethane 1,2- dichloroethane methylene chloride methylene chloride methylene chloride

CuCI/KC1/SiO2 Cr203/Si02 MnO/ Am bersorb@572 COO/ Ambersorb@572 Cu2Ca8(P04)6(0H)2

0.1YOPt/A1203 W03/A1203 mordenite B zeolite Y mordenite A ZSM-5 mordenite C Si02/A1203 Na-Y Ti02/S04 ZrOJS04

/Y-A1203

30000 15366 5900 5900 3600 3600 3600 3600 2400 2210 2210

300 300 184 184

10000 loo00 2000 2000 2.885 x lo5 mol/l 398 6700 6000 6000 10000 10000 loo00 10000 1716 959 959

10000 10000 loo00 96-98g/m3

540 540 540 330

400 400 400 400

methylene chloride 1,2-dichlorobenzene methylene chloride benzene

1500 1500 100

2361 1367 1120

350 175 150

methylene chloride carbon tetrachloride formaldehyde

Cr-Y co-Y 1%Pt/ hydrophobic fluorinated carbon (60% fluorine) CuC1/KCl/SiO2 CuCl/Si02 Cr03/KC1/Si02 0.5YOPd/30%V205

Inlet

98.4 96.7 99.9 99.6 100 92.0 60.0 92.2 86.4 90.1 85.6 72.9 51.6 83.5 100 90.6

32 2680 468 816 990 1440 2610 4840 283 0 90

33 24 68 68 66 66 66 66 93 81 81

88 88 74 74 100

88 88 88 10

100 100 97.4 100

0 0 260 0 160 330 2 8 0

91 91 39 100 100 100

0 0

Ref.

0

Conversion

/%

Outlet

Concentration / ppm GHSV/h-'

Temp. / "C

voC(S)

(cont.)

Catalyst

Table 3

c.. P

g 5.

o\

2: The Destruction Of Volatile Organic Compounds

147

148

Catalysis

hopcalite catalyst before the operating temperature increased above 500°C. The results of the former study indicate that the noble metal catalyst is more resistant to deactivation, however, many studies have found the reverse to be true [lS]. It is therefore evident that catalyst lifetime and deactivation are dependent on many factors which must be assessed individually for different catalysts in different applications.

5

Conclusions

Many different catalysts have been used for the vapour phase destruction of volatile organic compounds, the wide variety of systems is illustrated by the comparison of their performance in table 3. Two broad categories can be identified and these are supported precious metals and metal oxide based systems. Supported metal catalysts most commonly consist of platinum or palladium or a combination of both dispersed on a suitable support. In general these catalysts show high activity for combustion although they are susceptible to deactivation by poisoning from compounds containing sulfur, phosphorous, halogens or other metals. Metal oxide catalysts are usually transition metal oxides and are used in both supported and non-supported forms. Other oxides besides transition metals are also employed along with other types of catalysts such as halides, phosphates and zeolite based systems. Within the field of heterogeneous catalytic combustion for environmental applications there is considerable scope for catalyst improvement. New legislation is continually being introduced by national governments worldwide and this results in ever increasing demands for improved catalyst performance. It is important that high activity systems are developed which can treat a variety of compounds, such as those containing sulfur, phosphorous and halogens, and that such systems show high long term stable activity with resistance to deactivation.

6

2.

3. 4. 5. 6. 7. 8. 9.

References N. Mukhopadhyay, E.C. Moretti, Current and Potential Future Industrial Practices for Controlling Volatile Organic Compounds, American Institute of Chemical Engineers, Center for Waste Control Management, New York, 1993. M.S. Jennings, M.A. Palazzolo, N.E. Krohn, R.M. Parks, R.S. Berry, K.K. Fidler, Catalytic Incineration for the Control of Volatile Organic Compound Emission, Pollution Technology Review, (Noyes Ed.), No. 121, (1985). M.J. Molina, F.S. Rowland, Nature, 249, (1974), 810. Environmental Protection Agency, US Clean Air Act, 1990, USA. B.A. Tichenor, M.A. Palazzolo ,Environ. Prog., 6(3), (1987), 174. A.A. Zieba, T. Banaszak, R. Miller, Appl. Catal. A, 124, (1995), 47. T. Banaszak, R. Miller, M. Zembrzuski, JAPCA, 37, (1987), 1434. R. Miller, M. Zembrzuski, Staub., 46, (1986), 4. J.C. Mackie, Rev. Catal. Sci. Eng., 33, (1991), 169.

2: The Destruction Of Volatile Organic Compounds 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

149

M. Vassileva, A. Andreev, S. Dancheva, N. Kotsev, Appl. Catal. A, 49, (1989), 25. S.K. Gangwal, M.E. Mullins, J.J. Spivey, P.R. Caffrey and B.A. Tichenor, Appl. Catal., 36, (1988), 231. A. Musialik-Piotrowska, K. Syczewska, Environ. Prot. Eng., 15, (1989), 117. D.M. Papenmeier, J.A. Rossin, Ind. Eng. Chem. Res., 33, (1994), 3094. A.A. Barresi, G. Baldi, Ind. Eng. Chem. Res., 33, (1994), 2964. J.J. Spivey, J.B. Butt, Catal. Today, 11, (1992), 465. R.K. Sharma, B. Zhou, S. Tong, K.T. Chuang, Ind. Eng. Chem. Res., 34,1995,4310. G.I. Golodets, Heterogeneous Catalytic Reactions Involving Molecular Oxygen, Studies in Surface Science and Catalysis, Elsevier, Amsterdam, 15, (1982). J.J. Spivey, Ind. Eng. Chem. Res., 26, (1987), 2165. J. Volter, G. Lietz, H. Spindler, H. Lieske, J. Catal., 104, (1987), 375 W.R. Patterson, C. Kemball, J. Catal., 2, (1963), 465. R. Prasad, L.A. Kennedy, E. Ruckenstein, Catal. Rev. Sci. Eng., 26, (1984), 1. D. Brennan, D.O. Hayward, B.M. Trapnell, Proc. Roy. SOC.(London), A256, (1960), 81.

23. 24. 25. 26.

J. C. Lou and S. S. Lee, Applied Catalysis B: Environmental 12 (1997) 111-123 H. Nagata, T. Takakura, S. Tashiro, M. Kishida, K. Mizuno, I. Tamori, K. Wakabayashi, Appl. Catal. B, 5, (1994), 23. G.M. Bickle, T. Suzuki, Y. Mitarai, Trans. IChemE., 70B, (1992), 44. G.W. Cordonna, N. Kosanovich, E.R. Becker, Platinum Metals Review, 33, (1989), 46.

27. 28. 29. 30.

A.I. Dalton, S. Sircar, U.S. Patent 4,346,019, August 24, (1982). C.F. Cullis, B.M. Willatt, J. Catal., 83, (1983), 267. R. Shelef,G.W. Graham, Catal. Rev. Sci. Eng., 36, (1994), 433. M. Haruta, A. Ueda, S. Tsubota, R.M. Torres-Sanchez, Catal. Today, 29, (1996), 443.

31.

M. Haruta, S. Tsubota, T. Kobayashi, A. Ueda, Y. Nakahara, J. Catal., 144, (1993),

32. 33. 34. 35.

F. Boccuzzi, A. Chiorino, S. Tsubota, M. Haruta, Catal. Lett., 29, (1994), 225. B. Mendyka, A. Musialik-Piotrowska, K. Syczewska, Catal. Today, 11, (1992), 597. M. Skoglundh, L.O. Lowendahl, J.E. Ottersted, Appl. Catal., 77, (1991), 9. J. R. Gonzalez-Velasco, A. Aranzabal, R. Lopez-Fonseca, R. Ferret, J. A. GonzalezMarcos, Applied Catalysis B: Environmental, 24 (2000) 33. N. Bose, S. M. Senkan, Combust. Sci. Technol. 35 (1983) 187 B-H. Jang, S-S. Lee, T-H. Yeon, J-E. Yie, Korean Journal of Chemical Engineering,

175.

36. 37.

15(5) (1998) 516.

45.

K.T. Chuang, S. Cheng, S. Tong, Ind. Eng. Chem. Res., 31, (1992), 2466. K.T. Chuang, B. Zhou, S. Tong, Ind. Eng. Chem. Res., 33, (1994), 1680. S.K. Agarwal, J.J. Spivey, D.E. Tevault, Appl. Catal. B, 5, (1995), 389. J.A. Rossin, M.M. Farris, Ind. Eng. Chem. Res., 32, (1993), 1024 G.C. Bond, N. Sadeghi, J. Appl. Chem. Biotechnol., 25, (1975), 241. G.R. Lester, International Patent Application WO 90/13352, November 1990. G.R. Lester, R.E. Marinangeli, Development and Characterisation of oxidation catalysts for air purification, Report to the US Army by Signal Research Centre Inc., CRDEC-CR-87050,1987. M.M. Farris, A.A. Klinghoffer, J.A. Rossin, D.E. Tevault, Catal. Today, 11, (1992),

46. 47.

501. L. Borko, Z. Schay, L. Gucci, Appl. Catal. A, 130, (1995), 157. C.P. Hubbard, K. Otto, H.S. Gandhi, K.Y.S. Ng, J. Catal., 144,(1993), 484.

38. 39. 40. 41. 42. 43. 44.

Catalysis

150 48. 49.

C.P. Hubbard, K. Otto, H.S. Gandhi, K.Y.S. Ng, J. Catal., 139,(1993), 268. A. Ishikawa, S. Komai, A. Satsuma, T. Hattori, Y. Murakama, Appl. Catal., 110, (1994), 61.

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. 79. 80.

K. Muto, N. Katada, M. Niwa, Appl. Catal., 134, (1996), 203. M. Niwa, K. Awano, Y. Murakami, Appl. Catal.,7, (1983), 317. R.F. Hicks, H. Qi, M.L. Young, R.G. Lee, J. Catal., 122, (1990), 280. K. Otto, J.M. Andino, C.L. Parks, J. Catal., 131, (1991), 243. L.O. Lowendahl, J.R. Otterstedt, Appl. Catal., 59, (1989), 123. B. Chen, C. Bai, R. Cook, J. Wright, C. Wang, Catalysis Today, 30 (1996) 15-20 T. Aida, R. Higuchi, H. Niiyama, Chem. Lett., (1990) 2247 M. A. Centeno, M. Paulis, M. Montes and J. a. Odriozola, Appl. Catal. A., 234 (2002) 65 S. Minico, S. Scire, C. Crisafulli, R. Maggiore and S. Galvagno, Appl. Catal. B, 28 (2000) 245. V.D. Sokolovskii, Rev. Sci. Catal. Eng., 32, (1990), 1. T. Seiyama, Catal. Rev. Sci. Eng., 34, (1992), 281. T. Seiyama, N. Yamazoe, K. Eguchi, Ind Eng. Chem., Prod. Res. Dev., 24, (1985), 19. H. Arai, T. Yamada, K. Eguchi, T. Seiyama, Appl. Catal. A, 26, (1986), 265. G. Saracco, G. SchibiliaH. Ianibello, G. Baldi, Appl. Catal. B, 8, (1996), 229. T-R. Ling, Z-B. Chen, M-D. Lin, Catal. Today, 26, (1995), 79.

L.A. Isupova, V.A. Sadykov, S.F. Tikhov, O.N. Kimkhai, O.N. Kovalenko, G.N. Kustova, I.A. Ovsyannikova, Z.A. Doubii, G.N. Kryukova, A.Y. Rozouskii, Catal. Today, 27, (1996), 249. S. Imamura, Catal. Today, 11, (1992), 547. S. Imamura, H. Tarumoto, S. Ishida, Ind. Eng. Chem. Res., 28, (1989), 1449. S. Imamura, T. Shiomi, S. Ishida, K. Utani, H. Jinda, Ind. Eng. Chem. Res., 29, (1990), 1758. S. Karmakar, H.L. Greene, J. Catal., 151, (1995), 394. G-L. Li, I. Tatsumi, M. Yoshihiko, T. Yusaka, Appl. Catal. B, 9, (1996), 239.

R. W. van den Brink, P. Mulder, R. Louw, G. Sinquin, C. Petit, J-P. Hindermann, Journal of Catalysis, 180 (1998) 153-160 G. Clet, J. M. Goupul, D. Cornet, Bulletin Societe Chimie Francais, 134 (1997) 223 M. W. M. Hisham and S. W. Benson, Journal of Physical Chemistry, 99 (1995) 6194 R.S. Drago, K. Jurczyk, D.L. Singh, V. Young, Appl. Catal. B, 6, (1996), 155. G.C. Grunewald, R.S. Drago, J. Am. Chem. SOC.,113, (1991), 1636. G.J. Hutchings, C.S. Heneghan, I.D. Hudson, S.H. Taylor, ACS Symposium Series, Heterogeneous Hydrocarbon Oxidation, Eds. B.K. Warren, S.T. Oyama, ACS Books, Washington D.C., 638, (1996), 58. G.J. Hutchings,C.S. Heneghan, I.D. Hudson, S.H. Taylor, Nature, 384,(1996), 341. N. Watanabe, H. Yamashita, H. Miyadera, S. Tominaga, Appl. Catal. B, 8, (1996), 405. M. Vassileva, A. Andreev, S. Dancheva, Appl. Catal. A, 69, (1992), 221.

M. Vassileva, E. Moroz, S. Dancheva, V. Ushakov, A. Andreev, Appl. Catal. A, 112, (1994), 141.

81. 82. 83. 84. 85.

X-Z. Jiang, L-Q. Zhang, X-H. Wu, L. Zheng, Appl. Catal. B, 9, (1996), 229. M.I. Vass, V. Georgescu, Catal. Today, 29, (1996), 463. Y-M. Kang, B-Z. Wan, Appl. Catal. A, 114, (1994), 35. S.K. Agarwal, J.J. Spivey, J.B. Butt, Appl. Catal. A, 82, (1992), 259. H.A. Jones, H.S. Taylor. J. Phys. Chem., 27, (1927), 623.

2: The Destruction Of Volatile Organic Compounds

86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99.

151

H-G. Linz, K. Wittstock, Catal. Today, 27, (1996),237. D. Klissurski, E. Uzunova, K. Yankova, Appl. Catal. A,95, (1993), 103. M.L.H. Green, R.M. Lago, S.C. Tsang, J. Chem. SOC.,Chem. Commun., (1992), 365. R.M. Lago, M.L.H. Green, S.C. Tsang, M. Odlyha, Appl. Catal. B, 8, (1996), 107. S. Chatterjee, H.L. Greene, J. Catal., 130, (1991), 76. S. Chatterjee, H.L. Greene, Y.J. Park, J. Catal., 138, (1992), 179. S. Chatterjee, H.L. Greene, Appl. Catal. A, 95, (1993), 139. B. Ramachandran, H.L. Greene, S. Chatterjee, Appl. Catal. B, 8, (1996), 157. H.G. Stenger, Jr., G.E. Buzan, J.M. Berty, Appl. Catal. B, 2, (1993), 117. H.L. Greene, D.S. Prakash, K.V. Athota, Appl. Catal. B, 7, (1996),213. S.K. Agarwal, J.J. Spivey, J.B. Butt, Appl. Catal. A, 81, (1992),239. S.K. Agarwal, Appl. Catal. B, 3, (1994),N13. G.M. Bickle, T. Suzuki, Y. Mitarai, Appl. Catal. B, 4, (1994), 141. K.Y. Lee, M. Houalla, D.M. Hercules, W.K. Hall, J. Cata1.,145, (1994),23.

3 CO Oxidation Over Supported Au Catalysts BY MAYFAIR C. KUNG", COLLEEN K. COSTELLO AND HAROLD H. KUNG

1

Introduction

In the past decade the interest in supported Au catalysis has grown tremendously because of the unusual catalytic activities exhibited by these catalysts when they contain small nanoparticles of Au [13. Low temperature CO oxidation has been studied intensively over these catalysts. Despite this accelerated pace of study, important issues such as the correlation between preparation variables and catalytic activity, the nature of the active site, and the reaction mechanism are still topics of ongoing debate. Resolution of these issues requires additional research. There have been various interesting and insightful recent reviews written concerning the aforementioned subjects [1-41, In this article, we attempt not to duplicate these review articles, but to integrate the most recent developments in this area with our own results and present our perspective of CO oxidation over supported Au catalysts. The results referred to is for this reaction. Occasionally, results of selective CO oxidation (CO oxidation in the presence of a high concentration of hydrogen) are mentioned. They will be referred to as SCO specifically. The discussion will focus on (1) preparation method and the effect of support, (2) the nature of active site, and (3) the reaction mechanism and deactivation pathways.

2

Preparation of Supported Au Catalyst

A major hurdle towards understanding the unusual low temperature CO oxidation activities of supported Au catalysis is that there is a wide variation in the reported CO oxidation activities over these catalysts [S]. This arises because these catalysts are very sensitive to the preparation procedures. Most supported Au catalysts are prepared with the chloride-containing chloroauric acid precursor, although there have been recent attempts to prepare these catalysts with precursors such as dimethyl gold acetonate [6], gold-phosphine complexes [7], and gold ethylene diamine complexes [S]. This discussion will focus on the complications that arise with chloride-containing precursors.

Catalysis, Volume 17 0 The Royal Society of Chemistry, 2004 152

3: CO Oxidation Over Supported Au Catalysts

153

A major problem associated with using a chloride-containing precursor is that residual chloride has a negative impact on the CO oxidation activity [9,lO]. Chloride facilitates the agglomeration of Au particles during heat treatment [9,10], and it directly suppresses the catalytic activity by poisoning the active site [lo]. Deposition-precipitation using a solution of HAuCL is a commonly used method to prepare small particles of supported Au [ll]. In this process, the support is the nucleation initiator [121: the negatively charged Au complexes Au(OH)J&-,- are attached to the positively charged, protonated hydroxyl groups of the support. In the synthesis solution, besides the anionic Au complex, there are also negatively charged C1- anions due to hydrolysis of AuC14-,that are also adsorbed on the surface to an extent that depends on the surface charge [131. Therefore, the seriousness of chloride contamination depends on the isoelectric point (IEP) of the support [10,14]. Often, the synthesis solution is controlled at pH 7, which is slightly below the isoelectric point (IEP) of a typical A1203but above that of TiO2. At this pH, the degree of protonation of the surface hydroxyls is higher for A1203than for Ti02, which corresponds to a higher probability of chloride contamination. Thus, the impact of chloride on Ti02may be a lot less severe than on A1203. The fact that the deposition-precipitation procedure is often conducted at a pH near the IEP of alumina may account for the wide variation in activity of Au/A1203[5,15,16]. Perhaps this is also the reason why Griesel et al. reported that it was difficult to reproduce the preparation of small Au particles on A1203, but that the reproducibility can be improved by adding MgO [171. The optimal temperature and atmosphere to treat the catalysts after deposition-precipitation for maximum catalytic activity is another area where there has been no agreement between various researchers. The calcination temperature and atmosphere can affect the oxidation state of the Au catalyst. This will be discussed in detail in the section on the nature of the Au active site. Another effect of heat treatment is the removal of chloride. A reductive treatment in the presence of H2 for catalysts with a high residual chloride content is generally beneficial [9,10]. Since chloride is highly mobile in the presence of moisture, in the selective catalytic CO oxidation in hydrogen (SCO) an inductive period was observed for a Au catalyst with a high chloride content at the low temperature of 373 K [lo]. Presumably, during the induction period, the water formed in the reaction removes the chloride from the active site. There is little discussion in the literature about the effect of light during preparation. Beer et al. [18] have shown that when silver zeolites were illuminated in the presence of water, reduction of Ag occurred with concomitant evolution of molecular oxygen. It is conceivable that photo-reduction of supported Au catalysts can occur during synthesis and storage. In our laboratory, we have prepared Au/Ti02 by deposition-precipitation in the dark, and the catalyst was colorless as prepared before calcination. XANES results indicated that Au remained as Au"'. However, when exposed to light in ambient air, it changed to a purple color, indicating some reduction of the ionic gold. Likewise, if the preparation was conducted in the presence of light, the resulting sample was purple.

154

3

Catalysis

Nature of Au Active Site

Many reasons have been advanced for the unique catalytic properties of supported Au catalysts. They, as discussed presently, include Au particle size, the oxidation state of Au, and the strength of interaction between the Au particle and the support. Researchers in the community generally accept that small Au particle size is desirable for high activity, but it is not clear what the optimal dimension of the Au particle is and whether it differs for different support or pretreatment of the Au catalysts. Okumura et al. [6] used CVD method to prepare Au/Ti02, Au/Si02, and Au/A1203 catalysts. The average Au diameters of these catalysts were 3.8,6.5, and 3.5 nm, respectively, and the turnover frequencies (calculated by dividing the reaction rate by the number of surface Au atoms) were 0.02,0.02, and 0.01 s-', respectively. Considering the approximations used in estimating the turnover frequencies and that activity varies by orders of magnitude in other cases (see e.g. summary data in reference 5), the small variation suggests that similar activities were obtained over a range of Au particle sizes and different supports. Contrary to this work, Grunwaldt et al. [191 reported a strong support effect. They prepared Au/Ti02 and Au/Zr02catalysts by adsorption of Au colloid onto the support. For low Au loading samples, there were no significant differences between the two catalysts as characterized by XPS and HRTEM (average Au particle size was about 2 nm on these catalysts). However, the as prepared catalysts (dried in vacuum at 323 K and stored in air) had very different activities. For Au/Ti02 the CO conversions were 21% and 100% at 305 and 353K, respectively, whereas the Au/ZrOa catalyst was inactive at both temperatures. Even over the same support and using the same preparation method of deposition-precipitation, the results do not support an unequivocal optimal Au particle size. Bamwenda et al. [20] prepared a series of Au/Ti02 catalysts using the deposition-precipitation method. Their mean Au particle diameters and kinetic parameters of CO oxidation are shown in Table 1. Table 1

Mean particle diameters of AulTi02 and kinetic parameters for CO oxidation [taken from 201

Catalyst

Dmetal (nm)

0.5% Au-DP 0.7% Au-DP 1.8% Au-DP 2.3% Au-DP 3.1Yo Au-DP

3.5 f 1.1 3.1 f0.7 2.7 f0.6 2.5k0.6 2.9 f0.5

(K) 320 282 253 235 235

TOF at 300 K (s-')

Ea (kJlmoI)

3.7*10-2 3.4*10-2 1.2*106.8*10-2 2.6*10-'

27 19 18 20 27

'

*T,,* is temperature for 50% conversion.

A plot of TOF at 300 K versus average Au diameter appears to suggest that there is a dramatic dependence on the Au particle size, with a sharp maximum at around 2.9 nm. However, since the apparent activation energy are different for different samples, the activity of the 2.5 nm sample became similar to the 2.9 nm

3: CO Oxidation Over Supported Au Catalysts

155

sample at 235 K. Comparing other reaction data from the same set obtained by Haruta et al. [21], it appears that Au/Ti02calcined at 873K and with relatively large Au particle size of 6.7 1.4 nm had Tlj2(263 K) lower than the Au catalysts with diameters in the 3 nm range (0.5 and 0.7 Au-DP, Table 1).Thus, at this point in our knowledge of Au catalyst preparation and testing, we may conclude only that small Au particles are needed for high activity, but it is premature to decide on an optimal diameter. Small Au particle size is not the only necessary criterion for high activity. For the same Au particle size of around 6 nm, a Au/Ti02 catalyst prepared by deposition-precipitation [2 11 was four orders of magnitude more active than those prepared by photo-decomposition[20]. Haruta and coworkers [11,211 noticed that significant improvement in catalytic activities was observed when a mechanically mixed Ti02 powder and colloidal Au was heated to 873K. Concomitant with the activity change, morphological changes of the Au colloid took place, which suggested the existence of strong interaction between Au and Ti02. Therefore, they proposed that strong interaction of Au particles with the support is necessary for high activity [11,211. Grunwald et al. [191prepared catalysts via deposition of Au colloid on Ti02 and observed that the dried catalyst was already active for low temperature CO oxidation. They suggested that structural rearrangement of the colloidal Au is unlikely at such a low temperature, and no intimate structural gold/oxide interface is necessary for low temperature CO reaction. Regardless of whether there is a need for strong interaction between Au and the support, the Au-support interface is of definite importance because gold powder alone is very inactive for CO oxidation. Bollinger and Vannice [22] demonstrated this by depositing TiOz onto 10 pm Au crystallites. This combination became active for CO oxidation. These crystallites are too large to expect any quantum size electronic effect to occur, suggesting that catalytic active sites are associated with the Au-metal oxide interface. Their results imply that the role of the support goes beyond simply controlling the Au particle size. Another aspect that is intensely debated is the oxidation state of Au at the active site. Many of the results that support oxidic or metallic Au as the catalytic entity were derived from activation studies of the as-prepared catalyst in which researchers followed the evolution of catalytic activities and correlated them to the corresponding oxidation state changes at various activation temperatures and atmospheres. Oxidation states of Au were characterized with XPS [19,22,23,24], EXAFS, XANES [25,26,27], Mossbauer [28,29], and TPR [30]. One complication in the interpretation of the results from these experiments is that one must decouple the effect of oxidation state changes from other parallel changes taking place during thermal treatment of the catalyst. Generally, thermal treatment increases Au particle size and possibly promotes stronger Ausupport interaction in certain preparations [13. Catalysts prepared via the colloidal route may have carbonaceous impurities [31], and carbon removal by thermal treatment may be partly responsible for higher activities after calcination [19]. Guczi et al. [32] observed that high temperature vacuum treatment of their model Au/Fe0,/Si02/Si (111) caused reduction of the support and induced migration of partially reduced FeO, species onto the surface to partially cover

156

Catalysis

the Au particles. Another complication with some of these studies is that the catalysts were characterized after activation but not in use under reaction conditions. The oxidation state of the catalyst in situ may be different. Many researchers have proposed that the active site is metallic Au. Guczi et al. 1321studied CO oxidation over a Au/FeOX/SiO2/Si(111)model sample prepared by pulse laser deposition and concluded that metallic Au is important for catalytic activities. Grisel et al. [24] could find no evidence for Au"+ in active samples by XPS, though they allowed for the possibility that some ionic Au could be present at levels below the detection limit. Choudhary et al. [33] also were unable to detect the presence of ionic Au in active samples by XPS, though they too reported that the Au peaks obtained were broad, and the weight loading of Au may have been too low to detect the presence of small amounts of ionic gold. Others, such as Haruta [25] and Wolf et al. [34], noted that after calcination at 400°C, the samples should be all metallic gold, yet they still exhibited activity for CO oxidation. They concluded, therefore, that metallic gold is the active site. Based on spectroscopic analyses indicating the presence of ionic Au in the most active samples, some researchers proposed that ionic Au is necessary for high CO oxidation activity, though there is still no consensus whether Au"' or Au' is important. Several groups [23,28,29,35] have observed the presence of Au"' in active samples by Mossbauer, XPS, and XANES. Minico et al. [36] studied CO oxidation over Au/Fe203 using FTIR. They concluded that Au' species are more active than metallic Au. However, Au' was not stable and became irreversibly reduced during the reaction. Kobayashi et al. [37] also reported that the most active Au/Mg(OH)2 catalyst was found to contain the highest concentration of Au' as detected by 1 9 7 AMossbauer ~ spectroscopy. Park et al. [35] studied Au on different supports and concluded from XAF, XPS, and activity studies that oxidized Au is more active than metallic Au for CO oxidation. After extensively reviewing the available literature, Bond and Thompson [2] proposed that the active site consists of an ensemble of AuOH and metallic Au. Independently, our group has also arrived at a similar model based on our studies of catalyst deactivation [16,38,39,40]. Figure 1shows a schematic picture of our proposed active site. We observed that a Au/A1203 catalyst that had been calcined at 350°C could be reversibly deactivated by thermal treatment in a dry atmosphere at a temperature as low as 100°C. The activity of a catalyst deactivated in this manner could be regenerated by exposure to water at room temperature [17,38,39]. It is unlikely that such mild heating would cause severe sintering of particles. It is even more unlikely that room temperature exposure to water could effectthe re-dispersion of agglomerated Au particles. We interpreted these observations as an indication that the active site contains hydroxyl groups, and deactivation due to heating in a dry atmosphere is due to dehydroxylation. The low temperature needed to deactivate the catalyst suggests that the hydroxyl group is probably associated with Au. Consistent with the picture that AuOH is a necessary part of the active site is the fact that Au/A1203 catalyst is very sensitive to the presence of chloride. Cl- replacement of the OH group would destroy part of the ensemble necessary for activity. It was also noticed that a

3: CO Oxidation Over Supported Au Catalysts

157

Au+ support Figure 1

Model of active site for CO oxidation. From re5 16

partial reversal of C1- poisoning was possible by moisture. This is probably due to hydrolysis of Au+Cl- by H20 to regenerate the active site. A chloride content as low as Cl/Au = 0.0006 (atom ratio) was sufficient to cause detectable inhibition of catalytic activity if the chloride adsorption capacity of alumina was first reduced by adsorbed phosphate [lo]. The small Cl/Au ratio needed for the detectable poisoning effect suggests that the concentration of the Au active site is very low and that it will be a challenge to see Au-OH spectroscopically on Au/A1203. In this model of the active site, metallic Au functions to activate oxygen. In some earlier proposals, oxygen is activated on the support or at the Au-perimeter interface [24,41], the fact that Au/A1203, Au/Si02, and Au/MgO can be highly active CO oxidation catalysts suggests that activation probably occurs on Au. Our recent studies on the activation of as-prepared catalysts also indicated that a sample that is practically all ionic Au is inactive for CO oxidation. An asprepared Au/Ti02 catalyst prepared in the dark contained within error all ionic Au as characterized by XANES and TPR [26]. By comparing the XANES results with Au foil and Au'" acetate, qualitative information about the oxidation state of Au in the catalyst can be obtained. Samples containing Au"' displayed a very strong white line intensity about 4 eV above the LIIIedge that is absent in metallic Au. Also, the spectrum of metallic Au had a peak at 25eV above the edge that is absent in Au"' spectra. By comparing these peaks, it is possible to determine whether a sample is composed of mostly metallic or ionic Au. The as-prepared Au/Ti02 catalyst containing only Au(II1) was inactive for CO oxidation at -78°C. However, the catalyst was very active at 0°C and in situ XANES results showed significant reduction of Au under reaction conditions (Fig. 2). Concomitant with the appearance of metallic Au, the catalyst became active for CO oxidation at -78°C. Our investigation into the activation of Au/A1203catalyst also supported the need for the simultaneous presence of ionic and metallic Au. The as-prepared samples that were dried at room temperature contained no detectable metallic Au and were inactive for CO oxidation at room temperature. The catalyst can be made active by treating it with H2 at 100°C followed by a mixture of H2 and H 2 0 . However, HZalone at 100°C is not effective. It is possible that the presence of H 2 0 in the treatments is needed to prevent complete reduction of Au and to form the metallic and ionic Au

158

Catalysis

ensembles needed for high CO oxidation activity. H20 may also remove chloride from poisoning the active sites. Other activation studies also support this ensemble model of ionic and metallic Au. Using XANES and EXAFS, Guzman and Gates [27] found that, prior to the CO oxidation reaction, a sample dried in He at 100°Ccontained mostly AurlI, whereas one dried in He at 300°C contained mostly Auo.The sample containing Au"' had the highest initial catalytic activity, but both samples reached the same pseudo steady state. At this point, XANES showed that both samples contained Au' and Au', suggesting that an ensemble of these species is needed for CO oxidation activity. Margitfalvi et al. [30] have also proposed that an ensemble of Au,"+Aun is needed based on correlation between UV-VIS, FTIR, and TPR characterizations and activity studies. They suggested that this ensemble may exhibit behavior similar to ensembles of Snn+-Ptn.In a study of Au/Y-type zeolites with surface-modified acidity and Au/Fe/Y, all the highly active catalysts contained a mixture of Au" and Aux+[23]. Dried Au-Fe catalysts prepared by a reverse co-precipitation technique contained poorly crystallized Fe5H08 and possibly AuOOH.xH20, as determined by 1 9 7 AMossbauer ~ spectroscopy, and were found to be much more active than the calcined samples with Au metal particles on crystalline hematite platelets [28,29]. However, the characterization was not performed on the reaction-used catalyst. The existence of an induction period during the activity tests suggested that the catalyst may be changing under reaction conditions.

4

Reaction Mechanism

Haruta and coworkers [13 have suggested that different reaction mechanisms may be operative at different temperatures for CO oxidation. Extrapolating their suggestion, it is possible that different mechanisms occur on different active sites. The discussion below will focus only on the mechanism applicable to reactions near room temperature. Several mechanisms have been proposed in the literature. They can be classified into two categories: those that occur entirely on the Au particles and those that involve the support. Based on FTIR and TAP reactor studies, it is generally agreed that CO is adsorbed weakly and reversibly on Au particles [42,43,44]. However, there are many proposals for the subsequent steps of reaction. Based on their FTIR investigation of surface species generated in CO, C0-l6O2 and C0-1802 feeds, Boccuzzi et al. [44] proposed that CO oxidation over Au/Ti02 and Au/ZnO takes place by two parallel pathways. Coadsorption of CO and 0 2 produced a blue shift in the CO IR stretch from 2106 to 2116 cm-'. Furthermore, a shoulder peak at 2133 cm-' was observed and ascribed to CO on oxidized Au. In a more recent study [45], by comparing CO adsorption on Au/TiO2 catalysts with 5 nm and 10 nm Au particles, the same group concluded that the lower frequency stretch could be ascribed to CO coadsorbed with atomic oxygen, whereas the higher frequency band was CO coadsorbed with molecular oxygen on Au step sites. The proposed OC-Au-0, intermediate

3: CO Oxidation Over Supported Au Catalysts

159

1 - - .Uncalcined 1 ' -After CO Ox

1.58

I

1.38

%

1.18

[

I

I

C

.-0

CI

2 .z

a =

B-Y 4Q

E

0.98 0.78 I

0.58

I I

0

I

0.38

#

I I

0.18

11.80

11.85

11.90

11.95

12.00

12.05

12.10

12.15

Photon Energy, keV

Figure 2

XANES of uncalcined Au/Ti02 (dashed line) and the same sample after exposure to the CO oxidation feed (1% CO, 2.5% 02,in He) at room temperature (solid line)

decomposes to yield C02. Although unsupported Au is incapable of activating 0 2 , and there is still no direct experimental evidence that supported Au nanoparticles can dissociate 0 2 , results of DFT calculations [46,47] suggest that it is energetically possible. Boccuzzi et al. [44] also observed a slower pathway in which CO adsorbed at the border of Au support interface reacted with surface lattice oxygen from the support. They proposed that the former mechanism appears to be more relevant for the high activity in CO oxidation. Other groups have proposed that the support provides the activated oxygen species for CO oxidation [15,22,24,25,41,48]. Liu et al. studied [41] Au supported on TiOz and Ti(OH)4withdiffuse reflectance infra-red Fourier transform spectroscopy (DRIFTS) and observed no shift in the adsorbed CO band freTherefore, they proposed that the 0 2 adsorbed quency upon introduction of 02. on the support is the primary source of oxygen. They detected a superoxide signal with ESR spectroscopy and proposed that this may be important for CO oxidation. However, from their data, the time required for the superoxide ESR signal to disappear after introduction of CO took minutes. It would seem that the reactivity of this species is probably too low to account for the dominant low temperature pathway. The mechanisms of Bond and Thompson [2] and Kung and coworkers [5,17,38,39] are similar. They are derived from the picture of an active site consist of Au-OH and metallic Au, as shown in Figure 1. Figure 3 shows our proposed reaction mechanism [5,38]. CO is adsorbed on the Au cation and is inserted into

160

Figure 3

Catalysis

Mechanism of low temperature CO oxidation on Au catalysts. From ref 40

the hydroxyl group to form a hydroxycarbonyl. Stable hydroxycarbonly complexes of many group VIII metals have been prepared [49] and their formation from the CO and OH- ligands is enhanced by lower electron density at the metal ion [ S O ] . Oxidation of the hydroxycarbonyl will result in the formation of the bicarbonate which is decarboxylated to C 0 2 , and the active site Au-OH is regenerated. As suggested by Boccuzzi et al. [44] it is possible that multiple reaction pathways are operative during CO oxidation over supported Au catalysts and that one of them is the most relevant to the high CO oxidation activity at low temperatures. A possible explanation for the wide variation in the CO oxidation activation energy (see for e.g. Table 1) is that different reaction pathways are more predominant in different catalyst preparations. If so, it follows that different surface species could be detected and determined to be important by different research groups. The most relevant evidence for the important CO oxidation pathways would be those derived from studies of a highly reactive catalyst. However, it is not clear whether the reactivity of the catalyst remains unchanged after pelletization for FTIR study. Au/A1203 catalyst was very light in color but turned black, after subjecting it to high pressure for pelletization, indicating agglomeration of Au. On other Au catalysts such as reduced Au/Ti02, the color is already quite dark, and it is not clear whether the same phenomenon of agglomeration occurs. Therefore, the most desirable approach is to conduct spectroscopic studies in situ while simultaneously monitoring the activity of the catalyst, and confirm the relevance of observed species by following their changes to the corresponding changes in the reaction rates. Quantum chemical calculations offer insight on the reaction mechanism. Thus far, such calculations were performed on metallic Au clusters, and the possible

161

3: CO Oxidation Over Supported Au Catalysts

participation of ionic Au has not been explored. The general conclusion is that Au atoms on step edges appear to be much more reactive than atoms on terraces [46,47,51]. However, even on the edge atoms, dissociative adsorption of oxygen molecule into adsorbed atomic oxygen may not be energetically favorable. The support may play an active role. It provides a metal-support interface where adsorption of oxygen becomes favorable [Sl], and it may stabilizes a 0-0-CO intermediate [52]. The oxidation reaction occurs on Au steps in two steps: CO + 02+C02+ 0,and CO + O+CO2 [Sl].

5

Catalyst Deactivation

A unique feature of CO oxidation over supported Au catalysts is that this reaction is sensitive to moisture. The effect of moisture content in the feed on the reactivity of the catalyst depends on the nature of the support. For Au/Ti02, the effect of moisture is very pronounced. Date and Haruta [53] reported that the catalytic activity at room temperature was more than an order of magnitude lower with -0.1 ppm of H 2 0 in the feed stream compared with the optimal concentration of -200 ppm. At moisture contents higher than 6000 ppm, the reaction rate again decreased. Despite the large variations in the rates of reactions at different moisture concentrations, the activation energy remained constant suggesting that the nature of the active site was the same at low and high moisture content. Baiker’s group also observed similar effects of moisture over Au/Ti02 [19] and proposed that the inhibition effect of water at high moisture content is due to adsorption of water on the active site or due to pore filling (capillary condensation). Both of these research groups reported using a cooling trap in their line to control the moisture content, but this practice has not been commonly employed in many other studies. This may be a source of discrepancies in the CO oxidation activities between various research groups. The water effect is not due to water gas shift reaction [4]. The activity of a Au/A1203 catalyst at 300 K declined rapidly over the first ten minutes, followed by a much slower decrease until a pseudo steady state is reached [38,39]. The high initial activity is catalytic in nature. In one experiment, it was determined that the moles of COzformed (in the first five minutes of reaction) was 4.7* lo3times the total Au present. This initial loss of activity could be prevented completely by introducing 1.5% H2O in the reaction feed or by the presence of H2 [16,38,39]. The deactivated catalyst can be regenerated by exposure to a flow of 1.5% H 2 0 in He, pure H2 at room temperature, or by flowing a mixture of 1% CO, 0.5% 0 2 , and 49% H2 (SCO feed) over the catalyst [38,39]. However, other gases, such as 0 2 , CO, or He cannot restore the catalyst activity. We believe that the deactivation is not due to dehydroxylation of the active site. The amounts of H 2 0that the catalyst was exposed to during regeneration by H 2 0 or by the SCO feed were many orders of magnitude different. Instead, we proposed that it is due to the formation of inactive carbonate at the active site by deprotonation of the bicarbonate intermediate, transferring the proton to an adjacent hydroxyl group on the alumina [5,40].

-

-

162

Catalysis

+

Au-CO~H Al-OH

Au-COyAl

+-

+ H20

(1)

When an Au/A1203 catalyst was examined by DRIFTS, the intensities of the carbonate peaks were lower in a SCO feed compared with a CO O2 feed. Several other research groups had also cited carbonate poisoning as the cause for deactivation during CO oxidation. Bulushev et al. [54] proposed that carbonate, formate, or adsorbed oxygen species were the cause of deactivation during CO oxidation over Au/FeO,/C because the catalyst could be regenerated by treatment in H2 at 300°C. For similar reasons, Margitfalvi et al. [30] proposed that carbonate was the cause for deactivation during CO oxidation over Au/MgO. Lin et al. [23] proposed that Au/Fe/Y deactivated during CO oxidation by carbonate formation because they observed by XPS large amounts of carbonate on the catalyst after reaction. Thermal treatment also had significant effects on the catalytic activities. For Au/A1203,a mild heating to 100°Cin He deactivated the catalyst, but the loss of activity can be completely restored by flowing 1.5% H20 in He over the sample [38,39]. For an as-prepared Au/Ti02 that had not been heated, the sample was highly active for room temperature CO oxidation without further pretreatment. The activity was stable. However, if the sample was calcined to 350"C, it deactivated noticeably with time-on-stream, although its initial activity was still high. The deactivated sample can be regenerated by water vapor treatment, similar to Au/A1203. In light of the important roles of H2and H 2 0in the CO oxidation reaction, we examined the kinetic isotope effects of these species [5,40]. Similar to H2, D2 in the SCO reaction feed was also effective in preventing deactivation, but the rate of CO oxidation was only about 70% of that in H2. This corresponds to an apparent kinetic isotope effect for CO oxidation of 1.4 0.2. Also, the rate of D2 oxidation was much slower than the rate of H2oxidation, resulting in a selectivity toward CO oxidation of 86% versus 77% in the presence of H2. Moreover, regeneration of the catalyst with the SCO feed containing D2 restored only 65% of the activity that could be restored after treatment with a H2-containing SCO feed for the same exposure time. Similar differences were obtained when we attempted to regenerate a CO-oxidation deactivated catalyst by treatment in a flow of pure, dry D2 at room temperature. The subsequent initial CO oxidation activity was only 70% of that which could be achieved by treatment in H2. On the other hand, when H 2 0was replaced by D20 in the CO oxidation feed, the same steady activity was observed, and there was no discernable isotope effect. Similarly, regeneration of a catalyst deactivated by the CO oxidation reaction with 1.5% H20 in He gave the same subsequent initial activity as regeneration with D20 in He. No isotope effect was observed. These results suggest that H2 and H20 regenerate the catalysts by different mechanisms. Possibly, H2 regenerates the active site by hydrogenolysis of the carbonate to a formate and a hydroxyl. The formate can be further oxidized to a bicarbonate, returning the active site to the catalytic cycle. This scheme involves activation of H2 and hydrogenation of the carbonate. Both of these steps would exhibit a

+

3: CO Oxidation Over Supported Au Catalysts

163

deuterium isotope effect, as was observed experimentally. H 2 0 regenerates the active site by hydrolysis of the carbonate to a bicarbonate and hydroxyl. Only a weak deuterium isotope effect would be expected for this reaction, as observed experimentally.

6

Conclusion

In the literature, there are many reports that the activity of a Au catalyst depends on the support. However, in view of the fact that the activation and preparation procedure as well as the reaction conditions (e.g. water partial pressure) for the optimal performance may be different for different supports, more studies need to be carried out to determine if the apparent support effect can still be observed for the optimal performance of various samples. Anions such as chloride and phosphate have a negative impact on the activity, and therefore the ease of contamination of the various supports by these types of anions as a result of the synthesis procedure may give the appearance that a gold catalyst is sensitive to the nature of support. Not all studies report TOF or rate of CO oxidation. Judging from the widely different W/F used for activity tests (range 3 to 6*10-3 g.s.cc-'),Au catalysts on the same support prepared by different groups can be quite different in activity. Consequently, it is not surprising that in-situ characterization of these catalysts results in such different proposals of active sites and reaction mechanisms. It seems that, at the present stage of knowledge, the best description of the active site is one that consists of an ensemble of Au-OH and metallic gold atoms and that the fraction of ionic gold needed for activity is very low. However, direct evidence of the active site and reaction mechanism still need to be generated, and much better understanding of the energetics of the reaction pathway is needed to explain the unusually high activity of these catalysts.

Acknowledgement.- This research was supported by the EMS1 program of the NSF and Department of Energy Office of Science (CHE-9810378) at Northwestern University Institute of Environmental Catalysis. References 1. 2. 3. 4. 5. 6.

7.

8.

M. Haruta and M. Date, Appl. Catal. A: General 222 (2001)427. G. C. Bond and D. T. Thompson, Gold Bulletin 33 (2000)41. G. C. Bond and D. T. Thompson, Cat. Rev.-Sci Eng. 41 (1999) 319. M. Haruta, Cattech 6 (2001)102. H. H. Kung, M. C. Kung, and C. K. Costello, J. Catal. 216 (2003)425. M. Okumura, S. Nakamura, S. Tsubota, T. Nakamura, M. Azuma, and M. Haruta, Catal. Lett. 51 (1998)53; M. Okumura, S. Nakamura, S. Tsubota, T. Nakamura, M. Haruta, Stud. Surf. Sci. Catal. 118 (1998) 277. A. I. Kozlov, A. P. Kozlova, H. Liu, and Y. Iwasawa, Appl. Catal. A: Gen. 182 (1999) 9. D. Horvath, M. Polisset-Thfoin, J. Frissard, and L. Gucci, Solid State Ionics 141

164 9. 10. 11.

12. 13. 14. 15. 16.

17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.

30. 31.

32. 33.

34. 35. 36. 37.

38.

Catalysis

(2001) 153. S. D. Lin, M. Bollinger, and M. A. Vannice, Catal. Lett. 17 (1993) 245. H.-S. Oh, J. H. Yang, C. K. Costello, Y. Wang, S. R. Bare, H. H. Kung, and M. C. Kung, J. Catal. 210 (2002) 375. S. Tsubota, D. A. H. Cunningham, Y. Bando, and M. Haruta, in ‘Preparation of Catalysts VI,’ (G. Poncelet et al. Eds. ) p. 227, Elsevier Science, Amsterdampew York 1995. M. Che, Stud. Surf. Science and Catal. 75 (1993) 32. C. F. Bates Jr. and R. E. Mesmer, ‘The Hydrolysis of Cations’, p. 282 (1976) Wiley and Sons, N. Y. Y.-J. Chen, C.-t. Yeh, J. Catal. 200 (2001) 59. M. M. Schubert, V. Plzak, J. Garche, and R. J. Behm, Catal. Lett. 76 (2001) 143. C. K. Costello, M. C. Kung, H.-S. Oh, Y. Wang, and H. H. Kung, Appl. Catal. A, 232 (2002) 159. R. J. H. Grisel and B. E. Nieuwenhuys, J. Catal. 199 (2001)48. R. Beer, F. Binder, and G. Calzaferri, J. Photochem. Photobiol. A. Chem. 69 (1992) 67. J.-D. Grunwald, C. Keener, C. Wogerbauer, and A. Baiker, J. Catal. 181 (1999)223. G. R. Bamwenda, S. Tsubota, T. Nakamura, and M. Haruta, Catal. Lett. 44 (1997) 83. S. Tsubota, T. Nakamura, K. Tanaka, and M. Haruta, Catal. Lett. 56 (1998) 131. M. A. Bollinger and M. A. Vannice, Appl. Catal. B: Environ. 8 (1996) 417. J.-N. Lin, J.-H. Chen, C.-Y. Hsiao, Y.-M. Kang, and B.-Z. Wan, Appl. Catal. B: Environ. 36 (2002)19. R. J. H. Grisel, C. J. Westrate, A. Goossens, M. W. J. Craje. A. M. van der Kraan, B. E. Nieuwenhuys, Catal. Today 72 (2002) 123. M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M. J. Genet, and B. Delmon, J. Catal. 144 (1993) 175. Kungs’ group, unpublished results. J. Guzman and B. C. Gates, J. Phys. Chem. B 106 (2002)7659. R. M. Finch, N. A. Hodge, G. J. Hutchings, A. Meagher, Q. A. Pankurst, M. R. H. Siddiqui, F. E. Wagner, and R. Whyman, PCCP 1 (1999)485. N. A. Hodge, C. J. Kiely, R. Whyman, M. R. H. Siddiqui, G. J. Hutchings, Q. A. Pankhurst, F. E. Wagner, R. R. Rajaram, S. E. Golunski, Catal. Today 72 (2002) 133. J. L. Margitfalvi, A. Fasi, M. Hegedus, F. Lonyi, S . Gobolos, and N. Bogdanchikova, Catal. Today 72 (2002) 157. M. Maciejewski, P. Fabrizioli, J.-D. Grunwald, 0.S. Becker, and A. Baiker, PCCP 3 (2001) 3846. L. Guczi, D. Horvath, Z. Pastel, L. Toth, Z. E. Horvath, A. Karacs, and G. Peto, J. Phys. Chem. B 104 (2000) 3183. T. V. Choudhary, C. Sivadinarayana, C. C. Chusuei, A. K. Datye, J. P. Fackler, Jr., and D. W. Goodman, J. Catal. 207 (2002)247. A. Wolf and F. Schuth, Appl. Catal. A: Gen 226 (2002) 1. E. D. Park and J. S. Lee, J. Catal., 186 (1999) 1. S. Minico, S. Crisafulli, A. M. Visco, and S. Galvagno, Catal. Lett. 47 (1999) 273. Y. Kobayashi, S. Nasu, S. Tsubota, and M. Haruta, Hyperfine Interactions, 127 (2000)95. H.-S. Oh, C. K. Costello, C. Cheung, H. H. Kung, and M. C. Kung, Stud. Surf. Sci and Catal. 139 (2001) 375.

3: CO Oxidation Over Supported Au Catalysts

39.

40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54.

165

H.-S. Oh, C. K. Costello, C. Cheung, H. H. Kung and M. C. Kung, in Proceedings of the papers presented at the Pacific Chem. Conference, December 2000 and at the North American Catalysis Society Meeting, June 2001. C. K. Costello, J. H. Yang, H. Y. Law, Y. Wang, J.-N. Lin, L. D. Marks, M. C. Kung, and H. H. Kung, Applied Catal. A: Gen. accepted. H. Liu, A. I. Kozlov, A. P. Kozolova, T. Shido, K. Asakura, and Y. Iwasawa, J. Catal. 185 (1999)252. M. Olea, M. Kunitake, T. Shido, and Y. Iwasawa, PCCP 3 (2001)627. A. K. Tripathi, V. S. Kamble, and N. M.Gupta, J. Catal. 187 (1999)332. F. Boccuzzi, A. Chiorinio, A. Tsubota, and H. Haruta, J. Phys. Chem. 100 (1996) 3625. F. Boccuzzi and A. Chiorino, Stud. Surf. Sci. Catal. 140 (2001) 77. M. Mavrikakis, P. Stoltze, and J. N-rskov,Catal. Lett. 64 (2000) 101. N. Lopez and J. K. N P I ~ S ~J.OAmer. V , Chem. SOC.124 (2002)11262. L. Guczi, D. Horvath, Z. Paszti, and G. Pet6, Catal. Today 72 (2002)101. P. C. Ford and A. Rokicki, Adv. Organomet. Chem. 28 (1988) 139. D. F. Gross and P. C. Ford, J. Amer. Chem. SOC.,107 (1985)585. Z.-P. Liu, P. Hu, and A. Alavi, J. Amer. Chem . SOC.124 (2002) 14770. L. M. Molina and B. Hammer, Phys. Rev. Lett. 90 (2003)206102. M. Date and M. Haruta, J. Catal., 201 (2001) 221. D. A. Bulushev, L. Kiwi-Minsker, I. Yuranov, E. I. Suvorova, P. A. Buffat, and A. Renken, J. Catal. 210 (2002)149.

4 Coke characterization BY C.A. QUERlNl

1

Introduction

The formation of coke deposits leading to catalyst deactivation has been a challenge for catalytic technology in many hydrocarbon processes. The development of the fluid catalytic cracking (FCC) process and the continuous catalytic reforming (CCR) are examples of such developments. In these cases, the catalyst, the operating conditions, and the continuous coke removal are integrated and work in a delicate quasi-steady state. In any case, the effective management of catalyst deactivation and catalyst regeneration is the key in many heterogeneously catalysed processes. The optimization of such complex processes requires the characterization of the coke deposits in order to understand the effect of the operational variables on these deposits and, therefore, minimize its formation and develop effective regeneration strategies. In the last decade, there has been an increase in the research effort in this field. Coke characterization has been included in many papers where deactivation is a major issue. Several characterization techniques have been used to study coke deposits and to obtain information regarding reaction mechanism, deactivation mechanism, and regeneration conditions. One of the most widely used techniques is temperature-programmed oxidation. Because of its simplicity and utility, this technique has been widely accepted and used in the characterization of coke in a large variety of catalytic systems. In this work, the different techniques that have been used for coke characterization on heterogeneus catalysis are presented, using examples of applications reported in the literature. The objective is to present the type of information that each technique provides regarding coke composition, localization, morphology and kinetics of coke gasification, with related references to guide the researcher to previous studies in the field of coke characterization, and more specifically, coke deposits formed on catalytic surfaces. Some of the techniques limited to single crystals or polycrystalline foils, such as low energy electron diffraction (LEED), He scattering, core electron energy loss spectroscopy (CEELS), and metastable deactivation spectroscopy (MDS), are not presented in this work. A. Bell' reported some examples of application of these techniques. It is not the intention to fully review all the literature regarding coke characterization, but to present selected examples of application of different techniques in Catalysis, Volume 17 0 The Royal Society of Chemistry, 2004

166

4: Coke characterization

167

some of the most important catalytic systems. The examples for each technique are presented either for the most important reactions on which the technique was applied (e.g. naphtha reforming, dehydrogenation, cracking, etc), or for a given type of catalyst if several studies involving different reactions are discussed (e.g., zeolites, nickel catalysts, etc). Two Tables have been included at the end of the chapter, after the presentation of all the techniques. The first one reports references for several processes where coke characterization is addressed, including the techniques used in each study (Table 5). The second one gives a summary of the information that can be obtained with each technique (Table 6). 2

Temperature-Programmed Techniques

There are several temperature-programmed techniques that have been used to characterize coke. The difference among them is the composition of the carrier gas. When oxygen is present in the carrier, the technique is called temperatureprogrammed oxidation (TPO); if the carrier contains hydrogen, it is called temperature-programmed hydrogenation (TPHy); if only an inert gas (helium or nitrogen) or C 0 2is the carrier gas, it is called temperature-programmed gasification (TPHe, TPC02). 2.1 Temperature-Programmed Oxidation (TPO). - This technique has been extensively used to study coke in many different catalytic systems. It provides direct information regarding coke oxidation rate. In addition, it is possible to obtain useful information such as: location, composition (hydrogen/carbon ratio) and morphology (highly dispersed or multidimensional particles). The information regarding coke oxidation kinetics is relevant to properly design regeneration processes. The TPO analysis can be performed in different ways, using different detection methods:

detection of C 0 2by a thermal conductivity detector (TCD) after it is separated from oxygen and water in a GC column2; - detection of CO, C02 and hydrocarbonaceous compounds after methanatied; - quantification of C 0 2with a mass spectrometep; - monitoring temperature increment above a reference sample, in differential thermal analysis (DTA) equipment5; - measurement of weight loss in thermal gravimetric analysis (TGA) equipment6. - detection of C 0 2by FTIR'22,'23. -

The detection of C 0 2with a TCD requires that the gases coming out of the cell be fully oxidized, in order to transform CO and other hydrocarbons released from the catalysts into C02.Since the technique involves the separation of gases in a GC column, its disadvantages are that it is a noncontinous analysis and that it misses the fine structure, the peak height and its temperature. Besides, the

168

Catalysis

sensitivity of a TCD is not very high. The use of a mass spectrometer makes it possible to sample at shorter intervals, but it is not as simple as TCD. Additionally, it is difficult to determine if other fragments of hydrocarbons are released from the catalyst during the analysis. The differential thermal analysis is complicated by signals due to the oxidation of catalyst components (e.g. metals), sulfur or contaminants. In the case of conventional TGA analysis, plug flow through the sample is often not possible and therefore it is very difficult to study the coke oxidation kinetics. Besides, signals due to change in weight because of other reasons such as oxidation of catalytic components and dehydroxylation, complicate the analysis. The use of a methanator after the cell improves the sensitivity and leads to a continous analysis. It also allows to determine the amount of compounds released from the catalysts during the thermal treatment, in different carrier gases either under the form of C O + C02 or as hydrocarbons. 2.1.1 Determination of H/C Ratio. To determine the H/C ratio of a coke deposit, COzproduction and O2consumption are analysed as the sample temperature is raised. The difference between the C02 produced and the 0 2 consumed, is assigned to water production. With this information, the general formula CH, of coke is obtained7. This method has been applied to many different catalytic systems, such as 1-butene skeletal isomerization', naphtha reforming7,n-hexane isomeri~ation~ and isobutane alkylation". In some cases, water and C 0 2 have been collected during the T P O in cold traps, and then, upon heating detected chromatografically with a TCD'. In this case, special care must be taken when working with zeolites, since the water released from the catalyst itself could make the results confusing. One of the ways to estimate this, is to run a blank with the same sample, but using helium as carrier gas instead of oxygen*. 2.1.2 Determination of Coke Locution. The T P O technique allows the determination of the coke location on supported metal catalysts, such as naphtha reforming. Since the metal, typically platinum promoted with rhenium, iridium, tin, or germanium, has a catalytic effect for coke burning, the T P O profile displays two main peaks. The low temperature peak is due to the oxidation of the coke directly deposited on the metal particle, or in its ~ i c i n i t y ~In. ~this . way, it is possible to study the effect of catalyst formulation and operational conditions on the formation of coke on the metal and on the support. Platinum and palladium were used to facilitate the regeneration of zeolite catalysts used in isobutane alkylation. The TPO profiles in these cases display a characteristic peak, very sharp, that is not usually found in any other system, due to the presence of a large amount of hydrocarbonaceous deposits located very close to the metallic 2.1.3 Determination of Oxidation Kinetics. The reaction rate between oxygen and coke depends upon the oxygen partial pressure, the composition, morphology and location of coke, and the catalytic effect of catalyst components and impurities. Therefore, the study of the kinetics of coke oxidation provides information regarding coke characteristics. Additionally, it is necessary to

169

4: Coke characterization

properly design the regeneration process by coke burning. A linear combination of power-law kinetic expressions has been applied to analyse multipeak TPO profiles on supported metal catalyst^'^?'^, using fixed coke reaction orders, and considering that the reaction rate is proportional to coke concentration with a given reaction ~ r d e r ' However, ~. the reaction rate is proportional to the exposed carbon surface area, according to the following expre~sion'~, which is valid for each type of coke present on a given catalyst: ri = Aj . exp(-Ei / R T ) . Sj . P02m

(1)

where rj is the production rate of C02, from the ith coke type, Si is the exposed surface area of a particular type of carbon, P02 is the partial pressure of oxygen, rn is the oxygen reaction order, and Ei and Ai are the activation energy and preexponential factor respectively, for the ith type of coke. Taking into account that Si is related to the initial exposed surface area of coke (Sio) at a given coke conversion (Xi)by:

Si = Sjo (1 - X j y

(2)

where n is coke reaction order, and relating Siowith the initial coke concentration CiO results in: Yj

=

Ai, . exp(-Ej / R T ) . Cjo . (1 - Xi). . P0zm

(3)

The total oxidation rate is the sum of the contributions of expressions like (3) for each type of coke. There are four parameters associated with each type of coke, and therefore with each peak in the T P O profile: Ei, A,,, and Cio, plus the carbon reaction order. The latter is not neccesarily the same for all the types of coke, and is associated with the size and the shape of the coke particles. As coke is burnt off, the size and the shape change and therefore, the coke reaction order will change during regeneration. This has been used to obtain information regarding coke morphology, as discussed below. With this model, the influence of heating rate, oxygen concentration, coke particle size, coke concentration, etc, have been simulated and compared with experimental results. To obtain the parameters in equation (3), from experimental data, it has been suggested that more than one T P O profile must be used in the deconvolution in order to obtain reliable parameters. This model has been used to obtain kinetic parameters for coke combustion in an acetylene hydrogenation ~atalyst'~. 2.2.4 Determination of Morphology. Coke can be deposited on a catalyst with a large variety of morphologies. The size and the shape of the coke particle affect the oxidation kinetics. When coke has a three-dimensional structure, and since the oxidation occurs on the coke particle surface, the order with respect to coke results in less than first order16. In a limiting case, where the coke is deposited covering the support in a multilayer arrangement, the coke reaction order would be zero. This is because as coke is burnt-off, new carbon atoms are exposed and therefore the reaction rate does not change as coke content decreases. On the other limiting case, when all the carbon atoms are exposed, e.g. when it is forming

Catalysis

170

a monolayer, the reaction order would be one. In any other intermediate situation, the reaction order would be between 0 and 1.This allows one to obtain a first approximation to coke rn~rphology'~>'~. In order to distinguish this change in morphology more easily, and to gather information regarding mono or multilayer configuration of coke deposits, partial burn experiments were designed17.The TPO is carried out up to a given intermediate temperature, and then it is held constant. If coke reaction order is close to zero, at constant temperature the reaction rate would be constant. Figure 1 A shows TPO profiles of three different catalysts, with different coke content. Figure 1 B shows the profiles obtained in TPO experiments heating up to temperatures between 450 and 500°C. The higher the coke content, the more stable is the signal when the constant temperature is reached, in spite of using higher temperature for the heavily coked catalysts. Samples taken from a commercial naphtha reforming reactor between the 4th day and the last day of the cycle (day 208) were analysed with this procedure'*. The catalyst was Pt-Re-S/A1203.Figure 2 shows some of these results. It can be clearly seen that the catalyst with high coke content perfectly displays a constant oxidation rate at constant temperature, which indicates that there is an important fraction of the carbonaceous deposits that is forming a tridimensional structure. On the other hand, the sample taken from the reactor the 4th day of operation, when reaching the same temperature does not display a constant reaction rate. It rather shows an exponential decay, which is the typical behavior of a first order reaction. 2.1.5 Determination of Coke Amount. The amount of coke can be obtained from

TPO analysis. TGA provides a straightforward information regarding the weight of material that is being gasified from the catalyst, although the result is A

B

Pt

0 Figure 1

200 400 600 0 Temperature, "C

2000 4000 Time, sec

( A ) TPO for Pt-Re/Al,O, (4.11 wt% C ) , Pt/A1203(1.24 wt% C ) , and Pt-ReS/A1203(13.0 wt% C ) catalysts, ( B ) partial burn experiments: heating up to 450°C for Pt, 490°Cfor Pt-Re and 500°Cfor Pt-Re-S. t, indicates time at which an acceleration of the rate of decay of COz production occurs. From Ref 17

4: Coke characterization

171

i Tr450OC I

i

OO

4000

8000

12000

Time, sec TPO profiles for Pt-Re-S/Al2OI1,coked in a commercial reformer. Temperature was increased up to 450 "C,and then held constant

somewhat complicated due to weight changes caused by oxidation of catalytic components, dehydration and dehydroxylation. The TPO technique carried out using the methanator and FID provides a very simple and precise method to quantify the amount of carbon on the catalytic surface. Since all the compounds gasified during the analysis are converted in CH4 and detected with the FID, a simple calibration allows the determination of the coke percentage on the catalysts. This technique has been used to detect coke content well below 0.1 wt% with high precision. The use of a TGA/FTIR technique is also a sensitive technique to determine small amounts of coke122.For coke quantification, care must be taken in order to fully transform the compounds released from the catalyst (CO, hydrocarbons) to C02, which is the compound being monitored by the FTIR detector.

2.2 TPO Studies of Different Catalytic Systems. - Figure 3 and 4 show TPO profiles of several coked catalysts used in various hydrocarbon processes: hydrogenation, reforming, fluid catalytic cracking, and methanol to olefins (Fig. 3), dehydrogenation, alkylation and 1-butene isomerization (Fig. 4). Profiles for graphite and amorphous carbon are included as reference. It is interesting to note that even in the case of the pure amorphous carbon, the TPO peak is non-symmetric. This is an indication that the carbon reaction order is less than 0neI4. The peak is more symmetric when the carbon reaction order is one. Another interesting feature can be observed on the profile of the coked Pt/Si02 dehydrogenation catalyst. In this case, the very sharp peak that appears at 200 "Capproximately, corresponds to the coke that is related to the metal. This peak

172

Catalysis

Reforming

I

I

200 Figure 3

I

I

1

I

400 600 Temperature, "C

I

I

800

I

1000

TPO projles of several coked catalysts, using methanatorlFID conjguration

200 400 600 Temperature, "C Figure 4

TPO projles of several coked catalysts, using rnethanatorlFID conjiguration. (a) Pt/Si,O, 0.8 %C, isobutane dehydrogenation, (b) ferrierite, 9.3 %C, 1-

butene isomerization (c) PtlLaH Y, 12 %C, isobutane alkylation (d) LaH Y, 12%C, isobutane alkylation (From Ref. 3)

4: Coke characterization

173

can also be clearly distinguished in the case of the alkylation catalyst that contains Pt. 2.2.1 Naphtha Reforming. Coke deposited on naphtha reforming catalysts has been extensively analysed using TPO The main information obtained from these studies was the location of coke, either on the metal or on the support. The distribution of coke on a commercially coked Pt-Re-S/Al2O3-C1 catalyst as a function of time-on-stream was followed in this way2'. It is understood that the coke assigned to the metal particles, also corresponds to coke that is in the vicinity of such particles. In many cases however, it is not easy to distinguish the two burning regions, and therefore the assignment of coke to metal or to support is not easy. This is the case of catalysts with high coke content22. 2.2.2 Parafins Dehydrogenation. Coke formed on silica supported platinum and platinum-tin catalysts, during the dehydrogenation of isobutane, was studied by TP026,using a mass spectrometer for the detection of C02. The addition of tin to platinum decreases the amount of coke formed on the promoted catalyst as compared to the non promoted catalyst. The effect of the preparation method, hydrogen/isobutane ratio and regeneration treatments on the amount of coke was also addressed. A better characterization of the hydrocarbonaceous deposits was obtained using the methanator-FID configuration, since in this case, all the carbon-containing compounds are detected. With this technique, it was possible to clearly distinguish the location of the hydrocarbonaceous deposits left on the catalyst, and specifically how the tin suppressed the formation of coke on the metal function. These deposits are released at low temperature either as hydrocarbons or partially oxidized hydrocarbons, and therefore they are not detected as C 0 2in the mass ~pectrometel-2~. The profile for Pt/Si02is shown in Figure 4. The sharp peak that corresponds to coke deposits on the metal function was not observed with the mass spectrometer by following the 44 mass. Resasco et a12*have studied the dehydrogenation of isobutane over heavily sulfided Ni catalysts at 600°C. The coke was analysed by DSC, TPO and TPHy. The thermograms obtained by DSC display two exotherms at about 457°C and 587°C. Based on these results, they estimated the combustion heat for catalysts containing different amounts of coke. A clear trend was found. Samples containing carbon deposits below 3-4 wt% had a heat of combustion higher than 110 kcal/mol of C , while those containing relatively large amounts of carbon had a heat of combustion of about 90 kcal/mol of C, which is equivalent to the heat of combustion of graphite. The former value indicates that the coke contains some H associated with the carbon. The TPO profiles also displayed two peaks, that were designated coke I (maximum at 457-477 "C) and coke I1 (maximum at 587°C). The amount of coke I reached a constant value when the coke content was 2-3 wt%. This saturation value of carbon I was about 1wt%, and corresponded to a C/Ni molar ratio near unity. Carbon I1 increased steadily as a function of total coke content.

174

Catalysis

2.2.3 Cracking. The combustion mechanism of coke deposited on cracking catalysts was studied using TPO, after methanation of CO and C02. In this experimental setup, two flame ionization detectors were used: one detected CO plus C02, and the other detected only CO after C 0 2is absorbed in a ~ c a r i t eIt~ ~ ~ ~ ~ . was found that there were up to three peaks in the rate of C02 evolution and one CO peak. The lowest temperature C 0 2 peak, which was very small, was attributed to highly reactive coke or coke in the vicinity of trace metals. The other two overlapping C 0 2 peaks occurred at high temperatures. The first of them, coincided with the CO evolution peak. It has beens suggested that these observations are due to changes in the combustion mechanism, caused by a change in the rate limiting step as temperature is increased. Using the same experimental setup, J.M. Kanervo et a131 obtained kinetic parameters for different models. No significant difference among the models could be found, and therefore no conclusions about the reaction mechanism could be drawn. In another the TPO were carried out by measuring the oxygen consumption after trapping the water and the C02 in cold traps. In this study, the effect of contaminants metals (V, Ni) on coke oxidation rate was studied. The TPO profiles were deconvoluted empirically using four gaussian curves. 2.2.4 1-Butene Isomerization. The skeletal isomerization of 1-butene on ferrierite was studied with TP033-35. Reaction temperature, 1-butene partial pressure, and space velocity affected the amount of coke deposited on the catalyst, which was between 5 and 9 wt% after 3 hours. The TPO profiles displayed two main peaks, one between 250 and 450°C, and the other at high temperatures. Figure 4 shows a TPO of a ferrierite coked at 380°C. As the reaction temperature was increased, the size of the second peak became bigger than the first one. This second peak is mainly aromatic coke, formed both during the TPO analysis when heating above the reaction temperature, and during the reaction, specially at high temperatures, e.g. 400°C. It was found that to fully remove the coke, temperatures higher than 600°C were necessary. It took 15 h at 660°C to decrease the amount of coke below 0.1%. The H/C ratio was determined by measuring the total amount of water and the C02 released during TP08. It was found that this ratio decreases as a function of TOS, and that during the TPO analysis, the H/C ratio varied from as high as 4 at 300°C to as low as 0 at 600°C. This is in agreement with the previous finding, that the coke reorganizes during the TPO, changing from the initial composition rich in aliphatic compounds, to an essentially heavy aromatic coke at higher temperatures. 2.2.5 Isobutane Alkylation. The deactivation of solid acid catalysts due to coke deposition is the cause of not having as yet, a commercially available process for isobutane alkylation with C4 olefins, using solid acid catalysts. The coke on these catalysts have been characterized with TPO The TPO profiles on zeolites used in this reaction, displayed two well defined burning zones. One peak below 300"C,and the other at high temperatures. The relative size of these peaks depends on the zeolite and the reaction temperature. In the case of the mordenite, the first peak was the most important, and in the case of the Y-zeolite, at 50°C or

4: Coke characterization

175

higher reaction temperature, the second peak was more important. Figure 4 shows a TPO that corresponds to a LaHY zeolite, coked at 80°C. It has been demonstrated that the second peak is generated during the TPO analyses. Using other characterization techniques, such as FTIR and 13CNMR (see below) it was demonstrated that the coke is originally aliphatic. Therefore, as temperature is raised during the TPO analysis, the coke changes its structure from aliphatic to aromatic, this process being catalysed by the acid sites. This change makes it very difficult to regenerate the catalyst by burning the coke off. High temperatures, above 500"C, are needed to fully remove the coke. The first peak of the TPO can be eliminated by heating in an inert gas or just by applying vacuum. The H/C ratio has been determined for a LaHY zeolite". Below 550"C, the oxygen consumption is higher than the C 0 2 production. Above this temperature, it is essentially the same, which indicates that the coke remaining on the catalyst after a thermal treatment at this temperature left a highly dehydrogenated residue. 2.2.6 Methane Dehydro-aromatization. Methane dehydroaromatization on Mo/HZSM-5 catalyst has been investigated using a combination of temperature-programmed oxidation, hydrogenation, gasification with C02, and thermal gravimetric analysis37.Experiments with C 0 2 were carried out, because it was found that it reacts with the surface inert carbon species to regenerate CO, and therefore the stability is improved. The TPO profiles displayed two maxima at 507°C and 587 "C. The TPHy experiment indicated that only the high temperature coke is partially eliminated by hydrogen, while the C 0 2 decreases the amount of both types of coke. A treatment with C 0 2 followed by H2 leads to a partial regeneration of the catalyst. 2.2.7 Residue Hydrotreating. - A coked Ni-Mo/y-A1203commercial catalyst used in the hydrotreating of a heavy petroleum fraction was studied using TPO, following weight changes in a TGA apparatus3! The catalyst coked for 100 h showed only a broad peak at 460°C, which was assigned mainly to C 0 2and to a lesser extent, to SO2.It has to be pointed out that the thermogravimetric analysis is complicated with this type of catalyst since the combustion products consist of CO, C02,SO, SO2,NO, NO2,and H20. TPO of catalysts coked for 1100,2100 and 3100 h displayed several peaks. One at 25O-27O0C,that could correspond to SO2, another at 365430°C corresponding to COZ, CO, NO, and SO2, and the third one at 610-730°C due to a non-reactive coke and the SO2 formed from metallic sulfides deposited on the catalyst.

2.3 Temperature-ProgrammedHydrogenation (TPHy). - The hydrogenation of coke during a temperature-programmed analysis has been used to characterize the coke and to study catalyst regeneration. Characterization of coke by TPHy is of special interest when hydrogen is one of the reactants, e.g. reforming, dehydrogenation, methanol synthesis, and Fischer-Tropsch synthesis. The overall deactivation rate is the difference between the coke formation rate, and the coke gasification rate. If the former is greater than the latter, coke accumulates on the catalyst. If the gasification rate is greater than the formation rate, no coke

176

Catalysis

is formed. Therefore, in the above mentioned systems, hydrogen plays a key role in the control of the deactivation rate. 2.3.1 Coke on Nickel Catalysts. The different types of carbon formed on nickel catalysts have been characterized according to its reactivity with hydrogen. In this case, many different types of coke were identified, as reviewed by C.H. bar tho lo me^^^^‘", each of them having different TPHy profiles. The adsorbed atomic carbon (CJ, which is the most reactive, displays a maximum in the TPHy profile around 200°C. On the other hand, the graphitic carbon, the least reactive, displays the maximum between 550°C and 850°C. Since the coke forming and gasification reactions have different activation energies39p0, there is a safe region of temperature where it is possible to operate the reaction in order to have a gasification rate higher than the formation rate. The coke formed on a Ni/A1203 catalyst during n-hexane cracking, was characterized by TPHy41. The effect of alkali and alkaline-earth promoters was studied. No significant differences were observed among the profiles in this case. The coke formed on sulfided nickel catalysts was studied28.The TPHy profiles indicated that only a small fraction of the coke can be removed with H2, approximately 10%. The profile displays a small peak at 267"C, and a larger one at about 817°C. These results demonstrate the refractory nature of the carbon deposits toward hydrogen, and therefore, that there is little benefit in life characteristics to be gained by increasing the partial pressure of feed hydrogen. 2.3.2 Regeneration Studies with TPHy. During the alkylation of isobutane with C4 olefins using solid acid catalysts, coke formation deactivates the catalysts. This coke requires high temperatures to be eliminated with oxygen. Because of this, a metal was incorporated to the solid acid catalysts, and hydrogen was used during the regeneration at mild condition^'^^' 1!42. Platinum and palladium were used to analyse the hydrogenation of coke deposits on LaHY zeolite4*.The TPHy studies indicated that the coke could not be completely removed from the deactivated catalyst by using hydrogen, even if temperatures as high as 620°C were used. 2.4 Temperature-Programmed Gasification. - The reactivity of the coke has been studied using temperature-programmed analysis, with carrier gases different from oxygen. The presence of hydrocarbons that desorb upon heating from alkylation catalysts has been studied using helium36. It was found that an important fraction of the coke is eliminated in this way. Also, helium was used as carrier gas to study the oxidation of coke by surface hydroxyl groups on alumina supported ~atalysts~t'~. This reaction occurs at high temperature, above 600°C. Larsson et all5studied coked Pd/a-A1203used in acetylene hydrogenation. They used argon as carrier gas to distinguish if one of the peaks in the TPO profile was due to hydrocarbon desorption or to coke burning. Coke deposited on sulfated zirconia catalyst during n-butane isomerization was studied using TPHe, using the methanator/FID c~nfiguration~~. The treatment in helium up to 650°C removed all the coke from the catalyst. This

4: Coke characterization

177

experiment was carried out both going through the methanator and by-passing the methanator. In the latter case, very small amount of products were detected, which means that the compounds released from the coked catalyst during the heating in helium are C O and/or C 0 2 .This was explained considering that the coke was oxidized at some redox sites on the catalyst surface, these groups being probably sulfate and lattice oxygen ions. The use of COz in the carrier gas has been used in some case^^^'^'. 3

Electron Microscopy

The localization, nature and structure of coke deposits have been examined with electron microscopy. In some cases, such as reforming catalysts, there are few papers where the use of electron microscopy has been reported 44-46. In other processes, such as in those cases where carbon whistles are formed, there is a large number of studies using this technique. In this case, the morphology of the carbon deposits are easily distinguished from the catalyst, and the interpretation of the analysis is easier than in other cases where coke is distributed on the surface of the catalyst (e.g. naphtha reforming catalysts). Typically, the electron microscopy alone does not provide much information, and is generally used in combination with related spectroscopies. 3.1 Naphtha Reforming. - Espinat et a145observed a heavily coked reforming catalyst (27% coke) and concluded that there are three types of zones on the catalyst surface: a coke-free zone; a slightly-coked zone with a layer of pregraphitic carbon, having a thickness of few atomic layers and different degrees of organization; and highly coked zones, in which the support is buried in the coke which is up to several hundred A thick. In this zone, the coke is poorly organized. Gallezot et a146using conventional TEM, concluded that the coke deposited on the catalyst (Pt/A1203)probably has a disordered structure. Cabrol and Oberlin4 used a previously developed technique to identify planar aromatic ring structures. They combined the analysis of the scattered beams with lattice fringe techniques, selected area electron diffraction, and bright and dark field analysis. They suggested that in their Pt/A1203coked catalyst, the coke is deposited as a stack of 2 or 3 aromatic planar ring structures less than 10 A in size lying approximately flat on the alumina crystal face. Some of these carbon units gather also into porous aggregates a few thousand Angstroms in size on the catalyst. One of the limitations of these technique is that the TEM cannot see nonaromatic functional groups, and therefore each stack can be surrounded by an unknown amount of carbon and considerably enlarge the size of the real carbonaceous individual units. T.S. Chang et a14’ carried out in situ electron microscopy studies on model supported platinum catalytic systems. They showed that upon reaction in a hydrocarbon environment, the metal particles were capable of hydrogasifying carbon in their inmediate vicinity.

3.2 Coke on Nickel Catalysts. - Nickel catalysts are well known for their ability to form whisker (or filamentous) carbon. The mechanism of formation of

178

Catalysis

such whisker carbon has been studied over the last 30 years, and is the main route for carbon formation in steam reforming. The understanding of the mechanism of whisker carbon formation led to the development of improved catalysts to control the carbon formation. Additionally, the special properties of certain type of these carbon fibers, also known as ‘nanotubes’, have been extensively characterized. The nanotubes have the graphite planes parallel to the access of the fiber, whereas the whisker carbon formed during steam reforming have the graphite planes paralell to the axis. The thermodynamics of the system formed by Ccsl,CO, H2, CH4, COZ, and H20 is fully described by C.H. B a r t h ~ l o m e w ~ ~ aby n dJ.R. Rostrup-Nielsen and J. Sehested48. The carbon deposits obtained on Ni/A1203 and Ni/MgO catalysts during steam reforming of n-butane were studied with high resolution electron microscopy (HREM)49.In some cases optical diffraction was carried out to check the degree of graphitization of the carbon structures. Three kinds of carbon deposits were found: true filaments, tubes, and shells. Temperature influences the form in which the carbon is formed, and their degree of graphitization. The deposits produced at low temperatures (below 500 “C for Ni/MgO and below 600°C Ni/A1203)consist of true filaments. At higher temperatures (600°C and 680”C, respectively) tubes with carbon layers nearly parallel to the tube axis and close shells are created. According to this study, the carbon shells occurred only on large metal particles of 100 nm approximately. 3.3 1-Butene Isomerization. - The coke formed on ferrierite during 1-butene skeletal isomerization was observed with TEM”. The images show that for a ferrierite containing 9.1 wt% coke, the platelets are bordered by a layer of amorphous material which is probably coke. 4

Electron Energy Loss Spectrocopy (EELS)

Electron energy loss spectroscopy involves the excitation of core electrons by primary electrons. This technique provides analytical and structural information, similar to that given by X-ray absorption spectro~copy~~. Recording the EELS spectra with an electron spectrometer attached to a transmission electron microscope offers the advantage of having a high-resolution image of the specific zone analysed. The spatial resolution of the analysis can be as small as 1 nm2, therefore, it is possible to detect the location of the coke. Usually, the spectra of the coked catalyst is compared with that of reference compounds in order to obtain information about the coke s t r ~ c t u r e ~For ~ *this ~ ~ .reason, the technique provides qualitative information regarding the type of coke present on the catalyst. This technique has been used only in few cases to characterize coke deposits on heterogeneus catalysts.

4.1 Naphtha Reforming. - Gallezot et a146studied the location and structure of coke with this technique associated with electron microscopy, as above cited, in a

4: Coke characterization

179

study of naphtha reforming catalyst. They found that coke coverage around a metal particle extended as far as 20 nm, the support being covered with patches of amorphous carbon. The disadvantage of this technique, is that in a real (commercial)reforming catalyst, the metal particles were very small and distributed all over the support, as acknowledged by the authors. Since they were so close to each dther it was impossible to check how far from a given particle the coke coverage was extending, and the technique is not sensitive enough to check if they were covered by carbon. Therefore, a model catalyst, with metal particles in the range of 3-6 nm was prepared, and it was possible to scan individually by the electron probe in the STEM.

4.2 1-Butene Isomerization. - The STEM-EELS technique was used to check coke location on spent ferrierite crystal, after skeletal 1-butene isomeri~ation~~. Ferrierite crystals are very appropiate for STEM-EELS line scans, as they are plate-like. The measurements were performed in perpendicular directions over the ferrierite crystal, following lines parallel to the 8 MR channels and to the 10 MR channels. The carbon to zeolitic oxygen ratio was measured as a function of the distance to the edge of the crystal. A large increase in this ratio was found in the line parallel to the 8 MR towards the edge of the crystal for a catalyst containing 2.5 wt% coke. The increase in the 10 MR line was not as important. From this, it was concluded that the pore inlets of the 8 MR channel were largely blocked by carbonaceous deposits. For a ferrierite containing 6.8 wt% coke, the pores in the 8 MR channels were completely blocked, as no oxygen signal was detected at the edge of the crystal, which indicated that on the outer surface a full carbon layer was formed. The amount of carbon in the 10 MR direction did not change in the same way. Even though an increase in the carbon to oxygen ratio was found also in this direction, based on other results (micropore and adsorption measurements) it is most likely that the pore inlets remained accessible for butenes. A Parallel EELS (PEELS) analysis of coke on ferrierite platelets was used by de Jong et a15' to verify that the amorphous layer observed on the borders of the crystals was effectively amorphous carbon. 4.3 Other Reactions on Zeolites. - Coke deposits on USHY, H-OFF, and HZSM-5 zeolites were characterized by EELS52.The EELS spectra were taken on small areas during a short period of time to minimize specimen drift and electron beam-induced damages. Because of this, the spectra were too noisy to proceed with a data reduction, and therefore no quantitative treatment of the EELS spectra was attempted. A qualitative analysis was carried out, comparing the spectra taken on the coked zeolite (or on the insoluble coke) with the EELS spectra taken on reference carbon compounds of known structure. Figure 5 shows the results. It was found that in HZSM-5 and H-OFF the coke forms an external envelope around the zeolite crystal and stands as an empty mold after zeolite extraction. Its structure is similar to that of coronene (polyaromaticpregraphitic). In USHY zeolite, the coke has a structure more like pentacene (linear polyaromatic).

Catalysis

180 B

A

300

Figure 5

5

400

eV

300

400

eV

EELS spectra. ( A ) taken on reference compounds, afer background subtraction, (a) graphite, (b) amorphous carbon, (c) coronene, (d) pentacene; ( B ) coke deposited on the external surface of zeolites, (a) U S H Y , (b) H-OFF, (c) HZSM-5. From Ref. 52

Infrared Techniques (FTIR, DRIFTS)

Infrared techniques have been used in many systems to study coke deposits. The main information that can be obtained with these techniques is the chemical identity of compounds that form the coke, such as olefinic, saturated or aromatic. Additionally, information regarding the location of coke can be obtained by following the signal of certain catalyst surface groups, such as Bronsted OH. This technique has been used to observe the deposition of carbonaceous materials on the working catalyst. IR flow reactor cells are now commercially available. The disadvantage of this techique is that often, it leads to limited information on the nature of carbonaceous deposits because of their complexity and of the difficulty to assign unambiguously an IR band to particular species53.Table 1 summarizes the IR bands assignments for coke characterization compiled from different sources. 5.1 Cracking. - Eberly et alS4studied the coke formed on a cracking catalyst with IR. The fundamental C-H stretches in the 3100 to 2800 cm-' were important. The aromatic C-H absorption appears at 3050 cm-', and methylene groups at 2930 and 2860 cm-'. The ratio of the absorbance of the infrared bands at 3050 and 2930 cm-' is a meassure of the aromaticity of the coke. In this case, it was found that the coke was formed by highly condensed aromatics of low hydrogen con tent. 5.2

Isobutane Alkylation. - The deactivation of a LaHY zeolite in the alkyla-

4: Coke characterization

Table 1

181

IR Bands in the 720 - 4000 em-' region of spectra, and their assignments

Band (cm-')

Assignment

Ref:

740-760 790-840 1036/1045 1110 1240 1336 1370-1380 1385 1390 1410 1420/1430 1437

Mono-substituted aromatics C H deformation in alkenes Y CC/CH3 linear butenes / iC4=bond IT linear butenes wagging CH2 long chain paraffins wagging CH2/iC,= dimeric branched chain oligomers aromatics-polyaromatics/6 CH3 olefinic species linear butenes linear butenes/ 6 = CH2 vinyl CH3asymmetric deformation (attached to aromatics) aromatic skeletal ring breathing /ti CH3 CH2deformation (attached to aromatics) y C = C aromatics y C = C aromatics/polyaromatics mycrocristalline polycyclic aromatics y C = C aromatics/polyaromatics C = C stretching in butenes iC4=bond J[; /olefinic species saturated CH stretching saturated CH stretching saturated CH / polyolefins + aromatic saturated CH/stretching CH/olefinic species Y CH aromatic

114 114 115 115 114 116 55,56 117 117 115,118 115,116 57,58

1450/1463 1471 1513/1524 1583/1595 1589 16 15/16 17 1622 1630/1625 2865 2925 2960 2970/2990 3045/3058

114,116 57,58 119 114,119 58 114,119 57,58 116,120 120 120 117,120 117,120 114,119

tion of isobutane with 1-butene was also studied with FT-IR technique55.The adsorption of 1-butene caused first the OH bands at 3558 (OH in sodalite) and 3624 cm-' (OH associated with A1 defects) to disappear, followed by the band at 3668 cm-' (OH in supercages), and ultimately also the vibration at 3742 cm-' of the terminal Si-OH groups was perturbed. These data indicate that there is chemical interaction between zeolite hydroxide groups and the reactant. Based on the analysis of the OH bands, the authors suggested that the less acidic OH'S in the supercages react more slowly with 1-butene than the other two O H types. The unreacted 1-butene was followed with the absorption band at 1630 cm-', and the amount of branched chain oligomers with the band at 1370 cm-'. The coke formed during the alkylation reaction is found to be formed mainly by aliphatic oligomeric species, with C-H vibrations in the 2750-3040 cm-'. When the coked catalyst was heated, the signal that corresponds to the silanol and the OH groups progressively appeared. The higher the acidity of the functional group, the higher the temperature needed to restore the IR absorption band, and therefore to remove the carbonaceous residues from the acid sites. As the treatment temperature was increased, a small band at 3050 cm-' evidenced the formation of aromatics. The technique was very useful in this study to determine

182

2932

0.4

8C CR

9 0

8

a

2880

137:

I I

a'

1

3500

\

3d00

2dOO

2600

Wavenumber (cm-l) Figure 6

*

1469 1390 148 1642

0.2

0

Catalysis

1560

FTIR spectra of (a) fresh and (b) deactivated Beta zeolite, at 75°C.From Ref: 56

how the acid sites deactivates during reaction and regenerates upon thermal treatment. However, at very low coke content, the technique has not enough sensitivity to analyse the coke left on the catalyst, and it is not possible to follow the whole regeneration procedure with the IR spectroscopy. Nivarthy et a156studied the carbon deposits on H-Beta zeolite during isobutane alkylation. Figure 6 shows the FTIR spectra of the fresh and the deactivated catalyst. They found that the hydrocarbonaceous deposits are highly branched multiple alkylate species, in agreement with Flego et a155.A very small signal at 1590 cm-', attributed to polyaromatic coke, was also detected. The bands shown in Figure 6 at 1372,1487,and 1642cm-' are attributed to paraffinic oligomers. 5.3 1-Butene Skeletal Isomerization. - The coke formed on acid catalysts during 1-butene skeletal isomerization has also been studied with IR techniques. Finelli et a133-34 studied the coke on ferrierite catalysts using DRIFTS. The bands assignments are essentially the same as those used by Flego et als5, and are summarized in Table 1. In this aliphatic and olefinic groups and aromatic rings were detected. The proportion of the aliphatic to aromatic coke, changes with the reaction temperature. The higher the reaction temperature, the higher the proportion of aromatic type of coke. Reaction temperature was varied between 350 and 400°C. Xu et a18studied the same reaction system at 420°C, and found that coke is mainly aromatic, with an increasing amount of highly polycyclic aromatics with time on stream. The assignment of IR bands used in this study corresponds to that reported by Barbes et aP7and Ghosh et a P , which are

4: Coke characterization

183

also included in Table 1.

5.4 Butene Dehydrogenation.- The coke formed on Cr203-A1203 catalyst during the dehydrogenation of butene to butadiene was studied by FT-IR59.In this study, curve fitting procedures were used to improve the information provided by FTIR spectroscopy. The aromatic C = C stretching region (1630 - 1500 cm- ') and the aliphatic C-H bending region (1450-1300 cm-') were used. A band at 1632 cm-' was assigned to stretching vibrations of double bonds in polyolefinic structures. The assignments are essentially the same as those presented in Table 1. Additionally, they observed bands at 900-500 cm-' region due to metal oxygen bonds, and at 3450 cm-'due to hydroxyl groups of the catalyst and adsorbed water. It was found that the coke deposited at low temperature (480°C) is formed by polyolefinc chains. As the reaction temperature increases, the aromatization of coke takes place. The aromatization was followed by the 1590 cm-' band. 5.5 Other Reactions on Zeolites. - Ghost et a15' investigated the nature of coke deposits on different types of zeolites, such as faujasite, mordenite and ZSM-5. Coke was deposited during propene reactions. It was found that the chemical nature of coke depends on the type of pore structure of the zeolite. At low temperature, below 327"C, mainly large branched oligomers were detected in faujasite, while smaller oligomers were detected for mordenite and ZSM-5. This was estimated based on the relative intensities of the overlapping 2960 and 2930 cm-' assigned respectively to the CH3- and -CHI- asymmetric stretching vibrations of the oligomer, and arbitrarily assuming that the oligomers formed in the HZSM-5 and mordenite are linear. At higher temperatures, polyaromatics were found both on faujasite and on mordenite. Small aromatic molecules were also found on the mordenite, and were the only type of compounds detected on ZSM-5. 6

Laser Raman Spectroscopy

6.1 Classical Laser Raman Spectroscopy (LRS). - 6.1.1 Naphtha Reforming. Espinat et a160used Laser Raman Spectroscopy to study coke formed on naphtha reforming catalysts. Two main features are mentioned by the authors for this technique: the high sensitivity that allows the analysis of catalysts with low coke content (0.3 - 0.5 wt%), and the possibility of following the graphitization of amorphous carbon. The technique may provide information regarding coke structure (pregraphitic or highly organized) and on the average dimension of the crystallite, as long as a monophasic carbon is produced, which of course is not always the case. Spectra obtained for Pt/A1203catalysts with coke varying from 0.29 to 27.3 wt% coke were very similar and of nearly the same intensity. On one hand, it shows the sensitivity of the technique, on the other, that there is no correlation between signal intensity and amount of coke. Well crystallized graphite displays a line at 1581 cm-'. Pregraphitic carbons show another broad

184

Catalysis

composite band with a maximum at about 1355 cm-'. The spectra obtained on coked reforming catalysts have two main bands centered at about 1600cm-' and at 1350 cm-'. The latter is about 50 % of the intensity of the former. The spectra is similar to that of pregraphitic carbons. The 1350 cm-' band is composite, having a shoulder at 1250 cm-'. Another not resolved band appears in between the two main peaks at about 1500 cm-'. All the metal containing catalysts (Pt, Pt-Re, Pt-IR supported on chlorinated alumina) displayed fluorescence. This fluorescence indicated that the carbon had a polyaromatic character. As acknowledged by the authors, the interpretation of the spectra is not straightforward, and the presence of different structures and particle sizes complicates the analysis, and further data are needed for a more complete interpretation. The surface fluorescence becomes a particular serious problem when the catalyst is contaminated with carbonaceous species, and therefore there are not many situations in which this technique has been applied to study coke on catalysts. 6.1.2 n-Butane Isomerization. The fluorescence was also observed by D. Spielbauer et a161in the study of a sulfated zirconia catalyst. The sample used in the n-butane isomerization reaction had a strong fluorescence, and therefore baseline subtraction was carried out. The following bands appeared in the coked catalyst: 300 cm-' due to deformation vibrations of CC3 groups, 725 cm-' due to deformation vibrations of aromatic C-H groups and strectching vibrations of CC3groups, 950 cm-' due to deformation vibrations of olefinic C-H groups, and 1610 cm-' due to C = C stretching vibrations of olefins and aromatics. After the reaction, the sample was regenerated in oxygen at 480°C. The Raman spectra of the fresh and regenerated samples were identical.

6.2 UV-Raman Spectrometry (UV-RS). - A new technique has been recently developed to overcome the above mentioned difficulty, using ultraviolet Raman s p e c t r o s ~ o p f ~It- ~was ~ . found that by using an ultraviolet wavelength below 260 nm the Raman scattering enhanced and the fluorescence background is avoided62.C. Li and P.C. Stair62used UV-RS to study coke on ZSM-5, USY, and Pt/A1203.Figure 7 shows the UV Raman spectra obtained for ZSM-5 and USY zeolites, coked with propene. The UV-RS spectra recorded for ZSM-5 coked with C3H6show three groups of Raman bands at 1390, 1635, and 3000 cm-' approximately. These bands are assigned to CH, deformation modes, the C = C streching, and stretching mode of C-H bonds in adsorbed olefinic species, respectively. As the reaction temperature is increased, the band positions are shifted slightly to lower frequencies. This means that the hydrocarbon species remain similar to those formed at lower temperature, but a fraction of them transforms into more stable species through polymerization and/or dehydrogenation. At longer time-on-stream at 500"C, the bands at 3000 cm-' drastically decrease, the band at 1625 cm-' become narrower and shifted to 1615cm-', and the band in the 1390 cm-' region become sharper. These results clearly indicate that chemical conversion of the adsorbed olefinic species toward poly-

185

4: Coke characterization

USY

ZSM-5 0

6

1600

2400

3200

Raman Shift, cm-' Figure 7

11600

2.400

3iOO

Raman Shift, cm-'

U V Raman spectra of coke species formed in ZSM-5 and U S Y zeolites. Catalysts coked with propene. From Ref. 62

merized olefinic species and aromatic species take place. In the case of the USY zeolite, the spectrum obtained at room temperature is almost identical to that for ZSM-5. However, at high temperature the spectrum changes significantly. The band at 2980 cm-' decreases dramatically, and the band at 1635 cm-' shifts to 1595 cm-'. This indicates that aromatic coke is more easily formed on the USY zeolite than in the ZSM-5, and that more dehydrogenated species are present in the former. In the case of the Pt/AI2O3catalyst, the spectrum displays two bands, at 1400 and 1600 cm-', which are attributed to polyaromatic species, in agreement with Espinat et aim. The coke formation in methanol to hydrocarbons conversion over zeolite H-MFI was also studied with UV-RSM.To distinguish between the signals that correspond to CH deformations and to CC stretches, experiments were carried out with deuterated methanol (CD30H). The bands assignments used in this study are summarized in Table 2. It is concluded that cyclopentadienyl species are intermediates in the formation of polyaromatic hydrocarbons. By comparison with pure polynuclear aromatics UV-RS spectra, it is suggested that coke

Table 2

Raman band assignments of retained hydrocarbons

Raman sh$t (cm-')

Raman band assignments

1605- 1615 1360- 14 10 1200- 12 10 1545-1550 1483

Ring stretches of polyaromatic species Ring stretches of polyaromatic species C-C stretches of polyaromatic species C =C stretches of conjugated olefins In-phase C = C stretch of cyclopentadienyl species

186

Catalysis

has a chain-like structure (similar to anthracene and pentacene) rather than 2-D sheet-like structure (like pyrene or coronene).

7

Dissolution of the Support and Solvent Extraction

The dissolution of the support and the solvent extraction are closely related. In many ~ t ~ d iafter e ~the~dissolution ~ ~ ~ ~of the - ~support ~ with a strong acid, the coke is extracted with different solvents. In some other cases, the solvent extraction is carried out directly on the coked ~ a t a l y s t s ~ ~The ~ ~combination ~ @ ~ ~ ~ of ~~~-~~ coke extraction followed by support dissolution was also used in several studies. The dissolution of the support has as a major disadvantage, the fact that the coke could be modified during this procedure. Even though Magnoux et a16*verified in their system that the acid treatment did not modify the coke, it is questionable whether this is true in every situation. In fact, they checked this by impregnation of 1-tetradecene and 9-methylphenantrene on an inert support (SiOz),dissolving the support with 40 % HF, and analysing the products dissolved in C12CH2. In this case, it was found that the GC of these compounds were identical to that of the starting compounds. Guisnet and M a g n ~ u xfound ~ ~ that on H-Erionite, the acid treatment with HF eliminated highly volatile compounds that were trapped on the pores of the zeolite. Also, on H-Offretite it was found that the more volatile compounds were eliminated during this treatment, and also during the evaporation of the methylene chloride, the more volatile compounds were eliminated69.Also, it must be kept in mind that the identification of coke species after dissolution of a catalyst coked for a given time-on-stream, is not necessarily related to the deactivation process, which in many cases occurred at the beginning of the reaction, and therefore cannot lead to a meaningful description of the deactivation process74. Several solvents have been used to obtain information regarding coke functionality, according to the solubility in a given solvent. An elution scheme has been proposed using 8 different solvents, as indicated in Table 375.Alfonso et a176 used a sequential solvent extraction procedure, using n-hexane, chloroform and toluene, to study coke composition on a Pt-Sn dehydrogenation catalyst.

Table 3

Coke functionality and solvent ajfinity

Solvent or mixture

Eluted compounds

1 Hexane 2 Hexane-benzene (85:15) 3 Chloroform 4 Chloroform-diethyl ether (90:10) 5 Diethyl ether-ethanol (97:3) 6 Methanol 7 Chloroform-ethanol (97:3) 8 Pyridine-ethanol

Saturated hydrocarbons Aromatic hydrocarbons Polar aromatics, non-basic heterocycles Monophenols Basic nitrogenated heterocycles Highly functional molecules Polyphenols Molecules with a high content of N and 0

~

Elution Scheme from Ref. 75.

~

~

~~

~~~~~

4: Coke characterization

187

7.1 Naphtha Reforming. - Coke formed on naphtha reforming catalysts was analysed by XRD after dissolution of the alumina support with 40 wt% H F mixed with 37 wt% HCl at 60°C45.Large amounts of catalyst had to be treated to obtain enough coke for further analysis. Even though the structure of the graphitic coke seems to be unaffected by this treatment, it is likely that certain types of coke can be strongly modified in the presence of such strong acidic media. 7.2 Coke on Zeolites. - A ZSM-5 catalyst used in n-hexadecane cracking, was treated with 40 wt% H F after two successive extraction steps with C C 4 and CH2C12. The released coke was extracted with CH2C12to separate ‘soluble’ and ‘insoluble’ coke residues72.The insoluble fraction was studied with NMR (see section 9). Guisnet and M a g n ~ u xstudied ~ ~ several zeolites following a procedure that involved support dissolution with H F at 40% in order to liberate the inner coke. Then the coke was treated with CH2C12. The soluble fraction was analysed with several techniques, such as GC, HPLC, NMR, IR, MS and GC-MS. The nonsoluble coke was analysed to determine the H/C ratio by TPO, and physically characterized by electron microscopy and EELS. In this study they characterized by this method the coke deposited on USHY, HMOR, HZSM-5, and HErionite, during the cracking of n-heptane at 450°C. They found that at low coke content, it is non-aromatic with a high H/C ratio, and that it is always more aromatic for large pore zeolites than for small or intermediate size pores. It was also found that most of the soluble coke molecules are too volatile and too weakly basic to be located on the outer surface, and therefore they should necessarily be located in the pores. In this study, nitrogen and n-hexane adsorption measurements were carried out (see section 13). The same procedure was used to study the coke formed from propene and isobutene on 5A adsorbentP. A solvent extraction technique using two different solvents (CC4and CH2C12) was used to obtain information regarding coke location and composition in a ZSM-5 zeolite catalyst during the cracking of h e ~ a d e c a n eThe ~ ~ . composition of the soluble fraction was determined by GC-MS analysis. A first extraction with CCh, which is a molecule too large to enter the pores, removed the coke deposited on the external surface. A second extraction with CH2C12which supposedly can enter unblocked pores, removed material accessible within the pore system. Finally, the zeolite was treated with 40%HF to destroy the framework. The coke thus obtained was again treated with C12CH2.The coke that is not dissolved, is called ‘insoluble’fraction. The soluble fraction was grouped into the following compound types: alkanes, alkenes, substituted benzenes and substituted naphthalenes. The latter were the largest molecules found in the soluble fraction of the coke in this small-pore zeolite. In large pore zeolites, larger molecules were found in CH2C1267~77. Alkenes were not found on the external surface. The most significant contribution to the soluble fraction were the substituted benzenes, while the naphthalenes were very low. The concentration of these latter compounds increased with time-on-stream, which implies that they are products of the coking reaction and that do not react further. Because of

188

Catalysis

their size these compounds may lead to some pore blockage. Coke deposited on HY zeolite during n-heptane reaction at 450°C, was completely dissolved by C12CH2after dissolution of the support with HF6'. This is not usual, since in most studies there is a fraction of insoluble coke. The coke formed on PtUSHY and PtHMOR catalysts during benzene hydrogenation at 80°C was also completely solubilized in C12CH2 after support dissolution with 40% HF7'. In this case, since the coke was deposited at low temperature, such a high solubility could be expected. The coke deposited during the isobutane alkylation with C4 olefins on zeolites was extracted with a mixture of methanol/toluene, and with CI2CH2.None of these solvents were effective to remove the coke. Only a very small fraction was removed, which was attributed to the dissolution of external coke. The coke molecules formed inside the channels have a large size and therefore cannot diffuse out of the pores36.

7.3 Paraffins Dehydrogenation. - The coke deposited on a Pt-Sn-In/LiA120~commercialcatalyst, coked in a pilot plant reactor during the dehydrogenation of C 10-C13 n-paraffins, was A detailed quantitative solidstate NMR characterization of the coke deposits was carried out. It was found that the nature and composition of coke changes depending upon process parameters. The insoluble fraction in C12CH2 was essentially polyaromatic, while the soluble fraction was rich in alkylated mono- and diaromatics. In this study, the H/C ratio of the insoluble coke was determined by NMR. It was found that it was lower than that obtained by elementary analysis before the solvent extraction procedure, which also indicates that the insoluble coke is much more condensed than the soluble fraction. 7.4 Propene Oligomerization on Heteropoly-Acids. - Supported heteropolyacids have received considerable attention in a variety of reactions in the last few years. As with any other acid catalysts, coke deposition is a cause of catalyst deactivation. The coke formed on H3PW12040/Si02 catalysts during the propene oligomerisation was studied, by extracting with either toluene, cyclohexane, or dichloromethane7*.According to this study, the C12CH2 was the most effective one to remove coke, but mainly from the soft fraction of coke, defined as the coke that is removed in the TGA bewteen 170 and 370°C. As time on stream increases, the fraction of coke removed by C12CH2 decreases.

7.5 n-Butane Isomerization. - Coke deposited on a sulfated zirconia catalyst during the isomerization of n-butane, was extracted with different solvents, and the coke left on the catalyst analysed by TP043.The extraction with pure solvents had the objective of assesing the functionality of the coke by its solubility in specific solvents. Solubility in hexane and benzene was practically zero. Methanol and piridine were more effective in the extraction. According to these results, the only conclusion the authors could draw was that the coke was not completely paraffinic. In the extraction with piridine, after solvent evaporation, the solid was analysed by FD-mass spectrometry. Fragments up to m/z = 2737

4: Coke characterization

189

were found. Two maxima at m/z = 352 and 532 indicated the presence of polycondensed aromatic fragments of 25 to 40 carbon atoms.

8

Neutron Scattering and Attenuation

The scattering of neutrons is used similarly to X-rays or electrons scattering. The technique is especially sensitive to hydrogen since the scattering cross section for hydrogen is considerably greater than for any other atom. It is possible to study the vibrational transitions (inelastic scattering), the atomic and molecular diffusion (quasielastic scattering), and the particle (or pore) sizes (small-angle scattering)I2'j. A portable neutron source was employed to measure coke content and C/H ratio on a zeolite during cumene cracking79,and on chromia-alumina catalysts coked with butene". This technique has the advantage that the coke content and therefore the coke profile along the catalyst bed, can be measured in-situ. However, this calculation requires two assumption^^^. The first one is that the catalyst packing density is uniform along the reactor, and the second is that the C/H ratio is constant at all points within the reactor and at all times. It is necessary to know this value in order to calculate the coke content. The technique is more accurate at high coke content on the catalyst. At coke content below 1%, errors as high as 100% can be If the coke content is known, the technique can be applied to determine the C/H ratio". Small-angle neutron scattering (SANS) has been used to characterize a silicaalumina catalyst before and after coke deposition, during xylenes isomerization". This technique also requires the prior accurate knowledge of the density and composition of the scattering phases within the catalyst. Nitrogen adsorption isotherms, mercury intrusion measurements, and X-Ray diffraction data were collected in order to apply the SANS technique. Additionally, contrast variation experiments are carried out to obtain information on the pore size distribution and how coke might block some of these pores. This experiment consists in filling the pores with a fluid with the same scattering power as that of the catalyst matrix. In this case, a mixture of H 2 0 / D 2 0was used. It was found that the coke was deposited as a monolayer mainly in the porous structure of 3.3 nm, while it appears that no coke was formed in the micropores of less than 1 nm. 9

Nuclear Magnetic Resonance (NMR)

The NMR technique is a powerful technique to investigate the nature of carbonaceous deposits. As highlighted by de Jong et a150,the 13CCP/MAS - NMR can detect bindings between aromatic ringss2, alkyl fragments, and even tertiary carbenium ion-like speciess3.The information provided by NMR is not quantitative. Special care has to be taken with the coked catalyst sample after the reaction, to avoid the effects of oxygen or water on the relaxation process, and to avoid any confusion this may cause in interpreting the chemical structure of coke

190

Catalysis

on the basis of NMR r e s ~ l t s ’ ~ ~ ~ ~ . The 13CCP/MAS - NMR spectroscopy cannot distinguish between the olefinic and the aromatic carbonsS5,since signals due to these types of carbons appear in the same region of the spectra. This difference can be detected by UV-VIS. Certain types of coke, such as conducting coke or very slowly relaxing polyaromatics are not always fully detected by 13CCP/MAS - NMR spectroscopy55. 9.1 13CCP/MAS - NMR. - Table 4 summarizes the band assignment currently used for the identification of the functional groups of coke, using 13CNMR. 9.1 .I Coke on Zeolites. Holmes et a172used 13CCP/MAS - NMR to characterize the insoluble fraction of coke deposited on a ZSM-5 catalyst during the cracking of hexadecane, after dissolution of the support. The spectrum of coke deposits formed at 1 h on stream showed only peaks in the aliphatic region, at 13,22, and 30-32 ppm (using tetramethylsilane as reference). These signals correspond to methyl, a-methylene and internal CH2 groups, respectively. These assignments were previously reported by Lange et a183.After 6 h on stream, the spectra changed dramatically and no more peaks in the aliphatic region were observed. Only a broad hump around 144 ppm indicative of numerous polyaromatics species, appeared in the spectrum. The coke deposited during the cracking of n-nonane on USHY zeolite was also characterized with this technique84.The 13CCP/MAS - NMR spectra were recorded after 15,30,60 and 450 sec of reaction. Figure 8 shows these spectra. It was found that the coke contained both aliphatic and aromatic components. As time on stream increased, the intensity of the signal that corresponds to the sp3 region at 13 ppm (methyl groups) and at 30 ppm (tertiary carbon groups)

Table 4

13C CPIMAS - N M R signal assignments

Signal (ppm)

Assignment

Re$

13 15 19 20 21 23 30 and 32 30 30-50 125 128 130 135 138-141

methyl group connected with secondary carbon, aliphatic methyl group bound to paraffinic carbon methyl group linked to aromatic secondary carbon, aliphatic methyl group bound to olefinic carbon CH2group linked to methyl groups other paraffinic CH2groups methyl group at quaternary carbon secondary and tertiary paraffinic carbon aromatic C-H groups aromatic C-H groups sp2bonded carbon alkylated aromatics, carbon bridges between aromatic rings substituted aromatic carbons, bridged carbons in condensed aromatics, olefinic CH groups polyalkylated aromatic C - 0 bonds, aromatic

87 87 87

140-150 160

87 87 87 87 87 38 87 74 74 87 74 38

4: Coke characterization

Figure 8

191

I3C CPIMAS N M R spectra of coke at 15,30,60, and 450 s reaction time, on U S H Y- zeolite, coked with n-nonane. From Ref: 84

decreased relative to the sp2 region at 130 ppm signal. The secondary carbon groups at 20 ppm decreased more gradually. Similar assignments were used in other studies? Derouane et alg6studied the carbonaceous deposits on different zeolites, after reactions of either methanol or ethylene. The coke formed on the HZSM-5 zeolite contained mainly aliphatics and alkylbenzenics compounds, while on the H-Mordenite zeolite it contained polyaromatics compounds. The carbonaceous deposits formed on a LaNaY zeolite during the isobutane/butene alkylation reaction was studied with 13CCP/MAS - NMR spectroscopf7. The reaction was carried out between 80°C and 314°C. At 80°C, the spectrum showed only resonances in the paraffinic region (0-50 ppm), which was very interesting, since the only products at the end of the reaction were olefins. As reaction temperature increased, the intensity of the resonance in the aromatic/olefinic region (100-160 ppm) had a significant increase, while the paraffinic region changed. At low temperature (8OOC) the H/C ratio was 1.8, which could be due to a minor amount of unsaturated molecules. However, such species were not detected by 13CNMR, and therefore the low H/C ratio found at low temperature was explained by considering the presence of some multi-ring naphthenes. This statement could not be proved unequivocally with the data gathered in this study. The band assignment used by Weitkamp and Mai~nel.8~ is summarized in Table 4. Flego et a15' studied the alkylation of isobutane with 1-butene, using LaHY zeolite. They also found that the coke deposits are mainly paraffinic (10 to 40 pprn), with small amounts of olefinic/aromatic carbons (130 ppm). The I3CNMR chemical shifts cannot distinguish these two types of unsaturated species. The differentiation was done using UV-VIS, and found that only olefinic carbons

192

Catalysis

were present on this catalyst. On heating the coked catalysts, there is a progressive transformation of the aliphatic/olefinic carbons into aromatic. The influence of the acidity of ZSM-5 zeolites on coke composition was studied8*.I3CCP/M AS-NMR along with dipolar dephasing was used to estimate the fraction of aromatic carbon and variation in aliphatic components, as a function of the acidity which was changed by steam dealumination at different temperatures. The 13CCP/MAS-NMR with dipolar dephasing is used for identification and estimation of non-protonated aromatic carbons as well as freely mobile methyl groups. Figure 9 shows some of the spectra reported by Sahoo et a188.As the severity of steaming increases, the intensity of aromatic-olefinic band decreases, while all the aliphatic peaks gradually increase. The solid state 13C dipolar dephasing NMR spectra showed an increase in the spinning side bands, suggesting that the carbons present in the coke are highly rigid and that almost all are aromatic quarternary carbons. The transformation of acetone on HY and HZSM-5 zeolites was also studied by NMR technique^^^. On both types of zeolites, signals in the range of paraffinic carbons (-CH3,-C2H5)bonded to the aromatic rings were observed as a signal at 19.8 ppm. The signal at 130 ppm that corresponds to aromatic carbon was also observed in both zeolites. ZSM-5 displayed a signal at around 154 ppm, which was not observed on the HY zeolite. This signal was assigned to carbons belonging to an aromatic ring bonded to oxygen. 9.1.2 Residue Hydrotreating. A commercial Ni-Mo/y-A1203hydrotreating catalyst was used to process a heavy petroleum The coke left on the catalyst was characterized by 13CCP/MAS-NMR technique, in order to determine the carbon types present in the coke, after times on stream ranging from 100

250

Figure9

200

150

100

50

0

PPm

13C C P / M A S N M R spectra of spent ZSM-5 catalysts, (a) unsteamed, (b) 500°C steamed, and 13C dipolar dephasing N M R spectra, (c) unsteamed, (d) 500°C-steamed. From Ref: 88

4: Coke characterization

193

to 7400 h. The signals centered at 125 ppm and 160 ppm correspond to aromatic carbons (C-H and C - 0 bonds respectively). The first signal is found in all samples (from 100 to 3100 h on stream), while the second one is clearly observed after 2100 h on stream. Other resonance signals between 10 and 100 ppm, which correspond to aliphatic carbons, are found at short time on stream (100 h) and gradually disappear. A similar catalyst was used in the study of Diez et a171,deactivated with either a light catalytic cycle oil or a coal residuum. As the reaction temperature was increased from 335°C to 395"C, the amount of coke increased from 1.5 wt% to 3 wt% and the aromaticity from 40% to 70%, as measured by 13CCP MAS NMR. The aromaticity was calculated from the percentage of sp2 carbon signal. When using the coal residuum as feed, the aromaticity increased up to 73% and the coke content up to lo%, at 415°C.

9.2 'H NMR. - 'H NMR was also used to study HY zeolites coked with n-heptane at 450"C6'. The study discriminated among aliphatic or alicyclic protons, alkyl aromatic protons, aromatic protons, or protons linked to carbon near oxygen atoms, before and after oxidation treatment at 400°C. It was found that almost all the alkyl aromatic and aromatic protons are eliminated by the oxidation treatment, while the protons linked to saturated carbons are little affected, and signals characteristic of protons situated near oxygen atoms could be observed. The aromaticity of coke deposited on H-Offretite during n-heptane cracking at 450"C, increases with coke content. The percentage of aromatic protons increases at the expense of the percentage of aliphatic protons69. 9.3 '29XeNMR. - This technique has been used to study coke locationss on zeolite catalysts. The 129XeNMR spectra in the fresh USY zeolite has a sharp band at a chemical shift of 6 = 68 ppm. In coked USY zeolites, there is a broadening and an increase in the chemical shift. This is due to the decrease in the zeolitic void space due to coke, and to an additional contribution from xenon-coke interactions. A USY zeolite coked at 200°C during the n-propylbenzene disproportionation reaction, displays two overlapped peaks, at chemical shifts of 75 and 68 ppm, which indicates that the coke is heterogeneously distributed. In the case of zeolite coked at higher temperature, a single more symmetric peak appeared at 79 ppm, which indicates homogeneous coke depositsW.In this study8', the 129XeNMR spectra were obtained at different xenon loadings on USY and ZSM-5. On the latter, at low xenon loading two overlapping bands were observed, while at greater loadings a single line was observed. Based on these observations, the authors concluded that on the ZSM-5 catalysts, at low coke concentrations, the coke tends to deposit within the crystallite, while on USY zeolite the coke distribution depends on the reaction temperature. The technique was used to study coking and regeneration of a HY zeolite, using n-hexane or propylene, obtaining different levels of coke content on the zeolite, between 5 and 33 wtYo3'. 129XeNMR spectra were collected at different Xe loading, and the resonance frequency extrapolated to zero pressure was used

194

Catalysis

as reference. The spectra for the catalyst with 5 wt% coke, displayed a narrow and symmetrical line. It became much broader and asymmetrical for the sample with 15wt% coke, indicating that the xenon atoms were much less mobile. The presence of paramagnetic centres, the dipole-dipole interaction with the protons of the deposited coke and the chemical shift distribution, may also contribute to the width and asymmetry of the NMR line''. The spectrum of the most heavily coked sample, containing 33 wt%C was composed of a broad asymmetrical component at ca. 157 ppm and a narrow quasi-symmetrical line at ca. 10 ppm, when the adsorption was carried out at 937 torr of Xe. The first line was due to the strongly adsorbed xenon atoms in the modified supercages, whilst the other resonance was attributed to xenon atoms adsorbed in mesopores formed by the coke deposited between the crystallites. The variation of the chemical shift of the broad NMR lines as a function of the amount of adsorbed xenon on the different samples was used to discriminate between internal and external coke species. It was found that coke formed from propylene in large quantities (above 15 wt%) obstructs the access to some supercages. The presence of external coke was shown by the new NMR line, characteristic of xenon atoms adsorbed in macrocavities. Upon regeneration, the external coke was first eliminated. 9.4 29SiMAS NMR. - G r ~ t e et n ~a1~studied the deactivation of USHY catalyst during the cracking of n-hexene, with NMR techniques. The *'Si MAS NMR spectra indicated that there are significant interactions between the framework silicon atoms and the coke, which was observed as a considerable broadening of the Si(lA1) and Si(OA1) peaks with time-on-stream. This linewidths were fully restored upon regeneration of the catalyst by combustion. According to the authors, the possible reasons of this broadening were the structure, concentration, and distribution of the coke as well as the presence of free radical species in the coke. The broadening that occurs after 20 s time on stream, when structural changes in the coke have apparently ceased (as seen by 13CCP MAS NMR) and the coke loading per gram of catalyst is constant, suggest that these two factors are not contributors to the broadening after 20 s TOS. The linewidth remains unchanged when decreasing the delay, changing the relaxation time. This suggests that the free radicals were not a significant factor in this study. The other possible source of the broadening was the distribution of the coke molecules in the zeolite crystals. The analysis of the individual linewidths of the Si(lA1) and Si(OA1)peaks suggested that this is the factor that dominated the line broadening, 10

Auger Electron Spectroscopy (AES)

Auger electron spectroscopy (AES) was used in combination with secondary ion mass spectrometry (SIMS) to distinguish between four types of carbonaceous deposits, on metal foils (rhodium, iridium and platinum)92.The foils were coked by exposing to ethylene at low pressure. Auger spectroscopy can distinguish between molecular or carbidic on the one hand, and graphitic or amorphous carbon on the other. The Auger spectrum of carbonaceous deposits on a metal is

4: Coke characterization

195

a combination of three spectra: two of carbon and one of the substrate. The characteristic fine-structure of the carbon Auger spectra between 240 and 265 eV interferes with features in the spectra of the metals. The carbon Auger spectra in the energy range between 265 and 285 eV is not affected by features from the metals, and hence visual inspection of this part of the Auger spectrum can be used to draw qualitative conclusions on the type of carbon in the deposit. On the assumption that changes in the spectra of the constituents due to chemical interactions can be ignored, the fitting of the carbon spectrum on a metal with a linear combination of the appropriate base spectra, makes it possible to obtain more quantitative description. The amount of coke (as a fraction of monolayer) has been derived by following the intensity ratio of a carbon peak (273 eV) and a metal peak. Using the fitting, the fraction of amorphous/graphitic carbon in the deposit was estimated. It was found that the coke on Pt has the highest proportion of this type of carbon, followed by IR and then Rh. To discriminate between the molecular of carbidic carbon or between amorphous and graphitic, whose main difference is in the hydrogen content, Secondary Ions Mass Spectrometry was used. Goodman et a193studied the kinetics of surface carbide built up on a Ni(100) surface using AES. The signal C/Ni versus time was analysed as a function of reaction time. They found two regions of carbon build-up: a carbide region at C/Ni coverages less than 0.28 and temperatures below 377"C, and a graphite region at C/Ni coverages greater than 0.28 and temperatures above 377°C.

11

X-Ray diffraction (XRD)

Coke structure can be characterized by X-Ray diffraction analysis. This technique makes it possible to determine if there is coke with crystalline structure on the catalyst. However, the sensitivity of this type of determination is rather low, being it difficult to determine the fraction and/or amount of coke in the crystalline form. Support dissolution procedure was also used to analyse the coke free from support by XRD. Support dissolution procedure for coke XRD analysis, is more appropriate when the coke content on the catalyst is high, and as long as the strong acidic media used in the dissolution does not alter the coke structure. Coked catalysts and the coke after support dissolution were analysed with XRD by Espinat et a14'. They studied naphtha reforming catalysts, and compared the spectra of the fresh Pt/A1203 catalyst, and the catalyst containing 11.5 and 27.3 wt% coke. A characteristic band at 28 = 25"(d = 3.56 A) was the main difference between the spectra of the coked catalysts and the fresh catalyst. This line corresponds to line (002) of pregraphitic carbon (d(002)= 3.35 A). These results indicate that the coke on the reforming catalyst has indeed a certain degree of organization. This technique has low sensitivity in this system, since the carbon signal remains very weak even at high coke ratios and is often masked by the signal of alumina. The XRD analysis of coke after dissolution of the alumina by acid attack shows an intense diffraction band corresponding to line (002), which represents the stacking of the graphite sheets. The separation of these

196

Catalysis

sheets was 3.5 A and the crystal size close to 30 A, which corresponds to about 10 superimposed sheets. Other bands observed in the XRD spectra were at 28= 43.25", or d = 2.09 A, which corresponds to the sum of the two bands corresponding to (100) and (101). using Coke on reforming catalysts was previously studied by Figoli et XRD of both, the coked catalyst and the coke after alumina dissolution with hydrofluoric acid. In this study the objective was to compare the coke deposited in an industrial reforming reactor, with coke deposited in an accelerated deactivation test in the laboratory. The XRD spectra of both types of coke were very similar, suggesting the existence of interlayer structures separated by 3.46 A and 3.44 A in the samples coked in the plant and in the laboratory, respectively. The XRD spectra after alumina dissolution showed diffraction lines corresponding to inter-layer structures with separations of 3.46 A (002 plane) and 2.039 A (101 plane) for the coke deposited in the industrial reactor, and 3.44 A and 2.039 A for the coke deposited in the laboratory test. These distances are greater than those corresponding to graphite, which are 3.37 A (002 plane) and 2.027 A (101 plane). According to the authors, this suggests that the coke probably consists of unorganized aromatic systems and of alkyl chains and alicyclic moieties joined to a polynuclear aromatic system. Similar results concerning the coke structure were reported recently by J. Ruixia et all2', on Pd-La/spinel catalyst, coked during the production of 2,6 diisopropylaniline at 220°C. In the case of the ZSM-5 zeolite catalysts, it has been described that it undergoes a displacive transformation from monoclinic to orthorhombic when it occludes bulky ions, altering the relative peak positions and intensities in the XRD pattternsg4.The deposition of coke inside the channels was thus verified by means of changes in the XRD spectra of coked vs fresh catalystsg5.The XRD pattern of the coked catalyst is similar to that of the catalyst with the occluded template ions.

12

Secondary Ion Mass Spectrometry (SIMS)

SIMS is among the few surface sensitive techniques which are capable of detecting hydrogen content of the deposits. It has been used in combination with AES to analyse coke on metal foils92.In this study the peaks of the positive ions CHnf and the CHn-, CZHn-, and C4Hn- ions in the negative SIMS spectra were used. This is because other species present in only trace amounts such as Na, Al, K, 0 and C1, with high sensitivity may appear quite clearly in the SIMS spectra, and therefore limits the number of peaks which can be used to characterize the carbon deposit. The SIMS patterns of the above mentioned ions are indicators of the nature of the carbonaceous deposits. A qualitative conclusion is obtained by looking at the distribution pattern. The higher the reaction temperature, the lower the hydrogen content. However, it was not possible to obtain the real H/C ratio of the carbon deposit, since an empirical parameter was defined to follow the change in hydrogen content of the coke deposit. For example, when the

4: Coke characterization

197

carbon deposit was formed at 52"C,the Auger spectra indicated that the carbon was either molecular or carbidic. The high value of this parameter indicates that the deposit is highly hidrogenated, and therefore, mainly molecular carbon is present in the deposit. Similarly, the ratio of the intensities of C4- to Cz- ions was defined as the graphitization parameter, which correlates well with the tendency of the carbon to polymerize into graphitic structures. It is not clear if the presence of graphitic carbon on the deposit formed at low temperature (52=C)is due to electron-beam induced degradation, or to chemical reactions on the overlayer. This is another uncertainty present in this technique. SIMS was also used to study naphtha reforming catalyst^^^. It was found that the carbon profile along the pellet had wide irregularities. Two different regions could be distinguished, areas quite free of coke where the carbon signal was very low, and areas where the peak signal was close to the maximum. The size of these areas varied between 20 and 100 pm.

13

Sorption Capacity: Surface Area and Pore Volume

Adsorption measurements allow the determination of coke location. When the volume occupied by coke is much smaller than the volume inaccessible to adsorbates, it means that there is a pore blockage. However, in many cases the adsorption study is carried out at a different temperature than the reaction, and therefore diffusivity could be quite different. Another aspect that should be taken into account is that if the pretreatment for adsorption measurement requires temperatures higher than the reaction temperature, an important fraction of carbonaceous deposits could be stripped off the catalyst and, therefore, the pore volume measured in this way will be higher than the actual volume under reaction conditions. A modification in the coke structure might also occur under these circumstance^^^. 13.1 Coke on Zeolites. - Nitrogen and n-butane adsorption studies were carried out to obtain information regarding coke location on HZSM-5 catalystsg5. The n-butane adsorption was studied at 50°C, while the reaction temperature (acetone/n-butanol to aromatics) was 400°C. The decrease in n-butane adsorption capacity is compared with the residual aromatization activity. At low coke content, these curves are identical, and therefore the deactivation is due to site coverage. However, at higher coke content (beyond 2%) the curve that corresponds to the adsorption capacity decreases to a higher extent in some catalysts as compared to the residual aromatization activity. This is attributed to a higher mobility of the coke molecules at the reaction temperature (400°C).Therefore, if the adsorption study is carried out at a temperature different than the reaction temperature, a lack of correlation between the observed changes in adsorption capacity and the activity under reaction conditions, could be observed. The adsorption capacity loss has been calculated as the ratio between the volume occupied by coke (VR),and the volume not accessible to the adsorbate (VA). The first one, is estimated with the amount of coke and assuming a coke density of 1.2

198

Catalysis

g/cm3, which corresponds to polyaromatic coke. The ratio of these two volume allows one to determine the eventual blockage by coke of the access to part of the pore volumen. If VR/VAequals 1, there is no pore blockage and only site blocking occurs; if VR/VA is much lower than 1 there is pore blockage. Values close to one do not provide any definite information, since large uncertainty exists in coke density. Values greater than one have also been found, which is due precisely to this wrong estimation of the coke density. A. de Lucas et a19' measured the specific surface and micropore and mesopore area on fresh and coked HZSM-5 catalysts. The large decrease in micropore area indicated that the coke was mainly deposited on these types of pores, and that pore blocking occurred. Adsorption measurements of nitrogen (-196°C)and n-hexane (0°C)on USHY, H-MOR, HZSM-5 and H-Erionite that were coked with n-heptane at 450"C, were carried The inaccessible pore volume in each case (VA), and the volume occupied by coke (V,) were calculated. Typically, for low coke content the value is close to 1 (except for H-ERI), and at higher coke content it drops. The comparison of nitrogen and n-hexane results made it possible to discuss whether the coke inhibited the access of reactants to the active sites or not. The BET surface area was determined for both fresh and spent catalysts, during the isobutane alkylation with l - b ~ t e n eLaY ~ ~ . and Lao zeolites displayed a decrease in the BET area of 45 YOapproximately due to the coking, while the amorphous silica alumina a decrease of 33 %. This is the case where the pretreatment of the coked catalysts before the BET determination will eliminate some of the carbonaceous deposits, since the reaction temperature is typically below lOO"C, and the pretreatment for BET determination with zeolite catalysts, is usually around 250°C. TPO studies36clearly demonstrated that this treatment under vacuum eliminates a fraction of the coke, and therefore the real decrease in surface area due to coke deposition is larger than that measured by BET. Querini and Roa36also studied the coke on zeolites during isobutane alkylation, by measuring the surface area and pore volume. The reaction was carried out at temperatures up to 80°C. The pretreatment of the coke and the fresh catalysts used to determine the BET surface area were carried out using different heating rates and final temperatures. The amount of coke was also measured after the BET determination. It was found that the temperature of the pretreatment had a very important effect on the micropore volume and in the BET area. For example, if a deactivated mordenite is pretreated at 350"C, the amount of coke decreased from 5.6 to below 1.5wt% (depending upon the heating rate), and if pretreated at 200°C, the amount of coke decreased approximately to 3.85 wt%. In the former case, the micropore volume and the BET area decreased 9% approximately, while when the pretreatment was carried out at 200°C and a smaller amount of coke was removed, the decrease in micropore volume and surface area was approximately 50-60% that of the fresh catalyst approximately. In the case of a LaY zeolite, the heating rate had a major effect. In this catalyst, even at a high pretreatment temperature (350°C) the coke content decreased from the 11.2wt% of the deactivated catalyst to 7.5 wt% and 4 wt% for a heating rate of 40"C/min and 10"C/min,respectively. The BET area and the pore volume

4: Coke characterization

199

was approximately 57% of that of the fresh catalyst. At a pretreatment temperature of 2Oo0C,and a high heating rate, the pore volume and the BET area were negligible. This important difference between the mordenite and the Y zeolites is due to the deactivation mechanism. While the pores of the Y zeolites are essentially filled up with coke before the deactivation, the mordenite deactivates due to pore mouth blockage. When removing a small part of the coke in the mordenite, the whole pore structure is available for adsorption. In the case of the Y zeolite, the bulky coke molecules must gasify and diffuse out of the pore structure before releasing free surface area. If the heating rate is too high, there is not time enough for this to occur, and the aromatization of these coke precursors does not allow to liberate adsorption sites. Many other adsorbates have been used to test pore accessibility and adsorption capacity changes upon coke deposition. Some examples are trimethylamine and ethyldiisopropylamine on ZSM-597,m-xylene on USHY and HZSM-598, 3-methylpentane in H-Offretite99,xenon on ZSM-5 and USY zeolites8’. The effects that coke formed from propene and isobutene has on the adsorption capacity of a 5A zeolite, were studied by Boucheffa et allm.The decrease in the adsorption capacity of the zeolite with the amount of coke formed from propene, was independent of the coking temperature (100 to 420°C). On the other hand, at high coke contents (above 2 wt%) the coking temperature had a very pronounced effect on the adsorption capacity, when coke was formed from i-butene. The adsorption capacity of the coked catalyst was studied also after partial regenerations. Catalysts coked with propene, displayed the same adsorption capacity for a given coke content regardless if the catalyst was coked or if it was coked and regenerated. In the case of the catalysts coked with isobutene, the adsorption capacity was much higher for a given coke content, when the catalyst was partially regenerated. This indicated that in the case of the isobutene, coke initially deposits in all types of sites, but then it blocks the pores. When coke is partially burnt off, it causes a large increase in surface area, since now all the pore volume not occupied by coke molecules ise now accessible to nitrogen.

13.2 Residue Hydrotreating. - Nitrogen adsorption measurements were carried out on a Ni-Mo/y-AI2O3fresh and coked hydrotreating catalyst7’. From these experiments, BET surface area, pore volume, and pore size distribution were calculated. A small decrease was found both in surface area and pore volume when the catalyst was coked with a light catalytic cycle oil, which was approximately around 10%. Estimates of the thickness of the coke, assuming uniform coke layers and cylindrical pores, made it possible to calculate the pore size distribution, which agreed with the measured value. This is consistent with the conclusion that the deposits were nearly uniform and increasing in thickness with increasing amounts of coke. 13.3 Isobutane Dehydrogenation. - The pore size distributions of fresh and spent sulfided nickel catalysts were determined by a standard nitrogen desorption technique. Two catalysts with different pore size distribution were coked during the dehydrogenation of isobutane28.The pore volume was measured as a

200

Catalysis

function of the carbon content. It was found evidence of pore plugging (10-20 nm size range) when the coke content was 5 wt% for the catalyst with a mean pore diameter of 100 A, and when the coke was about 10 wt% in the catalyst with mean pore diameter of 115 A. In this case, the larger the mean pore diameter, the better the stability. 14

X-Ray Photo-electron Spectroscopy (XPS)

14.1 Coke on Zeolites. - XPS has been used to study coke on a ferrierite zeolite used in 1-butene skeletal i~omerization~~. The structure of the total coke deposits after extraction with toluene and pentane, was studied with XPS, obtaining the C/Si signal intensity ratios. The analysis of these results is based on the theory developed by Sexton et allo1.This theory gives a description of the attenuation of the Si signal due to carbon deposition inside the pores (channel filling, weak attenuation) followed by carbon built up at the external surface of the zeolites (surface coking, strong attenuation). They found that for all coke contents the XPS C/Si intensity ratio is above the line expected for channel filling only. Above a carbon content of 6 wt%, surface coking predominates. Nevertheless, the data presented in this paper5’ is rather highly scattered. Sexton et allo1studied coke formation on ZSM-5 during methanol conversion. They found that the C/Si ratio increases linearly at a low rate with internal coke formation, and exponentially when external coke is formed. According to their study, coke is formed initially inside the channels, without any significant amount of surface coke being formed. Up to 8 wt% coke there is no evidence for external coke formation. 14.2 Residue Hydrotreating. - A Ni-Mo/y-A1203 coked catalyst was studied with XPS, to determine the location of the coke deposits71.As coke accumulates on the catalyst surface up to lo%, the Mo concentration only slightly changes, and the Ni/Mo atomic ratio was constant at 0.3 in all the samples. On the other hand, the surface A1 concentration decreased sharply as the coke content increased. These data suggest that the coke was deposited preferentially on the alumina support rather than on the catalytically active components. 14.3 Isobutane Dehydrogenation. - The coke formed on a heavily sulfided nickel catalyst, used in the dehydrogenation of isobutane, was characterized by XPS28.The XPS spectra showed two different carbon states on the catalysts with low amounts of coke. One state can be ascribed to carbidic carbon (282 eV) and partially hydrogenated carbon species, CH, (285.3eV). Essentially one dominant feature was observed when the amount of carbon deposited became large. This was ascribed to graphitic carbon (283 eV). These results are in agreement with those found with TPO and DSC as above described. 15

Ultra Violet-Visible Spectroscopy (UV-VIS)

The UV-VIS spectroscopy can be used to determine the chemical identity of the

4: Coke characterization

201

coke componentes. UV-VIS, unlike NMR and IR, can easily detect alkyl and alkenyl carbenium ions, essentially due to its much higher sensitivity5'. This is usually carried out under vaccum, and therefore, the more volatile compounds could be lost under these conditions. During the ethylene oligomerization reaction in H-Mordenite, it was found that polyenylic compounds were formed"*. 15.1 Isobutane Alkylation. - Flego et a196sudied the alkylation of isobutane with 1-butene, using different catalysts. The nature of organic compounds was characterized with UV-VIS. Absorption bands with maxima at 300,370, and 460 nm were found. These bands are characteristic for alkenyl type carbenium ions at increasing unsaturation degree. Bands appearing at around 270 nm are attributed to the presence of cyclized olefinic intermediates. The polyaromatic compounds give rise to bands in the 470-700 nm range. All absorption bands were present in the spectra of the Lap zeolite, LaY zeolite, and amorphous silica alumina. In the case of the sulfated zirconia catalyst, no separated band could be distinguished. This study did not report at what temperature the reaction was carried out. Therefore, results cannot be directly compared with other studies in similar systems, in which only the aliphatic type of coke was found, as above discussed (see section 5.2). Flego et a155reported a second and more complete study on the characterization of LaHY zeolite, used in the isobutane alkylation reaction. Figure 10 shows the UV-VIS spectra obtained by Flego et a155,at different reaction temperatures and LHSV. In this case, bands at 315 nm due to monoenyl carbenium ions, were found to be more intense than the band at 370 nm, assigned to dienyl carbenium ions, in the case of a reaction temperature of 80 "C. At a higher reaction temperature, 100 "C, the intensity of these two bands is the same. From this, it is concluded that at higher reaction temperatures more alkenyl carbenium ions with greater unsaturation degree are generated. In this study, no aromatics were detected. Since more unsaturated carbenium ions were expected at longer contact time, the bands at 315 and 382 nm were compared for catalysts coke at different LHSV (Figure 10 11).As LHSV increases, the reactants have less time to take part in secondary transformations inside the cavities, and therefore the monoenyl carbenium ions dominate. 15.2 n-Butane Isomerization. - Sulfated-zirconia catalysts displayed bands at 245, 306, 366, and 400 nm after the n-butane isomerization reaction at 2OO0C6'. The band at 306 nm was attributed to the presence of allylic cation, following the work of Chen et allo3.The absorption bands at 366 and 400 nm correspond to polyenylic cations and aromatic compounds, respectively. The regeneration with oxygen was also followed with UV-VIS-DR (diffuse reflectance). It was determined that the different coke species transformed into polyciclic aromatic and condensed rings, as indicated by a very broad absorption band at 415 nm. Additional studies were conducted to elucidate the chemistry occurring during the oxidative regeneration process6'.

Catalysis

202

300 Figure 10

16

500 700 Wavelenght, cm-l

UV-VIS spectra of L a H Y zeolite used in a semi-batch reactor, isobutane alkylation with C4 olejins, ( I ) Influence of reaction temperature at LHSV=4.07 h-' , ( A ) 80 "C,( B ) 100°C; ( I I ) influence of L H S V a t 80 "C, ( A ) 2.27 h-', ( B ) 3.86 h-', ( C ) 4.07 h-' (from Ref. 55)

Electron Paramagnetic Resonance (EPR)

The EPR (or ESR) technique allows the study of the radicals that accompany the coke formation, and thus estimate roughly the amount of coke and obtain information regarding its nature. One of the advantages of this technique, is that it can be used both under static or under on-stream conditions. This technique has been used to study the carbonization of polyethylene over acidic zeoliteslW. Coked sulfated zirconia catalysts were analysed by EPR43.A strong band appeared in the coked catalyst at g = 2.0035 - 2.0044 with antisymmetric peaks at g = 2.0064 and g = 2.0014. This band was previously assigned to polyenylic and polyaromatic cations61.The signal is huge with no hyperfine splittings in the case of the completely deactivated catalyst. The total absence of hyperfine splittings was related to a high degree of carbonization of the radicals. Coke formation on H-Mordenite during olefins reaction was studied both under and under on-stream conditionslo6.Under static conditions, it was found that the carbonization of ethylene and propylene could be separated into two processes, depending upon the temperature. Below 227"C, the radicals of a low-temperature coke are formed and subsequently annihilated. These radicals do not appear to be precursors of the high-temperature coke. Above 227"C, highly unsaturated radicals of high-temperature coke are formed. The presence

4: Coke characterization

203

of reactive species in the gas phase, which is similar to the situation under on-stream conditions, may affect the carbonization of the low-temperature coke. This result is indicative of the relevance of conducting characterization experiments as close to the reaction conditions as possible. The experiments carried out under on-stream conditions, led to additional information. The low temperature radicals were found to be olefinic or allylic oligomeric species, and their formation was favoured by an increase in the acidity of the zeolite. In the formation of the high temperature coke radicals, which are highly unsaturated species, homolityc bond splitting of carbonaceous deposits may be involved. A good correlation was found between the number of radicals measured by EPR and the amount of coke formed at high temperature as measured by thermogravimetric analysis.

17

Coke Formation Rate

The amount of coke that is being deposited on a catalyst has been traditionally followed with conventional microbalances. However, due to the inherent limitations of this equipment, in which it is almost impossible to avoid feed by-pass effect and diffusional effects, this technique has not been very useful to determine coking kinetics as a function of feed composition. A recycle electrobalance reactor has been designed to avoid this undesirable effectlo7.The unit was designed to operate up to 500°C and at atmospheric pressure. A novel flow-through microbalance has been recently developed and used in catalyst deactivation studies. It is known as TEOM microbalance (tappered element oscillating microbalance). The basic principle of operation and the equations govening the operation are those of a cantilever beam mass-spring system. It has been first used to study coke formation on LaY and then it was applied to several other systems such as SAPO-34 and reforming catal y s t ~ ' ~HZSM-5 ~ ~ ~ ~ during ~, ethene oligomerization"', Pt-Cu/C during hydrodechlorination of 1,2 dichloropropane112,1-butene isomeri~ation~~~''~.

18

Concluding Remarks

The understanding of different issues related to coke, such as formation mechanisms, location, structure, gasification/oxidation kinetics, is needed in order to develop catalysts generating less coke and with higher coke tolerance, and to develop more efficient regeneration processes. Studies in several catalytic systems where deactivation due to coke deposition is a major issue, have been carried out. These studies include for example, bifunctional metallic/acid catalysts used in naphtha reforming, acid catalysts used in FCC, hydrotreating Ni-based catalysts, hydroisomerization Pt/zeolite catalysts, dehydrogenation catalysts, zeolites, etc. Each system has particular difficulties when addressing the coke characterization studies, and therefore different combination of techniques are necessary in order to gain the needed knowledge in this topic. Table 5

204

Table 5

Catalysis

Studies of coke characterization on diflerent catalytic systems

Reaction

Catalysts

Techniques

Refs

Naptha Reforming

Pt/Al,O,

TEM EELS-STEM, TEM CAEM, TGA, TPC02 BET, PV, SD, EM, XRD, SIMS, CTEM, STEM LRS TPO TEM, EELS 13CNMR TPO TPO, TPHy TPO, SD 13CNMR, TPO 13CNMR, BET, PV, PSD,XPS, SE I3CNMR, TGA, SE BET, PV, SE, TPO, IR PSD, XPS, TPO, DSC, TEM Review EPR,TPO, SE, TPHe UV-VIS,FTIR,'3CNMR

44 46 47 45

TG-DSC, UV-VIS, IR TPD, TPO, ET, PV, SE, TPHe TEM, XPS TPO, DRIFTS, TPHe PSD, H/C, TPO, IR,TPD, review FTIR SD, BET,PV,PSD IR, 13CNMR EPR,13CNMR '29Xe,'3CNMR, TGA, A1 TG, TPC02, TPO

96

Pt, Pt-Re, Pt-Ge, Pt-Sn, Pt-Ir Cracking

Pt-Re Y-zeolite

Hydrotreating Dehydrogenation

Pt - Sn/A1203

Steam Reforming n-C4 isomerization iC4 Alkylation

Ni Sulfated-Zirconia Y -zeolite Y, Beta, Sulfated-Zirconia Y, MOR,

1 -butene isomerization

Ferrierite

butene dehydrogenation

Cr203-A1203 Zeolites

ethene Polyet h ylene nProp y lBenzene dehydroaromatization Propene Oligomerization

HPA

60 22 52 84 29-31 32 65 38 71 73 76,121 28 39,48 43 55

36,lO 51 33,34 8 50 59 66-68 83 104 85 37

13CNMR,XPS,TPO,TGA, 78 SE

PV: pore volume, SD: Support Dissolution, EM: Electron microprobe, CAEM: controlled atmosphere electron microscopy, PSD: pore size distribution, SE: solvent extraction, AI: adsorption isotherms

presents examples of previous studies where coke characterization on catalysts was particularly addressed. The table includes the information regarding the type of reaction, catalyst, characterization techniques used in the study, and the references. In the Table, several examples of the multitechnique approach are included. All of them are discussed above, analysing separately the contribution of each technique to the understanding of coke charaterization.

205

4: Coke characterization

Table 6

Information obtained with each technique used in coke characterization

__________~

Technique

Application

References

IR, FT-IR, DRIFTS

- Functional groups of coke (nature)

8,33,55,56,58, 59,65,83,95,96 46,50 44,493 1 8,22,33,34,36,65 28,39,40,41 33,36 38,73,95,96

Acid site deactivation STEM- EELS Coke location Electron Microscopy Location, structure T P O TPHy TPHe -morphology, kinetic parameters, location TGA -coke reactivity -volatile compounds, oxygen content, lattice oxygen reactivity Coke Extraction -Composition of Soluble fraction -Coke location LRS UV-RS Structure (pregraphitic or highly organized) Average dimension of crystallite (monophasic carbon) Dissolution of Coke composition and structure Support XRD Crystalline structure coke location AES Amount of coke (approximately) Fraction of amorphous/graphitic Vs carbidic/molecular. SIMS Hydrogen content of the carbon deposits (qualitative) Sorption capacity - -coke location (pore mouth or site Pore volume analysis blocking) (uncertainty in coke density) -micro or mesopore blocking I3CCP/MAS - NMR -type of chemical groups in coke (nature) 'H NMR

129XeNMR 29SiMAS NMR XPS uv-VIS Neutron scattering and Attenuation TEOM (microbalance) EPR

36,43,65,66,70,72,73 60,6162,64 45,65,66,67,68,69,70 45,124 94,95 92,93 45,92 8,36,45,95-99,65 71,85 73,55,72,83,38, 85,71,74,87 65,69,70 71 85,9 1 74 5 1,71,101 102,96,61,103,55

-H/C distribution of hydrogen in coke (aliphatic, aromatic, etc) -Coke distribution -Coke distribution coke location (in zeolites) Functional groups of coke detection of carbenium ions coke content (requires the C/H be known) 79,80 C/H (requires the % C be known) 81 Coke distribution (requires additional experimental information) 108-113,50 weight gain Coke Nature, radicals during coke formation

104- 106,43,61

Table 6 sumarizes the type of information that each technique can provide to characterize the coke. The coke characterization area of heterogeneus catalysis is still receiving considerable attention in many catalytic systems, due to the relevant contribution it can provide to the overall improvement of the process.

206

Catalysis

Acknowledgement Thanks are given to Prof. Elsa Grimaldi for the revision of the English manuscript.

References 1. 2. 3. 4. 5. 6. 7.

8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35.

A. Bell, in ‘Catalyst Deactivation’, ed. E.E. Petersen and A.T. Bell, Marcel Dekker, 1987,p.235 J. Barbier, E. Churin, P. Marecot, J . Catal., 1990,126,228. S.C. Fung, C.A. Querini, J . Catal., 1992,138,240. S.M. Augustine, G.N. Alameddin, W.M.H. Sachtler, J . Catal., 1989,115,217. J.M. Parera, R.J. Verderone, C.L. Pieck, E.M. Traffano, Appl. Catal., 1986,23,15. J. Biswass. P.G. Gray, D.D. Do, Appl. Catal., 1987,32,249. J. Barbier, E. Churin, J.M. Parera, J. Riviere, React. Kinet. Catal. Lett., 1985, 29, 323. W.Q. Xu, Y.G. Yin, S.L. Suib, C.L. O‘Young, Journal ofPhysical Chemistry, 1995, 99,758. R.A. Comelli, S.A. Canavese, C.A. Querini, N.S. Figoli, Appl. Catal., 1999,182,275. C.A. Querini, Catal. Today, 2001,62, 135. C.L. Yang, US Patent 3,893,942 (1975) T.C. Basso, Z. Zhang, W.M.H. Sachtler, Appl. Catal. A: General, 1991,79,227. C.L. Pieck, R.J. Verderone, E.L. Jablonsky, J.M. Parera, Appl. Catal., 1989,55,1. C.A. Querini, S.C. Fung, Appl. Catal. A: General, 1994,117, 53. M. Larsson, J. Jansson, S. Asplund, J . Catal., 1998,178,49. R. Hughes, U.R. Shettigar, J . Appl. Chem. Biotechnol., 1971,21,35. C.A. Querini, S.C. Fung, Catalysis Today, 1997,37,277. C.A. Querini, in preparation J. Barbier, E. Churin, P. Marecot, J . Catal., 1990,126,228. J.M. Parera, R.J. Verderone, C.L. Pieck, E.M. Traffano, Appl. Catal., 1986,23,15. J.M. Parera, C.A. Querini, N.S. Figoli, Appl. Catal., 1988,44, L1. C.A. Querini, S.C. Fung, J . Catal., 1993,141,389. C.A. Querini, S. C. Fung, J . Catal., 1996,161,263. J. Barbier, St. Surf Sc. Catal., 1987, 34, 1. C.A. Querini, N.S. Figoli, J.M. Parera, Appl. Catal., 1989,52,249. S.M. Stagg, C.A. Querini, W.E. Alvarez, D.E. Resasco, J . Catal., 1997,168,75. P. Ferraris, C.A. Querini, Proc. XVI Iberoam. Symp. Catal., Vol 11, (1998),893 D.R. Resasco, B. K. Marcus, C.S. Huang, V.A. Durante, J . Catal., 1994,146,40. C.L. Minh, R. A. Jones, I.E. Craven, T.C. Brown, Energy and Fuels, 1997,11,463. C. Li, C.L. Minh, T.C. Brown, J . Catal., 1998,178,275. J.M. Kanervo, A.O.I. Krause, J.R. Aittanmaa, P.H. Hagelberg, K.J.T. Lipiainen, I.H. Eilos, J.S. Hiltunen, V.M. Niemi, Chem. Eng. Sci., 2001,56, 1221. 0. Bayraktar, E.L. Kugler, Appl. Catal. A: General, 2002,233,197. Z.R. Finelli, C.A. Querini, N.S. Figoli, R. A. Comelli, Applied Catalysis A: General, 1999,187,115. Z.R. Finelli, C.A. Querini, N.S. Figoli, R. A. Comelli, Applied Catalysis A: General, 2001,216,91. R.A. Comelli, Z.R. Finelli, N.S. Figoli, C.A. Querini, St. Surf: Sc. Catal., 1997, 111, 139.

4: Coke characterization

36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73.

207

C.A. Querini, E. Roa, Appl. Catal. A: General, 1997,163,199. H.Liu, T.Li, B. Tian, Y. Xu,Appl. Catal. A: General, 2001,213,103 M.A. Callejas, M.T.Martinez, T. Blasco, E. Sastre, Appl. Catalysis A: General, 2001, 218, 181. C.H. Bartholomew, Catal. Rev.-Sci. Eng., 1982,24 (l), 67. C.H. Bartholomew, Appl. Catal. A: General, 2001,212,17. I. Chen, F.L.Chen, Ind. Eng. Chem. Res., 1990,29,534. G. Panattoni, C.A. Querini, St. Surf Sc. Catal., 2001,139, 181. C.R. Vera, C. Pieck, K. Shimizu, C.A. Querini, J.M. Parera, J . Catal., 1999,187,39. R.A. Cabrol and A. Oberlin, J . Catal., 1984,89,256. D. Espinat, E. Freund, H. Dexpert, and G. Martino, J . Catal., 1990,126,496. P. Gallezot, C. Leclerq, J. Barbier, and P. Marecot, J . Catal., 1989,116, 164. T.S. Chang, N.M. Rodriguez, and R.T.K. Baker, J . Catal., 1990, 123,486. J.R. Rostrup-Nielsen, J. Sehested, St. Surf.Sc. Catal., 2001, 139, 1. E. Tracz, R. Scholz, T. Borowiecki, Appl. Catal., 1990,66,133. S. van Donk, J.H. Bitter, K.P. de Jong, Appl. Catal. A: General, 2001,212,97. K.P. de Jong, H.H. Mooiweer, J.G. Buglass, P.K. Maarsen, Stud. Surf Sci. Catal., 1997,111,127. P. Gallezot, C. Leclercq, M. Guisnet, P. Magnoux, J . Catal., 1988,114, 100 M. Guisnet, P. Magnoux, Appl. Catal. A: General, 2001,212,83. P.E. Eberly, Jr., C.N. Kimberlin, W. H. Miller, H.V. Drushel, Ind. Eng. Chem. Process Des. Dev., 1966,5, 193. C. Flego, I. Kiricsi, W.O. Parker Jr., and M.G. Clerici, Applied Catalysis A: General, 1995,124,107. G.S. Nivarthy, Y.He, K. Sesham, J. Lercher, J . Catal., 1998,176, 192. A.J. Barnes, J.D.R. Howells, J . Chem. SOC.,Faraday Trans. 2, 1973,69, 532. A. K. Ghosh, R.A. Kydd, J . Catal., 1986,100,185. J.V. Ibarra, C. Royo, A. Monzon, J. Santamaria, Vibrational Spectroscopy, 1995,9, 191. D. Espinat, H. Dexpert, E. Freund, G. Martino, M. Couzi, P. Lespade, F. Cruege, Appl. Catal., 1985,16, 3437. D. Spielbauer, G.A. H. Mekhemer, E. Bosch, H. Knozinger, Catalysis Letters, 1996, 36,59. C . Li, P.C. Stair, Stud. in Surf Sci. and Catal., 1996,40, 881. C. Li, P.C. Stair, Catalysis Letter, 1996,36, 119. Y.T. Chua, P.C. Stair, J . Catal., 2003,213,39. P. Magnoux, M. Guisnet, Appl. Catalysis, 1988,38, 341. P. Magnoux, M. Misk, G. Joly, S. Jullian, M. Guisnet, Stud. Surf Sci. and Catal., 1997,105,1835. M. Guisnet, P. Magnoux, Appl. Catal., 1989,54, 1. P.Magnoux, P. Roger, C. Canaff, V. Fouche, N.S. Gnep, M. Guisnet, Stud. Surf. Sci. Catal., 1987,34, 317. P. Magnoux, M. Guisnet, S. Mignard, P. CArtraud, J . Catal., 1989,117,495. V. Fouche, P. Magnoux, M. Guisnet, Appl. Catal., 1990,58, 189. F. Diez, B.C. Gates, J.T. Miller, D. J. Sajkowski, S.G. Kukes, Ind. Eng. Chem. Res., 1990,29,1999. S.M. Holmes, A. Garforth, B. Maunders, J. Dwyer, Appl. Catal. A: General, 1997, 151,355. S.K. Sahoo, P.V.C. Rao, D. Rajeshwer, K.R. Krishnamurthy, I.D. Singh, Appl. Catal. A : General, 2003,244,31 1

208 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 11 1. 112.

Catalysis W.A. Groten, B.W. Wojciechowski, B.K. Hunter, J. Catal., 1990,125,311. G. Alexander, I. Hasai, J. Chromatogr., 1981,217, 19. J.C. Alfonso, M. Schmal, R. Frety, Fuel Proc. Technol., 1994,41, 13. E.E. Wolf, F. Alfani, Catal. Reu. Sci. Eng., 1982,24, 329. I.V. Kozhevnikov, S. Holmes, M.R.H. Siddiqui, Appl. Catalysis A: General, 2001, 214,47. D.R. Acharya, M.R. Ghassemi, R.Hughes, Appl. Catal., 1990,58,53. C.L. Koon, D.R. Acharya, R. Hughes, J. Catal., 1990,126,306. D. R. Acharya, A.J. Allen, R. Hughes, Ind. Eng. Chem. Res., 1990,29,1119. W.A. Groten, B.W. Wojciechowski, B.K. Hunter, J. Catal., 1990,125,311. J.P. Lange, A. Gutsze, J. Allgeier, H.G. Karge, Appl. Catal., 1988,45,345. W.A. Groten, B.W. Wojciechowski, B.K. Hunter, J Catal., 1992,138, 343. S-B. Liu, S. Prasad, J.-F. Wu, L.-J. Ma, T.-C. Yang, J.-T. Chiou, J.-Y. Chang, T.-C. Tsai, J. Catal. 1993,142,664. E.G. Derouane, J.P. Gilson, J.B. Nagy, Zeolites, 1982,2,42. J. Weitkamp, S. Maixner, Zeolites, 1987,7, 6. S.K. Sahoo, N. Viswanadham, N. Ray, J.K. Gupta, I.D. Singh, Appl. Catal. A: General, 2001,205, 1. V. Bosacek, L. Kubelkova, J.Novakova, St. Surf: Sc. Catal., 1990,65,337. M.C. Barrage, J.L. Bonardt, J. Fraissard, Catal. Lett., 1991,5,436. T.Ito, J.L. Bonardet, J: Fraissard, J.B. Nagy, C. Andrk, Z. Gabelica, E.G. Derouane, Appl. Catal., 1988,43, L5. J.W. Niemantsverdriet, A.D. van Langeveld, Fuel, 1986,65,1396. D.W. Goodman, R.D. Kelley, T.E. Madey, J.M. White, J. Catal., 1980,64,479. D.M. Dibby,N.B. Millestone, J.E. Patterson, L.P.Aldridge, J. Catal., 1986,97,493. A. de Lucas, P. Canizares, A. Duran, A. Carrero, Applied Catalysis A: General, 1997, 156,299. C. Flego, L. Galasso, I. Kiricsi, M.G. Clerici, St. Surf: Sc. Catal., 1994,88,585. D.M. Bibby, C.G. Pope, J. Catal., 1989,116,407. B. Dimon, P. Cartraud, P. Magnoux, M. Guisnet, Appl. Catal. A: General, 1993, 101,351. S. Mignard, P. Cartraud, P. Magnoux, M. Guisnet, J. Catal., 1989,117,503. Y. Boucheffa, C. Thomazeau, P. Cartraud, P. Magnoux, M. Guisnet, Ind. Eng. Chem. Res., 1997,36,3198. B.A. Sexton, A.E.Hughes, D.M. Bibby, J. Catal., 1988,109,126. M. Laniecki, H.G. Karge, Proc. VIth Int. Symp. Heterogeneous Catalysis, Sofia, 1987, Part 2, p.129. F.R. Chen, G. Coudurier, J.-F. Joly, J.C. Vedrine, J. Catal., 1993,143,616. A. Gutsze, J.-P. Lange, H.G. Karge, J. Allgeier, J. Catal., 1988,113,525. J.-P. Lange, A. Gutsze, and H.G. Karge, J. Catal., 1988,114,136. H.G. Karge, J.-P Lange, A. Gutsze, and M. Laniecki, J . Catal., 1988,114, 144 H.C. Beinaert, R. Vermeulen, G.F. Froment, St. Surf: Sc. Catal., 1994,88,97 F. Hershkowitz, P.D. Madiara, Ind. Eng. Chem. Res., 1993,32,2969. S. C. Fung, C.A. Querini, K.Liu, D.S. Rumschitzki, T.C. Ho, St. Surf: Sc. Catal., 1994,88,305. K.Liu, S.C. Fung, T.C. Ho, D.S. Rumschitzki, Ind. Eng. Chem. Res., 1997,36,3264. D. Chen, A. Gronvold, H.P. Rebo, K. Moljord, A. Holmen, Appl. Catal. A: General, 1996,137, L1 W. Zhu, J. Zhang, F. Kapteijn, M. Makkee, J.A. Moulijn, St. Surf: Sc. Catal., 2001, 139,21.

4: Coke characterization

209

113. L.M. Petkovic, G. Larsen, Ind. Eng. Chem. Res., 1999,38, 1822. 114. J.L. Bellamy, The Infrared Spectra of Complex Molecules, 3rded., Chapman & Hall, London, 1975 115. N. Sheppard, C. de la Cruz, Advances in Catalysis, 1966,1,1. 116. S. Meijers, L.H. Gielgens, V. Ponec, J . Catal., 1995,156, 147. 117. C. Li, P.C. Stair, Catal. Today, 1997,33, 353. 118. W.Q. Xu, Y.G. Yin, S.L. Suib, J.C. Edwards,C.L.O'Young, J . Catal., 1996,163,232. 119. P. Andy, N.S. Gnep, M. Guisnet, E. Benazzi, C. Travers, J . Catal., 1998,173, 322. 120. M. Trombetta, G. Busca, S. Rossini, V. Piccoli, U. Cornaro, J . Catal., 1997, 168, 349. 121. J.C. Alfonso, D. A.G. Aranda, M. Schmal, R. Frety, FuelProc. Technol., 1995,42,3. 122. B. Li, R. Gonzalez, Catal. Lett. 1998,54, 5 123. F. Bauer, E. Geidel, W. Geyer. Ch. Peuker, Micro. Meso. Muter. 1999,29, 109. 124. N.S. Figoli, J.N. Beltramini, C.A. Querini, J.M. Parera, Appl. Catal., 1986,26,39 125. J. Ruixia, X. Zaiku, Z. Chengfang, C. Qingling, Appl. Catal. A: General, in press. 126. C.J. Wright, in Characterization of Catalysts (J.M. Thomas and R. M. Lambert, eds.), Wiley, New York, 1980, p. 169

5 Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds BY MOHAMMAD RAHMANI, KHASHAYAR BADII, MOSTAFA FAGHIHI, MEHRI SANATI, NEIL CRUISE, OW AUGUSTSSON, AND JAMES J. SPIVEY

1

Introduction

Understanding the deactivation processes that take place in oxidation catalysts used for volatile organic compound (VOC) abatement has both industrial and academic interest. The industrial importance of improving the deactivation resistance of catalysts used to remove VOC emissions is directly related to the economics of this process. The market for such equipment will grow significantly in the next few years. For example, in Europe the 'Solvent Emissions Directive' adopted by the EU's Environmental Ministers in 1999 seeks to reduce VOC emissions from operations using solvents by 67 % by 2007, based 1990 levels. The EU member states have now adopted these directives into national law. This directive will affect over 400,000 solvent users across more than 30 manufacturing sectors. Current estimates, from the European solvent industry, suggest that this directive will reduce emissions by 1.5 million tonnes a year and cost industry 80 billion € to implement. A portion of this cost will be used to introduce end-of-pipe solutions. At present, the dominant technology in Europe for abating airborne VOC emissions is thermal oxidation. Compared with catalytic oxidation, the thermal oxidation has several drawbacks. The main disadvantage is that the thermal oxidation operates at 800-900 "C compared with 250-350 "C for the catalytic alternative. The high operating temperature means that the burning chamber has to be designed for higher temperature, adding investment cost to the unit. The high operating temperature also means that more energy must be added to the process as supplemental fuel, compared to a comparable catalytic process. This contributes to a higher operating cost of the unit. Another disadvantage of the high operating temperature is that NO, can be formed in the flame, perhaps requiring additional emission controls. For catalytic oxidation, the cost of the catalyst, and its replacement due to premature deactivation, add to the cost. If deactivation can be minimized, catalyst replacement costs can be minimized. Catalysis, Volume 17 0 The Royal Society of Chemistry, 2004

210

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

21 1

All the above mentioned disadvantages for thermal oxidation compared with the catalytic alternative were understood when catalytic oxidizers were introduced outside the chemical process industry in the 1980s. Unlike the chemical process industry, which had long experience with oxidation catalysts, gases from processes that emit VOC such as printing and drying contain silicaorganic and phosphororgbriic compounds can rapidly deactivate noble metal catalysts. This led to significant problems with catalytic deactivation. In the worst performing units, the catalyst had to be replaced or regenerated within a few months. This led to a reluctance to use catalytic oxidation within the printing and coating industries, among others. At least in the near term, total replacement of the silica and phosphorous compounds in these industries is unlikely. Therefore, an understanding of the deactivation process resulting from these compounds is important. Despite improvements in VOC oxidation catalysts, there are still challenges. The emissions from a process using silica may contain as much as 0.1 mg Si/Nm3. At this level, 10 000 h of operation would lead to about 1.5 %(w/w) loading of Si on the catalyst. This is roughly the amount of silica that could be tolerated by a modern catalyst before its activity becomes too low to meet today’s strict emission standards. For economic reasons, this lifetime must be at least 20 000 h, and perhaps more. Even industries within the printing and coating sector that have silica emissions well below 0.1 mg Si/Nm3 hesitate to invest in catalytic units because a catalytic may restrict future operations in which the Si loading may be greater than this. A key challenge is to acquire a better understanding of the mechanisms for silica and phosphorus deactivation. This better understanding can then be used to produce catalysts with minimal deactivation even for the most severe emissions. From a scientific standpoint, there is an interesting relationship between the fouling and poisoning mechanisms that are typical of silica deactivation, and the poisoning mechanism that more accurately describes phosphorous deactivation. It is of interest to carefully review the dependence and interplay of these processes as a function of the reactant, catalyst support, pore structure, metal loading, and oxidation conditions. This review focuses on deactivation of VOC oxidation catalysts by Si and P compounds. Deactivation by both types of compounds are used as examples for the application of mathematical models describing the deactivation process. 2

Effect of Organo-silica Compounds

Organo-silica compounds are widely used in a range of industries such as printing, oil production, and mining as lubricants and polishing compounds. They also find domestic use in silicon rubber and aerosol sprays’’2because of their special physical properties, such as viscosity and dielectric constant. These compounds are also recognized as poisons/inhibitors in different catalytic processes, such as in atmospheric scrubbers, automobile exhausts, flammable gas

212

Catalysis

CH3 H3C

Figure 1

I

-SiI

CH3

I

0- Si-

I

CH3

The chemical structure of hexamethyldisiloxane ( H M D S )

detectors: hydrotreating (HDTp, and naphtha reforming.’ The most widely used compound is known as hexamethyl-disiloxane (HMDS), bis (trimethylsilyl)ether, or bis (trimethylsilyl)oxide, Figure 1.’ It is the monomer for other polysiloxanes that are the base of most silicon o i k 6 Also, methyl- and phenyl-substituted polyoxysilane (MPSP)’ and polydimethylsiloxane (PDMS)4have been used for investigation of catalyst deactivation by organo-silica compounds. Table 1 summarises studies on the deactivation of oxidation catalysts by organo-silica compounds. 2.1 Chemical Properties of HMDS. - HMDS is the most widely studied organo-silica compound for investigation of catalyst de-activation. HMDS contains six methyl groups, two silicon molecules, and one oxygen Its chemical and physical properties are given in several reference^.^^'

2.2 Deactivation Effect of HMDS on Oxidation Catalysts. - Studies show that the HMDS deactivates various catalysts in different ways. Gentry and Jones investigated the effect of added HMDS vapor on the oxidation of methane, CO, propene, and hydrogen6.The authors made a distinction between poisoning and inhibition. A species was considered a poison if, within the time-scale of the experiment, the reaction rate irreversibly decreases, i.e. it does not increase when the species is removed from the feed. If the rate increases when the species is removed, it is an inhibitor. In all tests, the concentration of VOC was 1% and HMDS was co-fed at a concentration of 40 ppm. The temperature was 523K for CO and 873K for other VOCs. Table 2 shows the effect of exposure to HMDS on the methane oxidation activity. Pt/A1203 was considerably more resistant to HMDS than any other catalyst/support combination. Table 3 summarizes the effect of HMDS vapor exposure at 873K to noble metal and supported catalysts on the rate of CO oxidation at 523 K. Pt wire and platinum and palladium on alumina were the most active catalysts, and silica supported catalysts were notably inactive. The time dependencies of the rates of three oxidation reactions at 873 K in the presence of HMDS vapor on Pt/alumina catalyst bead are illustrated in Figure 2. It can be seen that the rates of methane and propene oxidation are rapidly reduced in the presence of HMDS, but that there is negligible effect on hydrogen oxidation. Figure 3 shows the temperature dependence of the same reactions over a Pt/A1,03 catalyst in the absence of HMDS vapour both prior to and subsequent

Self supported polycrystalline platinum films, single crystals Pt( 110) and Pt( 111)

Pt wire, Pt supported on Oxidation of methane, alumina, unsupported Pt, propene, CO, hydrogen Pd-supported on alumina, silica and 13X zeolite

Pt/A120,

1996

1978

1999

Catalysts were contaminated using methyl- and phenyl-substituted polyoxysilane (MPSP), during reforming of cyclohexane

Adsorption and decomposition of HMDS on Pt, XPS, UPS and TDS study

Methane oxidation, HDMS

Platinum ribbons

1997

Compound(s)

Catalyst

Remarks

Refs.

Silicon changes the chemisorption properties of Pt, coke has a more important effect than silicon in reducing the activity of the metallic function, selective poisoning is proposed

5

1 The poisoning effects due to the dissociative adsorption of HMDS on catalytic activity of Pt ribbons in methane oxidation depended on the temp. of the interaction The influence of HMDS at 9 At low temp, a multilayer is high temp on all reaction formed which desorbs at 150 varied from poisoning K, with temp. increase several (methane oxidation), product appears in the gas inhibition (propene oxidation), phase, formation of negligible effect on hydrogen amorphous carbon and oxidation graphitization at above 700 K Silicon poisons both metalic Three types of behaviour can 6 and acid sites of Pt/Al,O, be distinguished; type-A, irreversible adsorption results in poisoning, type-B reversible adsorption results in inhibition, and type-C does not affect the reaction rate

~

c950K, overlayer of composition close to SiO,, at above 950 K the coating with sio2k0.2

~~

The decay eflect

Poisoning temp. 250 "C, The microscopy analyses reaction temp. 100suggest that organosilicon 500"C, 1 atm compounds containing in printing ink diffuse into the catalyst and deposit as silica particles in the micropores

100-600 "C

150-700 K

700-1 100 K

Reaction conditions

Deactivation studies on oxidation catalysts by organosilicon compounds

Year

Table 1

3

3

E

0

'ct

B

'cc

R

3

a

Fc!

Q-

k

Catalyst

Field-aged bead catalysts in a flexographic printing application

Ni- Mo/Al,O,, 2.7 wt% Ni and 9.9 wt% Mo

Supported palladium and Pt

1998

1993

1984

(cont.)

Year

Table 1

Methane- and butane oxidation, hexamethyldisiloxane

Industrial plant for hydrotreating of naphtha

Hexane oxidation

Compound(s)

Refs.

625-730 K

The combined use of 4 29SiMAS, 29SiCP/MAS, and 13C CP/MAS, NMR have led to a detailed and unambiguous characterization of the silicon species on aged naphtha HDT catalysts The organosiloxane appears 3 The effect of HMDS deactivation on oxidation of to be irreversibly adsorbed on methane and butane at the high-energy sites needed 625-730 K showed reversible to catalyze methane deactivation for butane oxidation but is reversibly oxidation, but much slower adsorbed on the lower energy recovery of methane activity sites which oxidize butane At 650 K, HMDS cause total and almost irreversible poisoning of catalytic beads for methane oxidation but has a much smaller effect on the activity with respect to butane oxidation

320-360 "C, 50 bar

The silica penetrates the 2 catalyst bead via a shell progressive mechanism and deposits in the micropores of the catalyst, deposition is non-selective and silica masks the noble metal active sites

Remarks

Silicone oil from the naphtha feed is converted to modified silica gel consists mainly of bulk SiO, carrying the surface species SiOH, Si(OH),, -SiCH,, = SiOHCH,, and = Si(CH,),

The decay eflect

350-800 "C

Reaction conditions

P

!2

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

Table 2

Average half-lives of methane oxidation catalysts in deactivation at 873 K by HMDS (from ref. 6)

Catalyst bead

Average half-lives, tllz (s)

Pt wire Pt/A1203 Pd/A1203 Pt/Si02 Pd/Si02 Pt/l3X Pd/l3X

Table 3

215

<5

108f 40 <10 -

<10 <10
Deactivation of carbon monoxide oxidation catalyst at 523K by HMDS (from ref. 6 )

Catalyst bead

Degree of poisoning (%)

Degree of inhibition (%)

Pt wire Pt/A120, Pd/A1203 Pt/Si02 Pd/SiO, Pt/l3X Pd/l3X

70+ 32f 77+ loo+ 38+ 73+ 23f

28f 17 14f 3 4+ 2

18 3 7 0 38 15 22

-

35f 10 6+ 8 38f 16

to HMDS exposure. It can be seen that the poisoning is permanent in the case of methane oxidation, but that HMDS has little effect at high temperature for hydrogen and propene oxidation. However, exposure to HMDS vapour affects the low temperature oxidation, resulting in a shift of the observed reaction rate maxima. The maximum reaction rate for hydrogen oxidation increased from 483 to 533K after one hour exposure to HMDS, and for propene oxidation from 503 to 613K. The maximum reaction rate for propene oxidation has shifted to approximately 663K after 2 hours of exposure. This study shows that the deactivation caused by HMDS vapour is selective, the deactivation depends on the chemical nature of the reactant and, presumably, reflects a difference in the types of active sites on the catalyst surface. Figure 3 shows total poisoning of methane oxidation, inhibition of propene, and a negligible effect on hydrogen oxidation. This is consistent with the hypothesis that HMDS is adsorbed irreversibly on the most active (type-A) sites, which are essential for the methane oxidation, and are responsible for the low temperature oxidation of propene and hydrogen. The decomposition of HMDS and the formation of a silicon-containing adlayer may cause this poisoning. Reversible deactivation (type-B sites) occurs on the less active sites, inhibiting the high temperature oxidation of propene, but not causing poisoning. It may be further suggested that hydrogen oxidation occurs via an Eley-Rideal mechanism. This mechanism does not require strong chemisorption of hydrogen and probably

Catalysis

216

-*-

-*.

30

45

60

Figure 2

Time dependency of the influence of HMDS on the reaction rate over Pt/A1203 at 873 K . -, H,; A-A, C3H,;V-V, CH4 (from ref. 6 )

Figure 3

Temperature dependency for oxidation over Pt/A1203.The curves illustrate the response of the catalyst to 1% concentrations in air of gases: , I I hydrogen; A-A, propene; V-V, methane. Initial response (open symbols) after exposure to 40 ppm HMDS for 1 hour (halfclosed symbols), and after 3 hours (closed symbols) (fromref. 6 )

Temperature f°Ct

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

217

involves activated oxygen species on sites which are virtually unaffected by HMDS. Cullis and Willatt3compared the effect of HMDS on the oxidation of methane and butane at 625-730 K. HMDS causes drastic and almost irreversible poisoning of catalytic beads for methane oxidation but has a much smaller effect on the butane oxidation a~tivity.~ In Fig. 4, a powdered 2.7 wt % Pd on y-Alz03catalyst was first exposed to butane/oxygen/nitrogen mixture for 1 hr, and then to a mixture of methane/oxygen/nitrogen for a similar period of time. Exposure to reactant mixtures containing HMDS caused a decreased in catalytic activity. The magnitude of the decrease in catalytic activity was dependent on the HMDS concentration, and deactivation was much larger for methane than for butane (Fig.4). For the purpose of recovery, the deactivated catalyst was purged with the mixture of hydrocarbon. It was found that the catalysts were almost irreversibly poisoned for methane oxidation, but much of the activity was recovered for butane oxidation, (Fig. 4). One explanation for the differences in the effect of HMDS might be that the oxidation of methane requires high energy sites for oxygen activation. But it seems that the high energy sites have been selectively poisoned by the HMDS in the initial part of the experiment. Under this scenario, the organo-silica compounds are adsorbed irreversibly on the high-energy sites that are needed to catalyse methane oxidation. The small but measurable increase in methane activity in the recovery cycle of Fig. 4 suggests that the poisoning is not entirely irreversible, and that there is a continuum of site activities, with a corresponding range in HMDS adsorption strengths. HMDS might adsorb also reversibly on

80

3

-t

. A

INITIAL ACTIVITY

-

M A C T I VAT ION

r _

RECOVERY

I

I

I

60

v

g, 0

Ly

t

20

0

20

40

60

80

0 Tima

Figure 4

20

40

60

80

0

20

40

(An)

The deactivation of a Pd/(A1203 + T h o 2 ) catalytic bead by HMDS and the subsequent recovery of catalytic activity, Temperature 800K. 0-0, Oxidation of methane (composition of reactant mixture: 2.5 mol% CH4, 20.5 mol% 02, together with 2.5 x 10-4 mol% HMDS during deactivation experiments, balance N 2 ) . 0- , Oxidation of butane (composition of reactant mixture: 0.75 mol% C4Hlo,20.9 mol% 02,together with 2.5 x l O'4 mol% HMDS during deactivation experiments, balance N 2 ) (from ref. 3)

218

Table 4

Catalysis

The efect of H M D S on the oxidation of methane and butane over catalyst Pd/(A1203 + Th02) beads at 800K (from ref. 3)

Concentration of H M D S (mslm')

Methane oxidation recovery

Butane oxidation recovery

(%)

(%)

23 113 282

0 0 0

100 50 0

the lower energy sites that are sufficient to oxidize the more reactive butane molecule. 2.3 Effect of HMDS Concentration. - Concentration of HMDS has an important effect on deactivation of the catalyst. At higher concentrations, the deactivation mechanism can change from inhibition to poisoning. Cullis and Willatt showed this behavior in the oxidation of b ~ t a n eIncreasing .~ the concentration of HMDS caused the complete deactivation of catalyst. Table 4 compares the effect of exposure to HMDS on the oxidation of methane and butane. This is consistent with the hypothesis that the high-energy sites are selectively poisoned first, followed by poisoning of the less active sites. However, at the highest HMDS concentration, the poisoning is irreversible. The cumulative exposure of HMDS results in the same activity loss for methane and butane ~xidation.~ Ehrhardt et. al." exposed a Pt wire to 0.5 pmol/cc HMDS in air then carried out catalytic oxidation of a 1% methane/air stream at various temperatures between 775 and 1025 K.' As expected, exposure to HMDS reduced the oxidation activity; e.g., the temperature required for the initial detectable oxidation increased from 810 K to 931 K. Their results are consistent with selective poisoning of the sites most active for methane oxidation before other sites are poisoned. It is also interesting that after exposing the Pt wire to varying concentrations of HMDS/air, the methane oxidation activity was the same for all concentrations when normalized to the total amount of HMDS to which the catalyst was exposed. Above 950 K, the deactivation was irreversible after a critical total amount of HMDS accumulated on the surface (8 x mol in this case). XPS analysis of the catalysts exposed to HMDS showed that the primary surface poison is SO2,with decreasing amounts of carbon on the catalyst surface at higher exposure temperatures. Exposing the catalyst to HMDS at different temperatures (using the same HMDS concentration) resulted in different catalytic activity, but essentially the same surface composition (primarily SOz). This suggests that HMDS decomposes to silica at these conditions, but selectively poisons the sites most active for methane oxidation first.

2.4 Effect of Catalysts and Supports. - Measurable support effects on deactivation by HMDS for otherwise comparable catalysts have been reported. For

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

219

instance, HMDS deactivates Pt and Pd for methane oxidation irreversibly in most cases but deactivation of powder Pd/A1203and Pd/Sn02catalysts has been shown to be rever~ible.’~~~~ Further conclusions can be drawn from the data in Table 3, which also suggest significant influence of the support in hydrocarbon oxidation. It is reasonable to assign the ‘degree of poisoning’ in Table 3 to Type A (high energy-irreversibly deactivated) sites, the ‘degree of inhibition’ to Type B (moderate energy-reversibly deactivated) sites, and the remaining sites as Types C sites (unaffected by HMDS). Significant differences are shown for CO oxidation on platinum wire, Pt/SiO2, Pt/A1203,and Pt/l3X. Type C sites seem to result from an interaction between platinum and the alumina support, since it is the only catalyst in Table 3 that has these types of sites. The Pt/Si02 catalyst is the only one that shows only Type-A sites, suggesting that on this support only high-energy sites are formed. The results for Pt/l3X suggest that all three types of sites are present. For the Pd catalysts, the results on Pd/A1203, like Pt/A1203, indicate the presence of all three types of behaviour; about 80% type-A, 5% type-B and 15y0 type-C. Pd/Si02and Pd/l3X catalysts also show all three types of sites, but with more equal distribution.6 Cullis and Willatt carried out Auger electron spectroscopy on catalysts poisoned or inhibited by HMDS (Table 5).3 This treatment was found to destroy the methane oxidation activity completely. The results show that silicon and carbon were only present on the outer surface and that removal of 1.5 nm of the surface layer by argon ion bombardment left little trace of these elements on the Pd/Sn02 catalyst. The results also show that the silica had penetrated below the surface of the bead in the case of Pd/(Th02 + y-A1203) catalyst, but its concentration decreased as successive atomic layers were removed. This is different from the case of Pd/Sn02 bead, in which the silica had only penetrated in the first few atomic layers and argon ion bombardment for 15 seconds removed all the silica as well as ~ a r b o n . ~ Cullis and Willatt3 showed that the natures of both the precious metal and catalyst support affect the depth of silica penetration. At 625-730 K, HMDS decomposition is strongly catalysed by transition metals, and under these condition the initial step is the cleavage of Si-CH3bonds. This is followed by liberation of gaseous methane and polymerisation of silicon to yield polyorgan~siloxanes.~ The extent to which both silicon and carbon, formed by the decomposition of HMDS, penetrate below the surface of the catalyst is markedly dependent on the nature of the both precious metal and the support. Because Pd absorbs silicon into its bulk, it is deactivated more readily than Pt. The Pd/Sn02 beads, which showed almost no Si penetration, were quite resistant to deactivation, but the activity recovery in this case was not as rapid as on the Pd/(A1203 ThO2) beads. This is consistent with a strong chemical interaction between the Pd and Sn02,which prevents significant penetration of the Si and mimimizes deactivation by Si3. Cullis and Willatt3also showed that the increased support porosity and metal dispersion provided the best means of increasing the poison resistance of Pt to high molecular weight poisons such as HMDS.3The bead form of the catalyst, as

-

+

220

Table 5

Catalysis

Auger Electron Spectroscopic Analysis of the surface of powder catalysts and catalytic beads exposed to HMDS in the presence of methane +oxygen mixture (from ref. 3) Elements presenP

Sample

Sn

0

Pd

C/Thb

Si

Powder catalysts 20 wt% Pd on SnO, As received After argon ion etching Si02 As received After argon ion etching

0.95 1.03

(1.0) (1.0)

0.16 0.21

0.07 -

0.44 0.05

-

(1.0) (1.0)

-

-

0.14 0.15

-

-

-

-

-

-

-

0.31 0.57 0.27

0.55 0.49 0.20

0.25 0.22 0.10

0.24 0.22 0.36

0.46 0.61 0.43

(1.0) (1.0) (1.0)

1.25 1.88 1.23

0.60

0.15

-

-

-

-

-

-

Catalytic beads Pd on A1203 + T h o 2 As received After argon ion etching (15s) After argon ion etching (390s) Pd on Sn0, As received After argon ion etching (15s) After argon ion etching (390s)

A1

-

-

-

aPeaks normalised to oxygen (1.0) bCarbon and thoria peaks could not be identified separately

opposed to the powder form, provides some protection from penetration of the HMDS into the fine pores of the catalyst, trapping fragments from the dissociative adsorption of HMDS on the outer layers of the catalyst. However, because the activity of even the unsupported Pt wire could be partially restored: Pt poisoning appears to be at least partially reversible at these conditions.

2.5 Effect of Deactivation Temperature. - The temperature at which the catalyst is exposed to silica poisons can affect the mechanism of deactivation. Ehrhardt et al.' studied the deactivation of platinum wires at different temperatures using HMDS. After exposing the catalyst to HMDS at different temperatures, they passed a mixture of 1% v/v CH4-air over the catalyst, increased the temperature and observed the oxidation activity as a function of temperature. They divided their results into two different categories according to the HMDS exposure temperature, below and above 950 K. When the temperature was kept below 950 K, methane activity was gradually and partially restored after the methane-air mixture passed over the catalyst. These results, coupled with XPS analysis, suggest that HMDS exposure below 950 K results in a uniform covering of the entire Pt surface (consistent with essentially zero activity immediately after exposure). This Si surface is then gradually and partially reorganized into

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

221

larger agglomerates, exposing active Pt sites. There is no loss of Si from the surface during this process. However, at initial HMDS exposure temperatures above 950 K, more refractory Si surface is produced. It does not appear to reorganize after the methane-air mixture is introduced. If the Pt wire containing this refractory layer is heated to 1070 K, methane oxidation activity is high. However, differences in the catalytic behavior at these temperatures (conversiontemperature dependence, oscillations, and transients) suggest a fundamentally different mechanism for the oxidation process. Ehrhardt et al.' believe this is caused by the catalytic activity of the Si02layer itself at these temperatures, as suggested in earlier work by these authors." The relatively high activation energy they calculate (- 130 kcal/mol) suggests a radical mechanism, consistent with the results of Lane and Wolf."

2.6 Effect of Reactor Design. - In practical systems, the reactor design can appreciably affect the deactivation process. For instance, silica compounds will deposit preferentially at the reactor inlet, leading to a gradient in catalytic activity along the reactor. Modeling of this process is essential to predicting the reactor life. It is especially important if there are changes in selectivity associated with the deactivation process, since products of incomplete oxidation may be emitted before the reactor is completely inactive. As an example, Libanati et d 2studied catalyst deactivation by exposure to HMDS in a commercial catalyst reactor. The composition of the reactant mixtures was 88% ethanol and 12% n-propyl acetate. The total inlet concentration was approximately 1500 ppm (as C1 hydrocarbon). The design conditions were an inlet catalyst temperature equal to 616K and a flow rate of 283 Nm3/min. Individual catalyst beads of approximately 3.1 mm diameter were packed in a bed with 17.8 cm in depth. The catalyst was Pt-Pd/y-A1203,impregnated to produce a 50-100 pm active metal 'eggshell' layer on the outside of 3.1 mmbeads. To investigate the deactivation phenomena as a function of position, containers (5.lcm diameter and 17.8 cm long) were packed with catalyst and placed at several locations in the bed. The containers were continuously aged for a period of 4.5-years.2After this long period, the containers were removed and the catalysts separated into seven fractions, each fraction representing approximately a 2.5 cm (1inch) bed depth. The activity tests were carried out on each fraction using 600 ppm of hexane in air at a flow rate of 1.90l/min (approximately, a space velocity of 11400 h-'). The results are shown as light-off curves in Figure 5. The oxidation activity increases with distance from the reactor inlet, as expected. Arrehenius plots for hexane oxidation data are shown in Figure 6. The pseudo-rate constant decreased with increasing silica loading, but the activation energy for samples at various locations in the bed was fairly constant. This suggests that the oxidation mechanism does not change with silica loading, i.e., the deactivation process is non-selectivefouling of the active metal sites. Fig. 7 shows that the pseudo-rate constant decreases with increasing silica loading on the catalyst. The deactivation is more rapid at lower silica loadings, probably due to the fact that the silica deposits on the outer part of the catalyst first,

Catalysis

222

Figure 5

Light-ofcurves for the oxidation of hexane in a laboratory reactor (from ref. 2) 8 7

5

2 6 c

f 4

3 2 1

Figure 6

Arrehenius plots for hexane oxidation data (from ref. 2)

fouling the ‘eggshell’ layer of the active metal initially. The residual activity of highly silica-deactivated beads was approximately 15YOof fresh activity. Silicon profiles for fresh and aged beads, measured by EDAX, showed that silica had penetrated into the 3.1 mm catalyst beads. The thickness and weight loading depended on the depth of catalyst in reactor and varied from 400 pm penetration and 5.9 wt% at the inlet to 67 p m penetration and 0.56 wt% at the exit from the sample canister, respectively. The results of electron microprobe profiles of silica in aged beads at various bed depths are shown in Figure 8. This profile is consistent with the results of Figure 6, suggesting that the outer (active) layer of the catalyst bead deactivates first, causing a more rapid loss of activity versus Si loading at low overall loadings than at higher loadings. Temperature

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

223

3 0

1

2

7

4

5

6

7

e

si Level {%)

Figure 7

Rate constant as a function of silica loading in bead (from ref. 2)

profiles within the catalyst beds, which would provide further insight into the deactivation, are not reported. The silica loading and thickness in beads measured by electron microprobe method are shown in Figure 8 and Table 6. Table 7 summarizes the pore structure of fresh and aged catalysts (top 1 inch). The surface area and micropore volume of the aged catalyst was lower than that of the fresh catalyst, suggesting that silica has been preferentially deposited within the micropores of the catalyst, probably as a gaseous silica compound rather than as a particulate. These gaseous precursor compounds diffuse into the bead catalyst and deposit as silica particles in the micropores. This deposition results in silica penetration further into the particle than the outer layer containing the active metals.

2.7 Mechanism of Deactivation. - The mechanism of deactivation by HMDS can be summarized as follows. The decomposition of HMDS is strongly catalysed by transition metals used for the oxidation of hydrocarbons. Under these conditions, the initial step is the cleavage of Si-CH3bonds. This reaction is followed by liberation of gaseous methyl groups and polymerisation of the residual species to yield polyorganosiloxanes. At temperature greater than 473 K, some of this polysiloxane is oxidized to Si02.It is generally assumed that the deposition of SiOzis largely responsible for the observed deactivation on the precursor metal catalysts3. Analysis of deactivated samples of a Pt/A1203catalyst by scanning electron microscopy and energy dispersive X-ray analysis indicates that silicon is dispersed across the metal surfaces rather than on the alumina support material. In another study comparing Pt on different supports, it was impossible to determine whether deposition occurred on the silica or zeolite supports.6 N

Catalysis

224

,

3rd inch

6th inch

n edge

4 -

Figure 8

7th inch 1,500 microns

Electron microprobe profiles of silica in aged beads at various bed depths (from ref. 2)

Colin et ~ 2 1 .studied ~ the adsorption and decomposition of HMDS in the absence of oxygen on various platinum surfaces at different temperatures at low pressure by XPS, ultraviolet Photoelectron Spectroscopy (UPS), and Thermal Diffuse Scattering (TDS). They also used Auger Electron Spectroscopy (AES) and Low-Energy Electron Diffraction (LEED) as surface characterization tools. They showed that a multilayer of silica was formed at low temperature and desorbs at 150 K, leaving on the surface an undistorted monolayer. With increasing the temperature, several products, including methane, lighter organodisiloxanes and hydrogen, appeared in the gas phase and methyl radicals were left on the surface. These radicals undergo progressive dehydrogenation, leading to the formation of amorphous carbon and graphitization above 700 K. The carbon residue was readily removed by an oxygen treatment at moderate temperature. Almost no silicon or oxygen remained on the surface above 300K. This result cannot explain the poisoning effect of HMDS on platinum sensors used to detect methane in coal mine and other applications. It shows that the mechanism of decomposition of HMDS in the absence of oxygen and at the low

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

Table 6

Silica loading and thickness in beads that were measured b y electron microprobe method (from ref. 2)

Bed section

Si wt% by I C P

Si layer thickness, p n

1st 2.5 cm section 2nd section 3rd section 4th section 5th section 6th section 7th section

5.9 5.2 4.7 2.9 2.38 0.5 0.56

400 31 1 258 22 1 236 106 67

Table 7

225

Pore structure offresh and aged catalyst at top 2.5 cm (from ref. 9)

Vtotal, cm3/g Vmac, cm3/g Vmio cm3/g drnac, A dmio A Density, g/cm3 Surface area, m2/g

Fresh

Aged

1.17 0.45 0.72 5500 120 0.66 245

1.o 0.41 0.59 6100 140 0.77 178

pressures necessary for the spectroscopic studies differs from the conditions typically require for VOC oxidation. The main reactions in these spectroscopic studies are:7 95 K 150-1 70 K 230 K

245 K 300 K

430-920 K

Multilayer formation, Multilayer desorption, Partial decomposition and Si containing products desorption, HMDSads-+(CH3)6-nSi20ads nCHjadS CH3ads CH2ads + Hads (CH3)6-,,Si20g 4-nHad,+(CH3)6 -.H,Si20g (n = lor2) Methane desorption, CH3ads + Hads+ CH4g Further dehydrogenation reactions and hydrogenation desorption, CHZads-) CHads + Hads Hads -k Hads-)HZg CH3ads -k Hads- CH4g Further dehydrogenation reactions, amorphous carbon and graphitization, CHadsjCads + Hads Hads -k Hadsj H2g

+

226

Catalysis

Studies in the presence of oxygen and HMDS in a plasma reactor also show progressive decomposition of HMDS on alumina foil.12Despite the substantial differences between plasma reaction conditions and those in VOC oxidation, similar results were observed. This study showed that an increase on the O2 concentration caused progressively more oxidation of the CH3 groups. The nature of the silica that was found on the alumina foil changed from an organosilicon polymer to an inorganic quartz-like refractory compound at higher oxygen concentrations. A similar transition was found by increasing the temperature, the balance of organic/inorganic Si content in the films becomes more inorganic at higher temperatures." In addition to HMDS deactivation, related studies have examined deactivation by other organosilicon compounds on supported metal catalysts. Kellberg et a1." investigated the deactivation of a commercial Ni-Mo/A1203 catalyst by polydimethylsiloxane (PDMS), which is added to the feed in coker processes operating at 593-633 K. The fresh catalyst contained 2.7 wt% Ni and 9.9 wt% Mo and had a surface area of 180 m2/g. Although the operating conditions in a coker are not the same as for VOC oxidation, it is interesting that they found that PDMS is oxidized to form silica gels on the catalyst surface. This gels consist of partially methylated silica compounds that deactivate the catalyst. The results of the quantitative analysis of the elements C, Si, S, Mo, Ni, Fe, and V by combustion and XRF analysis are given in Table 8 for six samples that were on stream for 10 months [the sample number is directly related to the proximity of the sample to the reactor inlet (1 = closest to inlet, etc)]! Comparison of these results to corresponding results in the absence of added PDMS (not shown here) suggests that silica alone is responsible for the rapid and irreversible deactivation? The combined use of 29Siand 13CMAS and CP/MAS NMR (Figure 9 and 10) led to a detailed and unambiguous characterization of the Si deposits on aged catalysts. Table 9 shows the results of 29SiMAS NMR spectra of samples 1,4, and 6 as 29Sichemical shifts. In this Table, Q2, Q3, and Q4 are for geminal silanol groups, silanol groups, and siloxane groups (bulk SO2),respectively. T2,T3, T4, and D4 are presented in Figure 11. The PDMS is converted to modified silica gel on the catalyst (Figure 11). The modified silica gel consisted mainly of bulk Si02 carrying the surface species SiOH, =SiCH3, = SiOHCH3, and = Si(CH3)2.4 The differences in relative signal intensities for the two C P spectra reflect variations C P dynamics for the various Si species (from ref. 4) These results are consistent with the studies of HMDS under oxidizing conditions. The methlyated silicon structure of PDMS progressively decomposes on the catalyst, producing a silica layer that fouls the catalyst, in a manner similar to HMDS. 3

Deactivation by Phosphorus Compounds

3.1 Introduction.- The effectof phosphorus compounds on poisoning of noble metal catalysts has been widely studied. This section deals with deactivation of

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

227

noble metal catalysts by phosphorus compounds under oxidising conditions. Examples of such deactivation phenomena include the catalytic oxidation of warfare agents, automotive exhaust gases, where phosphorus-containing lubricating oils can contaminate the exhaust catalyst. Catalytic deactivation may occur for a number of reasons, both chemical and physical in nature. Several authors have reported that chemical poisoning of the noble metal catalysts is the primary mechanism for phosphorus c o r n p ~ u n d s . ' ~ ~ ' ~ Nevertheless, inhibition also takes place. The difference between phosphorus inhibitors and poisons is that inhibitors absorb weakly on the surface and the process is often reversible. On the other hand poisoning is the irreversible loss of activity due to the strong chemisorption of the impurities in the feed on the catalytic active sites. In general, organophosphorus compounds show relatively low thermal stability. They decompose relatively quickly under the process conditions, either being converted to phosphorus pentoxide, or in the presence of water vapour to higher, condensed phosphoric acids. It is phosphorus compounds such as these which reach the surface of the catalyst. Most researchers have concluded that phosphorus poisoning is non-selective and can therefore additionally serve as a model for other non-selective poisons such as lead and zinc."J6 Most researchers have

Table 8

1 2 3 4 5 6

Resultsfrom X R F analysis of six samples of spent H D T catalysis (from ref. 4)

2.8 3.7 3.0 3.6 3.0 3.O

5.6 7.7 7.4 7.8 4.6 7.3

5.5 5.4 5.9 5.4 6.0 5.9

7.5 6.8 7.4 6.9 7.5 7.5

2.0 2.1 2.1 2.0 2.1 2.1

405 278 376 502 446 328

395 365 365 340 405 370

"Determined by combustion analysis

Table 9

29SiChemical Shifts"for samples 1 , 4 , and 6 (from ref. 4)

Sample

Q2

Q3

Q4

T2, 7-3

T4

0 4

lb

-89.8 -88.9 -90.0 -89.5 -89.6

-101.2

-106.6 -107.6 -107.9 -109.5 -108.0

-54.2 -55.6 -53.8 -54.4 -53.8

-61.8 -63.7 -64.1 -63.0 -63.9

-12 TO -20 -18.0 -14 TO -20 -18.1

4b 4' 6b 6'

d

-99.6 -100.7 -99.3

"Relative to an external sample of neat TMS bDetermined by 29SiMAS NMR 'Determined by 29SiCP/MAS NMR dResonancenot clearly resolved 'No signal detected

e

228

Figure 9

Catalysis

Proton-decoupled 2ySiM A S and I3C CPIMAS N M R spectra of samples 1 , 4 , and 6, and ordinary liquid-state 2ySi and 13C spectra of neat PDMS. (a) Proton-decoupled 2ySi N M R spectrum of PDMS 656 scans, 8-s repetition delay, v, 500 Hz). (b-d) 29SiM A S N M R spectra of samples 1 (b), 4(c), and 6(d) obtained using identical spectrometer conditions (8000 scans, 8-s repetition delay, v, 7.2 kHz) and plotted using the same vertical expansion. (e) Proton-decoupled I3C N M R spectrum of PDMS (521 scans, 4-s repetition delay, Y, 500 Hz). (f-h) I3C CPIMAS N M R spectra of samples 1 (f), 4(9), and 6(h) obtained using identical spectrometer conditions (I 6000 scans, 2-ms contact time, 4-s repetition delay, v, 5.0 kHz) and plotted using the same vertical expansion. Spinning sidebands from aromatic coke resonances are marked with asterisks (fromref. 4)

-

-

-

LJ

also concluded that phosphorus poisoning is a slow, essentially irreversible proce~s.'~J~ 3.2 The Influence of Phosphorus Poisoning. - Catalytic oxidation using noble metal catalysts has been used to reduce the concentration of unburned hydrocarbons, carbon monoxide pollutants released from internal combustion engines, and similar applications. It is well known that contaminants arising from lubricants, (P, Ca, and Zn) deactivate these catalysts. Phosphorus compounds in printing processes are the source of decay of noble metal catalysts used to control these emissions.

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

229

Figure 10

29Si MAS (a) and COIMAS (b,c) N M R spectra of sample 4 . The 29SiCPIMAS spectra were obtained using contact times of 1.0 ms (b) and 2.5 ms (c), respectively, and otherwise identical conditions (from ref: 4)

Figure 11

Structure of PDMS and surface species of modified silica gels (from ref. 4)

230

Catalysis

Tzou and Weller (1994) have studied the catalytic oxidation of dimethyl methylphosphonate (DMMP), a nerve gas stimulant, over laboratory-prepared Pt/A1203as a function of Pt loading (0.5 or 2.0%) and temperature (150,250, or 400°C). l7 A conventional flow system was used for the activity studies. They identified intermediate phosphorous-containing compounds in the reactor effluent after catalyst activity decline and looked for evidence of aluminum phosphate in the catalyst, which could be produced by reaction of alumina support with phosphoric acid. The deactivated catalyst was studied by several methods: scanning electron microscopy (SEM)-energy dispersive spectroscopy (EDS), infrared spectroscopy (IR), and by extracting water-insoluble phosphorus. The SEM-EDS studies gave no useful results. IR absorption was measured on samples that were mulled in mineral oil. Comparisons of IR spectra were made with samples of y -alumina and aluminum phosphate. Determination of total P in the deactivated sample, presumed to be present as water-insoluble aluminum phosphate, was made by standard wet chemical analysis dissolution in hot, dilute HC1 followed by colorimetric determination of ph0~phate.l~ As expected, the 2.0% Pt/A1203catalyst showed less rapid deactivation than the 0.5% catalyst, as measured by breakthrough time in the fixed bed reactor at 250°C (75 vs. 50 h). At 150°C the catalyst showed a very short breakthrough time of only 5-6 h and the C02 production never exceeded 50% of theoretical, suggesting incomplete oxidation even on the fresh catalyst. At 400 "C, a decline in C 0 2concentration with time, coupled with a lack of unreacted DMMP in the effluent, suggested that the phosphorous was retained within the catalyst. Measurements on the deactivated catalyst provided evidence that the P compounds present in the used catalyst was AlP04.The IR spectrum of deactivated catalyst showed a peak around 1100 cm-', which is characteristic of A1Po4but is not present in the spectrum of A1203.The measured weight gain of the catalyst, after water extraction and drying, agreed with the weight gain calculated from wet chemical analysis for total P in the catalyst, on the assumption that the P was present as AIPOs. An earlier studyI8using this same compound, DMMP, led to a mathematical model of the deactivation process. Graven et al.I8 studied the oxidation of DMMP vapor in a stream of air, or nitrogen, over platinum-alumina catalysts. A commercial catalyst and a number of laboratory-prepared catalysts were investigated over a range of temperatures from 573-773 K, residence times from 0.15 to 2.7 seconds. The average catalyst particle sizes varied from 0.31 to 2.4 mm. They found that the fresh catalyst showed a very high activity, but after a few hours on stream it deactivated to the point that measurable quantities of DMMP vapor appeared in the effluent.. The reaction products over the deactivated catalyst were methanol and phosphorus acid. The authors did not characterise the deactivated catalyst and did not identify the nature of the poison. An increase in reactor temperature led to a longer breakthrough time. As expected, the breakthrough time decreased with average catalyst particle diameter.I8 A closely related study on the actual warfare agent isopropyl methyl

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

23 1

flurophosphonate (Sarin or GB, a toxic warfare gas) by Baier and Weller” was carried out at temperatures between 573 K and 673 K. Catalytic decomposition of GB vapor over Pt/A1203 in air resulted initially in stoichiometric amounts of the total oxidation products C02, HF, H20 and H3P04. As the activity declined, the decomposition shifted to the dealkylation reaction. The catalyst bed was examined after the run and found to have increased in weight by 44% and to be partially agglomerated near the bottom of the reactor. The particles were similar in color to that of the starting catalyst. No significant increase in pressure drop through the bed was encountered during the run. Analysis of the used catalyst for total phosphorous and water-leachable acid showed that only about half of the material attributed to the weight increase could be removed by Soxhlet extraction. X-ray diffraction patterns were more diffuse on the used catalyst after leaching, suggesting a reduction in Pt crystallite size. Surface area measurements by N2 adsorption showed a decrease from 64 to 45 m2/g.The authors suggested that reaction of the product H3P04with A1203gives relatively non-crystalline A1Po4which fills some of the pores and cause deactivation.” Another study of phosphorous deactivation has been reported for automotive applications?’ In this study, the phosphorous concentration in the gas stream is much less concentrated than in the studies described above, but some similar results are reported. The study examined the effects of oil-derived contaminants on catalysts for use in high-mileage taxis. The contaminants, mostly phosphate compounds, deposit in a strong axial gradient from the inlet to the outlet of the monolithic catalyst channel. The catalysts were standard 400 cell/in2 monoliths (0.69 L volume), containing a two-layer washcoat with Pd dispersed throughout both layers. The bottom layer was primarily stabilized ceria and the top layer was primarily stabilized alumina. A detailed characterization study including X-ray fluorescence (XRF), X-ray diffraction (XRD), electron probe micro-analysis (EPMA), solid state nuclear magnetic resonance (NMR) and electron spin resonance spectroscopy (ESR) was used to identify the nature of these deposited contaminants. Results showed an overlayer of mixed phosphates containing Ca, Mg, Zn cations, and aluminum phosphates, and possibly cerium phosphates, within the washcoat. Also a thin band of silicon contaminant interspersed between the washcoat top layer and the overlayer. Together, these poisons combine to produce severe deactivation of this monolith catalyst. These results do not provide evidence for deactivation dominated by glaze formation on the surface of washcoat. While the XRD data showed substantial contributions from crystalline Ca, Mg, and Zn orthophosphate as well as 31P,NMR data showed little if any evidence for a broad peak between -10 to -20 ppm that has previously been associated with glassy phosphorus deposits. The phosphorus chemically associated with alumina in the catalyst, as determined by NMR, was the orthophosphate. The authors suggest that a band of an amorphous hydrated form of phosphate could be formed under operating temperatures of taxi catalysts (673-773 K).20 The activity tests over aged and fresh catalysts (using propylene oxidation as a measure of activity) showed that the inlet section of the catalyst never achieves complete propylene conversion. This suggests a diffusion limitation by an ex-

232

Catalysis

tremely thick phosphate overlayer (and possibly the silicon-containing layer). The middle section of the catalyst showed strong deactivation (relative to the fresh and dynamometer-aged catalysts) despite having almost no overlayer material. This indicates that phosphorus compounds in the washcoat (A1PO4and possibly cerium (111) phosphates) strongly affect the deactivation. The outlet section of the catalyst contains the least amount of phosphorus and exhibits activity similar to that of the thermally aged dynamometer catalyst. Gandhi et aLZ1have also studied deactivation by P contamination in the washcoat of the aged vehicle catalysts. They suggested that P arises from lubricant additives burned in the engine at high temperatures, which are released to the exhaust gases as solid particles. Kim et a1.” performed an analytical electron microscopy to study the microstructure and microchemistry of fresh and two vehicle-aged commercial pelleted automotive exhaust catalysts containing Pt/Pd/Rh. The two samples were selected with similar poison levels but having different catalytic activities. In their investigation, the radial distribution of poisons was also analysed by taking a series of EPMA area scans at points along a radial line near the pellet edge. The alumina support generally looked dense and was wrapped in a roughly 2 pm thick dark deposit. The major poison constituents Si, P, Pb, Mg and Zn were detected near the pellet surface. Except for Si, all other poison elements were absent below 7 pm. These authors also employed analytical electron microscopy (AEM) in particle by particle analyses of over 400 aged noble metal particles along a line tangent to the particle surfaces. This did not reveal the presence of a detectable quantity of any foreign species (perhaps not more that a few atomic layers thick). According to these authors, the physical aspects of poison accumulation appear to be similar in all aged samples. Other than the general conclusions concerning pore blockage, little can be said concerning the quantitative impact of poison microstructure and microchemistry on catalyst activity. Further microscopic studies of several areas on samples of both aged catalysts suggest that the surfaces be at least partially covered with glass-like poison layers and poison fragments. Many small (30-50 nm) particulates were observed near the pellet surface in higher magnification STEM images. EDS analyses suggested that these fragments are mainly zinc phosphate and/or lead phosphate contaminated with small amounts of several other elements. Once these layers are formed, most near-surface macro- and micro pores are blocked, inhibiting gas transport into and out of these pores that contain the noble metal particles. It is also interesting to note that STEM observations in this region of the catalyst reveal many very small noble metal particles. They concluded this high local density of small particles in the near-edge region may result either from (i) a reduction of the inherent sintering rate on the poison-modified surface or (ii) lower peak reaction temperatures in this region due to pore blockage and/or physical encasement of the noble metal particles. The poorer performance of this catalyst suggests the latter may be a more accurate description. This group also attempted to regenerate the catalyst23and found that washing an aged automotive exhaust catalyst with an aqueous solution of oxalic acid

5: Deactivation of Oxidation Catalystsfor VOC Abatement by Si and P Compounds

233

followed by calcination, results in the redistribution of phosphorus and lead based poison deposits, but leaves silicon- and sulfur- containing species in place. They also employed a high temperature chlorination treatment, but this process showed a little effect on poison radial distribution within the catalyst pellet. The washing treatment reduced the level of P and P b as measured by bulk chemical analysis; P is reduced from 0.26% to 0.21% and Pb is reduced from 0.15% to 0.10%. In addition, the washing treatment reduced the peak level of both P and P b at the edge of the pellet to less than half the aged sample level and redistributed these poisons over the volume of the pellet. The redistribution of P and P b was also demonstrated by electron probe microanalysis (EPMA) on vehicle-aged oxidation catalysts. This process is expected to reduce the degree of pore blockage due to glassy deposits of P- and Pb-containing materials which were observed in the surface region of the aged sample, thus improving the rates of reactions which are sensitive to mass transport within the pellet. They concluded that, because of significant solubilities of these materials in aqueous acids, the washing procedure dissolves and redistributes the components of these glasses, opening up the pore passages. On the other hand, the redistribution of poisons over a large volume of the pellet could reduce catalyst activity by contaminating more of the noble metal particles. The tendency for this to occur may be minimized by the ability of the alumina support to scavenge and react with both P (as a phosphate) and Pb (as an oxide). They showed the net effect of the washing treatment was to produce significant improvements in hydrocarbon and NO, activities but effectively no improvement in carbon monoxide activity. This suggests differences in the sites required to oxidize these corn pound^.^^ EPMA was also performed by Liu and Park to examine possible deactivation mechanisms of automotive catalysts.24These authors examined the glaze formation in the cross-sectional direction of four catalyst samples with the same formulation from in-use vehicles whose emission control capability had been severely reduced. They measured over 600 data points from dense glaze and contaminated y-alumina particles in these samples. A major effort was made to ensure that the data collected had adequate accuracy and were truly representative of the samples in investigation. Using Wavelength Dispersive Spectrometry (WDS), they found four major elements: phosphorus and zinc (from the oil additive zinc dialkyl dithiophosphate (ZDDP)),calcium (a detergent element in oil) and lead (residue from gasoline) in the deactivated samples. For morphological observation, in addition to scanning electron microscopy (SEM), they also performed secondary electron imaging (SE), and backscattered electron imaging (BSE) in the EPMA. From these observations they found in all samples a surface layer from 10 to 20 pm thick was found in the washcoat full of contaminant elements. A flaky crust on the channel surface was identified and a continuous glassy crust around 2 pm thick underneath this seemingly porous flaky crust. They observed the y-alumina particles in the washcoat immediately below the crust were also sintered by contaminants. The authors claimed from plotting the atomic ratio of Zn/P, Pb/P, and Ca/P versus atomic percent of (A1+ Ce; the main constituents of fresh catalyst), and using information from previous investigations, that these contaminants could be derived from ZDDP oil additives:

234

Catalysis

Pb3(P04)2,Ca3(P04)*,Zn2P207and AlP04. Their findings are summarized in Table 10. Liu and Park concluded that the deactivation was probably due to either the formation of an impervious glaze layer on the washcoat surface, sintering and gluing the y-alumina particles by glassy phase in the washcoat, or simply poisoning from lead. These results are consistent with the results of an earlier study by Hegedus et a1.,25who showed that phosphorus penetrates the catalyst pellet in a well-defined band. They believe that this is a diffusion-limited poisoning process. The conditions of their tests were 839K using a high P fuel. The measurement of phosphorus concentration along the pellet radius indicated that its profiles could be well approximated by a step function. They also showed that the phosphorus level of the inward progressing concentration profile reaches saturation such that as the phosphorus-containing poison precursor (presumably H3P04)enters the pores of the catalyst, diffuses past the already poisoned layer to react with the support and active sites at the leading edge of the phosphorous concentration profile, Figure 12. Also, they suggest that the interaction between the poison precursor and the catalyst was chemical rather than physical. These observations indicate that the phosphorus accumulation process is irreversible and essentially independent of the presence of an active metal component over the alumina support. Also, the electron microprobe observations suggest that the poison accumulation process is of the ‘progressive shell’ type in which the saturation concentration of the poison in the poisoned shell and the effective diffusivity of the poison precursor across this shell play an important

Important components of the contamination in various locations of the samples; (Liu and Park, 1993)

Table 10 _

Sample

_

_

Flaky crust

_

_

~~~

~

~

~

~

~ ~

~

~

From points straddled between a crust and y-grain surface and from points straddled Dense crust between sintering From points in y-grain underneath the agents’ and a grain phase but not deeper Paky crust surface than 10 pm

[] indicates the minor component;components not listed are of much lower concentration

~

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds i

Figure 12

235

Pelfet edge -1

Phosphorus concentration profiles along the radius of a poisoned catalyst pellet, measured by electron microprobe (from ref. 25)

They varied surface areas and repeated the tests under similar experimental conditions. The heights of the saturation plateaus were expressed in saturation phosphorus concentrations and were plotted against the N2BET surface areas of the catalyst pellets, Figure 13. The slope of the saturation concentration versus surface area line was about 2.2 x 1015 phosphorus atoms per cm2 of support surface, quite close to the number of atoms per cm2of solid surfaces. Therefore, this indicates a monolayerequivalent coverage of the alumina surface by phosphorus. The fact that there was very little change in the effective diffusivity upon poisoning by phosphorus compounds suggests that the poisons tend to deposit in a monolayer-like concentration over the surfaces of the poisoned shell. The simple pore-mouthpoisoning mechanism was adapted for catalyst deactivation by phosphorus Angele et a1.26investigated the effect of phosphorus compounds on poisoning of noble metal catalysts. They suggested that it is a typical example of nonselective poisoning. They studied the different catalysts after they had been contacted with phosphorous in a differential gas-flow fix bed reactor that allowed simultaneous activity tests. They used two types of catalysts. The first was a platinum wire gauze (wire diameter was 40 pm); all-metal catalyst (Cr-Ni steel wires with electrolyticallydeposited Pt black, promoters and stabilizers). The second was a series of three different supported catalysts: 0.1% Pt/a-A1203(approx. 5 m2/g), 0.5% Pt/y-A1203(approx. 50 m2/g), 0.2 % Pt/Si02 (approx. 130 m2/g).

Catalysis

236

0.

I

0.

Saturatiosl Concentration &p;fJ cat,)

Figure 13

Stape = 2.2 x 1 0 ' s

P AtomsJcn2

Phosphorus saturation concentration as a function of the N 2 BET surface area of the supported catalyst from electron microprobe analysis (from ref: 25)

Measurements were carried out with the model poison triethyl phosphate at 520-800 K. The results showed that the platinum wire gauzes were completely deactivated by even very small, unweighable quantities of poison. The wires could be regenerated completely with water and also annealing at 1100 K. The regeneration cycle was repeated several times, the platinum gauzes remained unchanged, whereas the Pt-black broke away and the catalyst was damaged. The supported catalysts were found to be far less sensitive to poisoning; considerable quantities of phosphorus were needed to achieve complete deactivation. Figure 14 shows relative activity versus poison content, expressed as the mass increase of the catalyst. The reaction rate decreased as the content of phosphorus increased and approached the saturation point.26 Depending on the temperature at which poisoning had taken place, the process of washing the phosphorus out of supported catalysts gives different results. When the catalysts had been annealed or heated to over 800K, no phosphate could be washed out; however, if they had been poisoned at lower temperatures <700 K, it was always possible to wash out a considerable percentage of the quantity of poison. Phosphate ions were found in the wash water. Pt remained inert in the presence of P4010and phosphoric acids until very high temperatures were reached. Therefore, they suggested that the poisoning of platinum wire gauze and all-metal catalysts was only possible through physical coverage of the surface, whereby cross-linked, glasslike coatings of phosphorus

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

237

pentoxide or polymeric phosphoric acids were formed. These could be hydrolysed and washed off with water. It could also be vaporized by heating; however, pentoxide forms an azeotrope at 94% P205and 6% H20 at 1140 K. Aluminum formed a considerable number of phosphates: AlP04, diphosphates, polymeric metaphosphate, acidic and basic phosphates. Some of these could contain different quantities of crystallization water. However, SO2 and P205had a greater tendency to form non-stoichiometric or glassy mixed oxides together. In the case of A1203,the coatings could be largely attributed to Alp04 if the system temperature was sufficiently high and the reaction time sufficiently long. Angele et ~ 1 showed . ~ that ~ BET surface area decreases with increasing phosphorus content for three different catalysts, Figure 15. There is a linear relationship up to a high poison content. The experiment also showed that phosphorus reduced the porosity of the catalyst linearly. Figure 16 shows that the mean pore radius has increased, as phosphorus content becomes larger. There can also be interaction between phosphorous and other poisons. Williamson et al. found that the individual effects of P and Zn on a three-way automotive catalyst were small compared to the combined presence of P and Zn.27They showed the presence of AlPO4on the catalyst containing 2.83% P and little Zn. Using a dynamotor to age the catalyst resulted in a catalyst containing

Figure 14

Decrease of activity (kinetic region) with increasing poison content (Ref. 26)

238

Figure 15

Catalysis

Decrease of the BE T-surface of diflerent catalysts with increasing poison content (Ref.26)

4.08% P; little Zn, and A1Po4 was detected. Phosphorous accumulated as a surface deposit on the outer edge of the washcoat, aged with zinc dialkyl dithiophosphate (ZDDP) and cresyl diphenyl phosphate (CDP). Using the CDP for aging the catalyst, they found that the aged catalyst had a P content of 6.51% P and a layer of 5 pm. 3.3 Support Effects. - The support may adsorb or react with a poison. Common examples can be taken from automobile exhaust catalysis, where P-containing impurities in the feedstream are quite reactive and react with the y-A1203support and a species of Al-0-P may be formed.” In the study of Tzou et al.,17 discussed earlier, measurement on the deactivated Pt/A1203catalyst provided evidence that the P compounds present in used catalyst were AlP04. In the work carried out by Baier et ~ 1 . on ’ ~ the catalytic decomposition effect of isopropyl methyl flurophosphonate (Sarin or GB a toxic warfare gas) vapor in air over Pt/A1203,the oxidation product H3P04reacted with the A1203support give relatively noncrystalline Alp04 that filled some of the catalysts pores and caused deactivation. For Pd supported on the monolithic samples, Rokosz et a1.” showed two major forms of phosphorus compounds, alumina phosphate within the washcoat and zinc, calcium and magnesium phosphates in an overlayer. The contaminants

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

Figure 16

239

Apparent increase of pore radii with increasing poison content for two diflerent catalysts (Ref: 26)

were mostly phosphate compounds. The washcoat bottom layer was primarily stabilized ceria and the top layer was primarily stabilized alumina. The formation of cerium phosphate and alumina phosphate was thought to be responsible for deactivation. They concluded that cerium (111) phosphate was formed, although evidence for the formation of cerium (111) phosphates was indirect, resulting from the coincident observation of an NMR-inactive species that decomposes at high temperatures with a concomitant growth in A1P04 concentration and a decrease in ESR signal associated with aggregated Ce+3species after the high temperature treatment. Both observations are consistent with NMR-inactive cerium (111) phosphate decomposing/reacting at high temperatures to provide aluminum phosphate and cerium dioxide. Liu and Park24 in their study on deactivated automotive catalysts in the presence of zinc dialkyl dithiophosphate (ZDDP) and Pb showed the formation of A1PO4 by reaction wit the support. The deactivation was due to several types of interaction with the alumina support, including the formation of an impervious layer on the washcoat or sintering of the y-alumina particles in the washcoat. 3.4 Mechanism and Kinetics. - Knowledge of catalyst poisoning and the resulting understanding of the decay mechanism can significantly improve the design of commercial catalysts. The three main mechanisms of catalyst poisoning are: poison adsorption, poison-induced surface reconstruction and compound

240

Catalysis

formation between a poison precursor and the catalyst. In most cases, the deactivation by phosphorous compounds may be attributed to the compound formation mechanism. Automobile fuel and lubricating oil contain impurities of phosphorous, which give rise to catalytic deactivation. It is assumed that contamination covers the active sites, or reacts with the support. In the latter case, it can be suggested that the poison precursor (H3P04)reacts with the alumina or cerium support and forms poisoned areas with close to monolayer coverage of the alumina surface by phosphorous. Two deactivation mechanisms have been suggested for phosphorus compounds.26The first mechanism is coverage of the surface of noble metal and supports by P4Ol0or polyphosphoric acid coatings. This is a reversible mechanism and the poison can be removed. The second mechanism occurs at high temperatures, where phosphorus produces a mixture of phosphates and mixed oxides by reacting with the carrier material. This mechanism is irreversible, and considerable quantities of phosphorus can be deposited in the catalyst, clog the pores, and cover the noble metal crystallites. Angele et al. suggest that this is non-selective catalyst poisoning.26 Graven et ~ 1 . 'studied ~ the oxidation of the warfare agent stimulant DMMP vapor in a stream of air, or nitrogen, over platinum-alumina catalysts. They found that the fresh catalyst shows a very high activity, but after a few hours on stream, deactivation occurs as measurable quantities of DMMP vapor appeared in the effluent. A pseudo steady-state kinetic study on the deactivated commercial catalyst showed that the oxidation reaction is first order with apparent activation energy of 7 to 8 kcal. This suggests intra-particle pore diffusion mass transport limitations at these conditions. The deactivation process presumably takes place when the active sites react with the poison precursor. While in many cases this is indeed true, the same phenomena can arise due to intra-particle transport effect. An empirical correlation was established for the catalyst deactivation behavior, as measured by the decrease of first-order rate constant, k, with time on stream using the following relationship;

k = koe-ut Where k is the apparent first order rate constant after time t and is the extrapolated rate constant at zero time, a is a deactivation rate constant. Graven et a1.18showed that the over-all DMMP conversion preceded about as rapidly in nitrogen as in air, and product analyses indicated that the initial reaction over deactivated catalyst was principally a hydrolysis reaction forming methanol and phosphorus acid. In the study by Kim et al.23a zinc phosphate phase and a negligible amount of lead phosphate were detected in the surface of catalyst, a pore mouth blockage model of poisoning was also suggested. Pore-mouth poisoning takes place when the outer portion of the catalyst has been poisoned by a sharply defined poison front. A diffusion controlled mechanism of deposition of phosphorus particles has been also proposed to explain the concentration of P contamination (arising from lubricant additives) in the external side of the washcoat and a mathematical

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

241

model has been developed to describe this deposition.**Often, a phosphate band that forms by reaction between phosphorous compounds and the support material has been responsible for the deactivation by phosphorus. The most serious form of catalyst poisoning by phosphorous compounds has been represented by the reaction between a poison precursor and the catalyst. Kiskinova and Goodman16explored the effect of electronegative poisons, including P, C1 and S , on the chemisorption of CO and H2 over Ni (100).The presence of these electronegative atoms on the surface reduced the adsorption rate, adsorption bond strength, and the adsorption capacity of the surface. They concluded that in the case of these electronegative poisons, the extent of poisoning was correlated with the increasing electronegativity of the sequence, P < S < c1.

4

Mathematical Modeling of Deactivation by Si and P-based Compounds

Mathematical modeling of catalyst deactivation has been addressed by many F r ~ m e n t , ~ ~H, e~g' -e~d~~ s ' ~and .~~~' authors, including B ~ t t ,bar ~ tho ~ -lo~me^,^^*^^ ~ P e t e r ~ e n , 4among ~ - ~ ~ others. Although many of the approaches to modeling of deactivation can be applied to any reaction, the focus of this section is to review the literature specific to the deactivation of oxidation catalysts. These modeling approaches have been applied to deactivation by organo-ph~sphorus'~-'~ and silica'-6compounds, which are generally recognized as poisons at the conditions of interest here. There are only few papers that have focused on modeling this phenomenon, with emphasis on poisoning by phosphorus ~ ~ m p o u n d s ~ ~ ~ ~ 4.1 Mathematical Approaches. - The complexity of mathematical modeling of catalyst deactivation is mainly due to developing kinetic equations of the deactivation phenomena and measurement or estimation of the various parameters. When two or more different deactivation processes occur at the same time, this adds another level of difficulty and complicates the interpretation of experimental results. There are two main approaches in modeling deactivation phenomena, a microscopic and a macroscopic approach. These include three levels of modeling, first at the individual catalytic site level, second at the particle level and third at the reactor level. At the microscopic level, we deal with information at the site, the chemical pathways leading to the deactivating agent and its interaction with the sites. Here the intrinsic kinetic aspects are important. At macroscopic level, we deal with the catalyst particle (modeling at particle level) and collection of particles (modeling at reactor level). Each catalyst particle is viewed as an entity. At particle level the results of intrinsic kinetics are coupled with the influence of catalyst morphology, and transport of heat, reactants and products in the interior of the particle. The final level is to couple these phenomena for a collection of particles with heat and mass transport and fluid flow in the reactor. This is modeling at reactor level, which studies the effects of deactivation on the reactor performance and ~ p e r a t i o n . ~ ~ . ~ ~ " ~

242

Catalysis

Fluid flow in

Reactor catalyst lifetime, reactor performance with time and >peration conditions, (poison loading and profile.. .)

Figure 17

chemical properties of

Modeling of catalyst deactivation

Figure 17 shows the general scheme for deactivation modeling. The rates of the main reactions and deactivation are important constituents of each model. It is clear that a mechanistic kinetic model is preferable although, because of complications included in deactivation mechanisms, the catalyst deactivation rate is typically characterized through empirical models. Once the kinetic model is developed, we can focus on the particle. If deactivation involves morphological changes such as pore blockage, the structure of pore network of the particle must be taken into account. Modeling tools for this are a ~ a i l a b l e . ~The ~ . modeling ~~.~~ at particle level has been widely studied.15~43~50~56-58 For the modeling at reactor level, classical heat, mass and simple momentum equations are coupled with the models that describe the deactivation at particle level. There is a wealth of open literature to develop and solve these equations for fixed bed reaCtorS.25,28,43,5 133,5639-61 With the computational facilities available, it is possible to develop and solve more sophisticated deactivation models. The results of mathematical modeling are used to predict catalyst lifetime as a function of time and reaction conditions. For poisoning, the accumulation of the poison and the concentration profile within the pellets and reactor can be estimated. The results can be used to design new catalysts.

4.2 Analytical and Numerical Methods. - Modeling deactivation at the particle and reactor levels involves solution of boundary value problems that result

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

243

(in simple cases) in a set of ordinary differential equations and (in rigorous cases) a set of partial differential equations. For some types of problems analytical or semi-analytical solutions can be ~ b t a i n e d . However, ~ ~ * ~ ~ for most problems of practical interest one has to use numerical solutions. In general, these equation sets are nonlinear and cannot be solved easily by using analytical methods. A discussion of modeling of the phenomena within catalyst particles, physical and numerical methods has been recently published by Kei1.58

4.3 Modeling of Catalyst Poisoning by Organosilicon Compounds. - There is little published information on the modelling of deactivation by organosilicon compounds. In a recent study Libanati and coworkers2have developed a simple reaction engineering model to perform sensitivity analysis on the effects of organo silicone (precursor) concentration on the deactivation rate and performance of commercial VOC catalytic incinerator. They used this model to design and size the industrial reactor to meet specific lifetime guarantees. However, they have not indicated what kind of model or assumptions they used. No other open literature on modeling the catalyst poisoning by organo silicon compounds has been found. 4.4 Modeling of Poisoning by Organo Phosphorous Compounds. - Hegedus and Cavendish have proposed a mathematical model for phosphorus poisoned automotive catalysts.43Using electron microprobe analysis they observed four different zones in a partially deactivated pellet. The schematic representation of the cross section of this pellet is shown in Figure 18. They used this model to predict the effective diffusivity in the partially impervious deposit (zone 4). They developed and solved the coupled mass balance equations for spherical pellets in a fixed bed reactor. To compare the model with experimental data, they measured the conversion of propane and propylene over partially poisoned pellets in a small isothermal packed bed reactor. For intrinsic kinetic rates of propane and propylene they employed the empirical rate equations proposed by Volts et U Z . ~and Hiam et The Hegedus model is briefly described below: Mass balance over pellet (composite domain): The mass balance of reactive species in the catalyst pellet is described by:

subjected to the following boundary conditions: at r = rl

-dCi =o dr

Catalysis

244

0

rl

r;? r3 R

Zone 1 : Unimpregnated core Zone 2,3: Noble metal impregnated shell Zone 3: Poisoned shell Zone 4: Partially impervious deposit

Figure 18

Representation of the cross section of a partially phosphorous poisoned catalyst pellet43

at r = r2 r = r2-

(4) r = r2+

at r = r3

where km,i is the bulk mass transfer coefficient (m/s) for reactant i. Mass balance over integral reactor: The integral fixed bed reactor is simulated by a cascade of cells. Each cell contains N catalyst pellets. Usually Vcellis selected such that its dimension in the flow direction is comparable to the diameter of catalyst pellet. N is calculated as following:

The reactant balance around cellj is:

where Ri is the total rate of reactant consumption by the single catalyst pellet:

Ri = 4njOR r2a(x)Ri dr, kmol/pellet.s and Ri is local reactant consumption (kmol/(m2of Pt).s). They solved these equations using a finite element method (Ritz-Galerkin

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

245

method). The effective diffusivity in impervious deposit shell (zone 4) was considered as an adjustable parameter to fit the experimental data. Figure 19 shows the comparison of measured and calculated propylene conversion on phosphorus-poisoned ~atalyst.4~ Curve 2 compares theory to the low temperature data by adjusting the reaction rate constant. Curve 3 was obtained by reducing the diffusivity of zone 4 by a factor of 25 from its original value, indicating that zone 4 is severely plugged by deposited poisons. Hegedus and Baron25used a shell-progressive poisoning model to study the deactivation of automotive catalysts by organo phosphorous. They employed the modeling approach of which has been generalized by B i ~ c h o f f . ~ ~ According to Olson model, a poison (supplied in the feedstock or formed as a by-product of the reaction system) accumulates on the catalyst pellet by a shell-progressive mechanism. Thus the fixed bed reactor has developed poremouth poisoning of a variable extent throughout the reactor. The poisoned fraction is considered as a fully deactivated shell but is permeable to reactant, product and poison species. It is also assumed that the effective diffusivity in the pellet is constant and does not change during poisoning. The mass balance equation for the poison is based on ion-exchange equations.68In dimensionless form these equations for an isothermal plug flow reactor are written as follows:

100 r

90 80

-

70 -

60

-

Conversion 5 0

-

C3”6

Measured

-

Calculated

(%I

,,

40-

30-

,

4

20

-

10I OA 300 400

1

500 T(K)

Figure 19

I 600

Propylene conversion over the poisoned catalyst. Curve 2 shows calculated conversion, curve 3 shows calculated conversion afer adjusting the diflusion coeflcient in zone 443

246

Catalysis

These coupled equations are subjected to the following boundary and initial conditions: at

<=0

at z = 0

Yp(<,z)= 1

(11)

q(<,z) = 0

(12)

<

where Yp,q, and are dimensionless concentration of poison-producing material in the fluid phase, volumetric fraction of the solid phase saturated with poison (fully deactivated), and the dimensionless distance through the bed, respectively, and z is the dimensionless time defined as:

@(q) indicates the mechanism of accumulation of the poison within the individual catalyst particles. @(q)has been derived by Olson for a shell-progressive poison accumulation in spherical catalytic pellet as follows:66 N

where:

a

= (1 - E-)

3DeL R2Y

kmR Bi = - (Biot number) De and kPR (Damkoehler number) Da = Dk

(17)

Hegedus and Baron" evaluated this model numerically and solved the equations for the poison distribution in the bed. They calculated the total poison content of the bed as:25

+-

a

a Da

Figure 20 shows the comparison of measured and calculated total phosphorus accumulation in a four stage fixed bed reactor used by Hegedus in a steady-state,

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

247

t

dhr)

Figure 20

Comparison of experimental and predicted total phosphorus accumulation, in the four-stage react^?^

accelerated phosphorus deactivation experiment. Excellent agreement between model and experimental results were obtained. In a recent modeling study, Angele and Krishner” have developed a model for poisoning of supported noble metal catalysts by phosphorus compounds in an isothermal fixed bed reactor. They have solved the model for a single pellet. Their model is based on the following assumptions: Isothermal condition. External mass resistance is negligible. The main and poisoning reactions are irreversible. The poisoning reaction if first order with respect to both the poison concentration and the adsorption sites. The main reaction is first order with respect to the concentration of reactant and active sites. The relationship between the deposited poison content and number of deactivated sites is linear. The poisoning rate with respect to the main reaction rate is low, so the quasi-steady state assumption is valid. Catalyst pellets are spherical of equal size. For pellets with different shapes the hydraulic radius of pellet is used. Using these assumptions the model can be written as follows:

2dYp cpdcp

+-d2Yp = @2pYpe dcp2

where:

248

Catalysis

de --- e.yP dz

(21)

with the dimensionless time variable: COP kP z=--t=*

c;

t t*

The boundary and initial conditions are: dYA -=o,

atcp=O

dcp

d YP =o dcp

-

These equations can be extended to cover a long fixed bed reactor by replacing the boundary condition (24) by balances for the fluid (one-dimensional plug flow reactor). This set of equations was solved using a numerical method developed by Schl0sser.6~ The residual catalyst effectiveness factor can be calculated for an individual pellet as follows:

They studied two limiting cases for this mathematical model. The first case was for a kinetic controlled mechanism (uniform poisoning and (Ip+O). In this case, the equation (20) is replaced by the condition Yp= 1 throughout the pellet and equation (21) is integrated:

(27) The integration of equation (19) alone leads to the following expression for effectiveness factor for spherical pellet:

e(z) = exp (-z)

3

q = --($A

44

*

coth (IA - 1)

and the rate of reaction is calculated by: -r0

(I* e-7/2* coth ((I e-'l2) - 1 (I*coth (I- 1

The second case was the shell model poisoning or diffusion controlled mechanism. They used the results of Haynes 70 to calculate the rate of reaction:

where, (J was the dimensionless radius of the un-poisoned core, which can be calculated as 50:

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

249

Angele and Kirchne9* carried out the same mathematical modeling for poisoning of a honeycomb catalyst by phosphorus compounds. They ignored axial diffusion compared to convection and simplified the transport equation as:

1 9' -

dYp cp2)*--

ds

d2Yp + _.1 dYp

=dcp2

rp dcp

where:

The balance between diffusion and reaction at the wall is calculated by:

where DaIIis a Damkoehler I1 group as:

khet

is the heterogeneous, area-related rate constant:

khet ' R DaIr = -

(37)

Df

At the beginning of the channel, the boundary condition is: Yp = 1 a t c = O a n d c p = 1, and at the center of the channel:

atcp=O

(38)

a YP -0 --

(39) drp The model can be solved by numerical methods. For a first order reaction, Brauer and S~hliiter'~ showed that:

where c'and C, were the mean concentration over the cross-section and the wall concentration, respectively. The coefficients B, and the factor m, are functions of

250

Catalysis

Da. In the study of Angele and Kirchner 28, the series term in equation (40) converges rapidly, so that the equation can be simplified to:

m land B1can be calculated as a function of Damkoehler number. Change in catalyst mass can be obtained by:

cfin equation (42) can be calculated from equation (41), so that:

For constant Cfoand Q,

w is also constant, and equation (44) can be integrated:

where the total throughput of poison, mtot,is calculated by: mtot= M.Q.Cfo-t

(46)

If (dmldz) is replaced by finite, short sections of honeycomb, Am/Az, and In (Am/Az) can be plotted versus z, yielding a straight line with the slope: d ln(Am/Az) w dz

-_

ml-Df

(47)

w 4R2 *

These quantities can also be estimated the mean relative activity:

- l 9 "-GZo a

(

t

F.=-exp w

ml.Df [ -------I)

j.L w-4R2 10

F and ml Df/4R2are constant and estimated from experiments.28 *

4.5 Optimization of Active Phase Distribution for Deactivating Systems. - One of the benefits of mathematical modeling of deactivation phenomena is to help develop new catalysts for processes involving poisoning, i.e. to develop poison resistant catalysts. Most of the catalysts used in VOC oxidation are noble metals dispersed on a high surface area support. These supports are generally spherical or cylindrical pellets, with the catalyst deposited on the external surface or uniformly within the pellet. It has been shown that the location of the catalyst in the support can have a significant effect on catalyst performance, especially when diffusional resistances during reaction are important. DeLancy 56 may have been the first to suggest the idea of an optimal catalyst activation policy for poisoning problems. He included economic analysis in his

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

251

model to balance the catalyst costs against the net return from reactant conversion. Later, Hegedus 44 introduced a procedure for improving the poison resistance of automotive exhaust catalysts toward lead and phosphorous. One of the aspects of that work was optimization of noble metal position inside the catalyst. Additional studies by Brunovska, Morbidelli, Vayenas, and Varma57772-75 discuss optimal distribution of the active metal within pellets and reactors. As an example, the general model formulation for a pellet undergoing poisoning is presented here73i76.Consider a catalyst pellet in which an irreversible reaction is taking place together with irreversible adsorption of catalyst poison. The rate of the poison adsorption is usually considerably lower that that of catalytic reaction (as in the case of organo phosphorus and organo silicon poisoning). If so, the quasi-steady-state approximation can be employed to write the mass balance equation for the reactant. The initial catalyst active metal distribution is a function of location, so during operation involving poisoning this distribution will also change with time. Defining an average active component density, we can write activity distribution as follows:

where

s:

a = (n + 1)

o(rp,.c = o)rpndcp

n is the characteristic pellet geometry. For cylindrical pellet n = 1 and spherical pellet, n = 2. From the definition of average active component density [equation (SO)] and activity distribution we can write:

This means the average initial active component density is equal to one for any distribution and geometry. The mass and energy balance in dimensionless form are as follows: Mass balance for main reaction:

Mass balance for poisoning reaction

Energy balance

The above equations are subject to the following boundary conditions

252

at

Catalysis

cp = 1,

Y,=Yp=w=l

(56)

Under poisoning conditions a balance for active sites, in terms of the activity distribution can be written as follows:

with initial condition:

The normalized effectiveness factor with respect to initial value of reaction rate computed at surface conditions and to the initial activity distribution (nonpoisoned catalyst pellet) is defined as following:

so with the above definition, the normalized effectiveness factor represents the mean reaction rate as a function of time. The total dimensionless reactant conversion up to complete catalyst deactivation is computed from the following equation:

1 1

H =

q((z)dz

0

Where dimensionless time is t = t / t d and t d is the time required for complete catalyst deactivation. The objective is to maximize the total reactant conversion. The method for solution of this optimization problem and some limiting cases can be found in B r u n o ~ s k aThe . ~ ~advantages of this technique can be seen in an early work by Hegedus and coworkers in improving the phosphorus-tolerance of automotive exhaust catalyst^.^^@.^^ Summary. - This analysis has shown that the general mathematical models for deactivation can be applied for analysis of deactivation by phosphorous compounds present in VOC oxidation. Using simplifying assumptions specific to that case, the poison loading as a function of time and position can be calculated. Limiting cases for various ratios of poisoning and main reactions, and transport within the particle, lead to models that describe the deactivation of these types of catalysts. Also, it has been shown that one benefit of developing such models is to design new catalysts more resistant towards phosphorus poisoning. 4.6

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

253

Nomenclature a aP B Bi C

c,*

D Da F H j

k kfn

L rn m M n n

dimensionless catalyst activity, local noble metal load, cm2Pt/cm3 equilibrium poison adsorbed amount coefficient, eqn (40) Biot number concentration poison concentration in a pellet that leads to total deactivation diffusivity Damkoehler number complex factor in Equation (48) average reactant conversion over catalyst lifetime defined by equation (60) summation index, eqn (40), (48) local reaction rate constant bulk mass transfer coefficient bed or reactor length eigenvalue, eqn (40) mass of catalyst molar mass of poison reaction order, eqn (34) integer, characteristic of pellet geometry, 1 for cylinder, 2 for sphere, eqn (50)

N 4

Q

r rP r0 R RA RP

R

R t t* td

V 21

-

W

Y Z

number of catalyst pellets in one cell fraction of the solid phase saturation with poison volumetric gas flow rate radial pellet coordinate poisoned reaction rate unpoisoned reaction rate radius of spherical catalyst pellet, honeycomb channel radius, eqn (33) dimensionless main reaction rate dimensionless poisoning rate specific reaction rate, mol/cm2Pt.s rate of reaction per one catalyst pellet, mol/pellet.s time time constant of poisoning reaction, eqn (22) characteristic deactivation time volume dimensionless temperature mean velocity dimensionless concentration reactor coordinate, axial

Greek a

parameter in eqn (14)

254

P A E

5

r r7 6 Y

0 0

T

4) rp Q, @A

Q,P

Catalysis

dimensionless heat of reaction difference bed void fraction dimensionless axial coordinate effectiveness factor normalized effectiveness factor fraction of unpoisoned surface interstitial velocity, eqn (13) dimensionless radius of the unpoisoned core, eqn (30),(31) concentration of available catalytically active sites, eqn (49) and (50) dimensionless time Thiele modulus dimensionless space coordinate poisoning mechynism function, eqn (10) = a[r@ACAo)]Z,IreactionThiele modulus, eqn (52) = a[ap/(DpCpotd)]~, poison Thiele modulus, eqn (53)

Subscripts 0 cell e

f het 1

P tot W

initial one mixing cell effective fluid phase heterogeneous component i poison total wall

Superscripts -

0 j sat

average inlet number of cell saturation

References 1. 2. 3. 4. 5.

J.-J. Ehrhardt, L. Colin, and D. Jamois, Sensors and Actuators B, 1997,40,117. C. Libanati, D.A. Ullenius, and C.J. Pereira, Appl. Catal. B: Envron., 1998,15,21. C.F. Cullis and B.M. Willatt, J. Catal., 1984,86, 187. L. Kellberg P. Zeuthen, and H.J. Jakobsen, J. Catal., 1993,143,45. M.O.G. Souza, P. Reyes, and M.C. Rangel, ‘Silicon poisoning of Pt/A1,0, catalysis in Naphtha Reforming’, Ed.: B. Delmon and G.F. Forment, ‘Catalyst Deactivation

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

6. 7. 8. 9.

10. 11. 12.

13. 14. 15. 16. 17. 18.

19.

20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32.

33. 34. 35. 36.

255

1999’, Proceedings of the 8th international symposium, Brugge, Belgium, Oct. 10-13, 1999,469-472. S.J. Gentry and A. Jones, J. Appl. Chem. Biotechnol., 1978,28,727. R.A. Holroyd, K. Itoh, and M. Nishikawa, Nuclear Instruments and Methods in Physics Research A, 1997,390,233. K.B. Borisenko, B. Rozsondai, and I. Hargittai, J. Mol. Structure, 1997,406, 137. L. Colin, A. Cassuto, J.J. Ehrhardt, M.F. Ruiz-Lopez, and, D. Jamois, Appl. Surf. Sci., 1996,99,245. Ehrhardt, J.J., L. Colin, A. Accorsi, M. Kazmieczak, I. Zdanevitch, Sensors and Actuators B, 1992,7,656. Lane, G.S. and E.E. Wolf, J. Catal., 1388,113, 144. M. Walker, K.M. Baumgartner, J. Feichtinger, M. Kaiser, and A. Schulz, Vacuum, 2000,57,387. D.D. Beck, J.W. Sommers, C.L. DiMaggio, Appl. Catal. B, 1997,11,257. T.N. Angelidis, S.A. Sklavounos, Appl. Catal. A, 1995,133, 121. Chemical Industries, Consulting Editors HeinZ Heinmann, Berekeley California, Volume 17, Catalyst Poisoning, by L. Louis Hegedus and Robert W. McCabe, ISBN : 0-824-7173-7, MARCEL DEKKER, INC., New York and Basel. M. Kiskinova and D.W. Goodman, Surf. Sci., 1981,108,64. T.Z. Tzou and S. W. Weller, J. Catalysis, 1994,146,370. W.M.Graven, S.W. Weller, and D.L. Peters, ‘Catalytic Conversion of an Organophosphate Vapor over Pt/alumina’, I&EC Process Design and Development, Vol. 5, N0.2, April 1966. R.W. Baier, S.W. Weller, ‘Catalytic and Thermal Decomposition of Isopropyl Methyl Fluoro-Phosphonate’, I&EC Process Design and Development, Vol. 6, N0.3 July, 1967. M.J. Rokosz, A.E. Chen, C.K. Lowe-Ma, A.V. Kucherov, D. Benson, M.C. Paputa Peck, R.W. McCabe, Appl. Catal. B: Environ., 2001,33,205. H.S. Gandhi, W.B. Williamson, J.L. Bomback, Appl. Catal. 1982,3,79. S. Kim, M.. J. D’Aniello, Jr., App. Catal., 1989,56,23. S. Kim, M.J. D’Aniello, Jr., Appl. Catal., 1989,56,45. D.R. Liu and J.S. Park, Appl. Catal. B: Environ, 1993,2,49. L.L. Hegedus, K. Baron, J. Catal., 1978,54,115. B. Angele, K. Krishner, Chem. Eng. Sci., 1980,35, 2089. W. Williamson, J. Perry, H.S. Ghandi and J.L. Bamback, Appl. Catal., 1985, 15, 277. B. Angele, K. Kirchner, Chem. Eng. Sci. 1980,35,2101. J. B. Butt, J. J. Spivey and S. K. Agrawal, Studies in Surface Science and Catalysis (1994),88(Catalyst Deactivation 1994),pp. 19-31. S. K. Agrawal, J. J. Spivey and J. B. Butt, Appl. Catal., 1992,81(2), 239. S. K. Agrawal, J. J. Spivey, G. B. Howe, J. B.Butt and E. Marchand, Studies in Surface Science and Catalysis (1991),68(Catal. Deact. 1991),pp. 475-8. J.J. Yu and J. B. Butt, Studies in Surface Science and Catalysis (1991), 68(Catal. Deact. 1991),pp. 343-50. V.K. Shum, W.M.H. Sachtler and J.B. Butt, Industrial & Engineering Chemistry Research, 1987,26(7), 1280. H. H.Lee, and J. B.Butt, AIChE Journal, 1982,28(3), 405. H.H. Lee, and J.B. Butt, AIChE Journal,1982,28(3), 410. D.M. Downing, J.W. Lee, J.B. Butt, AIChE Journal, 1979,25(3), 461.

256 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 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.

Catalysis

C.H. Bartholomew, Appl. Catal. A: General, 2001,212(1-2), 17. C.H. Bartholomew, Appl. Catal., 1984,91(23), 96. G.F. Froment, Appl. Catal. A, 2001,212,117. J.-W. Snoeck, G.F. Froment and M. Fowlest, J. Catal., 1997,169,240. J.-W. Snoeck, G.F. Froment and M. Fowlest, J. Catal., 1997,169,250. G.F. Froment, Catal. Today, 1999,52, 153. L.L. Hegedus, and J.C.Cavendish, Ind. Eng. Chem. Fund. 1977,16,356. L.L. Hegedus and J.C.Summers, J. Catal., 1977,48,345. L.L. Hegedus, J.C.Summers, J.C.Schlatter, and K.Baron, J. Catal., 1979,56, 321. E.E. Petersen, Studies in Surface Science and Catalysis (1997), 11 l(Cata1yst Deactivation 1997), pp. 87-98. M.A. Pacheco, and E.E. Petersen, J. Catal., 1986,98(2), 380. E.E. Petersen and M.A. Pacheco, ACS Symposium Series (1984),237(Chem. Catal. React. Model.), pp. 363-74. L. W. Jossens and E.E. Petersen, J. Catal., 1982,73(2), 366. B. Angele, K. Kirchnerand, E.G. Schlosser, Chem. Eng. Sci., 1980,35,2093. J.B. Butt, E.E. Petersen, ‘Activation, deactivation, and poisoning of catalysts’, Academic Press Inc., (1988). G. F. Froment, The modeling of catalyst deactivation by coke formation, in ‘Catalyst Deactivation 1991’ (C.H. Bartholomew and J.B. Butt, Eds.), pp. 53-83 (1991). G. F. Froment, K. B. Bischoff, ‘Chemical reactor analysis and design’, 2nd Ed., John Wiley & Sons, (1990) S., Arbabi and M. Sahimi, Chem. Eng. Sci., 1991,46,1739. S. Arbabi and M. Sahimi, Chem. Eng. Sci., 1991,46,1749. G.B. DeLancy, Chem. Eng. Sci., 1973,28,105 M. Morbidelli, A. Gavriilidis and A. Varma, ‘Catalyst design, optimal distribution of catalyst in pellets, reactors, and membranes’, Cambridge University Press, Cambridge, UK (2001). F. J. Keil, Chem. Eng. Sci., 1996,51(10), 1543. K.B. Bischoff, Ind. Eng. Chem. Fund., 1969,8,665. E. Wolf, and E.E. Petersen, Chem. Eng. Sci., 1974,29, 1500. E. Wolf, and E.E. Petersen, Chem. Eng. Sci., 1977,32,493. J. C. Gottifregi, and G. F. Froment, Chem. Eng. Sci., 1997,52(12), 1883. R. Aris, ‘The mathematical theory of diffusion and reaction in permeable catalysts’, Clarendon Press, (2 vols.), 1975. S. E. Voltz, C. R. Morgan, D. Liderman and S. M. Jacob, Ind. Eng. Chem., Prod. Res. Dev., 1973,12,294. L. Hiam, H. Wise and S. Chaikin, J. Cata1.,1968,10,272. J.H. Olson, Ind. Eng. Chem. Fund., 1968,7, 185. K.B. Bischoff, Ind. Eng. Chem. Fund., 1969,8,665. T. Vermeulen, Advan. Chem. Eng., 1958,2, 148. E.G. Schlosser, Chem. Ing. Technik., 1975,47,997. H.W. Jr. Haynes, Chem. Eng. Sci., 1970,25,1615. H. Brauer and H. Schlueter, Chem. Ing. Technik., 1965,37,1107. T. Bacaros, S. Bebelis, S. Pavlou and C.G. Vayenas, Optimal catalyst distribution in pellets with shell progressive poisoning: the case of linear kinetics, in ‘Catalyst Deactivation 1987’ (B. Delmon and G.F. Forment, Eds.), pp. 459-468,1987. A. Brunovska, M. Morbidelli and P. Brunovsky, Chem. Eng. Sci., 1990,45,917. P. Pranda and A. Brunovska, Chem. Eng. Sci.,1993,48(19),3423.

5: Deactivation of Oxidation Catalysts for VOC Abatement by Si and P Compounds

75. 76. 77.

257

P. Papageorgious,D. M. Price, A. Gavriilidis and A. Varma, J. Cata1.,1996,58,439. S. Pavlou and C.G. Vayenas, Chem. Eng. Sci., 1990,45,695. J. C. Summers and L. L. Hegedus, J. Catal., 1978,51, 185.

6 Microemulsion: an Alternative Route to Preparing Supported Catalysts BY S. ROJAS, S. ERIKSSON AND M. BOUTONNET

1

Introduction

A supported catalyst consists of one or several active component(s) deposited on a solid carrier with the aim both to achieve an optimal dispersion and to prevent sintering of the active phase. The preparation of supported catalysts is a complex process. Several aspects should be taken into account in order to obtain the appropriate catalyst for a given process, or in other words, to design a catalyst. A general procedure for catalyst preparation should be ruled out, since for every particular application different catalytic properties might be desirable. The physical and chemical properties of a catalyst, which may be related to the preparation procedure, will determine its catalytic performance’. As a general trend, a supported catalyst consists of an active component, in general one or several metal(s), deposited on a support, which as a rule should display a high surface area. In most cases, the metal phase has to be activated either previously or in situ, to obtain the active catalyst. In the best case, a highly dispersed catalyst is obtained. It is also desirable that the catalyst display a narrow particle size distribution, especially for structure-sensitive reactions, in which the final product distribution varies with the catalyst particle size. Impregnation, ion exchange, anchoring, grafting and heterogenization of complexes are among the most used methods for preparing heterogeneous catalysts. Recently, several reviews and books dealing with the description of such methods have In this paper we review several aspects of an alternative, although well established, method for the preparation of supported catalysts, the so-called microemulsion method. One of the major advantages of the microemulsion technique is that it allows the preparation of a metal-based catalyst displaying a narrow particle size distribution. This colloidal method allows the synthesis of catalytically active metal particles without any effect of a support during the particle formation giving particles with well-defined morphology. Some advantages of the microemulsion technique compared with the traditional method of impregnation in the case of metal catalyst preparation are reported in Table 1. Catalysis, Volume 17 0The Royal Society of Chemistry, 2004 258

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

Table 1

259

Preparation of metal catalyst particles: Microemulsion technique versus impregnation technique

Microemulsion method

No calcinations Low reduction temperature, ambient Bimetallic particles are easily formed in microemulsion. Some types of bimetallic (alloy) particles can be obtained at low reduction temperature Metal loading low due to the limited capacity of the microemulsion Precise control of the metal microstructure: size, composition, elemental distribution of each bimetallic particle The preparation is carried out in absence of support The reduction of the metal precursor take place in absence of the support The size of the particles is determined by the size of the micelles

Impregnation technique (Incipient wetness method) 0

0

High calcination at temperature (200-600°C) High reduction temperature (200-600°C)for both mono and bimetallic particles

Metal loading can be varied within a wide range No control of metal structure

The support influences the formation of the metal particles The size of the particles is determined by the amount of precursor in the solution and the pore size of the support

We will describe the current state-of-the-art of the microemulsion method for the preparation of metal-based catalysts. First, some general considerations concerning the nature of a microemulsion and its relation to the preparation of particles will be given. Then, both the preparation of solid oxides and metalsupported catalysts by microemulsion will be detailed. When possible, the properties of the solids prepared by microemulsion will be compared with those of their counterparts prepared by traditional techniques. Particular attention will be paid to the description of the catalytic properties of these solids. There is a large body of work in the field of organic synthesis, and enzyme catalysis in which microemulsion techniques play an important role. However this topic is not included in this paper, for that purpose several reviews are available, see for example those by Holmberg and Lawrence & Rees 435.

2

Formation of Nanoparticles in Microemulsions

It is not our purpose to get into a very detailed description of the properties of microemulsions, several reviews deal with this to pi^^,^. This chapter deals with the general description of a microemulsion, focusing on the properties that could affect the synthesis of a catalyst.

2.1 What is a Microemulsion? - As shown in Table 2, microemulsions, as well as emulsions, are colloidal systems. An emulsion is a thermodynamically unstable suspension of liquid droplets in a second immiscible liquid.

Catalysis

Table 2 Foams Xerogels Emulsion Dispersion Aerosols Smokes

Colloidal systems gas in liquid solid foams, gas in solid liquid in liquid solid in liquid liquid in gas liquid in gas

The term microemulsion was initially introduced in 1959 by Schulman who suggested the following definition: A microemulsion is formed on addition of an aliphatic alcohol (co-surfactant) to an ordinary emulsion'. Generally, microemulsions are defined as thermodynamically stable homogeneous mixtures of oil and water stabilised by surfactants and, in some cases, by co-surfactants. Thus, a water in oil microemulsion, i.e. a reversed micellar solution, is a transparent, isotropic and thermodynamically stable fluid in which nanometer-sized water droplets are dispersed in a continuous oil phase. Three factors distinguish a microemulsion from an emulsion: (i) the transparency, as the microemulsion is an optically isotropic solution, (ii) the thermodynamic stability of a microemulsion and the (iii) heterogeneity at the molecular level with droplets of the size 60-800 A (micelles). 2.2 Structure of Microemulsions. - A microemulsion can exhibit different phases depending on the temperature as well as the nature of the surfactant, co-surfactant, oil phase, and their relative concentrations': (i) Two phases, the microemulsion in equilibrium with the oil phase, (ii) Two phases, the microemulsion in equilibrium with the water phase, (iii)Three phases, the microemulsion in equilibrium with both oil and water phases, (iv) Single phase, oil, water and surfactant being homogeneously mixed. Once a microemulsion is formed, depending on the concentration of the different components, the mixture will consist of (i) spherical micelles either of an oil in water microemulsion, with aggregates called micelles, or a water in oil microemulsion with reversed micelle aggregates, (ii) rod-like micelles or bicontinuous phaseslo''l. Recently, Hellweg has reviewed the state-of-the-art of this topic'*. For the use of ionic type surfactants, particularly AOT, a recent review by Nave et a1 is re~ommended'~. Interconversion among different phases may be achieved by modifying the proportions of the different components. Figure 1 illustrates the phase transition as well as the influence of temperature on the stability of the system. For a given concentration of surfactant in the mixture, at a high water to oil ratio (figure l),a dispersion of small oil droplets surrounded by the surfactant's molecules (micelles) in water is found. The decrease of the water to oil ratio (moving from the left to the right side of the x-axis in figure 1) will induce the formation of a bi-continuous phase where water is bound to the hydrophilic group of the surfactant via hydrogen bonding. At a low water to oil ratio, reversed micelles are formed. A temperature increase will destroy the oil droplets while the water droplets will be destabilized by a temperature decrease. Although

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

26 1

T

t Micelles

Reversed Micelles

\ 100 %

100 9%

Water

Oil Bicontinuous phase

Figure 1

Schematic representation of transitionfiom W/Oto 01W by way of continuous structure for ternary system based on non-ionic surfactant. Injuence of temperature adapted from reference 1 1

not negligible, the effect of pressure is not considered to be of great importance in the behaviour of a microemulsion system.

2.3 Microemulsions as Synthesis Medium. - Since this paper deals with the synthesis of metal-based catalysts, we will focus on the description of water in oil microemulsions which have been recognised as suitable systems for the preparation of nanoparticles. Reversed micelles are defined as small aggregates formed by surfactant molecules surrounding a water core of a well-defined nanometer size. Potentially, the water droplets in the microemulsion may be considered as reaction media, microreactors, for the synthesis of particles. Reactants are confined within such dispersed droplets, provided that water-soluble precursors are used14. This structure has been shown to be the most suitable for the preparation of fine inorganic colloidal particles due to the size of the aggregates and their monodispersity and also due to the fact that most of the metal precursors are water-soluble. The water core of the reversed micelles will host the hydrophilic metal salts that will turn, after a reduction step, into metallic particles. The relation between the typical constituents of a microemulsion, oil, water and surfactant, may be represented by a ternary system like the one displayed in Figure 2. For that particular system, the area of microemulsion where the reversed micelles are formed is shown. The extension of the reversed micelle region within the ternary system depends, among other factors, on the nature of the surfactant (ionic, cationic, and non-ionic) and the solvent. 2.4 Some Relevant Aspects of Microemulsions for Particle Preparation. - The synthesis of the metal particles may be carried out in two different manners?

Catalysis

262

Hewdecane 100

-

100 CizE wt%

Figure2

50

\

#$$f

100

a) Phase diagram at 25°C of the ternary system hexadecane, water and Pentaethyleneglycoldodecylether ( C I 2 E 5 )b) . Reversed micelles with Pt4+ ions in the water core

either by mixing two reversed-micelle solutions, one containing the metal precursor and the other one containing the reducing (or precipitating) agent, or by adding a reducing agent such as hydrazine directly to the microemulsion containing the metal precursor. Figure 3 illustrates the preparation procedure in the case of Pd particles. Even though microemulsions are considered to be stable, they are dynamic systems in which droplets are involved in continuous collisions with each other. Some of those collisions result in coalesced drops that trend to break up since they are not thermodynamically stable. For further details see Li & Park and Agrell et al. and references therein15,16. As the particle formation will start in the interior of the droplet, its structure and its ability to exchange the micellar-containing material will influence the nature of the formed colloidal particle^'^^'^. An important feature of the preparation of metal-based catalysts is the size of the metal particles, which in turn depends on the size of the water droplets; water droplet size is influenced by the nature of the surfactant and the water content. Generally, a low water to surfactant ratio is required for the formation of reversed micelles, this value depending on the nature of the surfactant, i.e. the number and length of the hydrophilic chains. For a given surfactant, changes in the water to surfactant ratio, o,will give aggregates of different size and shape (spherical micelles, rod-like micelles and others). Consequently, this would give rise to metal particles displaying a wide range of size and shapes. This was shown, for instance, for the preparation of Cu particles formed in an Aerosol OT (AOT), iso-octane and water rnicr~emulsion'~ where the size of the Cu particles increases from 2 to 10 nm as the o ratio increases from 1 to 10. However, for the Ni2+-AOT/microemulsion system, a rod-like structure has been reported by Hellweg using light scattering study2'. Table 3 shows the influence of the surfactant, oil phase and reducing agent on

263

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

+

1. Surfactant molecules in solvent

2. Addition of Pd(NH3)Klz aq

3. Addition of hydrazine

1

Figure 3

Procedure of metal particle preparation from microemulsion. The obtained Pd particles displayed an average size of 7 nm

Table 3

Influence of several parameters on the particle size of Pd particles prepared by microemulsion. A comparison with the analogous particles prepared by impregnation is also given

Metal precursor

Particle size (nm)

PdC12

6-7 12 5 3.8-2.4 5

~

PdCl2 K2PdC14 PdC12 Pd(NH3)4 PdCl2 PdC12

5-8 3-10 12 5 3 and 11

Reducing agent

Surfactant Oil phase ~~~~~

Preparation method Reference ~~

NP-5

Cyclohexane Hydrazine

PEGDE AOT AOT AOT NP-5

Hexane n-Heptane Iso-octane Iso-octane Cyclohexane

Maripal 013/40

Cyclohexane NaH2P02

Hydrazine Hydrazine Hydrazine Hydrazine Hydrazine

~

ME IM ME ME ME ME ME IM ME HM

~~

85 21 119 120 15 96 121

the size of the Pd metal particles. The size of the Pd particles increases with the o ratio. At constant amount of water and reactants, i.e. metal salts, increasing the surfactant concentration will increase the number of reversed micelle aggregates. This means that the number of metal salt molecules per micelle will be lower and consequently the number of metal particles formed will increase. This was shown

264

Catalysis

for Cu particles where the size of the particles decreases from 5 to 3 nm when the concentration of the surfactant increases from 5.10-2to lo-' M19. Similar results could be obtained in the case of Pd particles16. The properties of a microemulsion will to a great extent depend on the nature of the surfactant. Surfactants may be non-ionic, anionic or cationic. Previous studies have shown that a suitable system for the preparation of metallic nanoparticles consists of a non-ionic surfactant such as pentaethyleneglycoldodecylether, hexane and water (Figure 2). A water-soluble precursor can be added to the system and thus, a reasonable amount of nanosized metal particles may be obtained. In some particular cases, systems based on ionic surfactants such as AOT or cationic surfactants such as cetyltriammonium bromide (CTAB) will give a lower solubility of the metal precursor2'. The nature of the oil phase is important in order to achieve a microemulsion system since the solubility of the water droplets is related to the nature of the oil phase. Also, the metal-reduction step may be influenced by the nature of the oil phase. In the case of metal particle preparation the choice of the metal precursor is of paramount importance. Obviously, water-soluble precursors are desired, generally transition metal salts, but even then different behaviours may be expected from different precursors. From Table 4 it can be observed that the solubility of chloroplatinic acid in a microemulsion is seven times higher than that of rhodium, iridium and palladium c h l o r i d e ~ ~ l - ~ ~ . The nature of the reducing agent may also play an important role in the properties of the synthesised particles. Among others, HZ, hydrazine and NaBH4 have been used in the preparation of metal particles, hydrazine oftenbeing the most suitable. For supported catalysts, the nature of the carrier plays an important role in order to obtain a well-dispersed catalyst. Oxide solids, either catalysts or supports, can be prepared by the microemulsion technique. The nature of the precursor and that of the hydrolysis agent determine the final properties of the materials. So far, we have just named several aspects that somehow may influence the nature of the catalysts. A more detailed description of the influence of those parameters, illustrated with some examples, is given in the following chapters.

Table 4

The solubility of diflerent transition metal salts in a microemulsion of surfactant (Pentaethyleneglycoldodecylether(PEGDE), CI2ES), iso-octane and water

Microemulsion composition

Concentration wt%

n-Hexane PEGDE Water

88 10 2

Metal salt(s)

Concentration of metal salt in the microemulsion mmolelkg

HZPtCI, IrC13 RhC13 5.6 PdC12 5.6

42 6 5.6 5.6

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

3

265

Metal Oxides by Microemulsion

3.1 Introduction. - Oxides are suitable as catalysts in redox as well as acidbase reactions since they are able to participate in the exchange of electrons, protons or oxide ions24.Oxides of the main group elements are used as acid-base catalysts and carriers whereas transition metal oxides often serve as redox catalysts and precursors for preparation of different active phases. One of the most important applications for oxide catalysts is in oxidation reactions. The use of nanoscale materials in the field of catalysis is of great interest since several material properties change significantly when the particle size reaches the submicron range. Attractive properties such as large surface area to volume ratio, high thermal stability and increased mechanical strength can be achieved. The preparation methods utilized in the production of nanomaterials include evaporation in an inert gas atmosphere, sputtering techniques, laser pyrolysis, plasma techniques, coprecipitation and the sol-gel method2’. In some cases, reverse microemulsion systems have also been used as reaction media for the synthesis process. In this chapter, the state-of-the-art of metal oxide preparation by the microemulsion technique is presented. The chapter is divided in two main parts. The first section treats the preparation of catalytically active metal oxides while the second part deals with the preparation of metal oxides mainly used as catalytic support materials.

3.2 Catalytic Oxide Materials. - 3.2.1 Ceria and Zirconia. Cerium oxide is a major component in today’s three-way exhaust catalysts (TWCs) due to its high oxygen storage capacity (OSC)26.This property is linked to the ability of Ce02to undergo rapid reduction/oxidation cycles and results in a higher conversion efficiency and resistance to thermal ageing. Masui et al. have prepared ultrafine particles of Ce02in a reverse microemulsion system27.By varying the concentration of the reactants cerium nitrate and ammonium hydroxide, the size of the particles in the range under 5 nm could easily be controlled. The catalytic activity of Ce02/A1203for the oxidation of CO was tested for catalysts prepared both by the microemulsion technique and by the coprecipitation method. The results presented in this work show that the catalysts prepared by the microemulsion method had a higher activity for CO oxidation when compared to catalysts prepared by the coprecipitation technique, despite the fact that equal surface areas are obtained by both preparation procedures. Masui and co-workers do not offer a clear explanation for this behaviour. Effects such as small size, morphology and high OSC are suggested to have a positive effect on the activity. The OSC effectiveness of CeO2 can be significantly improved by the addition of Zr0226.The preparation procedure for the mixed oxides is important since it affects the surface area, homogeneity and phase formation of the material2*.Also, it is well known that a high surface area is essential to achieve a high dispersion of active material, i.e. Pt and Rh, on the metal oxide. The influence of zirconium on Ce02has been investigated by Martinez-Arias

266

Catalysis

et al.29-31. Particles with a narrow size distribution and a moderate degree of composition heterogeneity were prepared. The material composition obtained was CeXZrl-,O2with x close to 0.5, with the crystalline material corresponding mainly to phase t". These properties are favourable since they will result in the highest OSC. However, according to literature, this composition should result in the formation of phase t' and not t". The authors suggest that this behaviour can occur because of the formation of extremely small particles. The mixed metal oxide exhibited a higher surface redox activity than both of the single oxides Ce02and Z r 0 2prepared by the same method. Furthermore, a high surface area, favouring the dispersion of active material, was obtained. Anderson and co-workers have prepared zirconia deposited on alumina by using the microemulsion technique32.Two microemulsions were mixed, one containing the metal precursor and the other one containing the precipitating agent, resulting in metal oxalate particle formation. To obtain the desired metal oxide, the metal oxalate particles were calcined at 350 "C. The material was used as a FCC catalyst additive for sulphur-compound cracking with the ability to reduce sulphur in gasoline. The zirconia-containing catalyst was not found to increase the yield of gasoline when compared to a reference catalyst. However, the catalyst showed a high sulphur-reducing ability. 3.2.2 Hexaaluminates. One of the most promising groups of materials for use as high-temperature combustion catalysts is h e x a a l ~ m i n a t e s ~These ~ . ~ ~ .materials are suitable as catalysts because of their resistance to sintering and their catalytic activity. Hexaaluminates have the general formula ABxA112-x019where A could be an alkali, alkaline earth or a rare earth metal and B could be a metal with similar size and charge as the A1 ion. The advantageous properties of the material are related to their crystal structure. These structures consist of spinel-structured blocks separated by mirror planes in which the large A1 ions are po~itioned~~. Barium hexaaluminate (BHA) has been prepared by a reverse microemulsionmediated sol-gel m e t h ~ d ~Several ~ . ~ ~ important . preparation parameters were investigated in this work. Nanoparticles with excellent thermal stability could be obtained under optimal preparation conditions when compared to conventional sol-gel derived materials. This stability improvement is believed to occur since crystallization to the desired hexaaluminate phase took place at a relatively low temperature. Furthermore, this material has been tested as a catalyst for methane c o m b u ~ t i o nThe ~ ~ . light-off temperature for 1 vol.% CH4 in air when using pure BHA as catalyst was observed to be 590 "C. This temperature could be lowered to 400 "C by depositing Ce02 on the catalyst.

-

3.2.3 Perovskites. Perovskites are a group of promising catalytic materials with the general formula ABO3, where A is a large cation. Properties such as high thermal stability and catalytic activity make these materials potential catalysts for high-temperature combustion reactions34. Gan et al. have successfully prepared single-phase perovskite-type LaNi03, La2Cu04and BaPb03 by the microemulsion technique3*.Nanosize La-Ni, LaCu and Ba-Pb oxalates with molar ratios of 1:1,2:1, and 1:1, respectively, could

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

267

be achieved. The required calcination temperature for obtaining single-phase perovskites was in general 100-250 "Clower when compared to metal oxalates prepared by the conventional precipitation method. Herrig and Hempelmann have prepared BaTi03, BaZr03,SrTi03and SrZr03 perovskites by hydrolysis of the corresponding metal alkoxides in a microemulsion system3'. A strong relationship between the hydrolysis rate and water content of the microemulsion was reported. 3.2.4 Spinels. Spinels are a group of materials with relevant properties in heterogeneous catalysis. They have the general formula AB204 where A represents a divalent metal ion and B corresponds to a trivalent metal ion. MgA1204 can for example be used as catalyst for reducing sulphur dioxide emissions or as catalytic support materiala. Nanosized spinels of type MA1204(M = Mg, Co, Ni, Cu) have been prepared by Meyer and co-workers4'. The materials were prepared in a w/o microemulsion by using heterobimetallic alkoxides, M[A1(OPt)4]2, as precursors. According to the results, by using a single-source precursor instead of stoichiometric mixtures of individual alkoxides a higher morphological purity of the spinel is obtained. Furthermore, the particle size is found to depend on the hydrophilic chain length. By increasing this length, the particle size could be decreased. The most relevant results on metal oxides have been summarized in Table 5.

3.3 Oxide Materials. - 3.3.1 Silica. Several research groups have investigated the preparation of Si02 by the microemulsion technique. The overall reaction resulting in the formation of silica can be represented as Si(OEt)4

+ 2H20

-

SiOz

+ 4EtOH

Arriagada and Osseo-Asare have compared the base-catalysed hydrolysis of tetraethoxysilane (TEOS) in microemulsion systems using both non-ionic and anionic surf act ant^^^'^^. The water to surfactant molar ratio (w) had a great effect on the silica nanoparticles prepared in both systems. The effect of ethanol on the stability of the microemulsion was investigated for the non-ionic system since this substance is produced by the hydrolysis reaction. The authors discovered that ethanol partitions preferentially either to the water pools or to the continuous oil phase depending on the value of w. The stability of the system is furthermore complicated as water is consumed during the reaction, therefore the value of w will vary with time. Additionally, the role of water as a stabilizing agent for the nanoparticles is recognized. Van Blaaderen and Kentgens have used the same system for preparation of SiOzU. The water to surfactant molar ratio was the main variable studied for the anionic system. The water molecules were found to bind strongly to the surfactant polar groups and the sodium counter ions at w values below 4. This behaviour resulted in an inhibition of TEOS h y d r ~ l y s i sAs ~ ~w . was increased from 5 to 9.5, the particle size increased and the size distribution decreased. Furthermore, the size distribution of the particles prepared by the anionic system was broader than for the non-ionic system.

0.9 20-50

CDBA/benzene/water

Triton X- lOO/hexanol/cyclohexane/ water Triton X- 1OO/hexanol/cyclohexane/ water AOT/cyclohexane/water

tetraethoxysilane (TEOS)

tetraethoxysilane (TEOS)

sodium orthosilicat or sodium metasilicate tetraethoxysilane (TEOS)

titanium tetrachloride

titanium tetrachloride

TiOC1,

Tetraisopropyl titanate

Tetrabutyl titanate

titanium tetrachloride

Tetrabut yl ti tana te

titanium n-butoxide, tetraet hoxysilane (TEOS)

Si02

Si02

Si02

Si02

Ti02

Ti02

Ti02

Ti02

Ti02

TiOz

Ti02

Ti02 on Si02

Support material

< 100

120 x 20 (needle shape) 50

Triton X- lOO/hexanol/cyclohexane/ water Triton X-45/cyclohexane/water

N

5 -

-

< 10

Span-80/cyclohexane/iso-pentanol/ water NP5 + NP9 (l:l)/cyclohexane/water

8.4e

6.23d

Support material

350-4Wb

Support material, photocatalyst Support material, photocatalyst Support material, photocatalyst Support material, photocatalyst Support material, photocat alyst Support material, photocatalyst Support material, pho tocat a1yst Support material, photocatalyst

Support material

Support material -

10-36

100-150

5-20

26-43

100-200

5-100

Support material

N

-

tetraethoxysilane (TEOS)

Si02

Support material

-

Catalytic use

30-70

NP-S"/cyclohexane/ammonium hydroxide/water AOT/decane/ammonium hydroxide/water NP-5"/cyclohexane/ammonium hydroxide/water NP-4"/heptane/ammonium hydroxide/water NP-9" + NP-S"/cyclohexane/ hydrochloric acid/water Atlox 4912/Isopar G/water or C 1 2 - 1 4 E 4 , 5 / i ~ ~waterc ~~tane/

tetraethoxysilane (TEOS)

Si02

Microemulsion system

Precursors

BET surface area (m' g-.')

Metal oxide

Particle diameter (nm)

Oxide materials prepared by microemulsion

Table 5

54

48

50

53

5132

47

46

49

44

45

43

42

41

40

Ref:

2

2.

%

cc

00

o\

N

zirconium oxynitrate

(a)

zirconium oxynitrate

Particle size of the metal oxalates before calcination.

Zr02

N

(b)

Support material Support material

-

Support material

-

8.W

10

The particle size varies depending on surfactant used.

Supportmaterial

-

8 500

Supportmaterial

Supportmaterial

-

65

Supportmaterial

Support material

22-41' -

Support material

50

-

-

-

-

Support material, photocatalyst Support material, photocatalyst Support material

5.9'

2 nm-10 mm

295-559

NP5 + NP9 (l:l)/cyclohexane/ammonia 5-20

NP5 + NP9 (2:l)/cyclohexane/oxalic acid zirconium n-propoxide Zr02 Berol02/cyclohexane/sulfuricacid or AOT/isooctane/water zirconyl nitrate, zirconium Span80/heptane/isopropylalcohol/ Zr02 acetate water or Arlacel83/cyclohexane/ isopropyl alcohol/water Igepal Co520/cyclohexane/ammonia zirconium n-propoxide, ZrSi04 tetraeth yl-orthosilicate Bri. 30/cyclohexane/water y203/zr02 zirconium oxynitrate AOT/propane/water aluminum nitrate Al(OH13 aluminum sec-butoxide AOT/n-heptane/water A1203

Zr02

Zr02

Zr02

from 2 titanium tetraisopropoxide TX-100, BL-2, BL-4.2 or AOT/cyclohexane or octane/water 8-140 zirconium tetra-n-butoxide NP-6/cyclohexane/sulphuric acid DOLPA/isooctane/ammonia and KC1 2 zirconium tetrabutoxide

Ti02

11.7-24.4

titanium tetraisopropoxide NP-S/cyclohexane/water

Ti02

66

65

64

63

57

62

58

59 61

60

56

55

2 2

pc

% m

L

ii

Y

%

cp

a9'

4

2

0

pc

5

Q

%I

rn

5.

k 9

a ;3

E g.

i;'

g

%

9

270

Catalysis

The kinetics of silica formation from TEOS in non-ionic reverse microemulsions has been studied by Chang and F ~ g l e rThey ~ ~ . concluded that the growth of silica particles is controlled by the rate of TEOS hydrolysis, which is found to be first order with respect to the aqueous ammonia concentration, approximately zero order with respect to water concentration and roughly one-half order with respect to the concentration of surfactant. The effect of alcohol chain length on particle size has been investigated by Esquena et al.46.They conclude that a longer alcohol chain will result in smaller silica particles. Gan et al. have used the precursors sodium orthosilicate and sodium metasilicate in the preparation of silica nanoparticles by hydrolysis in reverse micr~emulsions~~. The particle size was found to be affected by the pH and the precursor concentration. The pH also influenced the surface area; a lower pH resulted in a higher surface area. 3.3.2 Titania. Titania powders are used as pigments, catalytic supports, membranes, opacifiers, photocatalysts and fillers in industrial application^^^.^^. Titania particles have been prepared by a number of methods, such as hydrolysis, sol-gel, microemulsion and hydrothermal synthesis5'. Titania exists naturally in two tetragonal forms, the metastable phase anatase, and the stable phase rutile. On heat treatment, anatase transforms into rutile. The phase transition temperature depends on the starting materials and the preparation procedure. The effect of calcination temperature on the phase and particle size of Ti02has been investigated by Chhabra et The particles were obtained from a microemulsion consisting of water/Triton X-lOO/hexanol/cyclohexane and the catalytic activity for the photodegradation of phenol was tested. As a catalyst, only the anatase form of Ti02showed significant degradation of phenol, whereas the rutile form was completely inactive for this reaction. Kim and Hahn have utilized the same microemulsion system for the production of ultrafine t i t a r ~ i a ~ ~ . They discovered that higher concentrations of reactants promoted the transition of anatase to rutile by lowering the phase transition temperature. Joselevich and Willner produced ultra-small titania particles by hydrolysis of TiC14 in a nonionic w/o microem~lsion~~. The photosensitization of Ti02 was studied and a kinetic analysis showed that larger T i 0 2clusters are formed when the size of the water droplets are increased. Li and Wang also used this precursor to synthesise Ti02.The results show that anatase forms at temperatures from 200 to 750 "C and the phase transition from anatase to rutile occurs at temperatures above 750 0C52. The effect of several important preparation parameters on the size and stability of Ti02 particles prepared by hydrolysis of tetraisopropyltitanate has been A dramatic increase in the rate of particle aggregaobserved by Moran et a1.53-54. tion as the water to surfactant mole ratio increases was reported. Furthermore, the preparation of titania from tetrabutyltitanate by the reverse microemulsion method has been reported by Ju and co-worked5. Results from TEM analysis reveal an increase in particle size when o increased from 13.9 to 55.5 and then a particle size decrease when o increased from 55.5 to 110.9. TiOz has been prepared from the same precursor by Wu et al.50and Fu and Q ~ t u b u d d i n ~ ~ .

6: M icroemulsion:an Alternative Route to Preparing Supported Catalysts

27 1

Additionally, nanosized titania particles have been synthesized by hydrolysis of titanium tetraisopropoxide in a reverse r n i c r ~ e r n u l s i o n ~ ~ ~ ~ ~ . 3.3.3 Zirconia.Zirconia powder is useful in several advanced industrial applications due to special properties including low thermal conductivity, high fracture toughness, high mechanical strength and relatively high thermal expansion coefficient59.Nanosized zirconia has been prepared by various methods, such as co-precipitation, sol-gel, hydrolysis, thermal decomposition and hydrothermal processing@. Zirconia has been synthesized by hydrolysis of zirconium tetrabutoxide in the water pools of reverse microemulsions61.62. Kawai et al. investigated the relationship between particle formation and the solubilized states of water in the reaction system. The results indicate that monodisperse, spherical particles are more easily obtained in reverse and swollen micelles than in a w/o microemulsion. Crystalline Z r 0 2particles were achieved after calcination at 800 "C. The use of zirconium oxynitrate as precursor is also reported in the l i t e r a t ~ r eBy ~ ~reac~~~. ting a microemulsion containing this precursor with a microemulsion containing aqueous ammonia as water phase, the calcination temperature for obtaining crystalline zirconia could be reduced to 362 "C. The preparation of sulphated zirconia designed for catalyst supports was studied by Boutonnet et a1.64.Zirconia prepared in microemulsion showed a pure tetragonal structure compared with zirconia prepared by an impregnation precipitation procedure which also contained monoclinic phase. Platinum-promoted sulphated zirconia catalysts were prepared both in anionic and non-ionic microemulsions. Furthermore, the catalytic activity and selectivity for the isomerization of hexanes were tested. The catalysts produced by the microemulsion method showed a higher selectivity towards isomers but a lower activity when compared to catalysts prepared by impregnation technique. More recently, a study of zirconia synthesis from micro and macroemulsion systems has been c ~ n d u c t e dSpherical ~~. Z r 0 2particles ranging from tens of nanometers to a few micrometers were produced. Additionally, some mixed metal oxide materials containing zirconium have been prepared by using the microemulsion method. Zircon powder, ZrSi04,has been synthesized by using a microemulsion-mediated process6'. A stoichiometric composition of the particles was obtained after suitable pretreatment of the precursor metal alkoxides. The need for this treatment is probably due to the differences in hydrolysis rate of the metal alkoxides. Burgard and co-workers have prepared yttria-stabilized zirconia, Y203/Zr02,in a reverse microemulsiod6.An agglomerate-free amorphous powder with a particle size of 8 nm was obtained after bubbling ammonia through a microemulsion containing ZrO(N03)2and Y(N03)3 in the aqueous phase. 3.3.4 Alumina. Although alumina is one of the most common ceramics used in catalysis as support material, the preparation of alumina by the microemulsion technique has not been investigated to a great extent. The preparation of spherical Al(OH)3particles in the cores of reverse micelles

272

Catalysis

is reported by Matson et al.67.Particles of approximately 0.5 pm were formed by adding ammonia to a microemulsion consisting of aqueous Al(N03)3dispersed in a supercritical continuous propane phase. Byoun et al. were able to obtain narrow-sized alumina particles of 10 nm by hydrolysis of aluminium alkoxide in an anionic microemulsion6*.The powder characteristics were found to be dependent on the preparation conditions, i.e. water to surfactant molar ratio. The preparation of a mixed silica-alumina oxide by a microemulsion-mediated process has been reported69.In this work, porous particles with a spherical morphology were obtained. This material can be used as a carrier for catalysts or as a precursor of ceramic materials. The most relevant results on catalytic oxide support materials have been summarized in Table 6. 4

Metal-based Catalysts Prepared by Microemulsion

Introduction. - During the past few years, a number of papers dealing with the preparation of metallic particles from microemulsions have appeared. For example Pd, Ni, Pt, Ag and bimetallic Au/Pd were successfully prepared by this m e t h ~ d ~ 'For - ~ ~this . reason, it is of interest to report the research work done on this subject and its input on catalysis research. In this section we will focus on the description of several aspects relevant to the preparation of both catalytically active metal particles and metal-supported catalysts via the microemulsion technique. Regarding the metal supported catalysts, in some cases both the metallic particles and the support were synthesized by microemulsions. However, in general metallic particles prepared from microemulsions were deposited on commercial supports. The catalytic behaviour of these microemulsion-derived materials will be commented and, when possible, compared to catalysts obtained from traditional techniques under similar reaction conditions. Selected results concerning the study of the strong metalsupport interaction effect (SMSI) obtained with catalysts prepared by microemulsion will be detailed75.Several papers dealing with the preparation of immobilized metal particles on supports have been described although the catalytic behaviour of the solids was not studied. However, their potential catalytic ability led us to include those papers within this chapter. Also, the preparation and catalytic activity of some non-supported metal catalysts prepared from microemulsions will be described.

4.1

4.2 Unsupported Catalysts. - Most of the research work is focused on the preparation of noble metal catalysts using microemulsions in the metal particle synthesis. It is now well known that the early work of Boutonnet et a1 opened a new and simple way to preparing stable reduced noble metal particles from their salt precursors21.In their work, metallic particles of platinum, rhodium, palladium and iridium were obtained from their salt precursors, H2PtC&,RhC13,PdC12 and IrC13, respectively, in water in oil (w/o) microemulsions. For Pt and Pd, the most suitable reducing agent was hydrazine. In the case of Rh3+ and If3+the

tetraethoxysilane (TEOS) titanium tetrachloride

Si02

Tetrabutyl titanate

titanium tetrachloride

Tetrabutyl titanate

Ti02

TiOz

titanium tetrachloride

Ti02

TiOz

Ti02

TiOz

Ti02

Si02

Si02

Si02

Si02

5

120 x 20 (needle shape)

Triton X- 100/hexanol/cyclohexane/ water

<10

10-36

20-50

0.9

100-150

26-43 5-20

100-200

-5-100

30-70

Particle diameter (nm)

Span-SO/cyclohexane/iso-pentanol/ water NP5 + NP9 (l:l)/cyclohexane/water

Triton X- 1OO/hexanol/cyclohexane/ water Triton TiOC14 X- lOO/hexanol/cyclohexane/water Tetraisopropyl titanate AOT/cyclohexane/water

NP-5"/cyclohexane/ammonium hydroxide/water AOT/decane/ammonium hydroxide/water NP-5"/cyclohexane/ammonium hydroxide/water NP-4"/heptane/arnmonium hydroxide/water NP-9" + NP-5"/cyclohexane/hydrochloric acid/water Atlox 4912flsopar G/water or C12.14E4,5/isooctane/ waterc CDBA/benzene/wa ter

tetraethoxysilane (TEOS) tetraethoxysilane (TEOS) tetraethoxysilane (TEOS) tetraethoxysilane CrEOS) sodium orthosilicat or sodium metasilicate

Si02

oxide

Microemulsion system

Oxide catalyst prepared by microemulsion

Precursors

Metal

Table 6

-

-

8.4"

6.23d

< 100

350-4Wb

-

-

-

BET surface area (m2 g ')

Support material, photocatalyst Support material, photocatalyst Support material, photocatalyst Support material, photocatalyst Support material, photocatalyst Support material, photocatalyst Support material, photocatalyst

Support material

Support material Support material

Support material

Support material

Support material

Catalytic use

Ref:

$

r

c,

5

9

-

NP5 + NP9 (l:l)/cyclohexane/ammonia 5-20

zirconium oxynitrate

Support material Support material Support material

-

Support material

59

Support material -

8.P

64

Support material 22-41'

68

67

66

65

60

Support material

63

61

58 62

57

56

Re$

50

-

Support material

Support material

Support material, photocatalyst Support material, photocatalyst Support material

Support material, photocatalyst

Catalytic use

(a) Non-ionic surfactant polyoxyethylene nonylphenyl ether. (b) Prepared in acidic medium 1-alkanols and NH, were present during the reaction. (d) After calcination at 900 "C.(e) After calcination at 900 "C. (I)After calcination at 600 "C. (g) After calcination at 1400 "C.

Zr02

-

-

zirconium oxynitrate

from 2

TX-100, BL-2, BL-4.2 or AOT/cyclohexane or octane/water NP-6/cyclohexane/sulphuricacid

-8-140

11.7-24.4

NP-Slcyclohexane/water

-

5.9'

BET surface area (m2g-')

DOLPA/isooctane/ammonia and KC1 2

-50

Particle diameter (nm)

NP5 + NP9 (2:l)/cyclohexane/ 295-559 oxalic acid zirconium n-propoxide Berol02/cyclohexane/sulfuricacid ZrOz or AOT/isooctane/water ZrO, zirconyl nitrate, Span80/heptane/isopropylalcohol/water2 nm-10 mm zirconium acetate or Arlacel83/cyclohexane/ isopropyl alcohol/water 65 zirconium n-propoxide, Igepal Co520/cyclohexane/ammonia ZrSiO, tetraethyl-orthosilicate 8 Brij 30/cyclohexane/water Y203/ZrO2 zirconium oxynitrate aluminum nitrate AOT/propane/water 500 AWW3 aluminum sec-butoxide AOT/n-heptane/water 10 A1203

ZrO,

Zr02

Zr02

Ti02

Ti02

Triton X-45/cyclohexane/water

Microemulsion system

titanium n-butoxide, tetraethoxysilane (TEOS) titanium tetraisopropoxide titanium tetraisopropoxide zirconium tetra-n-butoxide zirconium tetrabutoxide

Precursors

Metal oxide

Ti02on Si02

(cont.)

Table 6

6

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

275

reduction was only achieved in the presence of hydrogen, which had to be previously activated for the reduction of the latter. It was stated that the Pd(NH3)4C12salt is a better precursor in order to obtain Pd particles than the PdC12 salt, since it is more soluble in water. However, PdC12is a more suitable precursor for the preparation of bimetallic catalysts76.Furthermore, the reduction was straightforward in the presence of hydrazine. Hydrazine was found to be the best reducing agent since the particle size distribution was optimal when compared with the one obtained for metals reduced with NaBH4, Hz or other reducing agents. The particle size distribution was found to be f 0.5 nm in the 3.0 nm range. In subsequent studies, the catalytic activity of platinum particles prepared via microemulsion routes was described. The hydrogenation of but-1-ene was performed using Pt particles in a microemulsion-based ~uspension'~. Such particles were prepared following the microemulsion procedure described by Boutonnet et a1.21.According to that method, Pt'" particles were reduced using either hydrogen or hydrazine in excess. For the latter case, it was observed that the remaining hydrazine behaved as a poison for the catalyst. Thus, it had to be removed by adding hydrogen peroxide in order to obtain some catalytic activity. Once hydrazine was removed, the catalysts were more active than the ones reduced with H2. The effect of the microemulsion composition was established revealing that the nature of the surfactant played a fundamental role in the catalytic behaviour of the metal. It was shown that catalysts prepared from non-ionic surfactant-based microemulsions were more active than those prepared from ionic ones. On the other hand, the solvent nature did not affect the catalytic behaviour of the metallic particles. Nonetheless, the activity of the suspended catalyst was poor when compared with its supported counterpart^^^. Another example dealing with the use of a microemulsion as the reaction medium can be found in a patent issued by Tinucci & P l a t ~ n eAlthough ~~. in this patent the catalyst was not prepared by microemulsion, the reaction takes place in a microemulsion system. The catalyst, a water-soluble rhodium complex, was formed in situ. The reactants, i.e. alkenes, were also the constituents of the oil phase of the microemulsion. Conversions higher than 70 % were obtained while the structure of the microemulsion was still maintained. Beyond that conversion level, the microemulsion was destroyed, probably because the oil phase was modified during the catalytic reaction. Further examples that deal with the preparation and catalytic activity of unsupported palladium nanoparticles in microemulsion are reported by Spiro & de Jesus 80-81.

4.3 Supported Catalysts. - A key point in the preparation of heterogeneous catalysts is the process of supporting the metal particles onto a carrier surface. Under optimal conditions, a well-dispersed metal may be obtained. However, during such a step it is not unlikely that metallic particles tend to agglomerate giving rise to a catalyst with particle size larger than 200 A and a broad particle size distribution. 4.3.1

Deposition of the Metals onto the Support. Two major approaches have

276

Catalysis

been studied in order to obtain metal supported catalysts prepared via microemulsion technique, preserving the metal particle size and a narrow size distribution as the desired target. The main approach consists of the destabilization of the microemulsion by the addition of an external agent such as tetrahydrofurane (THF), acetone or others. When a solid support, typically a solid oxide or carbonaceous material, is placed in the metal-containing microemulsion, the addition of such a chemical will result in a system destabilization, leading to the deposition of the metal particles on the support. The methodology was first reported by Boutonnet et al.78.They obtained a supported metal catalyst by depositing ultra-fine particles synthesized in a w/o microemulsion on pumice. The activity of this catalyst was tested in the but-lene hydrogenation reaction and compared with the activity of similar catalysts prepared by the traditional impregnation method both from aqueous and alcoholic solutions. The particle size was found to be in the 20-30 A range for the particles prepared from microemulsions and above 200 A for the classical impregnation counterparts. The behaviour of the Pt-based catalyst was found to depend on the preparation method. However, such a correlation between the activity and the preparation method could not be established in the case of Rh and Pd-based catalysts. A study of the metal-support interaction effect has been carried out for a Pt-Ti02catalyst prepared by microemu1sion82.Isomerization and cracking reactions of 2-methylpentane and hydrogenolysis of methylcyclopentane were chosen as model reactions for studying the influence of the catalyst nature on the strong metal-support interaction (SMSI) effect. A comparison between the catalyticbehaviour of similar catalysts prepared by microemulsion and incipient wetness method,was reported. The catalysts were reduced at 200 and 390 "C since it is well known that the reduction temperature plays a predominant role in the SMSI effect83.The behaviour of thePt-Ti02microemulsion-prepared catalyst was found to be similar to that of a Pt/A1203catalyst. Also, it was found that the microemulsion-prepared catalysts require a higher temperature of reduction to induce metal-support interactions. In a different approach, both the metal particles and the oxide support precursors were synthesised in microemulsion. In this case, the synthesis of the oxide support precursors was carried out in the microemulsion medium following a procedure similar to the sol-gel method84.The preparation is carried out as the following: a microemulsion containing the catalytic precursor is mixed with a microemulsion containing the support precursor, or the support precursor is added to the microemulsion containing the reduced metal particles. Then, a hydrolysis step will lead to a metal-supported catalyst. This procedure is illustrated in Figure 4. In the best case, metal particles surrounded by a monolayer of metal oxide is the result. The methodology allows the control of the particle size irrespective of the total metal loading. This approach was reported independently by Boutonnet et al.85and Kishida et al.86387.First, the Pt particles were prepared in a microemulsion solution according to the microemulsion technique, then the support precursor titanium

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

277

(3

n-octane FF,D GE

FED GL

t

n-octane I Z D GL

' i I '

U,rAml..ri c

Figure 4

-..-.-

.. .

Preparation of Pt particles (a) and Pt-TiOz particles (b), adapted from reference 83

isopropoxide, was added to the solution. After hydrolysis, Pt-Ti02particles were obtained by precipitation. Figure 4 illustrates the preparation pathway of the Pt and Pt-Ti02 particles. The catalytic behaviour of these solids, along with PtTi02and Pt-SO2obtained by traditional impregnation methods, was studied in the crotonaldehyde hydrogenation reaction. The selectivity of the reaction was found to be dependent on the nature of the catalyst. In the case of the catalyst prepared from the latter pathway, a metal-support interaction effect was observed. However, if the catalysts were prepared by supporting the Pt particles on the T i 0 2carrier, irrespectively of the preparation method of the metal particles, no metal-support interaction could be detected. In this last case, the catalytic results were similar to those displayed by Pt-Si02 or Pt in suspension catalysts, for which such an effect was not expected. Kishida et al., have reported the preparation of Rh particles in a w/o microemulsion87where the SO2 support was also prepared in the microemulsion environment. Once the reduced particles were synthesized (using hydrazine), an ammonia solution was added to the microemulsion. Then, TEOS (tetraethylorthosilicate) was added to the particle suspension which was heated to 40 "C.A precipitate containing the metallic particles was obtained. The solid was washed with ethanol and calcined in air at 500 "C.Thus, the catalyst had to be reduced again under hydrogen. The metal particle size was studied by both CO chemisorption and TEM spectroscopy. The particle size was found to be depend-

278

Catalysis

ent on the preparation conditions, particularly the o value. The higher the o ratio, the larger the particle size, as determined by TEM analysis. The particle size waswithin the 3.3 to 4.5 nm range for o ratios of 5.0 and 15.0, respectively. The authors found an inconsistency between the particle sizes when measured by TEM or CO chemisorption, suggesting that Rh particles can be covered by the carrier. The metal particle size of catalysts prepared by microemulsion was smaller than those of analogous catalysts prepared following the traditional impregnation method. CO hydrogenation was used as test reaction for the prepared catalysts. Pt, Pd and Rh microemulsion-based catalysts displayed higher activity that their impregnation counterparts. The activity for the microemulsion-prepared catalysts seems to increase with the particle size; this was explained by taking into account that the smaller particles, obtained at low ci) ratios, were buried inside the carrier, thus lowering the accessible rhodium sites. There are more examples dealing with this effect, see for instance the work of Hanaokaet a1.88.The authors described the preparation of a Rh/Si02 catalyst both by microemulsion and impregnation techniques. Again, an inconstancy between the Rh particle size determined by CO chemisorption or measured by TEM, was reported. The catalytic behaviour of Rh/Si02catalysts was also tested in the CO hydrogenation reaction. The results showed how the activity of the microemulsion-based catalyst was higher than the activity of the impregnation catalysts, irrespective of the amount of Rh and of the particle size. However, the selectivity towards the C2+ oxygenated products was lower for the microemulsion-based catalyst compared with the selectivity shown by the traditionally prepared catalysts. During the study of the effect of the metallic content of Rh/Si02 catalysts, prepared by w/o microemulsion, in the CO hydrogenation reaction, Tag0 et al. found that the selectivity towards C2+oxygenated products increased with the rhodium content89.The selectivity was above 40 % for a catalyst with a Rh content of about 3.5 wt %. From their results, it was established that the product selectivity and the turn-over frequencies (TOF) were dependent on the rhodium content of the catalyst. This effect was neither ascribed to the particle size nor to diffusion phenomena. The catalysts displayed similar particle sizes, irrespective of the Rh content, and it was evident from the Thiele modulus that the reactions were carried out under reaction rate-controlled conditions. The remarkable effect of the metallic particle size in the CO hydrogenation reaction was evident when Fe-based catalysts were studiedg0.Several Fe/Si02 catalysts were prepared by impregnation and microemulsion techniques and their catalytic activity was tested at 260 "Cand 4.0 MPa. In spite of the general trend of Fe-based catalysts to produce hydrocarbons in the CO hydrogenation reaction", it was found that Fe catalysts prepared by microemulsion produced mainly C2+oxygenates, whereas the catalysts prepared by impregnation techniques produced mainly hydrocarbons. Table 7, adapted from Hayashi et al.90, displays the behaviour of Fe/Si02catalysts, either prepared by microemulsion or impregnation, in the CO hydrogenation reaction, as a function of the reaction temperature.

Cyclohexanol/ Cyclohexane Heptane See ref 29

Methanol production

Combustion

Pt/A1203

Cornbustion

Immobilization particles

Catalyst preparation

Pd

ZnS & Rh on Si02

Rh on Si02

N 2 0decomposition

Catalyst CO hydrogenation Catalyst phase transfer Ammonium salts Catalyst (S removal, Zn, Mn, Zr, Co A1203,Hydrotalcite cracking) Ti02

NO reduction by CO

C 0 2hydrogenation

Pt/A1203

Rh/Si02

Rh/Si02

Cyclohexane/n-Hexanol

Cyclohexane

PtPd/C Iso-octane

PEGDE/B02

Cyclohexane/n-octane

Hexanol

NP-5 CTAC

CTAC

Hexanol

H2PtC1, RuC13

2-40 Zn(N03)z/RhC13 8-4

Pd(N03)2

H2Pt C&,/RhC13,/ PdC12

See ref 29

See ref 29

35

5-13

3-5

RhC13.3H20

H2PtC1,

Zr(N03) Co(Act) Zr(oxalic)

RhC13

4-5

4.7-20

-

4-5

NP-5,10,15,20/CRhC13.3H20 3-8 lO,l5,20/0-10,15,20 Triton X-100 Ag(N03) NP-5/NP-9 RuC13.3H20 4 C~(N03)2*3H20

NP-23/NP-5

NP-5

Berol-050

See ref 29

See ref 29

Berol-02

1-Hexanol

Cyclohexane

Cyclohexane,/nhydrocarbons/n-alcohols Catalyst (NOx reduction) Heptane Hexanol

PtRu/C, Cell

Pt/C, Fuel

Pd/CeZr02/A1203 TWC See ref 29

1-Hexanol/l-butanol Cyclohexane

Catalyst CO hyd

Fe/Si02

Cu/ZnO

Pd/CeZr02/A1203 TWC

NP-5/NP-9

Cyclohexane

Electrocatalysts

PtRu/C

Precursors

NP-20/NP-lO/NP-5 Fe(N03)3TEOS

Surfactant'")

Oil phase")

Use

Particle size nm

Initial catalytic behaviour of FelSiOr catalysts in the CO conversion reaction, adapted from reference 90

Metallsupport

Table 7

98

96

111 30

89

104

110

94

96

109

105

101

100

108

102

90

106

Reference

2

s

9

Catalyst preparation, CO hydrogenation

Methanol production

Rh/Si02

Pd/Zr02, Ti02,

Cata1yst synthesis

CO hydrogenation

Cata1yst preparation

Hydrogenation

C 0 2hydrogenation

Isomerization and cracking Isomerization and hydrogenolysis Hydrogenation

Hydrogenation

Particle preparation

Fe powder

Rh/Si02

Rh/SiO

Pt/Ti02

Rh/Si02

Pt/Ti02

Pt, Rh, Pd/Pumice

Pt

Pt, Pd, Rh, Ir

Pt/&03

CO hydrogenation

Pd/Zr02

A1203

Use

(cont.)

Metallsupport

Table 7

Octane Hexadecane Hexane

n-Hexadecane

Hexadecane n-hexane

Hexadecane

Hexadecane

Cycclohexane

Octane

Cyclohexane/n-alcohol

1-Hexanol

Octane

Cyclohexene

Cyclohexane

Hexanol

NP-5

85

113

88

H,PtCI6*6H,O RhC13 PdC12 IrC13

CTAB NP-5

3-5 (")

19

77

H,PtC1,.6H,O NP-5/CTAB

3-4'"'

78

82

H2PtC16.6H2O 6 PdC12 5W (NH4)[RhClS.H20] 2-3

-

3-5/7-32(b) 87

-

1.8-6

92

112

NP-5

NP-5

Metal chlorides H2PtCI6.6HzO

NP-5

NP-5,7.5,10,15/ CTAB/CTAC/ AOT

RhC13aq

RhC13aq

CTAB

3.4/7.5(b)

1.5-

c 12E4

3-12

NP-5 FeC13-6H20 Fe(BF4)2.6H20

103

Zr(BuOk, T ~ ( B u O ) ~ Al(i-Pro)3

NP-5 PdC12

97

5-7

RuC13.3H20 PdCl2

CTAB

DDAB

~

Reference

Precursors

Sur$actant(a)

Particle size nm

E?,

k

0

00

N

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

28 1

Furthermore, higher conversions were achieved with the microemulsionbased catalysts. However, this trend was dependent on the reduction temperature; only when the catalysts were reduced above 600 "C could this effect be observed. Although not observed for the microemulsion-based catalysts, the particle size played an important role in the reaction selectivity for the impregnated catalysts. A lower amount of Cz+ oxygenates was detected for a larger particle size. The higher selectivity toward C2+ oxygenates for the smaller Fe particles was correlated with the reducibility of the Fe species. The total reduction of Fe for the microemulsion-derived catalysts could only be achieved at ahigher temperature when compared to the impregnated counterparts. Even after reduction in H2 at about 600 "C for 10h, a TPR analysis of the catalysts displayed a reduction profile ascribable to some remaining FeO species. From the TPR profiles it was clear that such species were more stable in the microemulsion-based catalysts than in the impregnated ones. The Fe 6+ species were proposed by the authors as responsible for the higher selectivity towards oxygenated products displayed by the microemulsion-prepared catalysts. However, when several supported Pd catalysts were tested in the C O hydrogenation reaction, the selectivity towards C2+oxygenates was found to be comparable for both the microemulsion and impregnation-prepared catalysts, in spite of the smaller particle size displayed by the former92. Binet et al. studied the effect of reduction temperature on the catalytic activity of Pd/Ce02 catalysts93.Pd particles were prepared by the microemulsion technique and were used as model catalysts, since they consisted of particles of the same morphology. The behaviour of the catalysts, concerning their accessibility of the metal phase towards CO adsorption, appeared to be independent of Pd loading. 4.3.2 Control of the Preparation of the Catalysts. 4.3.2.1 Efect of surfactant and oil phase. A very detailed study of the relationship between the structure of the metallic particles and the nature of the surfactant and the oil phase was reported by Hanaoka et al.94.They studied the influence of various surfactants (ionic and non-ionic) and organic solvents (n-hydrocarbons and n-alcohols of different chain lengths) on the metal particle size of several silica-supported rhodium catalysts prepared in w/o microemulsions. TEOS was used as the source of the silica support. Rh particles in the size range of 3.0-8.2nm were obtained depending on the nature of the surfactant irrespective of the Rh content of the catalyst. The control of the particle size was possible when using nonionic surfactant together with n-alcohol. It was observed that the particle size increases with increasing the carbon number of the n-alcohol and decreasing of the surfactant hydrophilic part. 4.3.2.2 Efect of the synthesis time. Hanoka et al. studied the influence of synthesis time, defined as the time from the hydrazine injection to the TEOS addition94.In nonionic surfactantlorganic solvents and anionic surfactant/organic solvents, the particle size was independent of the synthesis time; furthermore, the particle size range in such systems was very narrow. in cationic surfactant/organic solvent the rhodium particle size increased with the synthesis time.

282

Catalysis

4.3.2.3 Eflect of the metallic precursor. An important factor controlling the final particle size of the catalysts is the metallic precursor. Most papers on the subject report the use of water-soluble metallic salts as precursors. However, this is not always a straightforward option in order to obtain better catalysts. For instance, it was stated for Pt catalysts that a better control of both particle size and particle size distribution could be achieved by the preparation of Pt complexes, nanoparticles, formed in water in oil microemulsions95. 4.3.2.4 Eflect of the reducing agent. As stated above, the effect of the reducing agent could play an important role in the particle size distribution, thus when hydrazine was directly injected into the microemulsion containing the metallic precursor, the particle size was smaller than observed when other reducing agents such H2 or NaBH4 were used. This feature is detailed in a paper by Hanoka et al.96.A major role of the hydrazine is to stabilize metallic particles in the form of metal-hydrazine complexes2’preventing metallic aggregation. 4.3.2.4 Effect of the hydrolysis conditions. A detailed study of some of the parameters affecting the oxide precursor hydrolysis step was reported by Kishida et al.97.The effect of the hydrolysis conditions on the R value (the atomic ratio of surface rhodium in contact with the gas phase to total surface rhodium) was studied. Rh metallic particles were prepared in CTAB/1-hexanol w/o microemulsions from RhC13 salts and supported on SO2, obtained by the ammonia hydrolysis of TEOS precursors. The most important conclusion was that the R value, i.e. the location of the metallic particles in the support, could be customized by adjusting the hydrolysis conditions of the oxide precursor. For instance, the amount of Rh surface exposed to the gas phase increased with decreasing hydrolysis time. 4.3.2.5 Control of the textural properties. Surface area and calcination steps. Another crucial property to control during the preparation of heterogeneous catalysts is the surface area of the catalyst. Microemulsion techniques allow the control of the surface area in a wide range of values, while maintaining several parameters constant such as the average particle size and the metal loading. Kishida et al.98described the preparation of Rh/Si02catalysts by microemulsion. Rh particles were synthesised in w/o microemulsions by using hydrazine as reducing agent. TEOS was used as the support precursor. The surface area of the catalysts was controlled by controlling the rate of the hydrolysis step. A higher hydrolysis rate, i.e. the higher the NH3 or the lower the TEOS concentrations, resulted in a lower surface area. Area values between 60 and 600 m2/g were reported for catalysts maintaining both the Rh content and the particle size constant at values of about 1.5 wt% and 4-5 nm, respectively. Even when a well dispersed catalyst has been obtained after the deposition of the metal particles onto the support, metallic particles may be covered by organic residues, arising mainly from the surfactant molecules. This will lead to a non-accessible catalytic surface resulting in a decrease in the catalytic activity. This poisoning effect was illustrated during the preparation of a Pt/A1203catalyst9’ from a w/o microemulsion containing H2PtCls.6H20 and reduced by hydrazine. Metal particles were transferred onto the support by adding A1203to the suspension and destabilizing the system with THF. TEM analysis displayed

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

283

a narrow particle size distribution showing a maximum around 2 nm. The behaviour of the catalyst in the methylcyclopentane hydrogenolysis and 2methylpentane isomerization reactions was compared with that of Pt/A1203 catalysts prepared by impregnation technique. The catalytic activity of the microemulsion catalyst was lower than the one obtained with the catalysts prepared by impregnation and poorly dispersed. This feature led the authors to propose a non-selective poisoning effect by the presence of some remaining surfactant molecules blocking the active centres of the microemulsion-based catalyst. Obviously, it is necessary to remove the carbon residues from the catalysts prepared by microemulsion. This is usually achieved by submitting the catalysts to a controlled calcination step. Since this is mostly done under air, a subsequent reduction step is needed to obtain the reduced metal. Hayashi et al. studied the amount of carbon left on Fe/SiO2 catalysts, prepared by microemulsion, after submitting the samples to calcination in air during 5 hours at different temperatures%.At temperatures above 500 "C the amount of carbon in such catalysts was similar to the amount of carbon detected in a Fe/Si02 catalyst prepared following a conventional impregnation procedure after calcination under the same conditions. As a way of summarizing, we could say that a major advantage of the method resides in the possibility of an easy control of the particle size of the supported catalyst. This control may be achieved by the adequate selection of the structure of the surfactants and the organic solvents used within the microemulsion. As stated in this paper as well as is shown in Figure 5, when metal-supported catalysts are prepared by microemulsion technique, the metal particle size is smaller and the particle size distribution is sharper than that observed for analogous catalysts prepared by conventional methods such as impregnation or incipient wetness96.

4.4 Microemulsion vs Traditional Techniques. - The performance of catalysts prepared by microemulsion has been studied in heterogeneous catalytic processes. Here we will describe some examples in which catalysts prepared by microemulsion display a performance different from that of catalysts prepared by traditional synthesis routes. The preparation of controlled morphology Ce02-Zr02/A1203supports has been achieved by using microemulsion technique^^^. This technique allows the preparation of ceria-zirconia mixed oxides highly dispersed on an alumina carrier displaying a moderate uniform size. A definite control of the morphology of such a support is essential in order to prepare TWC since the oxygen storage capacity (OSC) and other effects are related to the properties of the noble metal ceria-zirconia interface. These properties are optimal at Ce:Zr atomic ratios close to 1. However, the redox properties depend on the geometrical structure of the mixed oxide. Thus, the preparation procedure is a key point in order to obtain adequate Ce-Zr based materials. Fernandez-Garcia et al. have described the synthesis and characterization of Pd/C~Zr1.,02/A1203three-way catalysts ( T W C S ) ' ~ ~by' ~Pd ' impregnation of a

284

Figure 5

Catalysis

TEM photographs of Rh-SiO? catalysts prepared by microemulsion and impregnation methods, adapted from reference 96

mixed-oxide support. The preparation of the unsupported oxide Zr-Ce is described elsewhere in this text29. Following a similar procedure, ceria-zirconia/alumina supports were made. This methodology allows the production of Zr-Ce/A1203supported mixed oxides with high surface area, above 160 m2g-’, suitable for achieving a good catalytic dispersion. This is of great importance since a major drawback of these mixed oxides materials is their low surface area. Also, cation stoichiometries close to 1:l are obtained by using the microemulsion method, providing catalysts with optimum redox and stability properties. Recently, the microemulsion technique has been explored as an alternative route to prepare Cu/ZnO catalysts’02.The Cu/ZnO catalyst was prepared in a microemulsion system using the oxalate route. In spite of the low surface are and poor Cu dispersion, the catalysts obtained by this procedure, when compared to catalysts prepared by a classical co-impregnation method, were more active in H2 production through partial oxidation of methanol. The authors propose this enhancement in the activity to be due either to the presence of some specific

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

285

crystallographic Cu planes or surface defects as actives sites of the reaction or to some strong interaction between a small part of the CuO and the ZnO lattice in the unreduced microemulsion. In this case the properties of the catalyst cannot be explained by the small size of the particles but by the special surface structure of the catalyst, arising from the special environment provided by the microemulsion mixture during its preparation. The reverse reaction, i.e. methanol synthesis, was studied over supported palladium catalysts prepared by Kim et al. using w/o micr~emulsions'~~. The microemulsion consisted of NP-5/cyclohexane/aqueous palladium solution for the formation of the microparticles. Hydrazine was added as the reducing agent. After the alkoxide precursor (zirconium tetrabutoxide, titanium tetrabutoxide or aluminium triisopropoxide) was added to the microemulsion, the hydrolysis was carried out at a pH in the 1.5-2 range. The particle size was 2.5-12 nm depending on the synthesis conditions. The highest CO conversion was obtained for catalysts of an average particle size of around 2.5 nm. The influence of several parameters were described. For instance, the authors showed that the support does not play an important role in the methanol selectivity,however Zr02-based catalysts displayed higher activities. Another example of a bimetallic catalyst prepared by the microemulsion technique is given by Zhang et a1.'04. Bimetallic Ru-Cu catalysts supported on SiOzwere synthesised in microemulsions. These catalysts displayed very interesting surface characteristics such as a narrow particle size distribution (3-5 nm), a higher surface area than the reference catalyst prepared by impregnation (408 and 198 m2/g, respectively), and a narrow pore size distribution. The catalysts were tested in the N 2 0decomposition reaction. The catalytic performance of the microemulsion catalysts was slightly higher than that of the reference. Recently, taking advantage of the very narrow size distribution of the metal particles obtained, microemulsion has been used to prepare electrocatalysts for polymer electrolyte membrane fuel cells (PEMFCs)"'. Catalysts containing 40 % Pt:Ru (1:1)and 40% Pt:Pd (1:1) on charcoal were prepared by mixing aqueous solutions of chloroplatinic acid, ruthenium chloride and palladium chloride with BerolO50 as surfactant in iso-octane. Reduction of the metal salts was complete after addition of hydrazine. In order to support the particles, the microemulsion was destabilised with tetrahydrofurane in the presence of charcoal. Both isolated particles in the range of 2-5 nm and aggregates of about 20 nm were detected by transmission electron microscopy. The electrochemical performance of membrane electrode assemblies, MEAs, prepared using this catalyst was comparable to that of the MEAs prepared with a commercial catalyst. Liu et al. have prepared PtRu/C catalysts from microemulsions and emulsions'06. Their results show how particles prepared from microemulsions displayed a lower particle size than the emulsion counterparts. Metallic particles displaying a small particle size, around a few nanometres, and narrow size distribution could easily be obtained via the microemulsion technique. Such particles have proved to be active in several classic heterogeneous catalytic processes, in some cases displaying a higher performance than the catalysts prepared following more traditional impregnation methods. Major

Na2PdC14

Ru

34

118 <10

119

117

79

114

Reference

-

116 7- 120'b'

115

-

Particle size nm

(a)

In the oil phase and surfactant columns, the / indicates the various components that can be used to prepare the microemulsions.(b) Rh particle size as a function of the determination method CO chemisorptionnEM. (') Particle size in solution.

Metal Oxides

BerolO50

Pt, Ir, Rh, Ru, Pd

Noble metal particle Hydrocarbons preparation Support and oxidation catalysts

PEGDE

Sodium n-dodecyl sulfate Triton X 1001 Ipegal CA 520

PdClZ FeC13,Fe(N03h cuc12

Na2PdC14 HAuC14

PEGDE/Genapol@ 26-L-60

Pentane Alkyl phenol

Precursors

Surfactant (a)

Oil phase

Supported noble Treating waste water metal Supported Pt and Pd Combustion catalysts Rhodium complex Hydroformylation Olefin + products medium Pd, Fe, Cu@) Oxidation catalyst Cyclopentane Cyclohexane Heptane, octane Ru/A1203 CO hydrogenation n-Hexane

Pd-Au/A1203or Si02 Vinyl acetate production

Use

Metal-based catalysts prepared by microemulsion

Metallsupport

Table 8

CO hydrogenation

Noble metal particle preparation Support and oxidation catalysts

Ru/A1203

Pt, Ir, Rh, Ru, Pd Hydrocarbons

BerolO50

PEGDE

Triton X loo/ Ipegal CA 520

Sodium n-dodecyl sulfate

Alkyl phenol

PEGDE/Genapola26-L-60

Surfactant (a)

Ru NazPdC1,

4 0

36

119

118

117

79

114

__

115 116 7-1 20‘”)

Reference

Particle size nm

PdC12 FeC13, Fe(NO3)3 cuc12

Na2PdC1, HAuC1,

Precursors

(a)

The / indicates the various surfactants that can be used to prepare the microemulsions.(b) Particle size in solution. ()‘ The microemulsionwas used as the reaction medium and catalyst support

Metal Oxides

Pd, Fe, CdC)

Treating waste water Combustion catalysts Hydroformylation medium Oxidation catalyst

Supported noble metal Supported Pt and Pd Rhodium complex Olefin + products Cyclopentane Cyclohexane Heptane, octane n-Hexane

Vinyl acetate production Pentane

Pd-Au/Alz03or Si02

Oil phase

Use

Relevant patents concerning some catalytic applications of materials prepared by microemulsion

Metal/Support

Table 9

288

Catalysis

advantages may be expected when catalysts prepared by means of microemulsion are used in structure-sensitive reactions. Effects ascribed to heterogeneous particle size distribution may easily be overcome. However, further improvements in the supporting step have to be accomplished if this method is to be used as a major route to preparing heterogeneous catalysts for industrial applications. Although not reflected in the literature reports, an optimisation in the recycling of the solvents used during the catalyst synthesis is mandatory for its further implementation at industrial level. However, there are several patents concerning the application of catalysts prepared by microemulsion. Tables 8 and 9 summarize the most relevant results obtained with catalysts prepared by microemulsion as found in the literature and patents, respectively.

5

Concluding Remarks

In this paper, a detailed description of the state-of-the-art of catalyst preparation by the microemulsion technique has been given. It has been shown that one of the major advantages of the microemulsion technique is that it allows the preparation of metal-based catalysts displaying a narrow particle size distribution in the nanosize range. This feature is of a great importance since these materials may have a wide range of applications. Nanoparticles can possess unique properties, quite different from those of relatively larger particles of the same material. In some cases the performance of the microemulsion-based catalysts was superior to that of the traditional catalysts regarding both activity and selectivity. However, for the use of the microemulsion technique in catalyst preparation, the nature of the catalytic process itself should be taken into account. For processes in which the particle size does not play an important role, the election of other techniques for catalyst preparation might be more appropriate. The special properties of the catalysts prepared from microemulsion systems include in addition to a small and uniform particle size also unique features related to the special environment in which particle formation takes place. However, further studies are still required in order to understand the relation between the mode of preparation and the properties of the particles. It is expected that nanotechnology in catalysis manufacturing will become a remarkable challenge for catalysis science"'. Potentially, microemulsion may play an important role in the development of nanomaterial science, providing an easy route to preparing metallic and oxide nanoparticles. Acknowledgments. - S. Rojas would like to thank the Spanish Ministerio de Educacion, Cultura y Deportes for a postdoctoral grant. The European Commission and the Swiss Government are acknowledged by S. Eriksson for financial support to the AZEP project; contract no ENKS-CT200 1-005 14

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

289

References 1. 2. 3.

4. 5.

6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 19. 20.

21.

22. 23. 24. 25. 26. 27. 28. 29. 30.

J.M. Thomas and W.J. Thomas in Principles and Practice ofHeterogeneous Catalysis, VCH, Weinheim, Germany, 1997. G. Ertl, H. Knotzinger and J. Weitkamp (Eds.) in Preparation ofsolid catalysts. Wiley-VCH, Weinheim, Germany, 1999, p 49. G. Ertl, H. Knotzinger and J. Weitkamp (Eds.) in Handbook of Heterogeneous Catalysis. Wiley-VCH, Weinheim, Germany, 1997, ch 2. B.K. Paul and S.P. Moulik, J . Disper. Sci. Technol., 1997,18,301. V. Chhabra, M.L. Free, P.K. Kang, S.E. Truesdail and D.O. Shah, Tenside. Surf: Det., 1997,34,3. W. Stoeckenius, J.H. Schulman and L.H. Prince, J . Phys. Chem., 1959,63,1677. P.A. Winsor, Trans. Faraday SOC.,1948,44,376. L.E. Scriven, in Micellization, Solubilisation and Microemulsions K. L. Mittal (Ed.), Plenum,New York, 1977,2, p 877. M.-J. Schwuger, K. Stickdorn and R. Schomaacker, Chem. Reo., 1995,95,849. T. Hellweg, Curr. Opin. Collod In., 2002,7, 50. S . Nave, J. Eastoe, R.K. Heenan, D. Steytler and I. Grillo, Langmuir 2000,16,8741. M.-P. Pileni, Adv. Colloid. Interfac., 1993,46, 139. Y. Li and C.-W. Park, Langmuir, 1999,15,952. J. Agrell, M. Boutonnet Kizling, Ulla Jacobson and J.C. Eriksson, in preparation to be submitted to Colloids and Surfaces. U. Natarajan, K. Handique, A. Mehra, J.R. Bellare and K.C. Khilar, Langmuir 1996,12,2670. S . Quintillan, C. Tojo, M.C. Blanco and M.A. Lopez-Quintela, Langmuir 2001,17, 7251. I. Lisiecki and M. Pileni, J . Am. Chem. SOC.1993,115,3887. 18 T. Hellweg and W. Eimer, Colloids Surfaces A 1998,136,97. M. Boutonnet, J.Kizling, P.Stenius and G.Maire, Colloid. Surface., 1982,5,209. R. Touroude, P. Bernhardt, G. Maire, J. Kizling, M. Boutonnet-Kizling and P. Stenius in Organized solutions, Surfactants in Science and Technology, Marcel Dekker, Inc., New York, 1992, p 357. J. Kizling, M. Boutonnet-Kizling, P. Stenius, R. Touroude and G. Maire in Electrochemistry in Colloids and Dispersions, VCH Publishers, Inc., New York, 1992, pp 3 3 3-344. John M. Thomas and Kirill I. Zamaraev (Eds.) in Perspectives in catalysis, Blackwell Scientific Publ., WHERE, 1992. C.N.R. Rao and B.Raveau, in Transition metal oxides : structure, properties and synthesis of ceramic oxides, Wiley-VCH, New York, cop. 1998. R.J. Farrauto and R.M. Heck, Catal. Today 1999,51,351. T. Masui, K. Fujiwara, K. Machida and G. Adachi, Chem. Mater., 1997,9,2197. J. KaSpar, P. Fornasiero and M. Graziani, Catal. Today, 1999,50,285. A. Martinez-Arias, M. Fernandez-Garcia, V. Ballesteros, L.N. Salamanca, J.C. Conesa, C. Otero and J. Soria, Langmuir, 1999,15,4796. M. Fernandez-Garcia, A. Martinez-Arias, A. Iglesias-Juez,C. Belver, A.B. Hungria, J.C. Conesa and J. Soria, J . Catal. 2000,194,385. A. Martinez-Arias, M. Fernandez-Garcia, A.B. Hungria, J.C. Conesa and J. Soria, Alloys Compd.,2001,323-324,605. P-O.F. Anderson, M. Pirjamali, S.G. J a r h and M. Boutonnet-Kizling, Catal. Today, 1999,53,565.

290 31. 32.

33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58.

59. 60. 61. 62. 63. 64. 65. 66. 67.

Catalysis

M.F.M. Zwinkels, S.G. Jaris, P.G. Menon and T.A. Griffin, Catalysis Reviews Science and Engineering 1993,35,319. E.M. Johansson, D. Papadias, P.O. Thevenin, A.G. Ersson, R. Gabrielsson, P.G. Menon, P.H. Bjornbom and S.G. Jaris, in Catalytic Combustionfor Gas Turbine Applications, Catalysis-Specialist Periodical Reports, Volume 14, Royal Society of Chemistry, Cambridge 2001 A.J. Zarur, H.H. Wu and J.Y. Ying, Langmuir, 2000,16,3042. J.Y. Ying and A. Zarur, US Patent 6413489. B1. A.J. Zarur and J.Y. Ying, Nature, 2000,403,6765. L.M. Gan, L.H. Zhang, H.S.0 Chan, C.H. Chew and B.H. Loo, J. Muter. Sci. 1996, 31,1071. H. Herrig and R. Hempelmann, Muter. Lett., 1996,27,287. M.R. Alvarez Lopez, M.J. Torralvo Fernandez, C. Mas Carbonell, and C. Otero Arean, J. Muter. Sci. Lett., 1993,12, 1619. F. Meyer, A. Dierstein, Ch. Beck, W. Hartl, R. Hempelmann, S. Mathur and M. Veith, Nanostruct. Muter., 1999,12,71. F.J. Arriagada and K. Osseo-Asare, Colloids Surf:, 1992,69, 105. F.J. Arriagada and K. Osseo-Asare, J . Colloid Interface Sci., 1995,170,8. A. van Blaaderen, and A.P.M. Kentgens, J. Non-Cryst. Solids, 1992,149,, 161. C-L. Chang and H.S. Fogler, AlChE J., 1996,42,3153. J. Esquene, Th.F. Tadros, K. Kostarelos and C. Solans, Langmuir, 1997,13,6400. L.M. Gan, K. Zhang and C.H. Chew, Colloids Surf.,A, 1996,110,199. V. Chhabra, V. Pillai, B.K. Mishra, A. Morrone and D.O. Shah, Langmuir, 1995,11, 3307. E.J. Kim and S-H. Hahn, Muter. Sci. Eng. A, 2001,303,24. M. Wu, J. Long, A. Huang and Y. Luo, Langmuir, 1999,15,8822. E. Joselevich and I. Willner, J. Phys. Chem., 1994,98,7628. G.L. Li and G.H. Wang, Nanostruct. Muter., 1999,11,663. F.D. Moran, J.R. Bartlett, J.L. Woolfrey, G.A. Bowmaker, R.P. Cooney, Ceramic Processing Science and Technology 1995,27. F.D. Moran, J.R. Bartlett, J.L. Woolfrey, G.A. Bowmaker and R.P. Cooney, J . Sol-Gel Sci. Techno1.,1997,8,65. X. Ju, P. Huang, N. Xu and J. Shi, J. Membr. Sci., 2002,202,63. X. Fu and S . Qutubuddin, Colloids Surf: A, 2001,179,65. E.J. Kim and S-H. Hahn, Muter. Lett., 2001,49,244. Y. Mori, Y. Okastu and Y. Tsujimoto, J. Nanopart. Res., 2001,3,219. C.Y. Tai, M-H. Lee and Y-C. Wu, Chem. Eng. Sci., 2001,56,2389. J. Fang, J. Wang, S-C. Ng, C-H. Chew and L-M. Gan, Nanostruct. Muter., 1997,8, 499. T. Yamaki, H. Maeda, K. Kusakabe and S . Morooka, J. Membr. Sci., 1993,85,167. T. Kawai, A. Fujino and K. Kon-No, Colloids Surf: A,1996,109,245. J. Wang, L.S. Ee, S.C. Ng, C.H. Chew and L.M. Gan, Muter. Lett., 1997,30, 119. M. Boutonnet Kizling and F. Regali, Stud. Surf: Sci. Catal., 1998,118,495. P. Tartaj and L.C. De Jonghe, J. Muter. Chem., 2000,10,2786. D. Burgard, C. Kropf, R. Nass and H. Schmidt, Mat. Res. Soc. Symp. Proc., 1994, 346, 101. D.W. Matson, J.L. Fulton and R.D. Smith, Muter. Lett., 1987,6,31. Y-S. Byoun, C-S. Oh and S-C. Choi, Ceramic Processing Science and Technology 1995,21. L. Balducci, R. Ungarelli and C. Paolo, US Patent 5731261 (1998)to Enichem S.p.A

6: Microemulsion: an Alternative Route to Preparing Supported Catalysts

68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99.

29 1

K. Holmberg, Adu. Colloid. Interfac., 1994,51, 137. M.J. Lawrence and G.D. Rees, Adu. Drug Deliver Rev., 2000,45,89. Ch-Ch. Wang, D-H. Chen and T-C. Huang, Colloid. Surface. A, 2001,189,145. D.H. Chen and S.H. Wu, Chem. Muter. 2000,12,1354. J. F. Rivadulla, M.C. Vergara, M.C. Blanco, M.A. Lopez-Quintela and J. Rivas, J. Phys. Chem. B,1997,101,8997. Ya-Ping Sun, P. Atorngitjawat and M. J. Meziani, Langmuir, 2001,17,5707. M.-L. Wu, D.-H. Chen and T.C. Huang, Langmuir 2001,17,3877. M. Boudart, Adu. Catal., 1969,20,153. M. Boutonnet, Private communication. M. Boutonnet, J. Kizling, R. Touroude, G. Maire, and P. Stenius, Appl. Catal. 1986, 20,163. M. Boutonnet, J. Kizling, V. Mintsa-Eya, A. Choplin, R. Touroude, G. Marie and P. Stenius, J. Catal. 1987,103,95. L. Tinucci and E. Platone, EP 0 380 154 A2. 1990. M. Spiro and D.M. de Jesus, Langmuir, 2000,16,2464. D.M. de Jesus and M. Spiro, Langmuir, 2000,16,4896. A. Dausher, R. Touroude, G. Maire, J. Kizling and M. Boutonnet-Kizling, J. Catal., 1993,143,155 S.J. Tauster, S.C. Fung and R.L. Garten, J. Am. Chem. SOC.,1978,100,170. C.J. Brinker and G.W. Scherer in Sol-Gel Science: The physics and chemistry of sol-gel processing, Academic Press, Inc., California, 1990. M. Boutonnet Kizling, C. Bigey and R. Touroude, Appl. Catal. A, 1996,135, L13. M. Kishida, T. Fujita, K. Umakoshi, J. Ishiiyama, H. Nagata and K. Wakabayashi, J. Chem SOC.Chem Comm., 1995,91,763. M. Kishida, K. Umakoshi, J. Ishiyama, H. Nagata. and K. Wakabayashi, Catal. Today, 1996,29,355. T. Hanaoka, W-Y. Kim, M. Kishida, H. Nagata and K. Wakabayashi, Chem. Lett., 1997,645. T. Tago, T. Hanaoka, P. Dhupatemiya, H. Hayashi, M. Kishida and W. Wakabayashi, Catal. Lett., 2000,64,27. H. Hayashi, L.Z. Chen, T. Tago, M. Kishida and K. Wakabayashi, Appl. Cat. A,2002,231,81. M.E. Dry, Catal. Today, 2002, 71, 227. W-Y. Kim, T. Hanaoka, M. Kishida and K. Wakabayashi, Appl. Catal. A., 1997, 155,283. C. Binet, A. Jadi, M. Boutonnet-Kizling and J.C. Lavalley, J. Chem. SOC.Faraday Trans., 1992,88,2079. T. Hanaoka, T. Hatsuta, T. Tago, M. Kishida and K. Wakabayashi, App. Catal. A, 2000,190,291. M. Ikeda, S. Takeshima, T. Tago, M. Kishida and K. Wakabayashi, Catal. Lett. 1999,58, 195. T. Hanoka, H. Hayashi, T. Tago, M. Kishida and K. Wakabayashi, J. Colloid Interf. Sci., 2001,235,235. M. Kishida, K. Ichiki, T. Hanaoka, H. Nagata and K. Wakabayashi, Catal. Today, 1998,45203. M. Kishida, T. Hanaoka, H. Hayashi, S. Tashiro and K. Wakabayashi, in Preparation of Catalysts, Delmon et a1 (Eds.), VII, Elsevier, Amsterdam 1998,265. M. Boutonnet, J. Kizling, R. Touroude, G. Marie and P. Stenius, Catal. Lett., 1991, 9,347.

292

Catalysis

100. M. Fernandez-Garcia, A. Martinez-Arias, A. Iglesias-Juez, A.B. Hungria, J.A. Anderson, J.C. Conesa and J. Soria, Appl. Catal. By2001,31,39. 101. A. Martinez- Arias, M. Fernandez-Garcia, A. Iglesias-Juez, A. B. Hungria, J.A. Anderson, J. C. Conesa, and J. Soria, Appl. Catal. By2001,31,51. 102. J. Agrell, K. Hasselbo, K. Jansson, S.G. Jaris and M. Boutonnet, Appl. Catal. A,2001,211,239. 103. W-Y. Kim, H. Hayashi, M. Kishida, H. Nagata and K. Wakabayashi, Appl. Catal. A,1998,169,157. 104. K. Zhang, C.H. Chew, S. Kawi, J. Wang and L.M. Gan, Cat. Lett., 2000,64,179. 105. M.J. Escudero, E. Hontaiion, S. Schwartz, M. Boutonnet and L. Daza, J . Power Sources, 2002,106,206. 106. Z . Liu, J.Y. Lee, M. Han, W. Chen and L.M. Gan, Journal ofMaterials Chemistry, 2002,12,2453. 107. E.C. Derouane, Catal. Commun., 2000,l,vii-viii. 108. J. Rymes, G. Ehret, L. Hilaire, M. Boutonnet and K. Jirhtova, Catal. Today,2002, 75,297. 109. E. Pocoroba, L.J. Pettersson, J. Agrell, M. Boutonnet and K. Jansson, Top. Catal., 2001,16/17,407. 110. A. Martinez-Arias, M. Fernandez-Garcia, A. Iglesias-Juez, J.A. Anderson, J.C. Conesa and J. Soria, Appl. Catal. B; 2000,28,29. 111. W.T. Ford and G. Liu, Langmuir, 2000, 16, 8814. 112. A. Martino, M. Stoker, M. Hicks, C.H. Bartholomew, A.G. Sault and J.S. Kawola, Appl. Catal. A, 1997,161,235. 113. M. Kishida, T. Hanaoka, W.Y. Kim, H. Nagata and K. Wakabayashi, App. Surf: Sci, 1997,121/122,347. 114. R.S. Tanke, WO 9733690,1997. 115. G.V. Odegova, T.O. Zajkova, V.V. Boldyrev, L.A. Pavljukhina and S.A. Savintseva, RU 21 15474,1997. 116. M. Boutonnet-Kizling, S. Jaris, L.J. Pettersson, U. Winter and M. Zwinkels, WO 9709114,1997. 117. L. Irme, R. Schomacker and J. Daun, US 5710342,1998. 118. H. Abrevaya and W. M. Targos, US 47. 119. B.H. Robinson, A.N. Khan-Lodhi and T. Towey in Structure and Reactivity in Inverse Micelles, Elsevier, Amsterdam and New York, 1989, p. 198. 120. M.L. Wu, D.H. Chen and T.C. Huang, J . Colloid. Interf: Sci., 2001,243, 102. 121. M. Lade, H. Mays, J. Schmidt, R. Willumeit and R. Schomaker, Colloid. Surface A, 2000,163, 3.

7 Catalysis of Acid/Alde hyde/Alco hol Condensations to Ketones BY KERRY M. DOOLEY

1

Introduction

This review covers the catalytic literature on condensation reactions to form ketones, by various routes. The focus is on newer developments from the past 15 years, although some older references are included to put the new work in context. Decarboxylative condensations of carboxylic acids and aldehydes, multistep aldol transformations, and condensations based on other functional groups such as boronic acids are considered. The composition of successful catalysts and the important process considerations are discussed. The treatment excludes enantioselective aldehyde and ketone additions requiring stoichiometric amounts of enol silyl ethers (Mukaiyama reaction) or other silyl enolates, and aldol condensations catalyzed by enzymes (aldolases) or catalytic antibodies with aldolase activity. It also excludes condensations catalyzed at ambient conditions or below by aqueous base. Recent reviews on these topics are those of Machajewski and Wong, Shibasaki and Sasai, and Lawrence.'-3 The enzymatic condensations produce mainly polyhydroxyketones. The Mukaiyama and similar reactions require a Lewis acid or Lewis base as catalyst, and the protecting silyl ether or other group must be subsequently However, in some recent work the silane concentrations have been reduced to catalytic amounts (or even zero); this work is discussed. Although much work has been done to determine the reaction pathways for ketone syntheses at high temperature from acids, aldehydes and alcohols, the mechanisms are still incompletely understood. In part this is due to the great variety of reactions that can occur on metal oxides and with transition metal complexes in solution. In particular, there are three likely condensation routes to ketones starting from acids and aldehydes, one predominating at high temperatures (roughly > 650 K) consisting of the attack of an adsorbed ketene or alkyl decomposition product on a surface carboxylate, the others predominating at lower temperatures with acyl carbenium or enolate surface intermediates as the attacking species. In the latter two cases multiple possibilities for ketone formation exist, depending upon whether the target carbonyl group in the aldol condensation is dehydrated or whether it reverts back to the carbonyl. If the former occurs, the catalyst must include a selective hydrogenation function, to Catalysis, Volume 17 0 The Royal Society of Chemistry, 2004

293

294

Catalysis

hydrogenate the a,P-enone. If the latter occurs, ketone formation is hypothesized to occur through decarboxylation of the initial enolate's carbonyl group.

2

Decarboxylative Condensation, Acids

Non-symmetric or symmetric ketones can be produced by the decarboxylative condensation of carboxylic acids. Rajadurai has also reviewed the process and possible reaction mechanisms! The general reaction is: RCOOH

+ RCOOH

--*

RCOR'

+ H20 + C02

Esters, especially methyl esters, can also be used as feedstock. The product ketones are useful as solvents for elastomers, polyvinyl acetate and other plastics, as extraction solvents, and as intermediates in making pesticides, herbicides, and pharmaceuticals. For many of the commercial products, acetic acid is one of the reactants. Certain aldehydes can also be condensed catalytically (acid/aldehyde or aldehyde/aldehyde) to give ketones. From the patent literature, most previous work on acid condensation reactions involved experiments using supported mixed metal oxides with weakly acidic and basic sites, possibly along with smaller amounts of strongly basic oxide additives. Most catalysts exhibit either low or moderate reducibility at typical reaction conditions of >650 K and 0.1-2.0 MPa.6-13 Such materials include the amphoteric or weakly basic oxides such as CeOz, Mn02/Mn304, Fe2O3, ZnO, Ti02or Stronger basic oxides, such as La203 or the mixed oxides formed from Mg/Al hydrotalcites, can be quite active at short times onstream,2' but the activity decays due to irreversible carbonate formation, and the product distributions of typical ketonizations have not been reported. Metal oxide catalysts of high basicity and low M - 0 bond strengths are poor catalysts for decarboxylation?" favoring the formation of bulk acetates. Oxides such as La203 or Nd2O3 containing a wide distribution of both acidic and basic tend to catalyze cracking reaction^,'^ and reduction to the alcohols. Stronger base doping (4 wt%) of certain catalysts, such as Z r 0 2or TiO2, has the possible advantage of inhibiting coke f ~ r m a t i o n . ~ . ~ ~ Several mechanisms have been proposed for acid decarboxylation to ketones. Gonzalez et al. and Pestman et al.11924proposed similar schemes (Scheme 1) involving the coupling of adsorbed ketene and carboxylate intermediates, shown here for an acetic acid/carboxylic acid condensation. A carboxylate intermediate is formed on a coordinatively unsaturated metal site, and a surface ketene intermediate is formed on another site by the dehydrogenation of the acid. The surface ketene then reacts with an adsorbed carboxylate to ultimately form the ketone, eliminating COZ. The details of the coupling step are not clear; it is likely that more than one distinct step is involved. Some factors which support this mechanism, on certain catalysts, are as follows. First, if true, at least one of the reactants must contain an a-hydrogen atom or atoms to allow formation of the ketene intermediate. Pestman et al." used acids containing three to no a-hydrogen atoms; as the number of a-

7: Catalysis of AcidlAZdehydelAlcohol Condensations to Ketones

295

hydrogen atoms decreased, the formation of ketone did also, and no ketone was formed if there were no a-hydrogen atoms in either reactant. Second, a-unsaturated ketones are formed at low conversions from carboxylic acids over many metal oxides that are good ketonization catalysts, either at temperatures (600-660 K) slightly lower than those typically associated with ketonization, or at short contact times."^^^-^^ Third, using acetic ["C] acid and trimethylacetic acid, it was found that the carbonyl in the resulting ketone arose not from the acetic ["C] acid, but rather from the trimethylacetate only. In some TPD studies, no carboxylates could even be detected on certain oxide surfaces at >-650 Finally, it was found that acetone itself did not exchange its H-atoms with a deuterated Ti02 surface," but acetic acid did exchange its a-hydrogens, suggesting the absence of enol or enolate intermediates. However, the materials used in these studies (e.g., pure TiOz)were often quite different from what appear to be the better ketonization c a t a l y ~ t s . ~However, ~ - ~ * even over the better catalysts there are strong indicators of a 'ketene'-type mechanism, namely the presence of ketene and other unsaturated ketones. Metal oxides that have adjacent cations with coordination vacancies in certain common surface planes, like some TiOz and ZnO surfaces, can catalyze ketonization.26*29-30 This suggests that a bimolecular surface reaction is rate determining. However, not all such catalysts are very selective for ketonization, especially at high conversion. Kuriacose and co-workers proposed a mechanism based on the condensation of two adsorbed carboxylates, or a carboxylate and a less ionic The aldolspecies, following an aldol condensation-type pathway (Scheme 2).31-32 type attack on the carbonyl group of the less enolizable reactant leads to decarboxylation. The details of C 0 2formation are unclear, but C 0 2would have to arise from the carbonyl of the more enolizable reactant as depicted in Scheme 2. Okumura and Iwasawa also concluded, from studies of acetic acid reacting on Zr02/Si02catalysts, that a possible mechanism for ketonization involved the reaction of two carboxylate-type intermediates, one less nu~leophilic.~~ The other possible mechanism suggested by the above researchers consisted of reaction of an adsorbed acyl carbenium ion (RCO+) with an adsorbed carboxylate to give ketonization p r o d u ~ t s . ~Acyl l - ~ ~carbenium ions can be formed in the presence of strong Lewis acids. Kuriacose and co-workers proposed that at c673 K a ketene can react with H + to form such an ion, which then reacts with adsorbed a ~ e t a t e . ~ On l-~~ a Pd/Ce02 or Co/Ce02 catalyst, the ketonization observed at lower temperatures has also been attributed to reaction between a surface acetyl and an alkyl g r o ~ p . There ~ ~ - ~are ~ even ketonization reactions in solution which may occur according to this mechanism. For example, the ring-closing of diacids such as 2,Sdimethyladipic acid to 2,5-dimethylcyclopentanone can take place at less than 573 K with phosphate and pyrophosphate salts as catalysts, in Cloor heavier alkane solvents.36The ratio of acid to catalyst must be low at these conditions, < 5. Under similar conditions, 3-acyltetrahydrofurans can undergo a dual intramolecular condensation followed by decarboxylation to give alkylcyclopropyl, alkylketones; here the reagent ratio to the preferred iodide catalysts is even smaller, 2, and the selectivity lower.37 Therefore, it is seen that the main feature distinguishing the proposed mechanK.'9926

-

-

CataI ysis

296

H0' R)=o+

#

O\M/O\M/O

+ *OH + CO:!

Scheme 1

(3)

Ketene Mechanism with Acetic Acid Forming Surface Ketene Intermediate (* = unspeciJedsite)

isms is the identity of the electrophilic or other species attacking an adsorbed carboxylate. Most commercial ketonizations catalyzed by metal oxides are performed at temperatures >650 K, because at lower temperatures there are operational problems due to formation of heavy tars from the consecutive aldol condensations and Michael additions of the ketones, which are favored at these condition^.'^^'^^^^ Therefore the acyl carbenium ion route may not apply to these processes. However, lower temperature, less selective processes and liquid-phase reactions in highly acidic media could take place by an acyl carbenium route. At high-temperature conditions, the product distributions of typical decarboxylative and aldol condensations vary with temperature, time on stream, and catalyst age. Several ketone isomers can be p r ~ d u c e d . With '~ acetic acid, e.g., C& (e.g., methylhexanone, pentan-2-one, 3,3'-dimethylbutan-2-one) ketones and alkylphenols arise from acetone a l d ~ l i z a t i o n . 'An ~*~ important ~ cyclic product in low temperature acetone aldolization is isophorone (2-cyclohexen-1-one, 3,5,5'-trimethyl),formed by the aldol condensation of acetone with mesityl oxide, followed by 1,6-Michael addition.15J9In reactions with acetic acid, we have observed 2-cyclohexen- 1-one, 3,5-dirnethyl;' which is probably a cracking product of isophorone, and small amounts of isophorone itself. Cracking to produce

7: Catalysis of AcidlAldehydelAlcohol Condensations to Ketones

Scheme 2

297

Carboxylate mechanism (v = vacancy)

even-numbered carbon products is common for acetone reacting on metal oxides. Other ketone products can be formed by ring expansion. For example, in the condensation reaction of cyclopropane carboxylic acid (CCA) to methylcyclopropyketone (MCPK), the c8 ketone 1-acetyl, 2-methyl-1-cyclopentene is formed, which probably results from an aldol reaction of acetone and MCPK. Other acetone/methylketone aldol products are also observed at high conversion. Detailed analysis of the product distributions demonstrates the selectivity of supported C e 0 2and other rare-earth oxides (REOs) for unsymmetric ketonization. For example, if acid condensation were statistically random, the ratio of the self-condensation products to the unsymmetric product MCPK should be 0.20 for a 4/1 molar acetic acid/CCA. Actual observed ratios were as low as 0.03.28Low ratios were attainable only after a certain time on stream, which leads to the question considered below of what is happening to supported REOs in the first few days of operation.

298

Catalysis

We have performed pulse reactor experiments using standard and isotopically labeled feeds to further explore acid and aldehyde decarboxylative condensation m e c h a n i ~ m s ? ~The - ~ experiments used small amounts of an active 10 wt.% Ce02/Ti02catalyst, at low conversions and over a temperature range of 663-723 K, for the reactions of acetic acid and CCA, and associated aldehydes, to MCPK and side products. For acid condensation, the results showed a negative order in MCPK with respect to acetic acid (-0.5), and a positive (+ 0.3) order with respect to CCA. As expected, the addition of either D 2 0or deuterated acetic acid slowed down the reaction rates, because H-D exchange between water and acetic acid is relatively fast at these temperatures. The isotope effect for formate decomposition on various metals has been studied extensively, and found to vary from -2-15 for HCOOH/DCOOD and from 1-2.5 for HCOOH/HCOOD!l For the reactions studied here the isotope effect was 2.1 for the CH3COOH/CD3COOH feeds and 1.4 for the H 2 0 / D 2 0 feeds. Taking into account that almost all of the experiments performed on formate decomposition were done at temperatures less than 473 K, and that isotope effects decrease with increasing temperature, the kinetic isotope effects calculated here are within the expected range for carboxylate decomposition.

-

3

Decarboxylative Condensation, Aldehydes and Alcohols

Aldehydes and alcohols can be oxidized to their corresponding carboxylic acids on certain oxide s ~ r f a c e s . ~ OCarboxylate J~*~~ formation can also take place by the breakdown of certain adsorbates, such as an ester to carboxylate and alkoxy, or by hydride transfer from one adsorbate to another, as takes place in the Cannizzaro disproportionation reaction of aldehydes to alkoxy and c a r b ~ x y l a t e . ~ ~ ~ ~ ~ The oxidation reactions are less likely on hydroxylated surfaces, because the surface is less nucleophilic.44Hydrogenation of an acid to an aldehyde can take place on oxides of intermediate M - 0 bond strength at temperatures of 623-723 K,45 and at lower temperatures on the Group VIII metals with low M - 0 bond ~ t r e n g t h . 4Obviously, ~~~~ once the carboxylates are formed, condensation to ketones could occur according to one of the mechanisms given above. When a good hydrogenation metal (eg, Pt, Pd, Rh) is added to the typical MO,, ketonization and surface acetate formation is suppressed in the absence of 0 2 . 3 4 - 3 5 A shift to decarboxylation plus hydrogenation occurs, giving alkanes, CO, C02, and H20. This is in part due to hydrogenation of the acids to aldehydes, which are then more reactive in aldol ~ondensations.4~ Ketonization is suppressed to a lesser degree or not at all when the poorer (in terms of ability to hydrogenate) d-type transition metals were used (e.g., CU).~* Surfaces with basic sites form enolates from both the aldehydes and ketones, leading to multiple aldol condensations and Michael addition^.'^*^*-^^ Candidate molecules must be enolizable, i.e., contain an a-hydrogen atom. Aldol condensation / Michael addition products cover the range from a$-unsaturated aldehydes, saturated aldehydes, hydrogenated products (alcohols),and the heavier aromatics resulting from multiple condensation^.'^^'^^^^ The presence of coordina-

7: Catalysis of AcidfAldehydelAlcohol Condensations to Ketones

299

tion vacancies in two adjacent cations is not required for the aldol condensation reaction to o c ~ u r . ~ ~ J ” Another competing reaction that can occur on strongly acidic or basic surfaces, mostly at lower temperatures, is the Tischenko cross-e~terification.~~ This reaction can be written as follows for aldehydes: RCHO

+ R’CHO

+

RCOOCH2R’

(1)

where R is the more electron withdrawing group. This reaction has been found to take place at temperatures as low as 323 up to 473 K.53*60-62 It is similar in inception to the Cannizzaro disproportionation reaction (two aldehydes reacting to form an acid and an alcohol). Acids formed through hydrolysis of these esters can deactivate the strongly basic sites. At higher temperatures the reverse Tischenko reaction is a source of aldehyde byproducts in condensation reactions when using ester feeds.13 On metals with stable lower oxidation states, but not completely reduced, dehydration/dehydrogenation reactions become preferred. An example reaction is reductive coupling (the McMurry r e a ~ t i o n ) : ~ ~ ~ ~ ~ - ~ 2 R’RC = 0

+ nMx+

+

RR’C = C R R

+ 2 [O] + nM(x+4/n)+

(2)

The surface must undergo a 4-electron oxidation, and so a strong reducing agent must be present. Most suboxides of Ti02 are capable of reductive co~pling;3’7~~ this reaction is also possible on CeO, and other partly reduced REOs. When H2 is present, the reductive coupling of ketones and aldehydes containing fewer a-hydrogens is favored over dehydration/dehydrogenation. Notwithstanding the complexity of the above chemistry, there have been several studies of aldehyde condensations at high temperatures, sometimes starting from alcohols, that produce ketones in high yield in preference to either single aldol condensation products or multiple aldol condensation / Michael addition products.20*23~34-35~48~65-70 For example, the reaction of 1-propanol to 3pentanone at 773 K and 723 K has been studied using Ce02/Fe203and La203 catalysts, respe~tively.6~-~~ Both groups proposed a mechanism in which the alcohol is dehydrogenated to the aldehyde, which undergoes aldol condensation (step 1 below) to 3-hydroxy-2-methylpentanal.They infer that the decomposition of the aldol product proceeds by means of oxidative decarboxylation (step 4) Both CO and C02, mainly C02, were rather than deformylation (step observed. The steps could be written in general terms as ([O] is surface oxygen): 5).66770

+ R’CH2CHO RCH(OH)CH(CHO)R’ RCH(OH)CH(CHO)R’ + [O]+ RCOCH2R’ + CO2 + H2 RCH(OH)CH(CHO)R’ RCOCH2R + CO + H2 RCHO

--.*

+

(3) (4)

(5)

Kamimura et al. speculate that the extra oxygen needed for the oxidative decarboxylation came from the decomposition of 1-propanol to propane, while Kamimura et Claridge et al. proposed that any surface oxygen could be al. reported good results for symmetric ketone formation using other primary

300

Catalysis

alc0hols.6~If a branching group were present in the 2-position, the selectivity to symmetric ketone greatly decreased. In the case of benzyl alcohol, which has no a-hydrogens, no symmetric ketone was produced. This suggests that the reaction proceeds through either an aldol condensation route or 'ketene' decarboxylation route. The finding that propanal reduces the iota (Pr6OI1)phase of praseodymium to the C-phase (Pr2O3) sesquioxide is possible evidence of the involvement of a surface oxide in step (4). Of note was that when using Ce02,the major products were CO and hydrocarbons with no ketones. This was attributed to the different structure of Ce02 and the presence of more acidic sites relative to Pr6011.65 But this could also be due to the higher activity of Ce02,which is equally effective for ketonization but at lower temperature^.^^^^^.^^ Plint et al. proposed a slightly different pathway, based on their study of several alcohols and oxygen to the ketones over Ce02/Mg0.67,7' They postulated that the alcohol is oxidized to the aldehyde, then to carboxylate, which undergoes aldol condensation with an aldehydic species followed by decarboxylation. The final reactions are depicted in Scheme 3. Again, the exact details of the last step are uncertain. For butanol, 4-heptanone was the major product, butanal was a significant product and smaller amounts of C8aldol products were present. It was proposed that Ce02converted the alcohol to the acid, while aldol condensation took place on the more basic MgO. Water addition increased the amount of butanal formed, and also the amount of aldol product. The problems with such a pathway are that it does not explain: (1)the absence

p-"

Scheme 3

Aldol condensationldecarboxylation of two adsorbed carboxylates

7: Catalysis of AcidlAldehydelAlcohol Condensations to Ketones

301

of higher aldol products or 2-ethylhexenoic acids; (2) the high ketone/primary aldol product ratio observed under conditions where the opposite would be expected, namely, high alcohol/Oz ratios; (3) the fact that the production of the intermediate aldols did not increase with respect to space velocity while in some cases 'final products' such as ketones or hydrocarbons did. In an alternate 'ketonization' mechanism, it was proposed that an alcohol is dehydrogenated to an aldehydic surface intermediate, which can be rapidly oxidized to an adsorbed carboxylate, even on reduced Ce02. The adsorbed carboxylates then condense to form ketone, C02, and hydroxy, presumably according to Scheme All of these experiments were performed using ethanol as the starting material, except for Idriss et al., who used a~etaldehyde.~~ The temperatures for ketonization ranged from 490-770 K. The proposed reactions can be partly written as follows (a = adsorbed): 1.20734-35748

-

O(S)+ CH3CHO" + CH3CHOO"

(6)

CH3CHOOa+ CH3COO" + Ha

(7)

2 CH3COOa+ CH3COCH3"

+ C02" + O(S)

(8) This pathway is supported the following observations. First, going from a primary to a secondary alcohol reactant promotes aldolization but inhibits ket~nization;~' according to Scheme 3 both should be inhibited. Second, there is an absence of 2-propanol when starting from CO and H2;722-propanol could have been oxidized to give the acetone. Third, diisopropylketone, an aldol condensation product often found under similar conditions when starting from CO and H2,68is absent. Fourth, the addition of even small amounts of metals (e.g., Co) to REOs greatly enhances ketonization while reducing primary aldol production at <700 K.34Finally, the reactions of aldehydes over similar catalysts at lower temperatures and partial pressures, where the primary aldol condensation products would be expected to predominate, typically produce a more complex mix of aldol condensation, dehydrogenation, dehydration, dehydrocyclization and reductive coupling products.'9~34~63*73 It is hard to see how the condensations are more selective to the single decarboxylated aldol product (one aldol condensation only) at the more extreme conditions of higher temperatures and partial pressures. A further problem with Scheme 3 and with reactions (3)-(5) at these hightemperature (>673 K) conditions is found in the behavior of catalysts such as Ce02or Ce02/Mg0. For these materials, little primary aldol product is formed from the reaction of 1-propanol regardless of the conversion, while the more basic MgO gives almost entirely primary aldol product (3-hydroxy-2-methylpentanal), again regardless of c o n v e r ~ i o nUnless . ~ ~ the Ce02is absolutely necessary to catalyze the final decarboxylation (in reaction (4) or Scheme 3), this evidence suggests a parallel, non-aldol route at these conditions. In other words, it is suggested that the problem with the strong bases of high M - 0 bond strength (e.g., MgO) is not that they cannot form carboxylates (MgO forms bulk acetates at these conditions,)," it is that they cannot undergo the reduction needed to generate a ketene-type intermediate. However, further studies with 13C-labeled

302

Catalysis

feeds are probably necessary to settle this question. Limited studies have been made to determine if an added oxidant could replenish the surface oxygen removed in aldehyde oxidations, when they occur. Thermodynamic calculations show the oxidation of aldehyde using water as an oxygen source would be favorable. 2 RCHO

+ H20

+-*

R2CO

+ C02 + 2H2

(9)

AH298 (acetaldehyde) = -27.5 kJ mol-' Experiments have shown that both ketone and COZ production increase dramatically when water is added. Claridge et al. performed preliminary studies using water in decarboxylative condensation, and found that a 1/1 propanal/water feed led to an increase in pentan-3-one and C 0 2 formation, over Fukui et al. have shown that adding water to an isobutyraldehyde feed (0.6-0.75 mol water/mol aldehyde) increased conversion and the selectivity to diisopropyl ketone over Zr02.23Plint et al. found that adding 20% water to their butanol/02 feed gave a high yield to 4-heptanone; however, with time on stream the selectivity to 4-heptanone d e c r e a ~ e d . ~ ~ When cyclopropanecarboxaldehyde(CCAld) was condensed with acetic acid in pulse reactor experiments, the only primary product arising from CCald was the acid CCA.39-40 The ketone (MCPK) production rate was -4 times slower with acid/aldehyde than acid/acid feeds. An aldol route to MCPK could proceed through either of the following reactions: CH3COCH3 + C3HSCHO CH3CHO

+ C3HsCHO

+

C3HSCH(OH)CH2COCH3

(10)

+

C3HSCH(OH)CH2CHO

(11)

Both intermediates could conceivably decompose to MCPK by oxidative decarboxylation to give COXand water or by a concerted decarboxylation reaction to acetaldehyde starting from the first intermediate, or to formaldehyde starting from the second intermediate. However, neither intermediate, nor their dehydration products, nor acetaldehyde, formaldehyde, or CO were found even in trace quantities. Therefore, it appears that in this case as well the ketone is not being produced by an aldol reaction, but rather by a decarboxylative condensation reaction of the aldehyde and acetic acid, using oxygen from the surface as needed. When water was added, an increase in ketone formation was observed when comparing runs performed at the same CCald partial pressures. The reaction order for water was estimated to be 0.2. With oxygen present a much lower selectivity was attained in acetic acid/CCAld condensations than with either water present or with no additive. Although CCA production was greater, this did not enhance the production of ketones. It was impossible to say whether the inhibition resulted from more total oxidation or from deactivation of active sites for condensation. Plint et al. found that the yield to 4-heptanone from butanol was maximized at a butanol/oxygen ratio of 6/1, using a 40% Ce02/Mg0 catalyst, at one particular flow rate.67At other flow rates the optimum ratio varied from 3/1 to 1/1. They also found that butanol conversion was positive order with respect to oxygen. But they con-

-

7: Catalysis of AcidlAZdehydelAlcohol Condensations to Ketones

303

cluded that the product distributions differed to such an extent that no correlation between selectivity to 4-heptanone and oxygen content was possible. Claridge et al. studied the reaction of propanal to 3-pentanone using O2or N 2 0 as They observed that at 773 K total oxidation occurred. The data from our work show that an increase in CCald/02ratio increases the production of ketone, suggesting total oxidation is the more probable cause of the lower ketone yields. Experiments with acid/aldehyde feeds were also conducted using deuterium oxide and acetic-d3-acid-hin place of water and acetic acid, r e s p e c t i ~ e l y .The ~~-~ addition of D 2 0had a significant effect on the reaction rate, so that a reaction involving water is most likely a slow step. The addition of acetic-d3-acid-halso decreased the ketone turnover ratio slightly, so the breaking of a C-H bond in acetic acid could also be a slow step in the mechanism, or there could be two parallel overall reactions, one using and one not using water. The kinetic isotope effects for these reactions were 1.5 for CH3COOH/CD3COOHand 6.7 for H20/D20. As with the acid/acid reactions, these kinetic isotope effects are in the expected range of carboxylate decomposition, based on data for formic acid.4l For aldehyde/aldehyde feeds (CH3CHO/CCald), the side products from CCald were 2-cyclopropyl-1-butene, and 3-cyclopropyl-1-butene, both formed from the aldol condensation of CH3CH0with CCald, but also some CCA. With such feeds the production of ketone was greatly decreased, indicating the necessity of at least one easily formed carboxylate for decarboxylative condensation. The addition of water increased the rate of production of ketone, again suggesting that water is supplying oxygen to the surface to produce carboxylate intermediates. There were also a number of side products produced from CH3CH0, by aldol self-condensation. The aldol condensations involving either CH3CH0 or CCald are in this case faster than their decarboxylative condensation reactions. From thermodynamic calculations it is obvious that while all the reactions discussed above are thermodynamically possible, certain reactions are more probable. For example, thermodynamic calculations suggest that the condensation of acetic acid/CCald is more likely to proceed to C 0 2 as a byproduct, and this is in fact what was observed. When looked for, more C02 is found than CO in the products of aldehyde/aldehyde c o n d e n s a t i o n ~ .It~ ~ is*also ~ ~ important to note that the I& values for the acid/aldehyde and aldehyde/aldehyde formation of the ketone are both typically larger than for the acid/acid reaction. From this one might expect the turnover rate for ketone production to be larger for these two feeds. However, this is not what is observed. Again, this indicates some involvement of an adsorbed carboxylate, which cannot be formed directly from an aldehyde. 4

‘One-Step’ Aldol Condensations to Ketones

Lower temperature processes and catalysts exist for so-called ‘one-step’ aldol condensations to produce ketones by condensation of two ketones or of a ketone

304

Catalysis

and an aldehyde to give initially the aldol product, followed by dehydration to the a,P-unsaturated ketone (enone), and then by hydrogenation to the saturated ketone; all of these can occur using the same catalyst. In some cases it even appears possible to start with the corresponding alcohols which are first dehydrogenated to the ketone^;'^'^^-^^ this presumably obviates the need for H2 addition. When H2is used, a typical ratio to ketone or aldehyde is between 0.25 and 1. When an alcohol is used to supply the hydrogen, higher molar ratios to the ketone can be used. Higher selectivities to the product of a single condensation (e.g., methyl isobutyl ketone from acetone) are obtained by using H2 pressures well above atmo~pheric,7~ or by the addition of 5 wt% recycled product of the single aldol conden~ation.~~ The alcoholic dehydrogenation reaction requires a transition metal component unless it is possible to operate above -470 K, where moderately basic catalysts also become active for this r e a ~ t i o n . ~An ~ - ~example ' catalyst would be a MgA10, hydrotalcite of low A1 content. These catalysts can activate Hz/D2 exchange at these temperatures. The alcohol adsorbs dissociatively to surface alkoxide and hydroxy, a step requiring proximity of a basic and a Lewis acid There follows a hydride abstraction, mediated by another Lewis acid group, and heterolytic association to give H2. When a ketone and an aldehyde are condensed in a cross-aldol reaction, excess ketone must be used to avoid aldehyde self-condensation, and even so the surface is essentially saturated with aldehyde at low temperature. Exceptions are reactions of easily enolizable ketones at ambient and lower temperatures.8l Therefore increasing the temperature often increases selectivity as well, a situation reminiscent of decarboxylative condensation, by allowing for greater ketone adsorption. An enolate formed from an unsymmetric ketone can be of two types; the less electronegative (usually more substituted) side is deprotonated preferentially by many homogeneous transition metal complexes with N-donor or halide ligand~,8~ in- ~agreement ~ with local hard-soft acid-base But steric hindrance plays a greater role with heterogeneous catalysts, to such an extent that the less substituted side is often deprotonated preferentially.s8*85 Some examples of 'one-step' processes for methylketone production are as follows.

-

-

(1) Methylpentylketone from the acetone/isopropanol/n-butanol reaction at -520 K, low pressure, LHSV 1, with major side reactions to acetone and heavier condensation The primary reaction is:

CHCOCH3

+ CH3(CH2)30H -> CH3CO(CH2)4CH3 + H20

(12)

(2) Methylketones or methylvinylketones from a ketone/aldehyde/H2 reaction at -430-470 K, 1-2 MPa H2,86or at higher temperatures with an alcohol as hydrogen source,59~74-76~s7 or at <430K if an unsaturated or ketoalcohol product is d e ~ i r e d . Examples ~ ~ - ~ are ~ ~acetone ~ ~ ~+ ~ butanal ~ ~ ~ ~+ heptan-2-one or citral + methyl ethyl ketone to n-methylpseudoionone. (3) Cyclopentanones + aldehyde alkylcyclopentanones at -430 K, 5 MPa

-

7: Catalysis of AcidfAldehydefAlcohol Condensations to Ketones

305

-

H2.89 (4) Methylisobutylketone from acetone at 390-470 K, 1-3 MPa H2.49'76-77,90-95 Conversions are usually 30-40% maximum and selectivities to c6 aldol products as high as 90% have been claimed. Lower H2 pressures are possible at lower conversion^?^-^^

-

Selectivity is often a problem with 'one-step' processes. Conversions and temperatures are kept relatively low to avoid successive aldol condensations and hydrogenations. Controlling the hydrogenation is difficult, and much of the catalyst development work has been directed here rather than toward the aldol condensation/dehydration steps. Alloying of Pd (the most commonly used hydrogenation component) with most other metals can either decrease hydrogenation activity or favor hydrogenation of the ~ a r b o n y lEven . ~ ~ for Pd/Cu there can be significant hydrogenation of the C = O to alcohol unless the H2 partial pressures or the conversions are Another problem appears to be the inhibition of nearby basic sites by the metal component, often rendering strongly basic catalysts useless in 'one-step' processes.56~s8~99 In summary, the 'one-step' processes would appear to be candidates for advanced reactor designs that enhance gas/liquid mixing (e.g., microchannel reactors) or H2 solubility (e.g., at supercritical, single-phase conditions), or reactant mobility on heterogeneous surfaces (e.g., nanoparticle mixtures of metals and basic oxides within shapeselective supports).'@' The synthesis of enones, usually a,P-unsaturated, and hydroxyktones using heterogeneous catalysis has barely been addressed (homogeneous catalysis is another story, see below). Two-step processes predominate in the patent literature, with the aldol condensation base-catalyzed and the ketol dehydration acid-catalyzed; either reaction can be conducted in the liquid or gas phase. These processes are not especially selective, but there have been many years of process development behind them. There has been some work on compressing the above process to one step using a solid base whose dehydration capability may be due to Lewis acid sites, e.g., alkali-exchanged zeolites.'o1These attempts do not appear to have resulted in reasonable selectivities, because further dehydration and condensation reactions occur. An exception is the synthesis of methyl or phenyl vinylketones by condensation of a precursor met hylketone with methanol/02 or formaldehyde. This process can take place with >95% selectivity at <20% conversion on Fe/Ag/A1203at 600-650 K,lo2or at >70% conversion with a secondary amine hydrochloride ( 40/1 aldehyde/amine)/Nb205catalyst mixture at 390-420 K,lo3 or at lower selectivity (- 40-50%) but almost complete conversion using typical H-form zeolites at 670-730 K.'04 Most commercial enone synthesis methods are not applicable to all desired enones. The acetylene condensation/hydration method can only make enones with even numbers of carbon^,'^^ while formaldehyde condensation preferentially gives branched products at higher molecular weights.lMOther methods use expensive oxidation reagents and start with the allylic alcohols.'07Yet other methods use even more expensive reagents,lo8-lwor electrochemical methods

-

306

Catalysis

with relatively low yields and complex product distributions."' Converting the enone to the dihydroxyenone is possible by nucleophilic epoxidation followed by selective hydrolysis. The epoxidation reaction is well known."' A peroxide dissolved in a strongly basic medium must be used (Weitz-Scheffer reaction). The selective hydrolysis of the epoxide to the dihydroxy compound proves more difficult. The trick is to avoid conversion to the trio1 or to a diketone, especially under the basic conditions.

5

Lower Temperature Condensationsto Ketones

The high temperatures and/or byproducts often associated with the reactions discussed above have led to attempts to produce ketones by condensations at near ambient conditions. Most of the catalysts for these reactions are soluble transition metal complexes, and there have been few accompanying studies of the stability and reusability of the complexes. For example, the condensation of boronic acids (mostly phenylboronic acids) with anhydrides takes place at <373 K to give the corresponding methylketones in high yields using [Rh(C2H4)C1]2 (70-85% yields),'12 or Pd(OAc)2/pivalicanhydride with added phosphines (e.g., PPh3, giving up to 97% yield^)."^-^'^ The reactions were not successful with simpler Rh or Pd salts, or with complexes containing bidentate ligands. One good feature of the process is its tolerance of certain other functional groups (e.g., methoxy, acetyl) in the boronic acid, and the ability to substitute the acid for the anhydride, albeit with lower yields. The process is highly solvent-dependent, however; for example, addition of water to THF is greatly beneficial when using the Pd-phosphine cataly~ts."~ Catalyst stability in these complex solvent mixtures is often a problem. The Mukaiyama aldol condensation of a ketone and an aldehyde has been extended to the reduction or in some cases even elimination of the silyl ethers which are commonly used as protecting and structure-directing groups for the ketone-derived enolate. Where ketones are non-symmetric, highly regioslective and sometimes stereoselectiveadditions are possible using typical soluble Lewis acids such as TiC14, usually with added N-donor l i g a n d ~ . " ~ -Metal ~ ' ~ (e.g., B) triflates and aluminum binaphthol (BINOL) alkyls have also been u ~ e d . ' ' ~ - ' ~ ~ The 'catalytic' addition of a silylating reagent is still beneficial; Yoshida et al. used a trimethylsilyl chloride (TMSCl) to limiting reactant ratio of -0.05 in regioselective reactions of aldehydes with a,a-dimethylketones (to a,a-dimethyl, P-hydroxyketones), a-chloroketones (to a-chloro, P-hydroxyketones) and aoxyketones (to a-oxy, P'-hydroxyketones).'16In most cases the increases in yields to the regioselective product upon addition of TMSCl were substantial. While none of these studies reported on long-term catalyst stability or reusability, and temperatures below ambient (consequently, low rates) were typically required, the ketone/aldehyde ratios (e.g., 10/1)are no greater than used in non-regioselective heterogeneous processes. Typical solvents are THF, diethylether and acetonit rile. Enantioselective additions are facilitated with chiral Lewis acids or L-proline;

307

7: Catalysis of AcidlAldehydelAlcohol Condensations to Ketones

2 + Ru AH+ BINOLcat.

R

Scheme 4

HOJ

THF

R

-

Ph

Ph

6H

6H

syn-3

anti-3

A low-temperature aldol condensation to dihydroxyketone

for example, 2-hydroxyacetophenone can add to aldehydes to give the syn-3 and anti-3 dihdroxyketones (Scheme 4), using complexes derived from LaLi3tris((S)binaphthoxide) (S-BINOL) or an (S,S)-BINOL dinuclear Zn complex,12oor a dinuclear 0x0-bridged Zn-complex with a complex semi-crown phenoxide ligand.121Acetophenone addition to aldhehydes to give P-hydroxyketones can be catalyzed in a regio- and stereoselective manner by this same dinuclear Zn complex,'22by LaLi3tris((R)-binaphtho~ide),123 or by an (S,S)-hydrobenzoin (a 1,2-diphenyldiol)complex of Ca[N{ Si(CH3)3}2]2 in 3:1 molar ratio.'24While here either a Ph3P=S, trimethylsilyl, or KSCN adduct or cocatalyst was still required, the typical ratio of limiting reagent to trimethylsilyl, Ph3P= S, or KSCN was lO/l, 5/1, and 33, respectively. Good regio- and stereoselectivities were obtained for a wide range of alkyl, aralkyl and alkenyl R-groups, with the temperatures kept to <280 K (usually ~ 2 4 0K for the reaction of hydroxyacetophenone). Again, the reactions are very slow - 24 h reaction times are common - and even then yields sometimes disappointing. The triphenylphosphine sulfide probably helped displace the product ligand from the catalyst; the trimethylsilyl probably formed a silyl ether group with the active enolate in-situ; the KSCN apparently became part of an aggregated Ca-complex that comprised the working catalyst. Tridentate BINOL catalysts can be derived from titanium tetra(isopropoxide), BINOL ligands, and hindered amine bases. These catalysts have also been shown to provide good yields and ee's for the aldol reactions at low temperatures of 2-methoxypropene with several aldehydes to the P-hydroxyketones, but an acid workup of the products is required.'25 The use of L-proline as an aldolase mimic has produced extraordinary results in cross-aldol reactions of acetone and aldehydes and of dissimilar aldehydes, giving P-hydroxyketones with ee's > 80Y0.'~~-'~' From limited results it appears that L-proline is as active or more active than the catalysts mentioned above (below ambient temperatures, with good yields in <24 h), and exhibits considerN

I

Ph

N

308

Catalysis

able indifference to solvent, although polar, non-hydrogen bonding solvents work best. Catalyst usage is high (10-30 mol%) and reutilization has not been studied. An oxidative decarboxylation at ambient temperature is that of the a-hydroxy, a-trifluoromethyl acids or esters to the corresponding trifluoromethylketones, using bidentate Co-amide catalysts, with yields >80%.128Again, these are slow reactions requiring relatively low reactant/catalyst ( 25/1) ratios, and an aldehyde promoter which is partly oxidized itself. The impetus for low temperature aldol processes in the absence or minimization of organic solvent has resulted in several new approaches. For example, Dewa et al. describe an aldehyde-ketone condensation catalyst that can be suspended in water, prepared by reacting lanthanum tris(isopropoxide) with anthracenebisresorcinol (ABR).'29A polymeric aquo complex with several 1,230 water linked ABRs and two (LaOH)2+groups is formed, coordinating molecules and relatively stable to decomposition. The base strength of this complex approached that of aqueous Na2C03. Phase-transfer catalysis has also been employed in the area of specialty ketone production. C-dialkylated and C-trialkylated ketones can be synthesized from the corresponding alkylacetyl ketone using the alkyl iodides under phase transfer conditions, e.g., with KOH/18-crown-6. These are most useful when another group in the ketone reactant is a furan, thiophene, pyrrole (N-alkylation also takes place here), pyridine, or similar group. Typical temperatures are from ambient to 350 K, and some of the syntheses work with alkenyl and alkylsilane iodide reactants as well as simple alkyl groups.'30 Yields were mostly >50%. While not strictly a condensation, aromatic ketones can also be produced from fluorenes, indenes, xanthenes, etc. by first chlorinating them at the 9-position, then oxidizing at the same position with a strong acid at near ambient phase transfer conditions, with a tetraalkylammonium salt as ~ata1yst.l~~

-

-

-

6

Catalyst Properties- Decarboxylative Condensations

While several studies have compared various metal oxides for decarboxylative condensation reactions of acids and aldehydes, most were limited in scope due to the low conversions and/or partial pressures employed. In particular, the detrimental effects of water, often present in the feeds and always a product of the reactions, have usually been ignored. Regenerability of spent catalysts has also been given scant attention. In previous work using Ce02 and other rare-earth oxide (REO) catalysts, ketonization selectivity was found to be low compared to that of aldolization, cracking or reductive coupling reaction^.'^^^^^^^*'^^ However, in recent work it has been determined that many supported REOs are selective, long-lived decarboxylative condensation catalysts at high conversions, especially when acetic acid is one reactant.'0~12*27-28~40 Among the more common oxides outside the transition metal oxides, it was found that Ce02- or ZrOl-containing catalysts demonstrate the best overall

309

7: Catalysis of AcidlAldehydelAlcohol Condensations to Ketones

behavior for ketonization of acids/aldehydes at high c o n v e r ~ i o n Hi . ~gh~ ~ ~ ~ ~ ~ loading Ce02/A1203materials prepared by incipient wetness impregnation give STYs (space-time yields) in excess of 4 wt ketone/(wt cat-h) for over 14 h continuous operation without regeneration for methylcyclopropylketone and 2-undecanone s y n t h e s i ~ , 2 ~and - ~ ~for 9 ~acetone." ~ STYs in the range 1.8-2.0 have been reported for 3-pentanone, 6-undecanone and 7-tridecanone synthesis.1° With periodic regeneration these materials show no loss in STY over a few weeks of use, and in fact the selectivity improves through the first few regenerations, suggesting that interaction between the R E 0 and the support takes place. Deactivation seems to be associated with coking rather than excessive reduction, because the oxidation state of the pre-reduced (more active) catalyst and the used catalyst are similar. REOs such as Pr203 and Sm203,which are less reducible than Ce02at typical reaction conditions, have been reported to be more selective for the ketonization of but we have found them to be less active and less stable at high conversions. We and others have also tried to optimize the support and preparation method for Ce02-supported catalyst^,'^*'^^^^^^^ both unpromoted and those promoted with alkali or alkaline earths. A1203 appears to be the best support in terms of overall behavior, although Ti02is close. In contrast to suggestions from other work: we and others have found that alkali and alkaline earth promotion does not improve high loading Ce02-based catalysts, neither by eliminating undesired reactions (e.g., lactonization or cracking of acids) nor by decreasing rates of coke formation.65While some enhancement of Z r 0 2 activity with Nadoping has been this appears to be the result of better stabilization of the active tetragonal phase in Zr02. Mixing Ce02 with a d9 or dl0 transition metal (with the possible exception of Co), or with a metal oxide of wider base strength distribution (e.g., CaO), is usually detrimental unless the reactants are resistant to dehydrogenation, isomerization and cracking reactions, e.g., acetic acid. Therefore desirable catalysts (e.g., supported CeO2 or Zr02)appear to be characterized by a relatively amphoteric or weak base nature, a narrow distribution of surface basi~ity,'~~'~-'~ and relatively poor dehydrogenation activities. There is often little relationship between acid/base site distributions in supported MO, and classical measures of cation Lewis acidity/basicity such as electronegativity or the isoelectric point of charge (IEP). However, in general it is true that a cation of high electronegativity is likely to have more acid sites at the 'strong' end of the scale (as measured by heat of adsorption for an amine or NH3) than one of low electronegativity. The reverse is true for basic sites. In terms of relative base strengths, it is generally agreed that, e.g., ZnO >> Ce02 > ZrO2. Relative to typical supports such as A1203 (IEP 7-9) and Ti02(anatase, IEP 6-6.5), supported Z r 0 2 (IEP -6-6.7), Nd203, Ce02 (IEP 6.7-6.8) and ZnO (IEP 8.7-9.7) all have more basic sites per unit area. But these oxides are also characterized by a relative absence of strongly basic sites, especially when compared to M0,'s such as Li/A1203,Ca/A1203or Sr/La203.'6~18*22~'34 Z r 0 2 actually shows some Lewis acid sites by pyridine adsorption at 773 K, and 4-7 p,mol/m2 NH3 a d s ~ r p t i o n , 'while ~ ~ in comparison Ti02 normally gives many more acidic than basic sites upon titrations.20Ce02is similar to Zr02.

-

-

-

-

-

3 10

Catalysis

For a typical catalyst such as supported Ce02, the optimal loading is g Ce02/m2.The results of calculations for the Ce02 loading -0.7-0.9 x corresponding to a monolayer leads to a range of results. For example, assuming equal percentages of (11l), (110) and (100) planes exposed, with half of the Cexf coordinatively unsaturated, gives a surface density of 1.0 x g Ce02/m2, which corresponds to the transition density from mostly two-dimensional rafts to three-dimensional crystallites as determined by TEM and EPR.'36For 75% of the Cex+coordinative unsaturated, a surface density calculation gives a result similar to the density (- 1.7 x calculated assuming that the low-temperature reduction peak in TPR measurements corresponds to removal of 25% of surface oxygens to create a surface similar to Ce203.'37Taking into account the inability of incipient wetness impregnation to fill the smallest pores of the support, and the inaccuracy of the TPR calculations (it is difficult to deconvolute the surface reduction from the bulk reduction peak), it is probable that the optimum Ce02 loading corresponds to near one monolayer of C e 0 2 0 n the supports. Ce02crystallizes in the fluorite (fcc) structure with octahedral coordination of the cations. This structure is maintained upon reduction to at least CeOl.7and sometimes 10wer.'"O-'~~ Metastable phases of varying compositions, all with defective fluorite structures, have been observed, suggesting that the diffusion of oxygen vacancies into the bulk is rapid at reduction temperatures; this is a reason why the surface and bulk reduction peaks in the TPR spectrum overlap. The (defective) fluorite structure is retained even to high degrees of reduction in CeOX/Zr0,.'"O The gradual increase in ketonization activity of these catalysts is probably the result of reduction and spreading of Ce02 from 'islands' to a more two dimensional layer structure, and some intercalation in the support to ultimately form, e.g., CeA103or CeTi03.141-'43 Redox activity in the temperature range of interest is important because the mechanisms of ketonization all require reduction/oxidation of the surface. Reduction of high surface area Ce02 itself (to Ce203)typically begins at 600 K with a 'surface' peak maximum of 750-770 1020-1 100 K, and final stoichiometry at >1100 K, a 'bulk' peak maximum of K of Ce01,75-1.8.1401144-146 The amount of reduction at 723-773 K can vary from 0.01-0.13 O/CeOz, depending on factors such as surface area and heating rate.'39,145-146,147 Reoxidation at 450 K is rapid.'39Partly reduced Ce02surfaces are very strongly dehydr~genating,'~~ and generate alkyl moieties as sometimes hypothesized in the 'ketene' mechanism. This is especially true at lower pressure (of acid or aldehyde) conditions, as present, e.g., in TPR experiments or with alcohol feed^.^^'^' Surface reduction of >1 monolayer of Ce02/A1203with H2 takes place at 560-900 K (peak max. 730-860 K),'38,'433'469'49 and therefore within the range of acid condensation. This is a lower temperature range than for bulk Ce02,and a higher-temperature bulk peak is also sometimes seen. Rogemond et al. report reductions of as high as 0.24 mols O/mol Ce02associated with this peak.'49The total amount of reduction is also higher than for Ce02alone, to CeO1.54-1.58/A1203 at >1100 K.'& Rapid redox cycling ('oxygen storage') experiments to mimic redox

--

-

-

-

311

7: Catalysis of AcidjAldehydejAlcohoI Condensations to Ketones

reactions showed that the supported CeO, would exist at steady-state as Ce01,98-1.99 at 773 K.'433'46 Mixed Ce/REO's (e.g., La, Y, Hf) and Ce/Zr mixed oxides are often more easily reduced than Ce02 itself. These ions are either similar in size to Ce3+ or smaller, providing for increased oxygen mobility. However, the (defective) fluorite structure is retained during the reduction unless the doping exceeds several For degree of reduction vs. temperature, some of the differences in the literature are undoubtedly due to differences in the intimacy of mixing between the oxides. Better reducibility, however, does not necessarily translate into enhanced hydrogenation or ketonization ability. With the exception of La203,most of the other REO's are poorer hydrogenation catalysts than Ce02itself,'51and Ce02is much better than 1/1 Ce02/Zr02for typical ketonizations.28Fornasiero et a1 found that upon repeated redox processes the 1/1 mixture sintered;lMthis may be why Ce/Zr oxides made poorer acid condensation catalysts. The TPR behavior of the mixed oxides supported on A1203is unclear. XRD observations suggest that while reduction takes place, any formation of CeA103 is inhibited by the Zr02.152 If further increases in dehydrogenation activity are desirable in the ketonization process, a d6-d8transition metal can be supported on the active oxide l a ~ e r . ~There ~ . ' ~can ~ be lost yield to saturated h y d r o c a r b o n ~ . ' ~Most ~ ' ~ ~of the commonly used transition metals exhibit limited miscibility with REOs. An exception is C U . ' At ~ ~ambient conditions, the Cu is mostly dispersed as Cu+; some Cu+ can be maintained under reducing conditions even at 723 K.'55When supported on A1203,CeO, suppresses formation of inactive CuA1204up to very high temperature. Of the transition metal oxides, Fe203and Mn02/Mn304appear to be the best single metal ketonization catalyst^.'^-'^^^^ They are also quite effective when used in combination with Ce02in the alcohol to ketone dehydrogenation/condensations, more so than either the d6 or d9 metals.65Both oxides are easily dispersed on A1203.There is some disagreement over the active state of the Mn-containing catalysts. Pestman et al. reported that irreversible reduction to MnO takes place, while most others consider the active form to be a mixture of the IV to I1 formal oxidation states." The interconvertibility of the Mn-oxides at high temperature is probably important in their ability to fulfill the varied requirements of the dehydrogenation/condensation reactions. In contrast to the REOs, Zr02, etc., Nb2O5 and related group VB oxides are highly acidic (IEP O S ) , especially in the presence of water. The dispersion of Nb2O5 is particularly good on A1203,forming a distorted NbO6 octahedral 2-D 1 a ~ e r .This l ~ ~ structure is not dehydrated even upon drying until 770 K, and the acidity is thought to arise from a hydroxyl adjacent to the surface Nb = 0 bond. This and related mixed-metal oxides are poor ketonization catalysts, resulting in products more characteristic of aldolization, and excessive isomerization of the reactant acids and product ketone^.^^-^^ Strong basicity and low M - 0 bond strengths result in bulk acetate formation; pure oxide catalysts of this type are relatively inactive for ketoni~ation.6~"~~~

-

%.138,1409150

-

-

312

Catalysis

Oxides with a wide distribution of acid and base sites (e.g., La203/La(OH)3, Nd2O3, or mixed oxides of the alkaline earths with Ce02) tend to catalyze cracking of aldehydes and ketones:* or at <670 K to give multiple aldol condensation~:~-~ this ~ *usually ~~ results in rapid deacti~ation.'~.~~ The strong bases in addition cannot desorb the product C02. While the key requirement for dissociation of acids or alcohols on surfaces is the presence of coordinatively unsaturated surface metal cations, weakly basic surface sites are also needed to abstract and bind the protons originating from the a d ~ o r b a t e . ~ All ~ - ~of' the above suggests that it is the weakly basic, slightly reducible oxides which should provide the best combination of properties for the acid condensation reaction. The 01,binding energy in XPS is a good indicator of M - 0 bond strength, which in turn is a good indicator of relative activities in reactions such as these where activation of the reactant is rate-limiting. For example, for aldol condensations simple oxides of very high M - 0 bond strength are relatively inactive even if they possess the requisite base strength^.'^.^'

7

Catalyst Properties - 'One-Step' Aldol Condensations

The catalysts for these demanding ketone syntheses must possess at least mild basicity/acidity and a selective hydrogenation component(s). The presence of Lewis acid sites, which act as acceptors for the enolate intermediates that are formed at the basic sites, is beneficial when basic sites predominate. However, the majority of the selective catalysts appear to be mostly acidic in nature; catalysts with strongly basic sites are typically non-selective to except at low (- 310-330 K) temperatures concomitant with low reaction rates and the inability to dehydrate the ketoalcohol (e.g, diamino-functionalized MCM-41, or quaternary ammonium ion exchange p ~ l y m e r ) . ~ ~ . ' ~ ~ Cu/Cr, Cu/Fe, Cu/Zn, Ni/Co, and M o p oxides appear to be acceptable hydrogenation metal pairs. The acid to aldehyde hydrogenation is known to be selective on Cu and Fe oxides.45Cr and Fe oxides by themselves are usually insufficient for either complete dehydration or subsequent hydr0genation,8~.'~~ except at higher tempearatures. Mg/Al hydrotalcites (often, layered double hydroxides derived from hydrotalcites) are typically far more selective for the single aldol condensation and subsequent ketol dehydration than other primarily basic oxide catalysts. Hydrotalcite activities are comparable to all but very strong bases such as KF/A1203. Isomorphous substitution of transition metals for A1 in hydrotalcites does not result in selective hydrogenation at the low temperatures where single aldol condensations are prevalent. Therefore the hydrotalcites are best used for making a,P-unsaturated k e t o n e ~ . ~ ~The - ~ ~correct y ~ ~ ,thermal ~ ~ ~ activation procedure and Mg/Al ratio are important in generating a large number of acidic and Calcination basic sites, and so high activity at low ternperatures.54-56~57-58~92J59-160 at >700 K followed by rehydration is a suitable activation procedure; the optimal Mg) ratio is usually -0.25-0.33. More Lewis acid or more weakly Al/(Al basic Bronsted sites (-OH) typically enhance catalytic activities for both the

+

7: Catalysis of AcidlAldehydelAlcohol Condensations to Ketones

313

condensation and dehydration reactions, but too many Lewis acid sites can reduce selectivity due to multiple aldol condensations and/or Michael additions, e.g., isophorone from acetone. Both site density increases can be effected by increasing the Al/(Al + Mg) r a t i ~ . ~ Layered ~ - ~ ' double hydroxides derived from anions other than the carbonates can exhibit greatly different acid-base properties, e.g., those derived from exchange of some borate or silicate anions are primarily weak Lewis acids.80 Considering single metals rather than alloys, copper functions effectively for the selective hydrogenation of the a$-unsaturated product, in combination with moderately basic mixed oxides such as Ce02,MgO/A1203(high Al/Mg ratio) or K/Mg/CeO,, as long as H2 partial pressures are relatively low. One way to accomplish this is to use partly or wholly alcoholic feeds, which generate their own surface hydrogen. Copper appears to have little effect on acid/base strength distributions, and in particular does not inhibit neighboring basic ~ i t e s . ~ ~ ? ' ~ ~ However, it is difficult to control further aldol condensations with these materials, even at <40% conversion. Recent patents claim that control is enhanced by recycle of reaction intermediates such as mesityl oxide and methyl isobutyl carbinol,78which may function by the selective poisoning of stronger basic sites. The majority of the catalysts quoted in the literature (especially the patent literature) for 'one-step' syntheses use Pd, Ru, Pt, or Ni as the hydrogenation Component,49,77,90-91,93-94,96,98,162-165 with Pd typically the most selective metal. The 'aldol' component of the catalyst can be a basic ion exchange resin, an alkali- or alkaline earth-exchanged zeolite, a mild base supported REO, a hydrotalcite, or even more amphoteric oxides such as A1203or A1POis;95.98.163 however, mostly acidic solids are even more prevalent. Effective solid acids include Nb2O5.xH20, Sn02,phosphates, transition metal halides, H-form zeolites, silicoaluminophosphates (SAPOs), or acidic [usually sulonated poly(styrene-co-divinylbenzene)] ion-exchange resins. Most 'basic' zeolites actually consist of weakly basic and Lewis acid sites, as long as the alkali or alkaline earth ion remains associated with the framework. For example, Cs-exchanged zeolites behave as weaker bases, than, e.g., most hydrotalcite~.~~ The Pd/zeolite or other molecular sieve catalysts have the advantage of easier regeneration when compared to most other catalysts in the list above. Zeolites with almost no ion exchange sites (e.g., silicalite) actually require little added basic oxide to function successfully. For Pd supported on moderately basic oxides such as ZrOz, partial silanization of the surface increased selectivities to the primary ketone These results suggest that strongly basic sites are best neutralized even when using supported Pd, etc. catalysts. The same appears to be true of strongly acidic sites; both types of sites catalyze Tischenko cross-esterification at low (<373 K) temperature^:^ and alcohol dehydration and subsequent oligomerization of the alkene product at higher temperature^.'^^ While there is no question that the stronger acidic and basic oxides can speed up the typically slow aldol step, it is also true that few Bronsted-type acidic or basic sites are necessary. For example, dealumination of Pd/H-FAU or Pd/SAPO-11 molecular sieve catalysts was shown to increase activity by a factor of 5 in the acetophenone to 1,3-diphenylbutan-l-one or by -20% in

3 14

Catalysis

the reaction of acetone to MIBK,'63although in both cases a more open pore structure was also implicated. For Pd supported on the basic Mg/Al hydrotalcites (actually, the mixed oxides, as the hydrotalcite structure was almost completely destroyed), the catalyst activities are high ( - 3 mol/(h.g) at 413 K and WHSV = 500 h-' for the acetone reaction to MIBK), but the selectivities rather l0w.9~Both hydrogenation (isopropyl alcohol) and multiple aldol byproducts were found. While multiple condensation and oligomerization products can be desorbed or oxidized by regeneration in air, a further problem with the supported transition metal catalysts appears to be long-term operation. Obtaining a high dispersion of the metal is important in maximizing the ratio of the alkene to carbonyl rates of hydrogenation. The metal often cannot undergo repeated regenerations without gradual sintering, and so loss of both activity and selectivity. There are competing claims as to which amorphous metal oxide or zeolite support can minimize the sintering. The support can also fulfill a shape-selective function, either by facilitating ketone removal or by preventing the desorption of multiple condensation p r o d u ~ t s . 9For ~ - ~example, ~ the large 12-ring aperture of SAPO-5 (7.3 x 7.3 A) allows for easier ketone desorption, greatly increasing its selectivity toward ketone (vs. alkane) products, when compared to SAPOs with 10-ring apertures. However, the larger aperture also resulted in decreased MIBK/DIBK ratios in acetone condensation. An alternative to these catalysts might be ternary (or even more complex) mixed oxides such as Mo/V/Sn/WO, or V-Si-P catalysts (vanadium and phosphorus oxides supported on SOz, with V/Si <-0.1). These catalysts possess mildly basic sites (the acid sites are more numerous, some of fairly high strength), and have been shown to be selective for anhydride/HCHO, acid/HCHO, and aldehyde aldol condensations at conditions similar to those of other 'one-step' processes.166-168 The key to successful operation again appears to be the elimination of the strongest acidic and basic sites, the concentration of which is highly dependent on the amounts of promoters such as phosphates and group IV-V oxides. Tanner et al. have shown that for the V-P-0 the almost complete elimination of basic -OH groups, associated with surface V4+,improved selectivities to aldol vs. dehydration products such as alkenes and alkanes.

References 1. 2. 3. 4. 5. 6. 7. 8.

T.D. Machajewski, C.-H. Wong, Angew. Chem. Int. Ed., 2O00,39,1352-1374. M. Shibasaki,H. Sasai, Top. Stereochem.,1999,22,201-225. N.J. Lawrence, J. Chem. SOC.Perkin Trans. 1,1998,1739-1749. T. Mukaiyama, N. Iwasawa, R.W. Stevens, T. Haga, Tetrahedron, 1984,40,13811390. S.E. Denmark, S.K. Ghosh, Angew. Chem, Int. Ed., 2001,40,4759-4762. S. Radjadurai, Catal. Rev.-Sci.Eng., 1994,36,385-403. R.L. Cryberg, R.M. Bimber, U.S.Patent 4,570,021(1986). M. Kaufhold, W. Kleine-Homann,Ger. Patent DE 3,637,788 A1 (1988).

7: Catalysis of Acid/Aldehyde/AlcohoZ Condensations to Ketones 9. 10. 11. 12. 13. 14. 15. 16. 17. 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. 44. 45. 46.

315

C. Schommer, K. Ebel, T. Dockmer, M. Irgang, W. Holderich, H. Rust, U.S. Patent 4,950,763(1990). M. Glinski, J. Kijenski, A. Jakubowski, Appl. Catal. A, 1995,128,209-217. R. Pestman, R.M. Koster, A. van Duijne, J.A.Z. Pieterse, V. Ponec, J. Catal., 1997, 168,265-272. M. Glinski and J. Kijenski, Appl. Catal. A, 2OO0,190,87-91. R. Klimkiewicz, H. Teterycz, H. Grabowska, I. Morawski, L. Syper, B.W. Licznerski, J. Am. Oil Chem. SOC.,2001,78,533-535. J.P. Brunelle, Pure Appl. Chem., 1978, 50,211-232. S. Lippert, W. Baumann, K. Thomke, J. Mol. Catal., 1991,69, 199-214. X. Zhang, A.B. Walters, M.A. Vannice, J. Catal., 1995,155,290-302. T. Yamanaka, K. Tanabe, J. Phys. Chem., 1975,79,2409-2411. A. Auroux, A. Gervasini, J. Phys. Chem., 1990,94,6371-6379. S. Luo and J.L. Falconer, J. Catal., 1999,185, 393-407. H. Idriss, E.G. Seebauer, J. Mol. Catal., 2000,152,201-212. K. Parida, J. Das, J. Mol. Catal. A, 2000, 151, 185-192. A. Gervasini, G. Bellussi, J. Fenyvesi, A. Auroux, J. Phys. Chem., 1995, 99, 51175125. M. Fukui, S. Hayashi, T. Okamoto, I. Koga, T. Inui, U.S. Patent 3,966,822 (1976). F. Gonzalez, G. Munuera, J.A. Prieto, J. Chem. SOC.Faraday Trans. 1, 1978, 74, 1517-1 529. T. Imanaka, K. Adachi, Y. Okamoto, S. Terenishi, Nippon Kagaku Kaishi, 1974, 862-866. K.S. Kim, M.A. Barteau, J. Catal., 1990,125,353-375. S.D. Randery, Ceria Based Catalysts for Assymetric Ketone Production, M.S. Thesis, Louisiana State University, Baton Rouge (1999). S.D. Randery, J.S. Warren, K.M. Dooley, Appl. Catal. A, 2002,226,265-280. M.A. Barteau, J. Vac. Sci. Technol. A, 1993,11,2162-2168. M.A. Barteau, Chem. Rev., 1996,96,1413-1430. R. Swaminathan, J.C. Kuriacose, J. Catal., 1970,16,357-362. S.S. Jewur, J.C. Kuriacose, J. Catal., 1977,50,330-341. K. Okumura, Y. Iwasawa, J. Catal, 1996,164,440-448. H. Idriss, C. Diagne, J.P. Hindermann, A. Kiennemann, M.A. Barteau, J. Catal., 1995,155,219-237. A. Yee, S.J. Morrison, H. Idriss, J. Catal., 1999, 186,279-295. M. Alas, M. Crochemore, U.S.Patent 5,600,013 (1997). M. Kaufhold, U.S. Patent 5,763,627 (1998). K.V. Ramanamurty, G.S. Salvapati, J. Sci. Indus. Res., 2OOO,59,339-349. T.S. Hendren, Kinetics of Catalyzed Acid/Acid and Acid/Aldehyde Condensation Reactions to Non-Symmetric Ketones, M.S. Thesis, Louisiana State University, Baton Rouge (2001). T.S. Hendren, K.M. Dooley, Catal. Today, 2003,85,333-351. A. Ozaki, Isotopic Studies of Heterogeneous Catalysis, Kodansha, Tokyo, 1977, p. 168. X.D. Peng, M.A. Barteau, Langmuir, 1989,5,1051-1056. R. Pestman, R.M. Koster, A. van Duijne, J.A.Z. Pieterse, V. Ponec, J. Catal., 1997, 168,255-264. J.E. Rekoske, M.A. Barteau, 1999, Langmuir, 15,2061-2070. X.D. Peng, M.A. Barteau, Catal. Lett., 1992, 12,245-254. M. Ichikawa, CHEMTECH, 1982,674-680.

316

Catalysis

47. 48.

H. Orita, S. Naito, K. Tamaru, J. Catal., 1984,90,183-193. A. Kiennemann, H. Idriss, R. Kieffer, P. Chaumette, D. Durand, Ind. Eng. Chem. Res., 1991,30,1130-1138. G. Heilen, A. Nissen, 0. Woerz, U.S. Patent 4,270,006 (1981). J.M. Vohs, M.A. Barteau, J. Phys. Chem., 2001,95,297-302. K.H. Schulz, D.F. Cox, J. Phys. Chem., 1993,97,647-655. H. Idriss, K.S. Kim, M.A. Barteau, J. Catal., 1993,139, 119-133. H. Tsuji, F. Yagi, H. Hattori, H. Kita, J. Catal., 1994, 148,759-770. A. Corma, V. Fornes, R.M. Martin-Aranda, F. Rey, J. Catal., 1992,134,58-65. K. Koteswara Rao, M. Gravelle, J. Sanchez Valente, F. Figueras, J. Catal., 1998, 173,115-121. J. Lopez, R. Jacquot, F. Figueras, Stud. Surf. Sci. Catal., 2000,130 (Proc. Int. Cong. Catal.), 491-496. M.J. Climent, A. Corma, S. Iborra, J. Primo, J. Catal., 1995, 151,60-66. M.J. Climent, A. Corma, S. Iborra, A. Velty, Green Chem., 2002,4,474-480. A. Corma, S. Iborra, J. Primo, F. Rey, Appl. Catal. A, 1994,114,215-225. K. Yokota, T. Nakamura, Y. Ishii, Kogyo Kagaku Zasshi, 1965,68,690-695. K. Saito, K. Tanabe, Nippon Kagaku Kaishi, 1974,6,1014-1018. K. Terelak, S. Trybula, Technol. Chem. Przelomie Wiekow, 2000,211-213. H. Idriss, K.G. Pierce, M.A. Barteau, J. Am. Chem. SOC.,1994,116,3063-3074. K.G. Pierce, M.A. Barteau, J. Org. Chem., 1995,60,2405-2410. Y. Kamimura, S. Sato, R. Takahashi, T. Sodesawa, M. Fukui, Chem. Lett., 2000, 232-233. J.B. Claridge, M.L.H. Green, S.C. Tsang, A.P.E. York, J. Chem. SOC. Faraday Trans. 1,1993,89,1089-1094. N. Plint, D. Ghavalas, T. Vally, V.D. Sokolovski, N.J. Coville, 1999, Catal. Today, 49,71-77. K.-I. Maruya, A. Takasawa, M. Aikawa, T. Haraoka, K. Domen, T. Onishi, J. Chem. SOC. Faraday Trans., 1994,90,911-917. J. Bussi, S. Parodi, B. Irigaray, R. Kieffer, Appl. Catal. A, 1998, 172, 117-129. S. Sato, R. Takahashi, S. Sodesawa, K. Matsumoto, Y. Kamimura, J. Catal., 1999, 184,180-188. N.D. Plint, N.J. Coville, D. Lack, G.L. Nattrass, T. Vallay, 2001, J. Molec. Catal. A, 165,275-281. D.J. Elliott, F. Pennella, J. Catal., 1989, 119,359-367. H. Madhavaram, H. Idriss, Catal. Today, 2000,6,309-315. J.B. Letts, U.S. Patent 4,739,122 (1988). D.K. MacApline, B.L. Williams, P.S. Williams, P.S. U.S. Patent 4,727,196 (1988). J.I. DiCosimo, G. Torres, C.R. Apesteguia, J. Catal., 2002,208, 114-123. P.Y. Chen, S.J. Chu, C.C. Chen, N.S. Chang, W.C. Lin, T.K. Chuang, U.S. Patent 5,059,724(1991). M.R. Bassett, B.T. Keen, U.S. Patent 5,925,796(1999). J.I. DiCosimo, C.R. Apesteguia, M.J.L. Gines, E. Iglesia, J. Catal., 2000, 190, 26 1-27 5. M. del Arco, S. Gutierrez, C. Martin, V. Rives, Phys. Chem. Chem. Phys., 2001, 3, 119-126. M.V. Chintamani, M.M.V. Kumar, Intl. Patent WOO056688 (2000). R. Mahrwald, B. Guendogan, J. Am. Chem. SOC., 1998,120,413-414. N. Iranpoor, F. Kazemi, Tetrahedron, 1998,54,9475-9480. R.K. Roy, N. Tajima, K. Hirao, J. Phys. Chem. A, 2001,105,2117-2124.

49. 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. 79. 80. 81. 82. 83. 84.

7: Catalysis of Acid/Aldehyde/Alcohol Condensations to Ketones

317

85. G.J. Kelly, F. King, M. Kett, Green Chem., 2002,4, 392-399. 86. A. Nissen, G. Heilen, E. Sapper, W. Fliege, A. Wittwer, U.S. Patent 4,212,825(1980). 87. R. Tanner, P. Gill, R. Wells, J.E. Bailie, G. Kelly, S.D. Jackson, G.J. Hutchings, Phys. Chem. Chem. Phys., 2002,4,688-695. 88. D. Muthusamy, C.-C. Wang, R.D. Swain, D.B. Litzen, W.R. Pledger, U.S. Patent 5,583,263(1996). 89. Gramlich, W.; Lengsfeld, W. U.S. Patent 5,081,309(1992). 90. B.-Y. Coh, J.M. Hur and H.-I. Lee, Proc. 15th North Am. Catal. SOC.Ann. Meeting, Chicago, 1997. 91. P.Y. Chen, S.J. Chu, K.-C. Wu, W.C. Lin, U.S. Patent 5684207 (1997). 92. Y.Z. Chen, C.M. Hwang and C.W. Liaw, Appl. Catal. A, 1998,169,207-214. 93. Y. Higashio and T. Nakayama, Catal. Today, 1996,28,127-131. 94. T. Ushikubo, Y. Kakuno, H. Tsujiura, U.S. Patent 5,149,881(1992). 95. K.D. Olson, U.S. Patent 4,201,562 (1987). 96. P. Magnoux, N. Lavaud, L. Melo, G. Giannetto, A.I. Silva, F. Alvarez, M. Guisnet, Stud. Suf. Sci. Catal., 130 (Proc. Int. Cong. Catal.), 2000,3011-3016. 97. S. Mahmoud, A. Hammoudeh, S. Gharaibeh, J. Melsheimer, J. Molec. Catal. A, 2002,178,161-167. 98. M. Martinez-Ortiz, D. Tichit, P. Gonzalez, B. Coq, J. Molec. Catal. A, 2003, 201, 199-2 10. 99. G.J. James, Intl. Patent WOO187812 (2001). 100. I. Leipprand, T. Lahrs, U.S. Patent Appl. 2002-0055655 (2002). 101. M. Huang, P.A. Zielinski, J. Moulod, S. Kaliaguine, Appl. Catal. A, 1994, 118, 33-49. 102. T.S. Brima, R.M. Hanes, U.S. Patent 4,980,513 (1990). 103. J. Pugach, J.S. Salek, US. Patent 5,004,839(1991). 104. W.A. Beavers, U.S. Patent 5,741,940(1998). 105. O.A. Tarasova, B.A. Trofimov, S.V. Amosova, A.G. Tesler, M.G. Voronkov, J. Appl. Chem. USSR, 1987,60, no. 2, pt. 2,415-416. 106. T.M. Che, U.S. Patent 4,590,298(1986). 107. C. Thebtaranonth, Y. Thebtaranonth, in ‘The Chemistry of Enones‘, ed. S. Patai and Z. Rappoport, John Wiley and Sons, New York, 1989, p.199. 108. M. Arisawa, Y. Torisawa, M. Kawahara, M. Yamanaka, A. Nishida, M. Nakagawa, M. J. Org. Chem., 1997,62,4327-4329. 109. P. Chiu, S.T. Wong, Synth. Comms., 1998,28,4513-4516. 110. A.P.E. York, S. Hamakawa, T. Hayakawa, K. Sato, T. Tsunoda, K. Takehira, K., J. Chem. SOC.Faraday Trans., 1996,92,3579-3586. 111. Duval, D.; Garibaldi, S. in ‘The Chemistry of Enones’ (S. Patai and Z. Rappoport, eds.). John Wiley and Sons, New York, 1989, p.281. 112. C.G. Frost, K.J. Wadsworth, Chem. Commun., 2001,2316-2317. 113. L.J. Goosen and K. Ghosh, Angew. Chem. Int. Ed., 2001,40,3458-3460. 114. L.J. Goosen and K. Ghosh, Eur. J. Org. Chem., 2002,3254-3267. 115. C. Palomo, M. Oiarbide, J.M. Garcia, Chem. - A Eur. J., 2002,8,36-44. 116. Y. Yoshida, N. Matsumoto, R. Hamasaki, Y. Tanabe, Tetrahedron Lett., 1999,40, 4227-4230. 117. Z. Han, S. Uehira, H. Shinokubo, K. Oshima, J. Org. Chem., 2001,66,7854-7857. 118. D.A. Evans, P.J. Coleman, B. Cote, J. Org. Chem., 1997,62,788-789. 119. I. Simpura, V. Nevalainen, Angew. Chem. Int. Ed., 2000,39,3422-3425. 120. N. Yoshikawa, N. Kumagai, S. Matsunaga, G. Moll, T. Ohshima, T. Suzuki, M. Shibasaki, J. Am. Chem. SOC.,2001,123,2466-2467.

318

Catalysis

121. B.M. Trost, H. Ito, E.R. Silcoff, J. Am. Chem. SOC.,2001,123,3367-3368. 122. B.M. Trost, H. Ito, J. Am. Chem. SOC.,2000,122,12003-12004. 123. Y.M.A. Yamada, N. Yoshikawa, H. Sasai, M. Shibasaki, Angew. Chem. Int. Ed. Engl., 1997,36,1871-1873 124. T. Suzuki, N. Yamagiwa, Y. Matsuo, S. Sakamoto, K. Yamaguchi, M. Shibasaki, R. Noyori, Tet. Letts., 2001,42,4669-4671. 125. E.M. Carreira, W. Lee, R.A. Singer, J. Am. Chem. SOC.,1995,117,3649-3650. 126. A. List, R.A. Lerner, C.F. Barbas 111, J. Am. Chem. SOC.,2000,122,2395-2396. 127. A.B. Northrup, D.W.C. MacMillan, 2002, J. Am. Chem. SOC.,124,6798-6799. 128. G. Blay, I. Fernandez, A. Marco-Aleixandre, B. Monje, J.R. Pedro, R. Ruiz, Tetrahedron, 2002,58,8565-8571. 129. T. Dewa, T. Saki, Y. Aoyama, J. Am. Chem. SOC.,2001,123,502-503. 130. E. Abele, E. Lukevics, Chem. Heterocyclic Comps., 2001, 37,5-14. 131. Walters, M.E. Int. Patent W09309082 (1993). 132. H. Idriss, M. Libby, M.A. Barteau, Catal. Lett., 1992, 15, 13-23. 133. K.M. Parida, H.K. Mishra, J. Mol. Catal. A, 1999,139,73-80. 134. P. Carniti, A. Gervasini, A. Auroux, J. Catal., 1994,150,274-283. 135. V. Quaschning, J. Deutsch, P. Druska, H.-J. Niclas, E. Kemnitz, J. Catal., 1998,177, 164-174. 136. A. Martinez-Arias, M. Fernandez-Garcia, L.N. Salamanca, R.X. Valenzuela, J.C. Conesa, J. Soria, J. Phys. Chem. B, 2000,104,4038-4046. 137. M.F.L. Johnson, J. Mooi, J. Catal., 1987,103,502-505. 138. A. Trovarelli, Catal. Rev.-Sci. Eng., 1996,38,439-520. 139. S. Bernal, G. Blanco, M.A. Cauqui, G.A. Cifredo, J.M. Pintado, J.M. RodriguezIzquierdo, Catal. Lett., 1998, 53,51-57. 140. C.K. Narula, L.P. Haack, W. Chun, H.-W. Jen, G.W. Graham, J. Phys. Chem. B., 1999,103,3634-3639. 141. J.Z. Shyu, W.H. Weber, H.S. Gandhi, J. Phys. Chem., 1988,92,4964-4970. 142. A. Dauscher; P. Wehrer; L. Hilaire, Catal. Lett., 14,171 (1992). 143. M. Haneda, T. Mizushima, N. Kakuta, A. Ueno, Y. Sato, S. Matsuura, K. Kashara, M. Sato, Bull. Chem. SOC.Japan, 1993,66,1279-1288. 144. P. Fornasiero, G. Balducci, R. Di Monte, J. Kaspar, V. Sergo, G. Gubitosa, A. Ferrero, M. Graziani, J. Catal., 1996, 164, 173-183. 145. V. Laachir, A. Perrichon, E.C. Badri, J.C. Lamotte, J.E. Lavalley, L. Fallah, F. Hilaire, E. Le Normand, G.N. Quemere, 0. Sauvion, J. Touret, J. Chem. SOC. Faraday Trans., 1991,87,1601-1609. 146. H.C. Yao, Y.F. Yao, J. Catal., 1984,86,254-265. 147. P. Kanta Rao, K.S. Rama Rao, S. Khaja Masthan, K.V. Narayana, T. Rajiah, V. Venkat Rao, Appl. Catal. A, 1997,163,123-127. 148. R.M. Ferrizz, G.S. Wong, T. Egami, J.M. Vohs, Langmuir, 2001,17,2464-2470. 149. E. Rogemond, R. Frety, V. Perrichon, M. Primet, S. Salasc, M. Chevrier, C. Gauthier, F. Mathis, J. Catal., 1997, 169, 120-131. 150. P. Fornasiero, R. Di Monte, G.R. Rao, J. Kaspar, S. Meriani, A. Trovarelli, M. Graziani, J. Catal., 1995,151,168-177. 151. Kh.M. Minachev, Y.S. Khodakov, V.S. Nakhshunov, J. Catal., 49, 1977,207-215. 152. M.H. Yao, R.J. Baird, F.W. Kunz, T.E. Hoost, J. Catal., 1997,166,67-74. 153. H. Grabowska, R. Klimkiewicz, L. Syper, J. Wrzyszcz, Top. Catal., 2000, 11/12, 298-292. 154. R.M. Nix, T. Rayment, R.M. Lambert, J.R. Jennings, G. Owen, J. Catal., 1987,106, 2 16-234.

7: Catalysis of AcidlAldehydelAlcohol Condensations to Ketones

319

155. M. Fernandez-Garcia, E. Gomez Rebollo, A. Guerrero Ruiz, J.C. Conesa, J. Soria, J. Catal., 1997, 12, 146-159. 156. J.M. Jehng, I.E. Wachs, J. Phys. Chem., 1991,95,7373-7379. 157. B.M. Choudary, M. L. Kantam, S.P. Bandopadhyay, F. Figueras, A. Tuel, J. Molec. Catal. A, 1999,142,361-365. 158. L. Nondek and J. Malek, Coll. Czech. Chem. Commun., 1979,44,2384-2392. 159. J.I. DiCosimo, V.K. Diez, C.R. Apesteguia, Appl. Clay Sci., 1998, 13,433-449. 160. W.T. Reichle, J. Catal., 1985,94,547-557. 161. M.J.L. Gines, E. Iglesia, J. Catal., 1998, 176, 155-72. 162. T.J. Huang and W.O. Haag, U.S. Patent 4,339,606 (1982). 163. S.-M. Yang, Y.M. Wu, Appl. Catal. A, 2000,192,211-220. 164. M.R.A. Dutra and R.R. Soares, Proc. 16th North Am. Catal. SOC.Ann. Meeting, Boston, 1999, PII-123. 165. J.J. Spivey, B.W.-L. Jang, R. Subramanian, D. Olsen, T. Devon and R.D. Culp, Proc. 16th North Am. Catal. SOC.Ann. Meeting, Boston, 1999, p. 50. 166. L.J. Velenyi and A.S. Krupa, U.S. Patent 4,380,672 (1984). 167. M.R. Gogate, J.J. Spivey, J.R. Zoeller, Catal. Today, 1997,36,243-254. 168. J.J. Spivey, M.R. Gogate, J.R. Zoeller, R.D. Colberg, Ind. Eng. Chem. Res., 36, 4600-4608 (1997).

8 Turnover Frequenciesin Metal Catalysis: Meanings, Functionalities and Relationships BY JAMES G. GOODWIN, JR., S O 0 KIM AND WILLIAM D. RHODES

1

Introduction

It would be highly desirable to be able to determine the intrinsic site activity of a catalyst in order to evaluate its site efficiency relative to other catalysts active for a reaction of interest - unbiased by the amount of catalyst used, by the concentration of active sites per weight or volume of catalyst, or even, in the ideal, by the concentration of reactants or products. This intrinsic activity would thus reflect the catalytic essence of the reaction site. Use of such a ‘site activity’ would permit discrimination between a poor catalyst candidate and an excellent one for further study that, for whatever reason, might have only a few ‘super active’ sites. One could work to determine the nature/structure of those very active sites and seek a way to design a catalyst having an increased concentration of such sites. Initially, catalyst rates were expressed as rates per weight of catalyst. With the advent of surface area measurement using physisorption (BET), rate per total surface area of the catalyst could be determined, although a comparison of catalysts based on such a measure in general was grossly flawed. Use of chemisorption, especially where one of the reactants of interest was used, offered a more reasonable way to compare rates. In 1968, Michel Boudart [11proposed the use of turnover frequency (TOF,h,,), based on chemisorption, as an excellent way for comparing catalysts. This quantity has units of sec-‘ and represents, ostensibly, the number of molecules reacting per (chemisorption) site per second. This measure has become a standard in fundamental studies of catalytic reactions, especially for metal catalysts. In the last 25 years, use of isotopic tracing has permitted the determination of the average catalyst surface residence time for reaction (z) of a reactant to form a product molecule under reaction conditions for a number of reactions. The reciprocal of this quantity, l/z, has units of sec-’ and, consequently, expresses a form of turnover frequency ( T O F I ~ Although ~). often referred to in the literature (and here) as SSITKA (steady-state isotopic transient kinetic analysis), the use of isotopic tracing only requires for interpretation purposes that reaction not be changing significantly during the isotopic transient period. Recent reviews [2,3, 4) provide details about the technique and references to many research studies using SSITKA. Catalysis, Volume 17 0The Royal Society of Chemistry, 2004 320

8: Turnover Frequencies in Metal Catalysis

32 1

This paper discusses the relationship of the measure of TOF based on chemisorption (TOFchem) to the one from SSITKA (TOFITK) and discusses the relationship of both of these TOFs to the ‘true’ site TOF of a catalyst. The relationship of these TOFs to an ideal site activity is also discussed. TOFITK and TOFchem for several structure-sensitive and structure-insensitive reactions (ethane hydrogenolysis, ammonia synthesis, methanol synthesis, and methanation) on metal catalysts are compared. It should be noted that, for each reaction illustrated, these TOF values originate from the same experiment. The comparisons inherently comply with the guidelines set forth by Ribeiro et al. [S]. This paper also attempts to provide a better knowledge of the accuracy and reliability of TOFchem, an expression of specific rate which is a hallmark of fundamental papers on catalysis by metals and also much easier/cheaper to determine for most reactions than TOFITK.

2

Determinationof TOF Based on Chemisorption

TOF for heterogeneous metal catalysts has been traditionally based on chemisorption due to the ability of this technique to easily and often precisely count the number of surface metal atoms on many metal catalysts [13. Hydrogen and CO chemisorption have especially been used successfully to determine the number of accessible surface metal atoms. Even from the beginning, TOFchem was realized to be more a measure of specific rate rather than a true site activity [l]. Gates [6] has noted that the TOF based on chemisorption (TOFchem) does not accurately represent a true site activity since the catalytically active surface sites and the surface sites measured by chemisorption may not be the same. Determination of a true TOF is complicated for metal surfaces due to the unknown nature of active reaction sites. Many catalytic reactions on metal surfaces are structure-sensitive; the surface atoms on various crystal planes differ significantly in activity and/or the active sites involve ensembles of surface atoms. Even reactions that are structure-insensitive do not seem to occur on all potential sites. In addition, ‘sites’for different reactions on the same catalyst may be quite different in their makeup. Accordingly, it is recognized by most catalytic researchers that chemisorption measurements do not actually determine the number of catalytic sites for reaction [l]. As noted by Ribeiro et al. [SJ, the number of surface metal atoms is the maximum number of possible sites; hence, TOF as determined by chemisorption (TOFchem) represents a lower bound for the true site TOF. In fact, Ribeiro et al. [S] have given the name nominal turnover rate to TOFchemsince the actual number of active sites must be known to identify the true TOF.

3

Determination of TOF Based on SSITKA

The major shortcoming in TOF could be addressed if one were able to determine the actuaI number of active surface sites. Steady-state isotopic transient kinetic

322

Catalysis

analysis (SSITKA), initially developed by Happel [7], Bennett [S], and Biloen [9] and proven to be a useful technique for the in situ kinetic study of heterogeneous catalytic reactions {see references cited in [2,3,4]}, allows determination of the surface concentration of the most active reaction intermediates (Np) under reaction conditions. NP is related to the number of active sites, although some reactant molecules may desorb from active sites prior to reaction and at high reaction temperatures the active surface sites may not be saturated. The value of NP determined does not include intermediates sitting on very low activity sites; however, such sites do not contribute significantly to the overall reaction. By dividing the measured rate of reaction by Np, a TOF based on the concentration of active intermediates (TOFITK) is determined. IN SSITKA, Np and the mean surface residence time of these most active reaction intermediates (zp) are determined. After a step-change between two reactant streams containing different isotopes of a reactant without disturbing other reaction conditions or reaction (as long as an HJD2 switch is not used), the distributions of isotopically labeled products are monitored using a mass spectrometer. zp is first determined by integration of the normalized isotopic transient of a product relative to an inert tracer (usually Ar) that delineates gas phase hold-up (see Figure 1 for the case of methanation). NP is then calculated from Np = ~ p R p

(A)

where Rp is the rate of formation of the product [2,4]. It can be shown that the determination of NP is just based on a simple mass balance and requires no assumptions. Rearranging equation (A) gives Rate of product formed [active intermediates]

4

Relationship of TOFchemand TOFm to Site Activity

Ideally, a site activity would be equivalent to a rate constant and independent of reactant and product concentrations. It would be only a function of the catalyst and temperature. This would permit us to compare measures of such a quantity without having to take into account reactant and product concentrations. It is clear that most specific rates are not concentration dependent. So, it is interesting to consider how much TOFITK (l/zp) and TOFchemdeviates from an ‘ideal’ site activity (ie., to what extent they are dependent on concentration. Consider a simple surface reaction mechanism [lo, 111that can be used for the hydrogenation of CO to produce CH4, an essentially irreversible reaction: (a) CO(g) + * = *CO (b) * c o + * = * c + *o (c) H2 (8) + 2 * = 2 *H (d) *H + *CHx --* *CHx+l * (e) *CH3 + *H + CH4(8) + 2 *

+

[rate-limiting step] [X = 0-21

323

8: Turnover Frequencies in Metal Catalysis

(f) *O + *H = *OH + * (f) *OH + *H = HzO (g) + 2 * where * is a vacant site and *R indicates a molecule adsorbed on a site. Step (d), or perhaps (e), appears to be the rate-limiting step for methanation on Co, Ru, Ni, and Pt. Based on rate-limiting step (RLS) (d), the rate of reaction can be written in terms of surface concentrations as: rate of reaction = rate (RLS) = k [*HI [*CH,]

(C)

It is unimportant what 'X' is; however, CHx will constitute the bulk of the active surface intermediates containing carbon on the surface. Since the mechanism given above appears to be that valid for methanation on Co, Ru, Ni, and Pt, the TOF obtained from hydrogen chemisorption, TOFchem, for this reaction on these metals can, thus, be expressed as:

TOFChem

=

rate of reaction

[sd

=

k -[*HI

CsHl

[*CH,]

where [S,] represents the number of surface sites derived from chemisorption. Thus, TOFche, is a function of [SH], [*HI, and [*CHx], as well as temperature and the nature of the active sites. [*HI and [*CHx] are themselves functions of the partial pressures of the reactants, H2 and CO. This can be compared with the TOF determined by SSITKA. Recall that, in terms of the measured SSITKA parameters, 1

rate of reaction = -Np

(El

ZP

Np includes any adsorbed reactant molecules containing carbon that react plus 1.o

0.8

n

0.6

Ic, W

LL 0.4

0.2

0.0 0

10

20

30

40

50

60

T I M E (sec)

Figure 1

Typical SSITKA transients with area between "CH4 and Ar determining surface residence time ofmethane in tennediates

t,

324

Catalysis

any carbon-containing intermediates and adsorbed product molecules. Since all other reaction steps are faster than the rate-limiting step, [*CHx] can be approximated by Np. Thus,

1 rate of reaction = -Np = k [*HI [*CH,] w k [*HI Np ZP

(F)

Thus, rate NP

--

1 - = k [*HI ZP

Hence, TOFITK is a function mainly of [*HI (or partial pressure of hydrogen) in addition to temperature and the nature of the active sites. zp has, in fact, been shown to vary with H2partial pressure for methanation on Ru [12) and Co [13). Thus, for the simple reaction of methanation given above, TOFITK has, in general, less of a dependency on surface concentrations (and hence partial pressures of reactants) than TOFchem. The relationships of TOFchem, TOFITK, and true ‘site’ TOF are as follows: Rate Rate 5 [conc. surface metal atoms] [conc. active sites] Rate 5 [conc. active intermediates]

TOFChern 5 TOFSite ZS TOFITK

(1) Thus, TOFchem will always be smaller (often significantly S O ) than TOFsite, except where the concentration of metal surface atoms equals exactly the concentration of sites. TOFITK, on the other hand, will always be larger unless the concentration of intermediates (carbon-containing intermediates in the case of methanation) is equal to the concentration of sites. The only possibility of TOFITK ever being smaller than TOFsite is when there is significant readsorption of the product. However, usually in SSITKA experiments, care is taken to minimize readsorption by experiment and/or extrapolation [141. As mentioned earlier, a measure of site activity would, in the ideal, be only a function of the nature of the catalytic sites and the temperature and would be devoid of any dependence on reactant/product concentrations. In other words, site activity in its ideal would be essentially a rate constant, k. Let us now investigate this point for the three TOFs of interest. We will use for comparison purposes the mechanism for methanation given above. Table 1 gives the dependences previously determined for this reaction and assumed mechanism for TOFchem and TOFITK. TOFsite can be written as: rate k[*H][*CHx] TOFsite = - [Ssite] [Ssite]

(J)

where [Ssite] is the total concentration of active sites on the catalyst. [Ssite] is

8: Turnover Frequencies in Metal Catalysis

Table 1

325

Comparison of Dependences of TOFs and Ideal Site Activity for Methanation on Co, Ru, Ni, or Pt Dependence“

a

no Yes

no little

Yes

some

Assumes the mechanism given in the text.

probably somewhat greater than [*CHx]. [Ssite] might approach the value of [*CHx] for low reaction times before deactivation becomes significant, especially since the concentration of surface hydrogen is relatively low; however, there is usually a significantly amount of adsorbed CO during CO hydrogenation, most of which does not react. Much of this CO is probably adsorbed on nonactive metal sites. As can be seen in equation (J), in addition to being a function of temperature and the catalyst via k, TOFsite is also a function of [*HI (and hence the partial pressure of hydrogen) and to some degree [*CHx] (or the partial pressure of CO). Table 1 compares the dependences TOFchem, TOFITK, and TOFsite with that of the ideal site activity. As Table 1 clearly shows, all TOFs, even the ‘true’ or ‘site’TOF (TOFsite), are functions of reactant partial pressures to some degree. Thus, by its nature a TOF is not an ‘ideal’ measure of absolute site activity since it is not an intrinsic property of the catalyst. However, as long as comparisons are made for the same reaction conditions, it is a good relative comparison of site activity. TOFITK is somewhat closer to an ‘ideal’ site activity, at least for this example of methanation, than TOFsite since it has little dependence on the partial pressure of CO. In the next section, the values of TOFchem and TOFITK for a number of reactions will be compared.

5

Comparison of TOF-,

and TOFInrfor Actual Reactions

5.1 Methanation: a Classic Structure-insensitive Reaction. - Structure-insensitive reactions on a particular metal catalyst occur at about the same rate per exposed metal atom on metal particles of various sizes and on single crystals with different crystallographic orientations. Hydrogenation of an olefin is a good example of a structure insensitive reaction [11. Methanation, where dissociatively adsorbed CO is hydrogenated to CH4 and water, also appears to be essentially structure insensitive. These reactions are believed to occur possibly at a single metal atom on metal surfaces, although this is not absolutely certain. Catalysts for Fischer-Tropsch synthesis are often studied under methanation conditions

Catalysis

326

(high H2/C0 ratio) to understand the effects of promoters and supports in the absence of significant deactivation. Utilizing SSITKA as well as standard reaction, methanation on Ru, Co, Fe, Ni, and Pt catalysts has been extensively investigated [2,4]. Ru and Co are the most active CO hydrogenation catalysts [151. Ni and Fe have good activities,while Pt can be considered to be a relatively poor catalyst for this reaction. 5.1.1 Ru. 5.1.1.1 Effect of temperature and hydrogen pressure. Bajusz and Goodwin [12] studied the effect of hydrogen partial pressure and temperature on the surface coverage and activity of reaction intermediates on Ru/SiOzduring methanation. Both standard reaction and SSITKA data were collected (see Table 2). As Figure 2 shows, both the TOF based on H2 chemisorption (TOFH) and TOFITKwere relatively constant with TOS at 270°C and P H = 0.45 bar. The TOF ratio, shown also in Figure 2, was approximately constant at ca. 11 for 80 min. of reaction. For a hydrogen partial pressure of 0.18 bar (H2/CO = 5), the TOFs increased with temperature as would be expected (Figure 3). At this lower partial pressure of hydrogen, the TOF ratio was essentially constant over the whole temperature range (240-270°C),averaging ca. 24. However, at a higher

Table 2

Eflect of Temperature and Partial Pressure of H2 on Reaction and SSITKA Parameters for Methanation on RulSiO2 [ l 2 ]

TOS (min)

Part. Pres. ofH; (bar)

TOFHb Rate (polel ZM gls) s-l) (s)

240 250 260 270

5 5 5 5

0.18 0.18 0.18 0.18

0.6 0.9 1.2 1.7

5.6 8.4 11.2 15.9

7.4 5.6 3.2 2.8

0.14 0.18 0.3 1 0.36

4.4 5.0 3.8 4.8

4.1% 4.7% 3.6% 4.5%

240 250 260 270

5 5 5 5

0.72 0.72 0.72 0.72

1.4 2.3 3.6 5.3

13.1 21.5 33.7 49.6

3.8 3.6 3.1 2.6

0.26 0.28 0.32 0.38

5.3 8.3 11.2 13.8

5.0% 7.8% 10.5% 12.9%

240 240 240 240 240 240

5 5 5 5 5 5

0.14 0.18 0.25 0.35 0.45 0.72

0.4 0.5 0.6 0.8 1.0 1.4

3.7 4.7 5.6 7.5 9.4 13.1

8.8 7.4 6.1 5.0 4.3 3.8

0.1 1 0.14 0.16 0.20 0.23 0.26

3.5 3.7 3.8 4.0 4.3 5.3

3.3% 3.5% 3.6% 3.7% 4.0 %

270 270 270 270

5 25 50 77

0.45 0.45 0.45 0.45

2.9 2.5 2.4 2.3

27.2 23.4 22.5 21.5

4.0 3.9 3.7 3.9

0.25 0.26 0.27 0.26

11.6 9.8 8.9 9.0

10.9% 9.2% 8.3 % 8.4%

Reaction Temperature

("C)

TOFITK N M Surface ( p o l e l Coverage

(l/tM)

(s-9

s)

(%)

5.0%

"Pc0 = 0.036 bar. Number of exposed Ru surface atoms determined by H, chemisorption = 106.8 pmoles/g cat.

8: Turnover Frequencies in Metal Catalysis

Y

g c

327

-P-20x104

g

0.2

-

0.1

-

LL

e

0

20

40

60

80

TO (min)

Figure 2

Methanation on Ru/Si02 (TOS for T = 270°C and PH= 0.45)

partial pressure of H2 (PH = 0.72 bar, HJCO = 20) (Figure 4), while the TOFs increased with temperature as would be expected, their ratio actually decreased from 20 to 8. In Figure 5, it is apparent that both TOFs were direct functions of the hydrogen partial pressure. The TOF ratio decreased steadily from 30 to 20 as the hydrogen partial pressure was varied from 0.14 to 0.72 bar. The cause for this variation in TOF ratio was the effect of HZ partial pressure on the concentration of carbon surface intermediates (Figure 6) since that has a significant effect on TOFITK. While that concentration was essentially constant with temperature for a partial pressure of hydrogen of 0.18 bar, it was a strong function of temperature at the much higher partial pressure of of 0.72 bar, varying from 5 to 14 pmoles/(g cat) as the temperature was increased from 240 to 270°C. 5.1.1.2 Eflect of rnodiJiers. Chen and Goodwin [16] studied the effect of Cu decoration on surface abundance and intrinsic activities of the intermediates on Ru/Si02for catalysts having Cu/Ru atomic ratios of 0-0.5. The reaction rates as well as the abundances ( N M ) and lifetimes (zM) of CH4 surface intermediates at three H2/CO ratios for methanation on Cu/Ru/Si02 are shown in Table 3. For methanation over a wide range of reaction conditions, TOFITK ( l / z ~ and ) TOFchem(TOFH) differed by ca. one order of magnitude. The ratio of these values, the TOF ratio, is plotted as a function of catalyst modifier loading for the three reaction conditions in Figure 7. As shown in Figure 7, there was a steady

Catalysis

328 1

18x103 16~10~ 14~10~ n

5

h

E

12x103 10x10~ 8x103 6x10' 4x1O 3 2x10 *

0

,

-1

0.4 n

Ta

Y

0.3

Y

c

c

E

0.2 0.1 0.o

30 25 20 15

lo! 235

.

.

240

245

I

250

. 255

.

.

I

265

270

275

I

260

Temperature (OC)

Figure 3

Eflect oftemperature on methanation over Ru/Si02 (PH = 0.18)

decrease in the TOF ratio with an increase in Cu loading, except for H2/C0 = 15. Upon close examination, this was mostly due to a decrease in TOFITK. 5.1.2 Co. 5.1.2.1 La promotion. Vada et al. [17) applied SSITKA to investigate La3+promotion of 20 wt% Co/A1203 (La/Co atomic ratios of 0-0.1). Unlike Cu modification of Ru and probably Co, La3+ promotes methanation on Co catalysts. The data from the work by Vada et al. [17) is reproduced in Table 4, and the TOF ratios are plotted in Figure 8. Compared to Cu/Ru/Si02, TOFITK was more constant than TOFchem(TOFH)over the range of La composition and reaction conditions. However, it is interesting to note that for Co/A12O3 both TOFITK and TOFchm(TOFH)went through a maximum with increasing La/Co due to the promoting nature of La3+for CO hydrogenation on a Co catalyst. The presence of La3+initially increased the concentration of active sites as well as the reactivity of the most active sites resulting in an increase in the rate of reaction. The rate, however, decreased at the highest La3+ loading due to blockage of active sites becoming significant. The TOF ratio over the range of promoter loading and reaction conditions studied (see Figure 8) reflects an order of magnitude difference in TOFITK and TOFchem.

8: Turnover Frequencies in Metal Catalysis

329

50x103 r -

40x103

(II

Y

30x103 LL

g

20x103 10x103 0

0.3

0.1

0.0 25

235 240 295 250 255 260 265 270 275

Temperature (OC)

Figure 4

Effect of temperature on methanation over RujSi02 (PH = 0.72 bar)

La3+promotion of Si02-supported Co (20 wt% Co, La/Co atomic ratios of 0-0.75) was studied by Haddad et al. [18). Some of the results from this study are shown in Table 5. With the initial addition of La3+, the rate of reaction and TOFchm(TOFH)increased. However, the value of TOFITK( l/zM)did not change significantly suggesting formation of new active sites rather than an increase in activity of the sites. This is seen in the increase in number of active intermediates with La/Co ratio. As indicated by the TOF ratio in Table 5, T O F I T K was approximately an order of magnitude larger than TOFH. 5.1.2.2 Ru promotion. Kogelbauer et al. [191 applied SSITKA to investigate Ru promotion of Co/A1203catalysts for Fischer-Tropsch synthesis. Table 6 shows the resultant methanation data for Ru-promoted 20 wt% Co/&03. TOFITK (l/zM)was again about an order of magnitude larger than TOFchemdetermined from hydrogen chemisorption (TOFH). The TOF ratio was about the same (between 5-6) for the two different Ru/Co atomic ratios (0,0.015). The results of this CO hydrogenation study complement those noted above, suggesting that

Catalysis

330 J

12x104 -

-

n

'

F

(I)

I

-

8 ~ 1 0 ~ ~

4x103

A

0

--

0.3

-

0.2

-

0.1

-

r

'cr,

W

k

8I-

i

0.0

'

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Partial Pressure of H2 (bar)

Figure 5

Efect of H 2 Partial Pressure on methanation over Ru/Si02 ( T = 240°C)

the ratio of TOFITKto TOFchemis fairly robust over a wide range of experimental conditions and catalyst compositions. 5.1.2.3 Nature of silica support. Panpranot et al. [20] have studied the use of MCM-41 as a support of Ru-promoted Co FTS catalysts. In the study, a comparison was done of silica- and MCM-41 supported catalysts using SSITKA of methanation. As shown in Figure 9, the results for Ru-Co/Si02 were typical for supported Co catalysts. The value of TOFchem(in this case TOFH) declined with TOS as the catalyst partially deactivated during the initial reaction period. This was due to the rate decreasing while the number of surface atoms determined by chemisorption on the fresh catalyst remained constant. TOFITKremained essentially constant during the whole TOS, since the rate declined as the number of surface intermediates decreased. The TOF ratio for all TOS varied between 7 and 14. The TOFs for methanation on Ru-Co/MCM-41 behaved qualitatively similar (see Figure lo), except the value for TOFHwas much higher than normal for a Co catalyst at this temperature. This can be seen in the plot of the TOF ratio. The TOF ratio was much lower than typical for Co catalysts, ranging

331

8: Turnover Frequencies in Metal Catalysis

i

+PH = 0.16 bar

..o

........

12 l4

...O.'.'

"1 4

0 235

240

245

250

255

260

265

270

275

Temperature (OC)

Figure 6

Eflect of T and PH on N Mfor methanation on Ru/Si02

Table 3

CO Hydrogenation on Cu-Promoted Ru/SiOza 1161

CulRu (atomic)

Hydrogen Rate of CO Chemisorption Conv. (pmol H atomslg) (pnol/g/s)

TOFH (1 0-3 s-')

N M d

(pmol/g)

TOFU-K (1hM) (1 OP3 SKI)

{ HJCO = 5 ) 0.00 112.8 0.05 96.4 0.10 60.4 0.20 39.0 0.50 23.0

0.64 0.55 0.40 0.27 0.11

5.9 5.7 6.6 6.9 4.8

7.9 8.4 7.8 5.9 3.6

77 66 46 46 27

{ H 2 / C 0= 10) 0.00 112.8 0.05 96.4 0.10 60.4 0.20 39.0 0.50 23.0

1.04 0.9 1 0.61 0.48 0.17

9.6 9.4 10.1 12.3 7.4

7.7 9.6 7.2 6.9 3.8

129 92 78 70 40

(H2/CO=15) 0.00 112.8 0.05 96.4 0.10 60.4 0.20 39.0 0.50 23.0

1.49 1.29 0.82 0.65 0.23

13.8 13.3 13.6 16.6 10.0

9.8 10.6 8.5 5.5 2.4

143 118 90 90 87

a

Percent dispersion of Ru (Cu/Ru = 0.00) = 36%, d, = 2.4 nm. Static hydrogen chemisorption at 77 K. Reaction conditions: T = 240°C,P, = 1.8 atm, P , = 0.036 atm. Surface concentration of CH, intermediates.

Catalysis

332

0

1 0.0

0.1

0.2

0.3

0.4

0.5

0.6

Figure 7

CdRu (atomic ratlo) Methanation at 240°C on Cu/Ru/Si02

Table 4

CO Hydrogenation on La3+-PromotedCo/AE2O3[ I 71

Hydrogen Chemisorption" Rate of CO LalCo (molH C0nv.b (atomic ratio) atomslg) (pmollgls)

{ &/CO 0.00

TOFm TOF,y s-')

N M

(l/rM)

(pmollg)

( 1 O - j s-I)

=5 )

193 167 130

0.63 1.49 0.49

3.3 9.0 3.8

9.3 11.4 3.5

34 60 53

{ H 2 / C 0= 10) 0.00 193 0.05 167 0.10 130

1.23 2.90 0.92

6.4 15.9 4.4

9.6 15.9 4.4

77 97 100

{ HJCO = 1 5 ) 0.00 193 0.05 167 0.10 130

1.71 3.61 1.28

7.8 12.8 5.2

7.8 12.8 5.2

141 161 132

0.05 0.10

a

Static hydrogen chemisorption at 373 K. Reaction conditions: T = 220"C, P, = 1.8 atm, P,, =0.036 atm, TOS = 5 min.

between 1.5-2.4.The authors concluded, based on this and other evidence, that the use of MCM-41 resulted in some H2 chemisorption suppression at 373 K since TOFITKwas essentially identical for the MCM-41 and Si02 supported catalysts. 5.1.3 Fe. Recently, Sudsakorn et al. [21] studied the effect of activation method on methanation at 280°C (1.8 atm, H2/C0 = 10, Pco= 0.036 atm) on an Fe FTS catalyst. It was found that hydrogen pretreatment resulted in a catalyst

8: Turnover Frequencies in Metal Catalysis

333

42-

0.00

0.02

0.04

0.08

0.06

0.10

0.12

LdCo (atomic ratio)

Figure 8

Methanation at 220°C on La-promoted Co/AI2O3

Table 5

CO Hydrogenation on La3+-PromotedCo/Si02[ l 8 ]

LalCo (atomic)

Hydrogen Chemisorption" Rate of CO (pmole H2/g Conv. at SSh Co) (pmole/gCo/s)

TOFH (1 0 - 2 s-')

N M

(IIGf)

0.00 0.10 0.30 0.75

225 361 450 482

1.1 2.2 2.7 2.5

38 119 154 153

13 13 16 16

a

5 16 24 24

TOFITK

11.8 5.9 5.9 6.4

Static hydrogen chemisorption at 373 K. Reaction conditions: T = 220°C, PT = 1.8 atm, HJCO/He = 20/2/78 cc/min.

Table 6

Methanation on Ru-Promoted CoIA1203 [ I 91

RulCo (atomic)

Rate of CO Conv.a (woi/g/s)

TOFH'

(lIzM

(10-3

(10-3

0.00 0.015

7.1 26.6

15 20

a

TOFI TK (pmol/g) (1 0 - 2 SKI) TOFH

TOFi TK s-l)

TOFITK ~

1

93 95

Reaction conditions: T = 220 "C, PT = 1.8 atm, HJCO/He Based on static hydrogen chemisorption at 373 K.

)

6.2 4.8 =

20/2/78 cc/min.

that had higher activity for methanation than when the catalyst was pretreated with either CO or syn gas. However, since the hydrogen-pretreated catalyst also had an increased chemisorption ability, TOFchem based on CO adsorption for the hydrogen-pretreated catalyst was actually less than that for the CO-pretreated one (Figures 11-12).TOFITK on the other hand was essentially the same

Catalysis

334

A

40x10'

1

30x10'

-

20x104

-

lOxlO-3

-

r

0

W

P

4

0 0.4

-

0.3

-

0.2

2 --

0.1

-

n

r I

5 Y

E

e

0.0

0

50

100 150 200 250 300 350

TOS (min.)

Figure 9

Methanation on Ru-ColSi02

at ca. 0.15 s-l for both the hydrogen- and CO-pretreated samples, suggesting that the active sites on the differently pretreated samples must have been identical. The TOF ratios for the differently pretreated samples remained relatively constant with TOS at ca. 5.2 (CO-pretreated) and 10 (H2-pretreated). Ni. Although Ni was one of the first catalysts for methanation to be studied by SSITKA, not a lot of data is available for comparison of TOFs. However, Yang et al. [22] studied methanation on a 60 wt% Ni/Si02 catalyst and presented data suitable for making a comparison. At 225°C and H*/CO = 3, the TOFchem based on hydrogen chemisorption was 3 . 2 ~ 1 0 -s-' ~ while the TOFrTKcan be calculated to be 1 . 1 ~ 1 0 s-'. - ~ Thus, the TOF ratio was 34. 5.1.4

5.1.5 Pt. Methanation on Pt catalysts have been studied a number of times. Yang et al. [22] studied methanation on a 2 wt% Pt/Ti02 catalyst reduced at 225°C to avoid SMSI and chemisorption suppression. At the reaction temperature of 225°C and for H2/CO = 3, the TOFche, based on CO chemisorption was

8: Turnover Frequencies in Metal Catalysis

335

150x1O 3 n

-

r

cn 100x10-3 I

' LL

5Ox1C3

0 0.4 n P

8

0.3

Y

g

s1

0.2

0.1 0.0 5

I

st 4 F 3

'LI2

F

1 0

Figure 10

Methanation on Ru-CoIMCM-41

5.8~10-~ s-' while the TOFITK (calculated by l/z) was 0.5 s-'. Thus, the TOF ratio was 862. This might seem initially in error since all the TOF ratios for all the other metal methanation catalysts presented have been in the range of 6-30. However, in a more recent study by Bajusz et al. [23], this result was confirmed. Bajusz e t al. studied methanation over a 4.5 wt% Pt/Si02catalyst at 393°C and H2/CO = 12 using SSITKA. Figure 13 shows the variation in TOFs with TOS. The catalyst was found to decrease in activity with TOS. Consequently, TOFchem based on HZchemisorption on the fresh catalyst also decreased. TOFITK, however, remained relatively constant with TOS since it was based on the concentration of active intermediates at a given time. The TOF ratio increased somewhat with TOS from ca. 100 to above 200. Thus, methanation on Pt appears to produce a relatively high TOF ratio. 5.1.6 Methanation Summary. Based on extensive data, it would appear that active methanation metals (Ru,Co, Fe, and Ni) give TOF ratios

Catalysis

336

1 30x1O 3

8

20x104

L

J

P

10x10-3

0 0.20 y

0.15

t LL

g

0.10

0.00 --0 -5 -1 -- 8

0

2

6-

eF 4-

ti0

20

A 0

Figure 11

2

4

6

8

1

0

1

2

1

4

TO (h) Methanation on CO-Pretreated Fe at 280°C

(TOFITK/TOFChem)in the range of 6-30. It is noteworthy that the TOF ratio is approximately the same order of magnitude for a wide variety of Cu/Ru, La/Co, and Ru/Co loadings and for different H2/C0 ratios. It can be concluded that, when methanation is carried out on excellent methanation catalysts under a variety of experimental conditions and for different catalyst formulations, the TOFITK is approximately an order of magnitude larger than that determined based on chemisorption (TOFchem).Only, some form of chemisorption suppression, such as appears to be the case for Co on MCM-41, can cause this to be lower (ca. 1-2). The relatively poor methanation catalyst, Pt, on the other hand gives TOF ratios in the 100-800 range, an order of magnitude higher than those for the good catalysts. It is interesting to note that Pt shows surface coverages of intermediates of less than 1YOwhile the more active catalysts exhibit much higher surface coverages (3- 15 YO). 5.2 Methanol Synthesis. - Although some studies have suggested that MeOH

8: Turnover Frequencies in Metal Catalysis 25x10"

337

4

20x104

8

15x1W3

LL 10x10"

0.20 h r I

d

0.15

Y

f

2

0.10

0.00 14

0

5

10

15

20

25

TOS (h)

Figure 12

Methanation on H2-Pretreated Fe at 280°C

synthesis is dependent on metal particle size, this reaction is not considered to be strongly structure dependent. Vada and Goodwin [24j analysed Li+ promotion of MeOH synthesis on Pd/Si02. The aim of the study was to develop a better understanding of how Li+ promotion affects MeOH synthesis on Pd. Table 7 summarizes the results obtained therein that pertain to the issue being discussed. Vada and Goodwin noted that due to methanol readsorption on non-active or less active sites and/or adsorption to higher coverage on the active sites, the surface concentration of active intermediates ( N M e O H ) is usually a function of the partial pressure of methanol. Since readsorption of the reaction products occurs, TMeasured is the sum of the time for surface reaction (T& and the time spent in a readsorbed state (TRead). Thus, it is necessary to compare T M ~ O Hat the same P M e O H . As one can see in Table 7, when a comparison is made at similar low values of PM~OH in the reactor effluent, the TOFco differs from the estimation of TOFITK (f/TMeOH) by approximately one order of magnitude, similar to the structureinsensitive methanation reaction. The impact of product readsorption on TMeOH has been examined in greater

Catalysis

338 8x1O 3

6x10-3 I

$ c

4x103

2x10-3

0 n

1.2 Y

t

$

U

0.8

c 0.4

0.0

300 L= O 200

t

t 100

0

20

0

40

60

80

100

TOS (min)

Figure 13

Methanation on Pt/Si02 at 665°K

Table 7

Methanol Synthesis on Li+-Promoted Pd/Si02 [24]

Pd Li-Pd Pd Li-Pd

0.95 1.21 0.95 1.21

6.5 6.5 10 22.7

91 155 43 43

12 9 14.5 17.2

12.6 7.4 15.3 14.2

Pd = 5 wt% Pd/SiO,, atomic ratio of Li-Pd = (Li/Pd),, = 1. Based on static CO adsorption at 25°C. Reaction conditions:T= 220°C, P,,=0.52 atm, P, = 1.28 atm.

a

8: Turnover Frequencies in Metal Catalysis

339

8x10"

cn

Y

4x10" 2x10.3

-

n

r

u)

Y

0.6

t L

0.4

0.2

1

250

r

8

200

5t

150 100

i

50 0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

CdRu Atomic Ratio

Figure 14

Eflect of Cu modijcation of Ru/Si02 on ethane hydrogenolysis

detail in recent studies by Goodwin and coworkers. Correction of the effect of methanol readsorption on TMeOH was the focus of a study by Ali and Goodwin [14]. Table 8 shows the results from this study. The table lists two average surface residence times, TMeOH (measured value) and ToOMeOH (TMeOH corrected for readsorption), where

TooMeOH

= TRxn

CLI

Under the conditions of study, %O0MeOH (TRxn) was approximately one half the value of TMeOH. This difference results in the T O F ratio based on TMeasured ((I/TMeOH)/TOFCO)to be approximately one half of the value based on TRxn ((l/ToOMeOH)/TOFCO). It is important to note that once the readsorption effect has been subtracted out, l/ToOMeOH ( T O F I ~is~approximately ) an order of magnitude larger than TOFco (TOFchem), in good agreement with what was observed for methanation on Ru, Co, Fe, and Ni catalysts. Under similar reaction conditions,

a

0.07"

0.007d 0.054d 0.028d

80

48 154 103

0.875 0.146 0.351 0.272

126 153 230 156

MeOH Synthesis on Supported Pd Catalysts

70 96 83 102

Pd weight percents: Pd/SiO, = 5 wt%, Pd/NaY, Pd/NaHY, and Pd/HY = 4 wt%. Static CO chemisorption at 22°C. Reaction conditions:T = 220"C, P, = 1.8 atm, HJCO/He = 16/2/18 cc/min. Reaction conditions:T = 220"C, PT = 1.8 atm, HJCO = 24/2/10 cc/min.

Pd/Si02 Pd/NaY Pd/NaHY Pd/HY

Table 8

7.94 6.54 4.35 6.4 1

14.3 10.4 12.0 9.8

9.1 44.8 12.4 23.6

71.2 34.1 36.0

16.3

c251 c251 c251

c141

8: Turnover Frequencies in Metal Catalysis

341

Kim and Goodwin [25] studied MeOH synthesis on zeolite Y supported Pd catalysts. The study focused on the effects of Na+ cations and the concentration of acid sites on PdY catalysts. Table 8 also contains the results from this study. The TOF ratio (TOFITK/TOFChem)was in the range of 30-70, somewhat higher than for Pd/SiO2 and possibly due to their lower activities. This may be related to Pd being in a less desirable oxidation state in the zeolites. It has been suggested that this is key in determining Pd activity for MeOH synthesis [26,27,28,29].

5.3 Ethane Hydrogenolysis: a Classic Structure-sensitive Reaction. - Ethane hydrogenolysis, commonly used as a model to study carbon-carbon bond rupture, is very structure sensitive [30, 31, 321. Specifically, the metal catalyst activity depends on the spatial coordination of surface metal atoms [30] and appears to require site ensembles of ca. 12 metal atoms [31,32]. As a result, the TOFchem is significantly affected by the size and morphology of the supported metal crystallites that cause a significant variation in the distribution of the surface crystal planes exposed. The reaction mechanism for ethane hydrogenolysis is thought to involve dissociative chemisorption of both ethane and hydrogen on the surface, dehydrogenation of the adsorbed di-carbon species, breakage of the di-carbon intermediates into mono-carbon species, and hydrogenation of the mono-carbon intermediates into methane [33]. Goodwin and coworkers [33,34, 35, 361 applied SSITKA in studies of ethane hydrogenolysis on Ru/Si02 and Co/Si02. These studies provide measured reaction parameters (rates, intrinsic activities, lifetimes of surface intermediates of reaction, etc.) over a significant range of experimental conditions 5.3.1 Ru. In addition to elucidating the surface kinetics of the reaction and the heterogeneity of the active sites on the catalytic surface, studies of ethane hydrogenolysis on Ru catalysts included examination of the effect of temperature on abundance, intrinsic activities, and lifetimes of surface reaction intermediates on Ru/Si02 [33] and the effect of decoration of Ru by a modifier (Cu) on surface abundance and intrinsic activities of the intermediates [34]. Table 9 illustrates the data obtained for ethane hydrogenolysis on Cu/Ru/Si02, based on a study reported by Chen and Goodwin [34]. The data indicate that, while TOFITK ( l/zM)was essentially constant ( f7%), TOFH decreased significantly with Cu modification. For this structure-sensitive reaction, the ratio of TOFITK to that obtained based on hydrogen chemisorption (TOFH)was approximately 100-250 (Figure 14). Surface coverage of ‘available’ surface Ru atoms increased from 3% to 6.5% as the Cu loading on the catalyst increased. 5.3.2 Co. Studies were made into the modification of the catalytic properties of a 20 wt% Co/Si02 catalyst during secondary aqueous impregnation and drying [35] and as a result of La3+-promotion[36]. Table 10 shows the data for ethane hydrogenolysis on La3+-promoted 20 wt% Co/Si02, based on work reported by Haddad et al. [36]. The TOFITK was two to three orders of magnitude larger than that determined with hydrogen chemisorption, with the TOF ratio varying from 294-744 at 260°C. Furthermore, the ratio was approximately

342

Table 9

Cu/Ru (atomic) 0.00

0.05 0.10 0.20 0.50 a

Catalysis

Ethane Hydrogenolysis on Cu-Promoted Ru/Si02a [34] Hydrogen Chemisorptionb (PnolH Rate C2H6' atomslg cat) (nmol/g/s)

TOFH (1O-j s-')

ZM

112.8 96.4 60.4 39.0 23.0

7.0 4.2 3.3 3.2 2.6

1.50 1.44 1.64 1.58 1.66

759 405 200 122 60

TOFJl-K (s)

(1IZIM)

(1 0-j C J )

670 690 610 630 600

Percent dispersionof Ru (Cu/Ru = 0.00)= 36%, average Ru particle size = d, = 2.4 nm. Static hydrogen chemisorptionat 77 K. Reaction conditions:T = 18OoC,P, = 2.0 atm, HJC,H,JHe = 6/0.15/43.85 cc/min.

the same order of magnitude as was observed for ethane hydrogenolysis on Cu/Ru/Si02. The trend with modifier loading was also similar to that for Cu/Ru/Si02 with the TOF ratio monotonically increasing and approximately doubling as La3 loading increased (Figure 1 3 , while TOFITK (l/zM) remained fairly constant. The results for different pretreatments of Co/Si02 gave similar results, with the TOF ratio being ca. 300 [35]. The surface coverage in active intermediates was ca. 1 %. +

5.3.3 Ethane Hydrogenolysis Summary. For both Ru and Co catalysts, the TOF ratios were in the 100-800 range. This is not surprising since, assuming the experimentally determined 12 surface atom site, the ratio of TOFITK/TOFchem would be >12 even if all 12 atom ensembles were active and occupied by intermediates at all times. It is well known that the activity of different crystal planes of metal for ethane hydrogenolysis are orders of magnitude different. This is due to the fact that not just any 12 atoms are sufficient to form a site - they must be properly oriented. Therefore, adding in the crystallographic plane specificity, the TOF ratio manifested could be expected to reasonably exceed 100. The TOF ratio increased with addition of La or Cu modifiers since these modifiers tended to decorate the active metal surface. It is suggested that the major factor determining the increase in this ratio with modifier addition is the sequential faster removal (i.e.,blockage) of available surface ensemble reaction sites by the addition of surface modifying species than the sites available for chemisorption. This effect tends to reduce TOFchem more significantly than TOFITK. As shown in Tables 9 and 10, the initial addition of the modifier, as would be expected, had the most pronounced effect on reducing TOFchem (TOFH).This leads to the speculation that for surface-sensitive reactions that utilize, as is usually the case, large surface site ensembles (ca. 12 metal atoms for ethane hydrogenolysis on Ru), the ratio of TOFITK(l/zM) and TOFchem differ in value by ca. 2-3 orders of magnitude. Variations in TOFITK (l/zM)with modifier addition are probably primarily due to changes in the surface concentration of hydrogen at reaction conditions or, in rarer cases, promotion of the reaction sites.

8: Turnover Frequencies in Metal Catalysis

Table 10

Ethane Hydrogenolysis on La3+-PromotedCo/Si02[ 3 6 ]

~

~~

0.00

0.85

0.10 0.30 0.50 0.75

1.74 1.53 1.31

a

343

1.o 1.3 1.4 1.3 1.5

3.4 2.4 1.7 1.4 0.9

0.85

lo00 770 710 770 670

Reaction conditions: T = 26OoC,PT = 2.0 atm, HJC,HdHe = 6/0.15/43.85 cc/min.

-

4 ~ 1 0 ~ ( :Y

3x1C3

-

2x10"

-

1x109

-

(ID

v

ILI

P

1.o n F la

0.8

v

Y

t

0.6

e 1 LL

i

0.4 0.2

800

I

LL

0 600

tut LL

400 200

0

1 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

LdCo Atomic Ratio

Figure 15

Efect of La modijication of Co/Si02on ethane hydrogenolysis

5.4 Ammonia Synthesis. - Ammonia (NH3) synthesis is another well-known structure-sensitivereaction. The two most widely accepted mechanisms involve adsorption of nitrogen as the rate limiting step. A main question has been whether nitrogen adsorption is dissociative [37,38] or associative [38]. For NH3

344

Catalysis

synthesis on Fe, it has been determined that dissociatively adsorbed N2 constitutes the surface reaction intermediates [l]. Although Fe is the best understood NH3 synthesis catalyst, the isotopic transient kinetic work done on a commercial Haldor-Topsre Fe catalyst [39,40] does not have accompanying chemisorption results suitable for comparison of TOFITK and TOFchem. However, Nwalor and Goodwin [41] investigated the effects of K+ promotion on a Ru/Si02 catalyst using isotopic transient kinetic analysis and provided hydrogen chemisorption results as well. Both the Fe and the Ru studies indicated that product readsorption can have a large impact on the TOFITK determined based on TMeasured. Unfortunately, techniques were not extant at the time to account for readsorption. Since the degree of NH3 readsorption is even greater than that of MeOH, the TOF ratios determined for that study are problematic for interpretation. The TOF ratios calculated range from 1.4-9, significantly lower than might be expected for such a structure sensitive reaction. The surface coverage of exposed Ru surface atoms by active surface reaction intermediates and readsorbed product was 34% (assuming 1 active intermediate adsorbed per Ru surface atom) - incredibly high for a structure sensitive reaction. Clearly, once ammonia was synthesized it was able to readsorb extensively on available Ru surface atoms. Thus, the SSITKA data available for ammonia synthesis on Ru is not of sufficient quality to make any reasonable comparisons of TOFchem and TOFITK.

6

Conclusions

Based on methanation and its hypothesized mechanism, it was shown that no TOF provides a site activity independent of all variables except temperature and catalyst. All TOFs, including TOFsite, possess some dependence on reactant concentrations. For the example reaction (methanation) cited, TOFITK possessed the least dependence on CO partial pressure. All of the TOFs contained a dependence on H2 partial pressure. It can be concluded that TOFchem 5 TOFsite 5 TOFITK.Thus, using steady-state isotopic transient kinetic analysis (SSITKA), TOFITK, the reciprocal of the average surface residence time of the intermediates producing product (I+),can be employed as a guide to how much TOFchem may differ from TOFsite. For a structure-sensitive reaction where readsorption of products is not a problem, such as ethane hydrogenolysis, the data indicated that TOFITK (l/zM)exceeds the turnover frequency determined from chemisorption (TOFchem)by approximately 2-3 orders of magnitude. In contrast, for a structure-insensitive reaction where readsorption is not a problem, such as methanation, the data indicated that TOFITK exceeds TOFchem by only ca. one order of magnitude. The results for MeOH synthesis showed a difference in TOFITK ( 1/Zo0MeOH)VS.TOFchem of 1-2 orders of magnitude after removing to a large degree the effect of MeOH readsorption. This result suggests that there may indeed be an element of structure sensitivity for this reaction but not as much as seen for ethane hydrogenolysis. Ammonia synthesis results were presented here as a cautionary note

8: Turnover Frequencies in Metal Catalysis

345

that extraneous effects (such as readsorption of products) must be accounted for before conclusions can be made based on TOFITK. The difference between TOFchem and TOFITK for both structure-insensitive and structure-sensitive reactions, without a doubt, reflects the difference in the concentration of adsorption sites (for hydrogen or CO) and that of reaction sites occupied by the most active intermediates. The larger difference in the TOF’s for structure-sensitive reactions is an outcome of ensemble size and geometric arrangement required for active reaction sites. Since the adaptation of TOF from enzyme catalysis (where the nature of active sites is known and where they can be counted) for metal catalysis, TOFchem has often been reported and compared with the knowledge that chemisorption measurements cannot accurately represent the active sites. The goals of this paper included providing a better understanding of the accuracy and reliability of TOFchem and reporting on the advantages of TOFITK to more accurately represent the true ‘site’ TOF. TOFITK is much more expensive and difficult to determine and will, thus, never replace TOFchem as a general expression of specific rate. Luckily, for both structure-sensitive and structure-insensitive reactions, the TOF ratio (TOFITK/TOFChem)is fairly consistent over a wide variety of reaction conditions and catalyst compositions. Since hydrogen or CO chemisorption measurements can be done easily, values of TOFchem can apparently be used to track TOF, as long as one understands that TOFchem may differ in value from site TOF by up to several orders of magnitude. Use of TOFITK, however, is preferred when interpretation of the rate at the site level is desired. Acknowledgments. - This paper is based upon work supported by the National Science Foundation under Grant No. CTS-0211495.

References 1. 2.

3.

4.

5. 6. 7.

M. Boudart and G. Djega-Mariadassou, Kinetics ofHeterogeneous Catalytic Reactions, Princeton University Press, Princeton, NJ, 1984. Sonia Hammache, James G. Goodwin, Jr., Schohn L. Shannon, and So0 Y. Kim, ‘In Situ Analysis of Reaction at the Site Level in Heterogeneous Catalysis,’ in Encyclopedia of Surface & Colloid Science, Marcel Dekker, Inc., New York, 2002. Pp. 2445-2454. Carroll 0. Bennett, ‘Experiments and Processes in the Transient Regime for Heterogeneous Catalysis,’ Advances in Catalysis, vol. 44, Academic Press, New York, 2000. Pp. 329-416. S.L., Shannon and J.G. Goodwin, Jr., ‘Characterization of Catalytic Surfaces by Isotopic-Transient Kinetics during Steady-State Reaction,’ Chemical Reviews 95, 677 (1995). F.H. Ribeiro, A.E.S. von Wittenau, C.H. Bartholomew, and G.A. Somorjai, ‘Reproducibility of Turnover Rates in Heterogeneous Metal Catalysis: Compilation of Data and Guidelines for Data Analysis,’ Catal. Rev.-Sci. Eng. 39,49 (1997). Gates, B. C. Catalytic Chemistry, John Wiley and Sons, New York, 1992. J. Happel, ‘Transient Tracing,’ Chem. Eng. Sci. 33, 1567 (1978).

346

Catalysis

8.

C.O. Bennett, ‘Understanding Heterogeneous Catalysis Through the Transient Method,’ incatalysis Under Transient Conditions (A.T. Bell and L. Hegedus, Eds.), Vol. 178, American Chemical Society, Washington, DC, 1982. Pp. 1-32. P. Biloen, ‘Transient Kinetic Methods,’ J. Mol. Catalysis 21, 17 (1983). A.M. Efstathiou, T. Chafik, D. Bianchi, and C.O. Bennett, ‘CO Chemisorption and Hydrogenation of Surface Carbon Species Formed after CO/He Reaction on Rh/MgO: A Transient Kinetic Study Using FTIR and Mass Spectroscopy,’ J. of Catalysis 147,24 (1994). M. Otarod, S. Ozawa, F. Yin, M. Chew, H.Y. Cheh, and J. Happel, ‘Multiple Isotope Tracing of Methanation over Nickel Catalyst,’ J. of Catalysis 84, 156 (1983). I.-G. Bajusz and J.G. Goodwin, Jr., ‘Hydrogen and Temperature Effects on the Coverages and Activities of Surface Intermediates during Methanation on Ru/Si02,’ J. of Catalysis 169, 157 (1997). Joongjai Panpranot, James G. Goodwin, Jr., and Abdelhamid Sayari, ‘Effect of Hydrogen Partial Pressure on the Surface Reaction Parameters during CO Hydrogenation on Ru-Promoted Silica-Supported Co Catalysts,’ J. of Catalysis, 2 13, 78-85 (2003). S.H. Ali and J.G. Goodwin, Jr., ‘Impact of Readsorption Effects and Their Removal from Surface Reaction Parameters Obtained by Isotopic Transient Kinetic Analysis: Methanol Synthesis on Pd/Si02,’ J. of Catalysis 171,339-344 (1997). M.A. Vannice, in Catalysis: Science and Technology, vol. 3, ed. by J.R. Anderson and M. Boudart, Springer-Verlag, Berlin, 1982. Pp. 139-198. B. Chen and J.G. Goodwin, Jr. ‘Isotopic Transient Kinetic Analysis of CO Hydrogenation on Cu-Modified Ru/Si02,’J. of Catalysis 158,228 (1998). S. Vada, B. Chen, and J.G. Goodwin, Jr., ‘Isotopic Transient Study of La Promotion of Co/A1203for CO Hydrogenation,’ J. ofcatalysis. 153,224 (1995). G.J. Haddad, B. Chen, and J.G. Goodwin, Jr., ‘Effect of La3+Promotion of Co/SiO, on CO Hydrogenation,’ J. ofCatalysis. 161,274 (1996). A. Kogelbauer, J.G. Goodwin, Jr., and R. Oukaci, ‘Ruthenium Promotion of Co/A1203Fischer-Tropsch Catalysts,’ J. of Catalysis. 160, 125 (1996). Joongjai Panpranot, James G. Goodwin, Jr., and Abdelhamid Sayari, ‘CO Hydrogenation on Ru-Promoted Co/MCM-41 Catalysts,’ J. of Catalysis 211,530 (2002). Kandis Sudsakorn, James G. Goodwin, Jr., and Adeyinka A. Adeyiga, ‘Effect of Activation Method on Fe FTS Catalysts: Investigation at the Site Level using SSITKA,’ J. of Catalysis, 213,204-210 (2003). C.-H. Yang, Y Soong, and P. Biloen, ‘A Comparison of Nickel- and PlatinumCatalyzed Methanation Utilizing Transient-Kinetic Methods,’ J. of Catalysis 94, 306 (1985). I.-G. Bajusz, Derk J. Kwik, and J.G. Goodwin, Jr., ‘Methanation on K+-Modified Pt/Si02:The Impact of Reaction Conditions on the Effective Role of the Promoter,’ Catalysis Letters 48, 15 1 (1 997). Sturla Vada and James G. Goodwin, Jr., ‘Analysis of Li Promotion of Methanol Synthesis on Pd/Si02 using Isotopic Transient Kinetics,’ J. Physical Chemistry 99, 9479 (1 995). S.Y. Kim and J.G. Goodwin, Jr., ‘Effectof Back-Exchange of Sodium Cations on PdY Catalysis,’ unpublished results. Y. Kikuzono, S. Kagami, S. Naito, T. Onish, and K. Tamaru, ‘Selective Hydrogenation of Carbon Monoxide on Palladium Catalysts,’ Faraday Discuss. Chem. Soc. 72, 135 (1982).

9. 10.

11.

12.

13.

14.

15. 16. 17.

18. 19. 20. 21. 22.

23. 24.

25. 26.

8: Turnover Frequencies in Metal Catalysis 27. 28. 29.

30.

31. 32. 33. 34. 35.

36. 37.

38. 39. 40. 41.

347

G. van der Lee and V. Ponec, ‘On Some Problems of Selectivity in Syngas Reactions on the Group VIII Metals,’ Catal. Reu.-Sci. Eng. 29, 183 (1987). J.M. Driessen, E.K. Poels, J.P. Hindermann, and V. Ponec, ‘On the Selectivity of Palladium Catalysts in Synthesis Gas Reactions,’ J. of Catalysis 82,26 (1983). V. Pitchon, H. Praliaud, and G.A. Martin, ‘On the Role of Hydroxyl Groups in the CO Hydrogenation Into Methanol on Alkali-Promoted Pd/Si02 Catalysts,’ in Natural Gas Conversion (A. Holmen et al., Eds.), Elsevier Sci. Publ., Studies Surf. Sci. Catal. 61,265 (1991). D.W. Goodman, ‘Ethane Hydrogenolysis over Single Crystals of Nickel: Direct Detection of Structure Sensitivity,’Surface Science 123, L679 (1982). G.A. Martin, ‘The Kinetics of the Catalytic Hydrogenolysis of Ethane over Ni/Si02,’J. of Catalysis 60,345 (1979). G.A. Martin, ‘A Quantitative Approach to the Ensemble Model of Catalysis by Metals,’ Catal. Reu.- Sci. Eng. 30,519 (1988). B. Chen and J.G. Goodwin, Jr. ‘Isotopic Transient Kinetic-Analysis of Ethane Hydrogenolysis on Ru/Si02,’ J. of Catalysis 154, 1 (1995). B. Chen and J.G. Goodwin, Jr. ‘Isotopic Transient Kinetic Analysis of Ethane Hydrogenolysis on Cu Modified Ru/Si02,’ J. ofcatalysis 158,511 (1996). G.J. Haddad and J.G. Goodwin, Jr., ‘The Impact of Aqueous Impregnation on the Properties of Prereduced vs. Precalcined Co/Si02,’J. of Catalysis 157,25 (1995). G.J. Haddad, B. Chen, and J.G. Goodwin, Jr., ‘Characterization of La3+-Promoted Co/Si02 Catalysts,’ J. of Catalysis 160,43 (1996). G. Ertl, Kinetics of Chemical Processes on Well-dejned Surfaces, Springer, Berlin, 1983. M.I. Temkin and V. Pyzhev, ‘Kinetics of the Synthesis of NH3 on Promoted Fe Catalysts,’ Acta. Phys. USSR. 12,327 (1940). J.U. Nwalor, JG. Goodwin, Jr., and P. Biloen, ‘Steady-State Isotopic Transient Kinetic Analysis of Iron-Catalyzed Ammonia Synthesis,’ J. of Catalysis 117, 121 (1989). J.U. Nwalor, ‘Isotopic Transient Kinetic Analysis of Ammonia Synthesis Catalysis.’ Ph.D. Dissertation, University of Pittsburgh, Pittsburgh, 1988. J.U. Nwalor and J.G. Goodwin, Jr. ‘Isotopic Tracing Study of K Promotion of NH3 Synthesis on Ru,’ Topics in Catalysis 1994,1,285 (1994).

Catalysis: Vol 17 (Specialist Periodical Reports)

Catalysis: Vol. 19 (Specialist Periodical Reports)

Catalysis: v. 18 (Specialist Periodical Reports)

Catalysis, Volume 22 (RSC Specialist Periodical Reports)

Photochemistry (Specialist Periodical Reports) (Vol 13)

Chemical Modelling: Vol. 2: Applications and Theory (Specialist Periodical Reports)

Organosphophorus Chemistry (Specialist Periodical Reports) (v. 4)

Photochemistry, Volume 37 (Specialist Periodical Reports)

Heterocyclic Chemistry: v.3 (Specialist Periodical Reports) (Vol 3)

Heterocyclic Chemistry: v.4 (Specialist Periodical Reports)

Organometallic Chemistry, Volume 36 (Specialist Periodical Reports)

Organometallic Chemistry, Volume 31 (Specialist Periodical Reports)

Theoretical Chemistry,vol 2 (Specialist Periodical Reports) (v. 2)

Catalysis: A Review of Chemical Literature: Vol 15 (Specialist Periodical Reports)

Catalysis: A Review of Chemical Literature: Vol 16 (Specialist Periodical Reports)

Alkaloids Volume 4: Specialist Periodical Rep (Specialist Periodical Reports) (v. 4)

Nuclear Magnetic Resonance: Volume 37 (Specialist Periodical Reports)

Terpenoids and Steroids: Volume 12 (Specialist Periodical Reports)

Terpenoids and Steroids: Volume 3 (Specialist Periodical Reports)

Dielectric and Related Molecular Processes: Volume 1 (Specialist Periodical Reports)

Electrochemistry: v.8: A Review of Chemical Literature (Specialist Periodical Reports) (Vol 8)

Electrochemistry Vol 6: The Periodic Table of the Elements Wall Chart (Specialist Periodical Reports) (v. 6)

General and Synthetic Methods (A Specialist Periodical Report, Vol 11)

Dielectric and Related Molecular Processes, Vol. 1: A Review of Chemical Literature (Specialist Periodical Reports)

Organophosphorus Chemistry (Volume 18) Specialist Periodical Report

Gas Kinetics and Energy Transfer: v.4: A Review of Chemical Literature (Specialist Periodical Reports)

The Alkaloids: A Review of Chemical Literature: v. 8 (Specialist Periodical Reports)

Inorganic Chemistry of the Transition Elements: v. 4: A Review of Chemical Literature (Specialist Periodical Reports)

Electrochemistry: v.10: A Review of Chemical Literature (Specialist Periodical Reports)

Inorganic Chemistry of the Transition Elements, Volume 2 (Specialist Periodical Reports) (v. 2)

The Alkaloids: A Review of Chemical Literature: v. 6 (Specialist Periodical Reports)

Catalysis: Vol 17 (Specialist Periodical Reports)

Catalysis: Vol. 19 (Specialist Periodical Reports)

Catalysis: v. 18 (Specialist Periodical Reports)

Catalysis, Volume 22 (RSC Specialist Periodical Reports)

Photochemistry (Specialist Periodical Reports) (Vol 13)

Chemical Modelling: Vol. 2: Applications and Theory (Specialist Periodical Reports)

Organosphophorus Chemistry (Specialist Periodical Reports) (v. 4)

Photochemistry, Volume 37 (Specialist Periodical Reports)

Heterocyclic Chemistry: v.3 (Specialist Periodical Reports) (Vol 3)

Heterocyclic Chemistry: v.4 (Specialist Periodical Reports)

Organometallic Chemistry, Volume 36 (Specialist Periodical Reports)

Organometallic Chemistry, Volume 31 (Specialist Periodical Reports)

Theoretical Chemistry,vol 2 (Specialist Periodical Reports) (v. 2)

Catalysis: A Review of Chemical Literature: Vol 15 (Specialist Periodical Reports)

Catalysis: A Review of Chemical Literature: Vol 16 (Specialist Periodical Reports)

Alkaloids Volume 4: Specialist Periodical Rep (Specialist Periodical Reports) (v. 4)

Nuclear Magnetic Resonance: Volume 37 (Specialist Periodical Reports)

Terpenoids and Steroids: Volume 12 (Specialist Periodical Reports)

Terpenoids and Steroids: Volume 3 (Specialist Periodical Reports)

Dielectric and Related Molecular Processes: Volume 1 (Specialist Periodical Reports)

Electrochemistry: v.8: A Review of Chemical Literature (Specialist Periodical Reports) (Vol 8)

Electrochemistry Vol 6: The Periodic Table of the Elements Wall Chart (Specialist Periodical Reports) (v. 6)

General and Synthetic Methods (A Specialist Periodical Report, Vol 11)

Dielectric and Related Molecular Processes, Vol. 1: A Review of Chemical Literature (Specialist Periodical Reports)

Organophosphorus Chemistry (Volume 18) Specialist Periodical Report

Gas Kinetics and Energy Transfer: v.4: A Review of Chemical Literature (Specialist Periodical Reports)

The Alkaloids: A Review of Chemical Literature: v. 8 (Specialist Periodical Reports)

Inorganic Chemistry of the Transition Elements: v. 4: A Review of Chemical Literature (Specialist Periodical Reports)

Electrochemistry: v.10: A Review of Chemical Literature (Specialist Periodical Reports)

Inorganic Chemistry of the Transition Elements, Volume 2 (Specialist Periodical Reports) (v. 2)

The Alkaloids: A Review of Chemical Literature: v. 6 (Specialist Periodical Reports)

Recommend Documents