Catalysis: Vol. 19 (Specialist Periodical Reports)

Catalysis Volume 19 A Specialist Periodical Report Catalysis Volume 19 A Review of Recent Literature Editors J.J. Spi...

Author: J.J. Spivey

100 downloads 3422 Views 15MB 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

Catalysis Volume 19

A Specialist Periodical Report

Catalysis Volume 19 A Review of Recent Literature Editors J.J. Spivey, Louisiana State University, Baton Rouge, USA K.M. Dooley, Louisiana State University, Baton Rouge, USA

Authors David A Berry, US Department of Energy, West Virginia, USA David A Bruce, Clemson University, South Carolina, USA TV Choudhary, ConocoPhillips Company, Bartlesville, USA Todd H Gardner, US Department of Energy, West Virginia, USA DW Goodman, Texas A&M University, Texas, USA James G Goodwin Jr, Clemson University, South Carolina, USA JSJ Hargreaves, University of Glasgow, UK CGM Hermse, Eindhoven University of Technology, The Netherlands APJ Jansen, Eindhoven University of Technology, The Netherlands Yijun Liu, Clemson University, South Carolina, USA Dora E Lopez, Clemson University, South Carolina, USA Edgar Lotero, Clemson University, South Carolina, USA D McKay, University of Glasgow, UK Fernando Morales, Utrecht University, The Netherlands Dushyant Shekhawat, US Department of Energy, West Virginia, USA Kaewta Suwannakarn, Clemson University, South Carolina, USA Bert M Weckhuysen, Utrecht University, The Netherlands

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

ISBN-10: 0-85404-239-3 ISBN-13: 978-0-85404-239-5 ISSN 0140-0568 A catalogue record for this book is available from the British Library r The Royal Society of Chemistry 2006 All rights reserved Apart from any fair dealing for the purpose of research or private study for non-commercial purposes, or criticism or review as permitted under the terms of the UK Copyright, Designs and Patents Act, 1988 and the Copyright and Related Rights Regulations 2003, 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 UK, or in accordance with the terms of the licences issued by the appropriate Reproduction Rights Organization outside the UK. 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 0WF, UK Registered Charity Number 207890 For further information see our web site at www.rsc.org Typeset by Macmillan India Ltd, Bangalore, India Printed and bound by Henry Ling Ltd, Dorchester, Dorset, UK

Preface

Catalysis continues to be a strong and engaging area of research. New tools are being used to explore the complex processes taking place at the catalyst surface. Conversion of both traditional and new fuels to meet the challenge of clean energy is becoming more important. The reviews in this volume address these topics. Jim Goodwin, Jr. and colleagues at Clemson University (Edgar Lotero, David Bruce, and Kaewta Suwannakarn, Yijun Liu, and Dora Lopez) review the application of solid acid catalysts for the synthesis of biodiesel from renewable sources. Biodiesel is produced by the acid-catalyzed esterification of fatty acids derived from renewables such as vegetable oil. Although this esterification can be carried out using homogeneous acid catalysts, there are clear process advantages to using heterogeneous catalysts—provided the necessary activity and selectivity can be achieved. The authors assess both the current processes that are based on homogeneous catalysts, as well as recent studies of heterogeneous catalysts, which have not been extensively reviewed to date. Nitrides and oxynitrides represent a relatively new class of catalytic material. Justin Hargreaves and D. McKay (University of Glasgow, UK) show that these materials have only recently been explored for reactions (e.g., photocatalysis) beyond those that take advantage of their acid-base properties and their ability to mimic Pt-based catalysts. Tuning the acid-base properties of nitrides is possible by incorporating oxygen within their structure. Cobalt-based Fischer-Tropsch catalysts are the subject of continuing interest as large-scale Gas-to-Liquids plants come on line. Fernando Morales and Bert Weckhuysen (Utrecht University, the Netherlands) look specifically at the effects of various promoters for these catalysts, particularly Mn. The effect of these promoters in controlling the activity and selectivity of the overall reaction can be critical in the overall process economics. This chapter also looks at new spectroscopic techniques that have recently been used to study the effects of these promoters. The decomposition of methane is an important process since it can produce two valuable products: hydrogen and carbon filaments. Wayne Goodman (Texas A&M University) and Tushar Choudhary (ConocoPhillips) show that methane decomposition may be a viable alternative to conventional steam reforming as a source of hydrogen, without the formation of COx as a byproduct. The authors examine the effects of catalyst support and promoters, as well as the inevitable regeneration of the catalyst. The formation of carbon fibers, under certain conditions, makes this process an attractive one.

v

vi

Catalysis, 2006, 19, v–vi

Another route to hydrogen for fuel cell energy applications is the catalytic reforming of liquid fuels. In a review by authors from the US Dept. of Energy (Dushyant Shekhawat, Dave Berry, and Todd Gardner) and Louisiana State University (Jerry Spivey), the catalysts used for this reaction are examined. Among the key issues in this process are carbon deposition and sulfur poisoning. These deactivation mechanisms are widely recognized as barriers to the widespread use of catalytic reforming. The kinetics of the complex reforming process, which includes partial oxidation, steam reforming, and shift reactions, are also reviewed. Finally, the application of computational methods to the study of catalysis continues to increase dramatically. C.G.M. Hermse and A.P.J. Jensen (Eindhoven University of Technology, the Netherlands) present a review of the kinetics of surface reactions with lateral interactions. These methods can be used in predicting catalytic reaction mechanisms. In particular, the authors discuss the role of lateral interactions in adsorbed layers at equilibrium and the determination of lateral interactions from experiments—using the simulations to interpret experimental results. This chapter illustrates the increasing use of computational methods to understand and to design catalysts. I welcome to this volume my Co-Editor and colleague at Louisiana State University, Kerry Dooley. He is well-known to many in the catalysis community for his research in acid-base catalysis. Among other responsibilities, he will serve as Meeting Co-Chair for the upcoming North American Catalysis Society meeting, to take place in Houston on June 17–22, 2007. As always, comments are welcome. James J. Spivey Gordon A. and Mary Cain Dept. Chemical Engineering Louisiana State University Baton Rouge, LA 70803 [email protected] Kerry M. Dooley Gordon A. and Mary Cain Dept. Chemical Engineering Louisiana State University Baton Rouge, LA 70803 [email protected]

Contents Cover Image provided courtesy of computational science company Accelrys (www.accelrys.com). An electron density isosurface mapped with the electrostatic potential for an organometallic molecule. This shows the charge distribution across the surface of the molecule with the red area showing the positive charge associated with the central metal atom. Research carried out using Accelrys’ Materials Studios.

Promotion Effects in Co-based Fischer-Tropsch Catalysis Fernando Morales and Bert M. Weckhuysen 1

General Introduction 1.1 Fischer-Tropsch Synthesis 1.2 Scope of the Review Paper 2 Fischer-Tropsch Catalysis 2.1 Gas-to-Liquid Technology, Economic Impact and its Relevance to Society 2.2 Fischer-Tropsch Catalysts 3 Co-based Fischer-Tropsch Catalysts 3.1 Promotion Effects 3.2 Overview of the Promoter Elements Used in Co-based F-T Catalysts 4 Mn-promoted Fischer-Tropsch Catalysts 4.1 Mn-promoted Fe-based Fischer-Tropsch Catalysts 5 Concluding Remarks and Outlook 6 Acknowledgments References

vii

1

1 1 5 6 6 8 9 10 16 21 22 30 32 32

viii

Catalysis, 2006, 19, vii–x

The Catalysis of Biodiesel Synthesis Edgar Lotero, James G. Goodwin, Jr., David A. Bruce, Kaewta Suwannakarn, Yijun Liu and Dora E. Lopez 1 2

Introduction Overview 2.1 Vegetable Oils and Animal Fats 2.2 Reactions 2.3 Physicochemical Properties of Biodiesel 2.4 The Feedstock Issue 2.5 Processing Methodologies 3 Homogeneous Catalysis 3.1 Base-Catalyzed Synthesis 3.2 Acid-Catalyzed Synthesis 3.3 Integrated Acid-Base Biodiesel Synthesis 3.4 Existing Problems with Homogeneous Catalysts 4 Heterogeneous Catalysis in Biodiesel Synthesis 4.1 Catalysis by Metals, Metal Compounds and Supported Metal Complexes 4.2 Catalysis by Solid Bases 4.3 Catalysis by Solid Acids 4.4 Potential Problems with Heterogeneous Catalysts 5 Conclusions and Future Perspectives References

Catalysis with Nitrides and Oxynitrides J.S.J. Hargreaves and D. Mckay 1 2 3

Introduction Preparation of Nitride and Oxynitride Catalysts Catalytic Reactions with Nitrides and Oxynitrides 3.1 Ammonia Synthesis, Ammonia Decomposition and Hydrazine Decomposition 3.2 Amination and Ammoxidation 3.3 NO Removal 3.4 Hydrotreating and Hydrogenation 3.5 Base Catalysis 3.6 Photocatalysis 3.7 Use as Supports

41

41 43 43 44 46 47 48 49 49 55 60 62 63 64 67 72 77 78 80

84

84 85 89 89 92 93 94 96 98 99


3.8 Hydrogen Storage 4 Conclusions and Outlook 5 Acknowledgments References

Kinetics of Surface Reactions with Lateral Interactions: Theory and Simulations C.G.M. Hermse and A.P.J. Jansen 1 2

Introduction Basics of Lateral Interactions 2.1 The Mechanism of Lateral Interactions 2.2 Equilibrium Aspects 2.3 Effect of Lateral Interactions on the Kinetics 3 Theory 3.1 Including Lateral Interactions in the Kinetics 3.2 Analytical Expressions for Lateral Interactions 3.3 Experimental Determination 3.4 Calculating Lateral Interactions 4 Examples 4.1 NO/Rh(111) 4.2 Sulfate on Fcc(111) Surfaces 4.3 CO/Rh(100) 4.4 O/Pt(111) 4.5 Tartaric Acid on Cu(110) 5 Outlook 6. Acknowledgments References

Methane Decomposition: Production of Hydrogen and Carbon Filaments T.V. Choudhary and D.W. Goodman 1 2

Introduction Hydrogen Production 2.1 Catalytic Decomposition of Methane for Hydrogen Production 2.2 Step-wise Methane Reforming: Regeneration Issues

ix

102 102 103 103

109

109 110 110 114 119 120 121 133 135 137 143 143 145 148 151 153 155 157 157

164

164 166 166 172

x


3

Production of Carbon Filaments by Catalytic Methane Decomposition 3.1 Ni-based Catalysts 3.2 Fe and Co-based Catalysts 4 Concluding Remarks References

Catalytic Reforming of Liquid Hydrocarbon Fuels for Fuel Cell Applications Dushyant Shekhawat, David A. Berry, Todd H. Gardner and James J. Spivey 1

Introduction 1.1 Demands for Fuel Reforming Technology 1.2 Applications/Types of Reforming 1.3 Issues 1.4 Scope of this Chapter 2 Deactivation 2.1 Carbon Deposition 2.2 Sulfur Poisoning 3 Steam Reforming 3.1 Thermodynamics 3.2 Catalysts 4 Partial Oxidation 4.1 Thermodynamics 4.2 Catalysts 5 Autothermal Reforming 5.1 Thermodynamics 5.2 Catalysts 5.3 Exhaust Gas Reforming 6 Pyrolysis/Cracking 7 Plasma-Assisted Reforming 7.1 Non-Thermal Plasma 8 Supercritical Reforming 9 Prereforming 10 Kinetics 10.1 Reactivity of Hydrocarbons 11 Concluding Remarks References

176 176 179 180 181

184

184 184 185 188 190 190 190 203 206 207 210 214 215 215 218 218 221 231 232 235 235 235 237 239 242 243 244

Promotion Effects in Co-based Fischer– Tropsch Catalysis BY FERNANDO MORALES AND BERT M. WECKHUYSEN Department of Inorganic Chemistry and Catalysis, Utrecht University, Debye Institute, Sorbonnelaan 16, Utrecht 3584 CA, The Netherlands

1

General Introduction

1.1 Fischer–Tropsch Synthesis. – Franz Fischer, head of the Max-Planck Institut fu¨r Kohlenforschung in Mu¨lheim (Germany) and Hans Tropsch, a co-worker of Fischer and professor of chemistry in Prague (Czech Republic), Mu¨lheim (Germany) and Chicago (Illinois, USA), discovered in 1922 a catalytic reaction between CO and H2, which yields mixtures of higher alkanes and alkenes.1–20 This invention made it possible for Germany to produce fuels from its coal reserves and by 1938 9 Fischer–Tropsch (F–T) plants were in operation making use of, e.g., cobalt-based F–T catalysts. The expansion of these plants stopped around 1940, but existing plants continued to operate during World War II. It is worthwhile to notice that in 1944, Japan was operating 3 F–T plants based on coal reserves. Whilst being a major scientific as well as a technical success, the F–T process could not compete economically with the refining process of crude oil, becoming important starting from the 1950s. All this coincided with major discoveries of oil fields in the Middle East and consequently the price of crude oil dropped. Although a new F–T plant was built in Brownsville (Texas, USA) in 1950, the sharp increase in the price of methane caused the plant to shut down. Thus, due to bad economics F–T technology became of little importance for the industrial world after World War II and no new F–T plants were constructed. An exception was SouthAfrica, which started making fuels and chemicals from gasified coal based on the F–T process a half century ago due to embargoes initiated by the country’s apartheid policies. Till today, South Africa’s Sasol (South African Coal, Oil and Gas Corporation, Ltd.), building its first commercial F–T plant in 1955, is known as a major player in this field.21 It is remarkable to notice that there is today a renewed interest in F–T technology mainly due to: (i) The rising costs of crude oil. For some time now, the oil prices are well above $50 per barrel.

Catalysis, Volume 19 r The Royal Society of Chemistry, 2006

1

2

Catalysis, 2006, 19, 1–40

(ii) The drive to supply environmentally friendly automotive fuels, more in particularly, the production of synthetic sulphur-free diesel, especially interesting for the European car fleet. (iii) The commercialisation of otherwise unmarketable natural gas at remote locations. CO2 emission regulations will certainly lead in the future to a ban on natural gas flaring near crude oil production wells. This all has led to the recent decisions on major investments by big petrochemical companies, such as Shell22 and ExxonMobil,23 to built large scale F–T plants in Qatar. This will result in an important shift from crude oil to natural gas as feedstock for the production of fuels and chemicals in the decades to come.24–27 Industry projections estimate that by 2020 5% of the production of chemicals could be based on F–T technology with methane instead of crude oil refining operations. All this is especially promising in view of the long-term reserves of coal, which are estimated to be more than 20 times that of crude oil and coal is still used as the carbon source at the largest and economically successful F–T complex, namely the plants Sasol One to Three near Sasolburg in South Africa.21 A picture of a Sasol Fischer–Tropsch plant is shown in Figure 1.28 The stoichiometry of the F–T process can be derived from the following two reactions, the polymerization reaction to produce hydrocarbon chains (1), and the water-gas shift reaction (2): CO þ 2 H2 - –(–CH2–)– þ H2O

(1)

CO þ H2O 2 H2 þ CO2

(2)

Figure 1 Picture of a Sasol Fischer–Tropsch plant in South-Africa

Catalysis, 2006, 19, 1–40

3

The overall stoichiometry in case reaction (2) is completely driven to the right is: (3) 2CO þ H2 - –(–CH2–)– þ CO2 With DH227 ¼ 204.8 kJ, while the maximum attainable yield is 208.5 g of alkenes CnH2n per Nm3 of a mixture of 2 CO and H2 for complete conversion.29,30 The CO/H2 is usually called synthesis gas, or in short syngas. The production of syngas, either by partial oxidation or steam reforming, can account for over 60% of the total cost of the F–T complex since the gasification process is highly endothermic and therefore a high-energy input is required.29–31 It should also be clear that the carbon source used, being it either coal or natural gas, is available at low cost, while the gasification of methane is much more efficient than that of coal since coal simply has a much lower hydrogen content. The syngas produced is then fed into a F–T reactor, which converts it into a paraffin wax that is subsequently hydrocracked to make a variety of chemicals, at present mostly diesel, but also some naphtha, lubricants and gases. A scheme of the F–T reaction process, including syngas production and hydrocracking of the wax, is given in Figure 2. The F–T reaction involves the following main steps at the catalyst surface: (i) The adsorption and maybe dissociation of CO; (ii) The adsorption and dissociation of H2; (iii) Surface reactions leading to alkyl chains, which may terminate by the addition or elimination of hydrogen, giving rise to either paraffin or olefin formation. (iv) Desorption of the final hydrocarbon products, which can be considered as the primary products of the F–T process. (v) Secondary reactions taking place on the primary hydrocarbon products formed due to, e.g., olefin readsorption followed by hydrogenation or chain growth reinitiation. Various detailed mechanisms have been proposed and this matter still remains a controversial issue in the literature. Some of the scientific questions that arise are: (i) Does the adsorbed CO molecule first dissociate into chemisorbed carbon and oxygen atoms? The chemisorbed carbon formed can then be hydrogenated to surface methyl and methylene groups in subsequent steps. Chain growth occurs by stepwise addition of C1 monomers to a surface alkyl group. (ii) Is the adsorbed CO molecule hydrogenated to a CHO or HCOH species, which inserts in the growing hydrocarbon chain? (iii) Is CO directly inserted in the growing chain and then subsequently hydrogenated? It should be clear that a discussion on the F–T mechanism is beyond the scope of this paper and we refer the interested reader to several review papers on this topic.32–42,6,14 In this respect, it is noteworthy to mention the excellent

4

Catalysis, 2006, 19, 1–40

B

C A

(a)

Fischer-Tropsch reactor

Hydrocarbon product

Slurry Steam generation Naphtha Kerosine Diesel

Oxygen Natural gas

Syngas production unit

Waxy syncrude

Hydrocracking unit

Synthesis gas

(b)

Figure 2 (a) Picture illustrating the different steps in the Fischer–Tropsch process: syngas production (A), hydrocarbon formation (B) and hydrocarbon production and (C) product upgrade and (b) detailed flow chart of the Fischer–Tropsch process

updates by Dry on the challenges and technological implementations of Co F–T synthesis.17–20 The overall selectivity of the F–T process is intimately related to the production of methane, which is not economic, since the back conversion to

Catalysis, 2006, 19, 1–40

5

syngas encounters, severe thermal and yield penalties. Consequently, substantial research effors have been devoted to decrease the methane production by adjusting the catalyst composition. It is generally considered that the choice of the catalyst material is central to the F–T process. The latest generation of F–T catalysts are based on cobalt and the cobalt nanoparticles are usually supported on an oxide support, mostly silica, alumina and titania, while some promoters are added to the catalyst material in order to enhance the Co dispersion, e.g., some noble metals. Other metal oxide promoters are often added to the catalysts to improve the F–T selectivity, e.g., decreasing the methane production. Reducing the amount of promoter, especially in the case of noble metals, as well as the amount of cobalt are ways to reduce the catalyst production costs and it may be of no wonder that large research efforts in both academia and industrial laboratories have focused on finding the best performing, durable, but still cheap F–T catalyst formulation. Almost every industrial player in the F–T field has its own catalyst formulation, and is – as expected – very secretive about their exact composition of matter in the catalyst materials applied in pilot and/or industrial plants. The choice of the catalyst material is also related to the type of reactor used. In this respect, it is relevant to mention that Shell and BP use fixed bed reactors, whereas Sasol/Chevron and Exxon Mobil make use of slurry phase reactors. The latter plants require the continuous addition of catalyst material. 1.2 Scope of the Review Paper. – From the above reasoning it is clear that over the past decades a large number of studies have been reported on supported cobalt F–T catalysts. All these studies indicate that the number of available surface cobalt metal atoms determines the catalyst activity and attempts to enhance the catalytic activity have been focusing on two interconnected issues: (1) to reduce the cobalt-support oxide interaction and (2) to enhance the number of accessible cobalt atoms available for F–T reaction. It has been shown that the number of catalytically active cobalt atoms as well as their selectivity can be largely enhanced by the addition of small amounts of various elements, called promoters, to the catalyst material. The exact role of these promoters – as is the case for many other heterogeneous catalysts as well – remains often, however, unclear. The aim of this review paper is to give an extensive overview of the different promoters used to develop new or improved Co-based F–T catalysts. Special attention is directed towards a more fundamental understanding of the effect of the different promoter elements on the catalytically active Co particles. Due to the extensive open and patent literature, we have mainly included research publications of the last two decades in our review paper.43–177 In addition, we will limit ourselves to catalyst formulations composed of oxide supports, excluding the use of other interesting and promising support materials, such as, e.g., carbon nanofibers studied by the group of de Jong.178,179 The paper starts with an introduction in F–T catalysis, including some recent developments in gas-to-liquid technologies and an overview of the main F–T catalyst compositions. In a second part, we will focus on the effect of promoter

6

Catalysis, 2006, 19, 1–40

elements on Co-based F–T catalysis. A classification for the different modes of promotion effects will be proposed and each promoter element reported in literature will be accordingly evaluated. The obtained insights have led to guidelines to design improved Co-based F–T catalysts. A third part will deal with some highlights on the literature of Mn-promoted F–T catalysts and a comparison between supported and unsupported Mn-promoted Fe-, Ru- and Co-based F–T catalysts will be made. It will be shown that many advanced characterization techniques, including spectroscopy and microscopy, are necessary to reveal physicochemical insights in this complex catalytic system. The paper ends with some concluding remarks and a look into the future. 2

Fischer–Tropsch Catalysis

2.1 Gas-to-Liquid Technology, Economic Impact and its Relevance to Society. – At present, the main commercial interest in F–T is the production of high quality sulfur-free synthetic diesel fuels from natural gas, currently being flared at crude oil production wells.21–27 This renewed interest in F–T synthesis has not just only come about as a result of the abundant supply of natural gas, but also because of the global development of fuel supplies and environmental regulations to improve air quality in cities around the world. While the concept of a hydrogen fuel economy remains an important option for the more distant future, synthetic diesel is being promoted by the fuel industry as the most viable next step towards the creation of a sustainable transport industry. Some advantages of synthetic diesel are:

Low content of suphur and aromatic compounds High cetane number Low particulate formation Low NOx and CO emission

At the same time, increased efficiencies in the F–T process and the ability – based on past experience – to build large-scale plants to capture the economies of scale have made the F–T gas-to-liquid (GTL) technology attractive and competitive with the current crude oil refinery industries. It has been estimated that F–T GTL should be viable at crude oil prices of about $20 per barrel. For some time now the oil price has been well above $50 per barrel (more recently it has even topped above $70 per barrel), making it a very appealing technology for countries, having huge reserves of natural gas, but little local market for it and no major pipeline infrastructure to ship it to larger economies. Alternatively, such countries could crack ethane or propane to make ethylene or propylene and further convert it into polyethylene or polypropylene, which can then be shipped to more heavily populated areas in the world. All this holds for the Middle East countries and, e.g., Saudi Arabia is known to heavily invest in propane dehydrogenation plants and polypropylene production facilities, while Qatar is focusing on F–T GTL activities. These activities are concentrated near Ras Laffan in Qatar’s northern gas field,

7

Catalysis, 2006, 19, 1–40

holding 9% of the world’s proven gas resources.25 Table 1 gives a summary of the currently operated and recently announced F–T plants based on natural gas, together with the expected production levels and the industrial companies and countries involved.21–23,26 Industry projections suggest that by 2020 the total GTL capacity in the world could reach more than 1 106 bpd. Currently, there are two F–T plants operating on offshore methane. The first one is the Shell Bintuli plant in Malaysia, which produces 15000 barrels per day. The second one is the Moss Bay plant (PetroSA) located in South Africa. Recently, Sasol/Chevron, ExxonMobil and Shell announced major investments in F–T GTL plants.21–23 In addition, there are many small (mainly for local markets) and large (mainly for export) project proposals for F–T GTL projects on the table. Most of the large project proposals are in the Middle East Table 1

Country

Currently operating and recently announced F–T plants based on methane, together with the industrial companies and countries involved, the used Co F–T catalyst technology, the (expected) production levels and the (expected) year of start-up (barrels per day, bpd) Company or companies

Technology

Production level (bpd)

Start-up year

SouthAfrica

PetroSA

Sasol’s slurry phase technology

20 000

1992

Malaysia

Shell

Shell middle distillate synthesis (SMDS) fixed-bed technology

15 000

1993

Qatar

Sasol and Qatar Petroleum, in alliance with Chevron


34 000

2005 (2 other F–T plants are scheduled to operate in the coming years with the second F–T plant having a scale of 65 000 bpd)

Nigeria

Chevron Nigeria (Sasol/Chevron alliance) and Nigeria National Petroleum Company


34 000

2007

Qatar

Shell and Qatar Petroleum

Shell middle distillate 140 000 synthesis (SMDS) fixed-bed technology

Qatar

ExxonMobil and Qatar Petroleum

Advanced gas conversion for the 21th centure (AGC21) technology

154 000

2009 (first train of 70 000 bpd)/2010 (second train of 70 000 bpd) 2011

8

Catalysis, 2006, 19, 1–40

(Qatar), while the other envisaged projects are in Russia, Australia, Argentina, Egypt, Iran, Bolivia, Brazil, Indonesia, Malaysia and Trinidad. Especially, Russia is expected to have significant long-term potential for F–T GTL technology taking into account the huge country’s gas reserves. 2.2 Fischer–Tropsch Catalysts. – It is well known that all Group VIII transition metals are active for F–T synthesis. However, the only F–T catalysts, which have sufficient CO hydrogenation activity for commercial application, are composed of Ni, Co, Fe or Ru as the active metal phase. These metals are ordersof-magnitude more active than the other Group VIII metals and some characteristics of Ni–, Fe–, Co– and Ru-based F–T catalysts are summarized in Table 2. The exact choice of the active F–T metal to be used in a particular catalyst formulation depends on a number of parameters, including the source of carbon used for making syngas, the price of the active element and the end products wanted. F–T catalysts for the conversion of syngas made from a carbon-rich source, such as coal, are usually based on Fe. This is due to the high WGS activity of Fe, as given in reaction (2), so that less hydrogen is required and oxygen exits the reactor in the form of carbon dioxide. There are, however, new environmental considerations such as the greenhouse effect, which may preclude the future use of Fe precisely due to its high WGS activity. In the case of syngas production from hydrogen-rich carbon sources, such as natural gas, the preferred catalysts due to their lower WGS activities are based on Co or Ru. Nickel F–T catalysts, due to an easy dissociation of CO, possess too much hydrogenation activity, unfortunately, resulting in high yields of methane. At elevated pressure, Ni tends to form nickel carbonyl compounds (highly toxic), and the active component of the catalyst is lost from the F–T reactor. In addition, with increasing reaction temperature the selectivity changes to mainly methane with Ni. This tendency is also observed with Co– and Ru-based catalysts. Instead, with Fe, the selectivity towards methane remains low even at high reaction temperatures. Ru is the most active F–T element working at the lowest reaction temperature of, e.g., only 1501C, very high molecular weight products have been isolated. However, the very low availability and as a consequence the high cost of Ru makes the use of this element in large-scale industrial F–T applications questionable. This leaves Co and Fe as the most appropriate elements to prepare commercially interesting F–T catalysts and both systems have their own advantages Table 2

Overview of some characteristics of Ni–, Fe–, Co– and Ru-based F–T catalysts

Active metal

Price

F–T activity

WGS activity

Hydrogenation activity

Ni Fe Co Ru

þþþþ þ þþþ þþþþþ

þ þ þþþ þþþþþ

þ/ þþþ þ/ þ/

þþþþþ þ þþþ þþþ

Catalysis, 2006, 19, 1–40

9

and disadvantages. It is important to notice that Co is 3 times more active than Fe in F–T while its price is over 250 times more expensive. Because of the relatively low cost of Fe, fresh catalyst material can be added on-line to fluidizied bed reactors, and this practice results in long runs at high conversion levels. This luxury cannot be afforded for the more expensive Co F–T catalysts and so, it is vital that the minimum amount of Co is employed, while maintaining its high activity and long effective catalyst life. Co-based catalysts are preferred for the production of paraffins, as they give the highest yields for high molecular weight hydrocarbons from a relatively clean feedstock, and produce much less oxygenates than Fe catalysts. This is due to a higher hydrogenation activity of Co compared to Fe. On the other hand, if linear olefins are wanted as the end product, it is better to employ Fe-based F–T catalysts because there is less secondary hydrogenation of the primary formed olefins. However, Fe-based catalysts are known to produce aromatics and other non-paraffins, such as oxygenated compounds, as by-products. Another difference between Co and Fe is their sensitivity towards impurities in the gas feed, such as H2S. In this respect, Fe-based catalysts have been shown to be more sulfur-resistance than their Co-based counterparts. This is also the reason why for Co F–T catalysts it is recommended to use a sulphur-free gas feed. For this purpose, a zinc oxide bed is included prior to the fixed bed reactor in the Shell plant in Malaysia to guarantee effective sulphur removal. Co and Fe F–T catalysts also differ in their stability. For instance, Co-based F–T systems are known to be more resistant towards oxidation and more stable against deactivation by water, an important by-product of the FTS reaction (reaction (1)). Nevertheless, the oxidation of cobalt with the product water has been postulated to be a major cause for deactivation of supported cobalt catalysts. Although, the oxidation of bulk metallic cobalt is (under realistic F–T conditions) not feasible, small cobalt nanoparticles could be prone to such reoxidation processes. 3

Co-based Fischer–Tropsch Catalysts

While there have been much activity in the literature addressing Fe, Ru and Ni F–T catalysts, the largest body of papers and patents in the last three decades have dealt with Co-based F–T catalysts in attempts to make more active catalysts with high wax selectivities. It is, however, remarkable to notice that modern Co F–T catalysts are still very similar to the ones prepared by Fischer and co-workers; i.e., they consist of promoted cobalt particles supported on a metal oxide and most of, if not all, Co-based F–T catalyst compositions contain the following components: (i) Co as the primary F–T metal; (ii) A promoter metal, possessing noble metal behavior, e.g., Ru, Re, Pd, Pt, Rh and Ir. The main function of this promoter element is to facilitate the reduction of the cobalt nanoparticles; and as a consequence to increase the

10

Catalysis, 2006, 19, 1–40

number of active cobalt sites. As will be shown later these promoters have also other beneficial effects on the catalyst performance; (iii) An oxidic promoter elements, such as lanthanide and thorium oxide, ceria, zirconia, titania, vanadia, chromia and manganese oxides. However, their roles are much broader; and (iv) A high surface area oxide support, mostly alumina, titania and silica, although the use of supports such as ceria, zirconia, magnesia, gallia, silicaalumina, zeolites (e.g. zeolite Y, silicalite, ZSM-5 and ETS-10), ordered mesoporous oxides (e.g. MCM- and SBA-type materials having high surface area and a narrow pore-size distribution) and delaminated zeolites (e.g. ITQ-2 and ITQ-6) are also reported in literature.180–190 The role of the support material is rather well established. It provides mechanical strength and thermal stability to the Co nanoparticles, while facilitating a high Co dispersion. The choice of the support oxide largely determines the number of active Co metal sites stabilized after reduction, as well as the percentage of supported cobalt oxides that can be reduced to cobalt metal. This is due to a different Co-support oxide interaction. A strong Cosupport oxide interaction, as it occurs in the case of alumina and titania, favors the dispersion of the supported Co particles, but at the same time decreases their reducibility, leading to catalyst materials with a limited number of accessible surface Co metal sites. On the contrary, a much weaker interaction leading to a higher Co reducibility occurs for Co/SiO2 catalysts. In this case, the cobalt particles tend to agglomerate on the support surface during the thermal activation treatments resulting in a relatively low Co dispersion, and thus a low number of surface Co metal sites. Recent studies with ordered mesoporous oxides have shown that cobalt particles with defined particle sizes by confinement within the mesoporous channels are active for F–T catalysis.180–190 An increase of the average particle size of the supported Co particles was found with increasing pore size of the mesopores silica; these larger particles are more reducible and lead to catalyst materials with higher F–T activity. Similar effects have been observed for Co/SiO2 catalysts made from commercial amorphous silicas with increasing pore diameters.191 On the other hand, the origin of the promoter metal and metal oxide effects is not always clear, despite the many detailed characterization studies. In what follows, we will give first a possible definition of the different promotion phenomena described in literature, as well as their mode of operation. The second part deals with an extensive literature overview of the effect of each promoter element on the F–T activity, selectivity and stability of the active Co phase. The different modes of operation will be evaluated for each element. Special attention will be paid to noble metal and transition metal oxide promotion effects. 3.1 Promotion Effects. – The catalyst surface often contains substances that are added deliberately to modify the turnover rate for a given catalytic reaction.191–194 The simplest case being an additive that increases the rate per

Catalysis, 2006, 19, 1–40

11

site per second. It is, in this respect, useful to recall the concepts of catalyst promotion. Promoters are doping agents added to catalyst materials in small amounts to improve their activity, selectivity and/or stability.30 It is generally accepted that promoter elements may induce these beneficial effects in several manners. All this has led researchers to come up with a classification scheme for promoter effects and in the case of the Co F–T literature the following names (including often different definitions!) have been given to the different types of promotion: structural or structure promoters, electronic promoters, textural promoters, stabilizers and catalyst-poison-resistant promoters. Since many of the above-mentioned effects tend to overlap in practice, it is sometimes difficult to precisely define the observed function of a promoter. In addition, the degree to which additives modify a catalyst’s activity in the positive or negative manner is also dependent on the amount of the additive, the support oxide under consideration and the exact preparation method, causing them to act either as a promoter or a poison. In line with this reasoning, the term modifier should be more appropriate according to Paal and Somorjai.190 Finally, it is important to mention that promoter elements are mostly discovered in a serendipitous manner and this holds most probably also for the field of Co F–T catalysis. Only a few of them are expected to be the result of a priori catalyst design. In this review paper we have chosen to divide the family of promoter elements into two classes according to their intended function. Structural promoters affect the formation and stability of the active phase of a catalyst material, whereas electronic promoters directly affect the elementary steps involved in each turnover on the catalyst. The latter group of promoters affect the local electronic structure of an active metal mostly by adding or withdrawing electron density near the Fermi level in the valence band of the metal. This results in a modification of the chemisorption properties of the active metal. Hence, this affects the surface coverage of reactants and, as a consequence, the catalysis done by the metal. In addition to these two groups of promoters we have included in our classification the group of synergistic promotion effects. Although promoter elements are not considered themselves to be catalytically active, they may play other roles under F–T conditions. This may indirectly affect the behaviour of the catalytic active element, still dominating the overall catalytic performances of the catalyst material. We will now discuss more in detail the different promoter effects encountered in Co F–T catalysis.

3.1.1 Structural Promoters. The main functions of structural promoters are to influence the cobalt dispersion by governing the cobalt-support oxide interaction.30 A high Co dispersion results in a highly active Co metal surface and, therefore, in a high coverage by the reactants, and as a consequence an improved catalyst activity. Structural promotion may lead to an increased catalyst activity and stability, but in principle does not influence the product selectivity since it only increases the number of active sites in a catalyst material. This increase in active sites can be achieved by a stabilization of the

12

Catalysis, 2006, 19, 1–40

Co active phase due to the promoter element, which either avoids the formation of metal-support compounds, or prevents the agglomeration and sintering of the Co particles under F–T operation conditions. 3.1.1.1 Stabilizing the Support Oxide. Promoter elements can be added to the support oxide resulting in a decreased Co compound formation with the support oxide. This is illustrated in Figure 3A. More specifically, strategies should be followed to avoid the formation of either cobalt titanate, cobalt silicate or cobalt aluminate as a result of Co solid-state diffusion under reducing or regeneration conditions in the subsurface of these support oxides. Some transition metals, for example Zr or La, could act in such a way. A related problem is the reduction in support surface area. This is especially a problem in the case of titania, where the anatase polymorph is only stable under oxidative regeneration conditions from about 4001C to 7501C. The addition of Si, Zr and Ta as promoter elements may avoid or diminish surface collapse of the support oxide. 3.1.1.2 Glueing the cobalt particles on the support oxide. Some promoter elements can act as an oxidic interface between the supported Co particle and the support oxide, leading to an increased stability of the cobalt particles against sintering during reduction or oxidative regeneration. A plausible schematic representation of this promotion effect is shown in Figure 3B. 3.1.1.3 Promoters leading to increased cobalt dispersion. The addition of promoter elements may also lead to increased cobalt dispersion after preparation. In the absence of the promoters, relatively large cobalt crystals are formed, whereas, by adding these additives, smaller supported cobalt particles can be made. Such promotion effect is illustrated in Figure 3C. Related to this effect it is important to mention that small metal particles composed of a promoter element can dissociate hydrogen in the neighbourhood of a supported cobalt particle leading to the formation of atomic hydrogen that may spill over by diffusion to cobalt,30 as illustrated in Figure 3D. This can result in an enhanced degree of cobalt reduction and therefore a higher amount of surface cobalt metal atoms. The result of this promotion is an increase in the number of active sites and therefore a higher catalyst activity, leaving the catalyst selectivity unaltered. Noble metals, such as Re, Pt and Ru, are known to act in this manner. 3.1.2 Electronic Promoters. In contrast to structural promoters, electronic effects are much less obvious to be detected in an unambiguous manner. Electronic promotion can be best understood in terms of ligand effects. The surrounding (electronic) environment of an active Co site can be altered by the presence of a promoter element. This leads to an electronic donation or withdrawal leading to an increased intrinsic turnover frequency or change in product selectivity. Ligand effects may also result in a decreased deactivation rate by altering the adsorption/desorption properties of the reagents/reaction products. Electronic promotion can only occur when there is a direct chemical

Co3O4

Co3O4

Co3O4

H2

∆T

H2

∆T

H2

∆T

Support modified with promoter

Modified support

Co3O4

Oxidic support

Co3O4

Oxidic support

Co3O4

H2

∆T

Co0

Co0

Modified support

Co0

Co0

Co0

Oxidic support

Co0

Oxidic support

Co0

Oxidic support

Co0

Co0

Co0

Co0

Co0

(d)

(b)

Co

H Oxidic support

CoO

H H CoOx

Spillover effect by a noble metal promoter

Oxidic support

Co

H

Promoter forming an oxidic interphase

Figure 3 The different modes of action of structural promoters in Co-based Fischer–Tropsch catalysis: (a) structural promoter elements can lead to a decreased Co compound formation with the support oxide; (b) structural promoter elements can act as an oxidic interface between the supported Co particle and the support oxide; (c) structural promoter elements may lead to an increased cobalt dispersion; and (d) H2 spillover effect, leading indirectly to a higher dispersion of the supported Co particles

(c)

(a)

Co3O4

Stabilizer promoter

Oxidic support

Co3O4

CoTiO3

Catalysis, 2006, 19, 1–40 13

14

Catalysis, 2006, 19, 1–40

interaction between the promoter element and the cobalt active surface. It is important to mention that electronic metal-support effects have been found to exist in heterogeneous catalysis, but these effects should only play a minor role in Co F–T catalysis since the active Co particles are relatively large and the contact area between support and cobalt particle is therefore very small. Finally, electronic effects induced by promoter elements may be responsible for an increased resistance of the supported Co nanoparticles to re-oxidation or even their stability against deactivation in general. 3.1.2.1 Promoter metal oxide decoration of the cobalt surface. A first way to induce a ligand effect is to decorate the active cobalt surface with metal oxides. In this way, the catalyst surface properties are altered, resulting in improved selectivities and/or activities. It should be clear that a beneficial catalytic effect can only be obtained if the deposited metal oxides are not blocking (all) the active cobalt sites, which would lead to a decreasing hydrogen or CO chemisorption. The decoration effect of a supported Co particle by transition metal oxides is illustrated in Figure 4A. A similar effect may occur with the support oxide as decorating material. This effect is generally known as the ‘‘strong metal-support interaction’’ or SMSI effect.30,195 The SMSI effect is explained in Figure 4B. When metals supported on, e.g., titania are heated in hydrogen at relatively high temperatures, a dramatic decrease in hydrogen and CO chemisorption occurs. This observation is due to a partial encapsulation of the supported metal particle by the support oxide since reduced TiOx ensembles can migrate over the metal surface, leading to a (partial) decoration of the metal particle. 3.1.2.2 Cobalt-promoter alloy formation. Metal alloying or bimetallic alloy formation may also influence the activity and selectivity of Co F–T catalysts. Metal oxide promoter

Co Co

Co-promoter alloy

Co

Oxidic support

Oxidic support

A

C

TiOx Co3O4

TiO2

Co3O4 Co3O4

∆T H2

Co0

Co0 Co0

TiO2

B

Figure 4 The different modes of action of electronic promotors in Co-based Fischer– Tropsch catalysis: (A) promoter metal oxide decoration of the cobalt surface; (B) the SMSI effect; and (C) cobalt-promoter alloy formation

15

Catalysis, 2006, 19, 1–40

Indeed, cobalt and a promoter metal may form an integral metal particle deposited on the support oxide, altering the electronic properties of the surface cobalt metal atoms (Figure 4C). Depending on the promoter element added to the Co cluster, alloying might lead to an increased catalyst activity, selectivity, as well as stability. 3.1.3 Synergistic Promotion Effects. As was already mentioned promoter elements are not considered themselves to be catalytically active, but it is fair to say that this is not always the case. This promoter activity may indirectly affect the behaviour of the catalytic active element since it will alter, e.g., the local feed composition or may, due to its catalytic properties, influence the overall reaction product distribution. The following effects, illustrated in Figure 5, are expected to occur in a promoted Co F–T catalyst. 3.1.3.1 Water-gas shift reaction. The water-gas shift (WGS) reaction (reaction (2)) made by particles composed of a promoter element close to a supported cobalt particle leads to a change in the local CO/H2 ratio, which may affect the surface coverage of cobalt. As a result, both the activity and the selectivity of the catalyst can be altered. Some transition metal oxides are known to act as WGS reagents. 3.1.3.2 Hydrogenation/dehydrogenation reactions. The end products of the F–T process are a mixture of higher alkanes and alkenes. The promoter elements could show under F–T conditions some activities for hydrogenation or dehydrogenation reactions leading to a shift in the relative ratio of alkanes to alkenes. CO + H2O

CO2 + H2

CH3

Hydrogenation H

Co0

Co0

Co0

Co0

Oxidic support

CH2

Oxidic support WGS promoter

A

Hydrogenation promoter B

H2S

Co0

Co0 MxSy

Oxidic support C

Figure 5 Survey of possible synergistic promotion effects occurring in Co-based Fischer– Tropsch catalysis: (A) water-gas shift reaction, (B) hydrogenation/dehydrogenation reactions; and (C) H2S adsorption

16

Catalysis, 2006, 19, 1–40

3.1.3.3 Coke burning during regeneration. Co F–T catalysts deactivate due to coke formation blocking the active sites. This coke can be burned off by an oxidative treatment. The addition of promoter elements may decrease the temperature of this oxidative treatment, preventing the possible clustering of supported cobalt particles. 3.1.3.4 H2S adsorption reaction. As was already mentioned, Co F–T catalysts are prone to H2S poisoning and the addition of specific promoter elements may lead to an increased H2S tolerance. Crucial for this are the use of promoter elements, such as B and Zn, which make stable surface compounds with sulfur. 3.2 Overview of the Promoter Elements Used in Co-based F–T Catalysts. – The following chemical elements have been investigated as promoters in Co-based F–T catalysis: B, Mg, K, Ti, V, Cr, Mn, Ni, Cu, Zr, Nb, Mo, Ru, Rh, Pd, La, Re, Ir, Pt, Ce, Gd and Th.43–177,247–261 Taking the above described classification scheme, we have made an attempt to identify for every reported promoter element its beneficial effect for Co F–T catalysis. The result of this effort is summarized in Table 3. It is clear from this table that different promoter elements could have multiple modes of promotion action. Furthermore, in making this table we encountered several research papers giving opposite

Table 3

Overview of the promotion effects of different elements used in literature on the Co F–T catalyst performances Influence on catalyst

Promotion type

Promotion mode

Activity

Structural

Support stabilization

þ

þ

Cobalt glueing Cobalt dispersion increase

þ þ

þ þ

Decorating cobalt surface

þ

þ

þ

Cobalt alloying

þ

þ

þ

Watergas shift Hydrogenation/ dehydrogenation Coke burning H2S adsorption

þ

þ þ

Electronic

Synergistic

Selectivity

Stability

Element reported in literature to play a role in this promotion effect Mg, Si, Zr, Nb, Rh, La, Ta, Re, Pt B, Mg, Zr Ti, Cr, Mn, Zr, Mo, Ru, Rh, Pd, Ce, Re, Ir, Pt, Th B, Mg, K, Ti, V, Cr, Mn, Zr, Mo, La, Ce, Gd, Th Ni, Cu, Ru, Pd, Ir, Pt, Re B, Mn, Cu, Ce nra

þ þ

Ni, Zr, Gd B, Mn, Zn, Zr, Mo

a One may anticipate that hydrogenation and dehydrogenation reactions can be catalyzed by metals and metal oxides known to be active for this reaction. Examples are CrOx and Pt.

Catalysis, 2006, 19, 1–40

17

conclusions on the observed promotion effect. Such differences can only be explained by the different loadings and preparation methods employed in making the promoted Co F–T catalysts, as well as different F–T reaction conditions employed, leading to different conversions and selectivities. In what follows, we discuss some highlights of noble metal and transition metal oxide promotion of Co-based F–T catalysts. 3.2.1 Noble Metal Promotion Effects. In the group of noble metals; Ru, Re and Pt have been extensively studied as promoter elements for Co-based F–T catalysts, whereas other metals, such as Rh, Pd, Os and Ir, are only reported in limited occasions. Hence, we will focus our attention on the promotion effects induced by Ru, Re and Pt. 3.2.1.1 Ruthenium. Ru is the most studied noble metal promoter and it has been frequently showed to play a role both in structural and electronic promotion.43–84,146,149,155,157,160–168,170,174,176,177,254,260 In the early nineties, Turney et al. observed that the addition of Ru to Co/CeO2 catalysts drastically increased the Co F–T activity without modifying the catalyst selectivity.44,45 Results obtained from XPS and TPR indicated that Ru caused a decrease of the reduction temperature of supported Co3O4 nanoparticles. The authors proposed that Ru facilitates the reduction of cobalt via a hydrogen spillover from Ru to Co, thereby leading to an increase of the number of exposed Co0 sites and consequently, to an increase in the CO hydrogenation rate. This structural promotion of Ru has been shown to take place independently of the support material; i.e., the addition of Ru to Co/Al2O3,49 Co/SiO2157 and Co/TiO246 catalysts decreases the temperature at which CoOx is reduced to Co0 during activation, leading to catalysts with improved cobalt dispersions. The role of Ru as an electronic promoter has also been extensively investigated. In this respect, it is worthwhile to point out that Co–Ru catalysts exhibit exceptional high selectivities to C51 products and higher turnover rates compared to unpromoted Co catalysts.146 A remarkable insight into Ru promotion has been gained from the work of Iglesia et al.43,46 This group observed that at reaction conditions that favor the formation of higher hydrocarbons (i.e., high pressures and high conversions), the apparent turnover numbers on cobalt catalysts are independent of the support material, but are markedly increased by the addition of small amounts of Ru. They found large increases in the turnover rates and the C51 selectivity when Ru was added to Co/TiO2 catalysts in a ratio of Ru/Co o 0.008.46 The experimental results indicated that Ru inhibits the deactivation of the catalysts by keeping the Co surface ‘‘clean’’ and hence, preventing a carbon deposition on the Co particles. This promotion appeared to require an intimate contact between Co and Ru atoms, since a bimetallic nature of the active sites was found to exist and this nature was enhanced by oxidation treatments at high temperatures (4573 K). On the other hand, the higher C51 selectivities found for the Ru-promoted catalysts were discussed in terms of an increase in the Co site density. Apparently, a higher Co

18

Catalysis, 2006, 19, 1–40

site density inherent in the Co–Ru catalysts leads to diffusion-enhanced readsorption of a-olefins, which reverses the b-hydrogen abstraction termination step and thus, favors the formation of higher hydrocarbons.53 Many other groups have recently reported on the effects of Ru promoter in the performances of Co F–T catalysts. For example, Sun et al. observed that the addition of small amounts of Ru to Co/SiO2 catalysts increased the TOF, whereas it did not affect the CH4 selectivity.69 Ru appeared to be enriched at the metallic cobalt surface after reduction, modifying the adsorption properties of the Co0 sites. Price et al. investigated Co/TiO2 and Co–Ru/TiO2 catalysts with infrared (IR) spectroscopy making use of CO as a probe molecule.52 They concluded that their Ru–Co and Co catalysts had different surface structures and confirmed the existence of Co–Ru interactions. Finally, Hosseini et al. investigated the effect of Ru loading on a 20 wt% Co/Al2O3 catalyst,81 and observed that Ru promotion was only achieved for Ru loadings of 0.5 and 1.0 wt%. Higher Ru loading of 1.5 and 2.0 wt% led, however, to a decrease in the CO hydrogenation activity. On the other hand, the C51 selectivity remained almost unaffected by the Ru loading. 3.2.1.2 Rhenium. Re promotion has also been widely investigated in Cobased F–T catalysis.85–98,153,168,171,174,177,253 Re is regarded as a structural promoter and it has been frequently reported to increase the Co reducibility via a hydrogen spillover effect, leading to catalysts with enhanced Co dispersions. Re reduces to a metallic state at higher temperatures than Ru and therefore, can facilitate the second cobalt reduction step, from CoO to Co0.96 Hilmen et al. performed TPR experiments on Co/Al2O3 and Co-Re/Al2O3 catalysts, and on physical mixtures of Co/Al2O3 and Re/Al2O3, and they suggested that no direct contact between Re and Co is necessary to achieve the promotion by Re.86 Hence, Re might be in many cases located on the support material rather than decorating the surface of Co. In line with the above reasoning, Rygh et al. showed with DRIFTS studies that the presence of Re on Co/Al2O3 catalysts increased the amount of bridged CO-adsorbed species without any sign of electronic interaction between the two metals.89 Moreover, the oxidation or hydrogenation revealed bands arising from Re carbonyl species, suggesting that the Re was located at the catalyst surface. Nonetheless, other authors have reported the existence of bimetallic interactions between Co and Re. For example, Bazin et al. investigated the effect of Re on the structure of Re–Co/Al2O3 catalysts by EXAFS94 and the analysis of the Co K edge and Re LIII edge provided direct evidence for Re-Co bond formation. Their results suggest that Re prevents the formation of cobalt surface phases, such as cobalt aluminate, resulting in an increase of the catalyst activity. They also considered that Re prevents the agglomeration of small metal particles in oxidizing environments and thus their migration and sintering. In another paper, Ronning et al. showed from the Re LIII edge that small bimetallic particles were formed after reduction of Co–Re/Al2O3 catalysts containing 4.6 wt% Co and 2 wt% Re.91 More recently, Jacobs et al. have reported EXAFS data on Co–Re/Al2O3 systems.96 They proposed that a direct

Catalysis, 2006, 19, 1–40

19

contact of Re with the Co atoms existed, while evidence for Re–Re bonds was not observed. In addition, they observed that the average Co cluster size decreased with increasing Re loading. It is generally believed that the addition of Re to Co-based catalysts leads to an increase in the F–T activity as a result of the increased number of exposed sites, although the intrinsic activity of the Co sites is not expected to change. For instance, Li et al. showed that the addition of 0.34 wt% Re to Co/TiO2 catalysts with 10 wt% Co, resulted in the highest F–T conversion compared to umpromoted and Ru-promoted Co/TiO2 catalysts.168 On the other hand, the selectivity of the Co sites is generally expected to be not directly affected by the presence of Re,153 although some authors have reported improvements in the C51 selectivity when using Re.94,171 A possible explanation may be the influence of Re on the Co particle sizes distribution, which may also have an indirect effect on the selectivity due to the changes in the Co surface density. Some differences found in literature may also arise from comparisons of the catalysts at different conversions, which lead to significant variations in selectivities. 3.2.1.3 Platinum. Pt is another noble metal considered to play a role as structural promoter, and it is frequently reported to enhance the Co dispersion in supported Co F–T catalysts.99–108,150,153,160,162,164,166,167,176,177,258,260,261 In the early nineties, Zsoldos et al. reported XPS results on Pt-promoted Co/ Al2O3 catalysts, in which the Co reducibility was largely improved when combined with small amounts of Pt.99 They observed that the highly dispersed, but difficult to reduce surface cobalt particles were largely reduced in the Pt– Co/Al2O3 catalyst, whereas only the larger Co3O4 particles were reduced in the Co/Al2O3 catalyst. It was suggested that Pt would prevent the formation of Co aluminates during catalyst preparation and that the Pt might partially cover the surface of the Co metal and even form bimetallic Pt–Co particles. These findings were further evidenced in two following papers.100,101 A year later, Zsoldos and co-workers showed using XPS that Pt–Co/Al2O3 catalysts with a high Pt/Co ratio (Co atomic fraction of 0.2–0.5) contained CoPt3 bimetallic particles at the surface of the Co nanoparticles.102 The existence of Co-Pt interactions in these catalytic systems was further confirmed by Tang et al. making use of IR with CO as probe molecule.104 The results suggested that there was a strong interaction between Pt and Co atoms. Very recently, Jacobs et al. have reported on Co–Pt/Al2O3 catalysts investigated by EXAFS at the Pt L-edge, and they concluded that isolated Pt atoms interact with the supported cobalt clusters without forming Pt–Pt bonds.107 The Co K-edge was used to verify that the cobalt cluster size increased slightly for those systems with Pt promotion, in which the cobalt reduction extent increased by a factor of two and the formation of Co aluminates was prevented. Pt promotion has been also investigated with other support materials. For example, Schanke et al. studied the influence of small amounts of Pt (0.4 wt%), on the reducibility of Co/SiO2 and Co/Al2O3 catalysts containing 9 wt% Co, and observed that the presence of Pt decreased in all cases the reduction temperature of Co3O4, although the effect was more pronounced for

20

Catalysis, 2006, 19, 1–40

alumina-supported Co F–T catalyst.103 In this work the CO hydrogenation rates at pressures of 1 bar were found to be 3–5 times higher in the Pt-promoted catalysts than in the unpromoted counterparts. However, steady state and transient kinetic experiments indicated constant turnover frequencies for CO hydrogenation in all catalysts, independent of the support or the presence of Pt. Thus, the higher apparent turnover numbers for the Pt-promoted catalysts were considered to be due to a higher coverage of intermediates and not by an increase in the intrinsic activity. In addition, the F–T selectivity was not influenced by the presence of Pt. Similar findings were reported by Vada et al. They observed that the addition of 1 wt% Pt to a Co/Al2O3 catalyst containing 8.7 wt% Co, significantly increased the CO hydrogenation rate, whereas the selectivity was unaffected.153 Hence, it was argued that Pt increases the cobalt reduction extent and therefore the CO hydrogenation rate, although the average Co site activity is not altered. Finally, Das et al. have reported on Pt-promoted Co/Al2O3 catalysts with 25 wt% Co and have observed that Pt promotion does not alter the Co dispersion but increases the amount of Co reduced.258 Both unpromoted and Pt-promoted catalysts were found to decline in activity at the same rate. 3.2.2 Transition Metal Oxide Promotion Effects. Many transition metal oxides have been investigated as potential promoters for Co-based F–T catalysts.120–139,143–145,147–149,151–152,154,158–161,165,166,168–173,251,252,255–259 Although different promotion functions are proposed in the literature, mostly, transition metal oxides have been regarded as electronic promoters, having a direct influence on the intrinsic activity and/or selectivity of the Co sites. This effect is manifested through direct interaction of the supported cobalt particles with the transition metal oxides. Hence, the transition metal oxide promoters are thought to be spreading over the cobalt surface in submonolayer coverages, modifying the adsorption properties of the Co active sites. In this respect, it is worthwhile to highlight the work of Bartholomew et al.,118 in which the activity of supported Co nanoparticles was compared with different oxide supports. The following decreasing order in activity was found: Co/TiO2 4 Co/SiO2 4 Co/Al2O3 4 Co/MgO. Since the Co dispersion was found to be very similar in the first three catalysts, the enhanced activity in the case of Co/TiO2 appeared to be due to an electronic effect induced by the TiO2. This was attributed to the SMSI taking place with TiO2, as discussed in more detail before. Early in the nineties Ruiz et al. reported enhanced catalyst activities and increased selectivities to alkenes and higher hydrocarbons upon addition of V, Mg, and Ce oxides to Co-based F–T catalysts.151 These variations were attributed to electronic effects induced by the transition metal oxide. Similar results were obtained by Bessel et al. using a Cr promoter in Co/ZSM-5 catalysts.152 This group observed that the addition of Cr improved the catalyst activity, and shifted the selectivity from methane to higher, generally more olefinic, hydrocarbons. Based on H2 and CO chemisorption, as well as TPR and TPD results, they suggested that the promotion was caused by an interaction between the transition metal oxide and the cobalt oxide, which inhibits

Catalysis, 2006, 19, 1–40

21

cobalt reduction and improves Co dispersion. Furthermore, it was considered that Cr increased the Co–CO and decreased the Co–H bond strengths, resulting in a higher relative amount of CO/H absorbed on the Co particles. This effect would certainly lead to a lower hydrogenation rate during F–T synthesis. In other work, Kikuchi et al. reported the effect of Cr, Ti, Mn and Mo on the F–T performances of catalysts loaded with ultra fine Co particles.158 The addition of these promoters effectively enhanced the catalyst activity, lowered the methane selectivity and increased the C51 production. They attributed these effects to a structural promotion, causing a decrease of the Co particle sizes in the catalysts. However, an electronic effect induced by the transition metal oxide promoters was not considered. Another promotion effect induced by Mn and Mo was reported by the group of Belosludov.169 This group investigated the addition of Mn and Mo to Co-based F–T catalysts using computational chemistry. Interestingly, they found that the transition metal oxides under investigation improved the sulfur tolerance of the catalysts. In recent years, some more detailed studies on transition metal oxide promotion have been reported. For example, Mendes et al. have reported on the promotion of Nb on Co/Al2O3 catalysts, which were characterized using the temperature programmed surface reaction (TPSR) and diffuse reflectance spectroscopy (DRS) techniques.255 The results showed a transient behaviour of the catalysts, being very well distinguished the catalyst containing Nb2O5. The authors proposed that an existent Co0–Co21 interface is responsible for the methanation reaction, whereas a Co0–NbOx is responsible for the hydrocarbon chain growth. The relative amount of each species on the surface was found to influence the selectivity of CO hydrogenation. Another recent study by Xiong et al. has reported on the F–T catalytic performances of Zr-modified Co/Al2O3 catalysts.257 The CoAl2O4 spinel phase was detected in the prepared catalysts and its content decreased with increasing Zr loading. Thus, Zr appeared to inhibit the Co-support formation in the catalysts. Moreover, they observed improved activities and C51 selectivities upon the addition of Zr, and were attributed to the increase of the metal Co sites and reducibility. Similar results have been reported by Zhang et al. using a Mg promoter in Co/Al2O3 catalysts.256 They observed that the addition of Mg decreased the formation of a cobalt surface phase, which was detected by XPS. Small amounts of Mg were found to increase the catalyst activity by means of decreasing the formation of Co aluminates. However, large amounts of Mg caused a decrease in the reducibility of the catalysts due to the formation of MgO–CoO solid solutions.

4

Mn-promoted Fischer–Tropsch Catalysts

We now discuss the main literature results on Mn-promotion for unsupported as well as supported Fe-, Ru-, and Co-based Fischer–Tropsch catalysts. Such comparison between Fe-, Ru- and Co-based catalysts has been shown to be very useful because it places the role of Mn as a promoter in F–T catalysis in a broader perspective.

22

Catalysis, 2006, 19, 1–40

4.1 Mn-promoted Fe-based Fischer–Tropsch Catalysts. – 4.1.1 Unsupported Fe–Mn Fischer–Tropsch Catalysts. Iron-based F–T catalysts possess both hydrogenation and WGS activity, imposing a flexible option as a working catalyst for typically coal-derived CO-rich syngas conversion. Iron-based catalysts often contain small amounts of K and some other metals/metal oxides as promoters to improve their activity and selectivity. Mn has been widely used as one of the promoters for unsuppported Fe-based F–T catalysts, particularly in promoting the production of C2–C4 olefins.196–205 Initially Mn was discovered to lower the methane selectivity and increase the olefin selectivity. More specifically, it was reported by Kolbel and Tillmetz196 that ‘‘more than 50% Mn the remainder being Fe’’ or by Bussemeir et al.197,198 ‘‘about equal parts of Fe and Mn’’ led to these beneficial F–T effects. In contrast to these results, van Dijk et al.199 did not find any changes in the product formation when adding manganese oxide to an Fe F–T catalyst, although this work had been performed at atmospheric pressures, whereas the work of Kolbel and Tillmetz and Bussemeier was done at conditions close to industrial ones. The findings of Bussemeier et al. and Kolbel and Tillmetz were, however, later on confirmed by the work of Barrault and co-workers199,200 and a maximum increase in light olefin production was found for bulk chemical composition Fe/Mn ratios of 1. This group also explained the promotion effects in terms of electronic changes in the metal atoms by the surrounding manganese compounds; i.e., an electronic promotion effect. On the other hand, Barrault et al. also concluded that the beneficial effect of the promoter was related to the preparation method used, and more specifically to the formation of specific FexMnyOz precursor compounds. Other preparation routes were found to lead to other precursor species and, as a consequence, the final catalyst materials where also able to show less activity and selectivity. Calcination and reduction temperatures were found to be crucial in this respect. This important set of observations may explain the initial conflicting views on the promotion effect of Mn in Fe-based F–T catalysis. In a series of seminal contributions by the group of Baerns it was reported that both the activity and selectivity of Fe-based F–T catalysts are affected by the addition of manganese oxides.201,202 More specifically, the authors observed that (1) Mn-rich catalysts are rather resistant to deactivation; (2) the olefin to paraffin ratio is affected by the Mn content; the best selectivities towards olefins being obtained when Mn concentrations are about 15–20 wt%; (3) a reduction temperature of 4001C for the oxidic Mn–Fe catalyst precursors results in high initial activity, particularly for Fe-rich catalysts which then, however, deactivate rather severely as compared to when a 3001C reduction temperature is applied. The same authors also studied the changes in the catalyst materials after reduction and under conditions of Fischer–Tropsch synthesis. It was found that during reaction the catalyst material was transformed into Fe2MnO4 and partly into the Fe5C2 phase. Only minor amounts of Fe0 could be detected during F–T synthesis. Years later, Bian et al. showed that the pre-reduction of the catalyst with CO or H2 has a remarkable effect on the

Catalysis, 2006, 19, 1–40

23

catalyst performances.203 This was explained in terms of different types of iron carbide species formed. CO-reduced catalysts exhibit remarkably higher catalytic activity for F–T synthesis than the H2-reduced catalysts. Furthermore, CO reduction was shown to be more effective for the formation of iron carbide (FexC). The promotion effects of Mn on unsupported Fe-based F–T catalysts were also studied by Jensen and Massoth.204 These authors concluded that the incorporation of Mn chemically and electronically promotes the active Fe surface. More particularly, it appears to alter the CO hydrogenation reaction path by suppressing the direct formation of paraffins from the reactive intermediate, leading to the increased production of higher olefins. Finally, Das et al. also observed that the addition of moderated amounts of Mn promoter to unsupported Fe F–T catalysts promotes the catalytic activity as well as the selectivity towards lower alkenes.205 Due to the importance of the FexMnyOz precursor compounds to make true Mn-promoted Fe F–T catalysts, several groups have focused on the study of Fe, Mn-oxide compounds.206–210 For example, Hearne and Pollak studied the coprecipitation of iron and manganese oxides, and a mixture of two compounds was found to exist with the same Fe: Mn ratio.210 The two compounds are cubic spinels with the same crystallographic parameters; however, one compound is a defect spinel while the other is not. A similar work has been done by Delgado and co-workers.208 In another study, Goldwasser et al. synthesized five perovskite oxides containing La, K, Fe and Mn209 and observed that the activity of the solids in F–T synthesis is strongly dependent on the presence of Fe carbides. Mn was found to increase significantly the production of alkenes, and the combined presence of K and Mn resulted in a stable catalyst with a high yield of C2–C4 alkenes. More recently, Koizumi et al. observed that Mn has an additional beneficial effect in unsupported Fe-based F–T catalysts.211 These authors studied the sulfur resistance of Mn–Fe catalysts and they observed superior catalysts stabilities, especially when the catalysts were pre-reduced in CO. This group also used IR spectroscopy in combination with CO as a probe molecule to compare Fe and Mn–Fe catalysts. It was found that the addition of Mn led to the appearance of several well-resolved bands upon CO adsorption. The appearance of the bands arising from bridged-bonded CO on Fe0 indicated that the size of the Fe0 particles were clearly larger than in the case of the unpromoted catalysts. They attributed the decreased reactivity towards H2S to the observed increase in Fe0 particle size. 4.1.2 Supported Fe–Mn Fischer–Tropsch Catalysts. A much more limited number of studies have dealt with supported Mn-promoted Fe F–T catalysts. In this respect, it is worthwhile to mention the work of Xu et al.212 These authors added MnO to a Fe/silicalite catalyst and observed an enhanced selectivity towards light olefins. Meanwhile the yields for methane as well as for CO2 formation were almost unaffected by MnO addition. Moreover, the conversion of CO was also insensitive to the addition of the MnO promoter.

24

Catalysis, 2006, 19, 1–40

The results also indicated that the addition of MnO leads to an increased CO adsorption capacity. Das and co-workers also reported on the addition of Mn to silicalite supported Fe catalysts.205 They observed that the addition of Mn reduces the particle size of the iron oxide precursor, leading to an increase in alkene selectivity. Another important study is by Abbot et al.213 This group studied the addition of Mn to alumina-supported Fe catalysts and found that promotion led to an increase in the selectivity towards light olefins and a suppression of the selectivity towards methane. These effects were attributed to a change in the Fe dispersion (structural promotion) as well as to an electronic promotion. 4.1.3 Mn-promoted Ru-based Fischer–Tropsch Catalysts. To the best of our knowledge, the first claim on Mn-promoted Ru F–T catalysts was made in the patent literature by Kugler and co-workers.214 Furthermore, it seems that only the group of Hussain have investigated these catalyst systems in great detail.215–227 Hussain et al. observed that the addition of Mn to Ru/Al2O3, Ru/SiO2 and Ru/TiO2 catalysts produced new or improved active CO hydrogenation sites, which are responsible for the enhancement in the production of high molecular weight as well as unsaturated hydrocarbons. The authors discussed the addition of Mn in terms of an electronical and geometrical modification of the catalytically active Ru surface. In a continuation of their work, the same group used IR and CO as a probe molecule to characterize the metal surface of the supported Mn–Ru catalysts. The results indicated that Mn was present as a covering layer on the surface of the supported Ru nanoparticles. An increasing Mn loading led to a decreasing CO adsorption capacity, indicating that the excess of Mn masked the active Ru sites responsible for CO adsorption. No CO adsorption was found to occur on isolated Mn sites. In addition, it was found that the addition of Mn resulted in (1) the formation of a low frequency band in the IR spectrum and (2) a shift of the IR bands of Ru-adsorbed CO. Both observations were explained in terms of an electronic promotion effect, giving rise to different electronic properties of the surface Ru sites as well as changes in their local geometry. A necessary condition for promotion is that small Mn oxides decorate the surface of the supported Ru nanoparticles. Related to this decoration of the Ru surface, it is anticipated that the sites responsible for the production of CH4 are blocked by the addition of Mn. In order to gain additional information on the promotion effect of Mn, Hussain performed XPS, and static secondary ion mass spectrometry (SSIMS), to investigate the adsorption of adsorbed CO on the supported Mn–Ru F–T catalysts.223 XPS revealed that the addition of CO causes a negative binding energy shift of the Ru peak on silica- and alumina-supported catalyst systems. This shift was higher on the corresponding Mn-promoted catalysts. This was attributed to the combined effect of Ru and Mn on the CO adsorption geometry. On titania-supported systems the unusual shift in Ru binding energy is the result of the formation of TiOx covering the surface due to the SMSI effect. The SSIMS data further supported the findings of the XPS study and RuCO1, RuO1, MnO1 peaks were obtained. No evidence for CO adsorption

Catalysis, 2006, 19, 1–40

25

on Mn or MnO could be detected. All the experimental findings were consistent with the presence of linearly bonded CO on top of Ru nanoparticles supported on Al2O3 and SiO2. The SMSI effect in Mn-promoted Ru/TiO2 catalysts was studied in more detail making use of the SSIMS technique, as well as with TEM, and selective chemisorption experiments.224 The SSIMS technique revealed the presence of TiOx forming two new surface sites, TiOx–Ru and TiO–Mn. These species were found to be located at the immediate vicinity of the Ru nanoparticles. These new surface sites were considered to alter the electronic properties of the Ru metal surface and, as a consequence, the product selectivity. A similar study was performed by the same group on a model catalyst system; i.e., a Mn/Ru (0001) surface, prepared by sputtering Mn on a Ru (0001) surface at room temperature using a sputtered ion gun.222 The techniques of choice were SSIMS, EELS and TPD. It was concluded that Mn reduces the coverage of CO adsorbed on the surface by physically blocking the adsorption sites. The adsorbed CO molecule is predominantly linearly bonded, whereas EELS indicated a possible existence of an electronic interaction between the deposited Mn and Ru. This was reflected by the change in the CO stretching frequency to lower wavenumbers once Mn was deposited on the Ru surface. Tercioglu and Akyurtlu also studied the Mn promotion effect on Ru/Al2O3 catalysts.228 They observed that the three catalysts under study, 2.5 wt% Ru/ Al2O3, 6.3 wt% Mn-2.5 wt% Ru/Al2O3 and 6.3 wt% Mn-5 wt% Ru/Al2O3, showed different olefin selectivities. It was found that (1) the manganesecontaining catalysts had higher selectivity for ethylene, which was three times higher than the corresponding umpromoted catalyst, and (2) the olefin-toparaffin ratio was also affected by the presence of Mn and increased with increasing Mn content. The authors also noticed a similar influence of the Ru : Mn ratio on the formation of longer chain hydrocarbons. Finally, Shapovalova and Zakumbaeva studied Mn-promoted Ru/Al2O3 catalysts with adsorption microcalorimetry, XPS and IR.229 They concluded that Mn also acts as a structural promoter since it increases the dispersion of the Ru nanoparticles. 4.1.4 Mn-promoted Co-based Fischer–Tropsch Catalysts. – 4.1.4.1 Unsupported Co–Mn Fischer–Tropsch Catalysts. Van der Riet et al. were the first to report on the use of Mn as promoter in unsupported Co-based F–T catalysts.230 They reported on a stable Co-containing CO hydrogenation catalyst with a high selectivity for C3 hydrocarbons and suppressed CH4 selectivity. This finding was rationalized in terms of competing hydrogenation and oligomerization reactions of the primary hydrocarbon products. In continuation of this work, Hutchings et al. reported in a detailed investigation on the mechanism of CO hydrogenation catalyzed by a Mn–Co catalyst.231 The results of this study indicate that hydroformylation of a C2 surface intermediate cannot account for the high yields of propene and the low yields of methane observed. Based on their findings, a reaction mechanism for carbon–carbon bond formation was proposed involving a-hydroxylated metal-alkyl as an

26

Catalysis, 2006, 19, 1–40

important intermediate, the formation of which involves the coupling of a number of electrophilic and nucleophilic C1 surface intermediates. In-situ XRD studies of the same research group showed that the active catalysts contained bcc metallic Co supported on MnO.232 This uncommon phase is metastable and readily transforms into the stable fcc structure after exposure to air and slight pressure at room temperature. Based on this work and that of many other research groups233–239 it can be stated that oxidized Mn–Co unsupported catalysts are composed of mixed cobalt manganese spinels of Co2MnO4 and CoMn2O4, the ratio of which depends upon composition and pretreatment of the catalyst materials. Reduction of the catalysts in H2 results in a material that contains metallic cobalt, MnO and a certain amount of mixed spinels. An interesting study in line with the above description was reported by Liang et al.233 These authors investigated in detail the influence of the preparation method of Co/Mn oxides and their corresponding Co/Mn ratios on the hydrogenation of CO. Nanometer spineltype Co/Mn oxides with different Co/Mn ratios (Co3xMnxO4, 0 o x o 1.4), single phase composition and large specific surface area (470 m2/g) were prepared by the sol-gel method. These materials were compared with those prepared by nitrate decomposition and solid-state reaction methods. CO hydrogenation tests indicated that the nanometer Co/Mn oxide catalysts had much higher selectivities for light olefins and possessed a lower catalytic activity and methane production capability than the corresponding coprecipitated catalysts with the same composition. In-situ IR experiments on Co and Mn–Co F–T catalysts were performed by Jiang et al. to study in more detail the effect of the addition of Mn on the surface properties of the catalyst materials.237 The authors used CO and CO þ H2 as probe molecules under flowing conditions. It was found that metallic cobalt particles are formed on both Co and Mn–Co catalysts after reduction. This was manifested by the occurrence of the bands corresponding to CO adsorbed on Co0 sites. On the reduced Co sample, however, the molecularly adsorbed CO species were rather difficult to detect because of the rapid dissociation of CO on the fine active metallic cobalt particles formed. For the Mn–Co catalyst, however, distinct IR bands appeared and their intensities increased with increasing Mn loading. These bands were attributed to linearly, bridged, and multiply bridged-bonded CO. The authors considered that the sizes of the Co particles in the reduced Mn–Co catalyst were larger than those in the reduced Co catalyst, and that the stability of the Co particles could be remarkably enhanced with the incorporation of Mn promoter. This enhanced catalyst stability might be manifested by a decreased deactivation and an increased H2S tolerance. Keyser et al. studied Mn–Co F–T catalysts and found that, under industrial relevant conditions, the WGS activity of the catalysts increases with increasing Mn content, but decreases with increasing pressure.234,235 A lower olefin yield was also observed at high pressures. It was stated that structural changes in the cobalt spinel occur over a long period of time and are responsible for the increased hydrogenation activity and increased WGS activity. Mn seems in this

Catalysis, 2006, 19, 1–40

27

respect not to be able to stabilize the cobalt spinel structure in order to retain the olefin activity. Finally, Riedel et al. also noticed the WGS activity of MnO and the addition of this compound to Co F–T catalysts led to the production of CO2.238 In other words, Mn can be regarded as a water-gas shift promoter under F–T conditions. 4.1.4.2 Supported Co–Mn Fischer–Tropsch Catalysts. F–T synthesis of lower hydrocarbons on silicalite-1 supported Co and Co–Mn catalysts was reported by Das et al.110,119 Co3O4 was found to be the only phase present in Mn-free catalysts after calcination, while the addition of Mn favored the formation of a mixed spinel structure of the type (Co1xMnx)3O4. They considered that the addition of Mn decreased the reduction temperature of Co3O4 and presumably its particle size. The catalysts showed good activity and selectivity for light hydrocarbons in the C2–C4 range, particularly propene, and also showed very low WGS activity. Furthermore, the addition of Mn increased the CO conversion, while the selectivity towards alkenes slightly decreased. Another interesting use of zeolites in F–T catalysis is the addition of pentasil zeolites, e.g. ZSM-5, to the unsupported Mn–Co F–T catalyst in order to shift the product distribution.109 Two modes of operation were tested. The first was a single bed reactor with a mechanical mixture of the two components. The second mode of operation consisted of a dual bed approach with the Mn–Co and zeolite in separate reactors. This method of operation led to the formation of aromatic compounds, due to the transformation of olefinic and oxygen containing hydrocarbons. Additionally, high molecular weight hydrocarbons were cracked into lower alkanes. It was concluded that the dual bed arrangement with separate reactors was preferred since it allowed to adjust the optimum temperature for the F–T as well as the zeolite system and to regenerate the zeolite component independently. Zhang et al. studied the Mn promotion in Co/Al2O3 catalysts.116 It was found that the addition of Mn improves the catalytic activity, as well as the C51 selectivity, while the formation of methane and C24 hydrocarbons is significantly suppressed. They observed that Mn improved the dispersion of the active Co phase and also favored the formation of bridged-bonded CO as probed with IR. A small amount of Mn was also able to increase the H2 uptake, although it was again decreased with an excess of Mn. A very detailed characterization study on Mn-promoted Co/SiO2 catalysts was carried out by the group of Klabunde.112,113 They prepared their catalyst materials making use of the solvated metal atom dispersion (SMAD) technique. In this process metal atoms such as Co and Mn are solvated at low temperatures in toluene or other appropriate solvents. Upon warming the nucleation begins. The catalysts prepared were investigated with EXAFS and tested in the hydrogenation reaction of olefins. It was found that the reduced catalysts contained Mn in the oxidized state and Co in both the metallic and oxidized state. The presence of Mn might permit an appreciable increase in metallic Co. The EXAFS results also indicated that the most active Mn–Co/SiO2 catalyst had the largest amount of metallic cobalt and it was argued that the more

28

Catalysis, 2006, 19, 1–40

oxophilic metal, Mn, would scavenge oxygen, allowing Co to remain in the metallic state. Chemisorption studies revealed that the addition of Mn increased the Co dispersion and it was proposed that Mn, in addition to aiding in the dispersion, also influenced the catalytic performances by an electronic effect. The authors also noticed that Mn was most probably bound to the support as a highly dispersed MnO phase, while Co was stabilized in a metallic form on the MnO. The effects of Mn promoter on Co/TiO2 catalysts were investigated by Vob et al.117 They observed that the formation of CoTiO3 was more strongly evident in the presence of Mn. XPS revealed that Mn was much more dispersed than Co in the catalyst system and no appreciable shifts in Co2p XPS binding energies were noticed when Mn was present in the catalyst. Mn was in the reduced catalysts present as MnO and it was assumed that MnO had no influence on the catalytic properties and served as an additional support. Recently, Martinez et al. reported the use of mesoporous Co/SBA-15 catalysts promoted with Mn for the F–T synthesis.114 They observed that Mn favored the formation of long-chain n-paraffins (C101), while decreasing the selectivity towards methane. The Mn-promoted catalysts, however, turned out to be less active than the unpromoted ones. In recent years our research group has been studying in more detail the Mn promotion in F–T Co/TiO2 catalysts, using advanced characterization techniques, such as XAS, EXAFS, STEM-EELS and XPS.247–250 We found that the preparation method largely influences the state and location of Mn promoter in Co/Mn/TiO2 oxidized catalysts,249 and it was shown that the incipient wetness impregnation method leads to the formation of small MnO2 particles located on the TiO2 surface, and also a highly dispersed MnO2 phase. In contrast, the use of the homogeneous deposition precipitation method leads to the deposition of MnO2 species preferentially on top of the Co3O4 nanoparticles. This is evidenced by STEM-EELS, which revealed the elemental distribution of the Co/Mn/TiO2 catalysts in chemical maps. In addition, EXAFS results at the Mn K-edge indicated the formation of mixed oxide compounds between Co and Mn; i.e., the formation of Co3xMnxO4 solid solutions occurs to some extent, causing an elongation of the Mn–O interatomic distances. In a related work,248 we reported that Mn decreased the reducibility of the surface of the Co nanoparticles in Co/Mn/TiO2 catalysts, resulting in catalysts with more unreduced cobalt phase. These findings were revealed by XAS results at the Co L2,3 edges. It was reported that reduction of Co/TiO2 and Co/Mn/TiO2 catalysts, as well as bulk Co3O4 material, at conditions of 2 mbar and 4251C, resulted in different amounts of Co0 and CoO. The Co reducibility was strongly influenced by the TiO2 support and the Mn oxides, given that bulk Co3O4 was 100% reduced to Co0, whereas the TiO2-supported catalysts were only reduced to mixtures of CoO (60–75%) and Co0. This result was considered to be due to the SMSI effect occurring in these catalytic systems. The Mn-promoted catalyst was found to contain the least amount of Co0, indicating a decrease of the Co reducibility due to Mn. In other work using XPS,250 we reported similar effects of the Mn promoter on the Co

Catalysis, 2006, 19, 1–40

29

reducibility. Both a Co/TiO2 and Co/Mn/TiO2 catalyst were reduced in H2 at 3501C achieving high reduction degrees, although the Mn decreased the final amount of Co0. In addition, it was shown with quantitative XPS that the Mn oxides migrate upon reduction, from the top of the Co particles toward the TiO2 surface. The Mn oxides were shown to segregate after reduction and to be enriched at the surface of TiO2. Nonetheless, some remaining MnO was also observed at the surface of the reduced Co nanoparticles. Based on EXAFS and STEM-EELS,249 it was concluded that the Mn species might exist either in the form of MnO, which interacts with the Co nanoparticles, or in the form of Ti2MnO4, highly segregated at the outer layer of the TiO2 surface. These titanate compounds are expected not to play a role in the F–T reaction, but to behave as spectator species. On the contrary, we interpreted that the MnO particles observed in close interaction with the Co0 are able to induce an electronic promotion, resulting in the enhancement of the F–T selectivity. In this respect, we have reported in various papers significant changes in the F–T selectivitiy of Co/TiO2 catalysts as a result of the addition of Mn.247–250 The selectivity of the Mn-promoted catalysts has been shifted to higher C51 and lower methane production. In addition, the olefinic product was always increased in the case of the Co/Mn/TiO2 catalysts. All these findings suggest a different probability of chain growth versus chain termination, influenced by the Mn oxides. Furthermore, the Mn was found to lower the hydrogenation rate occurring under F–T conditions causing an increase of the olefin product formed in the F–T synthesis. Finally, it was observed by STEM-EELS that a Co/Mn/TiO2 catalyst prepared by IWI clearly decreased the Co particle size after activation. Co0 particles of 5–10 nm were covered by MnO after reduction, and this was manifested in the F–T reaction as a decrease in activity. Hence, it was concluded that even when Mn is not mixed with the Co3O4 in the calcined catalyst, upon reduction it may undergo changes, which lead to a covering and blocking of the small Co particles by the MnO phase.

4.1.5 General Observations Based on the Literature Survey. From the above discussions it is evident that Mn promotion has been explored in more detail for unsupported than for supported F–T catalysts. Regardless of the type of support and the type of active F–T metal, similar trends are found regarding the promotion effect of Mn. The following conclusions can be drawn from this literature survey: (1) Mn, when decorating as an oxide layer on top of the surface of the active metal, is clearly able to induce an electronic effect. This electronic promotion is evident from the effect on the catalyst selectivity, as well as on the CO IR adsorption properties (frequency shifts in metal-adsorbed CO). Moreover, XPS binding energy shifts have been observed, particularly for Ru-based F–T catalysts. Consequently, Mn–Co F–T catalysts possess decreasing hydrogenation properties compared to Co F–T catalysts, causing a shift in the product distribution towards more olefinic, and increasing the C51 production.

30

Catalysis, 2006, 19, 1–40

(2) Besides the electronic promotion effect, Mn may also induce structural and synergistic promotion effects and it has been occasionally reported to increase the metal dispersion and to improve the resistance of the catalyst material against H2S and re-oxidation during operation. Moreover, Mn possesses WGS behavior and therefore can be regarded as a WGS promoter. One could therefore state that Mn-promotion makes supported Co nanoparticles more ‘‘Fe-like’’. (3) In order to achieve all these promotion effects, the preparation method clearly plays a role and it may be of no wonder that many studies have focused on the formation of a MxMnyOz (with M ¼ active F–T metal; e.g., for Co, MnxCo3xO4) phase. This spinel oxide is broken up during reduction to make MnOx and a metallic surface. Due to the pre-existence of this Mn–M interaction, electronic promotion is much more easily achieved after reduction as well. It is worthwhile to mention that MnxCo3xO4 compounds are well studied in the literature, because they have important electrocatalytical properties. More specifically, spinel-type manganese oxides are widely used as precursors in the preparation of l-MnO2 ([&]A[Mn2]BO4], an oxide of technical interest due to its application as a cathode material for rechargeable cells.240–246

5

Concluding Remarks and Outlook

Fischer–Tropsch synthesis making use of cobalt-based catalysts is a hotly persued scientific topic in the catalysis community since it offers an interesting and economically viable route for the conversion of e.g. natural gas to sulphurfree diesel fuels. As a result, major oil companies have recently announced to implement this technology and major investments are under way to build large Fischer–Tropsch plants based on cobalt-based catalysts in e.g. Qatar. Promoters have shown to be crucial to alter the catalytic properties of these catalyst systems in a positive way. For this reason, almost every chemical element of the periodic table has been evaluated in the open literature for its potential beneficial effects on the activity, selectivity and stability of supported cobalt nanoparticles. The addition of promoter elements to cobalt-based Fischer–Tropsch catalysts can affect (1) directly the formation and stability of the active cobalt phase (structural promotion) by altering the cobalt-support interfacial chemistry, (2) directly affect the elementary steps involved in the turnover of the cobalt active site by altering the electronic properties of the cobalt nanoparticles (electronic promotion) and (3) indirectly the behaviour of the active cobalt phase, by changing the local reaction environment of the active site as a result of chemical reactions performed by the promoter element itself (synergistic promotion). Despite major research efforts, not so much fundamental insights exist in the origin and the exact mode of operation of these promotion effects. Furthermore, the same promoter can exhibit several effects on the catalyst performances. In addition, there are some contradictions in the literature on the proposed effects of specific promoter elements. Several reasons can be put forward to explain these general observations:

Catalysis, 2006, 19, 1–40

31

(1) Promoted cobalt-based F–T catalysts are from material scientist perspective very complex systems. As a consequence, most characterization techniques, including surface as well as bulk spectroscopies, are not suitable for discriminating between the multiple cobalt and promoter species present at the surface of the support oxide. The existence of spectator species, as a result of the formation of solid compounds between the cobalt oxide and support oxide (e.g. CoTiO3) and between the promoter element and the support oxide (e.g. Ti2MnO4), may further complicate this task. As a result, only detailed physicochemical insight can be obtained by making use of a combination of advanced characterization techniques. Since each technique has its own sensitivity towards the different active and spectator species present at the catalyst surface, a more complete picture will only emerge by combining the information gathered by the different spectroscopic techniques employed. (2) Promoter elements only exhibit their beneficial effect in a limited concentration range as the addition of the promoters in a too high amount may lead to a complete decoration of the active cobalt surface, and as a consequence, result in a decrease of the catalyst activity. In addition, the preparation method is crucial in achieving the envisaged promotion effect. Both arguments may be responsible for the conflicting views in the literature on the different promotion effects of a specific element added to the catalyst material. On the other hand, all this indicates that catalyst preparation tools should be improved to add promoter elements in a precise and controlled manner to the catalyst material. It is, however, important to recall that such synthesis tools should be at a later stage still attractive to the large-scale industrial production of cobaltbased Fischer–Tropsch catalysts. (3) It is far from easy to distinguish structural, electronic and synergistic promotion effects. Structural promotion is, in this respect, the most easily to observe. Most synergistic effects are also widely discussed in the literature in enhancing the catalytic performance of supported cobalt nanoparticles. Instead, promotion as a result of electronic effects are much more difficult to detect. The main reason is that one has to discriminate between the number of surface cobalt sites and the intrinsic activity of a surface cobalt site (turnover frequency). This is especially difficult in view of the complexity of the catalyst material. It also requires spectroscopic tools, which are able to detect changes in the electronic structure of the supported cobalt nanoparticles. (4) The characterization tools to investigate cobalt-based Fischer–Tropsch catalysts are mostly used to study the catalyst materials under conditions far from industrially relevant reaction conditions; i.e., in the presence of CO and H2, as well as of the reaction products, including H2O; at reaction temperatures and at high pressures. Since catalytic solids are dynamic materials undergoing major changes under reaction conditions it can be anticipated that the currently obtained information on the active site is at least incomplete. This holds also for the active state and location of the promoter element under reaction conditions. For example, an electronic effect on the cobalt active phase induced by a promoter element can maybe exist only at high pressures and will remain – due to the lack of the appropriate instrumentation – unnoticed to the catalyst

32

Catalysis, 2006, 19, 1–40

scientist. Therefore, major efforts should be directed to the development of advanced in-situ spectroscopy-microscopy techniques to study promoted Cobased Fischer–Tropsch catalysts in action. Summarizing, there are still many scientific challenges and major opportunities for the catalysis community in the field of cobalt-based Fischer–Tropsch synthesis to design improved or totally new catalyst systems. However, such improvements require a profound knowledge of the promoted catalyst material. In this respect, detailed physicochemical insights in the cobalt-support, cobalt-promoter and support-support interfacial chemistry are of paramount importance. Advanced synthesis methods and characterization tools giving structural and electronic information of both the cobalt and the support element under reaction conditions should be developed to achieve this goal.

Acknowledgments The authors gratefully acknowledge financial support from Shell Global Solutions, leading to the results reported in this review paper. B.M.W. also acknowledges financial support from NWO (Van der Leeuw and VICI grants) for his research on in-situ catalyst characterization. The work also benefited from fruitful discussions with H. Oosterbeek, H. Kuipers and C. Mesters of Shell Global Solutions. Finally, the authors thank E. Kuipers (Engelhard), A. Buchanan (Sasol) and A. Rautenbach (Sasol) for kindly providing us with Figures 1 and 2. Both pictures originate from Sasol Limiteds. References 1. A historical account on Fischer–Tropsch synthesis can be found on the web: http:// www.Fischer–Tropsch.org. 2. F. Fischer and H. Tropsch, Brennstoff-Chemie, 1926, 97. 3. E.J. Hoffman, Coal Conversion, The Energon Company, Laramie, Wyoming, 1970, p. 223. 4. J. Patzlaff, Y. Lieu, C. Graffmann and J. Gaube, Appl. Catal. A: General, 1999, 186, 109. 5. J.H. Gregor, Catal. Lett., 1990, 7, 317. 6. H. Schulz, Appl. Catal. A: General, 1999, 186, 3. 7. D.J. Duvenhage and T. Shingles, Catal. Today, 2002, 71, 227. 8. S.T. Rie and R. Kirshna, Appl. Catal. A: General, 1999, 186, 55. 9. B. Jager and R. Espinoza, Catal. Today, 1995, 23, 17. 10. P.J. Van Berge, S. Barradas, J. Van de Loodsdrecht and J.L. Visage, Petrochemie, 2001, 138. 11. T.H. Fleisch, R.A. Sills and M.D. Briscoe, J. Nat. Gas. Chem., 2002, 11, 1. 12. P.J. Van Berge, S. Barradas, J. Van de Loodsdrecht and J.L. Visage, Erdoel, Erdgas, Kohle, 2001, 117, 138. 13. J.G. Goodwin, Proc. of ACS Symp, Methane Upgrading, Atlanta, GA, USA, 1991, p. 156. 14. E. Iglesia, Appl. Catal. A: General, 1997, 161, 59. 15. E. Iglesia, S.C. Reyes, R.J. Madon and S.L. Soled, Adv. Catal., 1993, 39, 221. 16. M.E. Dry, in Catalysis Science and Technology, eds. J.R. Anderson and M. Boudart Vol. 1, Springer, Berlin, 1981, p. 159.

Catalysis, 2006, 19, 1–40 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. 47. 48. 49. 50. 51. 52. 53.

33

M.E. Dry, J. Chem. Technol. Biotechnol., 2001, 77, 43. M.E. Dry, Appl. Catal. A: General, 1996, 138, 319. M.E. Dry, Catal. Today, 2002, 71, 227. M.E. Dry, Appl. Catal. A: General, 2004, 276, 1. For more details we refer to the following website: http://www.sasol.com. Shell press release on March 9 2004 (http://www.shell.com). ExxonMobil press release on July 14 2004 (http://www.exxonmobil.com). J. Jacometti, A. Ekvall, Shell in the Middle East, issue 17, April 2002, 1–3. G. Coupe, Step on the gas, in The Engineer, August 6, 2004. F. Thackeray, Fischer–Tropsch GTL Approaches Treshold, in World Petroleum Congress, 2003, pp. 150–151. A.H. Tullo, Catalyzing GTL, Chem. Eng. News, 2003, 81, 18. The picture shows the Sasol Slurry Phase Distillate (SPDt) process plant, which converts at low temperatures synthesis gas to paraffins and specialty waxes. We express our gratitude to A. Buchanan and A. Rautenbach, both from Sasol, for kindly providing us with these figures. V. Ponec, in Handbook of Heterogeneous Catalysis, eds. G. Ertl, H. Knozinger and J. Weitkamp, Wiley-VCH, Weinheim, 1997, p. 1876. B. Cornils, W.A. Herrmann, R. Schlogl and C.H. Wong, Catalysis from A to Z, A concise encyclopedia, Wiley-VCH, Weinheim, 2000. J. Hagen, Industrial Catalysis, A practical approach, Wiley-VCH, Weinheim, 1999, p. 174. J.C. Geerling, J.H. Wilson and G.J. Kramer, Appl. Catal. A: General, 1999, 186, 27. A.A. Adesina, Appl. Catal. A: General, 1996, 138, 345. M.E. Dry, Appl. Catal. A: General, 1999, 189, 185. B. Jager, Stud. Surf. Sci. Catal., 1997, 107, 207. B. Davis, Fuel Proc. Techn., 2001, 71, 157. A.C. Vosloo, Fuel Proc. Techn., 2001, 71, 149. M.J. Overett, R.O. Hill and J.R. Moss, Coord. Chem. Rev., 2000, 206–207, 581. I. Wender, Fuel Processing Technology, 1996, 48, 228. R.L. Espinoza, A.P. Steynberg, B. Jager and A.C. Vosloo, Appl. Catal. A: General, 1999, 186, 13. R.B. Anderson, The Fischer Tropsch Synthesis, Academic Press, New York, 1984. A.T. Bell, H. Heinemann and W.G. McKee, Appl. Catal., 1982, 2, 219. E. Iglesia, S.L. Soled and R.A. Fiato, J. Catal., 1993, 143, 345. M. Hoang, A.E. Hughes and T.W. Turney, Appl. Surf. Sci., 1993, 72, 55. L.A. Bruce, M. Hoang, A.E. Hughes and T.W. Turney, Appl. Catal. A: General, 1993, 100, 51. E. Iglesia, S.L. Soled, R.A. Fiato and G.H. Via, Stud. Surf. Sci. Catal., 1994, 81, 433. J. Vaarl, J. Lahtinen and P. Hautojarvi, Surf. Sci., 1996, 346, 11. C.L. Bianchi, R. Carli, S. Merlotti and V. Ragaini, Catal. Lett., 1996, 41, 79. A. Kogelbauer, J.G. Goodwin and R. Oukaci, J. Catal., 1996, 160, 125. G.D. Zakumbayeva, L.B. Shapovalova, I.G. Yefremenko and A.V. Gabdrakipov, Petroleum Chem., 1996, 36, 428. A.R. Belambe, R. Oukaci and J.G. Goodwin, J. Catal., 1997, 166, 8. J.G. Price, D. Glasser, D. Hildebrandt and N.J. Coville, Stud. Surf. Sci. Catal., 1997, 107, 243. E. Iglesia, Stud. Surf. Sci. Catal., 1997, 107, 153.

34

Catalysis, 2006, 19, 1–40

54. M. Niemela, M. Reinikainen and J. Kiviaho, Stud. Surf. Sci. Catal., 1998, 118, 229. 55. C.L. Bianchi, S. Vitali and V. Ragaini, Stud. Surf. Sci. Catal., 1998, 119, 167. 56. M. Reinikainen, M.K. Niemela, N. Kakuta and S. Suhonen, Appl. Catal. A: General, 1998, 174, 61. 57. L. Guczi, G. Stefler, Z. Koppany, V. Komppa and M. Reinikainen, Stud. Surf. Sci. Catal., 1999, 121, 209. 58. P.M. Maitlis, R. Quyoum, H.C. Long and M.L. Turner, Appl. Catal. A: General, 1999, 186, 363. 59. Y. Borodko and G.A. Somorjai, Appl. Catal. A: General, 1999, 186, 355. 60. M. Kraum and M. Baerns, Appl. Catal. A: General, 1999, 186, 189. 61. Y.L. Zhang, D.G. Wei, S. Hammache and J.G. Goodwin, J. Catal., 1999, 188, 281. 62. L. Huang and Y. Xu, Catal. Lett., 2000, 69, 145. 63. J.L. Li, L.G. Xu, R. Keogh and B. Davis, Catal. Lett., 2000, 70, 127. 64. L. Huang and Y.D. Xu, Appl. Catal. A: General, 2001, 205, 183. 65. H.A.J. van Dijk, J.H.B. Hoebink and J.C. Schouten, Chem. Eng. Sci., 2001, 56, 1211. 66. G. Jacobs, Y. Zhang, T.K. Das, J. Li, P.M. Patterson and B.H. Davis, Stud. Surf. Sci. Catal., 2001, 139, 415. 67. B. Jongsomjit, J. Panpranot and J.G. Goodwin, J. Catal., 2001, 204, 98. 68. S. Hammache, J.G. Goodwin and R. Oukaci, Catal. Today, 2002, 71, 361. 69. S. Sun, K. Fujimoto, Y. Yoneyama and N. Tsubaki, Fuel, 2002, 81, 1583. 70. J. Panpranot, J.G. Goodwin and A. Sayari, J. Catal., 2002, 211, 530. 71. M.L. Turner, N. Marsih, B.E. Mann, R. Quyoum, H.C. Long and P.M. Maitlis, J. Am. Chem. Soc., 2002, 124, 10456. 72. J. Panpranot, J.G. Goodwin and A. Sayari, Catal. Today, 2002, 77, 269. 73. J. Panpranot, J. G. Goodwin and A. Sayari, J. Catal., 2003, 213, 78. 74. D. Bazin, I. Kovacs, J. Lynch and L. Guczi, Appl. Catal. A: General, 2003, 242, 179. 75. L.B. Shapovalova, G.D. Zakumbaeva and A.V. Gabdrakipov, Petroleum Chem., 2003, 43, 170. 76. M. Shinoda, Y. Zhang, Y. Shiki, Y. Yoneyama, K. Hasegawa and N. Tsubaki, Catal. Comm., 2003, 4, 423. 77. M. Johns, P. Collier, M.S. Spencer, T. Alderson and G.J. Hutchings, Catal. Lett., 2003, 90, 187. 78. H.A.J. van Dijk, J.H.B.J. Hoebink and J.C. Schouten, Top. Catal., 2003, 26, 111. 79. H.A.J. van Dijk, J.H.B.J. Hoebink and J.C. Schouten, Top. Catal., 2003, 26, 163. 80. F.M.T. Mendes, F.B. Noronha, C.D.D. Sluza, M.A.P. da Silva, A.B. Gaspar and M. Schmal, Stud. Surf. Sci. Catal., 2004, 147, 361. 81. S.A. Hosseini, A. Taeb, F. Feyzi and F. Yaripour, Catal. Comm., 2004, 5, 137. 82. L.B. Shapovalova, G.D. Zakumbaeva and A.A. Zhurtbaeva, Stud. Surf. Sci. Catal., 2004, 147, 373. 83. G.D. Zakumbaeva and L.B. Shapovalova, Petr. Chem., 2004, 44, 334. 84. M. Cerro-Alarcon, A. Maroto-Valiente, I. Rodrı´ guez-Ramos, A. Guerrero-Ruiz, D. Schanke, A.M. Hilmen, E. Bergene, K. Kinnari, E. Rytter, E. Adnanes and A. Holmen, Catal. Lett., 1995, 34, 269. 85. L.E.S. Rygh, I. Gausemel, O.H. Ellestad, P. Klaeboe, C.J. Nielsen and E. Rytter, J. Mol. Struc., 1995, 349, 325. 86. A.M. Hilmen, D. Schanke and A. Holmen, Catal. Lett., 1996, 38, 143. 87. A.M. Hilmen, D. Schanke and A. Holmen, Stud. Surf. Sci. Catal., 1997, 107, 237.

Catalysis, 2006, 19, 1–40

35

88. A.M. Hilmen, D. Schanke, K.F. Hanssen and A. Holmen, Appl. Catal. A: General, 1999, 186, 169. 89. L.E.S. Rygh and C.J. Nielsen, J. Catal., 2000, 194, 401. 90. A.M. Hilmen, E. Bergene, O.A. Lindvag, D. Schanke, S. Eri and A. Holmen, Catal. Today, 2001, 69, 227. 91. M. Ronning, D.G. Nicholson and A. Holmen, Catal. Lett., 2001, 72, 141. 92. L. Guczi, G. Stefler, Z. Koppany and L. Borko, React. Kin. Catal. Lett., 2001, 74, 259. 93. L. Guczi, L. Takacs, G. Stefler, Z. Koppany and L. Borko, Catal. Today, 2002, 77, 237. 94. D. Bazin, L. Borko, Z. Koppany, I. Kovacs, G. Stefler, L.I. Sajo, Z. Schay and L. Guczi, Catal. Lett., 2002, 84, 169. 95. L. Guczi, G. Stefler, L. Borko, Z. Koppany, F. Mizukami, M. Toba and S. Niwa, Appl. Catal. A: General, 2003, 246, 79. 96. G. Jacobs, J.A. Chaney, P.M. Patterson, T.K. Das and B.H. Davis, Appl. Catal. A: General, 2004, 264, 203. 97. C.J. Bertole, C.A. Mims and G. Kiss, J. Catal., 2004, 221, 191. 98. C. Aaserud, A.M. Hilmen, E. Bergene, S. Eric, D. Schanke and A. Holmen, Catal. Lett., 2004, 94, 171. 99. Z. Zsoldos, T. Hoffer and L. Guczi, J. Phys. Chem., 1991, 95, 798. 100. L. Guczi, T. Hoffer, Z. Zsoldos, S. Zyade, G. Maire and F. Garin, J. Phys. Chem., 1991, 95, 802. 101. G.M. Lu, T. Hoffer and L. Guczi, Catal. Lett., 1992, 14, 207. 102. Z. Zsoldos and L. Guczi, J. Phys. Chem., 1992, 96, 9393. 103. D. Schanke, S. Vada, E.A. Blekkan, A.M. Hilmen, A. Hoff and A. Holmen, J. Catal., 1995, 156, 85. 104. S. Tang, J. Lin and K.L. Tan, Surf. Int. Anal., 1999, 28, 155. 105. G.W. Huber, C.H. Bartholomew, T.L. Conrad, K.W. Woolley and C.G. Guymon, Am. Chem. Soc. Abstracts, 2000, 219, 261. 106. J. Li, X. Zhang, Y. Zhang, G. Jacobs, T. Das and H. Davis, Appl. Catal. A: General, 2002, 228, 203. 107. G. Jacobs, T.K. Das, P.M. Patterson, J. Li, L. Sanchez and B.H. Davis, Appl. Catal. A: General, 2003, 247, 335. 108. G. Jacobs, J.A. Chaney, P.M. Patterson, T.K. Das, J.C. Maillot and B.H. Davis, J. Synchrot. Rad., 2004, 11, 414. 109. D.J. Koh, J.S. Chung and Y.G. Kim, Ind. Eng. Chem. Res., 1995, 34, 1969. 110. D. Das, G. Ravichandran and D.K. Chakrabarty, Catal. Today, 1997, 36, 285. 111. G. Baurle, K. Guse, M. Lohrengel and H. Papp, Stud. Surf. Sci. Catal., 1993, 75, 2789. 112. B.J. Tan, K.J. Klabunde, T. Tanaka, H. Kanai and S. Yoshida, J. Am. Chem. Soc., 1998, 110, 5951. 113. K.J. Klabunde and Y. Imizu, J. Am. Chem. Soc., 1984, 106, 2721. 114. A. Martinez, C. Lopez, F. Marquez and I. Diaz, J. Catal., 2003, 220, 486. 115. H.Y. Gao, J.G. Chen, H.W. Xiang, J.L. Yang, Y.W. Li and Y.H. Sun, Chin. J. Catal., 2001, 22, 133. 116. J.L. Zhang, J. Ren, J.G. Chen and Y.H. Sun, Acta Physico-Chimica Sin., 2002, 18, 260. 117. M. Vob, D. Borgmann and G. Wedler, J. Catal., 2002, 212, 10. 118. C.H. Bartholomew and R.C. Reuel, Ind. Eng. Chem. Prod. Res. Dev., 1985, 24, 56.

36

Catalysis, 2006, 19, 1–40

119. D. Das, G. Ravichandran and D.K. Chakrabarty, Appl. Catal. A: General, 1995, 131, 335. 120. J.L. Li and N.J. Coville, Appl. Catal. A: General, 1999, 181, 201. 121. J.L. Li and N.J. Coville, Appl. Catal. A: General, 2001, 208, 177. 122. N.J. Coville and J. Li, Catal. Today, 2002, 71, 403. 123. J.L. Li and N.J. Coville, South African J. Chem., 2003, 56, 1. 124. J.L. Li and N.J. Coville, Am. Chem. Soc. Abstracts, 2003, 226. 125. S. All, B. Chen and J.G. Goodwin, J. Catal., 1995, 157, 35. 126. R. Oukaci, J.G. Goodwin, G. Marcelin and A. Singleton, Am. Chem. Soc. Abstracts, 1995, 209. 127. G.R. Moradi, M.M. Basir, A. Taeb and A. Kiennemann, Catal. Comm., 2003, 4, 27. 128. B. Jongsomjit, J. Panpranot and J.G. Goodwin, J. Catal., 2003, 215, 66. 129. W. Zhou, J.G. Chen and Y.H. Sun, Chin. J. Catal., 2004, 25, 467. 130. S. Vada, A.M. Kazi, F.K. Beduaddo, B. Chen and J.G. Goodwin, Stud. Surf. Sci. Catal., 1994, 81, 443. 131. S. Vada, B. Chen and J.G. Goodwin, J. Catal., 1995, 153, 224. 132. G.J. Haddad, B. Chen and J.G. Goodwin, J. Catal., 1996, 161, 274. 133. G.J. Haddad, B. Chen and J.G. Goodwin, J. Catal., 1996, 160, 43. 134. J. Vaari, J. Lahtinen, A. Talo and P. Hautojarvi, Surf. Sci., 1991, 251, 1096. 135. I. Puskas, T.H. Fleisch, J.B. Hall, B.L. Meyers and R.T. Roginski, Stud. Surf. Sci. Catal., 1993, 75, 2813. 136. M.K. Niemela and A.O.I. Krause, Catal. Lett., 1995, 34, 75. 137. A.F.Y. Alshammary, I.T. Caga, A.Y. Tata and J.M. Winterbotton, Eur. J. Sol. State Inorg. Chem., 1991, 28, 453. 138. X.P. Dai, C.C. Yu and S.K. Shen, Chin. J. Catal., 2001, 22, 104. 139. H.B. Shi, Q. Li, X.P. Dai, C.C. Yu and S.K. Shen, Stud. Surf. Sci. Catal., 2004, 147, 313. 140. E.A. Blekkan, A. Holmen and S. Vada, Act. Chem. Scand., 1993, 47, 275. 141. G.P. Huffman, H. Shan, J.M. Zhao, F.E. Huggins, T.E. Hoost, S. Halvorsen and J.G. Goodwin, J. Catal., 1995, 151, 17. 142. S.J. Jenkins and D.A. King, J. Am. Chem. Soc., 2000, 122, 10610. 143. G.W. Huber, S.J.M. Butala, M.L. Lee and C.H. Bartholomev, Catal. Lett., 2001, 74, 45. 144. H.K. Woo, R. Srinivasan, L. Rice, R.J. Deangelis and P.J. Reucroft, J. Mol. Catal., 1990, 59, 83. 145. J.E. Baker, R. Burch, S.J. Hibble and P.K. Loader, Appl. Catal., 1990, 65, 281. 146. A.L. Lapidus, A.Y. Krylova, M.P. Kapur, E.V. Leongardt, A.B. Fasman and S.D. Mikhailenko, Bull. Russ. Acad. Sci.-Div. Chem. Sci., 1992, 41, 45. 147. S.E. Colley, M.J. Betts, R.G. Copperthwaite, G.J. Hutchings and N.J. Coville, J. Catal., 1992, 134, 186. 148. A.F.Y. Alshammary, I.T. Caga, A.Y. Tata, J.M. Winterbottom and I.R. Harris, J. Chem. Tech. Biotech., 1992, 55, 375. 149. I.R. Harris, I.T. Caga, A.Y. Tata and J.M. Winterbottom, Stud. Surf. Sci. Catal., 1993, 75, 2801. 150. M.P. Kapoor, A.L. Lapidus and A.Y. Krylova, Stud. Surf. Sci. Catal., 1993, 75, 2741. 151. A.G. Ruiz, A.S. Escribano and I.R. Ramos, Appl. Catal. A: General, 1994, 120, 71. 152. S. Bessell, Stud. Surf. Sci. Catal., 1994, 81, 479. 153. S. Vada, A. Of, E. Adnanes, D. Schanke and A. Holmen, Top. Catal., 1995, 2, 155.

Catalysis, 2006, 19, 1–40 154. 155. 156. 157. 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.

37

J. Barrault and N. Biwole, Bull. Soc. Chim. Belg., 1995, 104, 149. M. Niemela and M. Reinikainen, Stud. Surf. Sci. Catal., 1998, 119, 161. D.O. Uner, Ind. Eng. Chem. Res., 1998, 37, 2239. K. Okabe, X.H. Li, T. Matsuzaki, M. Toba, H. Arakawa and K. Fujimoto, J. Jap. Petrol. Inst., 1999, 42, 377. E. Kikuchi, R. Sorita, H. Takahashi and T. Matsuda, Appl. Catal. A: General, 1999, 186, 121. A. Frydman, D.G. Castner, C.T. Campbell and M. Schmal, J. Catal., 1999, 188, 1. S. Sun, N. Tsubaki and K. Fujimoto, Chem. Lett., 2000, 2, 176. J.L. Li, L.G. Xu, R.A. Keogh and B.H. Davis, Am. Chem. Soc. Abstracts, 2000, 219. N. Tsubaki, S.L. Sun and K. Fujimoto, J. Catal., 2001, 199, 236. K. Okabe, X.H. Li, T. Matsuzaki, H. Arakawa and K. Fukimoto, J. Sol-Gel Sci. Tech., 2000, 19, 519. C.L. Bianchi, Catal. Lett., 2001, 76, 155. D.G. Wei, J.G. Goodwin, R. Oukaci and A.H. Singleton, Appl. Catal. A: General, 2001, 210, 137. G. Jacobs, T.K. Das, Y.Q. Zhang, J.L. Li, G. Racoillet and B.H. Davis, Appl. Catal. A: General, 2002, 233, 263. G. Jacobs, P.M. Patterson, Y.Q. Zhang, T. Das, J.L. Li and B.H. Davis, Appl. Catal. A: General, 2002, 233, 215. J.L. Li, G. Jacobs, Y.Q. Zhang, T. Das and B.H. Davis, Appl. Catal. A: General, 2002, 223, 195. R.V. Belosludov, S. Skahara, K. Yajima, S. Takami, M. Kubo and A. Miyamoto, Appl. Surf. Sci., 2002, 189, 245. T. Riedel and G. Schaub, Top.Catal., 2003, 26, 145. A. Martinez, C. Lopez, F. Marquez and I. Diaz, J. Catal., 2003, 220, 486. M. Wei, K. Okabe and H. Arakawa, J. Jap. Petrol. Inst., 2003, 46, 339. W.S. Yang, D.H. Yin, J. Chang, H.W. Xiang, Y.Y. Xu and Y.W. Li, Acta Chim. Sinica., 2003, 61, 681. F.M.T. Mendes, F.B. Noronha, C.D.D. Souza, M.A.P. da Silva, A.B. Gaspar and M. Schmal, Stud. Surf. Sci. Catal., 2004, 147, 361. N. Koizumi, K. Murai, T. Ozaki and M. Yamada, Catal. Today, 2004, 89, 465. K. Nagaoka, K. Takanabe and K. Aika, Appl. Catal. A: General, 2004, 268, 151. D. Bazin and L. Guczi, Stud. Surf. Sci. Catal., 2004, 147, 343. G.L. Bezemer, A. van Laak, A.J. van Dillen and K.P. de Jong, Stud. Surf. Sci. Catal., 2004, 147, 259. G.L. Bezemer, U. Falke, A.J. van Dillen and K.P. de Jong, Chem. Commun., 2005, 731. P. Concepcion, C. Lopez, A. Martinez and V.F. Puntes, J. Catal., 2004, 228, 321. A.Y. Khodakov, A. Griboval-Constant, R. Becchara and F. Villain, J. Phys. Chem. B, 2001, 105, 9805. A. Khodakov, A. Griboval-Constant, R. Bechara and V.L. Zholobenko, J. Catal., 2002, 206, 230. D. Yin, W. Li, W. Yang, H. Xiang, Y. Sun, B. Zhong and S. Peng, Microporous Mesoporous Mater., 2001, 47, 15. S. Suvato and T.A. Pakkanen, J. Mol. Catal. A: Chemical, 2000, 164, 273. J. Panpranot, J.G. Goodwing Jr. and A. Sayari, Catal. Today, 2002, 77, 269. A.Y. Khodakov, R. Bechara and A. Griboval-Constant, Appl. Catal. A: General, 2003, 254, 273.

38

Catalysis, 2006, 19, 1–40

187. A. Martinez, C. Lopez, F. Marquesz and I. Diaz, J. Catal., 2003, 220, 486. 188. D.J. Koh, J.S. Chung and Y.G. Kim, Ind. Eng. Chem. Res., 1995, 34, 1969. 189. Y. Ohtsuka, T. Arai, S. Takasaki and N. Tsubouchi, Energy and Fuels, 2003, 17, 804. 190. A.M. Saib, M. Claeys and E. van Steen, Catal. Today, 2002, 71, 395. 191. Z. Paal and G.A. Somorjai, in Handbook of Heterogeneous Catalysis, eds. G.Ertl, H.Knozinger and J. Weitkamp, Wiley-VCH, Weinheim, 1997, p. 1084. 192. Preparation of Solid Catalysts, eds. G. Ertl, H. Knozinger and J. WeitkampWileyVCH, 1999. 193. M.P. Kiskinova, Stud. Surf. Sci. Catal., 1992, 70. 194. The term SMSI was introduced by Tauster et al. (S.J. Tauster, S.C. Funk, R.L. Garten, J. Am. Chem. Soc. 1978, 100, 170) to denote the effect responsible for the drastic decrease in CO and H2 chemisorption on titania-supported metals after increasing the reduction temperature from 200 to 5001C. More details on this effect can be found in a review paper of Hadjiivanov and Klissurski (K.I. Hadjiivanov, D.G. Klissurski, Chem. Soc. Rev., 1996, 25, 61). 195. H. Kolbel and K.D. Tillmetz, Germ. Pat. Appl., 1976, 2, 507–647. 196. B. Bussemeier, C.D. Frohning and B. Cornils, Hydrocarbon Process., 1976, 56, 105. 197. B. Bussemeier, C.D. Frohning, G. Horn and W. Kluy, Germ. Pat. Appl., 1976, 2, 518–964. 198. W.M. van Dijk, J.W. Niemantsverdriet, A.M. Van der Kraan and H.S. Van der Baan, Appl. Catal., 1982, 2, 273. 199. J. Barrault and C. Renard, Appl. Catal., 1985, 14, 133. 200. J. Barrault, C. Forquy and V. Perrichon, Appl. Catal., 1983, 5, 119. 201. R. Malessa and M. Baerns, Ind. Eng. Chem. Res., 1988, 27, 279. 202. G.C. Maiti, R. Malessa and M. Baerns, Appl. Catal., 1983, 5, 151. 203. G.Z. Bian, A. Oonuki, N. Koizumi, H. Nomoto and M. Yamada, J. Mol. Catal. A: Chemical, 2002, 186, 203. 204. K.B. Jensen and F.E. Massoth, J. Catal., 1985, 92, 109. 205. C.K. Das, N.S. Das, D.P. Choudhury, G. Ravichandran and D.K. Chakrabarty, Appl. Catal. A: General, 1994, 111, 119. 206. J. Yang, Y. Lieu, J. Chang, Y.N. Wang, L. Bai, Y.Y. Xu, H.W. Xiang, Y.W. Li and B. Zhong, Ind. Eng. Chem. Res., 2003, 42, 5066. 207. Y. Liu, J. Yang, L. Bai, Y. Li, L.J. Liu, H.W. Xiang, Y.W. Li and B. Zhong, Chin. J. Catal., 2003, 24, 299. 208. S.M. Rodulfo-Baechler, S.L. Cortes-Gonzalez, J. Orozco, A.J. Mora and G. Delgado, Revista Mex. Fisica, 2003, 49, 195. 209. M.R. Goldwasser, V.E. Dorantes, M.J. Perez-Zurita, P.R. Sojo, M.L. Cubeiro, E. Pietri, F. Gonzalez-Jimenez, Y.N. Lee and D. Moronta, J. Mol. Catal. A: Chemical, 2003, 193, 227. 210. G.R. Hearne and H. Pollak, Hyperfine Interact., 1991, 67, 559. 211. N. Koizumi, K. Murai, T. Ozaki and M. Yamada, Catal. Today, 2004, 89, 465. 212. L.Y. Xu, Q.X. Wang, D.B. Liang, X. Wang, L.W. Lin, W. Cui and Y.D. Xu, Appl. Catal. A: General, 1998, 173, 19. 213. J. Abbot, N.J. Clark and B.H. Baker, Appl. Catal., 1986, 26, 141. 214. E.L. Kugler, S.J. Tauster, S.C.Fung, US. Pat. 4,206,134 (1980). 215. S.T. Hussain, J. Chem. Soc. Pakistan, 1993, 15, 97. 216. S.T. Hussain, J. Chem. Soc. Pakistan, 1993, 15, 110. 217. S.T. Hussain, J. Chem. Soc. Pakistan, 1993, 15, 162. 218. S.T. Hussain, J. Chem. Soc. Pakistan, 1993, 15, 234.

Catalysis, 2006, 19, 1–40 219. 220. 221. 222. 223. 224. 225. 226. 227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 237. 238. 239. 240. 241. 242. 243. 244. 245. 246. 247. 248. 249. 250. 251. 252. 253.

39

S.T. Hussain, J. Chem. Soc. Pakistan, 1994, 16, 87. S.T. Hussain, J. Chem. Soc. Pakistan, 1995, 17, 133. S.T. Hussain, J. Trace Microprobe Techn., 1996, 14, 353. S.T. Hussain, J. Trace Microprobe Techn., 1996, 14, 361. S.T. Hussain, J. Trace Microprobe Techn., 1996, 14, 367. S.T. Hussain, J. Trace Microprobe Techn., 1996, 14, 681. S.T. Hussain, Ads. Sci. Techn., 1996, 13, 489. S.T. Hussain and M.A. Atta, Turkish J. Chem., 1997, 21, 77. S.T. Hussain and F. Larachi, J. Trace Microprobe Techn., 2002, 20, 197. T. Tercioglu and J.F. Akyurtlu, Appl. Catal. A: General, 1996, 136, 105. L.B. Shapovalova and G.D. Zakumbaeva, Petroleum Chem., 2000, 40, 151. M. van der Riet, G.J. Hutchings and R.G. Copperthwaite, J. Chem. Soc., Chem. Commun., 1986, 798. G.J. Hutchings, M. van der Riet and R. Hunter, J. Chem. Soc., Faraday Trans. 1, 1989, 85, 2875. S.E. Colley, R.G. Copperthwaite, G.J. Hutchings, S.P. Terblanche and M.M. Thackeray, Nature, 1989, 339, 129. Q. Liang, K. Chen, W. Hou and Q. Yan, Appl. Catal. A: General, 1998, 166, 191. M.J. Keyser, R.C. Everson and R.L. Espinoza, Appl. Catal. A: General, 1998, 171, 99. M.J. Keyser, R.C. Everson and R.L. Espinoza, Ind. Eng. Chem. Res., 2000, 39, 48. S.M. Rudolfo-Baechler, S.L. Gonzalez-Cortes, J. Orozco, V. Sagredo, B. Fontal, A.J. Morea and G. Delgado, Mater. Lett., 2004, 58, 2447. M. Jiang, N. Koizumi, T. Ozaki and M. Yamada, Appl. Catal. A: General, 2001, 209, 59. T. Riedel, M. Claeys, H. Schulz, G. Schaub, S.S. Nam, K.W. Jun, M.J. Choi, G. Kishan and K.W. Lee, Appl. Catal. A: General, 1999, 186, 201. K. Guse and H. Papp, Fres. J. Anal. Chem., 1993, 346, 84. D.G. Wickham and W.J. Croft, J. Phys. Chem. Solids, 1958, 7, 351. J.L. Gautier, E. Rios, M. Gracia, J.F. Marco and J.R. Gancedo, Thin Solid. Films, 1997, 311, 51. J.M. Jimenez Mateos, J. Morales and J.L. Tirado, J. Solid. State Chem., 1989, 82, 87. E. Rios, P. Chartier and J.L. Gaultier, Solid St. Sci., 1999, 1, 267. E. Vila, R.M. Rojas and O. Garcia-Martinez, Chem. Mater., 1995, 7, 1716. E. Vila, R.M. Rojas, J.L. Martin de Vidales and O. Garcia-Martinez, Chem. Mater., 1996, 8, 1078. S. Naka, M. Inagaki and T. Tanaka, J. Mater. Sci., 1972, 7, 441. F. Morales, O.L.J. Gijzeman, F.M.F. de Groot and B.M. Weckhuysen, Stud. Surf. Sci. Catal., 2004, 147, 271. F. Morales, F.M.F. de Groot, P. Glatzel, E. Kleimenov, H. Bluhm, M. Havecker, A. Knop-Gericke and B.M. Weckhuysen, J. Phys. Chem. B, 2004, 108, 16201. F. Morales, D. Grandjean, F.M.F. de Groot, O. Stephan and B.M. Weckhuysen, Phys. Chem. Chem. Phys., 2005, 7, 568. F. Morales, F.M.F. de Groot, O.L.J. Gijzeman, A. Mens, O. Stephan and B.M. Weckhuysen, J. Catal., 2005, 230, 310. N.N. Madikizela-Mnqanqeni and N.J. Coville, J. Mol. Catal. A: Chemical, 2005, 225, 137. Y. Zhang, M. Koike and N. Tsubaki, Catal. Lett., 2005, 99, 193. S. Storsaeter, O. Borg, E.A. Blekkan, B. Totdal and A. Holmen, Catal. Today, 2005, 100, 343.

40

Catalysis, 2006, 19, 1–40

254. S.A. Hosseini, A. Taeb and F. Feyzi, Catal. Comm., 2005, 6, 233. 255. F.M.T. Mendes, C.A.C. Perez, F.B. Noronha and M. Schmal, Catal. Today, 2005, 101, 45. 256. Y.H. Zhang, H.F. Xiong, K.Y. Liew and J.L. Li, J. Mol. Catal. A: Chemical, 2005, 237, 172. 257. H.F. Xiong, Y.H. Zhang, K. Liew and J.L. Li, J. Mol. Catal. A: Chemical, 2005, 231, 145. 258. T.K. Das, G. Jacobs and B.H. Davis, Catal. Lett., 2005, 101, 187. 259. A. Infantes-Molina, J. Merida-Robles, E. Rodriguez-Castellon, B. Pawelec, J.L.G. Fierro and A. Jimenez-Lopez, Appl. Catal. A: General, 2005, 286, 239. 260. D.Y. Xu, W.Z. Li, H.M. Duan, Q.J. Ge and H.Y. Xu, Catal. Lett., 2005, 102, 229. 261. Z.W. Liu, X.H. Li, K. Asami and K. Fujimoto, Catal. Today, 2005, 104, 41.

The Catalysis of Biodiesel Synthesis BY EDGAR LOTERO, JAMES G. GOODWIN, JR., DAVID A. BRUCE, KAEWTA SUWANNAKARN, YIJUN LIU AND DORA E. LOPEZ Department of Chemical Engineering, Clemson University, Clemson SC 296340909, USA

1

Introduction

The term biodiesel was originally coined to describe unmodified vegetable oils or alkyl ester derivatives of vegetable oils or animal fats that could substitute for diesel fuel (DF). However, most bioderived oils and fats cannot be directly used as DF due to their high viscosity and propensity to incomplete combustion, and coking.1 Thus, the term has more recently been used to describe DF substitutes that consist of mono alkyl esters of long chain fatty acids prepared from renewable lipid feedstocks, such as vegetable oils and animal fats. Modern biodiesel, being made of smaller molecular species, does not show the unmanageable viscosity problems of unmodified oils and fats and can be burned more easily, reducing combustion and gumming problems. In addition, biodiesel can be used in diesel engines without engine modifications with comparable efficiency to the use of petroleum-base DF.2 For instance, pure biodiesel releases about 90% of the energy that petroleum-based diesel does; hence, its engine performance is nearly the same in terms of engine torque and horsepower. Two of the greatest advantages of biodiesel are its renewable nature and environmentally benign character. Biodiesel synthesis from biomass offers a means to produce a transportation fuel that is biodegradable and provides a pathway for the recycling of carbon dioxide emissions, thereby, greatly reducing the emission of green house gases. In general, biodiesel compares well to petroleum-based diesel (Table 1). However, biodiesel shows a number of characteristics that makes it more desirable than traditional petroleum-based diesel fuel: higher cetane number, no aromatics, almost no sulfur, and 10 to 11 percent oxygen by weight. As a result, the emission profile of biodiesel and biodiesel-diesel blends, compared to petroleum-based diesel, is much cleaner and has substantially lower sulfur emissions (Table 2).3 In biodiesel preparation, vegetable oils and animal fats are typically employed as biomass feedstocks, but waste greases such as yellow grease and brown grease can also be used. Thus, biodiesel synthesis provides a means for


41

42

Table 1

Catalysis, 2006, 19, 41–83

The American Society for Testing and Materials (ASTM) standards of maximum allowed quantities in diesel and biodiesel16

Property

Diesel

Biodiesel

Standard Composition Kin. viscosity (mm2/s) at 401C Specific gravity (g/mL) Flash point (1C) Cloud point (1C) Pour point (1C) Water, vol% Carbon, wt% Hydrogen, wt% Oxygen, wt% Sulfur, wt% Cetane number HFRRc, microns BOCLEd scuff (g)

ASTM D975 HCa (C10–C21) 1.9–4.1 0.85 60–80 15 to 5 35 to 15 0.05 87 13 0 0.05 40–55 685 3,600

ASTM D6751 FAMEb (C12–C22) 1.9–6.0 0.88 100–170 3 to 12 15 to 16 0.05 77 12 11 0.05 48–60 314 47,000

a Hydrocarbons. b Fatty Acid Methyl Esters. c High Frequency Reciprocating Rig. inder Lubricity Evaluator.

d

Ball-on-Cyl-

Table 2

Average B100 and B20 emissions compared to normal diesel3

Emission

B100

B20

Carbon monoxide Total unburned hydrocarbons Particulate matter Nitrogen oxides Sulfates Air toxics Mutagenicity

48% 67% 47% þ10% 100% 60% to 90% 80% to 90%

12% 20% 12% þ2% 20% 12% to 20% 20.0%

waste utilization and environmental protection. Currently, there is an urgent need to develop chemical catalysts that can advance biotechnology capabilities in terms of biomass processing given the complexity and component diversity of biomass feedstocks.4 Such catalysts are necessary to cost effectively obtain biobased products, including fuels. Although in the U.S. and elsewhere there are commercial processes for converting vegetable oils and animal fats to biodiesel, these processes are not for the most part cost-competitive with current petroleum-based DF production methods except with tax incentives. Primarily, there is a need to develop improved continuous processing technologies rather than batch processing technologies, to use heterogeneous catalysts to replace current homogeneous catalysts, and to incorporate catalysts that are effective for a broader spectrum of reactants and can tolerate higher levels of impurities. Some methodologies proposed in the past, such as the pyrolysis and cracking of oils and fats, can produce compounds that are smaller than their triglyceride

Catalysis, 2006, 19, 41–83

43

(TG) source and, therefore, suitable to be used as a fuel.5,6 However, pyrolysis is not very selective and a wide range of compounds is usually obtained. Depending on the TG source and the pyrolytic method employed, alkanes, alkenes, aromatic compounds, esters, CO2, CO, water, and H2 are produced in varying proportions. Oxygen removal from substrate molecules is another downside of pyrolytic production methods. Fuels obtained by pyrolysis of TGs have a lesser environmentally benign edge over petroleum-derived fuels in terms of oxygen content. Also, solid residues of ash and carbon can be created during TG pyrolysis, requiring additional separation steps. Catalytic cracking has been used in an effort to control the types of products generated by TG cracking, using a vast variety of catalysts,1,7 including: SiO2/Al2O3,8 NiMo/g-Al2O3,9 NiSiO2,10 Al2O3,11 MgO,11 composites of zeolites,12 and Al-MCM-41.13,14 However, a gasoline-like fuel is more likely to be formed than a diesel-like fuel.15 This paper reviews biodiesel synthesis by transesterification of triglycerides and esterification of free fatty acids present in vegetable oils and animal fats from a catalytic standpoint. The first half includes the most important aspects of current biodiesel production schemes that employ homogeneous alkali and acid catalysts. A complete review of these reactions and production methods is not included as they have been extensively reviewed elsewhere.5,6,16 In the second half, the subject of biodiesel synthesis using heterogeneous solid catalysts, such as metals, anchored metal complexes, solid bases and solid acids is reviewed in greater depth since this is the first review with this emphasis. In this part, both transesterification and esterification have been examined given the increasing importance of the latter as the amount of free fatty acids in the feedstock becomes larger, as is the case for low cost feedstocks. The reader is also encouraged to read other excellent reviews in the field with emphasis on homogeneous alkali-,6 acid-,16 enzyme-,5 and guanidine-,17 catalyzed conversions of lipid feedstocks.

2

Overview

2.1 Vegetable Oils and Animal Fats. – Oils and fats belong to an extensive family of chemicals called lipids. Lipids are bio-products from the metabolism of living creatures. As a result, they can be found widely distributed in nature. Their bio-functions are diverse, but they are most known for their energy storage capacity. Most lipids can easily dissolve in common organic solvents, meaning that they are hydrophobic. If a lipid is a solid at 251C, it is classified as a fat; otherwise, it is an oil. Typically, fats are produced by animals and oils by plants, but both are mainly made of TG molecules, which are tri-esters of glycerol (a triol) and free fatty acids (long alkyl chain carboxylic acids). Other glyceride species, such as di-glycerides and mono-glycerides, are obtained from TGs by the substitution of one and two fatty acid moieties, respectively, with hydroxyl groups (Figure 1). Typical fatty acid compositions of common vegetable oils and animal fats are shown in Table 3. As can be inferred, fats and oils are chemically equivalent; their differences arise from variations in their

44

Catalysis, 2006, 19, 41–83

Figure 1 Chemical structures of vegetable oils and animal fats (R1, R2, R3 ¼ alkyl groups)

compositional fatty acid building blocks. Fats have a greater percentage of saturated fatty acids as building units, while oils have more unsaturated ones. Fats and oils usually contain fatty acids in their free form as a result of spontaneous hydrolysis of the parent TG compounds. These free fatty acids (FFAs) are usually linear molecules with 4–24 carbon atoms that may be saturated or unsaturated with typically 1–3 C¼C double bonds. Other compounds, such as pigments, waxes, sterols, glycolipids, lipoproteins, hydrocarbons, long chain alcohols, carbohydrates and vitamins (E, A and D), can also be found in oils and fats in minor concentrations. 2.2 Reactions. – 2.2.1 Transesterification. The transesterification reaction, shown in Figure 2, involves TGs catalytically reacting with an alcohol (simple linear alcohols such as MeOH, EtOH, PrOH, and ButOH are generally used) to form glycerol (a by-product) and the mono alkyl esters that constitute the biodiesel fuel. In the reaction, a TG molecule reacts with three alcohol molecules sequentially in the presence of a catalyst (acid or base) to produce first a diglyceride (DG), then a monoglyceride (MG) and finally a glycerol (GL) product and three molecules of monoester. All reaction steps are reversible with the net equilibrium favoring the formation of products.18–20 The reaction can use either a base or an acid catalyst, but base catalysts are preferred because they give rise to a faster reaction under mild reaction conditions. 2.2.2 Esterification. The esterification reaction (Figure 3) involves the reaction of a FFA with an alcohol (usually a low molecular weight alcohol, such as MeOH, EtOH, PrOH, and ButOH) to produce an alkyl ester (biodiesel) and water. Either base or acid catalysts can be used for the reaction. However, base catalysts can only be used at high temperatures (or catalyst deactivation takes place by soap formation). More commonly, acid catalysts such as sulfuric acid are employed to carry out the esterification reaction under mild conditions.

1.7 4.8 2.4 1.7

0.1

14 : 00

Myristic

16 : 00 3.5 9.2 6.1 8.6 10.6 35.0 23.3 22.2 17.3 28.4 23.2 22.8

Palmitic

3.8 3.1

3.5 8.4 1.9

0.8

16 : 01

Palmitoleic 18 : 00 0.9 3.4 3.3 1.9 4.8 7.0 11.0 5.1 15.6 14.8 13.0 12.5

Stearic 18 : 01 64.4 80.4 16.9 11.6 22.5 44.0 47.1 42.3 42.5 44.6 44.3 42.4

Oleic

Fatty Acid Composition, wt.%

Typical fatty acid compositions of vegetable oils and animal fats

C-length:(no.¼bonds) Rapeseed oil Virgen Olive oil Sunflower oil Safflower oil Soybean Palm oil Choice white grease Poultry fat Lard Tallow Yellow grease Brown grease

Table 3

18 : 02 22.3 4.5 73.7 77.9 52.3 14.0 11 19.3 9.2 2.7 7.0 12.1

Linoleic

0.7 0.8

1.0 1.0 0.4

8.2

18 : 03 8.2 0.6

Linolenic

4.4 12.6 9.4 10.5 15.5 42.0 37.8 35.7 34.6 52.0 38.6 37.0

Sat. (%)

56 56 56 56 23 7 8 8 9 10 23 23

References

Catalysis, 2006, 19, 41–83 45

46

Catalysis, 2006, 19, 41–83

Figure 2 Transesterification reactions of glycerides with methanol

Figure 3 Esterification reactions of a FFA with methanol

2.3 Physicochemical Properties of Biodiesel. – The redox characteristics of biodiesel make it a reducing agent for materials, such as brass, bronze, cooper, lead, tin and zinc. For this reason, contact of biodiesel with these materials must be avoided.21 Materials such as aluminum, steel, fluorinated polymers and Teflon do not react with biodiesel and can be used to handle it. In addition, biodiesel shows mild solvent properties; hence, biodiesel contact with painted or varnished surfaces as well as rubber devices, such as hoses, seals and gaskets, may cause problems. Animal fats or plant oils can pass on to the final biodiesel product particular characteristics derived from their differences in composition.2 For instance, if biodiesel is produced from the transesterification of a representative vegetable

Catalysis, 2006, 19, 41–83

47

oil with a small alcohol such as methanol (see below), it will primarily contain compositional units from oleic and linoleic acids, which have unsaturated alkyl chains. In general, the length of the alkyl chains of the acid moieties and the degree of chain unsaturation correlate well with the cold flow performance of a specific batch of biodiesel fuel. For this reason, biodiesel fuels with higher levels of olefinic units or shorter alkyl chains tend to have lower melting points, which translate into better cold flow fuel performances, as determined by the fuel’s cloud-point and pour-point. Cloud-point refers to the temperature whereupon the fluid becomes cloudy, signaling the formation of small, suspended wax crystallites that are known to clog filters and fuel lines.1 Similarly, pour-point refers to the lowest temperature at which biodiesel can be handled without excessive formation of wax crystals. On the other hand, the storage stability of biodiesel is adversely affected by the presence of unsaturated alkyl components. The olefinic moieties in biodiesel fuel can undergo oxidative degradation via exposure to air with deleterious results, including formation of solids and gums.21,22 The degree of oxidative degradation has been shown to increase with fuel unsaturation. Contrary to what is seen with vegetable oils, biodiesel produced from animal fats has component units that are dominated by saturated alkyl species, and the observed biodiesel properties reflect the presence of these species. The fuel is more stable and degrades less in the presence of air. However, cloud-points and pour-points are higher, meaning that it may show poor performance at cold temperatures.

2.4 The Feedstock Issue. – Regardless of the latter mentioned disadvantages for biodiesel synthesized from animal fats, academic research has more often focused on the use of animal fats and waste greases, such as yellow grease and brown grease, given their more attractive prices, as an alternative to virgin oils. Still, however, production of vegetable oil amply surpasses production of any other type of lipid feedstock and, thus, dominates the production of biodiesel. In the United States alone, there are approximately 2.75 billion pounds of waste recyclable restaurant grease,23 11.64 billion pounds of animal derived fat, and 23.66 billions pounds of vegetable oil generated annually (Tables 4 and 5).23 Yellow grease, which is obtained from rendered animal fats and waste restaurant cooking oil, has free fatty acid levels of less than 15 wt% (6–15 wt%) and sells for $0.09–0.20/lb.23 Brown grease, also known as trap grease, refers to grease collected from traps installed in commercial, municipal or industrial sewage facilities which separate water from oil and grease in wastewater. In some circles, highly degraded animal fats are also classified as brown grease. This waste grease has FFAs levels greater than 15 wt% and can sell for $0.01–0.07/lb less than yellow grease. The use of cheap feedstocks such as brown and yellow grease (Table 6),23 the need for alternative uses of animal derived fats, the increasing rate of vegetable oil production and the development of new technology should make biodiesel more economical in the near future.

48

Table 4

Catalysis, 2006, 19, 41–83

An estimate of the total annual production of vegetable (US averages 1995–2000)23 Vegetable Oil Production (Billion pounds/yr)

Soybean Peanuts Sunflower Cottonseed Corn Others Total Veg. Oil

Table 5

18.34 0.22 1.0 1.01 2.42 0.669 23.659

An estimate of the total annual production of animal fats (US averages, 1995–2000)23 Animal Fats (Billion pounds/yr)

Edible Tallow Inedible tallow Lard & Grease Yellow Grease Poultry Fat Total Animal Fat

Table 6

1.625 3.859 1.306 2.633 2.215 11.638

Biodiesel feedstock pricing and impact on production cost (2004)23

Feedstock Crude Soybean Oil Tallow, Inedible Yellow Grease (o10% FFA) Brown Grease (420% FFA)

Price per Estimated Pounds of Feedstock Feedstock Cost per Gallon of Pound per Gallon of Biodiesel Biodiesel Produced $0.25

7.5

$1.88

$0.18 $0.12

7.5 8

$1.35 $0.96

$0.05

8

$0.40

2.5 Processing Methodologies. – Most biodiesel in the US is currently produced in batch reactors. Currently, batch plants typically produce between 500–10,000 tons of biodiesel per year. However, this number could be substantially improved with the use of continuous processing technologies, which could easily generate over 10,000 tons/year. At present, most continuous production plants are found in Europe, where the demand for biodiesel is higher. A few continuous processes for the production of biodiesel have recently been reported in the literature.24–28 The production of biodiesel is expected to prominently increase in the near future due to current concerns with the long-term supply of conventional hydrocarbon-based fuels and a growing demand for fuels from renewable sources.4 Significant biodiesel production will only be achieved by going to continuous processing.

Catalysis, 2006, 19, 41–83

49

Although continuous production plants can achieve higher biodiesel throughputs and are less costly to operate per biodiesel unit, batch plants are less expensive to build and can more easily be adapted to changing raw materials and reaction conditions. This flexibility is particularly important given the economic impetus to use diverse TG feedstocks for biodiesel production. Nevertheless, the current trend is toward the construction of continuous production plants given their higher production capacity and lower operational cost, in accordance with current biodiesel demand.29

3

Homogeneous Catalysis

3.1 Base-Catalyzed Synthesis. – 3.1.1 The Fundamentals. The accepted mechanistic route for transesterification under alkaline conditions is presented in Figure 4.6,17 The sequence of steps can be summarized as follows: first, the base catalyst reacts with alcohol producing the catalytically active species, RO (when the base is an alkali alkoxide, simple dissociation gives rise to the catalytically active species, RO). Second, a tetrahedral intermediate is formed by nucleophilic attack on a carbonyl carbon in the TG. Third, the tetrahedral intermediate breaks down into a fatty acid ester and a diglyceride anion. Fourth, proton transfer to the diglyceride ion regenerates the RO catalytically active species. This sequence is then repeated twice to yield first a monoglyceride intermediate and finally the glycerol product and biodiesel. In their pioneering work, Freedman et al. studied the kinetics of the alkalicatalyzed transesterification of soybean oil with methanol and n-butanol.18 Three regimes categorized the overall reaction process with the rate-limiting reaction step changing over time according to the observed reaction rate. Initially, the reaction was mass transfer limited because of the low miscibility of reagents; i.e., the non-polar oil phase was immiscible with the polar alcoholcatalyst phase, slowing down the reaction. As ester products were formed, these species acted as an emulsifying agent giving rise to a second rate regime that was kinetically-controlled and characterized by a sudden surge in product formation. Finally, in the last stages of reaction, a third regime was reached characterized again by a slower reaction rate. These authors used 6 : 1 and 30 : 1 alcohol-to-oil molar ratios for both methanol and butanol.18 As expected, a pseudo first-order reaction was found at large excess of alcohol for both alcohols. At low excess alcohol, however, the butanolysis reaction (301C) was second-order, but the methanolysis reaction (401C) was reported to be a combination of a second-order consecutive reaction and a fourth-order shunt reaction. The shunt reaction, in which three methanol molecules simultaneously attack a TG molecule, was adopted to better fit the kinetic data. However, such a reaction is highly unlikely. Nureddini et al.19 later found that the inclusion of a shunt mechanism was not necessary to fit the kinetic data of the transesterification reaction, and Boocock et al.30 showed that the shunt reaction assumption came as a misinterpretation of the observed kinetics. At low temperatures (20–401C) the multiphase methanolysis reaction

50

Catalysis, 2006, 19, 41–83

Figure 4 Homogeneous base-catalyzed reaction mechanism for the transesterification of triglycerides

is driven by oil solubility in the methanol-catalyst phase. Methanolysis as well as butanolysis should both be second order reactions at low excess alcohol, assuming that no diffusion limitations exist.30 Thus, the difference in initial reaction rates for the methanol and butanol systems can be traced back to the higher solubility of oil in the butanol-catalyst phase. Activation energies for the transesterification reactions involving methanol have been reported in the range of 6–20 kcal/mol.19,20 Reaction rate constants

51

Catalysis, 2006, 19, 41–83

(k) for the methanolysis of TG, DG and MG have also been calculated. In general, k values increased with temperature for TG to DG and DG to MG for both forward and reverse reactions. However, rate constants for the formation of glycerol, MG#GL, decreased with temperature. The values for the k(forward reaction) followed the series kMG,f4kDG,f4kTG,f at temperatures below 601C. Typical k values for all forward and reverse reactions are shown in Table 7.19 Homogeneous base catalysts used for the methanolysis of lipids include alkaline metal compounds, such as: CH3ONa and CH3OK, NaOH and KOH, and Na2CO3 and K2CO3. Alkaline metal alkoxides, such as CH3ONa, are the most effective catalysts for transesterification.30,31 As low as 0.5% (by weight of oil) of catalyst allows high yields (498%) in very short times. However, some studies have reported that NaOH has better activity than CH3ONa.32,33 In general, alkaline hydroxides catalysts (NaOH and KOH) are preferred because they are cheaper and easier to handle than alkaline alkoxides.34 When using hydroxides, small amounts of soap are expected to be produced due to the formation of water in the initial stages of reaction (step 1(b), Figure 4) leading to the hydrolysis of alkyl ester products to FFAs and consequently to soap formation. Formation of soap translates into catalyst deactivation and loss of productivity; hence, it must be avoided. In contrast, alkaline carbonates reduce the impact of soap production by forming bicarbonates instead of water (step 1(c), Figure 4).17 However, the carbonate anion is a weaker base, which translates into lower concentrations of the active RO species, slower reaction rates, and the need for higher amounts of the carbonate catalyst (2–3 wt%) in order to achieve yields comparable to those obtained with alkoxide or hydroxide catalysts. One important variable in the transesterification reaction is the molar ratio of reactants. The reaction’s stoichiometric ratio requires three moles of alcohol for each mole of TG to produce three moles of mono alkyl ester and one mole of glycerol. In general, the greater the molar ratio, alcohol:TG, the higher the conversion for a given time. An alcohol-to-oil molar ratio of 6 : 1 has been Table 7

Average reaction rate constants and activation energies at 501C for the methanolysis of soybean oil using NaOH21 Activation energy (kcal/mol)a

Reaction TG # DG Forward Reverse DG # MG Forward Reverse MG # GL Forward Reverse a

Rate constant (wt%.min)1

Arrhenius

Modified Arrhenius

0.050 0.110

13.2 9.9

11.7 8.5

0.215 1.228

19.9 14.6

18.4 13.4

0.242 0.007

6.4 9.6

7.9 11.0

Arrhenius relationship, k=AeEa/RT; modified Arrhenius relationship, k=ATneEa/RT, n ¼ 1.

52

Catalysis, 2006, 19, 41–83

suggested as optimal to achieve high conversions in reasonable times.31,35 Similar results were observed with the transesterification of oils with linear alcohols of higher molecular weight.36 Generally, transesterification is carried out near the boiling point of the alcohol, commonly methanol, at atmospheric pressure and under good stirring.37 Good mixing intensity is important when dealing with mixtures of small polar alcohols and hydrophobic vegetable oils or animal fats.19,38,39 As previously mentioned, when put in contact, methanol and TGs form a biphasic system giving rise to a diffusion-controlled reaction regime with poor reaction rates. To counteract phase separation, mechanical mixing is normally applied to achieve efficient mass-transfer among reactant phases. It has been reported that, for a soybean oil/methanol system at 701C and high mixing speeds (ca. 600 rpm), mass transfer limitations are almost non-existent.19 The kinetic-controlled regime, which follows the first mass-transfer controlled regime, is promoted by both the mixing intensity and the formation of biodiesel esters that act as emulsifying agents for the reaction mixture. The final reaction regime, which is also characterized by a slower reaction rate, appears for reasons still not well understood, but some authors have suggested it might result from a fall in catalyst concentration and/or a polarity effect due to methanol depletion, product formation and methanol-oil mixing.40 A reaction scheme using tetrahydrofuran (THF) as a co-solvent has been found to overcame the immiscibility problem of the small-alcohol/oil phases, eliminating the initial induction period for dilution of TGs in the methanol phase and yielding faster reaction times.30 Under these conditions reactor stirring was not required and both methanol and THF could be easily recycled by co-distillation since both have similar boiling points. In addition, glycerol separation was achieved faster due to the increased difference in phase polarity. Other organic co-solvents, such as toluene and methyltertiarybutylether (MTBE), have also been investigated with similar results.25,40,41 3.1.2 Base-Catalyzed Biodiesel Processing. Currently, most commercially available biodiesel is produced by base-catalyzed processes that employ NaOH as the active catalyst due to its lower cost. Base-catalyzed processes show procedural differences arising from the type of lipid feedstock employed and the pretreatment applied to it. In addition, they may differ in operational parameters, such as the use of co-solvents, reagents, temperature, pressure, batch or continuous processing, etc. One particular advantage of using alkaline catalysts is that they give rise to a relatively fast reaction. In general, basecatalyzed processes are carried out at low temperatures and pressures (60–651C and 1.4–4.2 bar) with low catalyst concentrations (0.5–2 wt%).24,42 Table 8 shows typical reaction conditions for base-catalyzed transesterification processes in biodiesel synthesis. Even though the base-catalyzed process seems operator friendly and economically possible, it suffers from a key limitation: only refined oils and pretreated fats with low concentrations of FFAs be used to produce biodiesel using homogeneous base catalysts. FFAs can react with the base catalyst giving

53

Catalysis, 2006, 19, 41–83

Table 8

Typical reaction conditions for biodiesel synthesis using homogeneous base catalysis24,44 Base-Catalyzed Biodiesel Synthesis

Feedstocks

Triglyceride mixtures with low free fatty acid contents (o0.5%) e.g., Refined vegetable oils þ Anhydrous short chain alcohol (generally, methanol) 6:1

Alcohol-to-oil molar ratio (recommended) Temperature 60–651C Pressure 1.4–4.1 bar Catalyst NaOH (most common) Catalyst concentration (by 0.5–2 wt% weight of lipid feedstock) Conversions Z 95% can be expected after 1 h reaction

Figure 5 (a) Reaction of the base catalyst with FFAs to produce soap and water, both undesirable by-products. (b) Ester hydrolysis due to reaction with water forming FFAs

rise to saponification (Figure 5a). The presence of soap causes an increase in viscosity and the formation of gels, which complicate the glycerol-monoalkyl ester separation process. Water also contributes to soap production since water can react with the monoalkyl ester product to form FFAs (Figure 5b). For these reasons, ideally, the lipid feedstock used in base-catalyzed transesterifications should not exceed a FFA content of 0.5% (by weight), and both alcohol and catalyst must be essentially anhydrous.32 Otherwise, production yields are proportionally affected. In order to keep feedstock prices low (refined oils are expensive), biodiesel plants generally refine crude oils themselves.43 The refining process consists of four steps: oil degumming, caustic treatment, bleaching and dewaxing (Figure 6). Oil degumming is simply a liquid-liquid extraction process that uses water to selectively remove polar organics such as phospholipids and lecithin. Following extraction, the two immiscible phases are commonly separated by centrifuging. If gums are not removed, they will give rise to a bottom layer with increased density and viscosity, making the oil difficult to handle. Caustic treatment of the degummed oil removes FFAs. During caustic treatment, the oil is washed with a dilute solution of sodium or potassium hydroxide

Figure 6

Steps involved in oil refining: (a) degumming, (b) caustic treatment, (c) bleaching, and (d) dewaxing

54 Catalysis, 2006, 19, 41–83

Catalysis, 2006, 19, 41–83

55

followed by centrifuging. Here, FFAs are separated as soap in the aqueous layer. Bleaching uses special clays that allow the removal of pigments, minerals, gums and FFA trace impurities. After use, the clays are separated by filtration and discarded. Finally, dewaxing removes unwanted waxes. Here, waxes are esters of long chain alcohols and fatty acids. During dewaxing, the oil is cooled down to a low temperature (6–81C), where waxes crystallize. Then, the oil is carefully heated to about 181C and filtered to extract the wax crystals. If waxes are left in the oil, the biodiesel obtained will exhibit higher pour and cloud points. After refining, the oil is ready for the base-catalyzed transesterification with, most commonly, methanol. Figure 7 depicts a simplified block flow diagram (BFD) for a typical biodiesel production process using base catalysis.24,44 In the first step, methanol and catalyst (NaOH) are mixed with the aim to create the active methoxide ions (Figure 4, step 1(b)). Then, the oil and the methanol-catalyst solution are transferred to the main reactor where the transesterification reaction occurs. Once the reaction has finished, two distinct phases are formed with the less dense (top) phase containing the ester products and unreacted oil as well as some residual methanol, glycerol, and catalyst. The denser (bottom) layer is mainly composed of glycerin and methanol, but ester residues as well as most of the catalyst, water, and soap can also be found in this layer. In general, a vacuum distillation step is used for methanol recycle prior to glycerin purification. Next, the remaining base catalyst in the crude glycerin is commonly neutralized with low cost mineral acids, such as phosphoric acid. This operation also converts the soaps back to FFAs. After neutralization, three distinct phases form: a low density (top) layer containing FFAs, a dense (bottom) liquid layer composed of glycerin, water and alcohol, and a third layer made of salt precipitates. These three phases are then separated with the nonglycerin layers being treated as waste. Glycerin is further purified by distillation to remove water and alcohol. This procedure yields 90–95% pure glycerin, which can be sold at market price. Crude biodiesel is initially purified by thoroughly washing the ester phase with water or by neutralization with a polyprotic mineral acid to eliminate base catalyst residues. Next, in a settling tank, an aqueous phase, salt precipitates, and biodiesel are separated. Another water washing step follows to further remove polar compounds that might still be present in the biodiesel product. Finally, the biodiesel is vacuum distillated at moderate-to-high temperatures (around 190–2701C) to comply with ASTM specifications (99.6% or purer).24 3.2 Acid-Catalyzed Synthesis. – 3.2.1 The Fundamentals. Homogeneous acid catalysts, such as sulfuric acid, phosphoric acid, hydrochloridric acid, organo sulfonic acids and others, can be used to catalyze the transesterification of TGs and the esterification of FFAs to produce biodiesel type monoesters.24,45–47 Nevertheless, because the acid-catalyzed transesterification is about 3 orders of magnitude slower than the alkali-catalyzed reaction for comparable amounts of catalyst,18 base catalysts have received the most attention in both the academic and industrial sectors. In addition, the corrosiveness of strong liquid acids and

Figure 7

Simplified BFD for a typical base-catalyzed process for the production of biodiesel (based on references

24,44

)

56 Catalysis, 2006, 19, 41–83

Catalysis, 2006, 19, 41–83

57

the environmental threat that they pose have also been deterrents to their use. However, given the versatility of acid catalysts to deal with FFAs, their use has been proposed as an economically viable alternative to base catalysts for the production of biodiesel from low cost feedstocks such as yellow grease.24,48 The accepted chemical mechanism for the homogeneous acid-catalyzed transesterification is illustrated in Figure 8.16,17 The sequence of steps can be summarized as follows: first, the TG carbonyl group is protonated by the acid catalyst. Second, the activated carbonyl group undergoes nucleophilic attack from an alcohol molecule, forming a tetrahedral intermediate. Third, solvent assisted proton migration gives rise to a good leaving group, promoting (fourth) the cleavage of the hemiacetal species (tetrahedral intermediate) and yielding a protonated alkyl monoester and a diglyceride molecule. Fifth, proton transfer regenerates the acid catalyst. This sequence is repeated twice to ultimately yield 3 alkyl monoesters and glycerol as products. Note how the key interaction boosting the catalytic effect is the protonation of the carbonyl group on the TG. Such catalyst-substrate interaction increases the electrophilicity of the adjacent carbonyl carbon atom, making it more susceptible to nucleophilic attack. Compare this to the base-catalyzed mechanism where the base catalyst takes on a more direct route to activate the reaction, creating first an alkoxide ion that directly acts as a strong nucleophile (Figure 4). Ultimately, it is this crucial difference, i.e., the formation of a more electrophilic species (acid catalysis) vs. that of a stronger nucleophile (base catalysis), that is responsible for the differences in catalytic activity. Compared to the base-catalyzed synthesis of biodiesel, fewer studies have dealt with the subject of acid-catalyzed transesterification of lipid feedstocks. Among acid catalysts, sulfuric acid has been the most widely studied. In the previously mentioned work of Freedman et al., the authors examined the transesterification kinetics of soybean oil with butanol using sulfuric acid.18 The three reaction regimes observed (in accordance with reaction rate) for basecatalyzed reactions were also observed here. A large molar ratio of alcohol-tooil, 30 : 1, was required in this system in order to carry out the reaction in a reasonable time. As expected, transesterification followed pseudo-first-order kinetics for the forward reactions (Figure 2), while reverse reactions showed second-order kinetics. Acid-catalyzed reactions require the use of high alcohol-to-oil molar ratios in order to obtain good product yields in practical reaction times. However, ester yields do not proportionally increase with molar ratio. For instance, for soybean methanolysis using sulfuric acid, ester formation sharply improved from 77% using a methanol-to-oil ratio of 3.3 : 1 to 87.8% with a ratio of 6 : 1. Higher molar ratios showed only moderate improvement until reaching a maximum value at a 30 : 1 ratio (98.4%).45,46,49 In general, alcohols, such as methanol, ethanol, propanol, butanol, amyl alcohol, etc., can be used with acid-catalyzed transesterifications to obtain high biodiesel yields.50 Methanol is preferred due to its low cost and wide availability, but bioderived ethanol would be ideal for the synthesis of a fully biogenerated fuel. On the other hand, the use of heavier alcohols like butanol

58

Catalysis, 2006, 19, 41–83

Figure 8 Homogeneous acid-catalyzed reaction mechanism for the transesterification of triglycerides

with higher boiling points could enable carrying out the synthesis of biodiesel at higher temperatures, enhancing oil-methanol solubility and reaction kinetics, while maintaining moderate-to-low pressures. Indeed, higher temperatures help minimize the initial mass-transfer controlled regime in transesterification and

Catalysis, 2006, 19, 41–83

59

promote more energized molecular collisions -two factors that translate into improved reaction times.31,34,51 For instance, when reacting oils having high FFA content with methanol and 1.7 wt% H2SO4 at 2401C and 70 bar, ester conversions over 90% were obtained in just 15 min.52 Higher temperature and pressure can also lead to the formation of unwanted products, such as alkyl ethers formed by the dehydration of alcohols. One advantage of acid catalysts over base catalysts is their low susceptibility to the presence of FFA in the starting feedstock. However, acid-catalyzed transesterification is especially sensitive to water concentration. Canakci and Van Gerpen showed that as little as 0.1 wt% water in the reaction mixture was able to substantially affect ester yields in transesterification of vegetable oil with methanol, with the reaction almost completely inhibited at 5 wt% water concentration.46 It was established that water content has to be kept under 0.5 wt% to achieve higher than 90% ester yield under their reaction conditions (601C, methanol-to-oil molar ratio 6 : 1, 3 wt% sulfuric acid and 96 h). Just recently, Kusdiana and Saka published their results on the effect of water on methyl ester formation by the transesterification of rapeseed oil with methanol using base and acid catalysts (1.5 wt% NaOH and 3 wt% H2SO4).53 Their studies found that water concentration was more critical in acid catalysis than in base catalysis, in agreement with the studies by Canakci et al.46 However, neither study addressed the cause for the observed difference in transesterification sensitivity to water using either base or acid catalysts. Sridharan and Mathai noticed that the transesterification of small esters under acid-catalyzed conditions was retarded by the presence of spectator polar compounds.54 Thus, given that water can form water-rich clusters around protons (solvent-proton complexes) with less acid strength than methanol-only proton complexes,55 some catalyst deactivation may be expected with increased water concentrations. Also, water-rich methanol proton complexes should be less hydrophobic than methanol-only clusters, thus making it more difficult for the catalytic species (H1) to approach the hydrophobic TG (and possibly DG) molecules and contributing to catalyst deactivation. Therefore, with water present in the feedstock or produced during the reaction in significant quantities, some catalyst deactivation can take place by hydration. Indeed, increased water concentration affects transesterification more than it affects esterification.46,53 This is due to the presence of polar carboxylic functional groups in FFAs that allows FFAs to more easily interact with polar compounds, facilitating the alcoholysis reaction.

3.2.2 Acid-Catalyzed Biodiesel Processing. The acid-catalyzed process does not enjoy the same popularity in commercial application as its counterpart, the base-catalyzed process, due to the need for longer reaction times. Nonetheless, the fact that acid catalysts can simultaneously carry out esterification of FFAs and transesterification of TGs is seen as a significant advantage, which could help in the processing of low-cost, low-quality feedstocks (generally high in FFAs), thereby lowering overall production costs.24,42

60

Catalysis, 2006, 19, 41–83

In general, acid catalyzed reactions are performed at high alcohol-to-oil molar ratios, low-to-moderate temperatures and pressures, and high acid catalyst concentrations. Table 9 summarizes reactions conditions proposed by Zhang et al. to prepare biodiesel from waste cooking oil using sulfuric acid as the catalyst.24 A simplified BFD of the acid process is shown in Figure 9. The acid-catalyzed biodiesel synthesis process involves the following. First, fresh methanol, recycle methanol and sulfuric acid are injected into the main reactor together with the lipid feedstock. There, simultaneous transesterification and esterification take place (reaction conditions listed in Table 9). Upon completion of the reaction, products and unreacted methanol are separated by distillation. Recovery and recycle of methanol are required due to the large methanol excess present that otherwise would hinder product separation and purification. Following methanol recovery, catalyst neutralization is carried out with CaO, producing CaSO4 precipitates and water. The precipitates are removed in a gravity separator and the liquid mixture is washed with water. At this point, the mixture is allowed to settle forming two liquid phases. The bottom layer, which is mainly glycerin, is distillated to obtain glycerin 92% pure. The upper phase with biodiesel, some water, methanol and unreacted oil is distillated under vacuum at moderate-to-high temperatures to avoid biodiesel decomposition, obtaining biodiesel 99.6% or purer. 3.3 Integrated Acid-Base Biodiesel Synthesis. – Currently, an integrated process comprising the acid-catalyzed pre-esterification of FFAs followed by the base-catalyzed transesterification of TGs is the preferred way to deal with feedstocks high in FFAs.49 An advantage of the integrated process is that it limits the use of sulfuric acid to only the initial feedstock pretreatment where acid-catalyzed esterification is more effective. The acid-catalyzed pre-esterification reduces the concentration of FFAs to levels amenable for the basecatalyzed transesterification process. Canakci and Van Gerpen recently demonstrated the application of a 2-step sulfuric acid-catalyzed pre-esterification process to eliminate FFAs in used cooking oil (yellow grease).56 Their two step procedure reduced FFA levels to below 1 wt%. The refined grease could then be processed using the fast base-catalyzed transesterification reaction to Table 9

Reaction conditions used by Zhang et al. in the acid-catalyzed synthesis of biodiesel from waste cooking oil24 Acid-Catalyzed Biodiesel Synthesis

Feedstock Alcohol-to-oil molar ratio Temperature Pressure Catalyst Catalyst load An oil conversion

Triglyceride mixtures with high free fatty acid contents (44%) e.g., waste cooking oil þ Methanol 50 : 1 801C 4.0 bar H2SO4 1.3 : 1 molar ratio of sulfuric acid to waste oil of 97% is expected after 4h of reaction

Figure 9

Simplified BFD for the acid-catalyzed process proposed by Zhang et al.24

Catalysis, 2006, 19, 41–83 61

62

Table 10

Catalysis, 2006, 19, 41–83

Typical reaction conditions for pre-esterification in a two step acidbase integrated process49 Acid-Base Integrated Process: acid catalyzed pre-esterification

Feedstock

(1) Lipid feedstocks with high free fatty acid contents (4 4%) e.g., Low quality tallow (yellow grease, brown grease) (2) Methanol

First esterification step Yellow grease Brown grease Reaction time Temperature Pressure Catalyst

Alcohol to oil molar ratio 20 : 1 20 : 1

Catalyst concentration 5 wt% 10 wt% 1h 60–701C 1.4–4.0 bar H2SO4

Second esterification step Alcohol to oil molar ratio Catalyst concentration Yellow grease 40 : 1 5 wt% Brown grease 40 : 1 10 wt% Reaction time 1h Temperature 60–701C Pressure 1.4–4.0 bar Catalyst H2SO4 Esterification should reduce free fatty acid levels to o1%

produce biodiesel. Water removal was necessary in-between pre-esterification steps to ensure a high FFA conversion, but other researchers have achieved high FFA conversions in only one pre-esterification step.24,57 Like transesterification, esterification rates and yields are increased by higher alcohol-to-FFA molar ratios, catalyst concentration, temperature, and pressure.56,58,59 Table 10 shows pre-esterification conditions based mainly on methodology disclosed by Canakci and Vangerpen with two consecutive pre-esterification steps.49 Figure 10 shows a BFD of the pre-esterification process, using reaction conditions shown in Table 10. As mentioned earlier, feedstocks high in FFAs, such as waste greases obtained from used cooking oil or low-grade animal fats, can be efficiently treated with the integrated process. When animal fats are used, feedstock pretreatment includes filtration using a cellulose filter to remove particulate matter, such as pieces of meat and bone. This is followed by steam distillation that denatures and degrades residual proteins in the grease and bleaching that removes spoiled proteins.43 After pre-esterification, transesterification is carried out using a typical basecatalyzed process as shown in Figure 7. 3.4 Existing Problems with Homogeneous Catalysts. – The use of homogeneous catalysts involved in biodiesel synthesis currently presents separation and catalyst recovery issues. For instance, process methodologies using homogeneous

Catalysis, 2006, 19, 41–83

63

Figure 10 Simplified BFD for the acid-catalyzed pre-esterification process24,49

alkali catalysts necessitate a multi-step preparation and purification protocol since such catalysts tolerate neither moisture nor FFAs. As previously mentioned, presence of water can constitute a major problem. Ma et al. reported that water and FFA contents of beef tallow had to be maintained below 0.06 wt% and 0.5 wt%, respectively, to carry out the transesterification reaction effectively under alkaline conditions.32 Thus, low water levels appear to be even more critical than FFA levels. On the other hand, most biodiesel production processes, using the alkalicatalyzed approach, have to extensively pre-treat virgin oils in order to achieve the high purity (FFAs o 0.5 wt%, anhydrous) required. For instance, in the US most biodiesel is produced from refined soybean oil, which accounts for 60– 75% of the total cost of biodiesel.2 Even though inedible animal fats, waste cooking oil, and brown grease can sell for significantly less than refined vegetable oils (Table 4), elevated FFA concentrations are a significant issue with these lipid feedstocks. Processing alternatives like the acid-base integrated process allows for the use of low cost, low quality feedstocks, but these technologies still require additional catalyst, processing steps and time, thereby eliminating some of the gains obtained by the use of lower cost feedstocks.

4

Heterogeneous Catalysis in Biodiesel Synthesis

Currently, nearly all biodiesel is produced via homogeneous catalysis. The use of homogeneous catalysts allows carrying out the transesterification of lipid

64

Catalysis, 2006, 19, 41–83

feedstock with small alcohols under mild conditions. However, the advantage of having a production process at low temperatures is most likely counteracted by the additional separation procedures the products require. Residual catalyst must be removed and catalyst loss is inherent, raising the overall cost of production. In contrast, processes using heterogeneous catalysts need higher temperatures to be as effective in a reasonable time. On the other hand, products do not necessitate complex separation procedures and, in most cases, catalysts can be recycled and reused for long periods of time. Many types of solid catalysts have been tested in esterification and transesterification reactions of fatty acids, TG feedstock and simple esters. Nonetheless, it is possible to group most catalysts in three general categories: metal, base, and acid catalysts. The following sections deal with these three groups accordingly. The reader should also be aware that some information provided in this section and in the sections below does not directly address the subject of biodiesel synthesis. Some discussions, for instance, are about the transesterification of simple esters or the production of monoglycerides by transesterification of vegetable oils; nevertheless, the information provided is relevant to the topic of biodiesel synthesis since knowledge of catalyst reactivity in these systems is directly applicable to reactions involving TGs and FFAs. 4.1 Catalysis by Metals, Metal Compounds and Supported Metal Complexes. – Most industrial reactors and high pressure laboratory equipment are built using metal alloys. Some of these same metals have been shown to be effective catalysts for a variety of organic reactions. In an effort to establish the influence of metal surfaces on the transesterification reactions of TGs, Suppes et al.60 collected data on the catalytic activity of two metals (nickel, palladium) and two alloys (cast iron and stainless steel) for the transesterification of soybean oil with methanol. These authors found that the nature of the reactor’s surface does play a role in reaction performance. Even though all metallic materials were tested without pretreatment, they showed substantial activity at conditions normally used to study transesterification reactions with solid catalysts. Nickel and palladium were particularly reactive, with nickel showing the highest activity. The authors concluded that academic studies on transesterification reactions must be conducted with reactor vessels where there is no metallic surface exposed. Otherwise, results about catalyst reactivity could be misleading. Many catalysts do not use metals in their pure reduced metallic forms. Anchored organometallic complexes are often analogs of homogenous catalysts fixed on a solid support. In particular, titanate complexes both in solution and in supported form have been found to be especially active in transesterifications of simple esters.61,62,63 It was proposed that titanates catalyze the transesterification reaction through a Lewis acid mechanism where the reactant ester and metal form a Lewis complex activating the carbonyl groups for a nucleophilic attack by the reactant alcohol. The tetrahedral intermediate that is formed breaks down into the product alcohol and an ester-metal Lewis

Catalysis, 2006, 19, 41–83

65

complex. The product ester desorbs from the Lewis site and the cycle is repeated (Figure 11).61 The reaction is first order in both reactants (ester and alcohol). In general, titanate species are very sensitive to moisture and the reaction requires all reagents to be anhydrous. Grafted titanates on silica are deactivated by water, but unlike homogeneous catalysts, these titanates are isolated on the support’s surface. This situation precludes them from forming stable titanium oxide species, which do not show nearly the same catalytic activity. Supported titanate complexes can be regenerated in isopropanol under reflux, reversing the hydrolysis reaction of the titanate anchored species. It was reported that catalytic activity could be completely restored by this treatment.61 The use of supported metal complexes in transesterification reactions of TGs is not new. An earlier patent claimed that supported metals in a hydroxylated solid could effectively catalyze transesterification.63 The catalyst preparation used an inert hydrocarbon solvent to attach transition metal alkoxide species to the support surface. The reaction, however, was carried out in the presence of water. The author claimed that water was essential in preparing materials with good catalytic activity. Among the metals employed, titanium catalysts showed the best activity. However, it was not clear from the preparation method if reproducibility could be easily achieved, an important requirement if such catalysts were to be commercially exploited. There are other metal complexes active in the transesterification of TGs that could be anchored to a catalyst support to yield an effective heterogeneous catalyst. For instance, Abreu et al. recently published work highlighting the use of metal complexes of Sn21, Zn21 and Pb21 ions as homogeneous catalysts in the transesterification reaction of various vegetable oils with alcohols.64 The catalytic performance of the complexes was compared to sulfuric acid at 601C. The reaction molar ratio was alcohol:oil:catalyst 400 : 100 : 1, respectively. In all cases the metal complexes outperformed sulfuric acid. In another example, a patent was disclosed that conveys the use of organotitanates assisted by zinc acetyl acetonate as catalysts in transesterification reactions of TG mixtures. The reaction conditions used were high temperatures (42001C) and high pressures. The process required the alcohol to be dried and the oil to be degummed and dried. These feedstocks specifications made this procedure similar to that for the homogeneous base-catalyzed process.65 In a different patent, metal compound catalysts made of a mixture of calcium and barium acetates (3 : 1 by weight) were used in the transesterification of oils and fats with high FFA content without a pre-esterification step.66 Again, high temperatures were required (42001C) and the oil as well as the alcohol had to be anhydrous. Organotin catalysts have been tested as well with good results for the transesterification of oils. Unfortunately, these compounds are very soluble in fatty acid esters, complicating their separation and adding to environmental concerns.67 Even though many of the catalysts mentioned above have shown good homogeneous phase catalytic activity, they also exhibit high toxicity, requiring highly efficient catalyst removal processes. In order for this type of catalysts to

Figure 11 Probable reaction mechanism for the transesterification of esters catalyzed by a supported titanate catalyst on silica (based on Blandy et al.)61

66 Catalysis, 2006, 19, 41–83

Catalysis, 2006, 19, 41–83

67

achieve greater use, they must be able to stay permanently anchored in/on a suitable solid support without compromising its catalytic activity. In addition, straightforward methodologies for their easy reactivation must be developed. A common problem with supported organometallic complexes is that their reactivity often does not mirror that of the species in solution. In general, these catalysts show lesser activity once they become fixed on a solid surface. In addition, heterogenized complexes tend to leach, hampering their reusability and releasing toxic material into the final product. 4.2 Catalysis by Solid Bases. – Classic heterogeneous base catalysts, where the solid contains either Lewis or Brønsted base sites, have been the most extensively tested solid catalysts for the transesterification reactions of TGs. For instance, zeolite X and microporous titanosilicate ETS-10 in their as-prepared forms and ion-exchanged with K1 and Cs1 have been examined for the transesterification of soybean oil with methanol and their catalytic activities compared.60 Non-thermally treated zeolites showed no activity, meaning that the basic sites in zeolites were probably poisoned from exposure to CO2 and moisture in the air during handling.68,69 For the K1-exchanged X zeolite, catalytic activity was higher than the Cs1-exchanged and Na1 forms, in that order. However, the same trend was not observed for ETS-10. The parent ETS10 material, which contains Na1 and K1 ions in approximately a 3 : 1 ratio, was the most active catalyst followed by the K1- and Cs1-exchanged materials. In all cases, the ETS-10 titanosilicate showed much higher activity than the X zeolite. The superior performance of ETS-10 was expected since it is known that ETS-10 is about four times more basic than NaX.70 Suppes et al. have noted that high catalytic activities for ETS-10 were obtained even though the soybean oil used in their experiments had a 2.6 wt% FFA concentration.70 However, when reactions were carried out with ETS-10 at 1001C in soybean oil with an excess of FFA (27 wt% oleic acid), poor conversions were observed. Their conclusion was that catalysts relying on strong basic sites were particularly inhibited by the presence of high concentrations of FFAs. Nevertheless, it was puzzling that low concentrations of FFA (2.6 wt%) did not affect the reaction in a similar way.60 In a previous patent, Bayense et al. reported the use of ETS-10 and ETS-4 for the transesterification of TGs with alcohols.71 In their studies, both batch and fixed bed reactors were used to conduct the reaction. Under batch conditions, reactions were carried out at 2201C, 21 bar, and a catalyst loading of 0.23 wt% based on total autoclave content. Experiments with soybean oil and methanol, at a methanol/oil ratio of 4.2, resulted in 69.0% total oil conversion with an ester yield of 52.6% for ETS-10 and 96.9% conversion with an ester yield of 85.7% for ETS-4. When tallow was used under the same conditions, conversion and ester yield were 30.6% and 19.1% for ETS-10 and 44.1% and 29.6% for ETS-4, respectively. The slightly better activity shown by ETS-4 might imply that only basic sites at the pore opening of these zeolites were catalyzing the reaction. The effective radius of the TG molecule is such that it can not enter through the small pore openings of ETS-4 (diameter of 3.7 A˚) or even ETS-10

68

Catalysis, 2006, 19, 41–83

(diameter of about 8 A˚).72 Methanol, however, can easily enter the pore structure of ETS-10 and possibly as well enter the pores of ETS-4 given its small size and the high temperature and pressure used.73 If methanol was moving freely in and out the pore structure of ETS-4, then a more likely explanation for the high activity of these two titanosilicates (and especially the high activity shown by ETS-4 with much smaller micropores) is the occurrence of a homogeneous-like reaction mechanism where the zeolite acts as a common Brønsted-type base. In such a mechanism, the first step involves the abstraction of a proton from the alcohol, similar to the homogeneously catalyzed reaction (Figures 12 and 13). Thus, the free alkoxide ions could reach the TG molecules outside the zeolite pores by leaching/diffusing out of the pores in the form of MeOM1 (M1 ¼ Na1, K1). This might also explain the detrimental effect of FFAs in the reaction media, as reported by Suppes et al.60 However, neither of the two previously cited works using ETS-zeolites have reported formation of soap, which would be expected if metal methoxide species were leaching out and reacting with FFAs. Small amounts of soap, on the other hand, would have not been easy to detect and could have been ignored. Alkaline earth metal oxides and hydroxides have also been tested in transesterification reactions. Ca(OH)2 did not show significant catalytic activity in the transesterification of rapeseed oil with methanol at conditions normally used to prepare biodiesel.74 Peterson et al. reported relative alcoholysis activities of a series of supported CaO catalysts under near reflux conditions of methanol-rapeseed oil mixtures at 6 : 1 molar ratios.75 Among the catalysts tested, the most active was CaO (9.2 wt% CaO) on MgO. For instance, in a 12 h reaction the total oil conversion using this catalyst was over 95%, similar to

Figure 12 Scheme for a homogeneous catalyst-like mechanism for strong Brønsted-type solid bases such as alkali ETS-10

Catalysis, 2006, 19, 41–83

69

the total conversion obtained with a homogeneous NaOCH3 catalyst. Unfortunately, this observation was based on thin layer chromatography results and was only valid in qualitative terms. In addition, the CaO/MgO catalyst produced great quantities of soap, an observation the authors did not properly address even though it was claimed that the reactions were run under anhydrous, FFA-free conditions. In any case, soap formation was a sign that most of the supported CaO was leaching out of the catalyst permitting homogeneous catalysis to take place. Other CaO catalysts, except CaO powder, showed very low or no activity under the reaction conditions reported. Even though CaO is considered a heterogeneous catalyst due to its insolubility in alcohols at room temperature, it is not clear if at moderate-to-high temperatures CaO might show some degree of solubility in methanol.74 This would mean that homogeneous rather than heterogeneous catalysis could be occurring when using CaO under more strenuous conditions. Indeed, CaO is soluble in glycerin. Thus, even if the catalytic process starts as purely heterogeneous, once small amounts of glycerol are present, it is possible that a homogeneous catalyzed reaction could also contribute to the catalytic activity. The issue of homogeneous versus heterogeneous catalysis needs to be more precisely addressed, especially when dealing with solids that may show some solubility under the high temperature-high pressure conditions often applied in systems using heterogeneous catalysts. Some authors have reported very low or no activity for MgO catalysts.74,75 This is unexpected given the strong base character associated with this solid.76 Strong base sites in solids are especially susceptible to catalyst pretreatment. Thus, failing to calcine the catalyst at high temperatures will not free the base sites from pre-absorbed CO2, affecting its catalytic activity. After pretreatment, special care must be taken to avoid extended exposure of the catalyst to air, since this will probably poison the strongest base sites with CO2. For instance, Leclercq et al. observed a strong dependence of MgO activity with calcination temperature. When MgO was calcined at 4501C in either air or nitrogen and used in the methanolysis of rapeseed oil at 651C (refluxing methanol, methanol-to-oil molar ratio of 75), the reaction took 22 h to reach a 65–69% total oil conversion. But, when MgO was calcined at 5501C, it only took the reaction 1 h to reach half the same conversion value.77 Hence, more activation of the strong base sites in MgO occurred at high temperatures. Accordingly, the alcoholysis reaction was dependent on the strength and number of the base sites involved in the catalysis. Similar findings have been reported by other authors. For example, Corma et al. compared the activity of Cs-MCM-41, Cs-Sepiolite, MgO and hydrotalcite (HT) in the transesterification of triolein with glycerol.78 These catalysts covered a wide range of strengths of base sites. The order in catalyst reactivity was MgO 4 HT 4 Sepiolite-Cs 4 Cs-MCM-41, in agreement with the base strength of the series. However, only the most basic catalysts (MgO and HT) exhibited high reaction rates and conversion levels at temperatures above 2001C. Similarly, Bancquart et al. used MgO, ZnO, La2O3 and CeO2 for the transesterification reaction of methyl stereate and glycerol. The reaction rate showed direct dependency with the base strength of each catalyst. Catalysts

70

Catalysis, 2006, 19, 41–83

with the strongest base sites showed the highest intrinsic activity.79 It was also noted that some base catalysts exhibited significant acid character (La2O3, CeO2 and ZnO, in that order). However, the impact of the catalyst bifunctionality on its reactivity was not clear. It was reported, though, that the acid character of a base catalyst such as La2O3 could have affected its selectivity. Acid sites may have been involved in important side reactions, such as the double dehydration of glycerol to produce acrolein. Acrolein has been reported as a possible important side product in reactions involving the condensation of GL molecules to produce GL ethers with solid base catalysts.80 Simple carbonate salts have also been used with success as base catalysts for transesterification reactions. For instance, the alcoholysis reaction of soybean oil and beef tallow with ethylene-, diethylene-, triethylene-glycol and glycerin was carried out using M2CO3 (M ¼ K, Na) and MCO3 (M ¼ Mg, Ca, Zn) base catalysts at temperatures above 2001C and alcohol/TG ratios 48. High triglyceride conversions (total conversion 4 95%) were achieved under these conditions in less than three hours.67 Sodium and potassium carbonates, however, catalyzed significant hydrolysis side reactions, which lowered their efficiency considerably. Mg, Ca and Zn carbonates produced a clean product. Further studies with packed-bed flow reactors were conducted with calcium carbonate because of its low cost, availability, and low solubility in fats and oils. In packed-bed flow reactors, catalyst activity was improved. This type of reactor allowed for higher catalyst loadings and improved catalyst-reagent contact. Conversions over 95% were achieved at temperatures Z 2001C and residence times of 18 min. However, experiments involving the same conditions but using ethanol instead resulted in very poor catalyst performance. Only when the temperature was raised to 2601C and a fraction of the fatty acid alkyl esters (biodiesel) was introduced as co-solvent were high conversions possible for the soybean/ethanol system. At temperatures Z 3001C, oil polymerization occurred, plugging the reactor. Although it is claimed in a recent patent application that carbonate catalysts can be used efficiently in the production of biodiesel from TG mixtures with high FFA contents,81 the fact that esterification of FFA produces water raises doubts about what kind of catalysis –homogeneous or heterogeneous- might really predominate. Such systems would have appreciable carbonate solubility. Carbonate species have shown leaching phenomena under mild conditions. Peterson et al. supported K2CO3 on MgO and Al2O3 with good results for the transesterification of rapeseed oil with methanol at 60–631C.75 However, the K2CO3 in K2CO3/Al2O3 leached into solution, and, most likely, the same occurred with K2CO3/MgO since the observed catalytic activities in both cases were very similar. Most probably, homogeneous catalysis was involved. Another catalyst which may also involve a homogeneous catalysis mechanism is barium hydroxide. Barium hydroxide has been reported to catalyze the methanolysis of rapeseed oil at 651C with over 80% of oil conversion in less than an hour.74,77 However, Ba(OH)2 shows non-negligible solubility in water, methanol and polyols.74,82 For instance, in alcohols Ba(OH)2 can form barium alcoholates (RO–Ba–OH and RO–Ba–OR). A barium alcoholate of formula

Catalysis, 2006, 19, 41–83

71

Ba(OR)2 (linear) is expected to have good solubility in oils. This may increase the solubility of alkoxide species in the oil, allowing close range contact between the catalytically active alkoxide species and the glycerides, resulting in a faster reaction rate. Indeed, when one compares Ba(OH)2 to well known homogeneous catalysts and the most active heterogeneous catalysts, the observed activity for Ba(OH)2 is closer to the catalytic activity shown by its homogeneous catalyst counterparts.60 Thus, most probably the catalytic activity shown by Ba(OH)2 is due to dissolved Ba-alkoxide species rather than basic sites at the surface of the Ba(OH)2 solid. Supports, such as MCM-41, have been used to increase the surface area and availability of base sites in catalysts. Furthermore, in reactions targeting the production of monoglycerides, selectivity towards this intermediate can be significantly enhanced. This is attributed to the shape selectivity effect resulting from the uniform pore size of the MCM-41 material. In particular, Barrault et al. prepared MgO/MCM-41 and MgO/AlMCM-41 catalysts for the transesterification reaction of methyl esters and glycerol.83 Here, glycerol adsorbed and accumulated inside the pores of (Al)MCM-41 supports, inhibiting the transesterification reaction. When pure silica MCM-41 was tried with pre-adsorption of the ester reagent, reactions showed improved activity and selectivity towards the monoglyceride with up to 80% monoglyceride yields, twice the amount reported using a homogeneous catalyst. The use of another supported base catalyst was disclosed in a recent patent. A ZnO/Al2O3 catalyst was used in the production of alkyl esters from the alcoholysis of oils. Reactions were carried out at high temperatures (above 2001C) and pressures in batch and continuous flow packed-bed reactors. High conversions were observed (over 80% total oil conversion) after 2h of reaction.84 Unfortunately, it is not clear up to what degree the ZnO/Al2O3 solid was responsible for the actual catalysis since the metallic surface of the reactor used was most probably contributing as well. For instance, in one case an ester yield of 91% was obtained in the presence of catalyst, while in the absence of catalysts under the same reaction conditions the yield was 60%. Alkaline metal oxide species confined in larger pore zeolites constitute another promising type of base catalysts. Zeolites can be modified by the decomposition of occluded alkali metal salts, forming intrazeolitic alkali metal oxides with very active basic sites.69 When NaX zeolites were impregnated with sodium acetate or sodium azide and subsequently calcined, confined sodium oxides species were obtained. Materials produced using sodium azide showed stronger basic sites than those obtained from sodium acetate. A distribution of base sites with different base strengths was produced with a population proportional to the amount of intrazeolitic sodium oxide. The use of NaOx/Nax zeolites in the transesterification of soybean oil with methanol (6 : 1 alcohol to oil ratio) at 1201C for 24 h resulted in 57–94% oil conversions.60 However, some catalyst leaching was also observed. In addition to transesterification reactions, solid base catalysts, including both simple oxides and zeolitic materials, have been used to carry out esterification reactions of fatty acids. These studies have mainly focused on the

72

Catalysis, 2006, 19, 41–83

production of monoglycerides from fatty acids and GL. Basic oxides, such as MgO and ZnO, were used in esterification with different organic solvents.85 The best results, in terms of selectivity, were obtained with ZnO and a diglyme solvent. Corma et al. also obtained high conversions (480%) and high monoester selectivities (480%) for the named reaction, using the basic forms of large pore zeolites.86,87 The good selectivity to monoglycerides was likely due to the shape selectivity or steric effect exerted by the zeolite pore structure on transition state configurations and intermediates formed in the interior of the zeolite. 4.3 Catalysis by Solid Acids. – Two aspects are considered here. The first aspect is concerned with transesterification reactions catalyzed by solid acids. Unfortunately, little research dealing with this subject has been reported in the literature. The second aspect deals with esterification reactions of carboxylic acids (or FFAs). This second part addresses an important characteristic of inexpensive TG feedstocks, i.e., high FFA content. Ideally, an active solid catalyst should be able to carry out transesterification and esterification simultaneously, thus eliminating pretreatment steps. It is likely that heterogeneous catalysts that perform well in esterification should also be good candidates for transesterification since the mechanisms for both reactions are quite similar. 4.3.1 Transesterification Reactions. The heterogeneous acid-catalyzed transesterification of TGs has not been investigated as much as its counterpart, the base-catalyzed reaction. Various solids are available with sufficient acid strength to be effective catalysts for the named reaction. Among the solid acids available are functionalized polymers, such as the acid forms of some resins, as well as inorganic materials, such as zeolites, modified oxides, clays, and others. Some of these solids have already been found to be effective in transesterification reactions of simple esters and b-ketoesters. In general, the application of solid acid catalysts to produce biodiesel from oils and fats has been largely ignored. The possibility of unwanted side reactions has been in part blamed for this fact.87 In particular, the double dehydration of the by-product glycerol to produce acrolein and water (reaction catalyzed by acids) has been a main concern. However, this reaction becomes important only at temperatures above 2501C.88 Actually, acrolein has been cited as one by-product in the etherification of glycerol to polyglycerol, using solid base catalysts.79 Acid sites on those catalysts were held responsible for the GL dehydration reaction. But, acid sites on base materials are not strong and, given that strong acids and high temperatures are usually required for the GL double dehydration, it is likely that a simple acid-catalyzed process was not by itself the cause for the formation of acrolein in those studies. Other studies dealing with the esterification of lauric acid with GL, using the acid form of zeolite b at 1001C, have reported GL ethers but no acrolein as GL-derived products.89 Thus, at temperatures below 2001C production of acrolein should

Catalysis, 2006, 19, 41–83

73

not be an important deterrent for the use of solid acid catalysts in biodiesel synthesis. Perhaps, a more important reason for the little research in this particular area is the slow reaction rate associated with acid catalysis in general. However, the ability of solid acids to catalyze both esterification and transesterification reactions simultaneously and the possibility for employing catalysts that are reusable and green, meaning that they do not pose a great environmental threat, are attractive aspects that make the study of these materials imperative. To the best of our knowledge, only a few publications have directly addressed the use of solid acid catalysts to prepare alkyl esters from alcohol and triglyceride mixtures. Mittelbach et al. tested a series of layered silicates in alcoholysis reactions of TGs.90 Reactions took place at 1501C with alcohols such as methanol, ethanol, 1-propanol and 2-propanol. Clays like montmorillonite were tested with and without acid activation, which was done by submerging the clays in a solution of sulfuric acid and methanol followed by methanol/water washing and drying at 701C. Montmorillonite KSF showed the highest activity for the alcoholysis of refined rapeseed oil with methanol. High conversions were obtained at high temperatures and high pressures using a 5 wt% catalyst loading in the reaction mixture. For instance, at 2201C and 52 bar an ester yield of almost 100% was obtained after 6 h. However, at this temperature, dimerization of alky esters was observed and the formation of GL ethers was also noted. At lower temperatures and pressures (1401C and 8.5 bar), side-reactions were not as important, but ester yields were affected (70% yield after 8 h). Compared to sulfuric acid, the clay catalysts produced a cleaner biodiesel due to their bleaching activity. Thus, unrefined oils or waste cooking oils could be employed as feedstock without pretreatment. However, the performance of the clays diminished with repeated use and catalysts had to be reactivated after each run to maintain peak performance, suggesting that some leaching of sulfuric acid took place. Furuta et al. tested a series of strong solid acids (alumina promoted sulfated zirconia, alumina promoted tungstated zirconia and sulfated tin oxide) for the transesterification of soybean oil with methanol at 200–3001C.91 Reaction yields over 90% were obtained for the alumina promoted tungstated zirconia at reaction times of 20 h using a flow reactor (T ¼ 2501C). The activity of the same catalyst was maintained for up to 100 h. In the past, hydrous oxides have been proposed as viable alternatives to replace homogeneous acid catalysts due to their good stability and insensitivity to common reaction pollutants, such as CO2 and water. Transesterification of simple esters, such as ethyl acetate, with alcohols using a hydrous SnO2 catalyst was able to be carried out under mild reaction conditions at reflux temperature with 10-fold excess ethyl acetate.92 Yields depended on the chemical characteristics of the alcohol. For instance, reagent mixtures with cyclohexanol were unreactive, and poor yields were observed with long alkyl chain linear alcohols. This suggests that the hydrophilic character of the hydrous SnO2 surface has an impact on catalyst performance.

74

Catalysis, 2006, 19, 41–83

Figure 13 Ketoester transesterification

The majority of publications on transesterification catalyzed by solid acids have focused on the reaction of b-keto esters. Transesterification provides an alternative route to synthesize these kinds of esters since direct preparation from b-keto acids is not a good option given that they can easily undergo decarboxylation. Table 11 provides a review of results in the literature relating to the transesterification of b-keto esters with alcohols. Several conclusions can be drawn from the information given in Table 11 and the references therein.93–97 For instance, zeolite activity showed some particular trends: larger pore zeolites and dealuminated zeolites showed higher yields, suggesting that catalytic activity depends on reagent intra-pore diffusion, acid strength and surface hydrophobicity. Simple esters and other types of esters (a-keto ester, a-halogenated esters and a,b-unsaturated esters) exhibited less transesterification activity than b-keto esters under the same reaction conditions. Sulfated tin oxide (STO) and amberlyst-15 showed dissimilar behavior when the alkyl chain in the alcohol was lengthened. Alcohols with longer alkyl chains showed a lower reaction yield on STO than on amberlyst15. It is likely that in these catalysts alcohol molecules cluster around acidic protons, creating modified active sites.98 Reaction, most probably, took place on these modified sites. Given that the swelled amberlyst-15 has larger pores than STO, the long alkyl chain alcohols probably made it more difficult for the Table 11

Transesterification reactions of b-keto esters with alcohols. Reactions were carried out with ethyl acetoacetate at 1101C with toluene as the solvent and continuous azeotropic removal of alcohola

Catalyst

Alcohol

Catalyst wt%b Reaction time Yield (%) Reference

Hb Hb Dealuminated Hb HY H-mordenite H-ZSM-5 H-ZSM-12 H-MCM-41 Kaolinitic clay Kaolinitic clay Sulfated tin oxide Sulfated tin oxide Amberlist-15 Amberlist-16

C12H25OH N-hexanol N-hexanol N-hexanol N-hexanol N-hexanol N-hexanol N-hexanol methanol N-pentanol N-butanol N-octanol N-butanol N-octanol

10 20 20 20 20 20 20 20 15 15 10 10 10 10

8 10 10 10 10 10 10 10 2 2 6 7 5.5 2

86 80 92 85 86 28 30 25 92 89 97 89 85 89

93 94 94 94 94 94 94 94 95 95 96 96 97 97

a All cited articles tested different b-keto esters with alcohols. Representative experiments were chosen and their results condense in the table. b By weight of ethyl acetoacetate.

Catalysis, 2006, 19, 41–83

75

ketoester molecules to approach the acid sites on STO than on amberlyst-15. It is possible that for STO only a few accessible catalytic sites were able to actually carry out the catalysis. 4.3.2 Esterification Reactions. The use of solid acids provides a practical substitute for homogeneous acid catalysts commonly employed to prepare alkyl esters. The use of homogeneous acid catalysts, such as sulfuric acid, and p-toluene- or methane-sulfonic acids, generally results in sulfur contamination of the final product, which upon combustion yield SOx compounds that are known pollutants. Many publications advocate the use of solid acid resins in esterification reactions. A comprehensive review of organic reactions catalyzed by resins is that of Harmer and Sun.99 However, thermal-stability above 1401C and lack of structural integrity at high pressures severely limit the applicability of organic resins as catalysts for esterification and transesterification. As expected, zeolites, given their variety in pore size and acid strength, are logical substitutes for homogeneous acid catalysts. Zeolites have been used in the production of monoglycerides from fatty acids and GL. Using zeolites for this reaction at temperatures up to 1801C and under reduced pressure (0.07 bar in order to remove the by-product water), fatty acid conversions between 66 and 92% have been achieved, reaching monoglyceride selectivities over 80%.86 Other studies conducted on the same type reaction with zeolite b at lower temperatures (1001C) showed lower yields.89 Catalyst activity increased with the Si/Al ratio, indicating that the reaction is influenced by acid site strength as well as by surface hydrophobicity. Likewise, strong solid acids like sulfated zirconia (SZ) and similar materials have been cited as excellent alternatives to liquid acids.100 For instance, SZ has been used in the preparation of 2-ethylhexyl dodecanoate from lauric acid and 2-ethyl hexanol. The reaction was carried out in a reactive distillation system, which allowed continuous separation of the water co-product when the system was operated at temperatures above 1001C.101 Water removal is essential for reaction progress not only because it drives the reaction to high product yields but also because, if water is allowed in the reaction media, formation of an aqueous phase (accompanied by phase segregation) could promote hydrolysis of the supported sulfate groups, causing leaching of the active acid sites and deactivating the catalyst. Trace amounts of water, however, were found to have no effect on catalyst deactivation. Other authors have also noted the effect of water on SZ catalysts. Corma et al. exposed calcined SZ to water vapor and followed the interaction of water and bonded SO3 groups using infrared (IR) spectroscopy.102 In particular, these authors noted that SO3 functional groups were very sensitive to water vapor. For instance, at a molar ratio of water-to-S of about 1.4, the SO3 moieties on the surface gave rise to sulfate species (SO42, HSO4 and H2SO4). In a liquid phase system, sulfate species in this form can easily be leached out from the catalyst. It has been recently claimed that a new SZ catalyst (UDCaT-5) with high sulfur content (9%) (prepared using chlorosulfonic acid) is resistant to the

76

Catalysis, 2006, 19, 41–83

presence of water. The new SZ was tested for sulfate lost, and no leaching was observed when the SZ solid was put in direct contact with water.103 A 76% conversion was observed when this catalyst was used in the esterification of acetic acid with p-tert-butylcyclohexanol at 901C and a 3 : 1 acetic-to-alcohol molar ratio (4 h). Catalyst activity was better than conventional SZ (conversion ¼ 53%) and no loss of activity was seen even after the same catalyst was used for three consecutive reaction experiments. Ardizzone et al. used the esterification of benzoic acid with methanol to test the catalytic performance of different SZ catalysts.104 Water had to be continuously removed from the reaction medium to shift the reaction equilibrium to product formation and to avoid catalyst deactivation by sulfate leaching. According to these authors, catalysts with a higher density of acid sites with pKa values in the range 14.2 to 5.6 performed better. Acid sites with pKa of 5.6 seemed to play an important role in intermediate adsorption. Catalyst reusability, however, was compromised. After reaction, both the type and population of acid sites changed. For instance, acid sites with pKa of 5.6 disappeared and the total acid site population decreased by ca. 30%. This is likely an indication that either sulfate leaching occurred or methanol-induced dehydration of the catalyst caused deactivation of the acid sites. Sulfated tin oxide (STO) is classified as one of the strongest solid acids (STO calcined at 5501C ranks first among sulfated metal oxides according to the Hammett function scale, H0 value ¼ 18).105 However, the use of STO has been more limited than that of SZ (calcined at 6501C, H0 value ¼ 16.1) due to preparation difficulties and poor yields. However, new preparation routes are making this catalyst more accessible, and recently its use has become more widespread.106,107 In a recent study by Furuta et al., STO was compared to SZ in the esterification of n-octanoic acid with methanol.107 The STO catalyst showed superior activity compared to SZ at temperatures below 1501C. For instance, STO approached a 100% ester yield at 1001C, while SZ required temperatures as high as 1501C to reached similar yields. In general, the solid acid-catalyzed esterification of alkyl acids with alcohols exhibits overall second order behavior, similar to the homogeneous acidcatalyzed reaction. Guner et al. conducted kinetic studies on the esterification of oleic acid with glycerol using sulfated iron oxide.108 The kinetic data could be effectively modeled as a second order reaction. In the gas phase esterification of acetic acid with ethanol over MCM-41, acetic acid was first protonated on the surface acid sites followed by reaction with ethanol in a Langmuir-Hinshelwood fashion.109 This reaction was also found to be second order. Similarly, in the esterification of lauric acid with 2-ethyhexanol on SZ, the reaction was found to be first order in the alcohol, the acid, and the catalyst.101 At high temperatures, the auto-catalyzed esterification (which is second order in the acid and first order in the alcohol) could not be ignored and had to be included in the calculations to find rate constants and reaction orders. Functionalized mesoporous materials with organosulfonic groups have also been used in the esterification of fatty acids with alcohols.110,111 The increased hydrophobic character of the catalyst surface resulted in high catalytic activity

Catalysis, 2006, 19, 41–83

77

and selectively toward monoglycerides in the esterification of FFAs with GL.110 Likewise, the esterification of FFAs in model mixtures of palmitic acid and soybean oil with methanol using various mesoporous supports for the sulfonic groups showed that the more hydrophilic HMS molecular sieve had the least catalytic activity.111 In addition, activated diffusion played an important role in catalyst activity. Catalyst activity increased with the support’s pore diameter and this effect was particularly pronounced for medium pore diameter (22–35 A˚) materials. For smaller pore materials, the accumulation of methanol and water around the hydrophilic acid sites may have hindered access of FFA molecules to them giving rise to a diffusion controlled reaction. Acid site strength was also an important factor influencing catalyst activity. For instance, organosulfonic groups with a phenyl moiety attached to the SO3H functional group instead of just an aliphatic chain, showed greater acidity and produced a more efficient catalyst with an esterification activity similar to sulfuric acid (reaction temperature ¼ 851C, conversion 490% at 1.5 h). However, the activity comparison was made for two catalysts with different pore diameters; hence, the effect of the acid strength on catalyst activity could have been overestimated due to activated diffusion. Nevertheless, characteristics like hydrophobicity, pore diameter, and acid strength, are directly correlated with catalyst activity and should be carefully considered in solid acid catalyst applications. 4.4 Potential Problems with Heterogeneous Catalysts. – Most research concerning the application of heterogeneous catalysts for biodiesel synthesis has focused on solid base catalysts and, although some of these catalysts show good promise, they do so at the expense of high temperatures and high pressures. High temperatures and, more importantly, high pressures translate into high equipment costs and hazardous working conditions. Pressure, in particular, is an important factor. At moderate-to-high temperatures, the vapor pressure of methanol and other low molecular weight alcohols substantially increases. For instance, in going from 651C to 1201C, the methanol vapor pressure increases from 1 bar to about 6.5 bars; but when temperature is increased to 1701C the pressure reaches almost 24 bars. However, if a purely liquid phase reaction were replace by a gas-liquid reaction phase (plus the solid catalysts), as is utilized for many other reactions, lower pressures could be used. Operating at temperatures around 120–1501C at pressures of 1–2 bars could allow for the continuous separation of unreacted methanol and water (if doing esterification and transesterication simultaneously), which would provide for a simple yet effective reaction strategy that could compensate for the additional cost of carrying out reactions at higher temperatures. An important issue concerning the use of heterogeneous catalysts for biodiesel synthesis is the lack of systematic research exploring the principles of solid catalyst activity for transesterification of TGs and esterification of FFAs with alcohols. For instance, the question about the true catalytic nature of some solid bases, called heterogeneous, remains unanswered. For example, the most active heterogeneous catalysts reported to date is Ba(OH)2. However, due to its

78

Catalysis, 2006, 19, 41–83

non-negligible solubility in alcohols like methanol, this catalyst most probably is operating through a homogenous reaction mechanism rather than a heterogeneous one. The same might be occurring with other solid bases that required higher temperatures to show measurable catalytic activity. Other active base catalysts like ETS-10 and ETS-4 with no solubility in alcohols, most likely carry out catalysis using a homogenous-like type mechanism, which would mean that their catalytic effect can be lost after the formation and leaching out of the catalytically active species MeO(Na1 or K1). Indeed, systematic research tackling these questions has to be conducted in order to more precisely assert the true catalytic nature of some of these active solid bases. Even though less research has been carry out dealing with the use of solid acids for biodiesel production, esterification reaction studies of carboxylic acids (FFAs) with alcohols has been investigated (keep in mind, esterification shares the same reactivity principles as transesterification when using acid catalysts). These studies suggest that the main problems for solid acid catalysts concern the diffusivity of large TG (and glyceride species in general) molecules through pores and cavities of solid materials and the deleterious effect that water (and some polar compounds) can inflict on acid sites. Water, for instance, can adsorb and cluster around acid sites (through the formation of strong hydrogen bonds), isolating and lowering the acid strength of these sites. Methanol and glycerol (once this is formed) can also have a similar effect. In addition, water can promote degradation and leaching of acid sites in sulfated based catalysts. One important problem with acid catalysts in general is their lower reaction rates compared to base catalysts. Given that solid acids have an even smaller population of acid sites per gram of catalysts as compared to a homogeneous acid catalyst like sulfuric acid, even lower rates are to be expected for solid acids on the same weight basis even though their rate per site might be equivalent. Thus, most likely, solid acid catalysts would need high temperatures, high catalysts loading, and/or larger reactors to obtain reasonable biodiesel production rates.

5

Conclusions and Future Perspectives

Biodiesel represents an increasingly important alternative to conventional petroleum-based fuels. However, biodiesel will not be able to completely replace petroleum-based diesel given the current worldwide demand for this energy source.42 Most likely, biodiesel will find widespread use in the form of blends with conventional diesel as is currently occurring in some countries in Europe.112,113 In the US, Minnesota is leading the way, and starting in June 2005, Minnesota has required that all diesel fuel contain at least 2% biodiesel, opening the door for a more favorable biodiesel market.114 Current biodiesel technologies are in many aspects simple, operator friendly and, within their boundaries, efficient. Homogeneous base catalysts, such as NaOH and MeONa (most commonly used in biodiesel synthesis), are fast and produce high biodiesel yields in relatively short reaction times. However, only

Catalysis, 2006, 19, 41–83

79

refined lipid feedstocks (free from FFAs) and anhydrous alcohols can be used with these catalysts, resulting in high biodiesel costs. In addition, further neutralization and washing processing steps are required to eliminate catalysts residues, which also increase biodiesel production cost. Although processing alternatives like the acid-base integrated process allows for the use of low cost, low quality feedstocks, these technologies require even more catalyst and processing steps, thereby eliminating some of the gains obtained by their application. To lower biodiesel costs and to enhance processing efficiency, current homogeneous catalysts could be replaced by solid heterogeneous catalysts that can be used along with continuous processing technologies (to replace batch processing). The application of heterogeneous catalysts to the production of biodiesel is only in its infancy. Most research on the use of solid catalysts has focused on solid base catalysts. Although some of these catalysts show good promise, they do so at the expense of high temperatures and high pressures. Cost wise, it would be ideal if efficient solid catalysts could work at temperatures below 1501C. Some solid bases, showing good catalytic activity, are most probably catalyzing transesterification reactions through a homogeneous molecular pathway rather than a truly heterogeneous one, due to their non-negligible solubility in alcohols, such as methanol. Also, for zeolitic-type materials like ETS-10 (Na, K), which has shown high catalytic activities for the transesterification of soybean oil with methanol at moderate temperatures, catalysis is probably occurring through a homogeneous-like reaction mechanism that should translate into heavy catalyst deactivation after only a few reaction cycles. On the other hand, solid acid catalysts have been largely ignored for biodiesel synthesis due to pessimistic expectations in terms of reaction rates and undesired side reactions. However, as research concerning the esterification of carboxylic acids with alcohols has shown, solid acids, indeed, should have a place in biodiesel synthesis. For instance, according to results obtained from esterification studies, a good solid acid catalyst for biodiesel synthesis should have large pores, a hydrophobic surface, good acid site strength, and high acid site concentration. An issue that deserves more attention is the future role of glycerol obtained as a by-product from the methanolysis of lipid feedstocks, given the increasing demand for and production of biodiesel. Glycerol accounts for approximately 10 wt% of all products generated from biodiesel synthesis. It has been reported that blending glycerol ethers with biodiesel gives rise to a fuel with improved physical chemical properties (lower cloud- and pour-points).115 The preparation of glycerol ethers involves the reaction of glycerol with isoalkenes, producing di- and tri-tertiary alkyl ethers, using an acid catalyst like amberlyst-15.115–117 Thus, a viable alternative to the low purity glycerin (glycerol) produced during biodiesel processing is the preparation of glycerol-based additives for biodiesel, making good use of the surplus production of glycerol and improving overall production economics. Finally, as indicated in this chapter, many opportunities exist for improving the synthesis of biodiesel. Use of solid catalysts and continuous processing with

80

Catalysis, 2006, 19, 41–83

flow reactors would go a long way in reducing the cost of production. Particularly, work carry out in this area needs to include in the future studies addressing solid catalyst reusability and regeneration in order to estimate the potential of specific catalysts for commercial application. References 1. G. Knothe, R.O. Dunn and M.O. Bagby, in ‘Biodiesel: the use of vegetable oils and their derivatives as alternative diesel fuels.’ USDA, Peoria, IL, 1996, Volume, p. 37. 2. J.A. Kinast and K.S. Tyson, in ‘Production of biodiesel from multiple feedstocks and properties of biodiesel and biodiesel/diesel blends. Final report’, NREL, Golden, CO, 2003, Volume, p. 57. 3. National Biodiesel Board, in ‘Biodiesel Emissions’, Jefferson City, MO, 2004. 4. USDA, in ‘Roadmap for biomass technologies in the United States’, USDA, 2002, p. 34, Volume. 5. H. Fukuda, A. Kondo and H. Noda, J. Biosci. Bioeng., 2001, 92, 405. 6. F.R. Ma and M.A. Hanna, Bioresource Technol., 1999, 70, 1. 7. F. Billaud, A.K.T. Minh, P. Lozano and D. Pioch, J. Anal. Appl. Pyrol, 2001 58, 605. 8. D. Pioch, P. Lozano, M.C. Rasoanantoandro, J. Graille, P. Geneste and A. Guida, Oleaguineux, 1993, 48, 289. 9. G.N. Darocha, D. Brodzki and G. Djegamariadassou, Fuel, 1993, 72, 543. 10. J. Gusmaõ, D. Brodzki, G. Djega-Mariadassou and R. Frety, Catal. Today, 1989, 533–544, 533. 11. J.R.S.G. DosAnjos, Y.L. Lam and R. Frety, Appl. Catal., 1983, 299. 12. F.A.A. Twaiq, A.R. Mohamad and S. Bhatia, Fuel Process. Technol., 2004 85, 1283. 13. F.A. Twaiq, N.A.M. Zabidi, A.R. Mohamed and S. Bhatia, Fuel Process. Technol., 2003, 84, 105. 14. F.A.A. Twaiq, A.R. Mohamad and S. Bhatia, Micropor. Mesopor. Mat., 2003 64, 95. 15. A.W. Schwab, G.J. Dykstra, E. Selke, S.C. Sorenson and E.H. Pryde, J. Am. Oil Chem. Soc., 1988, 65, 1781. 16. E. Lotero, Y. Liu, D.E. Lopez, K. Suwannakarn, D.A. Bruce and J.G.J. Goodwin, Ind. Eng. Chem. Res., 2005, 44, 5353. 17. U. Schuchardt, R. Sercheli and R.M. Vargas, J. Brazil. Chem. Soc., 1998, 9, 199. 18. B. Freedman, R.O. Butterfield and E.H. Pryde, J. Am. Oil Chem. Soc., 1986, 63, 1375. 19. H. Noureddini and D. Zhu, J. Am. Oil Chem. Soc., 1997, 74, 1457. 20. D. Darnoko and M. Cheryan, J. Am. Oil Chem. Soc., 2000, 77, 1263. 21. K.S. Tyson, in ‘Biodiesel handling and use guidelines’, NREL, Golden CO, 2001, Volume, p. 22. 22. R. Dunn, J. Am Oil Chem. Soc., 2002, 79, 915. 23. P. Talley, Render, 2004. 24. Y. Zhang, M.A. Dube, D.D. McLean and M. Kates, Bioresource Technol., 2003, 89, 1. 25. D.G.B. Boocock, in ‘Process for production of fatty acid methyl esters from fatty acid triglycerides’, US, 2004. 26. H. Noureddini, D. Harkey and V. Medikonduru, J. Am. Oil Chem. Soc., 1998 75, 1775.

Catalysis, 2006, 19, 41–83

81

27. V. Jordan and B. Gutsche, Chemosphere, 2001, 43, 99. 28. C.L. Peterson, J.L. Cook, J.C. Thompson and J.S. Taberski, Appl. Eng. Agric., 2002, 18, 5. 29. D. Anderson, D. Masterson, B. McDonald and L. Sullivan, Renewable Energy Management International Palm Oil Conference, Putrajaya, Malaysia, 2003. 30. D.G.B. Boocock, S.K. Konar, V. Mao and H. Sidi, Biomass Bioenerg., 1996 11, 43. 31. B. Freedman and E.H. Pryde, J. Am. Oil Chem. Soc., 1984, 61, 1638. 32. F. Ma, L.D. Clements and M.A. Hanna, T. ASAE, 1998, 41, 1261. 33. G. Vicente, M. Martinez and J. Aracil, Bioresource Technol., 2004, 92, 297. 34. M.J. Nye, T.W. Willianson, S. Deshpande, J.H. Schrader, W.H. Snively, T.P. Yurkewich and C.R. Frech, J. Am. Oil Chem. Soc., 1983, 60, 1598. 35. A.V. Tomasevic and S.S. Siler-Marinkovic, Fuel Process. Technol., 2003, 81, 1. 36. G. Steinke, S. Schonwiese and K.D. Mukherjee, J. Am. Oil Chem. Soc., 2000 77, 367. 37. R. Fillieres, B. Benjellounmlayah and M. Delmas, J. Am. Oil Chem. Soc., 1995, 72, 427. 38. A. Srivastava and R. Prasad, Renew. Sust. Energ. Rev., 2000, 4, 111. 39. F.R. Ma, L.D. Clements and M.A. Hanna, Bioresource Technol., 1999, 69, 289. 40. D.G.B. Boocock, S.K. Konar, V. Mao, C. Lee and S. Buligan, J. Am. Oil Chem. Soc., 1998, 75, 1167. 41. K. Krisnangkura and R. Simamaharnnop, J. Am. Oil Chem. Soc., 1992, 69, 166. 42. M.J. Haas, Lipid Technol, 2004, 16, 7. 43. M. Zappi, R. Hernandez, D. Sparks, J. Horne, M. Brough, S.M. Arora, and W.D. Motsenbocker, in ‘A review of the engineering aspects of the biodiesel industry.’ Mississippi Biomass Counsil, Jackson MS, 2003, Volume, p. 71. 44. J. Connemann and J. Fisher, International Liquid Biofuels Congress, CuritivaParana, Brazil, 1998. 45. E. Crabbe, C. Nolasco-Hipolito, G. Kobayashi, K. Sonomoto and A. Ishizaki, Process Biochem., 2001, 37, 65. 46. M. Canakci and J. Van Gerpen, T. ASAE, 1999, 42, 1203. 47. S. Siler-Marinkovic and A. Tomaseviv, Fuel, 1998, 77, 1389. 48. Y. Zhang, M.A. Dube, D.D. McLean and M. Kates, Bioresource Technol., 2003, 90, 229. 49. M. Canakci and J. Van Gerpen, T. ASAE, 2003, 46, 945. 50. A. Demirbas, Energ. Convers. Manage., 2003, 44, 2093. 51. K.S. Liu, J. Am. Oil Chem. Soc., 1994, 71, 1179. 52. A.K. Khan, ‘Research into biodiesel kinetics & catalyst development’, University of queensland, Brisbane, Queensland, 2002. 53. D. Kusdiana and S. Saka, Bioresource Technol., 2004, 91, 289. 54. R. Sridharan and I.M. Mathai, Sci. Indus. Resource, 1974, 33, 178. 55. F. Rived, I. Canals, E. Bosch and M. Roses, Anal. Chim. Acta, 2001, 439, 315. 56. M. Canakci and J. Van Gerpen, T. ASAE, 2001, 44, 1429. 57. D.G.V. Boocock, in ‘Single-phase process for production of fatty acid methyl esters from mixtures of triglycerides and fatty acids’, USA, 2003. 58. V.O.V. Sonntag, J. Am. Oil Chem. Soc., 1979, 56, 751A–754A. 59. C. Blandy, D. Gervais, J.L. Pellegatta, B. Gilot and R. Guiraud, J. Mol. Catal., 1991, 64, L1. 60. G.J. Suppes, M.A. Dasari, E.J. Doskocil, P.J. Mankidy and M.J. Goff, Appl. Catal. A-Gen., 2004, 257, 213.

82

Catalysis, 2006, 19, 41–83

61. C. Blandy, J.L. Pellegatta and B. Gilot, J. Catal., 1994, 150, 150. 62. H. Deleuze, X. Schultze and D.C. Sherrington, J. Mol. Catal. A-Chem., 2000, 159, 257. 63. J.F. White and J.C. Bertrand, in ‘Process for the production of esters’, US, 1977. 64. F.R. Abreu, D.G. Lima, E.H. Hamu, C. Wolf and P.A.Z. Suarez, J. Mol. Catal. AChem., 2004, 209, 29. 65. M. Gheorghiu, in ‘Process for the preparation of methyl fatty acid esters starting from natural oil or fat’, US, 1996. 66. H. Basu and M.E. Norris, in ‘Process for production of esters for use as a diesel fuel substitute using a non-alkaline catalyst’, US, 1996. 67. G.J. Suppes, K. Bockwinkel, S. Lucas, J.B. Botts, M.H. Mason and J.A. Heppert, J. Am. Oil Chem. Soc., 2001, 78, 139. 68. J. Weitkamp, M. Hunger and U. Rymsa, Micropor. Mesopor. Mat., 2001, 48, 255. 69. Y. Ono, J. Catal., 2003, 216, 406. 70. A. Philippou, J. Rocha and M.W. Anderson, Catal. Lett., 1999, 57, 151. 71. C.R. Bayense, H. Hinnekens and J. Martens, in ‘Esterification Process’, US, 1996. 72. W.J. Kim, M.C. Lee, J.C. Yoo and D.T. Hayhurst, Micropor. Mesopor. Mat., 2000, 41, 79. 73. G.Q. Guan, K. Kusakabe and S. Morooka, Micropor. Mesopor. Mat., 2001, 50, 109. 74. S. Gryglewicz, Bioresource Technol., 1999, 70, 249. 75. G.R. Peterson and W.P. Sacarrah, J. Am. Oil Chem. Soc., 1984, 61, 1593. 76. H. Hattori, Appl. Catal. A-Gen., 2001, 222, 247. 77. E. Leclercq, A. Finiels and C. Moreau, J. Am. Oil Chem. Soc., 2001, 78, 1161. 78. A. Corma, S. Iborra, S. Miquel and J. Primo, J. Catal., 1998, 173, 315. 79. S. Bancquart, C. Vanhove, Y. Pouilloux and J. Barrault, Appl. Catal. A-Gen., 2001, 218, 1. 80. J.M. Clacens, Y. Pouilloux and J. Barrault, Appl. Catal. A-Gen., 2002, 227, 181. 81. G.J. Suppes, in ‘Carbonate catalyzed alcoholysis of triglycerides’, US, 2003. 82. J.P. Cosman, R.J. Bettge, R.A. Elwell and R.A. Plepys, in ‘Process for purifying polyethers’, US, 2002. 83. J. Barrault, S. Bancquart and Y. Pouilloux, C. R. Chimie, 2004, 7, 593. 84. R. Stern, G. Hillion, J. -J. Rouxel and S. Leporq, in ‘Process for the production of esters from vegetable oils or animal oils and alcohol’, US, 1999. 85. Y. Pouilloux, S. Metayer and J. Barrault, Comptes Rendus De L Academie Des Sciences Serie Ii Fascicule C-Chimie, 2000, 3, 589. 86. A. Corma, M. Rodriguez, N. Sanchez and J. Aracil, in ‘Process for the selective production of monoesters of diols and triols using zeolitic catalysts’, EP, 1994. 87. A. Corma, S. Iborra, S. Miquel and J. Primo, in ‘Process and catalysts for the selective production of esters of fatty acids’, EP, 1998. 88. A. Neher, T. Haas, D. Arntz, H. Klenk and W. Girke, in ‘Process for the production of acrolein’, US, 1995. 89. M.D. Machado, J. Perez-Pariente, E. Sastre, D. Cardoso and A.M. de Guerenu, Appl. Catal. A-Gen., 2000, 203, 321. 90. M. Mittelbach, A. Silberholz and M. Koncar, Oils-Fats-Lipids 1995, Proceedings of the 21st World Congress of the International Society for Fats Research, The Hague, 1996, p. 497. 91. S. Furuta, H. Matsuhashi and K. Arata, Catal. Commun., 2004, 5, 721. 92. G. Waghoo, R.V. Jayaram and M.V. Joshi, Synth. Commun., 1999, 29, 513.

Catalysis, 2006, 19, 41–83 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117.

83

B.S. Balaji and B.M. Chanda, Tetrahedron, 1998, 54, 13237. M. Sasidharan and R. Kumar, J. Mol. Catal. A-Chem., 2004, 210, 93. B.P. Bandgar, L.S. Uppalla and V.S. Sadavarte, Green Chem., 2001, 3, 39. S.P. Chavan, P.K. Zubaidha, S. W. Dantale, A. Keshavaraja, A.V. Ramaswamy and T. Ravindranathan, Tetrahedron Lett., 1996, 37, 233. S.P. Chavan, Y.T. Subbarao, S.W. Dantale and R. Sivappa, Synthetic Commun., 2001, 31, 289. J.G. Goodwin, S. Natesakhawat, A.A. Nikolopoulos and S.Y. Kim, Catal. Rev., 2002, 44, 287. M.A. Harmer and Q. Sun, Appl. Catal. A-Gen., 2001, 221, 45. G.D. Yadav and J.J. Nair, Micropor. Mesopor. Mat., 1999, 33, 1. F. Omota, A.C. Dimian and A. Bliek, Chem. Eng. Sci., 2003, 58, 3175. A. Corma and H. Garcia, Catal. Today, 1997, 38, 257. G.D. Yadav and A.D. Murkute, J. Catal., 2004, 224, 218. S. Ardizzone, C.L. Bianchi, V. Ragaini and B. Vercelli, Catal. Lett., 1999, 62, 59. K. Arata, H. Matsuhashi, M. Hino and H. Nakamura, Catal. Today, 2003, 81, 17. H. Matsuhashi, H. Miyazaki, Y. Kawamura, H. Nakamura and K. Arata, Chem. Mater., 2001, 13, 3038. S. Furuta, H. Matsuhashi and K. Arata, Appl. Catal. A-Gen, 2004, 269, 187. F.S. Guner, A. Sirkecioglu, S. Yilmaz, A.T. Erciyes and A. ErdemSenatalar, J. Am. Oil Chem. Soc., 1996, 73, 347. R. Koster, B. van der Linden, E. Poels and A. Bliek, J. Catal., 2001, 204, 333. J. Perez-Pariente, I. Diaz, F. Mohino and E. Sastre, Appl. Catal. A-Gen., 2003 254, 173. I.K. Mbaraka, D.R. Radu, V.S.Y. Lin and B.H. Shanks, J. Catal., 2003, 219, 329. W. Korbitz, Renew. Energ., 1999, 16, 1078. V. Harman, in ‘Bio-Logical’, 2003. Minnesota Department of Agriculture, in ‘Governor Pawlenty establishes Minnesota Biodiesel Task Force’, ed. M.D.o. Agriculture, St Paul, Minnesota, 2003. H. Noureddini, in ‘System and process for producing biodiesel fuel with reduced viscosity and a cloud point below thirty-two (32) degrees fahrenheit’, US, 2001. J.H.S. Kesling, L.J. Karas and J.F.J. Liotta, in ‘Diesel fuel’, US, 1994. D.S. Bradin, in ‘Biodiesel fuel’, US, 1996.

Catalysis with Nitrides and Oxynitrides BY J.S.J. HARGREAVES AND D. MCKAY WestCHEM, Department of Chemistry, Joseph Black Building, University of Glasgow, Glasgow G12 8QQ, UK

1

Introduction

Advances in synthetic procedures are resulting in an expansion of accessible nitride materials.1–5 For example, it is only in recent years that the first nitrides possessing the spinel structure have been prepared,6–9 although previously oxynitrides with this structure were known.10,11 To date, primary catalytic interest5,12–14 in nitrides has centred around two general themes: (i) the well documented ability of some transition metal nitrides to exhibit catalytic properties which are similar to platinum group metals, and (ii) the acid–base properties of nitrides and oxynitrides. Up until now, catalytic studies have been directed upon the application of a relatively limited number of nitrides/ oxynitrides to a restricted number of reactions. However, recent years have seen an expansion in the range of reactions catalysed by nitrides along with the application of novel nitrides to established catalytic reactions. In this review, we detail a number of these developments in the context of previous studies. It is clear that the area of nitride and oxynitride catalysis is a vibrant field with a number of significant advances being reported lately. However, the catalytic behaviour of most of the reported nitride based materials is largely unknown and synthetic developments continue to markedly outstrip catalytic developments. Future exciting advances in this area can be anticipated. Nitrides can be sub-divided into ionic, covalent and interstitial types.15,16 An alternate general classification of nitrides, based on bonding classification, as ionic, covalent and metallic has also been applied. Ionic or ‘‘salt-like’’nitrides are formed by electropositive elements such as Li, Mg, Ca, Sr, Ba, Cu, Zn, Cd and Hg and possess formulae which correspond to those expected on the basis of the combination of the metal ion with N3 ions. A range of covalent nitrides are known and are exhibited by less electropositive elements such as B, S, P, C and Si. Interstitial nitrides are formed by some transition metals and refer to compounds which can be described in terms of the occupancy of interstitial sites in close packed metallic structures by nitrogen atoms. Oxygen can also be accommodated within these structures and a range of oxynitrides are known to


84

Catalysis, 2006, 19, 84–108

85

exist, along with carbonitrides wherein both nitrogen and carbon are distributed between interstitial sites. These subdivisions are, to an extent, arbitrary and nitrides/oxynitrides are further classified as binary (containing one metal within the main structure), ternary (containing two metals) and quaternary (containing three metals) etc. Arguably, nitrides and oxynitrides which are best described as interstitial have attracted the greatest attention in catalytic studies, although a number of oxynitrides in which partial substitution of the oxygen lattice within oxide structures by nitrogen has been performed have also attracted interest, particularly in terms of their acid-base character. There has also been a lot of interest in nitrogen doped oxides as photocatalysts. The main structure–types of the interstitial nitrides are given elsewhere.17 Of these, binary molybdenum nitrides have been the most well studied. Although a number of polymorphs have been characterised,18 it is the g-Mo2N phase which has been most widely reported in terms of its catalytic behaviour. The structure of this polymorph can be best described as a face centred cubic array of Mo atoms with nitrogen atoms occupying one half of the octahedral interstices. As discussed later on, there is frequently oxygen contained within this structure, particularly in the near-surface region, where the stoichiometry may differ from the bulk. It is only recently that the catalytic activity of ternary nitrides has begun to be reported, and of these Co3Mo3N has been the most widely studied. The structure of this nitride is face centred cubic and has been described in terms of interpenetrating diamond-type lattices.19

2

Preparation of Nitride and Oxynitride Catalysts

The major preparation routes to nitrides and oxynitrides have been documented elsewhere.17 Historically, a number of nitrides were prepared by direct reaction of the metal with nitrogen gas, a process generally requiring the use of high temperature (although, for example, lithium metal forms the nitride directly at ambient temperature) to produce low surface area nitrides. Nowadays, due to surface area considerations, it is rare to find catalysts prepared in this manner, with ammonolysis as described below being the preferred method. However, recently Antonelli and co-workers have reported the direct preparation of mesoporous titanium and niobium oxynitrides containing low levels of nitrogen by direct cleavage of N2 at room temperature.20–22 Generally, metal nitride and oxynitride catalysts have been prepared by the temperature programmed ammonolysis method. This technique was originally applied by Volpe and Boudart to MoO3 precursors to produce g-Mo2N which was pseudomorphic with the precursor, with surface areas up to 225 m2 g1.23 The resultant surface area of materials is known to be a strong function of the synthesis parameters as demonstrated in Table 1.18 Generally, high space velocities of ammonia and low temperature ramp rates are required for maximisation of surface area. The former has the effect of reduction in the partial pressure of water vapour generated on reaction, whilst the latter minimises sintering by solid-solid reaction. In a factorial study, Choi

86

Catalysis, 2006, 19, 84–108

Table 1

The influence of ammonolysis parameters on the surface area of molybdenum nitride adapted from reference 18 2

1

Amount of commercial MoO3 (Sg ¼ 2 m g ) precursor (g) Rate of temperature increase (K min1) -from 293 to 633 K -from 633 to 723 K -from 723 K to final T Final temperature (K) Step time (h) Ammonia flow rate (l h1) Specific surface area (m2 g1)

Mo2N–A

Mo2N–B

2–3

2–3

10 1 – 700 10–12 35 115–120

20 20 20 780 0.5 35 15–20

γ -Mo 2OyN1-y

γ -Mo2N (HSA)

MoO2

γ-Mo2N (MSA)

Mo

γ-Mo2N (LSA)

HxMoO3

MoO3 MoO2

Scheme 1 The inter-conversion pathways during ammonolysis of MoO3 adapted from reference 24

et al. have elucidated the influence of synthesis parameters on the structural properties of molybdenum nitrides prepared via this method.24 They have concluded that the use of low ramp rates combined with high space velocity channels the conversion of MoO3 through a g-Mo2OyN1y intermediate as summarised in the Scheme 1 above: (where LSA, MSA and HSA refer to low, medium and high surface area respectively). A number of studies have indicated that the molybdenum nitride prepared via temperature programmed ammonolysis contains a number of surface NHx residues (partially dehydrogenated forms of surface ammonia).25,26 In temperature programmed ammonolysis, samples are generally cooled in flowing nitrogen or ammonia – the choice of which can affect the surface NHx and hydrogen composition. This is followed by passivation at room temperature in a dilute oxygen mixture (typically 1% O2, although in some cases this can still lead to exotherms and a more dilute source may be preferable). In this latter stage, an oxynitride surface skin is formed which facilitates handling and protects the bulk nitride from reaction with air. Significant levels of oxygen are incorporated into the near surface region by this method27,28 and these are generally removed, to some extent, prior to catalytic application by hydrogen pre-treatment at ca. 4001C.24,26,28 In a detailed TPD and TPR study, Wei et al.

Catalysis, 2006, 19, 84–108

87

have shown that passivation removes some of the surface NHx species, with the more strongly held species being preferentially removed.26 Demczyk et al. have undertaken a high resolution transmission electron microscopy and XPS study on the surface structure of passivated and re-reduced g-Mo2N and have concluded that there are significant differences between the bulk and near surface structures.29 It is proposed that the surface structures of both of these sets of materials should be viewed as oxynitrides of general formula Mo2N3xOx with a body centred like structure. The temperature programmed ammonolysis procedure has been applied to a number of different molybdenum precursors. Jaggers et al. have made a comparison between the use of various oxide based precursors.30 MoO3, (NH4)6Mo7O24.4H2O, (NH4)2MoO4 and HxMoO3 were investigated and it was concluded that high surface areas result when the reactants and products were pseudomorphic (as in the MoO3 and (NH4)6Mo7O24 4H2O cases) and HxMoO3 was reported to yield higher surface areas than MoO3. Pseudomorphism has been reported in Volpe and Boudart’s pioneering preparation of g-Mo2N by ammonolysis.23 In a detailed study of the sequence of transformation, a topotactic relationship between the MoO3 parent and g-Mo2N was discerned, with intermediate MoO2 and oxynitride phases being formed. In Jaggers et al. study, the reaction was also proposed to occur via an oxynitride intermediate and hexagonal d-MoN (which can be described as hexagonal close packing of molybdenum with nitrogen atoms occupying all the octahedral interstitial sites) was observed to be formed in some cases. In their study of MoO3 and HxMoO3 precursors, Choi et al. did not evidence the formation of d-MoN intermediates/ products. However, Marchand and co-workers have reported that modified forms of this phase can be prepared by ammonolysis of MoS2 precursors.18 In addition, temperature programmed ammonolysis has also been applied to the preparation of supported g-Mo2N, e.g..31–33 Although temperature programmed ammonolysis is a convenient method for the preparation of nitrides and oxynitrides on the lab scale, Wise and Markel34 have drawn attention to the temperature gradient effects resulting from endothermic ammonia decomposition reaction, which can be considerable even on the small scale. In view of this, they propose that the use of N2/H2 mixtures would be preferable on the large scale to minimise these effects. Despite these considerations, this procedure continues to be the method of choice in many catalytic studies with, in addition to molybdenum nitrides, VN, W2N, TiN and NbN being prepared in this manner, e.g..35,36 This method can also be applied to mixed oxide precursors to yield ternary nitrides, e.g.37 and oxynitrides, e.g..38 Within the literature, it is notable that in a number of studies nitridation has been performed by isothermal treatment of the appropriate precursor under flowing ammonia. Examples of studies adopting this procedure have been the preparation of nitrogen-containing ZSM-5,39 nitrogen-containing NaY and the oxynitride phosphates AlPO-5,40 ZrPON,41 AlGaPON.42 Within these studies variation of nitridation temperature and/or duration has been shown to exert an influence over the final nitrogen content, e.g.41,43 which can, of course, have an important effect on the resultant catalytic properties43 (see Table 2).

88

Table 2

Catalysis, 2006, 19, 84–108

The influence of nitridation time on the nitrogen content of AlPONs adapted from reference 43

Composition

Nitridation Nitridation temperature (K) time (h)

Surface area (m2 g1)

Total N (wt.%)

Surface N (wt.%)

AlPO3.64N0.24 AlPO3.55N0.30 AlPO3.10N0.60 AlPO2.67N0.89 AlPO1.96N1.35 AlPO1.71N1.53

1073 1073 1073 1073 1073 1073

275 275 235 230 215 195

2.8 3.6 7.2 11.0 17.5 20.0

2.7 1.3 1.0 1.1 2.4 2.6

3 8 40 65 120 200

When controlled nitridation of surface layers is required, as for example in the modification of the chemical properties of the surface of a support, the atomic layer deposition (ALD) technique can be applied.44 This technique is based upon repeated separate saturating reactions of at least two different reactants with the surface, which leads to the controlled build-up of thin films via reaction of the second component with the chemisorbed residues of the first reactant. Aluminium nitride surfaces have been prepared on both alumina and silica supports by this method wherein reaction cycles of trimethylaluminium and ammonia have been performed with the respective supports, retaining their high surface areas.45 This method has been applied to the modification of the support composition for chromium catalysts supported on alumina.44 Within the catalysis literature, examples of nitridation methods other than reaction with ammonia or nitrogen at atmospheric pressure have been rare. Ammonolysis frequently leads to partial nitridation and, whether it is recognised or not, often the resultant compounds should be viewed as oxynitrides.15 However, the use of alternative preparation procedures can lead to unusual polymorphs of nitrides already known to possess catalytic activity which are therefore likely to be of catalytic interest in themselves. An example of this is the preparation of a-Mo2N, which possesses the a-Mo2C structure (an orthorhombic unit cell resulting from the ordering on interstitial nitrogen atoms in a hexagonally close packed array of molybdenum atoms). It has been claimed that this polymorph can be prepared by the metathesis reaction of MoCl5 with Ca3N2 in an excess of molten CaCl2 at elevated temperature:18 CaCl2

2MoCl5 þ 5=3Ca3 N2 ! Mo2 N þ 5CaCl2 þ 7=6N2 1123K

In another example, the high temperature reaction of calcium and strontium sub-nitrides with molybdenum foil has also recently been claimed to produce novel molybdenum ternary nitride oxides which contain isolated [MoN4]6 tetrahedra in an ordered sub-lattice of O2 and N3 anions.46 High temperature-high pressure syntheses are also yielding novel metastable structures, such as the spinel analogues mentioned previously.6,7 There is therefore an everincreasing base of nitrides and oxynitrides with diverse structures and potential catalytic interest.

Catalysis, 2006, 19, 84–108

3

89

Catalytic Reactions with Nitrides and Oxynitrides

In this section, the application of nitrides and oxynitrides as supports or catalysts themselves is outlined. The section is sub-divided by reaction/reaction type or application. Although emphasis is given to the more recent developments, where appropriate the historical context is given. 3.1 Ammonia Synthesis, Ammonia Decomposition and Hydrazine Decomposition. – The activity of nitrides and oxynitrides as catalysts for ammonia synthesis has been long recognised, e.g..47 Among some of the earlier publications in this area have been reports by King and Sebba48,49 on the use of vanadium oxynitride. In these studies, it was proposed that the active form of the catalyst was a vanadium oxide – nitride solid state solution, that lattice nitrogen was reactive towards hydrogenation to form ammonia and that the rate determining step of reaction was the adsorption of nitrogen into this state.48 Furthermore, for catalysts prepared at temperatures above 11001C it was found that there was a discontinuous activity increase in the 460–5001C reaction temperature range.48 In a subsequent paper, this was ascribed to disordering of the lattice oxygen and nitrogen species.49 Ammonia decomposition had been previously investigated over ‘‘vanadium nitride’’ by the same group,50 and over the temperature range studied (400–4801C), the kinetics were reported to be consistent with the Temkin-Pyzhev mechanism (in which nitrogen desorption is rate determining). However, it was noted that at hydrogen pressures below 200 mmHg, Temkin-Pyzhev kinetics were no longer obeyed. More recently, in his study using VN, Oyama51 has reported that above 4751C the kinetics are more consistent with the Tamaru mechanism (in which ammonia activation and nitrogen desorption are rate determining). This results in the rate being zero order in ammonia pressure at low temperature and first order at high temperature, whilst it is zero for both hydrogen and nitrogen. Rate parameters were calculated and were reported to be between those for Pt and Fe, which was taken as confirmation of the platinum group metal – like behaviour of VN. Structure-sensitivity for ammonia synthesis and decomposition has been claimed in a study by Choi et al.,52 with the surface area normalised activities increasing with decreasing surface area. Variation of stoichiometry as a function of particle size was proposed as a possible explanation. In this study, vanadium nitrides were also reported to possess lower activity than vanadium carbides which was explained on the basis of the greater acidity of the nitrides, adsorbing NH3 more strongly. The activity of molybdenum nitride in ammonia synthesis has been well established. Hillis et al.53 proposed that the rate determining step of the synthesis was one of the steps in the conversion of adsorbed nitrogen to gasphase NH3, rather than nitrogen adsorption itself. They also observed hydrogenation of bulk nitride to yield ammonia, but found it to be ca 50 times slower than the observed ammonia synthesis rate. These claims were subsequently investigated by Aika and Ozaki,54 who argued that nitrogen adsorption was rate determining. Volpe and Boudart55 measured the effect of variation of

90

Catalysis, 2006, 19, 84–108

Mo2N particle size on reaction rate and concluded that structure sensitivity was operative. Space-time yield ratios of 40 : 25 : 1 for 63, 13 and 3 nm particles were reported at atmospheric pressure and 4001C. Furthermore, NH3 was found to very strongly inhibit reaction rate. The interaction of NH3 with Mo(100) and Mo(100)-c(2 2)N surfaces56 and with b-Mo16N7, g-Mo2N, d-MoN films and g-Mo2N powders57 have been reported. In the former study, ammonia decomposition, which was observed to occur via stepwise dehydrogenation, was claimed to be less significant on the Mo(100)-c(2 2)N surface (argued to be a good model of the g-Mo2N surface) than the Mo(100) surface. The presence of surface nitrogen was found to reduce both the initial sticking coefficient and the saturation capacity with respect to the Mo(100) surface. In the latter study which employed TPD, NH3 decomposition was observed to produce two H2 desorption peaks, the lower temperature one being ascribed to surface hydrogen and the other to sub-surface hydrogen. The NH3 saturation capacity of the films was found to be in the order d-MoN o b-Mo16N7 o gMo2N and the behaviour of g-Mo2N films was proposed to resemble low surface area g-Mo2N powders, whilst the b-Mo16N7 and d-MoN films resembled high surface area g-Mo2N powders. Wise and Markel34 have reported that ammonia decomposition over Mo2N during temperature programmed synthesis fits a Tamaru rate expression. Recently, independent publications by Kojima and Aika58–61 and workers from Topsoe62–65 have reported on the high activity and stability of Cs/ Co3Mo3N based catalysts for ammonia synthesis. The activity has been found to be significantly higher than for the commercial iron based catalyst. Table 3 gives a comparison of the ammonia synthesis rates of Cs and K promoted catalysts at various promoter loadings (2, 5, 10 and 30 mol%) for low conversion at 4001C and atmospheric pressure.59 High activity was also evident at higher reaction pressure.61 Aika and Kojima have thoroughly investigated the influence of preparation,59 the reaction kinetics60 and reactant gas treatment.61 Cs was demonstrated to be the most effective promoter and it was argued that its effect was electronic in origin. It was also reported to retard the ammonia inhibition observed in this system. An optimum loading was evident since increasing alkali metal addition both decreased surface area and also suppressed Co3Mo3N phase formation, Table 3

The ammonia synthesis activities of various catalysts adapted from reference 59

Catalyst

Rate (mmol h1 g1)

Surface area (m2 g1)

Specific Activity (mmol h1 m2)

Fe–K2O–Al2O3 Co3Mo3N Co3Mo3N–K5 Co3Mo3N–K30 Co3Mo3N–Cs2 Co3Mo3N–Cs10

330 652 869 364 986 586

14 21 17 8 16 10

24 31 51 46 62 59

Catalysis, 2006, 19, 84–108

91

favouring less active Co and Mo2N phases instead. After passivation, pretreatment with N2/H2 at 6001C was found to yield higher activity catalysts than those pre-treated with either N2 or H2 alone. This was explained on the basis of more complete formation of the active Co3Mo3N phase.61 The same authors have published a paper which includes data on the lower activity Cs promoted Mo2N.66 In their study, Jacobsen and co-workers63,64 developed the catalyst based upon consideration of the volcano shaped relationship between the ammonia synthesis activity of various different metal catalysts and their nitrogen adsorption energy. Of the pure metals, Ru and Os were found to be best. It was argued that for optimum activity, two opposing factors were required – low N2 dissociation energy and low nitrogen surface coverage- and these could be achieved by alloying a metal with too high an adsorption energy (Mo) with one with too low an adsorption energy (Co). A theoretical study on this system has been reported and it was concluded that the only requirement of the nitrogen was to induce the correct ordering, with the (111) surface containing mixed Co– Mo sites being exhibited.64 In a subsequent microkinetic study, these authors reported that a good fit to ammonia synthesis kinetics could be made using nitrogen binding energy as the only adjustable parameter.65 This was taken to be consistent with the above idea, although it was recognised that this may be an over-simplification. Antonelli and co-workers22 have recently demonstrated that room temperature stoichiometric ammonia synthesis is possible with their mesoporous titanium and niobium oxide catalysts. In this study, they proposed that the ammonia species are formed via the reaction activated nitrogen with the underlying moisture of the support. Reversible, inter-conversion of N3 and NH2 species via exposure to moist air for aluminophosphate oxynitride catalysts has been observed by FTIR and XPS by Marquez and co-workers.67 There has been a lot of interest in the literature in the development of novel routes for the low temperature stoichiometric conversion of nitrogen to ammonia, e.g..68 However, in principle this could be realised by the nitridation of Li, followed by hydrolysis, although the kinetics would be very slow. Nitrides have also attracted interest as catalysts for hydrazine decomposition.69–72 This has applications in space technology for control of the altitude of space equipment such as satellites using thrusters. The traditional catalyst employed in these applications has been Ir/g-Al2O3. Alumina supported molybdenum nitride has been reported to be as active as supported Ir,72 as has tungsten nitride,69 although niobium nitride was less active.70 In an in-situ FTIR study of bulk and supported molybdenum nitride catalysts, Chen et al.72 have reported that below 4001C, hydrazine decomposes into ammonia and nitrogen, whereas above this temperature the ammonia produced further decomposes into nitrogen and hydrogen. Based on competitive adsorption studies with CO, they have concluded that surface molybdenum sites are the main locus of hydrazine adsorption. Passivated bulk Mo2N was observed to have a much lower activity than fresh Mo2N and a passivated catalyst which had been re-reduced. It was proposed that when the valence state of the Mo

92

Catalysis, 2006, 19, 84–108

species on the surface is lower than, or equal to, þ4, catalysts have activities comparable to Ir/g-Al2O3. 3.2 Amination and Ammoxidation. – There are comparatively few papers which have investigated the use of nitrides as catalysts for amination and ammoxidation. Neylon et al.73 have compared the activities of a range of nitrides with supported Pt and Ni catalysts for the amination of ethanol with ammonia in the presence of hydrogen. Although they were susceptible to deactivation, the nitrides were found to be very selective for the production of ethylamine. At low conversion the ethylamine selectivity was in the order: Mo2N 4 VN 4 W2N 4 NbN 4 TiN. At higher conversion, Mo2N was found to be more selective than Ni/SiO2 for the production of di- and tri- ethylamine, which was argued to be advantageous as these products are of greater value. On the basis of turnover frequency, the following activity order was proposed: VN 4 Mo2N 4 W2N 4 TiN 4 NbN. In the absence of ammonia, dehydration of ethanol occurred over the nitrides and, based on studies of the reactivities of potential intermediates, the reductive amination pathway was proposed to occur rather than the dehydration pathway as shown in the following Scheme 2 shown below. A Du Pont patent74 reports that tungsten nitride preferably supported on alumina or titania is an active catalyst for amination by reaction of an alcohol with an amine in the presence of hydrogen. Examples are given of the amination of monoethanolamine by diethylenetriamine and amination of monoethanolamine by ethylenediamine. Recently, amorphous high surface area vanadium aluminium oxynitrides have been reported as active catalysts for propane ammoxidation to yield acrylonitrile (AC) at atmospheric pressure.75–77 Optimal performance was achieved at 5001C using a C3H8:O2:NH3 molar ratio of 1.25 : 3 : 176 (see Tables 4 and 5). The space time yields of these catalysts have been reported to be much higher than for other catalysts reported in the literature. The presence of nitrogen within the catalyst was argued to be important on the basis that the performance of a vanadium aluminium oxide catalyst was much worse (acrylonitrile per pass yield ¼ 1.5%). The nitrogen content of the oxynitride catalyst was observed to increase with time on stream as shown in Table 5.76 It was argued that the mechanism of ammoxidation over this catalyst was different from other metal oxides, because propylene was not observed to be an

C2H4 -H2O CH3CH2OH CH3CH2NH2 +NH3

-H 2 CH3CHO

CH3CHNH

+H2

Scheme 2 The reaction pathways during ethanol amination over nitrides adapted from reference 73

93

Catalysis, 2006, 19, 84–108

Table 4

Catalyst

Comparison of propane ammoxidation performance of various catalysts adapted from reference 75 Reaction Conv. temperature C3H8 (1C) (%)

Mo–V–Nb– 440 Te–O V–Sb–W– 500 Al–O V–Sb–O 425 (Sb/V ¼ 5) Fe–Sb–O 500 (Fe:Sb ¼ 1 : 2) Ca–Bi–Mo 510 VAlON (V/ 500 Al ¼ 0.25)

Table 5

Select. ACN (%)

Yield ACN (%)

W/F g h/ Productivity mol (1 ACN/kg/ (C3H8) h) References

89.1

60.0

53.5

77

48

37

53

30

26.6

8

2036

0.9

80

22

23

5

740

1.51

81

15 59

63 50

9.5 29.5

12 8

384.6

31.6

78

164

79

177 812

82 75

The effect of time on line on the nitrogen content catalytic performance of VAlON, adapted from reference 76 Time (h) a

N content (wt.%) ACN selectivity (%) C3H8 conversion (%) a

0.5

2

5

10

12

24

1.8 12.3 56.9

3.1 24.7 58.6

3.8 35.1 58.6

4.2 46.0 60.8

4.5 54.5 60.8

4.5 56.6 60.8

Total nitrogen content.

intermediate under optimum reaction conditions.75,76 Increasing acrylonitrile selectivity with nitrogen content at constant conversion was ascribed to enhancement of surface basicity. The relatively high selectivity of this catalytic system for acetonitrile has been argued to be a further economic advantage.77 A subsequent TAP (temporal analysis of products) reactor study of this system163 has led to the proposal that strongly adsorbed lattice nitrogen in the form of NHx species are involved in the formation of acrylonitrile. More weakly held species such as co-ordinated NH3 or M–NH41 were proposed to be involved in non-selective reactions. Overall, the mechanism was described as being a double Mars – van Krevelen type, involving both oxygen and nitrogen from the lattice. 3.3 NO Removal. – A number of studies have investigated the potential of nitrides as catalysts for NO removal. NO reduction with hydrogen has been reported in a series of papers by Au and co-workers.19,83,84 Of the binary nitrides, the activities of VN, Mo2N and W2N have been compared.83 VN was found to be the least active and rapidly deactivated as a consequence of bulk oxidation, whereas both Mo2N and W2N were observed to possess activity with

94

Catalysis, 2006, 19, 84–108

the latter being higher and more stable. Both Mo2N and W2N underwent a degree of bulk oxidation which resulted in deactivation. By using a higher concentration of hydrogen to aid the removal of surface oxygen species, the rate of deactivation could be reduced. At a H2:NO ratio of 5 : 1, 480% conversion of a mixture containing 1% NO could be observed at 4001C using W2N over a period of 40 hours. In another study,19 the activities of Mo2N and Co3Mo3N were compared. Co3Mo3N was observed to be much more active than Mo2N and was resistant to oxidation. This was explained on the basis of structural considerations. Comparison, where possible, between the data reported in this study and the previous one, indicates that Co3Mo3N is more active than W2N. In all three of the studies, the nitrides have been investigated using XPS. The maximum oxidation states of the metals were observed which was attributed to the formation of surface oxides resulting from the passivation procedure. However, additional valencies were evident and, on this basis, the following valence states for the nitrides were proposed: Mod1 d ¼ 2 o d o 4,19,83 Wd1 d ¼ 0 o d o 484 or 483 and Vd1 d ¼ þ 2.83 For Co3Mo3N, only Mo data was reported and only oxidation state þ6 was evident.84 The surface NO species formed at room temperature over W2N were probed by DRIFTS84 and both N2O4 and chelating NO2 were evidenced in the as prepared sample. The same bands were apparent in the used sample, but with a much greater intensity of NO2. The activity of nitrides has also been reported for NO decomposition85 and reduction with CO.86 In the former study, Mo2N was investigated and was reported to have stable activity over 10 hours at a reaction temperature of 4501C. NO dissociation was proposed to occur and Mo2N which had been saturated with oxygen was still found to be active. It was proposed that Mo2OxNy was present under reaction conditions. Ni3N was investigated for NO and N2O reduction by CO at 2001C. Unlike N2O, for NO, a very high initial activity was observed which rapidly declined as a result of oxidation. The catalyst lifetime could be improved by reduction of the NO concentration, and a gas to surface diffusion barrier comprising powdered glass on top of the catalyst was also found to achieve this. It was proposed that, given the correct geometry and pore structure, a suitable support could also function as a diffusion barrier. 3.4 Hydrotreating and Hydrogenation. – Nitride and oxynitride catalysts have been extensively investigated as catalysts for hydrotreating (in particular hydronitrogenation and hydrodesulfurisation) and hydrogenation (for CO,87 aromatics,88 crotonaldehyde89 and alkenes90–93– the reduction of NO with H2 has been covered in the previous section). The use of nitrides, along with sulfides and carbides, as catalysts for hydroprocessing has recently been extensively reviewed by Furimsky14 and will not be discussed in detail here. Subsequently, Al-Megren et al.94 have published a comparison of the activities of bulk CoMo carbide, oxide, nitride and sulfide catalysts for pyridine hydrodenitrogenation. Of these, the sulfide catalysts were reported to possess more stable activity, with the carbide being next, followed

Catalysis, 2006, 19, 84–108

95

by the nitride and the oxide being the least stable. Some activation of sulfide was observed on testing, and this was ascribed to carbon incorporation. The initial activities of the carbide, nitride and oxide were reported to be similar and greater than that of the sulfide, although it had the greatest selectivities to cyclopentane and pentane. A comparison between the HDS (hydrodesulfurisation), HYD (hydrogenation) and HDN (hydrodenitrogenation) activities of vanadium sulfide, vanadium carbide and vanadium nitride has also been recently published.118 Both the carbide and nitride catalysts were found to contain a V2O3 phase. For HDS, the carbide and nitride have similar activity which was much greater than the sulfide, whereas the nitride possessed much lower activity for HDN. All three catalysts were found to exhibit similar performances in toluene hydrogenation. Both the carbide and the nitride were reported to be stable under hydroprocessing conditions. Ranhotra et al.,87 have compared the CO hydrogenation activities of the different phases of molybdenum carbide (both hexagonal and cubic) with gMo2N. Under the conditions tested, methane was the major product and the activities of the cubic carbide phase and the nitride were found to be similar. Olefin selectivity was higher over the hexagonal nitride phase. Both carbide phases were found to pass through an activity maximum, with declining CH4 selectivity, however, the nitride activity continually increased with increasing methane selectivity. Structure-sensitivity has been reported in the selective hydrogenation of crotonaldehyde with molybdenum nitride.89 The activities of bulk phase molybdenum nitride was compared with that supported on activated carbon and high surface area graphite. There was a crystallite size dependence, with the graphite supported catalyst being observed to be the most active. A relationship between the results of powder X-ray diffraction studies and activity data suggested that the more selective sites for crotylalcohol production were associated with the (200) planes of the nitride. Overall, the activity of the nitride catalysts was observed to be lower than that for Pt supported on graphite, although their selectivity was much higher. At 601C, some deactivation occurred after 25 minutes. Selective hydrogenation of benzene to cyclohexene has been investigated by Imamura et al.,88 using rare earth nitrides generated by thermal decomposition of amides. Both ytterbium and europium were studied, with the former being more active. A significant enhancement in selectivity was observed with ammonia treatment, but the overall activity declined. In another example of selective hydrogenation, Wu et al.,90 have characterised the surface intermediates present by in-situ FTIR during the hydrogenation of 1,3-butadiene to produce 1-butene over supported molybdenum nitride. Several types of adsorbed butadiene intermediate were observed – the p-, s- and dehydrogenated forms. The presence of the padsorbed form was argued to prevent further adsorption and reaction of butene. It was proposed that surface N atoms were responsible for weakening the interaction of the butadiene, leading to high selectivity. Hydrogenation of long chain linear alkadienes has been reported for molybdenum oxynitride synthesised by a method involving hydrazine pre-treatment of molybdenum oxide, followed by CaO treament to remove water.91 Optimum activity was

96

Catalysis, 2006, 19, 84–108

observed over a catalyst of composition MoO1.83N0.36 which was found to have both a better performance and better sulfur tolerance than palladium based catalysts. The activity was correlated to the presence of molybdenum sites with a valence between 4 and zero under reaction conditions. Propene hydrogenation has been reported over phosphorus containing molybdenum92 and tungsten oxynitrides93 prepared from the respective heteroployanion precursors. In each case, the presence of phosphorus was observed to enhance activity increasing both the number and strength of surface sites, making them more metallic-like, as had been proposed in previous studies of n-heptane isomerisation.94 3.5 Base Catalysis. – Nitrides and oxynitrides have attracted a lot of interest in terms of base catalysis. The activities of a range of oxynitrides for the Knoevenagal reaction, in which a carbonyl containing compound is reacted with a Z-CH2-Z 0 type compound, have been reported by Grange and co-workers. Amongst the systems tested have been nitrided aluminophosphates (AlPON),43,95 nitrided zirconophosphates (ZrPON),41,96–98 nitrided galloaluminophosphates (AlGaPON)99–101 and nitrided aluminovanadate (AlVON).102,103 The reaction between benzaldehyde and malononitrile has been most commonly investigated in these studies. In these materials, which are generally amorphous, surface Bronsted acid sites, mainly comprising P–OH species, and Lewis acid sites, resulting from surface cations, have been evidenced in addition to basic sites.101 It is usually found that increasing incorporation of nitrogen within the structures enhances both basicity (strength and number of sites) and catalytic activity, and therefore these can be adjusted by control of the amount of nitrogen incorporation. In the case of AlVON using DRIFTS CDCl3 measurements, the proton affinity of the strongest base sites was reported as 872 kJ mol1 which compares to 895 kJ mol1 for g-Al2O3.112 A number of these studies have sought to identify the nature of the basic sites and to characterise the forms of nitrogen present. In a number of cases, on the basis of DRIFTS experiments, the catalytic activity has been associated with the presence of specific sites, including: (i) surface hydrogenated groups, e.g. M–NH2 in AlGalPON,99–100,108 where the nitridation sequence has been proposed as M–O– NH41 and/or M–NH3 - M–NH2 - M–NH–M - N3, (ii) some oxygen species, in addition to N3 and NHx have been proposed as being very strong basic sites for ZrPON,96 and (iii) for AlVON, negatively charged oxygen species associated with the neutralisation of hydroxyl groups by reaction with ammonia (i.e. NH41O) have been identified as the catalytic base sites.103,112 The specific location of the various nitrogen species have been probed in the different catalysts using a number of techniques. For AlPONs, phosphorus atoms were observed to be preferentially nitrided with respect to aluminium atoms and phase segregation was reported, whereas for AlGaPONs nitrogen was incorporated into the first co-ordination sphere of P, Al and Ga atoms, although it was less favourable for Al.105 On this basis, at low levels of nitridation, nitridation of Ga–O–P bonds to Ga–N–P bonds was favoured over Al–O–P. Increasing Ga content was also reported to enhance nitrogen incorporation in these systems.

Catalysis, 2006, 19, 84–108

97

Nitridation was also shown to be specifically associated with V with respect to Al in AlVONs.106 Using a combination of DRIFTS, XPS and TGA, evidence has been reported for the formation of an intermediate metal-dinitrogen or azide species in the nitridation of AlVON.104 XAS measurements have demonstrated that mixed tetrahedra of the form XOxNy (where X ¼ P, Al, or Ga) are formed in the AlGaPON system.107,108 The extent of nitrogen incorporation, and therefore the basicity and catalytic activity, of these materials can be controlled by ammonolysis time and temperature. A study of the influence of temperature on nitrogen incorporation in the ZrPON system has been reported.41 It was shown that nitrogen substituted for oxygen above 5501C and the total nitrogen content increased with increasing temperature up to ca. 9001C. However, at higher temperatures, the cationic network was also affected with loss of phosphorus occurring at 9001C and reduction of zirconium at 13001C. Corma and co-workers113 have also investigated the effect of temperature on the incorporation of nitrogen species in AlPON. Using MAS NMR, they demonstrated that nitridation at 800–8501C generates P–NH2 groups by the reaction: O3Al–O–PO3 þ NH3 - O3–Al–OH þ H2N–PO3 subsequent interaction with ammonia at room temperature increased the Al coordination number to 5 and 6 by the following process: O3Al–O–PO3 þ 2NH3 - O3Al–OH þ H2N–PO3 At temperatures higher than 10001C, nitridation produced a tridymite like phase containing phosphorus atoms with more than one nitrogen in the first coordination sphere. A previous study by the same group114 had shown the presence of NH and NH2 by FTIR after standard nitridation. In this study, the ratio of secondary to primary amine groups was reported to increase with longer times of nitridation. The heterogeneity of surface species, and the unfavourable influence of nitrogen atoms in next nearest neighbour positions, was apparent in a study of the use of AlPON as a catalyst for arylsulfone synthesis.115 In this case, the formation of a-phenylsulfonylchalone was observed to pass through a maximum with increasing N content. Comparisons of the activity of the AlPON materials with the more widely applied MgO and hydrotalcite-derived base catalysts have been made and AlPONs performed favourably. It has been shown that although AlPONs have weaker basicity than MgO or hydrotalcite, and therefore lower activity, this can result in higher selectivity.114 Amongst other systems reported for the Knoevenagel condensation reaction have been, nitrogen incorporated ZSM-539 and high surface area silicon oxynitride.116 The ZSM-5 based -system potentially combines tuneable acid/ base properties with shape selectivity leading to the potential of selective conversion and/or unusual reaction pathways. The potential for Mo2N to function as a base catalyst has been demonstrated by Bej and Thompson109 in a study of acetone condensation. Isophorone, which is known to be produced from a base catalysed pathway, was a product

98

Catalysis, 2006, 19, 84–108

of reaction, as was mesityl oxide, which can be formed from metal or base catalysis, and mesitylene which was produced by an acid catalysed route. It was therefore deduced that Mo2N contained both surface acid and base sites and this was confirmed by CO2 and NH3 TPD. The surface area normalised acetone conversion rate was greater for Mo2N than both MgO and USY. Potassium loaded on a porous silicon nitride has been reported as a superbase catalyst.110 With appropriate pre-treatment, the catalyst was found to be highly active for 2,3-dimethylbut-1-ene isomerisation yielding 2,3-dimethylbut2-ene in close to 100% selectivity. In order for high activity to be demonstrated, 30 wt% potassium amide was loaded by impregnation. A trace of Fe2O3 was also added on impregnation. A silicon nitride synthesised via a silicon diimide precursor which was pre-treated at 10001C was found to be best and the catalyst was activated by heating in vacuo. The possibility that this resulted in active potassium nitride species via: 3KNH2 - K3N þ 2NH3 111

as proposed by others, was considered, since NH vibrational modes were lost in the FTIR spectrum on activation. However, it was concluded that the active phase was neither K nor K3N, but an intermediate phase. It was suggested that the well-defined porosity of the system could be used to develop shape-selective superbase catalysts. Although not strictly nitrides or oxynitrides, a significant amount of work has been done on the application of silica supported amines as catalysts, e.g..117 3.6 Photocatalysis. – Based on calculations, Asahi et al.,119 have recently demonstrated that doping titania (anatase) with nitrogen leads to active catalysts for visible light photocatalytic degradation of methylene blue and acetaldehyde. This observation has led to a flurry of research activity and, to date, nitrogen doped titania has been reported to possess visible light photocatalytic activity for oxidative degradation of propan-2-ol,120,121 NOx,122,123 4-chlorophenol,124 benzene,124 carbon monoxide,124 methyl orange132 and acetone.125 In general, consensus has been reached that the active phase of the catalyst is nitrogen doped anatase and that band gap narrowing is due to the localisation of N 2p states above the top of the O 2p valence band,126,127 as had originally been suggested.119 However, conflicting reports have been made on the effect of nitrogen doping on the rutile modification. In one study, doping rutile single crystals with nitrogen has been reported to cause a blue-shift of oxygen photodesorption,128 whereas others have reported a red-shift in the absorption edge of rutile powders.129 Theoretical studies126 support the former view and propose that although N 2p states reside in the band gap, the O 2p band undergoes a contraction, leading to a net increase in the band gap. The interpretation of the original suggestion of decomposition occurring via the oxidative pathways proposed by Asahi et al.119 has recently been questioned by Mrowetz et al..130 The authors have argued that there may be some contribution from the direct absorption of light by the methylene blue itself. They have also reported that the upper limit on the quantum yield of CO2 radicals from

Catalysis, 2006, 19, 84–108

99

oxidation of HCOO is below B2 105 at l 4 400 nm, although it was oxidised in the UV region. Due to the fact that the point of zero charge of the TiO2 surface is shifted to higher pH,121 they have also investigated the oxidation of NH3OH1 to NO3 and have similarly found that it is converted in the UV region, but not at all in the visible region. Visible light induced hydrophilicity has also been reported for nitrogen substituted titania films119,138 which is useful for the production of self-cleaning tiles and anti-fogging mirrors. A number of studies have investigated the effect of catalyst preparation. Amongst the various methods reported have been ball-milling P-25 with hexamethyleneamine.122 the use of 2,2 0 -bipyridine and aqueous ammonia precursor complexes,121 sputtering TiO2 with N2,119 ammonolysis of TiO2 powder, oxidation of TiN129 and hydrolysis of TiCl4 with ammonia, ammonium carbonate and ammonium bicarbonate.124 In general, it has almost universally been accepted that catalysts require pre-treatment at elevated temperature in order to exhibit activity. One exception to this has been a report of room temperature nitridation of titania colloidal particles with alkyl ammonium salts.131 This observation has recently been critically investigated.130 Following nitridation, catalysts have been reported to be yellow in colour and nitrogen incorporation in the lattice is generally accepted. However, on the basis of XPS and FTIR studies, one report has proposed that nitrogen is present in the form of hyponitrite.124 Co-doping TiO2xNx with lanthanum has been reported to inhibit sintering on thermal treatment.132 A study of the effect of nitrogen concentration on the photocatalytic activity of TiO2xNx powders for propan2-ol decomposition has concluded that when x increased in the range 0–0.019, the quantum yield decreases when visible irradiation is used.120 The same trend was observed with UV light which, in all cases, is reported to be more effective. Following the reports of the effect of nitrogen doping anatase, visible light photocatalysis has also been reported for SrTiO3 (NO elimination),133 MOxZnO (where M ¼ W, V or Fe for acetaldehyde decomposition),134 and TaON (for water splitting135,136 and methanol oxidation137). 3.7 Use as Supports. – A number of studies have investigated the use of nitrides as catalyst supports, the motivation for doing this has been enhanced thermal conductivity,139–142 greater inertness,143,144 modified basicity145–147 and increased hydrophobicity143 with respect to conventional supports. Garcia Cervantes et al.139 have compared the behaviour of Pd supported on SiC (17 m2 g1), Si3N4 (7 m2 g1), a-Al2O3 (11 m2 g1) and SiO2 (32 m2 g1) for 1,3butadiene hydrogenation. An emphasis was placed on the comparison of supports in terms of their thermal conductivity for this exothermic reaction. However, no correlation was found. Pd/SiO2 and Pd/Si3N4 were found to be similar in terms of deactivation, activity decreased rapidly and then stabilised. On the other-hand, Pd/Al2O3 and Pd/SiC did not stabilise within 20 hours on-stream. Even though Pd particle size was similar (4 nm), XPS indicated that the Pd in Pd/Si3N4 was more electron deficient than on other supports with a Pd 3d5/2 binding energy 1 eV greater than for Pd/SiO2 and Pd/SiC being

100

Catalysis, 2006, 19, 84–108

observed. Both Pd/SiO2 and Pd/Si3N4 exhibited 100% selectivity to butenes at 201C, after an induction period which was lower for Si3N4. The selectivity for Pd/SiC and Pd/ a-Al2O3 did not attain 100% within 18 hours of testing. The Si3N4 was mainly in the a- form although some b- was evident. In three further studies,140,141,143 Pd/Si3N4 has been investigated in terms of its methane oxidation behaviour. Methivier et al.140 have reported that these catalysts are highly stable even after being subjected to temperatures in excess of 8001C under reaction mixture. Again, electronic interaction between Pd and Si3N4 was observed and this was proposed to retard sintering and structural changes of the supported Pd particles. Methane combustion over Pd/Si3N4 supported on radiant panels has been reported by Cadete Sanos Aires et al.141 These panels use catalytic combustion to generate infra-red emission for paint drying and thermoforming of plastics. In this study, a comparison was made between Pd/a-Si3N4 and Pd/g-Al2O3 washcoats. a-Si3N4 supported catalysts were shown to operate over a long period of time (2500 hours) at high power density, although they were less active than the g-Al2O3 counterparts. This was ascribed to poor morphological properties and it was stated that there was scope for improving them. The short contact time partial oxidation of methane over Pd/Si3N4 in the 900–11001C temperature range has been reported by Monnet et al.143 For higher loading catalysts (i.e. 1.0 and 2.2 wt% Pt with average particle sizes of 5.0 nm and 6.4 nm respectively) above 9001C, small metal particles were observed to sinter and larger ones were observed to oxidise with loss via volatilisation occurring. A 0.045 wt% Pt catalyst (with an average particle size of 1.1 nm) proved to be stable at 9001 which was ascribed to metalsupport interactions. Boron nitride has been the subject of some studies.142,144 Lin et al.149 have made a comparison of the activities of Pt catalysts supported on two types of hexagonal BN – a higher crystallinity form and a higher surface area (i.e. 30 m2 g1 v. 2 m2 g1) lower crystallinity form. Pt particles were reported to bond more weakly to BN crystal faces than grain boundaries and more weakly to BN overall than to g-Al2O3 (100 m2 g1). Weaker support–Pt bonding was proposed to allow Pt particles to more easily remain in the reduced state, weakening Pt–O bonding on the surface of the particles and promoting benzene oxidation. On this basis, Pt/BN was suggested as an effective catalyst for reactions whose rate-determining step is oxygen activation. BN supported catalysts were also stated to be less susceptible towards chlorine poisoning than their g-Al2O3 counterparts. The same group have also reported on the use of Pt/BN for the catalytic destruction of VOC (principally gasoline vapour).142 In experiments run over a number of cycles, the activity was observed to increase after the first cycle, unlike the g-Al2O3 counterpart which deactivated. Using BN, no Pt sintering occurred and this was ascribed to the high thermal conductivity of BN, ensuring that no local hot-spots were formed. On the basis of XPS, the locus of Pt particle attachment was proposed to be surface boron oxide impurities. Taylor and Pollard150 have compared the activities of silica (194 m2 g1) and boron nitride (7 m2 g1) supported vanadium oxide catalysts for propane oxidation. The use of boron nitride was reported to significantly

Catalysis, 2006, 19, 84–108

101

enhance the selectivity towards acrolein. This was explained on the basis of the lower surface area and hydrophobicity of BN, in combination with the observation, by TPR, that the supported vanadium is in a chemically different form. Jacobsen144 has investigated the use of a BN support on the behaviour of Ba– Ru ammonia synthesis catalysts. He found no detectable deactivation over a 3500 hour test period. The selection of BN was made on the basis that it has a similar structure to graphite, which is usually used, and the incentive for testing was that it would be more resistant as methanation of the graphite can occur in some situations. Puurunen et al.45 have investigated the effect of surface nitridation of gAl2O3 supports by the atomic layer deposition process on the activity of chromium catalysts for isobutane dehydrogenation. Nitridation was observed to suppress activity and it was argued that oxide ions were more active for the dissociation of isobutane. In a series of papers investigating oxynitride supported Pt43,145–147and Pt–Sn43 for isobutane dehydrogenation, Grange and co-workers have demonstrated beneficial effects of enhanced basicity. Amongst the supports tested were AlPON,43,146 AlGaPON,145,147 and AlCrPON.145 In a comparison between platinum supported on AlPON, AlGaPON and AlCrPON, the order of surface area normalised activity was observed to be Pt/AlGaPON 4 Pt/AlCrPON 4 Pt/ AlPON and the enhanced activity with gallium substitution was ascribed to its effect on hydrogen recombination.145 In general, the dehydrogenation activity has been observed to correlate with the nitrogen content of the supports41,146,147 and XPS on the AlPON supported catalyst has shown that the solid becomes more basic (in terms of strength as well as site number) since the binding energies of Al, P, O and N decrease.147 The nitrogen content has also been observed to be influenced by the platinum precursor, with Pt(NH3)4(NO3)2 derived catalysts possessing a lower N content than those synthesised using Pt(acac)2.146 This was attributed to both the presence of water causing surface hydrolysis and the reaction between NO3 and adsorbed ammonia species to form N2. However, at comparable nitrogen content the activity of the Pt(NH3)4(NO3)2 derived catalyst was higher than their Pt(acac)2 counterparts, despite similar Pt dispersion. It was proposed that juxtaposed acid-base sites were responsible for isobutane activation and that the platinum function via assisting the loss of hydrogen.147 This explained the enhanced activity on increasing nitrogen incorporation. Pt/ZrPON has also been studied for heptane reforming and it has been shown that increasing nitrogen content promotes aromatisation and suppresses hydrogenolysis, whilst the dehydrogenation and isomerisation side-reactions are unaffected.148 This was also attributed to enhanced basicity (both number and strength of sites) on increasing N incorporation. Molybdenum nitride itself has been dispersed on platinum clusters dispersed in EMT zeolite.151 In this study, coverage of platinum particles by molybdenum deposition was investigated and was reported to suppress the activity for benzene hydrogenation and enhance the heptane conversion. With respect to bare Pt for heptane conversion, the aromatisation activity was reduced, whereas isomerisation and hydrogenolysis were observed.

102

Catalysis, 2006, 19, 84–108

3.8 Hydrogen Storage. – Within this area, the reversible reaction of certain nitrides with hydrogen has been investigated. Whilst this, in itself, is not a catalytic process, as detailed below some investigations have sought to apply catalysts to facilitate the rates of such processes. The potential of Li3N for hydrogen storage applications has been demonstrated by Chen et al.152 At 2551C, 9.3 wt% H2 was absorbed after 30 minutes. Under high vacuum, 6.3wt% could be released below 2001C, whilst the remaining 3 wt% was only released above 3201C. On the basis of XRD studies, the following pathway was proposed as operative: Li3N þ 2H2 2 Li2NH þ LiH þ H2 2 LiNH2 þ 2LiH which in principle could reversibly store 410 wt% H2. Subsequently, this system has attracted some interest.153–157 In particular, the second stage of the pathway has been investigated which can still store 6.5 wt% H2156and the favourable effect of a TiCl3 catalyst on the reverse process has been noted.156 Recently, the alternative Mg(NH2)2/LiH system has been investigated, where 7 wt% H2 could be stored and the reversibility of storage was reported to be complete.157 In addition, other nitride systems have also been investigated for hydrogen storage applications, including BN, e.g.164 and CN, e.g..165

4

Conclusions and Outlook

It is clear that there has been a resurgent interest in the use of nitrides and oxynitrides as heterogeneous catalysts in recent years. This has partly been driven by the use of temperature programmed techniques to produce high surface area materials. In this review, a number of exciting developments have been reported where the activities of nitrides/oxynitrides rival, if not better, current commercial catalytic systems. Whilst there has been an expansion in the areas of their application, these are still relatively limited and, it appears that much more remains to be discovered. Generally, interest has centred around the modified basicity and platinum group metal–like properties of such systems. However, it is generally unclear as to the extent which nitrides can be used as reservoirs of ‘‘active’’ nitrogen for nitrogen transfer reactions. Although nitrogen storage has recently been reported for transition metalintermetallic compounds,158 the extent to which a nitride may, e.g., participate in a Mars-van Krevelen type process still remains unclear. The possible analogy with this type of oxide behaviour had been raised in early studies of ammonia synthesis over uranium nitride159,160 and it appears that there may be such parallels with, the often related, carbide catalysts for methane oxidation.161 Very recently, since this review was completed, a ‘‘double’’ Mars-van Krevelen mechanism has been proposed for the production of acrylonitrile in propane ammoxidation over VAlON catalysts.163 Despite becoming more apparent from photocatalysis studies, the use of nitrogen as a ‘‘dopant’’ to modify the defect structure of oxides for thermally activated reactions is largely unexplored. Recent studies have detailed the unusual effects of low levels of nitrogen

Catalysis, 2006, 19, 84–108

103

incorporation on the ionic conductivity of yttria stabilised zirconia.162 Finally, the application of carbonitrides and other mixed ‘‘anion’’ nitrides seems to have been largely ignored. Further exciting progress with nitride-based materials can be anticipated.

Acknowledgments We would like to express our appreciation to Professor C. T. Au, Dr. C. Cellier, Professor E. Gaigneaux, Dr. J. Jizhong and Dr. T. Xiao for the kind provision of reprints. We are also grateful to Crystal Faraday and the EPSRC (GR/ S87300/01) for funding in the area of nitride catalysis. References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

19. 20. 21. 22. 23.

R. Niewa and H. Jacobs, Chem. Rev., 1996, 96, 2053. R. Niewa and F.J. DiSalvo, Chem. Mater., 1998, 10, 2733. E. Kroke and M. Schwarz, Co-ord. Chem. Rev., 2004, 248, 493. A. Simon, Co-ord. Chem. Rev., 1997, 163, 253. ‘‘The chemistry of transition metal carbides and nitrides’’, S.T.Oyama (ed.), Blackie Academic and Professional, Glasgow, 1996. A. Zerr, G. Miehe, G. Serghiou, M. Schwarz, E. Kroke, R. Riedel, H. Fuess, P. Kroll and R. Boehler, Nature, 1999, 400, 340. K. Leinenweber, M. O’Keffee, M. Somayazulu, H. Hubert, P.F. McMillan and G.W. Wolf, Chem.-Eur. J., 1999, 5, 3076. N. Scotti, W. Kockelmann, J. Senker, S. Trabel and H. Jacobs, Z. Anorg. Allg. Chem., 1999, 625, 1435. G. Serghiou, G. Miehe, O. Tschauner, A. Zerr and R. Boehler, J. Chem. Phys., 1999, 111, 4659. J.W. McCauley and N.D. Corbyn, J. Am. Ceram. Soc., 1979, 62, 476. J. Grins, P.-O. Koll and G. Svensson, J. Solid State Chem., 1995, 117, 48. S.T. Oyama and G.L. Haller, Catalysis, Spec. Per. Rep., Vol. 5, RSC, London, 1982, p. 333. P.W. Lednor (ed.), Catal. Today, 1992, 15. E. Furimsky, Appl. Catal. A: Gen, 2003, 240, 1. D.H. Gregory, J. Chem. Soc., Dalton Trans., 1999, 259. F.A. Cotton and G. Wilkinson, ‘‘Advanced inorganic chemistry’’, 5th Edn., Wiley, New York, 1988. S.T. Oyama, ‘‘The chemistry of transition metal carbides and nitrides’’, S.T. Oyama (ed.), Blackie Academic and Professional, Glasgow, 1996, Chapter 1, p. 1. R. Marchand, X. Gouin, F. Tessier and Y. Laurent, in ‘‘The chemistry of transition metal carbides and nitrides’’, S.T. Oyama (ed.), Blackie Academic and Professional, Glasgow, 1996, Chapter 13, p. 252. C. Shi, A.M. Zhu, X.F. Yang and C.T. Au, Catal. Lett., 2004, 97, 9. M. Vettraino, X. He, M. Trudeau, J.E. Drake and D.M. Antonelli, Adv. Funct. Mater., 2002, 12, 174. M. Vettraino, X. He, M. Trudeau and D. Anotnelli, Stud. Surf. Sci. and Catal., 2002, 141, 661. M. Vettraino, M. Trudeau, A.Y.H. Lo, R.W. Schurko and D. Antonelli, J. Am. Chem. Soc., 2002, 124, 9567. L. Volpe and M. Boudart, J. Solid State Chem., 1985, 59, 332.

104

Catalysis, 2006, 19, 84–108

24. 25. 26. 27. 28. 29. 30. 31. 32.

J.-G. Choi, R.L. Curl and L.V. Thompson, J. Catal., 1994, 146, 218. G.W. Haddix, D.H. Jones, J.A. Reimer and A.T. Bell, J. Catal., 1988, 112, 556. Z. Wei, Q. Xin, P. Grange and B. Delmon, J. Catal., 1997, 168, 176. E.J. Markel and J.W. van Zee, J. Catal., 1990, 126, 643. C.W. Colling, J.-G. Choi and L.T. Thompson, J. Catal., 1996, 160, 35. B.G. Demczyk, J.-G. Choi and L.T. Thompson, Appl. Surf. Sci., 1994, 78, 63. C.H. Jaggers, J.N. Michaels and A.M. Stacy, Chem. Mater., 1990, 2, 150. C.W. Colling and L.T. Thompson, J. Catal., 1994, 146, 193. M. Nagai, A. Miyata, T. Kusagaya and S. Omi, ‘‘The chemistry of transition metal carbides and nitrides’’, S.T. Oyama (ed.), Blackie Academic and Professional, London, 1996, Chapter 17, p. 372. Z. Wu, C. Li, Z. Wei, P. Ying and Q. Xin, J. Phys. Chem. B, 2002, 106, 979. R.S. Wise and E.J. Markel, J. Catal., 1994, 145, 335. M.K. Neylon, S.K. Bej, C.A. Bennett and L.T. Thompson, Appl. Catal. A: Gen., 2002, 232, 13. Y. Zhang, Y. Li, R. Raval, C. Li, R. Zhai and Q. Xin, J. Mol. Catal. A: Chem., 1998, 132, 241. S. Korlann, B. Diaz and M.E. Bussel, Chem. Mater., 2002, 14, 4049. C.C. Yu, S. Ramanathan and S.T. Oyama, J. Catal., 1998, 173, 1. C. Zhang, Z. Xu, K. Wan and Q. Liu, Appl. Catal. A: Gen., 2004, 258, 55. S. Ernst, M. Hartmann, S. Sauerbeck and T. Bongers, Appl. Catal. A: Gen, 2000, 200, 117. N. Fripiat and P. Grange, J. Mater. Sci., 1999, 34, 2057. S. Delsarte, A. Aroux and P. Grange, Phys. Chem. Chem. Phys., 2000, 2, 2821. A. Massinon, E. Gueguen, R. Conanec, R. Marchand, Y. Laurent and P. Grange, Stud. Surf. Sci. and Catal., 1996, 101, 77. R.L. Puurunen, S.M.K. Airaksinen and A.O.I. Krause, J. Catal., 2003, 213, 281. R.L. Puurunen, A. Root, P. Sarv, M.M. Viitanen, H.H. Brongersma, M. Lindblad and A.O.I. Krause, Chem. Mater., 2002, 14, 720. C.F. Baker, M.G. Barker, C. Wilson and D.H. Gregory, Dalton Trans., 2004, 1298. A. Ozaki and K. Aika, in Catalysis: Science and Technology, eds J.R. Anderson and M. Boudart, Springer-Verlag, Berlin, 1981, Volume 1, Chapter 3, p. 87. D.A. King and F. Sebba, J. Catal., 1965, 4, 253. D.A. King and F. Sebba, J. Catal., 1965, 4, 430. W.J. McGill and F. Sebba, J. Catal., 1963, 2, 104. S.T. Oyama, J. Catal., 1992, 133, 358. J.-G. Choi, J. Ha and J.-W. Hong, Appl. Catal. A: Gen., 1998, 168, 47. M.R. Hillis, C. Kemball and M.W. Roberts, Trans. Faraday Soc., 1966, 62, 3570. K.-I. Aika and A. Ozaki, J. Catal., 1969, 14, 311. L. Volpe and M. Boudart, J. Phys. Chem., 1986, 90, 4874. R. Bafrali and A.T. Bell, Surf. Sci., 1992, 278, 353. H.J. Lee, J.-G. Choi, C.W. Colling, M. Mudholkar and L.T. Thompson, Appl. Surf. Sci., 1995, 89, 121. R. Kojima and K.-I. Aika, Chem. Lett., 2000, 514. R. Kojima and K.-I. Aika, Appl. Catal. A: Gen., 2001, 215, 149. R. Kojima and K.-I. Aika, Appl. Catal. A: Gen., 2001, 218, 121. R. Kojima and K.-I. Aika, Appl. Catal. A: Gen., 2001, 219, 157. C.J.H. Jacobsen, M. Brorson, T. Sehested, H. Teunissen and E. Turnqvist, US Patent 6,235,676, 1999. C.J.H. Jacobsen, Chem. Commun., 2000, 1057.

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.

Catalysis, 2006, 19, 84–108

105

64. C.J.H. Jacobsen, S. Dahl, B.S. Clausen, S. Bahn, A. Logadottir and J.K. Norskov, J. Am. Chem. Soc., 2001, 123, 8405. 65. A. Boisen, S. Dahl and C.J.H. Jacobsen, J. Catal., 2002, 208, 180. 66. R. Kojima and K.-I. Aika, Appl. Catal A: Gen., 2001, 219, 141. 67. F. Marquez, R. Guil-Lopez, V. Fornes and A. Corma, Catal. Commun., 200021. 68. M. Dorr, J. Kassbohrer, R. Grunert, G. Kreisel, W.A. Brand, R.A. Werner, H. Geilmann, C. Apfel, C. Robl and W. Weigand, Angew. Chem. Int. Ed., 2003, 42, 1540. 69. J.A.J. Rodriguez, G.M. da Cruz, G. Bugli, M. Boudart and G. Djega-Mariadassou, Catal. Lett., 1997, 45, 1. 70. R. Brayner, G. Djega-Mariadassou, G.M. da Cruz and J.A.J. Rodrigues, Catal. Today, 2000, 57, 225. 71. X. Chen, T. Zhang, L. Xia, T. Li, M. Zheng, Z. Wu, X. Wang, Z. Wei, Q. Xiu and C. Li, Catal. Lett., 2002, 79, 21. 72. X. Chen, T. Zhang, M. Zheng, Z. Wu, W. Wu and C. Li, J. Catal., 2004, 224, 473. 73. M.K. Neylon, S.K. Bej, C.A. Bennett and L.T. Thompson, Appl. Catal. A: Gen., 2002, 232, 13. 74. R.C. Bowman, M.H. Tegen and G.E. Hartwell, US Patent 4, 1990, 024, 922. 75. M. Florea, R. Prada-Silvy and P. Grange, Catal. Lett., 2003, 87, 63. 76. M. Florea, R. Prada-Silvy and P. Grange, Appl. Catal. A: Gen., 2003, 255, 289. 77. R. Prada-Silvy, M. Florea, N. Blangenois and P. Grange, AIChE Journal, 2003, 49, 2228. 78. M. Vaarkamp and T. Ushikubo, Appl. Catal. A: Gen., 1998, 174, 99. 79. J. Nilsson, A.R. Landa-Canovas, S. Hansen and A. Andersson, J. Catal., 1999, 186, 442. 80. G. Centi, R.K. Grasselli, E. Patane and F. Trifiro, Stud. Surf. Sci. Catal., 1990, 55, 515. 81. M. Bowker, C.R. Bicknell and P. Kerwin, Appl. Catal. A: Gen., 1996, 136, 205. 82. J.S. Kim and S.I. Woo, Appl. Catal. A: Gen., 1994, 110, 173. 83. C. Shi, A.M. Zhu, X.F. Yang and C.T. Au, Appl. Catal. A: Gen., 2004, 276, 223. 84. C. Shi, X.F. Yang and C.T. Au, Catal. Today, 2004, 93–95, 819. 85. H. He, H.X. Dai, K.Y. Ngan and C.T. Au, Catal. Lett., 2001, 71, 147. 86. Y. Egashira and H. Komiyama, Ind. Eng. Chem. Res., 1990, 29, 1583. 87. G.S. Ranhotra, A.T. Bell and J.A. Reimer, J. Catal., 1987, 108, 40. 88. H. Imamura, T. Nuruyu, T. Kawasaki, T. Teranishi and Y. Sakata, Catal. Lett., 2004, 96, 185. 89. A. Guerrero-Ruiz, Y. Zhang, B. Bachiller-Baeza and I. Rodriguez-Ramos, Catal. Lett., 1998, 55, 165. 90. Z. Wu, Z. Hao, P. Ying, C. Li and Q. Xin, J. Phys. Chem. B, 2000, 104, 12275. 91. Y. Li, Y. Fan and Y. Chen, Catal. Lett., 2002, 82, 111. 92. P. Perez-Romo, C. Potvin, J.-M. Manoli, M.M. Chehimi and G. Djega-Maraidassou, J. Catal., 2002, 208, 187. 93. P. Perez-Romo, C. Potvin, J.M. Manoli, R. Contant and G. Djega-Mariadassou, J. Catal., 2002, 205, 191. 94. H.A. Al-Megren, T. Xiao, S.L. Gonzalez-Cortes, S.H. Al-Khowaiter and M.L.H. Green, J. Mol. Catal. A: Chem., 2005, 225, 143. 95. Y. Laurent, P. Bastians, R. Conanec, R. Marchand, Y. Laurent, L. Gandia, M. Montes, J. Fernandez and A. Odriozola, Stud. Surf. Sci. and Catal., 1995, 91, 381. 96. N. Fripiat, V. Parvulescu, V.I. Parvulescu and P. Grange, Appl. Catal. A: Gen., 1999, 181, 331.

106

Catalysis, 2006, 19, 84–108

97. N. Fripiat, R. Conanec, A. Auroux, Y. Laurent and P. Grange, J. Catal., 1997, 167, 543. 98. N. Fripiat, R. Conanec, R. Marchand, Y. Laurent and P. Grange, J. Eur. Ceram. Soc., 1997, 17, 2011. 99. S. Delsarte, A. Aroux and P. Grange, Phys. Chem. Chem. Phys., 2000, 2, 2821. 100. S. Delsarte and P. Grange, Stud. Surf. Sci. and Catal., 2000, 130, 3237. 101. S. Delsarte, V. Peltier, Y. Laurent and P. Grange, Stud. Surf. Sci. and Catal., 1997, 118, 869. 102. H. Wiame, L. Bois, P. L’haridon, Y. Laurent and P. Grange, J. Eur. Ceram. Soc., 1997, 17, 2017. 103. H.M. Wiame, C.M. Cellier and P. Grange, J. Phys. Chem. B, 2000, 104, 591. 104. H.M. Wiame, M.A. Centeno, L. Legendre and P. Grange, Stud. Surf. Sci. and Catal., 1998, 117, 879. 105. S. Delsarte, V. Serin, A.-M. Flank, F. Villain and P. Grange, J. Solid State Chem., 2002, 163, 163. 106. H. Wiame, L. Bois, P. Lharidon, Y. Laurent and P. Grange, Solid State Ionics, 1997, 101–103, 755. 107. S. Delsarte, F. Mause, J.-C. Lavalley and P. Grange, Catal. Lett., 2000, 68, 79. 108. S. Delsarte, V. Serin, A.M. Flank, F. Villain and P. Grange, J. Synchroton Rad., 1999, 6, 433. 109. S.K. Bej and L.T. Thompson, Appl. Catal. A: Gen., 2004, 264, 141. 110. S. Kaskel and K. Schlichte, J. Catal., 2001, 201, 270. 111. T. Baba, H. Hanada and Y. Ono, J. Chem. Soc., Faraday Trans., 1994, 90, 187. 112. H. Wiame, C. Cellier and P. Grange, J. Catal., 2000, 190, 406. 113. T. Blasco, A. Corma, K. Fernandez, V. Fornes and R. Guil-Lopez, Phys. Chem. Chem. Phys., 1999, 1, 4493. 114. M.J. Climent, A. Corma, V. Fornes, A. Frau, R. Guil-Lopez, S. Ibarra and J. Prino, J. Catal., 1996, 163, 392. 115. M.J. Climent, A. Corma, R. Guil-Lopez, S. Ibarra and J. Prino, Catal. Lett., 1999, 59, 33. 116. P.W. Lednor and R. de Ruiter, J. Chem. Soc., Chem. Commun., 1991, 1625. 117. K.A. Utting and D.J. Macquarrie, Appl. Catal. A: Gen., 2002, 232, 7. 118. P. Rodriguez, J.l. Brito, A. Albornorz, M. Labadi, C. Pfaff, S. Marrero, D. Moronta and P. Betancourt, Catal. Commun, 2004, 5, 79. 119. R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki and Y. Taga, Science, 2001, 293, 269. 120. H. Irie, Y. Watanabe and K. Hashimoto, J. Phys. Chem. B, 2003, 107, 5483. 121. M. Miyauchi, A. Ikezawa, H. Tobimatsu, H. Irie and K. Hashimoto, Phys. Chem. Chem. Phys., 2004, 6, 865. 122. S. Yiu, H. Yamaki, M. Komatsu, Q. Zhang, J. Wang, Q. Tang, F. Saito and T. Sato, J. Mater. Chem., 2003, 13, 2996. 123. T. Sano, N. Negishi, K. Koike, K. Takeuchi and S. Matsuzawa, J. Mater. Chem., 2004, 14, 380. 124. S. Sakthivel and H. Kisch, Chem. Phys. Chem., 2003, 4, 487. 125. T. Ihara, M. Miyoshi, Y. Iriyama, O. Matsumoto and S. Sugihara, Appl. Catal. B: Env., 2003, 42, 403. 126. C. Di Valentin, G. Pacchioni and A. Selloni, Phys. Rev. B, 2004, 70, 085116. 127. H. Kisch and W. Macyk, Chem. Phys. Chem., 2002, 3, 399. 128. O. Diwald, L. Thompson, E.G. Goralski, S.D. Walck and J.T. Yates, Jr., J. Phys. Chem. B, 2004, 108, 52.

Catalysis, 2006, 19, 84–108

107

129. T. Morikawa, R. Asahi, T. Ohwaki, K. Aoki and Y. Taga, Jap. J. Appl. Phys., 2001, 40, L561. 130. M. Mrowetz, W. Balcerski, A.J. Colussi and M.R. Hoffmann, J. Phys. Chem. B, 2004, 108, 17269. 131. J.L. Gole, J.D. Stout, C. Barda, Y. Lou and X. Chen, J. Phys. Chem. B, 2004, 108, 1230. 132. H. Wei, Y. Wu, N. Lun and F. Zhao, J. Mater. Sci., 2004, 39, 1305. 133. J. Wang, S. Yin, M. Komatsu, Q. Zhang, F. Saito and T. Sato, Appl. Catal. B: Env., 2004, 52, 11. 134. D. Li, H. Haneda, N. Ohashi, S. Hishita and Y. Yoshikawa, Catal. Today, 2004, 93–95, 895. 135. G. Hitoki, T. Taneka, J.N. Kondo, M. Hara, H. Kobayashi and K. Domen, Chem. Commun., 2002, 1698. 136. M. Hara, T. Takata, J.N. Kondo and K. Domen, Catal. Today, 2004, 90, 313. 137. S. Ito, R. Thampi, P. Conte, P. Liska and M. Gratzel, Chem. Commun., 2005, 268. 138. H. Irie, S. Washizuka, N. Yoshino and K. Hashimoto, Chem. Commun., 2003, 1298. 139. G. Garcia Cervantes, F.J. Cadete Santos Aires and J.C. Bertolini, J. Catal., 2003, 214, 26. 140. C. Methivier, J. Massardier and J.C. Bertolini, Appl. Catal. A: Gen., 1999, 182, 337. 141. F.J. Cadete Santos Aires, S. Ramierz, G. Garcia Cervantes, E. Rogemond and J.C. Bertolini, Appl. Catal. A: Gen., 2003, 238, 289. 142. J.-C. Wu, Z.-A. Lin, J.-W. Pan and M.-H. Rei, Appl. Catal. A: Gen., 2001, 219, 117. 143. F. Monnet, Y. Schuurmann, F. Cadete Santos Aires, J.C. Bertolini and C. Mirodatos, Catal. Today, 2001, 64, 51. 144. C.J.H. Jacobsen, J. Catal., 2001, 200, 1. 145. E. Gueguen, S. Delsarte, V. Peltier, R. Conanec, R. Marchand, Y. Laurent and P. Grange, J. Eur. Ceram. Soc., 1997, 17, 2007. 146. M.A. Centeno, M. Debois and P. Grange, J. Catal., 2000, 192, 296. 147. S. Delsarte, F. Mauge and P. Grange, J. Catal., 2001, 202, 1. 148. N. Fripiat and P. Grange, Catal. Lett., 1999, 62, 53. 149. C.-A. Lin, J.C.S. Wu, J.-W. Pan and C.-T. Yeh, J. Catal., 2002, 210, 39. 150. S.H. Taylor and A.J.J. Pollard, Catal. Today, 2003, 81, 179. 151. T. Becue, J.-M. Manoli, C. Potvin, R.J. Davis and G. Djega-Mariadassou, J. Catal., 1999, 186, 110. 152. P. Chen, Z. Xiang, J. Luo, J. Lin and K.L. Tan, Nature, 2002, 420, 302. 153. Y.H. Hu and E. Ruckenstein, Ind. Eng. Chem. Res., 2003, 42, 5135. 154. P. Chen, Z. Xiang, J. Luo, J. Lin and K.L. Tan, J. Phys. Chem. B, 2003, 107, 10967. 155. Y.H. Hu and E. Ruckenstein, J. Phys. Chem. A, 2003, 107, 9737. 156. T. Ichikawa, N. Hanada, S. Isobe, H. Leng and H. Fuji, J. Phys. Chem. B, 2004, 108, 7887. 157. H.Y. Leng, T. Ichikawa, S. Hino, N. Hanada, S. Isobe and H. Fuji, J. Phys. Chem. B, 2004, 108, 8763. 158. M. Ito, K. Machida, K. Hirose, T. Sakata, H. Mori and G. Adachi, J. Phys. Chem. B, 1999, 103, 9498. 159. N. Segal and F. Sebba, J. Catal., 1967, 8, 105. 160. N. Segal and F. Sebba, J. Catal., 1967, 8, 113.

108

Catalysis, 2006, 19, 84–108

161. T. Xiao, A. Hanif, A.P.E. York, Y. Nishizaka and M.L.H. Green, Phys. Chem. Chem. Phys., 2002, 4, 4549. 162. M. Kilo, M.A. Taylor, C. Argirusis, G. Borchardt, M. Lerch, O. Kaitasov and B. Lesage, Phys. Chem. Chem. Phys., 2004, 6, 3645. 163. M. Olea, M. Florea, I. Sack, R. Prada Silvy, E.M. Gaigneaux, G.B. Marin and P. Grange, J. Catal., 2005, 232, 152. 164. S.-H. Jhi and Y.-K. Kwon, Phys. Rev. B, 2004, 69, 245407. 165. X.D. Bai, D. Zhong, G.Y. Yang, X.C. Ma, S. Liu, E.G. Wang, Y. Chen and D.T. Shaw, Appl. Phys. Lett., 2001, 79, 1552.

Kinetics of Surface Reactions with Lateral Interactions: Theory and Simulations BY C.G.M. HERMSE AND A.P.J. JANSEN Laboratory of Inorganic Chemistry and Catalysis, ST/SKA, Eindhoven University of Technology, P.O. Box 513, Eindhoven 5600 MB, The Netherlands

1

Introduction

Interactions between adsorbates, so-called lateral interactions, on transition metal surfaces have at least been known for as long as diffraction techniques have revealed that adlayers can form very well-defined structures at low temperatures. The importance of these interactions for kinetics at higher temperatures has only more recently been acknowledged. There may be various reasons for that, both experimental and theoretical. To show the importance of the lateral interactions in a kinetic experiment it is necessary that a system is well-defined, so that other possible causes for the experimental results can be excluded. For the same reason a system should be simple, which also helps in the analysis of the experimental results, because the relation between lateral interactions and their effects is generally complex. Only in recent years all these conditions have been met.1–27 The kinetics of catalytic processes is normally described by reaction rate equations. Because reaction rate constants may differ by many orders of magnitude, these equations are often already very difficult to handle when one ignores lateral interactions. A justification for this may be that catalysis is often done at high temperatures. If the thermal energy is larger than the lateral interactions, the adlayer is disordered and the rate equations should be at least a good approximation. There are a number of reasons why this justification is not valid. First, not all catalysis takes place at high temperatures. In fact low temperatures, if possible, are preferred. Second, the magnitude of lateral interactions varies substantially. We will show that the lateral interactions between adsorbates that are close together are often much larger than the thermal energy. This means that at high coverages relevant to catalysis the adlayer is ordered even at high temperature. Third, we will show that even if there is disorder the lateral interactions may still be important. This is because there remains some correlation in the occupation of neighboring sites even Catalysis, Volume 19 r The Royal Society of Chemistry, 2006

109

110

Catalysis, 2006, 19, 109–163

when there is disorder, and it is this correlation that affects the kinetics. One should distinguish between long-range order, as in talking about an ordered or disorder adlayer, and short-range order, as in talking about this correlation. This paper deals with the following topics. We first define precisely what we mean by lateral interactions in Section 2, and discuss the various mechanisms that lead to these interactions. The section also presents an overview of how these lateral interactions show up in experiments. This includes equilibrium phenomena, but, as this paper focuses on kinetics, only in a condensed form and only as far as it is relevant to kinetics. We also introduce the restriction to lattice-gas models and where to look for off-lattice approaches. Section 3 deals with the theory of lateral interactions. This means a discussion of different approaches to describe the kinetics when there are lateral interactions, and methods to obtain quantitative data on these interactions. The former introduces our main method to study the effect of lateral interactions on kinetics: kinetic Monte Carlo (kMC) simulations. The latter includes the determination of lateral interactions from experiments. The reason why we put this in a section on theory is that an analysis of experimental data to extract lateral interactions usually involves simulating the experiment. Finally, Section 4 discusses in some detail a number of reaction systems that show various aspects of the influence of lateral interactions. To be specific the reduction of reactivity because vacant sites become energetically unfavorable to occupy due to strong repulsion, the formation of various adlayer superstructures, and island formation in the reduction of NO on Rh(111), the order-disorder phase transition and how it shows up in voltammograms of sulfate on (111) surfaces of fcc metals, the determination of lateral interactions from fitting temperature-programmed desorption experiments with kMC simulations of CO desorption from Rh(100), the feasibility of calculating lateral interactions for O/Pt(111) with density-functional theory, and the formation of chiral domains in tartaric acid on Cu(110).

2

Basics of Lateral Interactions

2.1 The Mechanism of Lateral Interactions. – Adsorbates on transition metals that are not too far apart interact in a similar way as atoms or molecules in the gas phase. There is also another contribution that is due to the fact that adsorbates change the geometric and electronic structure of the substrate. This second contribution is often stronger than the first. Lateral interactions are usually divided into two classes, direct and indirect. Interactions from the first class have a gas phase equivalent. Those of the second class have no gas phase equivalent. They are referred to as through-surface or surface-mediated. The direct interactions are subdivided further into electrostatic interactions, Van-der-Waals interactions, and hybridization between neighboring adsorbates. The indirect interactions are subdivided into electronic through-surface interactions and elastic interactions. Below we give a short overview of the different types of lateral interactions. These are discussed in more detail in references.28–30

111

Catalysis, 2006, 19, 109–163

2.1.1 Electrostatic Interactions. If the nature of the bond between metal and adsorbate is sufficiently ionic, then a dipole will exist along the surfaceadsorbate axis (Figure 1). For positive ions like sodium and potassium, the dipole will be directed out of the surface, while for negative ions like sulfur and chlorine, the dipole will point into the surface. A neighboring adsorbate will sense this dipole, and its binding energy will change accordingly. If both the adsorbates present are identical, then the dipoles formed will point in the same direction, and they will repel each other. This is the case for potassium in the presence of other potassium atoms, and for sulfur in the presence of other sulfur atoms. If one adsorbate is a cation, the other one an anion, then the dipoles will point in opposite directions, and they will attract each other. This is the case for coadsorption of cations with anions. Electrostatic interactions are small if one or both the adsorbates are not ionically bound to the surface.31,32 Note that although this is considered a direct interaction the substrate, being a metal, plays a role because it forms an image charge. Apart from a surfaceadsorbate dipole, also a dipole inside the adsorbed molecule may be present. This dipole is usually a lot smaller than for the case of ion adsorption. It does, however, influence the bonding of neighboring adsorbates, be it in a more subtle way. For example, in the case of CO on Rh(111) surfaces it has been shown that the internal C–O stretch frequency shifts due to dipole-dipole coupling with neighboring adsorbed CO molecules. A similar effect has also been seen for NO on Rh(111) and Pd(111).33–35 2.1.2 Hybridization between Neighboring Adsorbates. When two molecules are brought close to each other, their wavefunctions will mix to form new hybrids. This also happens for molecules adsorbed on surfaces. The hybridization is due to an overlap between wavefunctions on one adsorbate with wavefunctions on another adsorbate. This causes the bonding electronic states involved in the interaction to drop in energy (i.e., become more stable), while the antibonding electronic states increase in energy. The net interaction can be repulsive or attractive, depending on the occupation of the bonding and antibonding electronic states involved. The attractive case is referred to as bond formation. The repulsive case is often described as a steric interaction. This type of A

c

−

−

b

+

+

−

+

d

+

−

A

a +

+

−

−

Figure 1 Examples of electrostatic interactions. (a) A schematic side view of a surface, with two anions (indicated by the A’s) binding to it. (b) The same as in (a), but now showing the surface-adsorbate dipoles (large ovals) due to adsorption of the anions. Since the dipoles point in the same direction, they repel each other. (c) Similar, to (b), except that the adsorbates are now cations. (d) When cations and anions are mixed, their dipoles point in opposite directions, and they attract each other

112

Catalysis, 2006, 19, 109–163

interaction requires close proximity (o2 A˚). A good example of hybridization resulting in an attractive interaction is the formation of NO dimers on Ag(111) surfaces. A good example of hybridization resulting in a repulsive interaction is the formation of chiral domains for (R,R)-tartaric acid on Cu(110) because of hydroxyl groups interacting (see Section 4.5).36–38 2.1.3 Van-der-Waals Interactions. There always is bonding due to Van-derWaals interactions when bringing two molecules close. This effect is usually small compared to the other contributions (originating from both direct and through-surface interactions), and is therefore often neglected. One expects Van-der-Waals interactions to dominate only for closed shell systems (like adsorbed noble gases), since in this case all other types of interactions are negligible. An example of such a system is Xe on Pt(111).39,40 2.1.4 Electronic Through-Surface Interactions. An adsorbate changes the electronic structure of the substrate. This affects the propensity of the substrate for adsorption of other atoms and molecules. This is usually expressed as that each surface atom can only form a limited number of bonds with adsorbates. With only one adsorbate present, the bond is strongest. If more than one adsorbate binds to a surface atom, then the hybridization of the metal will be different, and the interaction with each adsorbate will be weaker than for the case of just one adsorbate present. This decrease in binding in energy increases with the number of adsorbates binding to the surface atom. This so-called electronic throughsurface interaction is always repulsive when two adsorbates bind to one surface atom. Electronic through-surface interactions where two adsorbates bind to neighboring surface atoms can be both repulsive and attractive. 2.1.5 Elastic Interactions. The adsorption of an atom or a molecule may also change the geometry of the substrate. Through-surface interactions which are due to an adsorbate-induced change in the substrate geometry are referred to as elastic interactions.41,42 To explain how an elastic interaction can arise, let’s take a closer look at a simple model for the adsorption of sulfate anions on the Cu(111) surface (Figure 2). The sulfate anion binds with two oxygens to the surface. However, the distance between the oxygens is larger than the distance between the copper atoms in the surface. To remove this mismatch, either the sulfate oxygens have to be pushed closer to each other, or the distance between neighboring copper atoms in the surface has to be stretched. Experimentally, the last is observed: the copper–copper distance increases when sulfate is bound to the surface.43–45 This stretching of the copper-copper distance of course comes at an energetical penalty. If a second sulfate anion binds to the surface, it will also bind best when the surface copper-copper distance is increased. If this second sulfate anion binds far from the first, the energetical penalty for stretching the copper-copper distance will be similar to the one for the first. However, if it binds close to the first sulfate anion, then the copper–copper distance has already been increased by adsorption of the first sulfate anion. The energetical penalty for stretching the copper–copper distance will therefore be

113

Catalysis, 2006, 19, 109–163

a

c b

Figure 2 A possible mechanism for the attractive elastic interaction between sulfate ions on Cu(111). (a) A schematic view of part of the Cu(111) surface, showing the surface copper atoms (filled circles). (b) Upon adsorption of sulfate (large oval), the coppercopper distance is stretched, and copper atoms (filled circles) are displaced from their original positions (open circles). In total 8 surface atoms are shifted. (c) If two sulfates bind close to each other, then in total 14 surface atoms are shifted. If two sulfates bind far away from each other, then 16 surface atoms (twice the amount for the case displayed in (b)) will be shifted with respect to the second metal layer. Less reorganization of the surface atoms is therefore required when the sulfates bind close to each other. There is thus an effective attraction between the sulfates

less, and there will be an effective attraction between the two sulfate anions. Elastic interactions in general can be long-range, and both attractive and repulsive in nature. This is because adsorption of molecules can cause significant reorganization of the surface atoms, sometimes even with surface reconstructions as a consequence.46,47 In the last two sections we have treated changes in the substrate geometry and in the electronic structure separately. It needs to be mentioned that when the substrate geometry changes the electronic structure of course also will change. Through-surface interactions are therefore usually composed of both an elastic and an electronic component. 2.1.6 Energy versus Entropy Effects. If we regard the part of a potential energy surface around a minimum corresponding to an adsorption onto a site, then the previous paragraphs describe a shift in energy of this part due to lateral interactions. The lateral interactions may also change the curvature of the potential energy surface. As a consequence the vibrations of the adsorbate will get different frequencies. This has been observed for various adsorbates.33–35,48–51 If we have a vibration with an excitation energy that is much larger than the thermal energy ( ho c kBT), as when we have a stretch vibration of a strong bond, then the dominant effect is a shift in the zero-point energy of that vibration. If ho o kBT, with and without lateral interactions, as e.g. for a slightly hindered rotation, then the zero-point energy is small, but there is an entropy effect because the occupation of excited states changes. One can even imagine a situation with ho o kBT without and ho c kBT with lateral

114

Catalysis, 2006, 19, 109–163

interactions: e.g., a rotation that becomes strongly hindered or even impossible. In such a case the lateral interactions decrease the entropy and increase the zeropoint energy. The effect of lateral interactions on kinetics is usually modeled via the energetics, although huge shifts in desorption temperatures as a function of coverage for NH3/Rh(111) have been assigned to entropy effects.52 There is another entropy contribution which should always be included in a kinetic model, and that is the configurational entropy of an adlayer. This entropy derives from the various ways that adsorbates may be distributed over the sites. At high temperatures the adlayer will be disordered because this gives a high configurational entropy. At low temperatures the energy determines the structure, which is then most often ordered with a low entropy. 2.2 Equilibrium Aspects. – Undoubtedly the most obvious manifestation of lateral interactions is the formation of ordered adlayers at low temperatures. Although this paper is not on equilibrium adlayer structures and phase transitions between them, some discussion of them is unavoidable if we want to understand the effect of lateral interactions on kinetics. Island formation typically means that only adsorbates at the edge of an island can react. Structured adlayers generally result from repulsive interactions. They show a reduced reactivity because vacant sites will not become occupied because of this repulsion. Phase transitions show a corresponding change in the kinetics, and may show specific experimental features: e.g., voltammetric experiments can show a peak at the potential where a phase transition takes place (see Section 4.2). We refer the reader to the paper of Patrykiejew et al. for a more extensive discussion of equilibrium phenomena.53 Because it is not necessary to describe the temporal evolution when dealing with equilibrium properties, a larger range of methods is available than for kinetics. The paper of Patrykiejew et al. also discusses off-lattice models, which, as we have mentioned before, we will not do here, because such models have hardly been developed for kinetics. (Exceptions are some kinetic off-lattice model for crystal growth,54–56 and for diffusion of clusters on metal surfaces.57,58) 2.2.1 Island Formation and Segregation. Island formation affects the kinetics because often only the adsorbates at the edges of islands can react with other molecules. (Exceptions are adsorbates that form islands and that can react with each other, and if there is an Eley-Rideal mechanism.) To form islands it seems necessary to have an attractive interaction that keeps adsorbates together. Indeed, if there is just one type of adsorbate an attractive interaction leads to island formation at low temperatures, whereas a repulsive one does not. If there is more than one type of adsorbate, then one can even have island formation if all lateral interactions are repulsive.59–61 The term segregation may be more appropriate for such a situation. To see how repulsive interactions can lead to islands of the same adsorbate let’s assume that we have atoms or molecules A and B, and let’s assume that they all repel each other. If the coverage is low so that all adsorbates can move so far apart that they don’t feel each other, then that is what will happen and

115

Catalysis, 2006, 19, 109–163

the adlayer will have randomly positioned A’s and B’s apart from the fact that they are well separated. If the coverage is increased and it is unavoidable that the adsorbates feel each other’s presence, then the adlayer structure either remains a random mixture or the adsorbates segregate. To show this we look at the limiting case that we have S sites that are all occupied and there are NA A’s and NB B’s. (NA þ NB ¼ S.) We also assume that the sites form a regular grid with each site having Z neighbors. As a model for the lateral interactions we assume that only nearest neighbors interact, although the phenomenon also occurs in more complicated models. The energy of the adlayer is given by E ¼ NAAjAA þ NBBjBB þ NABjAB þ NAeA þ NBeB

(1)

where NXY is the number of X–Y neighboring pairs, jXY the energy of such a pair (jXY 4 0 for all X and Y, because we have only repulsion), and eX the adsorption energy of an isolated adsorbate X. In this expression we can eliminate NAA and NBB by using so-called sum rules 2NAA þ NAB ¼ ZNA,

(2)

2NBB þ NAB ¼ ZNB.

(3)

This gives us 1 1 1 E ¼ NAB jAB jAA jBB þ NA ZjAA þ eA 2 2 2 1 þ NB ZjBB þ eB : 2

ð4Þ

If the numbers of A’s and B’s are fixed, then a structural change affects only the first term. If jAB o 12(jAA þ jBB), then a minimal energy is obtained if NAB is maximal. If jAB 4 12(jAA þ jBB), then the system wants NAB to be as small as possible. Put differently, if the repulsion between an A and a B is larger than the average of the repulsion between two A’s and the repulsion between two B’s, then the A’s and B’s will form islands. If not, they will mix. If there are more than two adsorbate types more complicated situations can arise. An example is given in Section 4.1.59 2.2.2 Ordered Adlayers. Lateral interactions for chemisorbed adsorbates are generally repulsive, although there have been reports of attractive interactions for adsorbates at distances of about twice the distance between neighboring sites.62,63 At low coverage repulsive interactions simply keep the adsorbates apart. The adlayer is disordered. At intermediate and high coverages a rich field of possible adlayer structures can be observed even when there is only one type of adsorbate. If the coverage is increased in the disordered low coverage structure the system will attempt to avoid the repulsive interactions as long as possible. The adlayer becomes less and less mobile and eventually ends up in a rigid structure with a larger unit cell than the substrate, but one in which the adsorbates do not

116

Catalysis, 2006, 19, 109–163

feel each other. For example, the sites of many atoms on a (100) surface of an fcc metal form a square grid and there is a strong repulsion between atoms on neighboring sites. As a consequence there is a c(2 2) structure (checkerboard structure), in which none of the adsorbates has a neighbor, but no adsorbate can be added or moved without having to pay the penalty of a strong repulsion.64,65 Avoiding some lateral interactions also occurs at other surfaces and if there are lateral interactions between adsorbates farther apart.66–68 Structures that are obtained by avoiding all repulsive interactions are found in so-called ‘‘hard’’ models.69 They already show quite some remarkable structures (see also Section 4.2), but real systems show even more diversity for three reasons. First, the lateral interactions are generally more complicated then just one interaction between pairs of adsorbates. A pair of adsorbates may affect each other differently depending on the distance between them, and lateral interactions need not be limited to pair interactions (see e.g. Sections 4.1 and 4.4).25,63,70 Second, different adsorption sites may play a role. At low coverages some sites will not be occupied, because they are energetically unfavorable. At high coverage however it may be better to occupy such a site if this avoids paying the higher price of a strong repulsive interaction between two adsorbates (see Section 4.1).33,50,65,71–80 Third, at high coverages the lateral interactions may become so strong that it is more important to minimize the energy of the interactions between the adsorbates than to have a good interaction with the substrate. The concept of sites is then no longer useful, and adlayer structures that are incommensurate with the substrate can be formed.53 This review will not discuss this last situation. 2.2.3 Phase Transitions. Island formation and ordered adlayers are found at low temperatures. At higher temperatures thermal motion can overcome the lateral interactions and lead to disordered structures. For kinetic experiments the transitions between structured and the disordered phases are important. Let’s take the example of atoms on a (100) surface of an fcc metal with strong nearestneighbor repulsion of Section 2.2.2 again. If we do a temperature-programmed desorption experiment starting with all sites occupied, we see the following. Initially we have a (1 1) structure. The adsorbates repel each other so desorption will take place at low temperature. If half of the adsorbates have desorbed, the remaining ones will form a c(2 2) structure, in which the adsorbates feel no repulsion. These adsorbates will desorb at a much higher temperature. So the temperature-programmed desorption spectrum will show two peaks: the first corresponds to a change from a (1 1) to a c(2 2) structure. The second corresponds to a change from c(2 2) to a bare substrate.81 During desorption the adlayer is disordered. Desorption starts and stops at ordered structures. Figure 3 shows schematically how such a system moves through its phase diagram. Note that this diagram shows that there are different levels of disorder. Although ‘‘(1 1) (bare) þ c(2 2)’’ and ‘‘(1 1) (full) þ c(2 2)’’ in the figure look disordered, the occupation of neighboring sites is still very low. At higher temperatures this order disappears. Peaks with similar origins can be observed in voltammetric experiments (see Section 4.2).

117

Catalysis, 2006, 19, 109–163

Temperature

disordered phase

(1x1) (full) + c(2x2) (1x1) (bare) + c(2x2)

c(2x2) 0

0.2

0.4

0.6

0.8

1

Coverage

Figure 3 Sketch of an example of the evolution of a system during a temperatureprogrammed desorption experiment in the system’s phase diagram. The fat line indicates the change of the temperature and coverage during the experiment, and the thin lines indicate the phase transitions (see text). The snapshots below the order-disorder transition line are taken during a simulation of the experiment. The coverages are 0.3, 0.5, and 0.7 ML. The snapshots above the order-disorder transition line show adlayers of 0.3 and 0.7 ML at high temperatures

2.2.4 Aspects of (Dis)Order. While discussing structured adlayers we have only looked at ways to minimize energy. We also mentioned the configurational entropy in Section 2.1.6. At high temperature this entropy becomes more important than the energy. As a consequence at a certain temperature there will be a transition from a low-temperature ordered to a high-temperature disordered phase. Many catalytic processes are done at high temperature. It is tempting to assume that the adlayer is then in the disordered phase and that the adsorbates are randomly distributed over the sites. This would simplify the kinetics enormously, as one can derive a set of simple rate equations for the coverages (see Section 3.1.2).82 One can even include lateral interactions in the description.83–85 There are two reasons, however, why one should hesitate to make the assumption of randomly distributed adsorbates. First, lateral interactions can be quite large and the order-disorder temperature may be much higher than expected. For example, the interaction between two CO molecules on neighboring sites on Rh(100) was determined to be about 24 kJ/mol (see Section 4.3).13,86 This gives an order-disorder transition temperature of about 1450 K. One can argue that such a strong repulsion will be avoided and that otherwise the interaction between CO molecules is small. This is indeed correct, but still the strong repulsion of nearest-neighbor adsorbates causes a strong correlation

118

Catalysis, 2006, 19, 109–163

for the occupation of neighboring sites, which makes the assumption that the adsorbates can be randomly distributed untenable. This is obvious for CO coverages above 0.5 ML. A second, more subtle, point is that even in a disordered phase there is a correlation in the occupation of two sites that have adsorbates that feel a mutual interaction. It is this correlation that is important for the kinetics, and not the order or disorder of the adlayer. The structures that we have discussed are characterized by a certain longrange order, which makes them identifiable by diffraction techniques. This long-range order is not what is important for the kinetics however. Suppose we look at desorption in O/Pt(111), and for the moment we ignore lateral interactions. The oxygen atoms preferentially adsorb at fcc sites, which form a hexagonal grid, and an exact expression for the rate of desorption is82 dhOi ¼ 6Wdes hOOi: dt

ð5Þ

Here hOi is the probability that a site is occupied by an oxygen atom (¼yO), hOOi is the probability that two neighboring sites are both occupied by oxygen, t is time, Wdes is the rate constant for desorption, and the factor 6 is the coordination number. For desorption in CO/Rh(100) including nearest-neighbor interactions we get the exact rate equation X X2 dhCOi ¼ WX1 ;X2 ;X3 ;X4 X1 CO X3 : X4 dt X ;X ;X ;X 1

2

3

ð6Þ

4

The X’s stand either for CO or a vacant site. The quantity with the angular brackets on the right-hand-side stands for the probability that there is a CO on a site with X1, X2, X3, and X4 on the four neighboring sites. The rate constant W for desorption depends on the occupation of these sites. An important point that we want to make here is that the probabilities on the right side of eqns. (5) and (6) can only be reduced to normal coverages, and the expressions can only be changed to normal rate equations, at the expense of making approximations. For the examples that we will discuss in this review such approximations are not good. It is not clear to us when such approximations are valid. They are implicitly made when rate equations are used for kinetic modeling. (Alternatively, one can say that kinetic modeling with rate equations is a phenomenological approach.) Recent experiments seem to indicate however that in general such approximations are flawed.6,25,87–93 That the multiple-site probabilities in eqns. (5) and (6) cannot be reduced to normal coverages means that the occupations of neighboring sites are correlated, or one can say that there is short-range order. This order should be distinguished from the long-range order that characterizes structural adlayers.94 If we have a high-temperature disordered structure, then we do not have long-range order. If we have lateral interactions, there will still be some short-range order however. We will discuss CO desorption from Rh(100) at high temperature more fully in Section 4.3. We will look at situations where the relevant lateral interactions are quite small, and not able to form an ordered

119

Catalysis, 2006, 19, 109–163

structure, but the effect on the kinetics is still obvious in the experiment, and may change the rate constant for desorption. The dominant effect of the lateral interactions is to change the activation energy for desorption. This leads to higher desorption rates even if there is no short-range order. CO/Rh(100) shows a very strong repulsion for CO molecules at nearestneighbor positions and much weaker repulsion between molecules farther apart. The occurrence of different lateral interactions is typical for real systems and leads to complicated order-disorder phenomena, especially if there are different types of adsorbates. Thermal motion of the adsorbates may overcome some lateral interactions but not others. The phase diagram can be quite complex even if there is only one type of adsorbate, and the effect on the kinetics can be profound. 2.3 Effect of Lateral Interactions on the Kinetics. – The effect of lateral interactions on kinetics has been studied for adsorption,1,95 desorption,19,67,68,89,92,96,97 diffusion,8,10,11 reactions,98–101 including ones that show oscillations and chaos,85 and hopping cascades.102 The discussion in Section 2.1 has shown that the presence of another adsorbate can change the adsorption energy of an adsorbate either by a direct interaction or by changing the geometric and electronic structure of the substrate. A decrease in adsorption energy makes desorption easier. An increase makes it harder. Changes in adsorption energy may also affect the reactivity of the adsorbate. If a reaction has an early activation barrier (i.e., a transition state that is close to the initial state of a reaction), then lateral interactions will influence the transition state about as much as the initial state, and the activation energy will remain the same. If a reaction has, however, a late barrier, then lateral interactions will have a different effect on the transition state than on the initial state, and the activation energy will change.103,104 The previous sections have dealt with equilibrium situations: i.e., minima of the (free) energy. For the kinetics we also need to know how lateral interactions affect transition states. There has hardly been any work done on this.105 From a theoretical point of view one can in principle use quantum chemical calculations just as one would for the stable states (see Section 3.4). The kinetic experiments of Section 3.3.3 depend on the activation energies and on the difference between the lateral interactions in the transition state and the initial state of a reaction. The experiments of Sections 3.3.1, 3.3.2 and 3.3.4 do not yield any information on the effect of lateral interactions on transition states. In our simulations we use the following pragmatic approach. For a reaction we have a model of the lateral interactions that tells us how the energies of the initial and the final states (both minima) depend on the lateral interactions. We then use the Brønsted-Polanyi relation to relate the shifts in the initial and final state to a change in the activation energy106,107 ð0Þ

Eact ¼ Eact þ aðDE DE ð0Þ Þ:

ð7Þ

120

Catalysis, 2006, 19, 109–163 barrier:

early

late

reactant(s) change in environment

product(s)

Bronstedcoefficient:

0.0

0.5

1.0

Figure 4 Sketches showing how changes in the lateral interactions affecting the reactants or the products change the activation energy (vertical arrows) depending on whether the barrier is early, late, or in between. The thin line indicates the reaction profile without lateral interactions, the fat line with interactions

Here Eact (E(0) act) is the activation energy with (without) lateral interactions, and DE (DE(0)) is the reaction energy with (without) lateral interactions. DE o 0 for exothermic and DE 4 0 for endothermic reactions. The Brønsted-Polanyi coefficient a varies between 0 and 1, and is a measure of how much the effect of the lateral interactions of the initial and final states influences the activation energy. The idea of the Brønsted-Polanyi relation is that if we have a transition state that resembles the initial state (a so-called early barrier), then lateral interactions will affect the transition state as much as the initial state, and the activation energy will not depend on the lateral interactions. So we choose a ¼ 0 in that situation. If the transition state resembles the final state (a so-called late barrier), then changes in the lateral interactions of the initial and final states will fully end up in the transition state, and we choose a ¼ 1 (see Figure 4).

3

Theory

The main progress in the theoretical study of the effect of lateral interactions on the kinetics of surface reactions in recent years has been in kMC simulations and in density-functional theory (DFT) calculations of lateral interactions. We deal with the latter in Section 3.4, and with the former in Section 3.1. Section 3.2 deals with analytical expressions to describe lateral interactions, and Section 3.3 deals with the experimental determination of lateral interactions. It may be surprising to put this in a ‘‘Theory’’ section, but, as we will see, it is usually necessary to do a simulation of the experiment to be able to say anything about

Catalysis, 2006, 19, 109–163

121

lateral interactions, which is why we have put this here. We would like to refer the reader to the excellent works of Lombardo and Bell for research prior to 1991,108 and of Zhdanov for a more extensive treatment of the analytical methods from Section 3.1.109 3.1 Including Lateral Interactions in the Kinetics. – We will discuss in this section the most common methods to study kinetics with lateral interactions. There are other analytical approaches that are occasionally used as well: e.g., transfer matrices.110–112 We skip them here as they have been not been applied to many systems. 3.1.1 The Lattice-Gas Model. If we have a single-crystal surface and the molecules are not too large, we can define positions, called (adsorption) sites, where the molecules adsorb. Different molecules may adsorb at different sites, but for single-crystal surfaces these sites always form a two-dimensional grid or lattice with a translation symmetry that is the same as that of the surface. If the coverage (i.e., the number of molecules) is not too large, then each molecule can also be regarded as being at a particular site. This is the situation that we will deal with here. At high coverages the molecules may push each other away from the adsorption sites. Figure 5 shows some examples of single-crystal surface and different lattices formed by particular adsorption sites. In molecular dynamics and also some

Figure 5 Examples of single-crystal surfaces, adsorption sites, and the corresponding lattices. The top row shows a (100) surface and the bottom row a (111) surface of an fcc metal. On the left two layers of the fcc metal are shown. Top sites for the (100) and hcp sites for the (111) surface are added in the middle. The intersections of the lattices on the right correspond to these sites

122

Catalysis, 2006, 19, 109–163

forms of Monte Carlo the position of atoms are precisely specified.113 It turns out that for surface reactions it is much more efficient to use a cruder model in which for each adsorbate only the site is specified at which it is adsorbed. Such a model is called a lattice gas model. It consists of a lattice or grid with each lattice or grid point having a label which tells us which adsorbate is found at the corresponding site. We will describe in Section 3.1.5 how this slight sacrifice in positional accuracy allows us to do simulations with time scales that correspond to those of real kinetic experiments (seconds or longer), whereas a method like molecular dynamics is generally restricted to nanoseconds. 3.1.2 The Mean-Field Approximation. The rate of a reaction when there are lateral interactions does not only depend on the reactants and temperature, but also on the occupation of the sites surrounding the sites where the reactants are found. As a consequence exact reactions rate equations contain probabilities of the occupation of clusters with many sites. We have already seen this for CO desorption in eqn. (6). To use this equation we have to express the 5-site probability on the right-hand-side in terms of 1-site probabilities (i.e., the coverages). The simplest way to do this is to approximate a multisite probability as a product of 1-site probabilities. This is called a meanfield approximation.94,114,115 For the 5-site probability in eqn. (6) this would mean

X2

X1 CO X3 X4

¼ hX1 ihX2 ihX3 ihX4 ihCOi:

ð8Þ

Here hXni is the probability of having an X on site n. If this does not depend on the site, then this is nothing but the coverage of X. With eqn. (8) the rate equations become a closed set that can be solved at least numerically. If the lateral interactions are pair interactions, then the right-hand-side of the rate equations can be simplified. We show this for the CO desorption, but similar simplifications are possible for more complicated reactions systems as has been shown by Makeev.83–85 The rate constant in eqn. (6) with pair interactions can be written in an Arrhenius form Eact ðX1 ; X2 ; X3 ; X4 Þ WX1 ;X2 ;X3 ;X4 ¼ n exp kB T

ð9Þ

with ð0Þ

Eact ðX1 ; X2 ; X3 ; X4 Þ ¼ Eact þ jX1 ;CO þ jX2 ;CO þ jX3 ;CO þ jX4 ;CO :

ð10Þ

Here E(0) act is the activation energy for an isolated CO molecule and jXn ;CO is the pair interaction between CO and Xn. (Note that Xn is either a CO molecule or a vacant site, in which case the interaction equals zero.) With this expression for

123

Catalysis, 2006, 19, 109–163

WX1 ;X2 ;X3 ;X4 eqn. (6) becomes " # ð0Þ X dhCOi Eact ¼ n exp hCOi ejX1 ;CO =kB T hX1 i dt kB T X ;X ;X ;X 1

2

3

4

jX ;CO =kB T 3 hX3 i ejX4 ;CO =kB T hX4 i ejX2";CO =kB T hX # 2i e ð0Þ

¼ n exp

Eact hCOi½1 þ ðejCO;CO =kB T 1ÞhCOi4 : kB T

ð11Þ

We have used here that Xn is either CO or a vacant site * and hCOi þ h*i ¼ 1. The main problem with this mean-field approximation is that at temperatures below the order-disorder transition there is a strong correlation between the occupations of sites. As will be shown in Section 4.3 jCO,CO for neighboring sites is about 24 kJ/mol, which corresponds to a thermal energy at 2880 K. This means that neighboring sites will not be occupied simultaneously at any realistic temperatures, and a more sophisticated approach is needed that describes this well. For weak interactions eqn. (11) may be acceptable, but one should be aware that there is some correlation in the occupation of sites even if there is no long-range order. 3.1.3 The Quasi-Chemical Approximation. The mean-field approximation ignores all correlation in the occupation of neighboring sites. This is incorrect when there is a strong interaction between adsorbates at such sites. The simplest way to include some correlation is to work with probabilities of occupations of two sites hXYi instead of one site hXi. Approximations that do this are generally called pair approximations (not to be confused with pair interactions). There are more possibilities to reduce multi-site probabilities as in eqn. (8) to 2-site probabilities than to 1-site probabilities. This leads to different types of pair approximations. The best-known approximation that is used for Ising models is the Kirkwood approximation, which uses for example115–117 hXYZi ¼

hXYihXZihYZi hXihYihZi

ð12Þ

for 3-site probabilities. This can be extended to probabilities of more than three sites.118,119 The drawback of the Kirkwood approximation for describing reactions is that the number of atoms is not conserved (see below and Section 3.1.4). The quasi-chemical approximation does not have this drawback. We will derive it here in a manner that has inspired its name, but there are other derivations possible.109 Suppose we have a pair of neighboring sites that each are either vacant or occupied by an adsorbate A. If we have two such pairs we can write AA þ ** " 2A*.

(13)

124

Catalysis, 2006, 19, 109–163

This equilibrium has an equilibrium constant so that hAAihi hAi2

¼ K:

ð14Þ

The value of K is determined by the energy of the different ways to occupy the pair of sites. Because hXYi is in equilibrium proportional to a Boltzmann factor we get hAAihi EAA þ E 2EA ¼ exp ; ð15Þ kB T hAi2 where EXY is the energy of an XY pair. This energy includes the adsorption energy of individual adsorbates and possibly lateral interactions. Expression (15) does not specify the 2-site probabilities. We also need sum rules. Their general form is X

hxyi ¼ hxi;

ð16Þ

y

where x and y each stand for one or more sites with occupations, and the summation is over all possible ways to occupy the sites of y. For the example related to eqn. (15) the sum rules are hAAi þ hA*i ¼ hAi,

(17)

hA*i þ h**i ¼ h*i.

(18)

For simple cases it is possible to derive analytical expressions for each of the 2-site probabilities in terms of 1-site probabilities or coverages, but in general we have to resort to numerical procedures. So far the quasi-chemical approximation has been shown as a way to deal with 2-site probabilities. For lateral interactions we are usually dealing with probabilities of many more sites (e.g., see the 5-site probability in eqn. (8)). The quasi-chemical approximation becomes then much more cumbersome and is hardly ever used. The approach in Section 3.1.4 presents a way that is straightforward to extend to larger numbers of sites, while it can be made to coincide with the quasi-chemical approximation for 2-site probabilities. 3.1.4 The Maximum-Entropy Principle. The mean-field and the quasi-chemical approximations can be extended to larger clusters. Using the derivation of Section 3.1.3 to obtain a quasi-chemical approximation for a cluster with more sites is quite cumbersome. In this section we present an approach that is new and that unifies various approximations to deal with multi-site probabilities. It is based on the maximum-entropy principle.120 Suppose we have a cluster of n sites with occupations X1, X2, . . ., Xn. We define an entropy X hX1 . . . Xn i lnhX1 . . . Xn i ð19Þ S X1 ... Xn

125

Catalysis, 2006, 19, 109–163

with hX1 . . . Xni the probability of having the occupations X1, . . ., Xn. The maximum-entropy principle states that we should maximize this entropy as a function of the probabilities. If the only restriction on the probabilities is that their sum must be equal to one, then we get hX1 . . . Xn i ¼

1 Nconf

ð20Þ

where Nconf is the number of all possible ways one can put adsorbates on the n sites. This is nothing but the fundamental hypothesis of statistical physics that says that all states are equally likely.94,114 If we have restrictions on the probabilities, then the expressions for the probabilities change. We will discuss the situation that the restrictions are linear equations for the probabilities. Suppose we have X

Am; X1 ... Xn hX1 . . . Xn i ¼ bm

ð21Þ

X1 ... Xn

as restrictions with m ¼ 1,. . ., Nres and Nres the number of restrictions. We can include them in the maximization using Lagrange multipliers.121 This means we define sS

Nres X m¼1

" lm

#

X

Am; X1 ... Xn hX1 . . . Xn i bm ;

ð22Þ

X1 ... Xn

with lm the Lagrange multipliers. We maximize s without restrictions: i.e., we equate the derivatives of s with respect to the probabilities to zero. This yields Nres X @s ¼ lnhX1 . . . Xn i 1 lm Am; X1 ... Xn ¼ 0: @hX1 . . . Xn i m¼1

ð23Þ

Together with the restrictions these are the equations we have to solve to get the probabilities and the Lagrange multipliers. We can rewrite (23) as " hX1 . . . Xn i ¼ exp 1

Nres X

# lm Am; X1 ... Xn :

ð24Þ

m¼1

Substitution of these expressions in the restrictions gives us equations for the Lagrange multipliers only. Solving them can in general not be done analytically, and one has to resort to numerical methods. In some cases the restrictions are such that we can give analytical expression for the probabilities and Lagrange multipliers. One such situation is when one wants to express the probabilities in terms of coverages. This can be

126

Catalysis, 2006, 19, 109–163

accomplished by using the sum rules as restrictions. X hX1 . . . Xn i ¼ hXk i:

ð25Þ

X1 ... Xk1 Xkþ1 ... Xn

The quantity on the right-hand-side is the probability that site k is occupied by an adsorbate Xk, which is nothing but the coverage of Xk for site k. There is such a restriction for each site in the cluster and each possible occupation. Following the same derivation as above gives us " # X ðkÞ hX1 . . . Xn i ¼ exp 1 lX k ð26Þ k ðkÞ lX k

where is the Lagrange multiplier for the restriction of site k and occupation Xk. The solution for the restrictions and the probabilities is given by hX1 . . . Xn i ¼ hX1 i . . . hXn i

ð27Þ

1 ðkÞ lXk ¼ lnhXk i n

ð28Þ

for the probabilities and

for the Lagrange multipliers. (The solution for the Lagrange multipliers is not unique.) To see that the restrictions are indeed fulfilled note that X hXm i ¼ 1 ð29Þ Xm

for all m. This expression means that a site must have one of the possible adsorbates or be vacant. We can recognize eqn. (27) as a mean-field expression. Reducing a multi-site probability to a probability of fewer sites is called a decoupling scheme. They have been used extensively in equilibrium statistical physics for models like the Ising model.94,114,115 The difference with models of surface reactions is that we have sum rules like eqn. (25), which conflicts with some of the decoupling schemes from equilibrium statistical physics: e.g., the Kirkwood approximation mentioned before is not consistent with sum rules. Valid decoupling schemes beyond the mean-field approximation can be derived in the same way as we have derived the mean-field approximation above. Suppose we have three sites with sites 1 and 2 and sites 2 and 3 being neighbors, whereas sites 1 and 3 are not. We want to retain the correlation in occupation of the neighboring sites so we reduce the 3-site probabilities only to 2-site probabilities. We use the sum rules X hX1 X2 X3 i ¼ hX1 X2 i ð30Þ X3

127

Catalysis, 2006, 19, 109–163

and X

hX1 X2 X3 i ¼ hX2 X3 i:

ð31Þ

X1

With sS

X

ð1;2Þ lX 1 X 2

X1 X2

X

"

ð2;3Þ

lX 2 X 3

X2 X3

" X X3

X

# hX1 X2 X3 i hX1 X2 i #

hX1 X2 X3 i hX2 X3 i

ð32Þ

X1

we get @s ð1;2Þ ð2;3Þ ¼ lnhX1 X2 X3 i 1 lX1 X2 lX2 X3 ¼ 0: @hX1 X2 X3 i

ð33Þ

The solution of eqns. (30), (31), and (33) is hX1 X2 X3 i ¼

hX1 X2 ihX2 X3 i ; hX2 i

ð34Þ

1 1 ð1;2Þ lX1 X2 ¼ lnhX1 X2 i þ lnhX2 i; 2 2

ð35Þ

1 1 ð2;3Þ lX2 X3 ¼ lnhX2 X3 i þ lnhX2 i: 2 2

ð36Þ

and

(Note again that the solution of the Lagrange multipliers is not unique.) The result (34) can be generalized as follows. Suppose we have an n-sites cluster with site 1 being a neighbor of only site 2 and we have the sum rule X hX1 X2 X3 . . . Xn i ¼ hX1 X2 i; ð37Þ X3 ... Xn

then hX1 X2 . . . Xn i ¼

hX1 X2 ihX2 . . . Xn i : hX2 i

ð38Þ

It is possible to repeat this procedure with the (n – 1)-site cluster of sites 2 to n. Suppose site 3 is also only a neighbor of site 2, then hX1 . . . Xn i ¼

hX1 X2 ihX2 X3 ihX2 X4 X5 . . . Xn i hX2 i2

;

ð39Þ

128

Catalysis, 2006, 19, 109–163

or suppose site 3 is only a neighbor of site 4, then

hX1 . . . Xn i ¼

hX1 X2 ihX3 X4 ihX2 X4 X5 . . . Xn i : hX2 ihX4 i

ð40Þ

These expressions, and also eqn. (34) make sense because hXkYli/hXli equals the conditional probability that there is an Xk at site k if there is an Xl at site l. Such an easy interpretation is not possible if we have e.g. three sites that are pairwise neighbors as for a hexagonal grid. Eqn. (34) and similar equations like (38), (39), and (40) have been used before by Mai et al. in models of CO oxidation.119 The maximum-entropy principle that we have used above does not include any effects of lateral interactions. We can include them as well if we extend to definition of the entropy in eqn. (19) to that of a free energy F

X

EX1 ...Xn hX1 . . . Xn iþkB T

X1 ...Xn

X

hX1 . . . Xn i lnhX1 . . . Xn i;

ð41Þ

X1 ...Xn

with EX1 ...Xn some energy that is a function of the occupations. The idea is to minimize this free energy as a functions of the probabilities. (Note that with respect to the definition (19) the entropy term has a different sign in eqn. (41). Therefore we maximize S and minimize F.) If there are no restrictions on the probabilities, except for the one that the sum of all probabilities should be one, we get EX1 ... Xn hX1 . . . Xn i / exp : kB T

ð42Þ

We see that this is nothing but a Boltzmann factor. The proportionality constant in this expression can be obtained by normalization of the probabilities. As with the maximum-entropy principle we are more interested in the situation where we have other restrictions. Suppose we look again at restrictions that are given by eqn. (21). Introducing Lagrange multipliers gives us " # Nres X X lm Am;X1 ... Xn hX1 . . . Xn i bm : fF ð43Þ m¼1

X1 ... Xn

We minimize f without restrictions. This yields @f ¼ EX1 ... Xn þ kB T½lnhX1 . . . Xn i þ 1 @hX1 . . . Xn i Nres X lm Am; X1 ... Xn ¼ 0: m¼1

ð44Þ

129

Catalysis, 2006, 19, 109–163

Together with the restrictions these are the equations we have to solve to get the probabilities and the Lagrange multipliers. Rewriting (44) gives us hX1 . . . Xn i ¼ e

1

Nres EX1 ... Xn Y lm Am; X1 ... Xn exp exp : kB T kB T m¼1

ð45Þ

Substitution of these expression in the restrictions gives us equations for the Lagrange multipliers only. Solving them can in general again not be done analytically, and one has to resort to numerical methods. Analytical expressions for the probabilities may be obtained from the maximum-entropy principle, but it may be necessary to make additional assumptions. For example, eqn. (25) is not sufficient to get the mean-field approximation. This can be seen as follows. Instead of eqn. (26) we get " ðkÞ # lX k EX1 ... Xn Y 1 exp hX1 . . . Xn i ¼ e exp : ð46Þ kB T kB T k Of course eqn. (27) would still fulfill the restrictions, but such a product of factors, each of which depending on one site, is not consistent with eqn. (46), because of the factor with EX1 ...Xn . Only when this energy is a sum of contributions of the individual sites do we get eqn. (27) again. Equation (46) does lead to another approximation that we have seen before. Suppose we have just two sites, n ¼ 2, and a site can only either be an adsorbate A or be vacant. Then (46) becomes " ð1Þ # ð2Þ lX 1 þ lX 2 EX1 X2 1 hX1 X2 i ¼ e exp exp ; ð47Þ kB T kB T where X1 and X2 are either A or *, with * indicating a vacant site. Instead of substituting this expression in eqn. (25) to determine the Lagrange multipliers, we can use it to derive hAAihi EAA þ E 2EA ¼ exp : ð48Þ hAihAi kB T The Lagrange multipliers have vanished because there are equals numbers of A’s and *’s in the numerator and denominator. The equation is identical to what we have obtained in Sections 3.1.3 for the quasi-chemical approximation. There are however two differences. First, we have another way to determine the 2-site probabilities. Instead of using eqn. (48) or (15) we can use the equation we get by substituting (47) in the restriction (25). Second, this way of determining the 2-site probabilities insures automatically that the solutions fulfill the sum rules. It is quite easy to see how we can extend the procedure in the previous paragraph to larger clusters. Such large cluster can either be found on the

130

Catalysis, 2006, 19, 109–163

left- or the right-hand-side of sum rules. To get equations similar to (48) from the equations corresponding to (47) we just have to insure that the same Lagrange multipliers (type and number) occur in numerator and denominator. This then leads to a quasi-chemical approximation for clusters with more than two sites. 3.1.5 Kinetic Monte Carlo Simulations. The approximations of the previous sections generally make it rather easy to interpret the results of a kinetic model. Their drawback is that it is very difficult to assess their accuracy. KMC simulations do not have this drawback. For a given reaction model the results of a kMC simulation are exact. There are three ways to derive kMC simulations. Originally kMC simulations were presented in the form of an algorithm that described how the adlayer, represented by a lattice-gas model, evolved. The changes were of course nothing but the reactions and diffusional hops of adsorbates, and each of these was described explicitly in the algorithm. Instead of reaction rate constants one used reaction probabilities. Consequently time was given in Monte Carlo steps. Such a step typically consists of one attempt at each site to do each reaction. Although it is possible to do good simulations in this way, it is inconvenient if one wants to make quantitative comparisons with experiments where one has rate constants, rates, and time in seconds. Nevertheless this way of doing kMC simulations has been successfully applied, and is still being applied, to the study of the effect of lateral interactions mainly on desorption.122–129 The second way to derive kMC was presented by Fichthorn and Weinberg.130 They have shown when it is possible to interpret equilibrium Monte Carlo simulations dynamically. This leads then to rate constants and kMC simulations in real time. The method has also been extended to deal with lateral interactions.81,131 Although these rate constants have been calculated using expressions from transition-state theory,132 it has, as far as we know, never been proven that this is correct. The method has also not been derived for situations in which the rate constants depend on time: e.g., when temperature is changed during an experiment. This is always treated by assuming that temperature is piecewise constant. In spite of these theoretical shortcomings, the work of Fichthorn and Weinberg is generally seen as the work that has led to present-day kMC simulations. The third way derives the kMC simulations from the master equation133 dPa X ¼ ½Wab Pb Wba Pa ; dt b

ð49Þ

where t is time, a and b are configurations of the adlayer, Pa and Pb are their probabilities, and Wab and Wba are rate constants that specify the rate with which the adlayer changes due to reactions. This equation can be derived from first principles in a manner that is very similar to the derivation of the expressions of variational transition-state theory.134–137 The difference is that in variational transition-state theory phase space is partitioned into a reactant

Catalysis, 2006, 19, 109–163

131

and a product part, whereas the master equation is obtained by partitioning phase space into regions that each correspond to an adlayer configuration.82,138,139 Monte Carlo methods to solve this master equation have been known for some time.140 It is also possible to use stochastic methods that have been developed to solve reaction rate equations,141,142 and which are normally called dynamic Monte Carlo simulations.143 The advantage of the third way is that the derivation of the master equation shows that the use of expressions of transition-state theory to calculate the rate equations is often correct. An exception is when the potential-energy surface is very flat, and it is important to know what the size of the region in phase space is that corresponds to a configuration.82 Another advantage is that it is also possible to solve the master equation analytically. This generally involves approximations. In this way kMC simulations can be related directly to methods like mean field: in particular, the relation between the rate constants in kMC simulations and in the analytical theories is fixed by theory. Finally, the master equation also holds for time-dependent rate constants. It has been possible, for example, to show that there are situations in which there is a finite probability that same molecules will never react even though the rate constants are not zero.144 Lateral interactions are also straightforward to include as we will show in Section 4. The master equation is also often used in statistical physics as a mathematical model to describe the evolution of systems ranging from sand piles, to forest fires, and earth quakes.145–149 All kMC algorithms generate an ordered list of times at which a reaction takes place, and for each time in that list the reaction that occurs at that time. A kMC simulation starts with a chosen initial configuration. The list is traversed and changes are made to the configuration corresponding to the occurring reactions. There are many algorithms that can be used to determine these reactions and their times.150 The various algorithms differ in how the reaction times are computed, how a reaction of a particular type is chosen, and how it is determined where on the surface a reaction takes place. An analysis of how computer time and memory scales with system size shows that there are only a few algorithms that are really useful.82,150–152 If there are lateral interactions, then there are many reactions with different rate constants. If there are S sites on which there may be adsorbates that affect the rate of a reaction, and each of these sites can be occupied in N different ways (vacant or by one of N – 1 different adsorbates), then there are NS different environments for a reaction. This need not mean that there are as many different rate constants, but the number of different rate constants certainly can become huge. Algorithms that determine the type of reaction (different rate constant means a different type of reaction in this context), and the place where such a reaction takes place separately will be inefficient, because there probably is only one place on the surface where a particular type of reaction can take place, which will take some time to find. Separating the determination of the time for the next reaction and which reaction actually takes place is often also inefficient. This is done by oversampling, i.e., assuming that all reactions have the same rate constant and then accepting a reaction with a probability that is equal to the ratio of its

132

Catalysis, 2006, 19, 109–163

actual rate constant and this assumed rate constant.82,150,151 Unless the lateral interactions are very small compared to the thermal energy, this probability is small, which makes the algorithm inefficient. This leaves us with the first-reaction method.82,141,151 The algorithm is as follows. The First Reaction Method (FRM) 1. Initialize Generate an initial configuration a. Set the time t to some initial value. Make a list Lrx containing all reactions. Generate for each reaction a - b in Lrx a time of occurrence tba ¼ t

1 ln r Wba

ð50Þ

with Wba the rate constant for the reaction and r a random deviate on the unit interval. Choose conditions when to stop the simulation. 2. Reaction Take the reaction a - a 0 with ta 0 a r tba for all b. If the reaction is enabled go to step 3. If not go to step 4. 3. Enabled update Change the configuration to a 0 . Change time to t ¼ ta0 a Remove the reaction a - a 0 from Lrx. Add new enabled reactions to Lrx and generate for each reaction a 0 - b a time of occurrence tba0 ¼ t

1 ln r: Wba0

ð51Þ

Skip to step 5. 4. Disabled update Do not change the configuration: a 0 is the same configuration as a. Remove the disabled reaction from Lrx. 5. Continuation If the stop conditions are fulfilled then stop. If not set a to a 0 and repeat at step 2. Here an enabled reaction is a reaction that is possible and has always been possible since it was put in the list of all reactions Lrx. A reaction is disabled if it

133

Catalysis, 2006, 19, 109–163

is not enabled. Disabled reactions are found in Lrx because they were possible at some time, but before they actually occurred the configuration changed in such a way that they were no longer possible. The computer time per reaction of this algorithms scales with system size as O(log S) where S is the number of sites in the system. (Note that for all kMC algorithms the total number of reactions in a system is of the order O(S). So for the first-reaction method the computer time for a whole simulation scales as O(S log S).) This logarithmic dependence originates from the datastructures, which are normally trees, that are used to store the reactions and their times.153 3.2 Analytical Expressions for Lateral Interactions. – To deal with the effects of lateral interactions in a whole adlayer it is necessary to have a concise description of the lateral interactions. This can be done in a number of ways, each corresponding to a different model of the lateral interactions. Such a model needs then further be specified by assigning values to the parameters in the model. This and the following sections discuss some models for the lateral interactions and experimental and theoretical methods to determine values for the interactions parameters. Analytical expressions for interactions can be regarded either as mathematical in origin or with a physical origin.154,155 The former have often a simple form and they hold for many if not all systems, but they may have many parameters. The latter are based on the mechanism that leads to the interactions, they have seldom a simple mathematical form, they generally hold only for a limited set of systems, but they may have only few parameters.156 An expansion in terms two-, three-, four-, and more-particle interactions (see Sections 3.2.1 and 3.2.2) is an example of a mathematical model for lateral interactions. As an example of a physical model we only present the bondorder-conservation model in Section 3.2.3 as this is the model that has been used most extensively in kMC simulations. 3.2.1 Pair Interactions. An interaction is something between at least two particles. So the simplest model is one that assumes that all interactions can be written as a sum of interactions between a pair of adsorbates. The adsorption energy of an adlayer can then be written as Eads; adlayer ¼

X n

ð0Þ

Eads ðnÞ þ

1X jðn; mÞ: 2 n;m

ð52Þ

Here E(0) ads(n) is the adsorption energy of adsorbate n in the absence of other adsorbates, and j(n, m) is the pair interaction between adsorbates n and m. The summations are over all adsorbates. The factor 12 corrects for double counting of pair interactions (j(n, m) and j(m, n) are the same interaction), and we set j(n, n) ¼ 0 for all n per definition.

134

Catalysis, 2006, 19, 109–163

If there is only one type of adsorbate, then it may be more convenient to use the adsorption energy per adsorbate X ð0Þ cn jn ð53Þ Eads; adsorbate ¼ Eads þ n

E(0) ads

where is the adsorption energy of an isolated adsorbate, jn is the pair interaction of type n, and cn is the number of such pair interactions per adsorbate. By a type of pair interaction we mean an interaction between adsorbates at a particular distance, in a particular direction, or with any other characteristic that may affect the interaction between them. Models with only pair interaction have only few parameters. It is not clear how appropriate a model with only pair interactions can be. For physisorbed systems with dominating direct interactions between the adsorbates such a model may be valid. For adsorbates on a transition metal surface (the kind of system we are here interested in) a through-the-surface mechanism probably dominates. It seem that such a mechanism would necessitate a more sophisticated model. Nevertheless, for the examples that we will discuss we have found that pair interactions are often quite satisfactory. 3.2.2 More-Particle Interactions. An extension of eqn. (52) is to use three-, four-, and more-particle interactions. X ð0Þ 1X 1X Eads; adlayer ¼ Eads ðkÞ þ jðk; lÞ þ jðk; l; mÞ 2 k;l 3! k;l;m k ð54Þ 1 X þ jðk; l; m; nÞ þ . . . : 4! k;l;m;n It is possible to show that such an expansion can be made as accurate as desired by including sufficient terms,154 but one rather has as few terms as possible, of course. If we have one type of adsorbate the adsorption energy per adsorbate can again be written as eqn. (53) provided that the summation extends not just over pair interactions, but also contains multiple-particle interactions. The coefficients cn stand for the number of such interactions per adsorbate. A major advantage of the expansion of the adsorption energy in terms of E(0) ads, pair, and multiple-particle interactions is that the adsorption energy depends linearly on the interactions parameters. This means that a determination of these parameters involves only equations of linear algebra. 3.2.3 Bond-Order-Conservation. Bond-order-conversation (BOC) is based on the observation that when an atom forms bonds with other atoms the energy gain is high for the first bond that is formed but then decreases with each subsequent bond. BOC is an attempt to quantify this observation.157 The bond order x in BOC is defined as the following function of the bond length R. x ¼ exp[a(RR0)].

(55)

135

Catalysis, 2006, 19, 109–163

Here R0 is the equilibrium bond length. The coefficient a, not to be confused with the Brønsted-Polanyi coefficient, is determined by writing the potentialenergy surface as V(R) ¼ Dex(x2)

(56)

with De the dissociation or bond energy. We see that this is a Morse potential. The coefficient a determines the curvature of the potential in the minimum, and can therefore be determined by assuring that the potential gives the correct vibrational stretch frequency of the bond. Note that if an atom forms only a single bond, then R ¼ R0, x ¼ 1, and V(R) ¼ De. The central assumption of BOC is that if an atom forms multiple bonds that X xn ¼ 1 ð57Þ n

holds, where the summation is over all these bonds and xn is the bond order of bond n. If there are N equivalent bonds with the same bond order, then xn ¼ 1/ N and the combined energy of all bonds is De(21/N). If we calculate how much each new bond that is formed decreases the energy we find that the Nth bond decreases the energy by De/(N(N1)). We see that already the second bond lowers the energy only by half the amount of the first bond. It is not necessary to restrict ourselves to bonds that are described by Morse potentials. We can regard eqn. (56) as a quadratic equation in x, use any form of the potential energy V(R) with the usual shape (i.e., a minimum, a repulsive barrier at short distances, and a monotonical increase at large distances), and determine x to get another definition of the bond order. This is called the unity bond index–quadratic exponential potential (UBI–QEP) method by Shustorovich and Sellers.20 The explanation for the reduced energy of additional bonds is very similar to the through-the-surface mechanism of lateral interactions (see Section 2.1). The first bond changes the electronic structure of the atom. As a consequence the electronic structure is less favorable for forming another bond, so that such a bond yields a smaller energy gain. Because this mechanism is so similar to the one for lateral interactions, it is natural to use BOC or UBI–QEP to describe such interactions. This has been done for kMC simulations first by Lombardo and Bell and more recently by Baranov et al. and by Hansen and Neurock.22–24,126 3.3 Experimental Determination. – To do an actual kMC simulation of a system, all the reaction rate constants need to be known. These rate constants can be obtained from both experiments and from theory. The most common experimental methods are temperature programmed desorption (TPD) experiments, scanning tunneling microscopy (STM), single crystal adsorption calorimetry, and low-energy electron diffraction, but there are others as well.15–17 There are two ingredients in all the methods treated below to determine lateral adsorbate interactions. First, one needs to know how many neighbors an adsorbate has. Second, one needs to estimate the change in binding energy

136

Catalysis, 2006, 19, 109–163

caused by these neighboring adsorbates. From these two pieces of information one can estimate the value of the lateral interaction. Some of the techniques below yield information both on the spatial distribution and the change in binding energy, and can therefore be used solely to determine the lateral interactions. Other techniques only yield information on the spatial distribution or the binding energy. In this case information from different techniques has to be combined to yield the values for the lateral interactions. Values for the lateral interactions can sometimes be directly extracted from the experiments. However, often these values are only obtained by fitting a lattice-gas model to the experimental results. It also needs to be mentioned that the estimates for these kinetic parameters are usually only obtained after assuming a certain model. 3.3.1 Scanning Tunneling Microscopy. One way to get estimates for lateral interactions is STM. Using STM one can determine the statistical distribution of the adsorbates over the surface. This distribution can be converted into a radial distribution function, which in turn can be converted into effective lateral interactions. The accuracy of the value of a certain lateral interaction depends on the number of times it has been observed on the surface. Strongly repulsive interactions (larger than a few times kBT) can therefore not be determined using this method. One has to be able to image the exact position of each adsorbate to obtain the radial distribution function. This requires a low adsorbate mobility, since a full scan typically takes one minute to do. This method is therefore mainly applied to atoms on low-temperature surfaces.7,9–11,158 The consequence of keeping the temperature low to suppress surface mobility is that the absolute value of the repulsive interactions that can be determined is also small, typically less than 10 kJ/mol at room temperature. 3.3.2 Low Energy Electron Diffraction. LEED allows one to study the ordered phases of an adsorbate. One usually compares the different ordered phases found for an adsorbate and the temperature ranges in which they appear. This involves transitions between different ordered phases and disorder-order transitions. These experimental results can then be fitted using a lattice-gas model, thus yielding values for the lateral interactions. Several ordered phases are needed for this method to be of practical use. The study as a function of temperature implies that the adsorbate must not desorb or decompose over this large temperature range. This technique is therefore limited to atoms on surfaces, like oxygen on Ru(0001) and Ni(111).2,159,160 3.3.3 Temperature–Programmed Desorption. Accurate estimates of (differences in) binding energies can be obtained by analyzing temperature programmed desorption traces.4,6,67,68,161,162 The differences in desorption temperatures can be directly related to differences in binding energies. It is, however, in general difficult to relate these differences in binding energies to lateral interactions, since the local adsorbate configuration (the number of adsorbates interacting with the desorbing molecule) is unknown.

Catalysis, 2006, 19, 109–163

137

Lateral interactions can only be determined when the local adsorbate configuration is known. This has been successfully used to extract lateral interactions for CO coadsorbed on a c(2 2)-N adlayer on Rh(100). Because the nitrogen atoms form an fixed arrangement (as verified by LEED), the number of nitrogen neighbors for each CO molecule is known to be four. The difference in desorption energy for the CO molecule could therefore be directly related to a pairwise interaction with the four neighboring nitrogen atoms.89 Instead of changing the temperature it is also possible to determine lateral interactions from isothermal experiments,163 or from multi-isotherm experiments in which the temperature is increased in steps.92,163,164 3.3.4 Single Crystal Adsorption Calorimetry. In temperature programmed desorption the binding (or desorption) energies are calculated using the desorption temperature. In SCAC on the other hand, these binding energies can be directly measured. The differences in binding energies due to lateral interactions can therefore also be accurately determined. The configuration of the adsorbate adlayer is unknown, however, and relating these differences in binding energies to lateral interactions is therefore – just like for TPD – difficult. If the local adsorbate configuration is known or guessed using another technique, then lateral interactions can be extracted. This has amongst others been done for CO on Rh(100) and Pt(111).1,12–14,95,165–167 3.4 Calculating Lateral Interactions. – With computer hardware becoming faster it is possible to do quantum chemical calculations on quite realistic models of adsorbates on transition metal surfaces. With software becoming more user-friendly, it is also getting easier. Such quantum chemical calculations for the systems that we are interested in are very computer intensive calculations. Still they have the major advantage over the determination of lateral interactions from experiments that one knows precisely the system one is dealing with, and one does not have to worry about possible effects of steps, impurities, shortcomings of apparatuses, et cetera. 3.4.1 Density-Functional Theory. Transition metals pose a problem for classical quantum chemical methods like self-consistent field (SCF), perturbation theory, configuration interaction (CI), and variations on these methods,168,169 because of the very large electron correlation. SCF underestimates binding substantially, and post-SCF methods are so expensive for transition metals that one can do a calculation only on models with few atoms. DFT on the other hand is relatively cheap: it is about as expensive as SCF.170 Moreover, with the development of the generalized-gradient approximations it is also reasonably accurate.171–173 A large majority of quantum chemical calculations on surface reactions on transition metals is done using DFT nowadays. DFT calculations come in two variations: cluster and periodic calculations. The difference is in the model of the system. Most calculations of lateral interactions have used periodic calculations,23–25,63,76,79,160,174–177 but a few

138

Catalysis, 2006, 19, 109–163

cluster calculations have been published as well.178 In a cluster calculation the system is modeled by a cluster of transition metal atoms representing the substrate and one or more adsorbates (see Figure 6). Such a model is treated as a large molecule. In a periodic calculation the system is modeled as a crystal with a large unit cell. A slab consisting of a number of layers of metal atoms represents the substrate. Adsorbates are put on the slab so that the slab plus adsorbates forms a periodic structure in two directions. The whole system is then obtained by stacking such slabs, but widely separated so that there is no interaction between them (see Figure 7). The rational of this model is that a plane-wave basis set can be used for the electronic structure. The advantage of such basis is that calculations take much less computer time than a calculation with a localized basis set of the same size.179 Plane waves also have the advantage that forces on atoms are easier to compute, which makes the optimization of the structure of a system, and the computation of vibrations easier. The drawback is that a localized basis set is often a better representation of the electronic structure near the nuclei so that it requires fewer basis functions. Concerning the question of how well a cluster or a periodic model represents a real system, one should be aware of artifacts caused by the edge of a cluster model, because the atoms at the edge can be much more reactive because they are undercoordinated. Periodic calculations can have the problem of unwanted lateral interactions between adsorbates. For the determination of lateral interactions this is generally not a problem. An advantage of a cluster calculation for the lateral interactions is that one can sometimes make a model that definitely has only one type of lateral interaction. For example, if one wants to compute the pair interaction between two adsorbates, one simply puts just these two adsorbates on the cluster. With a periodic calculation this is not possible: the periodicity always gives you an infinite number of adsorbates, and to get the pair interaction of interest one needs to make an assumption about the other lateral interactions that are in principle present. This means that in a periodic calculation one chooses one of

Figure 6 Example of a cluster model of two CO molecules on a cluster of 18 rhodium atoms representing a Rh(100) surface

Catalysis, 2006, 19, 109–163

139

Figure 7 Example of a periodic model of CO on a slab representing a Rh(100) surface. A unit cell containing several metal atoms and the adsorbate is tiled in the x, y, and z direction. This produces a metal slab with on one side molecules adsorbed. The slab extends along the xy plane, and is separated by some empty space from its image in the z direction

the analytical expression that we have seen in Section 3.2 and fits this expression to the results of the calculations. There are various ways in which people have done this. We assume that we model the lateral interactions with pair and multiple-particle interactions as in eqn. (53). This expansion has the advantage that the energy of a system depends linearly on the parameters of our model, and the equations determining these parameters will also be linear and easy to solve. Let’s suppose that we have done calculations on Nstr different adlayer structures and that we have obtained

140

Catalysis, 2006, 19, 109–163

E(n) calc for the adsorption energy per adsorbate in adlayer structure number n. We now want to find an expression ðnÞ

ð0Þ

Efit ¼ Eads þ

Npar X

cnm jm ;

ð58Þ

m¼1 (n) (0) so that E(n) fit agrees with Ecalc as well as possible. In this expression Eads is the adsorption energy of a single isolated adsorbate, jm is one of the pair or multiple-particle interactions (there are Npar of these), and cnm is the number of such interactions per adsorbate in adlayer structure n. Our task is to determine the Npar parameters jm and E(0) ads. The way to determine the lateral interactions with the fewest calculations is (n) (n) to calculate E(n) calc for Nstr ¼ Npar þ 1 adlayer structures. By setting Efit ¼ Ecalc we get just enough (linear) equations to solve the lateral interactions jm and E(0) ads, provided these equations are independent. That dependence of the equations is not unlikely as is discussed below based on the structures shown in Table 1. This figure shows adlayer structures of O/Pt(111) that might be used to compute pair interactions for nearest jNN, next-nearest jNNN, and nextnext-nearest neighbor distances jNNNN, and for three-particle interactions for three oxygen atoms at nearest-neighbor distances forming an equilateral triangle jtriangle, a straight line jlinear, and a bent line jbent. (These interactions have actually been computed with similar structures for O/Ru(0001).25) Suppose that we use the first, second, and fifth structure ((1 1), (2 2)-3O, and (3 2)-3O) to determine E(0) ads, jNN, and jNNN. We have three structures for three parameters. However, the equations for the lateral interactions are dependent, because the ratio between the number of nearest and next-nearest neighbor interactions are the same for all these structures. The same holds for the last four structures. So the separate lateral interactions can not be computed. If Nstr 4 Npar þ 1 then we have more information than we strictly need. This information can be used to get some idea of how well the lateral interactions can be determined. This can be done by taking Npar þ 1 adlayer structures to compute the parameters, substitute these in E(n) fit the other adlayer structures, and compare this with E(n) calc. One can also use all adlayer structures and determine the parameters with a least-squares procedure. The adsorption energy E(0) ads is obviously a different parameter from the lateral interactions jm. It is not uncommon to see people determine this parameter separately. This can be done by including a calculation of an adlayer structure for which one assumes that there are no lateral interactions. Once E(0) ads is known, the lateral interactions can then be determined as described above with the other adlayer structures.

3.4.2 Error Analysis. The simulations of desorption of CO from Rh(100) in Section 4.3 have shown that lateral interactions may be quite large (24 kJ/mol for nearest neighbors), but that there are also interactions that have a profound effect on the kinetics but that are quite small (1.1 and 0.9 kJ/mol for CO–CO pairs farther apart). It is hard to find quotes for numerical values of the errors

141

Catalysis, 2006, 19, 109–163

Table 1

All adlayer structures for which the adsorption energy of the oxygen atoms have been determined. Also given are the coverage y and the adsorption energy per oxygen atom Eads (in kJ/mol). Data is reproduced from references.70,176

(1×1)

(2×2)-3O

(√3 × √3)-20

(2×1)

(3×2)-3O

= 1.00 Eads,calc = -301

0.75 -333 (3× 2)-2O

0.67 -350 (2× 2)

0.50 -367 (2 × √3)

0.50 -365 (√7 × √7)-20

0.25 -402 (√7 × √7)

0.25 -387 (3×3)

0.25 -391

(3×3)-2O

0.33 -379 (3× 2)

= 0.22 Eads,calc = -387

0.17 -393

0.13 -395

0.11 -421

(√3 × √3)

= 0.33 Eads,calc = -381

in DFT for adsorbate-substrate interactions in the literature, but, in view of the smallness of some important lateral interactions, one should definitely ask oneself which lateral interactions one can calculate reliably using DFT. We have not been able to find any publication dealing with this. Instead people have focussed on determining which lateral interactions are important for determining the behavior of a system.25,63,70,176,180 As this question is about small lateral interactions, it is not clear if it can be answered using DFT. First note that when we have just enough adlayer structures to determine all parameters (Nstr ¼ Npar þ 1) that we cannot say anything about the errors we (n) have. We get per definition an exact match between E(n) fit and Ecalc one should have at least Nstr 4 Npar þ 1. Suppose then that this is the case, and that we have calculated adsorption energies per adsorbate E(n) calc but the real values are E(n) ads. We hope that the errors ðnÞ

ðnÞ

en Ecalc Eads

ð59Þ

142

Catalysis, 2006, 19, 109–163

are so small that they affect the determination of E(0) ads and the jm’s only little. These errors need not be independent. There may be a systematic component s that is the same for all adlayer structures, and a rest rn en s that is independent. For each adlayer structure we can now write ! X ðnÞ ðnÞ ð0Þ ðnÞ cnm jm rn Eads : Efit Ecalc ¼ Eads s þ ð60Þ m

We see that the terms that differ for various adlayer structures do not depend on the systematic error s. These terms are also the ones that contain the lateral interactions. This means that the systematic error does not affect the lateral interactions. They depend only on the smaller random errors rn. This means that lateral interactions can be determined better than one might suppose having some idea of the accuracy with which one can compute adsorption energies with DFT. The systematic errors cause only a shift in E(0) ads. Looking at errors as above also shows that singling out one adlayer struc(1) (0) ture, say n ¼ 1, to determine E(0) ads is not a good idea. Because then Efit ¼ Eads (1) (0) (n) (1) and one will set Ecalc ¼ Eads and work with Ecalc Ecalc to determine the lateral interactions. However ðnÞ ð1Þ ðnÞ ð1Þ ð61Þ Ecalc Ecalc ¼ Eads Eads þ ðrn r1 Þ: (1) (n) We see that E(n) calc Ecalc has a different error from Ecalc p.ffiffiffiIn fact, as the r’s are independent rn r1 has a standard deviation that is 2 times the one of rn. Consequently the errorspwe ffiffiffi will make in the determination of the lateral interactions will also be 2 as large. Singling out some adlayer structures to determine the parameters and others to determine the the errors also has drawbacks. First, which adlayer structures are chosen to get the parameters will influence the values one gets, so that there is a degree of arbitrariness. Using other adlayer structures to get some idea of the errors may lead to information on how well adsorption energies can be reproduced, but does not say anything about possible errors in the lateral interactions. However, there is a procedure, called leave-one-out, that generalizes this partitioning of all adlayer structures.70,181 It is useful, not only because it gives information on the errors of the parameters, but it also gives information on whether a parameter should be included or whether it leads to overfitting. An example of the use of this procedure is given in Section 4.4. If we treat all adlayer structures equally and we have Nstr 4 Npar þ 1, then it is natural to use a least-squares procedure: i.e., we minimize

Nstr h X

ðnÞ

ðnÞ

Efit Ecalc

i2

ð62Þ

n¼1

as a function of E(0) ads and the jm’s. This leads to a set of linear equations for these parameters called the normal equations.182 The theory of least-squares also gives error estimates of the parameters. Section 4.4 shows what the results

143

Catalysis, 2006, 19, 109–163

are for the lateral interactions in O/Pt(111) and how well they can be determined.

4

Examples

In this section we present some examples from our own research to illustrate the things that we have discussed above. 4.1 NO/Rh(111). – This system is a good example of how in a relatively simple system subtle combinations of lateral interactions and various sites with different adsorption energies can lead to a number of surprising adlayer structures. Experimentally there are two well-defined adlayer structures. At 1 2ML there is a c(4 2)-2NO structure with two NO molecules per unit cell. It has taken some time to determine the sites for these NO molecules, but it is now generally accepted that one NO is at an hcp site, but the other is at an fcc site. At 34ML there is a (2 2)-3NO structure with three NO molecules per unit cell.50,183 In this structure there are equal numbers of NO molecules at hcp, fcc, and top sites. Figure 8 shows the two structures. The fcc and certainly the top site are energetically less favourable than the hcp site. The reason that these sites become occupied is to avoid strong repulsive lateral interactions. Suppose we try to have all NO at their favourite hcp sites. Up to coverages of 14ML we can avoid all interactions between NO molecules closer then 2a, where a is the nearest distance between two Rh atoms in the top layer. (We assume that the lateral interactions in this system can be described by pair interactions. It has been shown that this is indeed a good model for the lateral interactions.184,185 We will also assume that the

top fcc hcp c(4x2)−2NO

(2x2)−2NO honeycomb

(2x2)−3NO

p(2x1)−1NO

Figure 8 Top: experimentally observed ordered structures of NO on Rh(111). The light grey circles indicate threefold bound NO, the dark grey circles top bound NO. Bottom: alternative ordered structures at 0.50 ML coverage

144

Catalysis, 2006, 19, 109–163

interactions are always repulsive.) Above 14ML we cannot avoid these interac1 tions. interactions between NO molecules closer than pffiffiffiUp to1 3ML we can avoid p ffiffiffi pffiffiffi 2a= 3. At 3ML we can form a 3 3 structure with only next-next-nearestneighbor interactions. Above 13ML we also get next-nearest-neighbor interactions. At 12ML we can make a p(2 1) structure shown in Figure 8. The energy of this structure per adsorbate equals. Ehcp þ jhh a ;

ð63Þ

where Ehcp is the adsorption energy of an NO molecule on a bare substrate and jhh a is the interaction energy between two NO molecules at hcp sites a distance a apart. The subscript x in jX x stands for the distance between the NO molecules, and the superscript X stand for the adsorption sites. This expression should be compared to i 1h i 1h fh pffiffi pffiffi þ Ehcp þ jfh E þ j ; ð64Þ fcc 2a= 3 2a= 3 2 2 which is the energy for the c(4 2) structure. If all the lateral interactions would be zero, then expression (64) would be larger than expression (63), because Efcc 4 Ehcp. The important term in expression (63) that changes this is jhh a , which, according to DFT calculations, equals about 26 kJ/mol, whereas the other lateral interactions are very close to zero.185 The reason why half the NO molecules sit at fcc sites is that in this way they can avoid this very strong repulsion. The price of a smaller adsorption energy at hcp sites is a small one to pay: Ehcp ¼ 248 kJ/mol versus Efcc ¼ 245 kJ/mol. There is a honeycomb (2 2)-2NO structure at 12ML with the same number of NO molecules at hcp and fcc sites (see Figure 8). This structure has an energy of 1 3 1 3 fh pffiffi pffiffi þ Ehcp þ jfh E j þ : ð65Þ fcc 2 2 2a= 3 2 2 2a= 3 The reason why this structure is less stable than the c(4 2) structure is that it pffiffi per adsorbate. The difference is small however, has one extra interaction jfh 2a= 3 and for coverages between 14 and 12ML a kMC simulation often shows patches of this (2 2)-2NO structure when the temperature is not too low (see Figure 9). The NO adlayer only exists at low temperatures. NO dissociates at slightly above 200 K at low coverage. The dissociation is suppressed completely if the coverage is too high: it will only start at about 450 K when NO also starts desorbing. The reason is part site blocking. A nitrogen and an oxygen atom together occupy two sites, whereas NO occupies only one, so there have to be vacant sites for the dissociation. But even if these vacant sites exist the dissociation may still be suppressed, because of strong repulsion between the adsorbates that increases the activation energy for dissociation.185 At intermediate coverages the dissociation starts between 200 and 450 K, then stops when the overall coverage becomes too high, and then restarts when new vacancies are formed due to NO desorption. The temperature-programmed desorption spectra also show a first-order-like N2 desorption peak

Catalysis, 2006, 19, 109–163

145

Figure 9 Typical configuration of the adlayer after adsorption of 0.47 ML NO, showing the onset of c(4 2)-2NO domain formation (indicated by the ovals and the zig–zag lines). In addition to the c(4 2)-2NO domains also small (2 2)-2NO honeycomb domains are present. Two of these domains are highlighted at the left of the figure

at 450 K besides the normal second-order one.98,186 This too can be attributed to the lateral interactions.185 Between 200 and 450 K the simulations show island formation even though all lateral interactions are repulsive. This is explained in Section 2.2.1. The interactions between the atomic adsorbates are similar so they mix well, but the interactions with NO are different. When few NO molecules have dissociated N þ O islands are formed. After most NO molecules have dissociated or desorbed the surface is mainly covered by nitrogen and oxygen and the remaining NO molecules form islands. Dissociation of NO takes place mainly at the boundary between the NO and the N þ O regions, because the repulsion between the adsorbates leads there to the necessary vacant sites. 4.2 Sulfate on Fcc(111) Surfaces. – The system of of sulfate anions on fcc(111) surfaces is a good example of how lateral interactions influence the kinetics of adsorption.187 The system described here relates to the electrochemical system of a sulfuric acid solution in contact with a single crystal electrode. By varying the potential of the electrode one can perform adsorption/desorption experiments. The anions are (partially) discharged during adsorption by transfer of electrons to the electrode. This current can be measured, and is a quantitative measure of the adsorption rate. The linear sweep voltammograms shown in this section display the adsorption current as a function of electrode potential, while the electrode potential is being increased.

146

Catalysis, 2006, 19, 109–163

Figure 10 shows the (111) substrate and the neighboring sites around a central bridge-bonded sulfate (in black). We model the interactions between the adsorbates in two different ways. First, we consider a shell of purely hard interactions, in which the simultaneous bonding of two anions to neighboring sites is simply excluded. These excluded neighboring sites are displayed in white in Figure 10. Next, we consider a second shell of neighboring sites with either finite attractive or finite repulsive interactions. These are displayed in grey in Figure 10. We first look at the case without second neighbor shell interactions (solid line in Figure 11). Going from more negative to more positive potential, the anion adsorbs between 0.1 and 0.1 V in a disordered phase (see the snapshot on the left in Figure 12). This results in a broad adsorption peak in the voltammogram. There is a disorder-order transition at 0.11 V, indicated by the sharp peak in the voltammogram. At this voltage the anion coverage rapidly increases from 0.18 MLpto ffiffiffi 0.20 pffiffiffi ML, which is the saturation coverage. Large adsorbate islands of a ð 3 7Þ structure are formed, as shown in the right panel of Figure 12. Interactions between the adsorbed anions strongly influence the adsorption isotherms. An attraction of 0.5 kBT causes the broad peak associated with adsorption in a disordered phase to sharpen and shift to lower potential (dotted line in Figure 11). At the same time the disorder-order transition peak looses intensity and also shifts to lower potential. The adsorbates on the surface group together in islands even at low coverages, but within the islands there is initially no ordering. Only atpaffiffiffi coverage of 0.18 ML does the disorder-order transition pffiffiffi take place and are ð 3 7Þ ordered domains formed. The islands are slightly smaller in size (100–500 adsorbates) than the ones obtained for the simulations without interactions. This is due to the attractions the adsorbates, pffiffiffi between pffiffiffi which effectively reduce the diffusion rate. Small ð 3 7Þ islands of different orientations are therefore less easily joined into larger islands, and packing faults are less readily removed. The sharpening up of the adsorption peak because of attractive interactions is due to (1) a shift in the adsorption/

Figure 10 Lateral interaction model on an fcc(111) lattice. The adsorbed sulfate anion binds to two surface atoms in a bridged fashion (black atoms), making bonding to the first shell of neighboring sites (white) impossible. There is a finite attractive or repulsive interaction with sulfate anions binding to the second neighbor shell (grey)

147

Catalysis, 2006, 19, 109–163

120 -0.5 kT 80 0.0 kT 40

current j ( µ A/cm 2 )

160

+0.5 kT 0

coverage θ (ML)

0.25 0.0 kT

0.2 0.15

-0.5 kT

+0.5 kT

0.1 0.05 0 -0.2

-0.1

0 0.1 potential E (V)

0.2

0.3

Figure 11 Simulated voltammogram (top) and adsorption isotherm (bottom) for the model with first neighbor shell exclusion and second neighbor shell interaction. Adsorption with an attraction of 0.5 kBT (dotted line) and a repulsion of þ0.5 kBT (dashed line) compared to the case without second neighbor shell interaction (solid line)

0.10 V, 0.18 ML

0.12 V, 0.20 ML

Figure 12 Snapshots of the surface during anion adsorption for the model without second neighbor shell interaction. Before the disorder-order (left panel) there is no pffiffiffi ptransition ffiffiffi ordering; during the disorder-order transition ( 3 7) islands grow p (not after ffiffiffi shown); pffiffiffi the disorder-order transition large islands dominate (right panel). The ( 3 7) unit cell is indicated in the right panel

148

Catalysis, 2006, 19, 109–163

desorption equilibrium, since the adsorption and desorption rate constants are influenced by the presence of neighboring adsorbates, and (2) the presence of more sites accessible for adsorption, since the empty sites are joined into adsorbate-free areas. A repulsion of þ0.5 kBT causes the broad peak associated with adsorption in a disordered phase to broaden even further (dashed line in Figure 11). The onset of adsorption is not influenced, since repulsive interactions only become apparent at higher coverages. Only at higher voltages does adsorption become more difficult, and the broad peak associated with adsorption in a disordered phase thus extends to higher potential. The disorder-order transition in this case retains its intensity, it only shifts pffiffito ffi higher pffiffiffi potential. After the adsorption sweep, large ordered islands of the ð 3 7Þ structure are formed, despite the repulsive interactions. These islands are comparable in size to the ones obtained for the simulations without interactions. The situation discussed here for the model without second neighbor shell interactions strongly resembles the adsorption from sulfuric acid solutions onto Pt(111) surfaces.188,189 The model with attractive second neighbor shell interactions strongly resembles the adsorption from sulfuric acid solutions onto Pd, Ir and Rh surfaces.190–192 These systems have been studied extensively since perchloric and sulfuric acid are the main electrolytes in use. 4.3 CO/Rh(100). – This system forms an example where we have determined lateral interactions by fitting temperature-programmed desorption spectra that were simulated using kinetic Monte Carlo to experimental spectra.86 For coverages below 12ML CO adsorbs at top sites, which form a square grid. CO desorption has a rate constant Eact Wdes ¼ n exp kB T

ð66Þ

with Eact ¼ E(0) act þ nNNjNN þ nNNNjNNN þ nNNNNjNNNN.

(67)

Here Eact is the activation energy for a CO molecule with nNN nearest-neighbors, nNNN next-nearest-neighbors, and nNNNN next-next-nearest-neighbors. E(0) act is the activation energy for an isolated CO molecule, n is the prefactor for desorption, jNN is the nearest-neighbor pair interaction, jNNN is the nextnearest-neighbor pair interaction, and jNNNN is the next-next-nearest-neighbor pair interaction. Note that the lateral interactions fully affect the activation energy, because the barrier for desorption is late and we have a ¼ 1 for the Brønsted-Polanyi coefficient. The fitting resulted in logðn des = sec1 Þ ¼ 12:2þ0:7 0:2 , ð0Þ þ8 þ0:4 Eact ¼ 121þ7 kJ=mol, f ¼ 24 kJ=mol, f ¼ 1:1 kJ=mol, and f NN NNN NNNN 1 3 0:2 ¼ 0:9þ0:3 0:3 kJ=mol. The nearest-neighbor interaction is much larger than the other two. It is also much larger than the thermal energy at the temperature of the experiment: the peak maximum temperature is about T ¼ 500 K. The other interactions are

Catalysis, 2006, 19, 109–163

149

smaller than the thermal energy. The large value of the nearest-neighbor interaction means that CO molecules will not occupy nearest-neighbor sites. At the coverage of 12ML the adlayer forms a c(2 2) structure. Another remarkable consequence is that jNN does not affect the temperatureprogrammed desorption spectra. If jNN would be the only interaction in the system, then the temperature-programmed desorption spectra would be perfect first-order spectra,193 because the molecules avoid this interaction. Consequently, jNN could only be determined approximately. The spectra are however very sensitive to jNNN and jNNNN, so that these small interactions could be determined quite well.86 There is a small shift of the desorption peak of about 24 K to lower temperature going from a coverage of 0.04 to 0.47 ML. This shift is caused only by jNNN and jNNNN. The temperature is much too high for these interactions to induce any ordering of the adlayer, but they still affect the kinetics. One reason is that they lead to some short-range order. Another is that even without short-range order the activation energy is changed. The latter is the most important. In the c(2 2) structure a CO molecules feels four jNNN and four jNNNN interactions, which decrease the activation energy by 8 kJ/ mol. With the prefactor of n ¼ 1.4 1012 s1 this allows us to estimate a shift of 27 K to lower temperature for the change in coverage from 0.04 to 0.47 ML, compared to 24 K experimentally. The difference between these numbers is due to the induced short-range order. The interactions jNNN and jNNNN keep the CO a bit farther apart so that a shift is lessened. The changes in the temperature-programmed desorption spectra as a function of initial coverage are small (see Figure 13), but it seems to be possible to use them to determine quite small lateral interactions, provided a good procedure is used to fit the simulated spectra to the experimental ones. Numerical mathematics provides us with many methods to look for minima of functions.182,194 By defining a function that stands for the difference of the spectra, one might think that it is possible to do the fit with one of these methods from numerical mathematics. However, kMC is a stochastic method. This means that the results are noisy. If we would want to use a standard method from numerical mathematics in the fitting procedure, we have to remove the noise, because such numerical mathematics methods generally cannot handle noisy data. This is only possible by doing simulations with large system sizes, which we want to avoid because of computational costs. We have therefore used evolutionary strategies.195,196 This is one of many optimization methods that mimic natural evolution. (Other better-known ones are genetic algorithms and genetic programming197–199). The method evolves a set of kinetic parameters (the prefactor and activation energy for desorption, and the lateral interactions) that best reproduces the experimental spectra. It not only can handle noisy data, but it also avoids the drawback of many minimization methods that end up in a local minimum close to the starting point. (We also tried Powell’s method and Simulated Annealing to fit the experimental spectra.182 Powell’s method managed to converge in spite of the noise, but generally gave bad fits, because it got trapped in the nearest local minimum. Simulated

150

Catalysis, 2006, 19, 109–163

Desorption rate

simulations experiments

0.47 0.43 0.33 0.25 0.17 0.11 0.08 0.04

400

450

500

550 400 Temperature (K)

0.47 0.43 0.33 0.25 0.17 0.11 0.08 0.04

450

500

550

Figure 13 Experimental (left) and simulated (right) temperature-programmed desorption spectra for CO/Rh(100). The values on the right of each set of curves indicate initial coverages. The thin curves on the right are simulated spectra with the lateral interactions switched off. The heating rate is 5 K/s89

Annealing gave fits of the same quality as ES, but seemed to be somewhat less efficient.) Another advantage of our optimization method is that it gives us an estimate of how well the kinetic parameters can be determined. The nearest-neighbor interaction is much stronger than the other interactions. It leads to a c(2 2) structure at a coverage of 12ML, but it hardly affects the temperature-programmed desorption spectra. This shows up in the fact that we get a very large error estimate for the nearest-neighbor interaction. So the optimization method indicates which parameters are important for an experiment, because only those can be determined with small errors. A model with both next-nearest-neighbor and next-next-nearest-neighbor interactions is a little, but clearly, better than a model without next-nextnearest-neighbor interaction. There is however a clear correlation between these two interactions. Variations in the next-nearest-neighbor and next-nextnearest-neighbor interaction change the spectra only little as long as the sum of these interactions is kept the same. If we allow the lateral interactions to vary over a too large range during the optimization, then we occasionally get adlayer structures in the kMC simulations that differ from those found experimentally. This does not mean necessarily that very different temperature-programmed desorption spectra are

151

Catalysis, 2006, 19, 109–163

obtained. For example, a very good fit to the experimental spectra was obtained with jNN ¼ 2.4 kJ/mol, jNNN and jNNNN ¼ 1.5 kJ/mol. These pffiffiffi ¼ 1.3 pffiffikJ/mol, ffi lateral interactions yield a ð 2 2 2Þ-2CO structure at low temperature. To avoid getting unwanted structures it is possible to impose restrictions on the range of lateral interactions in the optimization procedure. 4.4 O/Pt(111). – This section shows the pros and cons of using DFT calculations to obtain values for lateral interactions.70,176 The DFT calculations were done with a surface model consisting of a supercell with a slab of five metal layers separated by five metal layers replaced by vacuum. The Generalize Gradient Approximation of Perdew and Wang (PW-91), a 5 5 1 grid for Brillouin zone sampling and a cut-off of 400 eV was used. We have focussed in particular on the error analysis as has also been discussed in Section 3.4.2. Table 1 shows all adlayer structures that have been used to obtain the lateral interactions. For adlayer structure n (n Z 1) we have fitted the adsorption energy per oxygen atom E(n) ads with ðnÞ

ð0Þ

Efit ¼ Eads þ

X

lim itsm cnm phim ;

ð68Þ

where E(0) ads is the adsorption energy for an isolated oxygen atom, jm is a lateral interactions parameter, and cnm is the number of such interactions per oxygen atom in adlayer structure n. We looked at models for the lateral interactions that included pair interactions jNN, jNNN, and jNNNN, and 3-particle interactions jlinear, jtriangle, jbent, and all possible subsets of these interactions (see Figure 14). Table 2 shows results of a number of models for the lateral interactions. They are obtained by a linear regression procedure. We use the leave-one-out method to see how reliable these results are.181 The idea is to do linear regression with

Figure 14 Definition of the pairwise interactions jNN (bottom-middle), jNNN (bottomleft), jNNNN (bottom-right), and the 3-particle interactions jlinear (top-right), jtriangle (top-left), jbent (top-middle)

152

Catalysis, 2006, 19, 109–163

Table 2

Adsorption energy for an isolated oxygen atom, lateral interactions, and the leave-one-out error for different models of the lateral interactions. If a lateral interaction is not specified it is not included in the model. All quantities are in kJ/mol

E(0) ads 396.3 393.9 396.8 393.3

jNN

1.2 1.5 1.6 2.1

19.9 20.8 26.0 29.7

2.3 2.0 1.5 1.8

7.0

jNNN

jNNNN

jlinear

Rloo

5.5 0.9 5.0 0.8 4.8 1.2

1.9 0.9

6.1 2.0 7.2 1.8

3.1 3.7 4.6 6.1

14

2456 1456 1236

6.5

Rloo [kJ·mol-1]

6.0

1235

13 15 16

1

5.5

126

5.0

125 123

4.5

12356 123456

12 1256

12345 12346 12456

4.0 1234 1245

3.5 124

3.0

1246

Figure 15 The leave-one-out error (in kJ/mol) as a function of the model of the lateral interactions. The numbers indicate the different lateral interactions: jNN(¼1), jNNN(¼2), jNNNN(¼3), jlinear(¼4), jtriangle(¼5), and jbent(¼6). The lines are guides to the eye. They connect models that differ by one lateral interactions parameter. A fat line indicates that adding a parameter improves the model

all structures except one, and then looked at the leave-one-out error R2loo ¼

i2 1 X h ðkÞ ðkÞ Eads;pred Eads;calc ; Nstr k

ð69Þ

where E(k) ads,pred is the energy of the adlayer structure k obtained with the lateral interactions that have been determined without that structure, and Nstr is the number of adlayers structures in the summation. This error indicates how well a model predicts the energy. Adding parameters will not necessarily decrease this error. An increase indicates overfitting. Figure 15 shows how the leave-one-out error changes with the model of the lateral interactions. It is clear that the nearest-neighbor and next-nearestneighbor interactions are the most important. These two interactions already provide a good description of all the adlayer structures. The next-next-nearestneighbor interaction does not improve the model, but remarkably the linear

Catalysis, 2006, 19, 109–163

153

3-particle interaction does. Our results indicate that it is not justifiable to determine more lateral interactions as was done for this system before,63 and also for the similar O/Ru(0001).25 Table 2 gives the lateral interactions for some important models. The reason why some lateral interactions cannot be determined is that they are too weak and consequently would have huge relative errors. The root-mean-square error of the model with only nearest-neighbor and next-nearest-neighbor interactions is 3.9 kJ/mol. The one of the model that also includes the linear 3-particle interaction is 2.9 kJ/mol. Using these errors as estimates for the errors in DFT calculations gives the error estimates of the lateral interactions in table 2. These should be regarded as lower estimates. If we would use a larger estimate for the errors in DFT then the errors of the lateral interactions would increase proportionally. Note that, because the term (n) E(0) ads is always the same in the expressions for Efit independent of the adlayer structure, possible systematic errors in DFT (s in Section 3.4.2) only affect E(0) ads and not the lateral interactions. To further assess the quality of the models of the lateral interactions that we have developed, we have used kMC to simulate TPD experiments. The main kinetic parameters for desorption are not the lateral interactions, but the activation energy and the prefactor for desorption. From a given set of lateral interactions we have determined these parameters by fitting the simulated TPD spectra to the experimental ones using Differential Evolution in a procedure described for CO desorption from Rh(100) in Section 4.3.86,201 Adsorption experiments indicate that there is a precursor for adsorption.202 This also means that there is a transition state not far from the surface. Consequently, the adsorption energy and the activation energy for desorption will be different, which is why we fit the activation energy. Another consequence of this transition state is that lateral interactions may affect the energy of this transition state differently from the way they affect the adsorbates on the surface. In general we determine this by fitting the Brønsted-Polanyi parameter a (see eqn. (7)) just like the prefactor and the activation energy. Figure 16 shows the experimental and simulated TPD spectra for four initial coverages. We see that there is an excellent agreement between the experimental and the simulated spectra. 4.5 Tartaric Acid on Cu(110). – The system of tartaric acid on Cu(110) shows how a combination of direct and through-surface interactions can yield a complex ordered structure.38,203,204 Preadsorption of tartaric acid has been used to perform enantioselective catalysis on ordinary heterogeneous catalysts. Scanning tunneling microscopy studies of the adsorption of tartaric acid chiral modifiers on a model catalyst surface, Cu(110), have revealed that each isomer forms different chemisorption domains, where the domain of the (R,R)isomer is the mirror image of the domain formed by the (S,S)-isomer (see Figures 17 and 18).205–207 The formation of such mirror image ordered domains for molecules of opposite chirality is also known for adsorption of amino acids.208,209

154

Desorption rate

Catalysis, 2006, 19, 109–163

600

700 800 Temperature (K)

900

Figure 16 Simulated and experimental temperature-programmed desorption spectra for O/Pt(111). The solid lines are experimental spectra. The crosses indicate simulated spectra for a model of the lateral interactions with nearest and next-nearest pair interactions, and also a linear 3-particle interaction. The O2 is formed from two atoms at nextnearest-neighbor positions. The kinetic parameters are E(0) act ¼ 206.4 kJ/mol, n ¼ 2.5 1013 s1, a ¼ 0.773, jNN ¼ 19.9 kJ/mol, jNNN ¼ 5.5 kJ/mol, and jlinear ¼ 6.1 kJ/mol. In each plot the curves from top to bottom are for initial oxygen coverage of 0.194, 0.164, 0.093, and 0.073 ML, respectively. The heating rate is 8 K/s200 O O C * H C OH

=

*

HO C H C O O [001]

HO H H OH * * C C C C O O O O

[110] [110] top view

[001] side view

Figure 17 The adsorption geometry of (R,R)-bitartrate on Cu(110). The dashes protruding from the tartrate oval represent the tartrate hydroxyl groups sticking out

(9 0,1 2) unit cell

[001]

[110]

trough in direction

Figure 18 The (9 0,1 2) structure for (R,R)-bitartrate on Cu(110). The overall (9 0,1 2) unit cell is indicated with a solid line. The trough in the ½1 14 direction can provide chiral adsorption sites on the otherwise achiral Cu(110) surface

155

Catalysis, 2006, 19, 109–163

These chiral domains of tartaric acid are found to be formed by its bitartrate form. They follow a 2-dimensional ordered structure. In the case of the (R,R)isomer the supramolecular assembly can be described by the following matrix notation, which defines the unit cell of the adlayer unambiguously in terms of the unit cell of the substrate: MR;R ¼

9 1

0 2

ð70Þ

This matrix notation will be referred to as (9 0,1 2) in the text. The molecules form extended molecular rows along the ½114 direction. These parallel rows are assembled in groups of three, each group being separated from the next by an empty space (trough) (see Figure 18). These troughs can provide a chiral adsorption site on the otherwise achiral Cu(110) surface. DFT calculations were used to quantify the different interactions. All interactions were found to be repulsive. The most repulsive interactions were along the copper rows (in the ½1 10 direction), due to a through-surface interaction between carboxylate groups of different tartaric acid molecules binding next to each other. In addition to these interactions we have proposed the existence of an adsorbate-induced surface stress which reduces the binding energy when more than three tartaric acid molecules bind to the same copper row. This surface stress causes the empty troughs in the (9 0,1 2) ordered structure.38 The chirality of the (9 0,1 2) structure is caused by the hydroxyl groups sticking out from the adsorbed tartaric acid molecules. In the mirror image of the (9 0,1 2) structure, the (9 0,1 2) structure, hydroxyl groups of neighboring tartaric acid molecules are forced close to each other. This causes a decrease in binding energy. In the absence of these hydroxyl groups there is no difference in energy between the two mirror ordered structures, and both are therefore observed for succinic acid (HOOC–CH2–CH2–COOH) on Cu(110).204 The formation of the (9 0,1 2) structure is displayed in Figure 19. Since all interactions are repulsive, the (9 0,1 2) structure only forms at coverages close to 0.167 ML. It forms as a consequence of the avoidance of strongly repulsive interactions. At around 0.15 ML, the tartrate orders in rows in the ½114 direction. This is similar to the (9 0,1 2) structure, except for the larger spacing in the ½1 10 direction (indicated by the arrows above the panel). Upon increasing the coverage, the rows collapse to form the (9 0,1 2) structure (left and right of panel a). At first the structure thus formed has defects, since at some positions tartaric acid molecules are missing. Finally, when after continued adsorption all defects are filled, a perfect (9 0,1 2) structure is formed, as shown in panel b.

5

Outlook

The foregoing has dealt with what we do know about and what we can do with lateral interactions. This might give the impression that this is a mature and well-researched area. This is however far from the case. Especially if we look at

156

Catalysis, 2006, 19, 109–163

a

b

Figure 19 Snapshots of simulated adsorption of (R,R)-bitartrate on Cu(110). (a) the (9 0,1 2) structure (upper left of the picture) forms through compression of wider spaced rows (middle of the picture). The difference in row spacing is indicated by the arrows above the picture. (b) Under continued adsorption, a perfect (9 0,1 2) structure is obtained; the unit cell is indicated

quantitative information, very little is known about lateral interactions. We would like to end this paper with a short overview of the main obstacles to progress in this area. For experiments the problem is mainly the interpretation of the results. The formation of islands or superstructures, shifts or additional peaks in TPD spectra, and heats of adsorption as a function of coverage are all examples that clearly show the presence of lateral interactions, but it is far from straightforward to say how strong the lateral interactions are between adsorbates. We have argued above that one needs to do simulations to get this information, but there is still little experience with this, especially as far as fitting of simulation results to experimental data is concerned. Moreover, even a relatively simple system shows that it is quite possible to reproduce experimental results with an incorrect model.185 If coverages become high and the lateral interactions become of the same order of magnitude as the differences in adsorption energies for different sites, the problem may easily become intractable, even without further complicating factors as steps and contaminations. More experiments on simple, well-defined systems combined with good simulations of the same are needed. Kinetic Monte Carlo simulations are extremely useful to study lateral interactions, but they have their own problems. The main one is diffusion. This is generally much faster than the reactions that one is interested in. Consequently most computer time is spent on this diffusion, making a simulation inefficient or even infeasible. A possible solution is to reduce the rate of diffusion from its real rate to a smaller one that still equilibrates the adlayer and gives a simulation that is possible to do. With lateral interactions this can not always be done, because they often hinder the restructuring of the adlayer causing very long relaxation times. Methods to speed up kMC simulations are currently being developed,152,210–213 but it is still very much early days to say how useful and general these methods are. Another problem with kMC simulations is that they use a lattice-gas model. Even if we have a single-crystal surface, this may not be appropriate when the

Catalysis, 2006, 19, 109–163

157

coverage is high and the lateral interactions are so strong that they force adsorbates away from the normal adsorption sites. For a real catalyst consisting of small metal particles on a support a lattice-gas model is of course totally inappropriate. KMC simulations without a lattice are possible, but such simulations are much less efficient, because they need to compute reactions and their rate constants during the simulation instead of using a list of them that is given on input.54,57 It is not clear however how useful these methods are for real catalysts. Kinetic studies generally use phenomenological rate equations where the effect of lateral interactions is hidden in effective rate constants. The drawback of these equations is that they have only a very limited range on which they are valid. (Usually they cannot be extrapolated outside the range that is sampled in experiments to which the rate equations have been fitted.) What is sorely needed is an analytical approach that includes lateral interactions properly. It need not have to be applicable at all reaction conditions: even if its use is somewhat restricted, it would still be very useful because it will be so much easier to handle than kMC simulations. Rate equations like (11) might be appropriate at high temperatures. Again, more experience is needed. The advantage of calculating lateral interactions with DFT is that it is clear what the system looks like, although this does not mean that it is always clear which lateral interactions are present. The problem with DFT is how its errors propagate to the lateral interactions. The most important lateral interactions are those with energies of the same order of magnitude as the thermal energy. This energy is generally too small for DFT. There are also larger lateral interactions, e.g., when adsorbates are close together. Systems often avoid these interactions. This means that it will be difficult to extract them from experimental results. They are still important, because they do determine the behavior of the system. They should be easier to calculate with DFT. There has hardly been any analysis of just which lateral interactions can be calculated with DFT. In fact, the analysis of O/Pt(111) Section 4.4 is the only one that addresses this question that we know of. Here too more work is necessary.

Acknowledgments We would like to thank all the people who have worked with us on various systems showing the effects of lateral interactions. We are especially grateful for many stimulating discussions with Drs. J.J. Lukkien and M.T.M. Koper, and Profs. R.A. van Santen and J.W. Niemantsverdriet. Our research was financially supported by the National Research School Combination Catalysis (NRSC-C). References 1. M. Smedh, A. Beutler, M. Borg, R. Nyholm and J.N. Andersen, Surf. Sci., 2001, 491, 115. 2. N.V. Petrova and I.N. Yakovkin, Surf. Sci., 2002, 519, 90. 3. C.M. Chan, R. Aris and W.H. Weinberg, Appl. Surf. Sci., 1978, 1, 360. 4. A.M. de Jong and J.W. Niemantsverdriet, Surf. Sci., 1990, 233, 355.

158

Catalysis, 2006, 19, 109–163

5. A.M. de Jong and J.W. Niemantsverdriet, Vacuum, 1990, 41, 232. 6. D.L.S. Nieskens, A.P. van Bavel and J.W. Niemantsverdriet, Surf. Sci., 2003, 546, 159. 7. L. O¨sterlund, M.Ø. Pedersen, I. Stensgaard, E. Lægsgaard and F. Besenbacher, Phys. Rev. Lett., 1999, 83, 4812. 8. T. Zambelli, J. Trost, J. Wintterlin and G. Ertl, Phys. Rev. Lett., 1996, 76, 795. 9. J. Trost, T. Zambelli, J. Wintterlin and G. Ertl, Phys. Rev. B, 1996, 54, 17850. 10. S. Renisch, R. Schuster, J. Wintterlin and G. Ertl, Phys. Rev. Lett., 1999, 82, 3839. 11. J. Wintterlin, J. Trost, S. Renisch, R. Schuster, T. Zambelli and G. Ertl, Surf. Sci., 1997, 394, 159. 12. L. Vattuone, Y.Y. Yeo and D.A. King, J. Chem. Phys., 1996, 104, 8096. 13. R. Kose, W.A. Brown and D.A. King, J. Phys. Chem. B, 1999, 103, 8722. 14. Q. Ge, R. Kose and D.A. King, Adv. Catal., 2000, 45, 207. 15. E.G. Seebauer, A.C.F. Kong and L.D. Schmidt, J. Vac. Sci. Technol. A, 1987, 5, 464. 16. E.G. Seebauer, A.C.F. Kong and L.D. Schmidt, J. Chem. Phys., 1988, 88, 6597. 17. E.G. Seebauer and C.E. Allen, Prog. Surf. Sci., 1995, 49, 265. 18. V.P. Zhdanov, J. Pavlicek and Z. Knor, Catal. Rev. Sci. Eng., 1988, 30, 501. 19. V.P. Zhdanov, Surf. Sci. Rep., 1991, 12, 183. 20. E. Shustorovich and H. Sellers, Surf. Sci. Rep., 1998, 31, 1. 21. E. Hansen and M. Neurock, Chem. Eng. Sci., 1999, 54, 3411. 22. S.P. Baranov, A. Abramova, A.V. Zeigarnik and E. Shustorovich, Surf. Sci., 2004, 555, 20. 23. E.R. Hansen and M. Neurock, Surf. Sci., 2000, 464, 91. 24. E. Hansen and M. Neurock, Surf. Sci., 1999, 441, 410. 25. C. Stampfl, H.J. Kreuzer, S.H. Payne, H. Pfnu¨r and M. Scheffler, Phys. Rev. Lett., 1999, 83, 2993. 26. K. Honkala, P. Pirila¨ and K. Laasonen, Phys. Rev. Lett., 2001, 86, 5942. 27. B. Hammer, Surf. Sci., 2000, 459, 323. 28. D.H. Wei, D.C. Skelton and S.D. Kevan, J. Vac. Sci. Technol. A, 1994, 12, 2029. 29. J.K. Nørskov, In ‘Coadsorption Promotors and Poisons’, volume 6. Elsevier, Amsterdam, 1993, pp. 1–27. 30. M.L. Merrick, W. Luo and K.A. Fichthorn, Prog. Surf. Sci., 2003, 72, 117. 31. N.D. Lang, S. Holloway and J.K. Nørskov, Surf. Sci., 1985, 150, 24. 32. D.F. Ogletree, R.Q. Hwang, D.M. Zeglinski, A. Lopez Vasquez de Parga, G.A. Somorjai and M. Salmeron, J. Vac. Sci. Technol. B, 1991, 9, 886. 33. R. Linke, D. Curulla, M.J.P. Hopstaken and J.W. Niemantsverdriet, J. Chem. Phys., 2001, 115, 8209. 34. D. Loffreda, D. Simon and P. Sautet, Chem. Phys. Lett., 1998, 291, 15. 35. D. Loffreda, D. Simon and P. Sautet, Surf. Sci., 1999, 425, 68. 36. C.I. Carlisle and D.A. King, J. Phys. Chem. B, 2001, 105, 3886. 37. L.A.M.M. Barbosa and P. Sautet, J. Am. Chem. Soc., 2001, 123, 6639. 38. C.G.M. Hermse, A.P. van Bavel, A.P.J. Jansen, L.A.M.M. Barbosa, P. Sautet and R.A. van Santen, J. Phys. Chem. B, 2004, 108, 11035. 39. B. Lehner, M. Hohage and P. Zeppenfeld, Phys. Rev. B, 2002, 65, 165407. 40. B. Lehner, M. Hohage and P. Zeppenfeld, Chem. Phys. Lett., 2003, 379, 568. 41. R. Brako and D. Sˇokcˇevic´, Surf. Sci., 1998, 401, L388. 42. R. Brako and D. Sˇokcˇevic´, Surf. Sci., 2000, 469, 185. 43. M. Wilms, P. Broekmann, C. Stuhlmann and K. Wandelt, Surf. Sci., 1998, 416, 121.

Catalysis, 2006, 19, 109–163

159

44. P. Broekmann, M. Wilms, A. Spaenig and K. Wandelt, Progr. Surf. Sci., 2001, 67, 59. 45. M. Lennartz, P. Broekmann, M. Arenz, C. Stuhlmann and K. Wandelt, Surf. Sci., 1999, 442, 215. 46. R.J. Gelten, A.P.J. Jansen, R.A. van Santen, J.J. Lukkien, J.P.L. Segers and P.A.J. Hilbers, J. Chem. Phys., 1998, 108, 5921. 47. P. van Beurden, B.S. Bunnik, G.J. Kramer and A. Borg, Phys. Rev. Lett., 2003, 90, 066106. 48. L.-W.H. Leung, J.-W. He and D.W. Goodman, J. Chem. Phys., 1990, 93, 8328. 49. J.S. Villarrubia and W. Ho, J. Chem. Phys., 1987, 87, 750. 50. I. Zasada, M.A. Van Hove and G.A. Somorjai, Surf. Sci., 1998, 418, L89. 51. S.-J. Han, C.-W. Lee, H. Yoon and H. Kang, J. Chem. Phys., 2002, 116, 2684. 52. F. Frećhard, R.A. van Santen, A. Siokou, J.W. Niemantsverdriet and J. Hafner, J. Chem. Phys., 1999, 111, 8124. 53. A. Patrykiejew, S. Sokolowski and K. Binder, Surf. Sci. Rep., 2000, 37, 207. 54. F. Much, M. Ahr, M. Biehl and W. Kinzel, Comp. Phys. Comm., 2002, 147, 226. 55. M. Biehl, F. Much and C. Vey. Off-lattice kinetic Monte Carlo simulations of strained heteroexpitaxial growth. http://arXiv.org/, paperno. cond-mat/0405641, 2004. 56. M. Biehl, Lattice gas models and kinetic Monte Carlo simulations of expitaxial growth. http://arXiv.org/, paperno. cond-mat/0406707, 2004. 57. G. Henkelman and H. Jońsson, J. Chem. Phys., 2001, 115, 9657. 58. G. Henkelman and H. Jońsson, Phys. Rev. Lett., 2003, 90, 116101. 59. H. Xu and K.Y.S. Ng, Surf. Sci., 1996, 365, 779. 60. C.B. France and B.A. Parkinson, J. Am. Chem. Soc., 2003, 125, 12712. 61. B. Riedmu¨ller, D.C. Papageorgopoulos, B. Berenbak, R.A. van Santen and A.W. Kleyn, Surf. Sci., 2002, 515, 323. 62. J.L. Gland and E.B. Kollin, J. Chem. Phys., 1983, 78, 963. 63. H. Tang, A. Van der Ven and B.L. Trout, Phys. Rev. B, 2004, 70, 045420. 64. D.-J. Liu, J. Chem. Phys., 2004, 121, 4352. 65. S. Shen, F. Zaera, D.A. Fischer and J.L. Gland, J. Chem. Phys., 1988, 89, 590. 66. W.D. Mieher, L.J. Whitman and W. Ho, J. Chem. Phys., 1989, 91, 3228. 67. L.J. Whitman and W. Ho, J. Chem. Phys., 1989, 90, 6018. 68. M. Kiskinova, A. Szabo and J.T. Yates Jr., J. Chem. Phys., 1988, 89, 7599. 69. R.J. Baxter, ‘Exactly Solved Models in Statistical Mechanics’, Academic Press, London, 1982. 70. A.P.J. Jansen and W.K. Offermans, In ‘Computational Science and Its Applications—ICCSA 2005’, eds. O. Gervasi et al., Springer, Berlin, 2005, pp. 1020–1029. 71. J. Lahtinen, K. Kauraala, J. Vaari, T. Vaara, P. Kaukasoina and M. Lindroos, Phys. Rev. B, 2001, 63, 155402. 72. M. Kinne, T. Fuhrmann, C.M. Wheland, J.F. Zhu, J. Pantfo¨rder, M. Probst, G. Held, R. Denecke and H.-P. Steinrlick, J. Chem. Phys., 2003, 117, 10852. 73. N. Ikemiya, T. Senna and M. Ito, Surf. Sci., 2000, 464, L681. 74. J. Lahtinen, J. Vaari and K. Kauraala, Surf. Sci., 1998, 418, 502. 75. A. Beutler, E. Lundgren, R. Nyholm, J.N. Andersen, B.J. Setlik and D. Heskett, Surf. Sci., 1998, 398, 117. 76. M. Birgersson, C.-O. Almbladh, M. Borg and J.N. Andersen, Phys. Rev. E, 2003, 67, 045402. 77. F. Bondino, G. Comelli, F. Esch, A. Locatelli, A. Baraldi, S. Lizzit, G. Paolucci and R. Rosei, Surf. Sci., 2000, 459, L467.

160

Catalysis, 2006, 19, 109–163

78. A. Eichler, Surf. Sci., 2003, 526, 332. 79. C.D. Lund, C.M. Surko, M.B. Maple and S.Y. Yamamoto, Surf. Sci., 2000, 459, 413. 80. F.G. Requejo, E.L.D. Hebenstreit, D.F. Ogletree and M. Salmeron, J. Catal., 2004, 226, 83. 81. B. Meng and W.H. Weinberg, J. Chem. Phys., 1995, 102, 1003. 82. A.P.J. Jansen. An introduction to Monte Carlo simulations of surface reactions. http://arXiv.org/, paperno. cond-mat/0303028, 2003. 83. A.G. Makeev and M.M. Slinko, Surf. Sci. Lett., 1996, 359, L467. 84. A.G. Makeev, N.M.H. Janssen, P.D. Cobden, M.M. Slinko and B.E. Nieuwenhuys, J. Chem. Phys., 1997, 107, 965. 85. A.G. Makeev and B.E. Nieuwenhuys, J. Chem. Phys., 1998, 108, 3740. 86. A.P.J. Jansen, Phys. Rev. B, 2004, 69, 035414. 87. A.P. van Bavel, C.G.M. Hermse, M.J.P. Hopstaken, A.P.J. Jansen, J.J. Lukkien, P.A.J. Hilbers and J.W. Niemantsverdriet, Phys. Chem. Chem. Phys., 2004, 6, 1830. 88. M.J.P. Hopstaken and J.W. Niemantsverdriet, J. Chem. Phys., 2000, 113, 5457. 89. A.P. van Bavel, M.J.P. Hopstaken, D. Curulla, J.J. Lukkien, P.A.J. Hilbers and J.W. Niemantsverdriet, J. Chem. Phys., 2003, 119, 524. 90. V. Bustos, R.O. U nac, F. Zaera and G. Zgrablich, J. Chem. Phys., 2003, 118, 9372. 91. F. Zaera, S. Wehner, C.S. Gopinath, J. L. Sales, V. Gargiulo and G. Zgrablich, J. Phys. Chem. B, 2001, 105, 7771. 92. F. Zaera and C.S. Gopinath, J. Chem. Phys., 2002, 116, 1128. 93. C. Stampfl and M. Scheffler, Phys. Rev. B, 2002, 65, 155417. 94. R. Becker, ‘Theorie der Wa¨rme’, Springer, Berlin, 1985. 95. Y.Y. Yeo, L. Vattuone and D.A. King, J. Chem. Phys., 1996, 104, 3810. 96. A.M. de Jong and J.W. Niemantsverdriet, J. Chem. Phys., 1994, 101, 10126. 97. W. Widdra, P. Trischberger, W. Frieß, D. Menzel, S.H. Payne and H.J. Kreuzer, Phys. Rev. B, 1998, 57, 4111. 98. H.J. Borg, J.F.C.-J.M. Reijerse, R.A. van Santen and J.W. Niemantsverdriet, J. Chem. Phys., 1994, 101, 10052. 99. M. Aryafar and F. Zaera, J. Catal., 1998, 175, 316. 100. C.S. Gopinath and F. Zaera, J. Catal., 1999, 186, 387. 101. M.J.P. Hopstaken and J.W. Niemantsverdriet, J. Vac. Sci. Technol. A, 2000, 18, 1503. 102. A.J. Heinrich, C.P. Lutz, J.A. Gupta and D.M. Eigler, Science, 2002, 298, 1381. 103. J.E. Leffler, Science, 1953, 117, 340. 104. G.S. Hammond, J. Am. Chem. Soc., 1955, 77, 334. 105. B. Hammer, Phys. Rev. B, 2001, 63, 205423. 106. J.N. Brønsted, Chem. Rev., 1928, 5, 231. 107. M.G. Evans and M. Polanyi, Trans. Faraday Soc., 1938, 34, 11. 108. S.J. Lombardo and A.T. Bell, Surf. Sci. Rep., 1991, 13, 1. 109. V.P. Zhdanov, ‘Elementary Physicochemical Processes on Solid Surfaces’, Plenum, London, 1991. 110. S.H. Payne and H.J. Kreuzer, Surf. Sci., 1998, 399, 135. 111. H.J. Kreuzer, S.H. Payne, P. Jakob and D. Menzel, Surf. Sci., 1999, 424, 36. 112. J.-S. McEwen, S. H. Payne, H. J. Kreuzer, M. Kinne, R. Denecke and H.-P. Steinru¨ck, Surf. Sci., 2003, 545, 47. 113. M.P. Allen and D.J. Tildesley, ‘Computer Simulation of Liquids’, Clarendon, Oxford, 1987. 114. D.A. McQuarrie, ‘Statistical Mechanics’, Harper, New York, 1976.

Catalysis, 2006, 19, 109–163

161

115. L.D. Landau and E.M. Lifshitz, ‘Statistical Physics Part 1, Course of Theoretical Physics’, Vol. 5, Pergamon Press, Oxford, 1980. 116. R.J. Glauber, J. Math. Phys., 1963, 4, 294. 117. J.G. Kirkwood, J. Chem. Phys., 1935, 3, 300. 118. H. Mamada and F. Takano, J. Phys. Soc. Japan, 1968, 25, 675. 119. J. Mai, V.N. Kuzovkov and W. von Niessen, J. Chem. Phys., 1994, 100, 6073. 120. N. Wu, ‘The Maximum Entropy Method’, Springer, Berlin, 1997. 121. R.A. Adams, ‘Calculus’, Addison-Wesley, Don Mills, Ontario, 1995. 122. V.P. Zhdanov, Surf. Sci., 1997, 392, 185. 123. D. Gupta and C.S. Hirtzel, Chem. Phys. Lett., 1988, 149, 527. 124. D. Gupta and C.S. Hirtzel, Surf. Sci., 1989, 210, 322. 125. S.J. Lombardo and A.T. Bell, Surf. Sci., 1988, 206, 101. 126. S.J. Lombardo and A.T. Bell, Surf. Sci., 1989, 224, 451. 127. J.L. Sales and G. Zgrablich, Phys. Rev. B, 1987, 35, 9520. 128. M. Silverberg, A. Ben-Shaul and F. Rebentrost, J. Chem. Phys., 1985, 83, 6501. 129. M. Silverberg and A. Ben-Shaul, J. Chem. Phys., 1987, 87, 3178. 130. K.A. Fichthorn and W.H. Weinberg, J. Chem. Phys., 1991, 95, 1090. 131. B. Meng and W.H. Weinberg, J. Chem. Phys., 1994, 100, 5280. 132. R.A. van Santen and J.W. Niemantsverdriet, ‘Chemical Kinetics and Catalysis’, Plenum Press, New York, 1995. 133. N.G. van Kampen, ‘Stochastic Processes in Physics and Chemistry’, North-Holland, Amsterdam, 1981. 134. J.C. Keck, J. Chem. Phys., 1960, 32, 1035. 135. J.C. Keck, Adv. Chem. Phys., 1967, 13, 85. 136. P. Pechukas, In ‘Dynamics of Molecular Collisions, Part B’, ed. W. Miller, Plenum Press, New York, 1976, pp. 269–322. 137. D.G. Truhlar, A.D. Isaacson and B.C. Garrett, In ‘Theory of Chemical Reaction Dynamics, Part IV’, ed. M. Baer, CRC Press, Boca Raton, 1985, pp. 65–138. 138. A.P.J. Jansen, Comput. Phys. Comm., 1995, 86, 1. 139. R.J. Gelten, R.A. van Santen and A.P.J. Jansen, In ‘Molecular Dynamics: From Classical to Quantum Methods’, eds. P.B. Balbuena and J.M. Seminario, Elsevier, Amsterdam, 1999, pp. 737–784. 140. K. Binder, ‘Monte Carlo Methods in Statistical Physics’, Springer, Berlin, 1986. 141. D.T. Gillespie, J. Comput. Phys., 1976, 22, 403. 142. D.T. Gillespie, J. Phys. Chem., 1977, 81, 2340. 143. J. Honerkamp, ‘Stochastische Dynamische Systeme’, VCH, Weinheim, 1990. 144. M.T.M. Koper, J.J. Lukkien, A.P.J. Jansen, P.A.J. Hilbers and R.A. van Santen, J. Chem. Phys., 1998, 109, 6051. 145. F.C. Alcaraz, M. Droz, M. Henkel and V. Rittenberg, J. Phys., 1994, 230, 250. 146. G. Grinstein and C. Jayaprakash, Comput. Phys., 1995, 9, 164. 147. C.J. Pe´rez, A. Corral, A. Dı´ az Guilera, K. Christensen, and A. Arenas. On selforganized criticality and synchronization in lattice models of coupled dynamical systems. http://xxx.lanl.gov, paperno. cond-mat/9601102, 1996. 148. V. Privman, ‘Nonequilibrium Statistical Mechanics in One Dimension’, Cambridge University Press, Cambridge, 1997. 149. H. Hinrichsen, Critical phenomena in nonequilibrium systems. http://xxx.lanl.gov, paperno. cond-mat/0001070, 2000. 150. J.P.L. Segers, ‘Algorithms for the Simulation of Surface Processes’, Ph.D. thesis, Eindhoven University of Technology, 1999.

162

Catalysis, 2006, 19, 109–163

151. J.J. Lukkien, J.P.L. Segers, P.A.J. Hilbers, R.J. Gelten and A.P.J. Jansen, Phys. Rev. E, 1998, 58, 2598. 152. A.P.J. Jansen and J.J. Lukkien, Catalysis Today, 1999, 53, 259. 153. D.E. Knuth, ‘The Art of Computer Programming, Volume III: Sorting an, Searching’, Addison-Wesley, Reading, 1973. 154. J.N. Murrell, S. Carter, P. Huxley, S.C. Farantos and A.J.C. Varandas, ‘Molecular Potential Energy Functions’, Wiley-Interscience, Chichester, 1984. 155. P.G. Mezey, ‘Potential Energy Hypersurfaces’, Elsevier, Amsterdam, 1987. 156. H.J. Kreuzer, S.H. Payne, A. Drozdowski and D. Menzel, J. Chem. Phys., 1999, 110, 6982. 157. E. Shustorovich, Surf. Sci. Rep., 1986, 6, 1. 158. V.P. Zhdanov and B. Kasemo, Surf. Sci., 1998, 415, 403. 159. C. Schwennicke and H. Pfnu¨r, Phys. Rev. B, 1997, 56, 10558. 160. P. Piercy, K. De’Bell and H. Pfniir, Phys. Rev. B, 1992, 45, 1869. 161. A. Szabo, M. Kiskinova and J.T. Yates Jr., J. Chem. Phys., 1989, 90, 4604. 162. R.J. Behm, K. Christmann, G. Ertl and M.A. Van Hove, J. Chem. Phys., 1980, 73, 2984. 163. L.D. Peterson and S.D. Kevan, J. Chem. Phys., 1991, 94, 2281. 164. C.G.M. Hermse, A.P. van Bavel, B.E. Nieuwenhuys, J.J. Lukkien, R.A. van Santen and A.P.J. Jansen, Langmuir, 2005, 21, 8302. 165. J.T. Stuckless, N. Al-Sarraf, C. Wartnaby and D.A. King, J. Chem. Phys., 1993, 99, 2202. 166. Y.Y. Yeo, L. Vattuone and D.A. King, J. Chem. Phys., 1997, 106, 392. 167. Y.Y. Yeo, L. Vattuone and D.A. King, J. Chem. Phys., 1997, 106, 1990. 168. R. McWeeny and B.T. Sutcliffe, ‘Methods of Molecular Quantum Mechanics’, Academic Press, London, 1976. 169. A. Szabo and N.S. Ostlund, ‘Modern Quantum Chemistry: Introduction to Advanced Electronic Structure Theory’, McGraw-Hill, New York, 1982. 170. R.G. Parr and W. Yang, ‘Density-Functional Theory of Atoms and Molecules’, Oxford University Press, New York, 1989. 171. J.P. Perdew, In ‘Electronic Structure of Solids 91’, eds. P. Ziesche and H. Eschrig, Akademie Verlag, Berlin, 1991, p. 11. 172. A.D. Becke, J. Chem. Phys., 1993, 98, 5648. 173. B. Hammer, L.B. Hansen and J.K. Nørskov, Phys. Rev. B, 1999, 59, 7413. 174. D.R. Jennison, P.A. Schultz and M.P. Sears, Phys. Rev. Lett., 1996, 77, 4828. 175. J.-S. McEwen, S.H. Payne and C. Stampfl, Chem. Phys. Lett., 2002, 361, 317. 176. A.P.J. Jansen and W.K. Offermans, J. Comput Meth. Sci. Eng., 2002, 2, 351. 177. C.J. Zhang, P. Hu and M.-H. Lee, Surf. Sci., 1999, 432, 305. 178. D. Curulla, R. Linke, A. Clotet, J.M. Ricart and J.W. Niemantsverdriet, Phys. Chem. Chem. Phys., 2002, 4, 5372. 179. C.J. Cramer, ‘Essentials of Computational Chemistry’, Wiley, Chichester, 2004. 180. N.A. Zarkevich and D.D. Johnson, Phys. Ref. Lett., 2004, 92, 255702. 181. D.M. Hawkins, J. Chem. Inf. Comput. Sci., 2004, 44, 1. 182. W.H. Press, B.P. Flannery, S.A. Teukolsky and W.T. Vetterling, ‘Numerical Recipes. The Art of Scientific Computing’, Cambridge University Press, Cambridge, 1989. 183. C.-T. Kao, G.S. Blackman, M.A. Van Hove, G.A. Somorjai and C.-M. Chan, Surf. Sci., 1989, 224, 77. 184. A.P.J. Jansen, C.G.M. Hermse, F. Frechard and J.J. Lukkien, In ‘Computational Science—ICCS 2001’, eds. V.N. Alexandrov, J.J. Dongarra, B.A. Juliano, R.S. Renner and C.J.K. Tan, Springer, Berlin, 2001, pp. 531–540.

Catalysis, 2006, 19, 109–163

163

185. C.G.M. Hermse, F. Frechard, A.P. van Bavel, J.J. Lukkien, J.W. Niemantsverdriet, R.A. van Santen and A.P.J. Jansen, J. Chem. Phys., 2003, 118, 7081. 186. M.J.P. Hopstaken, W.J.H. van Gennip and J.W. Niemantsverdriet, Surf. Sci., 1999, 433–435, 69. 187. C.G.M. Hermse, A.P. van Bavel, M.T.M. Koper, J.J. Lukkien, R.A. van Santen and A.P.J. Jansen, Surf. Sci., 2004, 572, 247. 188. M.T.M. Koper, J.J. Lukkien, N.P. Lebedeva, J.M. Feliu and R.A. van Santen, Surf. Sci., 2001, 478, L339. 189. A.M. Funtikov, U. Stimming and R. Vogel, J. Electroan. Chem., 1997, 428, 147. 190. L.-J. Wan, T. Suzuki, K. Sashikata, J. Okada, J. Inukai and K. Itaya, J. Electroan. Chem., 2000, 484, 189. 191. T. Senna, N. Ikemiya and M. Ito, J. Electroan. Chem., 2001, 511, 115. 192. L.-J. Wan, S.-L. Yau and K. Itaya, J. Phys. Chem., 1995, 99, 9507. 193. J.W. Niemantsverdriet, ‘Spectroscopy in Catalysis’, Wiley-VCH, Weinheim, 2000. 194. J. Stoer and R. Bulirsch, ‘Introduction to Numerical Analysis’, Springer, New York, 1993. 195. H.-P. Schwefel, ‘Evolution and Optimum Seeking’, John Wiley, Chichester, UK, 1995. 196. Z. Michalewicz, ‘Genetic Algorithms þ Data Structures ¼ Evolution Programs’, Springer, Berlin, 1999. 197. D.A. Goldberg, ‘Genetic Algorithms in Search, Optimization, and Machine Learning’, Addison-Wesley, Reading, Massachusetts, 1989. 198. J.R. Koza, ‘Genetic Programming: One the Programming of Computers by Natural Selection’, MIT Press, Cambride, MA, 1992. 199. W. Banzhalf, P. Nordin, R.E. Keller and F.D. Francone, ‘Genetic Programming: An Introduction’, Morgan Kaufmann, San Francisco, CA, 1998. 200. D.H. Parker, M.E. Bartram and B.E. Koel, Surf. Sci., 1989, 217, 489. 201. D. Corne, M. Dorigo and F. Glover, ‘New Ideas in Optimization’, McGraw-Hill, London, 1999. 202. C.T. Campbell, G. Ertl, H. Kuipers and J. Segner, Surf. Sci., 1981, 107, 220. 203. S.M. Barlow and R. Raval, Surf. Sci. Rep., 2003, 50, 201. 204. V. Humblot, M.O. Lorenzo, C.J. Baddeley, S. Haq and R. Raval, J. Am. Chem. Soc., 2004, 126, 6460. 205. M.O. Lorenzo, S. Haq, T. Bertrams, R. Raval and C.J. Baddeley, J. Phys. Chem. B, 1999, 103, 10661. 206. M.O. Lorenzo, C.J. Baddeley, C. Muryn and R. Raval, Nature, 2000, 404, 376. 207. M.O. Lorenzo, V. Humblot, P. Murray, C.J. Baddeley, S. Haq and R. Raval, J. Catal., 2002, 205, 123. 208. R.B. Rankin and D.S. Sholl, Surf. Sci. Lett., 2003, 574, L1. 209. R.B. Rankin and D.S. Sholl, Surf. Sci., 2004, 548, 301. 210. M. Tammaro, M. Sabella and J.W. Evans, J. Chem. Phys., 1995, 103, 10277. 211. D.-J. Liu and J.W. Evans, Phys. Rev. B, 2004, 70, 193408. 212. M.A. Katsoulakis, A.J. Majda and D.G. Vlachos, J. Comp. Phys., 2003, 186, 250. 213. R. Rico-Martı´ nez, C.W. Gear and I.G. Kevrekidis, J. Comp. Phys., 2004, 196, 474.

Methane Decomposition: Production of Hydrogen and Carbon Filaments BY T.V. CHOUDHARYa AND D.W. GOODMANb a ConocoPhillips Company, Bartlesville Technology Centre, Bartlesville 74004, USA b Department of Chemistry, Texas A&M University, College Station, TX77843, USA

1

Introduction

Hydrogen, presently, finds application as a chemical rather than a fuel in commercial operations. However, being a non-polluting source of energy, hydrogen is predicted to be the ‘‘fuel of the future’’.1 One of the most potential applications for hydrogen is to power fuel cells. Major automobile manufacturers are currently working towards developing fuel cell vehicles; such vehicles are expected to significantly curtail the pollution from the transportation sector. Fuel cells, because of their modular nature, can be utilized to provide heat and electricity not only to single homes but also to provide a large amount of electricity to a large grid network. Fuel cells can be broadly classified into two types; high temperature fuel cells such as molten carbonate fuel cells (MCFCs) and solid oxide polymer fuel cells (SOFCs), which operate at temperatures above 923 K and low temperature fuel cells such as proton exchange membrane fuel cells (PEMs), alkaline fuel cells (AFCs) and phosphoric acid fuel cells (PAFCs), which operate at temperatures lower than 523 K. Because of their higher operating temperatures, MCFCs and SOFCs have a high tolerance for commonly encountered impurities such as CO and CO2 (COx). However, the high temperatures also impose problems in their maintenance and operation and thus, increase the difficulty in their effective utilization in vehicular and small-scale applications. Hence, a major part of the research has been directed towards low temperature fuel cells. The low temperature fuel cells unfortunately, have a very low tolerance for impurities such as COx; PAFCs can tolerate up to 2% CO, PEMs only a few ppm, whereas the AFCs have a stringent (ppm level) CO2 tolerance. Methane, due to its abundance and high H/C ratio (highest among all hydrocarbons) is an obvious source for hydrogen. Steam reforming of methane represents the current trend for hydrogen production. Other popular methods Catalysis, Volume 19 r The Royal Society of Chemistry, 2006

164

Catalysis, 2006, 19, 164–183

165

of hydrogen production include autothermal reforming and partial oxidation. However, all these processes involve the formation of large amount of COx as by-product.2–4 Hydrogen generated by these conventional methods can be utilized by the low temperature fuel cells only if COx (CO for PEM and CO2 for AFC) are completely eliminated (to ppm levels) from the stream prior to its introduction into the fuel cell. The process required to eliminate CO from the hydrogen produced in the steam reformer is briefly described below. The steam reformer products containing B10% CO (depending on the feedstock and conditions employed) are passed into water gas shift reactors (WGSs) where CO is reacted with water to form CO2 and hydrogen.5 Generally two WGS reactors are used in series (high temperature and low temperature) to minimize the amount of water. The WGS shift reactors are extremely bulky. Finally, the CO content is reduced to a few ppm in the preferential oxidation reactor (PROX). The hydrogen can be introduced in the fuel cell only after this circuitous procedure of removing CO. AFCs would additionally require removal of CO2 to ppm levels. Also, it is known that high levels of CO2 in the hydrogen stream can be detrimental for the performance of PEM fuel cells.6 Other conventional process of hydrogen production such as partial oxidation and auto-thermal reforming also entail similar procedures for COx removal. Removal of COx to ppm levels from the hydrogen stream makes the process extremely complex and bulky and thereby prohibits the use of the existing hydrogen production technology for use in vehicular and small-scale stationary fuel cell applications. Hydrogen production routes, which do not require complex COx removal procedures, are therefore desired for fuelling low temperature fuel cells. Recently, there has been a great deal of interest in investigating the catalytic decomposition of natural gas (whose major constituent is methane) for production of hydrogen. Since only hydrogen and carbon are formed in the decomposition process, separation of products is not an issue.7 The other main advantage is the simplicity of the methane decomposition process as compared to conventional methods. For example, the high- and low-temperature water-gas shift reactions and CO2 removal step (involved in the conventional methods) are completely eliminated. This review will address the following topics related to the methane decomposition process: (a) fundamentals of methane decomposition, (b) effect of support and promoters on the methane decomposition process (c) alternate reactor design for improving the process yields and (d) catalyst regeneration. Catalyst regeneration is extremely important for the practical application of the clean hydrogen production process; issues related to catalyst regeneration by steam, air and CO2 will be summarized separately. Since hydrogen production via methane decomposition is a relatively new field there are several unresolved issues. This review will attempt to bring forth these issues. Under certain process conditions, high yields of carbon filaments can be obtained on the catalyst during the catalytic decomposition of methane. Currently, there is a great interest in these carbon filaments, as the unique properties exhibited by these materials can be exploited in a number of applications such as catalyst support, energy storage devices, selective adsorption agents and

166

Catalysis, 2006, 19, 164–183

reinforcement materials. In this review we will focus on factors that influence carbon filament formation rates/yields and properties. This review will not address the (relatively low yield) synthesis of specialty single walled carbon nanotubes; a recent review on this topic can be found elsewhere.8 2

Hydrogen Production

The theoretical hydrogen formation reaction via decomposition of methane can be represented as: CH4 - 2H2 þ C

DH1073 ¼ 90.1 kJ/mol(CH4)

This moderately endothermic process results in the formation of 2 moles of hydrogen per mole of methane consumed above a certain threshold reaction temperature. A gradual catalyst deactivation is expected due to the accumulation of carbon on the catalyst. The catalyst can be regenerated by removing the carbon on the catalyst in a separate step. Thus, hydrogen production by this approach involves two distinct steps: (a) catalytic decomposition of methane and (b) regeneration of catalyst. ðaÞ CH4 !2H2 þ C ðbÞ C þ H2 O=O2 =CO2 !COx þ H2 and clean catalyst surface At the outset, this section will address studies related to the catalytic methane decomposition step and then subsequently describe the work undertaken on the combined step-wise reforming (two step) process. 2.1 Catalytic Decomposition of Methane for Hydrogen Production. – The methane decomposition reaction for hydrogen production has garnered considerable interest in the past 4–5 years. The recent interest in this approach for producing hydrogen stems from the stringent requirement of CO-free hydrogen for the proton exchange membrane fuel cells. For vehicular and small scale stationary applications, it is necessary that the fuel reformer be compact; this is difficult for the conventional processes since high CO conversion efficiencies require large water gas shift reactors. Recent studies have addressed different issues such as methane decomposition fundamentals, support/promoter effects and reactor design. The ensuing discussion will show that while some interesting issues about catalytic methane decomposition (as a method for generating pure hydrogen) have been uncovered, a significant amount of work still needs to be undertaken for better understanding this process. 2.1.1 Methane Decomposition Fundamentals. The fundamentals of methane decomposition have been extensively investigated on model-single crystal catalysts;9,10 an exhaustive review on this subject can be found elsewhere.11 Herein, only the studies undertaken on the fundamentals of methane

Catalysis, 2006, 19, 164–183

167

decomposition reaction on high surface area catalysts will be described. Otsuka and co-workers used hydrogen-deuterium exchange studies to investigate the methane decomposition reaction mechanism over Ni/SiO2.12 An isotopic effect was observed for the CH4/CD4 decomposition on the catalyst. Furthermore, substituted H–D methanes were not observed when a CH4–CD4 mixture was decomposed over the Ni/SiO2 catalyst. Based on this the authors suggested that the first C–H bond cleavage was the rate determining step for the decomposition of methane to carbon and hydrogen. In line with this theory, the authors also observed a reverse isotopic effect between hydrogen and deuterium when the carbon deposited on the catalyst was hydrogenated back to methane. The methane decomposition mechanism was further studied by performing the following set of experiments sequentially: (i) decomposition of 12CH4 (ii) decomposition of 13CH4 (iii) hydrogenation of deposited carbon. The studies showed that the carbon that was deposited last was hydrogenated first; thus indicating that there was no significant scrambling between the carbon atoms. Using an array of catalyst characterization techniques, the same group further investigated the structural changes of the Ni species in the Ni/SiO2 catalyst during the methane decomposition reaction.13 Prior to the reaction, the Ni species on the catalyst were in the metallic state. The Ni metal particles were found to aggregate [X-Ray Diffraction (XRD) studies] as soon the catalyst was contacted with methane. Following this initial aggregation at the onset of the methane decomposition reaction, no significant change in the structure of the Ni species was observed until towards the end of the reaction. During the rapid catalyst deactivation stage, Ni K-edge X-Ray Absorption Near Edge Structure (XANES) studies indicated the formation of Ni carbide species. As will be discussed later, along with Ni/SiO2 the Ni/TiO2 catalyst is also a promising catalyst for the methane decomposition reaction. Zein et al.14 have very recently investigated the kinetics of methane decomposition on a Ni/TiO2 catalyst. Their studies suggested a first order rate law for the decomposition reaction and activation energy of 60 kJ/mol. Interestingly, their studies indicated that the methane adsorption step on the catalyst surface was the rate determining step. This is in contradiction to studies on the Ni/SiO2 catalyst wherein,12 the scission of the first C–H bond was proposed as the rate determining step for the methane decomposition reaction. Carbon-based catalysts have also been considered for the methane decomposition reaction.15 Yoon and co-workers have recently investigated the kinetics of methane decomposition on activated carbons as well as on carbon blacks.16,17 In case of activated carbons the authors observed mass transport effects in the catalyst particles and also significant pore mouth plugging. The reaction order was found to be 0.5 and the activation energy was found to be B200 kJ/mol for the different activated carbon samples. On the other hand, for

168

Catalysis, 2006, 19, 164–183

the carbon black catalysts a reaction order of near unity was observed and the activation energy was lower than that observed for the activated carbons. Recent studies have shown that unreduced Ni catalysts,18 depending on the synthesis procedure, are also efficient for the hydrogen production reaction. Since most fundamental studies have been undertaken on reduced Ni catalysts (Ni0), it will be interesting to investigate methane decomposition fundamentals on unreduced Ni catalysts. 2.1.2 Effect of Support. Methane decomposition on various Ni-supported catalysts has been extensively investigated by the Goodman group.19,20 These studies were directed towards understanding the role played by the support in determining the nature of surface carbon and CO content in the hydrogen stream. Time on stream methane activity studies at a reaction temperature of 823 K revealed comparable initial methane decomposition activities for the Ni/ HY, Ni/HZSM-5, Ni/SiO2 and the Ni/SiO2/Al2O3 catalysts.20 However, unlike the Ni/SiO2, Ni/HY and Ni/SiO2/Al2O3 catalysts, which showed methane conversion activity for several hours, a rapid deactivation (in ca. 1 h.) was observed in case of Ni/HZSM-5. Transmission Electron Microscopy (TEM) images of Ni/HZSM-5 catalyst after the reaction showed an encapsulating type of carbon (Figure 1), which explained the rapid deactivation of the catalyst. On the other hand, carbon filaments (Figure 2) were observed in case of Ni/SiO2, Ni/HY and Ni/SiO2/Al2O3 catalysts.19 The presence of the Ni particle at the apex of the carbon filaments elongates the life-time of the catalyst. It is

Figure 1 TEM image of Ni/HZSM-5 after methane decomposition at 823 K19

Catalysis, 2006, 19, 164–183

169

Figure 2 TEM image of carbon filaments formed after methane decomposition at 823 K on Ni/HY19

noteworthy that while there was rapid deactivation of the Ni/HZSM-5 catalyst at 823 K, the catalyst had a much greater stability at 723 K. TEM images revealed the presence of carbon filaments at lower temperatures (723 K), which resulted in greater catalyst stability for methane conversion. X-Ray photoelectron Spectroscopy (XPS) of the spent samples showed the presence of carbidic and graphitic carbon at low reaction temperatures (r723 K), whereas only the graphitic species were observed at higher temperatures. This is in excellent agreement with studies on model Ni catalysts (single crystal).21 Previously it has been noted that methane decomposition may lead to CO formation via reaction of the carbonaceous residue with the oxygen of the oxide support.22 Since the CO content in the hydrogen stream is a critical parameter for the PEM fuel cells, it is necessary to achieve an accurate quantification of CO (ppm levels). Although this aspect has been neglected in most studies, in our studies particular attention was devoted towards the CO quantification issue.19,20 Quantitative estimation of CO (to ppm levels) was achieved by utilizing the analysis system showed in Figure 3. The effluents from the reactor were first introduced in the thermal conductivity detector (TCD) for detection of hydrogen and methane; Ar was employed as a carrier gas. Analysis of CO was carried out by converting it into methane in a methanizer prior to its introduction in a flame

170

Catalysis, 2006, 19, 164–183

Figure 3 Schematic of the experimental set-up employed to study the methane decomposition reaction; methanizer coupled with the FID was used to quantify ppm levels of CO in the hydrogen stream20

ionization detector (FID). Essentially 100% conversion efficiency (CO to methane) was achieved by operating the methanizer at 673 K with large amounts of hydrogen. This was accomplished by routing all of the hydrogen flow to the FID through the methanizer. An auxillary flow of carrier gas was employed to obtain the optimum carrier/hydrogen ratio for maximum detector sensitivity. The CO formation rates showed a common trend for all the catalysts; high initial rates that rapidly decreased with time and finally stabilized.20 The rate of CO formation was found to increase with increasing temperatures and decrease with increasing gas space velocities (decrease in contact time). The space velocity effect was especially pronounced in the initial period of the methane decomposition reaction. The CO content in the hydrogen stream at a reaction temperature of 823 K was ca. 50, 100 and 250 ppm for Ni/SiO2, Ni/SiO2/ Al2O3 and Ni/HY respectively after the CO-formation rates had stabilized. Diffuse Reflectance Infra-Red Spectroscopy (DRIFTS) studies showed the presence of approximately 55 m-moles of hydroxyl species at 823 K on 0.1 g of the Ni/SiO2 catalyst. The hydroxyl groups on the support were held responsible for CO formation during the methane decomposition reaction. The authors proposed that supports with a greater content of reactive hydroxyl groups at a given methane decomposition temperature would show higher CO formation. It should be noted that some of the support hydroxyl groups may also be involved in CO2 formation, further complicating the issue. Since CO2 is relatively benign to the PEM fuel cell, the CO2 content was not measured in our study.19 However, the knowledge of CO2 content is important to estimate the effect of support hydroxyl groups on the CO formation. Extensive studies involving different catalyst supports (effect of temperature on hydroxyl groups) coupled with methane decomposition studies (quantitative detection of ppm levels of CO and CO2) under different process conditions will be needed to obtain a satisfactory understanding about the influence of support on the CO formation.

Catalysis, 2006, 19, 164–183

171

Takenaka et al. have investigated the effect of several supports (Ni-based) on the activity and stability for the methane decomposition reaction.23 While the catalysts with SiO2, TiO2 and graphite supports showed high activities and long life-times, the catalysts with Al2O3, MgO and SiO2–MgO2 supports were found to be inactive. Characterization of the catalysts by XRD and XANES revealed that in case of the active catalysts Ni was present in the metallic state, whereas in case of the inactive catalysts Ni formed an oxide compound with the support. Studies on catalysts with different silica supports showed that the catalyst activity/stability was also dependent on the pore structure of the support; the silica support devoid of pore structure was found to enhance the catalyst activity/stability. While a large number of studies have been undertaken related to the effect of support on hydrogen production activity/stability, the influence of Ni particle size has not been studied in detail. Otsuka and co-workers have recently reported that the Ni metal particles within a specific size range (60–100 nm) for a Ni/SiO2 showed longest catalytic life for the methane decomposition reaction.24 However, more information is desirable on this topic; for example it will be interesting to systematically investigate the effect of particle size on the hydrogen production rate. Since the hydrogen production rate and the catalyst stability are both important to the practical application of this process, it is important that future studies address both these aspects simultaneously. 2.1.3 Bimetallic Catalysts and Promoters. Shah and co-workers compared the methane decomposition activities and stabilities for monometallic (Pd, Mo or Ni) and bimetallic M–Fe (M ¼ Pd, Mo or Ni) catalyst above 673 K.25 Their studies showed that the bimetallic M–Fe catalysts produced hydrogen at significantly higher rates than the monometallic (M) catalysts. The Pd– Fe catalyst was found to be the most active methane decomposition catalyst at 973 K. Chen et al. have investigated the effect of Cu content on the methane decomposition activity and stability of bimetallic Ni–Cu/Al2O3 catalysts.26 The 2Ni–1Cu–Al catalyst was found to be superior to the 15Ni–3Cu–2Al, 3Ni– 3Cu–2Al and 1Ni–1Cu–1Al catalysts; high activity required optimized levels (not too high and not too low) of Cu in the catalyst. The authors believed that the introduction of Cu (especially at high levels) transformed the catalyst into a quasi liquid state between 973 and 1013 K thus making them less stable. Similar to this study, Li and co-workers (who investigated a series of Ni–Cu–Nb2O5 catalysts) also observed that optimized levels of Cu were required to maximize hydrogen yields. The best catalyst (65Ni–25Cu–5Nb2O5) gave a yield of 7274 mol H2/mol Ni.27,28 The methane decomposition reaction is severely constrained by equilibrium. A few studies have also been undertaken to circumvent the equilibrium constraints.29,30 Otsuka and coworkers used the addition of CaNi5 to Ni/SiO2 for cheating equilibrium.29 The physical mixture of CaNi5 and Ni/SiO2 showed greater than equilibrium methane (decomposition) conversion due to the hydrogen absorption property of CaNi5.

172

Catalysis, 2006, 19, 164–183

2.1.4 Reactor Design. The studies discussed until this point were all undertaken in conventional fixed bed reactors; this segment focuses on methane decomposition studies in membrane and fluidized bed reactors. Application of membrane reactors for hydrocarbon reforming is known to increase the yield of hydrogen.31 In line with this, Ishihara et al. observed greater than equilibrium methane conversions using a 90%Pd10%Ag hydrogen permeable membrane reactor during the decomposition of methane over Ni/SiO2 catalyst.30 The permeated hydrogen was swept by argon (Ar) gas and an increase in the Ar flow-rate was found to enhance the hydrogen production. The positive effect in the hydrogen yields was more pronounced at higher temperatures (4700 K) due to higher hydrogen permeability at these temperatures. A constant conversion of 70% (10% CH4 in N2) was observed at 773 K over a time period of 60 h. Contact times larger than 50 g-cat h mol1 and sweep Ar flow rates higher than 200 ml min1 were found to be favorable for the process. Utilization of sweep gas on the permeate side results in dilution of the permeated hydrogen; this can be a serious limitation for producing pure hydrogen. Also, membrane reactors have a tendency to get fouled; hence when considering a membrane reactor it is important to address the fouling issue. Recently Weizhong and co-workers have used a two stage fluidized bed reactor to study the methane decomposition reaction over a Ni–Cu/ Al2O3 catalyst.32 The temperature in the lower stage of the reactor was held constant at 773 K, while the temperature in the upper stage was controlled between 773 K and 1123 K. Operation at higher temperatures is desired as it increases the hydrogen production rates. Unlike the fixed bed reactor/single stage fluidized bed reactor studies, wherein a rapid catalyst deactivation was observed at 1123 K, the catalyst showed significantly lower deactivation rate in the two stage fluidized bed reactor (upper stage at 1123 K and lower stage at 773 K). The authors believed that the two stage temperature operation decreased the disparity in carbon production and diffusion rates (which was responsible for rapid catalyst deactivation in fixed bed/single stage fluidized bed reactors operating at high reaction temperatures). While this is an interesting concept, the suggested reactor design may be too complex for practical operation. 2.2 Step-wise Methane Reforming: Regeneration Issues. – The catalyst is gradually expected to deactivate due to accumulation of carbon on the catalyst surface during the methane decomposition reaction. This means that after a certain reaction time period, the catalyst has to be regenerated or replaced with a new catalyst (expensive approach). The latter approach could be used for synthesizing carbon filaments with high yields. However, frequent replacement of catalyst is not a practical approach for hydrogen generation. It is therefore essential to employ the regeneration strategy for hydrogen production. The hydrogen production process therefore consists of two steps (a) methane decomposition (Step I) and (b) catalyst regeneration (Step II). This segment, which will focus on the combined hydrogen production process (stepwise reforming), has been sub-divided by the type of regeneration gas used

Catalysis, 2006, 19, 164–183

173

(steam/air/CO2). Issues related to regeneration efficiency and energy efficiency will be addressed in this section. 2.2.1 Step-wise Reforming with Steam Regeneration. The step wise methane steam reforming process, which involves the catalytic decomposition of methane in step I and steam regeneration in step II, has been investigated by our group at relatively low reaction temperatures.33,34 The studies were performed in a pulse mode so as to ensure accurate quantitative analysis of the carbon removed in the regeneration step. No catalyst deactivation was observed in this study; in contrast consequent pulsing of methane without intermittent regeneration (i.e. without Step II) showed an exponential deactivation of the catalyst. The amount of surface carbon removed varied from 92% to 100% (of the amount deposited in Step I) in the various cycles; 95% of the carbon was removed on an average. The CO content in the hydrogen produced in step I was less than 20 ppm. The average amount of hydrogen produced per mole of methane consumed in Step I was 1.1, thus indicating the presence of hydrocarbonaceous residue on the catalyst surface. Recent neutron vibrational studies have revealed the presence of methylidyne (CH), vinylidene (CCH2) and ethylidyne (CCH3) species on Ni–based surfaces after methane dissociation at low temperatures (o673 K).35 The ethylidyne species were found to be less stable than the vinylidene and methylidyne species with increasing methane decomposition temperatures. Amiridis and co-workers employed a continuous flow reactor to study the step-wise steam reforming process.36 In the first step, methane was decomposed over 15% Ni/SiO2 catalyst at 923 K and space velocity of 30000 h1 for 3 h. In the second step, the catalyst was regenerated with steam until no hydrogen was observed in the product stream. Ten reaction cycles performed as described above showed no significant decrease in catalytic activity.36 XRD patterns collected after individual cycles suggested that a large fraction of the carbon deposited in Step I was removed in the regeneration step. It is noteworthy that there was no significant change in the crystallite size of Ni during the reactionregeneration cycles. In agreement our recent studies on a pulse mass analyzer balance have indicated that ca. 75% of the surface carbon (deposited in Step I on Ni/Al2O3/SiO2 at 823) can be removed during the steam regeneration step at 823 K in the continuous flow mode.20 Choudhary and co-workers have investigated the step-wise steam reforming process in two parallel reactors;37,38 methane decomposition and carbon gasification were carried out simultaneously by switching a methane containing feed and steam containing feed between the two reactors at pre-determined time intervals. Amongst the various Ni supported catalysts (ZrO2, MgO, ThO2, CeO2, UO3, B2O3, MoO3, HZSM-5, Hb, NaY, Ce(72)NaY and Si–MCM–41) screened for this cyclic reaction, Ni/ZrO2 and Ni/Ce(72)NaY were found to be the most suitable catalysts. The degree of carbon removal by steam increased significantly on increasing the regeneration temperature from 773 K to 873 K.39 Since issues related to pressure drop are extremely important for practical operation, the pressure drop across the reactor was

174

Catalysis, 2006, 19, 164–183

also monitored in this study. An exponential increase in pressure drop was observed across the reactor during the methane decomposition reaction after a certain threshold of carbon deposition was exceeded. This study highlights the importance of optimizing the switch time between the methane decomposition reaction and regeneration reaction (optimizing run lengths for the two processes). Both methane decomposition and regeneration by steam are endothermic processes, and hence the step-wise steam reforming (like conventional methane steam reforming) is expected to be an energy intensive process. The hydrogen generation process is expected to be more energy efficient for air/oxygen based regeneration; studies related to step-wise air/oxygen methane reforming are summarized in the next section. 2.2.2 Step-wise Reforming with Air/Oxygen Regeneration. The step-wise reforming with air has been employed in the past by Universal oil products (Hypro Process).40 The process utilized a 7% Ni/Al2O3 catalyst in a fluidized bed reactor-regenerator. Catalytic decomposition of methane occurred in the fluidized bed reactor at B1150 K followed by regeneration with air at B1475 K in the fluidized bed regenerator. The product stream consisted of 93–95% hydrogen and unreacted methane. We have also recently investigated the reaction/regeneration (by air) cycles on Ni/HZSM-5 at 723 K in a fixed bed reactor.20 In this case, the methane decomposition step was performed for 1 h following which the catalyst was regenerated using an oxidation–reduction cycle. There was no apparent decrease in catalytic activity throughout the 12 cycles studied at 723 K. Similarly, Zein and Mohamed observed stable catalytic activity for six methane decomposition-regeneration cycles on a 15MnOx–20NiO/TiO2 catalyst.41 Monnerat et al. have investigated the methane decomposition and air regeneration process over a Ni gauze catalyst (Ni-grid with Raney type outer layer).42 Their studies revealed an optimal reaction performance when the cycle consisted of 4 min of reaction period followed by 4 min of regeneration period. In a second study, Mirodatos and co-workers investigated the same process on a Pt/CeO2 catalyst at 673 K under forced unsteady-state conditions.43 No CO was detected in the products under these conditions in either the cracking or the oxidative regeneration steps. Utilization of air can effectively increase the energy efficiency of the process as the exothermic regeneration step can be employed to drive the endothermic hydrocarbon decomposition step. However on the flip side, air regeneration may lead to sintering of the catalyst especially in fixed bed reactors. Villacampa et al. investigated several reaction-regeneration cycles on a co-precipitated Ni/ Al2O3 catalyst.44 Although, the initial activity for hydrogen was recovered after each regeneration step, the regenerated catalyst had a significantly higher deactivation rate. This effect was most prominent after the first catalyst regeneration. The increase in deactivation rate was attributed to the sintering of Ni. Otsuka and co-workers, on the other hand, observed an excellent stability for Ni/Al2O3 and Ni/TiO2 and catalysts for the step wise reforming

Catalysis, 2006, 19, 164–183

175

process.45 This may be related to the following differences in the catalyst reactivation procedures between the two studies: (i) The regeneration temperature was significantly lower in the latter study (smaller exotherm expected). (ii) Also no reduction step was employed in between the regeneration and the reaction step in the latter study (smaller exotherm expected). Based on this, it is apparent that the exothermic catalyst reactivation reactions need to be appropriately controlled to avoid Ni sintering/catalyst deactivation. While steam regeneration appears to be more benign for the catalyst life (several reaction-regeneration cycles), regeneration by air is more energy efficient. It may therefore be interesting to perform the regeneration using a combination of steam and air (oxy-steam regeneration). This topic deserves attention in future investigations. 2.2.3 Step-wise Reforming with CO2 Regeneration. While most of the work to date has focused on regeneration by steam/air/oxygen, few studies involving regeneration by CO2 have also been undertaken. Takenaka et al. investigated the step-wise reforming reaction with CO2 on Ni/SiO2, Ni/Al2O3 and Ni/TiO2 catalysts; the methane decomposition reaction was carried out at 823 K, while the carbon gasification by CO2 was performed at 923 K.46,47 The supports played a crucial role in determining the hydrogen production stability for the process. A gradual decrease in the hydrogen yield (total hydrogen produced in each cycle) was observed for the Ni/SiO2 catalysts during consecutive reactionregeneration cycles. However, there was no decrease in the hydrogen yield for the Ni/Al2O3 and Ni/TiO2 catalysts for several consecutive reaction regeneration cycles. The author claimed a 495% conversion of the carbon to CO in the regeneration step. The structural changes of Ni species for the different catalysts occurring during the consecutive reaction-regeneration cycles were monitored by XANES, XRD and Scanning Electron Microscopy (SEM) to enhance the understanding of the role played by the different supports.48 Based on these studies the authors arrived at the following conclusions: (i) Ni particles in the 60–100 nm range are most effective for the methane decomposition reaction. (ii) While the fresh Ni/SiO2 catalyst had Ni particles in the 40–100 nm range, consecutive reaction-regeneration reactions led to sintering/agglomeration of Ni particles (4200 nm were observed). The Ni sintering was responsible for the inferior performance of the Ni/SiO2 catalyst. (iii) The Ni/Al2O3 and Ni/TiO2 catalyst on the other hand maintained the optimized size distribution of Ni particles throughout the consecutive reaction-regeneration cycles and hence were superior catalysts. Due to its highly endothermic nature, regeneration by CO2 is unfortunately a very energy intensive process.

176

3

Catalysis, 2006, 19, 164–183

Production of Carbon Filaments by Catalytic Methane Decomposition

Carbon filaments (CF), due to their unique properties, have the potential to be used in several applications such as selective adsorption agents, in energy storage devices, catalyst supports and reinforcement materials.49–56 In the last few years significant efforts have been directed towards optimization of the process condition for CF formation. Researchers have investigated CF formation on different metal-based catalysts. In this section, the studies related to production of CF (mainly issues related to rates/yields and quality) have been categorized based on the metals used for catalyzing the methane decomposition reaction. This section will not address issues related to the CF growth mechanism; detailed information about the CF growth mechanism may be found in reference.53 3.1 Ni-based Catalysts. – Ni-based catalysts have been by far the most investigated catalysts for the CF formation via methane decomposition. This may be attributed to the high CF yield obtained on Ni-based catalysts. CF yield is defined as the total amount of CF formed per gram of catalyst at complete deactivation. Since the catalyst has to be essentially replaced for subsequent CF formation, from an economics point of view it is desirable to achieve extremely high CF yields. Shaikhutdinov and co-workers have exhaustively investigated CF formation yields on co-precipitated Ni-alumina and Ni-Cu-alumina catalysts.57,58 The amount of CF formed per gram of the catalyst was found to increase with increasing Ni content in the Ni-alumina catalyst. However, the CF yield was found to be radically small for pure Ni powder.57 In good agreement, studies by Toebes et al. also showed negligible CF formation from methane decomposition on unsupported Ni catalysts.59 Low CF yields from methane decomposition on unsupported Ni catalyst have been attributed to the presence of large Ni particles (50–1000 nm) with low index planes, since low index planes are incapable of dissociating the unreactive methane molecules. Li et al. employed a Ni-alumina catalyst prepared from Feitknecht compound for maximizing CF yields from methane.60 Similar to the work by Shaikhutdinov and co-workers,57 the total amount of CF formed was found to increase with increasing Ni content of the catalyst. The total amount of CF formed was dependent on the reduction temperature as well as the reaction temperature. Although the rate of CF formation increased at higher temperature there was a decrease in the total yield of CF due to rapid deactivation of the catalyst. Ermakova and co-workers manipulated the Ni particle size to achieve large CF yields from methane decomposition.61,62 The Ni-based catalysts employed for the process were synthesized by impregnation of nickel oxide with a solution of the precursor of a textural promoter (silica, alumina, titanium dioxide, zirconium oxide and magnesia). The optimum particle size (10–40 nm) was obtained by varying the calcination temperature of NiO. The 90% Ni–10% silica catalyst was found to be the most effective catalyst with a total CF yield of 375 gCF/gcat. XRD studies by the same group on high loaded Ni-silica

Catalysis, 2006, 19, 164–183

177

showed that the nickel particles seemed to ‘‘self organize’’ to the optimum size (30–40 nm) during the course of the methane decomposition reaction, i.e. smaller particles underwent sintering to form larger particles whereas larger particles were found to undergo dispersion.63 The authors proposed that deactivation occurred when the distance separating the metal particles (present on filament ends) increased to an extent such that reversible merging and dispersion of the Ni particles was prevented. Several studies have considered the addition of Cu to Ni-based catalysts for enhancing the methane decomposition CF yields. Studies by Shaikhutdinov and co-workers showed that the addition of Cu decreased the rate of CF formation but greatly increased the stability of the catalyst.57 The maximum CF yield (240 gCF/gcat) was obtained for a 3% Cu-87% Ni catalyst. Li et al. also studied the doping effects of Cu on the CF formation.64 In this work, addition of small amounts of Cu not only increased the total amount of CF formed, but also increased the growth rate at 873 K. This was unlike the work by Shaikhutdinov et al. where addition of small amount of Cu had decreased the growth rate but increased the overall CF yield by significantly increasing the life time of the catalyst.57 Addition of large amounts of Cu had a detrimental effect on the performance of the catalyst at relatively low methane decomposition temperatures, however the high Cu content catalyst was found to be the most effective catalyst for CF formation at higher temperature when methane was co-fed with hydrogen. Reshetenko et al. also used the Feitknecht compound as a precursor for preparing copper(8–45%) promoted Ni catalysts and carried out a detailed investigation of the methane decomposition reaction.65 The highest CF yield (525 gCF/gcat) was obtained on the 75Ni-15Cu/ Al2O3 catalyst at 898 K. In general, Ni-based catalysts in their reduced (Ni0) forms are used for CF generation from methane. However, some recent studies have shown that it may not be necessary to pre-reduce the Ni catalysts. Qian and co-workers observed methane conversions approaching equilibrium on an unreduced Ni– Cu/Al2O3 catalyst in a fluidized bed reactor.66 The corresponding methane conversion for the reduced catalyst was significantly lower from the onset of the reaction. The CF yields were also considerably higher for the unreduced catalysts. The authors believed that in situ reduction of the lattice oxygen (in case of the unreduced catalyst) provided energy for the endothermic methane decomposition process. Also, since some the hydrogen produced in the reaction was consumed in situ, this assisted in shifting the equilibrium in the direction of CF formation. This is in contrast to recent fixed bed reactor studies by Suib and co-workers, wherein a significantly higher initial methane conversion for the reduced Ni catalyst (40% Ni/SiO2 catalyst prepared from nitrate salts) was observed as compared to the unreduced version.18 However, these studies do suggest that unreduced catalysts depending on the synthesis procedure may provide large CF yields. Since the pre-reduction treatment is an important process parameter for CF formation, it would be worthwhile to obtain a detailed understanding apropos the effect of catalyst reduction on methane conversions/CF yields.

178

Catalysis, 2006, 19, 164–183

The large surface area of CF makes it an attractive candidate for catalyst support material. Avdeeva and co-workers used CF with different textural properties as supports for Ni to study the methane decomposition reaction.67,68 Ni supported on the CF, which was obtained from methane decomposition on a Ni–Cu/Al2O3 at 898 K, showed the highest yield for secondary carbon (224 gCF/gNi). Highly porous CF supports were found to be most effective for the secondary generation of CF. Compared to studies related to CF optimization fewer studies have been undertaken related to the nature/quality of the CF formed from the methane decomposition process. Studies by Baker and co-workers on Ni and Ni–Cu catalysts revealed that the structural characteristics of CF were strongly influenced by the nature of the catalyst particles.69,70 While particles rich in Ni resulted in the formation of smooth filaments, Cu-rich alloy particles gave rise to filaments having a spiral conformation. The filament size (25–100 nm) was found to be strongly dependent on particle size of the catalyst. The morphology and surface structure of CF (F1) produced on Ni-alumina catalysts and carbon (F2) produced in case of Ni-Cu-alumina catalysts were studied by scanning tunneling microscopy (STM) and High-Resolution Transmission Electron Microscopy (HRTEM) by Shaikhutdinov et al.58 The carbon surface of the filaments was found to be rough and was formed by misoriented edge planes of graphite crystallites. In case of the F1 the basal graphite planes lay inclined to the fiber axis, whereas the basal planes were perpendicular to the filament axis for F2. HRTEM micrographs indicated a closed layer structure on the edges for F2, which was contrary to the open structure observed for graphite crystallites. Kuvshinov and co-workers observed that the CF texture could be modified by changing the CH4:H2 feed ratio.71 Their work also suggested that the surface area of carbon growth centers was an important parameter for determining the maximum CF yield on Ni catalysts using pure methane. From a practical view point it is essential to have an excellent understanding of the CF yields in relation to desired CF properties (surface area, structural/ mechanical properties etc). Unfortunately this aspect of CF production has been seriously neglected. In a couple of studies, the surface area of the CF formed during methane decomposition process has been related to the CF yields.65,72 These studies clearly show that the surface area of the CF and CF yields are both strongly dependent (however in a different way) on the catalyst and process conditions. The BET surface area of the CF obtained on the catalyst, which showed highest CF yield (catalyst: 75Ni–15Cu/Al2O3 and yield: 525 gCF/gcat), was 233 m2/g. On the other hand, the CF with highest surface area (286 m2/g) was obtained with a 45Ni–45Cu/Al2O3 catalyst, which had a corresponding CF yield of only 118 gCF/gcat. While the CF BET surface area was found to decrease with increasing methane decomposition temperatures, the CF yield was found to pass through a maximum with increasing reaction temperature.72 The above study clearly demonstrates the importance of optimizing the CF yields and CF quality simultaneously.

Catalysis, 2006, 19, 164–183

179

3.2 Fe and Co-based Catalysts. – Catalysts which are not Ni-based have low CF formation rates/yields. However, it is interesting to consider metals such as Fe and Co since the properties of the CF depend on the metal employed for catalyzing the methane decomposition reaction. Ermakova and co-workers have investigated the methane decomposition reaction on un-supported Fe powder as well as supported Fe catalysts.73 With the exception of SiO2 support, the other supports (Fe/Al2O3, Fe/Al2O3, Fe/ TiO2) gave similar CF yields as unsupported Fe powder (17 gCF/gcat). The interesting behavior of Fe/SiO2 system motivated them to investigate it in greater details.74 Their studies showed that the silicates depending on their concentration in the catalyst could have either a promoting effect or an inhibiting effect on CF formation. The Fe/SiO2 catalyst with optimal silicate content showed a yield of 45 gCF/gFe. It should be noted that although the yield of CF on Fe-based catalysts is small, Fe-based catalysts produce predominantly thin walled CF (considered to be more valuable than other CF). Bennissad et al. investigated CF formation on Fe-based catalysts using CH4– H2 mixtures at temperatures to 1423 K.75,76 Under these conditions thicker fibers (ca. 1 m) were obtained, but when heating was stopped at 1323 K, the normal structure of CF was observed. Shah et al. investigated CF formation on bimetallic Fe–M (M ¼ Pd, Mo or Ni) catalysts.25 The bimetallic catalysts were found to be more active for the CF formation than the corresponding monometallic catalysts. While only CF formation was observed at the methane decomposition reaction temperature range of 973–1073 K, amorphous carbon and carbon flakes were observed concomitant with CF at reaction temperatures above 1173 K. Otsuka and co-workers have recently investigated the structural changes of Fe species and the nature of the CF formed during the methane decomposition reaction on Fe2O3/Al2O3 and Fe2O3/SiO2 catalysts.77 XANES studies showed that during the methane decomposition reaction the smaller sized Fe particles were transformed into Fe3C while the larger particles were converted into carbon atom saturated g-Fe species. The supports had a profound effect in determining the nature of the CF formed in the process; multi-walled CF and chain like carbon fibers were formed on Fe/Al2O3, while CF composed of spherical carbon units were formed along with chain like carbon fibers on Fe2O3/SiO2 catalysts. Avdeeva and co-workers studied the methane decomposition reaction on Coalumina catalysts with varying concentration of Co.78 The CF formation was found to be maximized at 60–75% content of Co. No induction period was observed for the Co-alumina catalysts, which was contrary to their previous experience involving Ni-alumina catalysts.57 Also in this case a different variety of filaments (not observed on Ni-based catalysts) with hollow-like core morphology were observed. Takenaka et al. have recently investigated the effect of supports on CF formation for Co-based catalysts.79 The Co/Al2O3 and Co/ MgO catalysts were found to be superior to the Co/SiO2 and the Co/TiO2 catalysts for CF formation. Based on catalyst characterization studies the

180

Catalysis, 2006, 19, 164–183

authors claimed that the 10–30 nm range for Co particles was preferred for CF formation. The authors also found that the temperature had a significant influence in determining the nature of the CF. Multi-walled CF were formed in the temperature range of 873–973 K, whereas helically coiled and bamboo-like CF were preferentially formed at 1073 K. Smith and co-workers investigated the effect of metal support interaction on the CF formation on a series of Co-silica catalysts.80 The metal support interaction was manipulated by addition of either BaO, La2O3 or ZrO2 to silica. The rate of catalyst deactivation was found to increase with the increase in the metal support interaction. Competition between CF formation and encapsulating carbon formation controlled the catalyst deactivation rate. In case of the catalysts with high metal support interaction, the encapsulating carbon formation was dominant and hence led to a rapid deactivation of the catalyst. It is unfortunate that very few studies have been undertaken which relate the methane decomposition process conditions and CF yields to the CF quality (surface area, structure/texture, mechanical strength etc). From a process application view point it is extremely important to comprehensively investigate this aspect of CF formation. 4

Concluding Remarks

Catalytic methane decomposition has received considerable attention in recent years. The reaction has been investigated for two main applications (a) production of hydrogen and (b) synthesis of carbon filaments. The important conditions necessary for clean hydrogen production are as follows: (i) High conversion of methane (to COx-free hydrogen) to avoid costly product separation. (ii) Absence of pressure drop issues across reactor (iii) Effective regeneration of catalyst (iv) Stable life of catalyst over several cycles While the methane decomposition step has been extensively investigated, unfortunately less attention has been devoted to other aspects. To avoid a pressure drop it is important to optimize the run lengths for the methane decomposition step and the regeneration step. In order to assess the commercial viability of the process, it is important to study the process over several cycles (4100). If the metal-based catalysts are not reduced in between the decomposition and deactivation stage, significant amount of CO may be formed due to reaction of the metal-oxide with methane during the initial stages of the methane decomposition reaction. Most studies have employed argon gas with a thermal conductivity detector to analyze product gases. Such an analysis procedure is not astute for accurate quantification of ppm levels of CO/CO2. Special attention should be paid towards this analysis as the main advantage claimed in this process is the production of clean hydrogen (without further need for purification).

Catalysis, 2006, 19, 164–183

181

Following are the important conditions necessary for carbon nanotube synthesis (i) High yield of CF (ii) CF with desired properties While several studies have been undertaken to optimize the CF yield, only a few studies have addressed the quality factor for CF. It is important to tune the synthesis procedures such that CF yields/properties are consistent with the desired applications. References 1. Web-site: http://www.eere.energy.gov/RE/hydrogen.html. 2. J.N. Armor, Appl. Catal., 1999, 176, 159. 3. V.R. Choudhary, A.S. Mamman and S.D. Sansare, Angew Chem. Int. Ed. Engl., 1992, 31, 1189. 4. J.R. Rostrup-Nielson, Catal. Today, 1993, 18, 305. 5. J.M. Thomas and W.J. Thomas, ‘‘Principles and Practice of Heterogeneous catalysis’’, VCH, Weinheim, 1997, p. 545. 6. N. Vanderborgh, Private Communication, 2000. 7. T.V. Choudhary, C. Sivadinarayana and D.W. Goodman, Chem. Eng. J., 2003, 93, 69. 8. M. Terrones, Ann. Rev. Mater. Res., 2003, 33, 419. 9. Y.-N. Wang, R.G. Herman and K. Klier, Surf. Sci., 1992, 279, 33. 10. T.V. Choudhary and D.W. Goodman, J. Mol. Catal., 2000, 163, 9. 11. T.V. Choudhary, E. Aksoylu and D.W. Goodman, Catal. Rev. Sci. Eng., 2003, 45, 151. 12. K. Otsuka, S. Kobayashi and S. Takenaka, J. Catal., 2001, 200, 4. 13. S. Takenaka, H. Ogihara and K. Otsuka, J. Catal., 2002, 208, 54. 14. S.H. Zein, A.R. Mohamed and P.S. Sai, I&EC, 2004, 43, 4864. 15. N.Z. Muradov, Energy and Fuels, 1998, 12, 41. 16. M.H. Kim, E.K. Lee, J.H. Jun, S.J. Kong, G.Y. Han, B.K. Lee, T.-J. Lee and K.J. Yoon, Int. J. Hydrogen Energy, 2004, 29, 187. 17. E.K. Lee, S.Y. Lee, G.Y. Han, B.K. Lee, T.-J. Lee, J.H. Jun and K.J. Yoon, Carbon, 2004, 42, 2641. 18. R.A. Couttenye, M.H. De Villa and S.L. Suib, J. Catal., 2005, 233, 317. 19. T.V. Choudhary, C. Sivadinarayana, C. Chusuei, A. Klinghoffer and D.W. Goodman, J. Catal., 2001, 199, 9. 20. T.V. Choudhary, C. Sivadinarayana, A. Klingerhoffer and D.W. Goodman, Stud. Surf. Sci. Catal., 2001, 136, 197. 21. D.W. Goodman, R.D. Kelley, T.E. Madey and J.M. White, J. Catal., 1980, 64, 479. 22. P. Ferreira-Aparicio, I. Rodriguez-Ramos and A. Guerrero-Ruiz, Appl. Catal. A: Gen., 1997, 148, 343. 23. S. Takenaka, H. Ogihara, I. Yamanaka and K. Otsuka, Appl. Catal. A: Gen., 2001, 217, 101. 24. S. Takenaka, S. Kobayashi, H. Ogihara and K. Otsuka, J. Catal., 2003, 217, 79. 25. N. Shah, D. Panjala and G.P. Huffman, Energy and Fuels, 2001, 15, 1528. 26. J. Chen, Y. Li, Z. Li and X. Zhang, Appl Catal. A: Gen., 2004, 269, 179.

182

Catalysis, 2006, 19, 164–183

27. 28. 29. 30. 31. 32.

J. Li, G. Lu, K. Li and W. Wang, J. Mol. Catal. A, 2004, 221, 105. J. Li, G. Lu and K. Li, Chem. Lett., 2004, 33, 652. K. Otsuka, S. Kobayashi and S. Takenaka, Appl. Catal. A: Gen., 2000, 190, 261. T. Ishihara, Y. Miyashita, H. Iseda and Y. Takita, Chem. Lett., 1995, 2, 93. E. Kikuchi, Cattech, 1997, 1, 67. Q. Weizhong, L. Tang, W. Zhanwen, W. Fei, L. Zhifei, L. Guohua and L. Yongdan, Appl. Catal. A: Gen., 2004, 260, 223. T.V. Choudhary and D.W. Goodman, Catal. Lett., 1999, 59, 93. T.V. Choudhary and D.W. Goodman, J. Catal., 2000, 192, 316. C. Sivadinarayana, T.V. Choudhary, L. Daemon, J. Eckert and D.W. Goodman, Angew Chem. Int. Edit., 2002, 41, 144. R. Aiello, J.E. Fiscus, H.-C.Z. Loye and M.D. Amiridis, Appl. Catal. A: Gen., 2000, 192, 227. V.R. Choudhary, S. Banerjee and A. M. Rajput, J. Catal., 2001, 198, 136. V.R. Choudhary and A.M. Rajput, US Patent 2004118047, 2004. V.R. Choudhary, S. Banerjee and A.M. Rajput, Appl. Catal. A: Gen., 2002, 234, 259. K. Cox and K. Williamson, Hydrogen: Its technology and implications, CRC Press, Boca Raton, FL, 1977, Vol. 1. S.H.S. Zein and A.R. Mohamed, Energy and Fuels, 2004, 18, 1336. B. Monnerat, L. Kiwi-Minsker and A. Renken, Chem. Eng. Sci., 2001, 56, 633. C. Marquez-Alvarez, E. Odier, L. Pinaeva, Y. Schuurman, C. Millet and C. Mirodatos, Stud. Surf. Sci. Catal., 2001, 136, 483. J.I. Villacampa, C. Royo, E. Romeo, J.A. Montoya, P. Del Angel and A. Monzon, Appl. Catal. A, 2003, 252, 363. K. Otsuka, S. Takenaka and H. Ohtsuki, Appl. Catal. A: Gen., 2004, 273, 113. S. Takenaka and K. Otsuka, Chem. Lett., 2001, 3, 218. S. Takenaka, Y. Tomikubo, E. Kato and K. Otsuka, Fuel, 2004, 83, 47. S. Takenaka, E. Kato, Y. Tomikubo and K. Otsuka, J. Catal., 2003, 219, 176. R.T.K. Baker, Carbon, 1989, 27, 315. G.E. Gadd, M. Blackford, S. Moricca, N. Webb, P.J. Evans, A.N. Smith, G. Jacobsen, S. Leung, A. Day and Q. Hua, Science, 1997, 277, 933. P. Chen, X. Wu, J. Lin and K.L. Tan, Science, 1999, 285, 91. N.M. Rodriguez, M.S. Kim and R.T.K. Baker, J. Phys. Chem., 1994, 98, 13108. N.M. Rodriguez, J. Mater. Res., 1993, 8, 3233. V.A. Likholobov, V.B. Fenelonov, L.G. Okkel, O.V. Goncharova, L. Avdeeva, V.I. Zaikovskii, G.G. Kuvshinov, V.A. Semikolenov, V.K. Duplyakin, O.N. Baklanova and G.V. Plaksin, React. Kinet. Catal. Lett., 1995, 54, 381. K.P. De Jong and J.W. Geus, Catal. Rev.-Sci.-Eng., 2000, 42, 481. V.B. Fenelonov, L.B. Avdeeva, V.I. Zheivot, O.V. Goncharova and L.G. Pimneva, Kinetics and Catalysis, 1993, 34, 483. L.B. Avdeeva, O.V. Goncharova, D.I. Kochubey, V.I. Zaikovskii, L. Plyasova, B.N. Novgorodov and S.K. Shaikhutdinov, Appl. Catal. A: Gen., 1996, 141, 117. S.K. Shaikhutdinov, V.I. Zaikovskii and L.B. Avdeeva, Appl. Catal. A: Gen., 1996, 148, 123. M.L. Toebes, J.H. Bitter, A.J. van Dillen and K.P. Je Dong, Catal. Today, 2002, 76, 33. Y.D. Li, J.L. Chen and L. Chang, Appl. Catal. A: Gen., 1999, 163, 45. M.A. Ermakova, D.Y. Ermakov, G.G. Kuvshinov and L.M. Plyasova, J. Catal., 1999, 187, 77.

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.

Catalysis, 2006, 19, 164–183

183

62. M.A. Ermakova, D.Y. Ermakov and G.G. Kuvshinov, Appl. Catal. A: Gen., 2000, 201, 61. 63. M.A. Ermakova, D.Y. Ermakov, L.M. Plyasova and G.G. Kuvshinov, Catal. Lett., 1999, 62, 93. 64. Y.D. Li, J.L. Chen and L. Chang, J. Catal., 1998, 178, 76. 65. T.V. Reshetenko, L.B. Avdeeva, Z.R. Ismagilov, A.L. Chuvilin and V.A. Ushakov, Appl. Catal. A: Gen., 2003, 247, 51. 66. W. Qian, T. Liu, F. Wei, Z. Wang and Y. Li, Appl. Catal. A: Gen., 2004, 258, 121. 67. S.K. Shaikhutdinov, L.B. Avdeeva, B.N. Novgorodov, V.I. Zaikovskii and D.I. Kochubey, Catal. Lett., 1997, 47, 35. 68. T.V. Reshetenko, L.B. Avdeeva, Z.R. Ismagilov and A.L. Chuvilin, Carbon, 2004, 42, 1143. 69. M.S. Kim, N.M. Rodriguez and R.T.K. Baker, J. Catal., 1991, 131, 60. 70. M.S. Kim, N.M. Rodriguez and R.T.K. Baker, J. Catal., 1992, 134, 253. 71. G.G. Kuvshinov, Y.I. Mogilnykh, D.G. Kuvshinov, V.I. Zaikovskii and L.B. Avdeeva, Carbon, 1998, 36, 87. 72. T.V. Reshetenko, L.B. Avdeeva, Z.R. Ismagilov, V.V. Pushkarev, S.V. Cherepanova, A.L. Chuvilin and V.A. Likholobov, Carbon, 2003, 41, 1605. 73. M.A. Ermakova, D.Y. Ermakov, A.L. Chuvilin and G.G. Kuvshinov, J. Catal., 2001, 201, 183. 74. M.A. Ermakova and D.Y. Ermakov, Catal. Today, 2002, 77, 225. 75. F. Benissad, P. Gadelle, M. Coulon and L. Bonnetain, Carbon, 1988, 26, 61. 76. F. Benissad, P. Gadelle, M. Coulon and L. Bonnetain, Carbon, 1989, 27, 585. 77. S. Takenaka, M. Serizawa and K. Otsuka, J. Catal., 2004, 222, 520. 78. L.B. Avdeeva, D.I. Kochubey and S.K. Shaikhutdinov, Appl. Catal. A: Gen., 1999, 177, 43. 79. S. Takenaka, M. Ishida, M. Serizawa, E. Tanabe and K. Otsuka, J. Phys. Chem. B, 2004, 108, 11464. 80. X. Li, Y. Zhang and K.J. Smith, Appl. Catal. A: Gen., 2004, 264, 81.

Catalytic Reforming of Liquid Hydrocarbon Fuels for Fuel Cell Applications BY DUSHYANT SHEKHAWAT,a DAVID A. BERRY,a TODD H. GARDNERa AND JAMES J. SPIVEYb a National Energy Technology Laboratory, US Department of Energy, 3610 Collins Ferry Road, Morgantown, WV 26507-0880, USA b Department of Chemical Engineering, Louisiana State University, S. Stadium Drive, Baton Rouge, LA 70803, USA

1

Introduction

1.1 Demands for Fuel Reforming Technology. – The recent interest in fuel cell technology has intensified the focus on fuel reforming. Fuel cells are in essence a continuously operating battery that generates electricity as long as fuel, such as H2, and an oxidant are supplied.1 Since the fuel is not burned, there are significantly less emissions than commonly associated with the combustion of fossil fuels. The high efficiency of fuel cells leads to more electric power per unit of fuel consumed.2 Figure 1 compares the energy efficiency of various fuel cells to those of conventional technologies.3

Figure 1 Energy efficiency of various processes3 (adapted from ref. 3) Catalysis, Volume 19 r The Royal Society of Chemistry, 2006

184

Catalysis, 2006, 19, 184–254

185

This and other attributes make fuel cell technology appealing. For example, the proton exchange membrane (PEM) fuel cell is being pursued by a number of companies because of its low operating temperature, response to transients,4 and compact size, which make it desirable for a number of residential,5 commercial, and military6 applications. However, challenges remain before H2 can be widely used.7,8 Nonetheless, clear environmental and efficiency benefits provide incentives for the development of these technologies9,10 and have attracted the attention of many researchers.11–15 The U.S. Department of Energy has a stated goal of replacing 2–4 quads of U.S. energy with H2 by 2010.16 Current U.S. production of H2 is about 20 billion lb/yr (corresponding to about 1.1 quads), almost all of which is used for chemical synthesis and processing.17 If used solely to produce energy, this could fuel 20–30 million cars, or power 5–8 million homes.17,18 Unfortunately, the lack of widely available alternative sources of H2 dictates that the H2 (or H2-rich synthesis gas for high-temperature fuel cells) be derived from hydrocarbon fuels. Depending on the application (stationary, central power, remote, auxiliary, transportation, military, etc.), there are a wide range of conventional fuels, such as natural gas (methane),19 propane,20,21 butane,22 light distillates, methanol,23 ethanol,24 propanol,25 dimethyl ether,26,27 naphtha, gasoline,28 diesel,28–31 biodiesel,32 naval distillate fuel (NATO F-76),33,34 kerosene,35 and jet fuels36,37 that could be used in reforming processes to produce H2. Alcohol-based fuels such as methanol are widely studied for the production of H2 for fuel cells because they can be reformed at relatively low temperatures and are free of sulfur compounds. However, methanol is corrosive to several metals, rubberized components, gaskets, and seals, and thus cannot be employed by existing infrastructures without significant modifications. Fisher-Tropsch fuels and biofuels, which are free of sulfur, aromatics, metals, or other toxics,38 are being considered for reforming applications. FT fuels can be delivered through the existing infrastructure, which could make them more practical and economical for use in fuel cells than other fuels. Kerosene-type fuels can be very practical in countries like Japan, because of their low price and well-developed infrastructure.39 Liquid fuels, such as gasoline and diesel, are globally attractive because of their existing infrastructure, higher well-to-wheel efficiencies, and higher energy densities;40 they are the focus of this chapter. Several studies have recently reported on the technical viability and economics of fuel cell systems for various applications.9,41–63 A variety of designs are under development, but all rely on a H2-rich gas stream. Transportation and storage of H2 is one of the key issues that must be overcome,64 and is an important reason for the interest in reformed liquid fuels, which can take advantage of a well-developed existing infrastructure to provide H2. 1.2 Applications/Types of Reforming. – The type of reformer has a significant impact on overall efficiency and operating characteristics of a fuel cell system. Reformers can either be thermally activated or catalytic.65 The processes considered in this chapter are limited to catalytic. Depending on fuel, mode

186

Catalysis, 2006, 19, 184–254

of reforming, and catalyst type, temperatures for catalytic systems typically range from 6001C–1,0001C.28,35,66–69 Higher temperatures required for noncatalytic reforming have a higher tendency towards unwanted nitrogen oxide formation, require more costly materials of construction, and result in lower system efficiency due to the larger temperature mismatch between the fuel cell and reformer. There are three predominant modes of reforming; partial oxidation (POX), steam reforming (SR), and autothermal reforming (ATR). All three involve oxidation of the hydrocarbon fuel to produce a H2-rich synthesis gas. The coupling of the reformer to the fuel cell can influence both system configuration and reformer operation. For example, low-temperature fuel cells like PEMs require a H2-rich stream with virtually zero CO, which acts to poison/deactivate the catalyst on the fuel cell anode. Figure 2 shows a representative process configuration, which consists of three steps: reforming of the liquid fuel into syngas, water-gas shift (WGS), and preferential oxidation of CO. However, in the case of high temperature fuel cells like solid oxide fuel cells (SOFCs) that can utilize CO as a fuel, downstream WGS and preferential CO oxidation are not required and simplify the system. Consider the generalized expression for the POX reaction: n m n Cn Hm þ ðO2 þ 3:76N2 Þ ! nCO þ H2 þ 3:76 N2 ð1Þ 2 2 2 Frequently, steam may be present during POX due to maximize selectivity toward combustion, or it may be specifically added as the sole oxidant in SR, which is described by the following reaction: m Cn Hm þ nH2 O ! nCO þ n þ ð2Þ H2 2 These reactions are followed by the establishment of the WGS equilibria: CO þ H2O 2 CO2 þ H2

Figure 2 Generic on-board fuel processor

(3)

Catalysis, 2006, 19, 184–254

187

During ATR, air (or O2), fuel, and steam react to form a syngas mixture with a net heat of reaction near zero at temperatures of 600–8001C (A more thorough discussion can be found in Section 5). A number of reactions take place in this step, including POX and SR.11 The primary reactions (i.e., those leading to CO and H2) can be represented as follows: CmHn þ O2 þ H2O - CO þ H2

(4)

Thermal integration is an important aspect of the system. Figure 3a depicts a fuel cell coupled to a catalytic POX reformer. This reformer thermally matches very well with the 8001C SOFC technologies being developed. However, both the reformer and fuel cell involve exothermic oxidation reactions, and heat is in

Figure 3 Thermal integration of reformer with fuel cell system (a) Catalytic partial oxidation reformer, (b) Catalytic steam reformer, and (c) Autothermal reformer

188

Catalysis, 2006, 19, 184–254

excess for both. This does not allow for a high degree of thermal integration and limits overall system efficiency. There are some favorable characteristics though. The POX reactor has very fast kinetics, which minimizes the size of the reformer. Startup of the reformer can be extremely quick for POX, although the heat-up or ramp rate requirements of the SOFC will probably limit ultimate startup times for the system. Catalytic SR represents the other end of the reforming spectrum (Figure 3b). Fuel is mixed with steam instead of O2 in the reformer, to produce a very H2rich synthesis gas. The reaction is highly endothermic and therefore, it requires an external source of heat to maintain the reactor temperature. The amount of energy required is generally on the order of 22% of the lower heat of combustion of a liquid fuel.66 The heat from the fuel cell can drive the reformer, so overall system efficiencies can be maximized. However, startup can be difficult, due to needed external heat transfer. Transient response may be slow as well. The kinetics of SR is also significantly slower than POX and therefore requires a much larger reformer for the same throughput. A water delivery system is also needed for SR, which can add cost and complexity to the system. Finally, SR may not be efficient in reforming hydrocarbons heavier than light naphtha because of their tendency to form carbon even at higher steam to carbon ratios.70 ATR is actually a hybrid of POX and SR (Figure 3c), and is arguably the most thermally efficient means of producing a H2-rich gas stream from liquid fuels.71 In this type of reactor, fuel, air, and steam are mixed and reacted in a thermally neutral manner such that the endothermic SR reaction is thermally balanced by the exothermic POX reactions (i.e., no net heat gain/loss). It has been observed that at relatively low space velocities (4,000 to 10,000 h1), ATR can be divided into two distinct zones: a POX zone and a SR zone.72 In the first zone, POX preferentially occurs. Here, reactions (1) and (3) principally take place. Within this zone, sufficient heat is generated to sustain the highly endothermic SR reaction that occurs within the subsequent SR zone. The principal reactions occurring in this zone are reactions (2) and (3). ATR offers high thermal integration with the fuel cell. Fast startup via POX also enhances flexibility. In terms of transient response, ATR fits between POX and SR systems. Although steam and oxygen to carbon ratios can be calculated for a given fuel to achieve thermoneutrality, practical steam to carbon ratios of 0.8 have been suggested as a point of relatively safe, carbon-free operation.73 1.3 Issues. – Despite their logistical appeal as feedstocks for H2 production, gasoline and diesel fuel are relatively difficult to reform for a variety of reasons related to both catalyst and reactor design. These fuels are complex variable mixtures of hundreds of hydrocarbon compounds containing mainly olefins, saturates and aromatics. The differing boiling points of these components can make fuel introduction and reactant mixing into fuel processors difficult. Vaporization is usually not considered a viable option, due to the pyrolytic and carbon-forming tendencies of some hydrocarbon species at the elevated temperature required to vaporize all components. In several efforts, automotive

189

Catalysis, 2006, 19, 184–254

fuel injectors have been used to introduce and facilitate mixing of air and liquid fuels.74 The advantage to using fuel injector technology is that the feed is usually injected at ambient conditions, thus reducing thermal fragmentation, and the mixing that is provided minimizes regions of combustible fuel-air mixtures in the reactor where the temperature is above the autoignition temperature of the fuel. Still, care must be taken to avoid unwanted heating in the mixing zone above the hot catalyst bed. But most of the issues involve the catalyst system itself. The catalyst must be active and selective for the fuel of choice, stable, and resistant to poisoning and attrition while subjected to variations in flow, temperature, and pressure.41,75 For successful operation at commercial scale, the reforming process must be able to achieve high conversion of the hydrocarbon feedstock at high space velocities, as well as high H2 and CO selectivities. The reforming catalyst has to meet performance targets (see Table 1) as identified by U.S. DOE before it becomes feasible for use in the fuel reformers of transportation fuel cell systems.64,76,77 According to Krumpelt et al.,77 a reforming catalyst will have to process the feed at a GHSV of 200,000 h1 with a hydrocarbon conversion greater than 90%, a H2 selectivity greater than 80%, and have a lifetime of 5,000 h. This can be a significant challenge given the many hurdles facing the catalyst. Some fuel components may have detrimental effects on the catalysts as well as on the reaction rates.78–83 For example, aromatics present in diesel decrease the rate of reforming of the paraffins because the aromatics are strongly adsorbed on the metal active sites.84 Aromatics in liquid fuels can also contribute significantly to the carbon formation, as compared to the paraffins and cycloparaffins. Increasing reaction temperatures to reform these difficult compounds can trigger excessive sintering of the catalyst or vaporization of the active metal itself, thus rendering the catalyst ineffective and short-lived. The desire to lessen or avoid water addition and management in ‘‘dry’’ fuel cell systems worsens deactivation by carbon formation. Sulfur, present in most hydrocarbon fuels, also represents a poison to these metal-based catalytic processes. These deactivation processes are covered in greater detail in Section 2. Even the process of experimental setup and measurement can be an issue. In a fixed bed laboratory reactor at reforming temperatures (near 8001C), the following sequence of reactions is thought to take place.85,86 Very near the Table 1

Target and Status for the Fuel Processor in 50 kWe (net) Fuel Cell Systems76

Characteristics

Units

2001 Status

2005 Target

2010 Target

Power density Cost Durability Cold start-up timea

W/L $/kW Hours Min

500 85 1000 o10

700 25 4000 o1

800 10 5000 o0.5

a

To max power from þ201C ambient temperature.

190

Catalysis, 2006, 19, 184–254

reactor inlet, O2 is quickly consumed to produce primarily CO2 and H2O. Next, SR (reaction (2)), dry reforming (reaction (5)), m H2 ð5Þ 2 and WGS (reaction (3)) take place. It is extremely difficult to study these reactions independently because of the experimental complexity of measuring the reactants and products at various points in the reactor. This problem was recognized even in the earliest work on the POX of methane.87 This complexity, coupled often with significant rates of deactivation, make the task even more complicated. Moreover, the highly exothermic nature of the complete oxidation reaction must also be accounted for, because it leads to hot spots at the reactor inlet. These hot spots can lead to pyrolytic reactions such as cracking, which complicate not only the study of individual reactions but even the measurement of the true reaction temperature.88,89 Clearly, these factors may account for some of the differences in the literature with studies carried out at nominally similar conditions. Cn Hm þ nCO2 ! 2nCO þ

1.4 Scope of this Chapter. – The purpose of this paper is to review recent accomplishments in the reforming of liquid fuels, primarily diesel, gasoline, and the model compounds that represent them, including experimental as well as theoretical studies. A significant amount of work has been reported in the last 15–20 years on fuel reforming. The progress achieved in the rapidly growing field of H2 production, primarily for fuel cells, is reflected in the number of recent reviews.41,66,90–101 Several studies have been conducted to compare different fuels for H2 production,102–105 while others focus on the sustainability of the H2 as a source of energy for the future,106–109 and the technical viability and economics of fuel cell systems for various applications.9,41–63 Many effects/ studies involving methane reforming may be applicable and/or extend to liquid hydrocarbon fuels. The literature is quite extensive for methane reforming but will not be covered, as this chapter will focus on recent work using liquid fuels. Also, only reforming to a H2-rich synthesis gas will be covered. No attempt will be made to cover the large body of work that exists for WGS or preferential oxidation of CO to CO2, as required by low-temperature fuel cells.

2

Deactivation

Deactivation of catalysts in the reforming of liquid fuels is caused principally by two processes: the formation of carbon-containing deposits and sulfur poisoning. This section examines the thermodynamics and the literature dealing with these processes. 2.1 Carbon Deposition. – The term ‘‘carbon deposition’’ is used in the literature to refer to the accumulation of two distinct types of carbon-containing materials on the catalyst surface, both of which lead to deactivation.110–113 The first can be described as elemental carbon, formed directly by one of several

191

Catalysis, 2006, 19, 184–254

reactions involving either the decomposition of hydrocarbons in the feed stream, or reactions of CO formed in the reforming process. The second is called ‘‘coke’’—a term defined here to mean two types of higher molecular weight compounds: (i) polycyclic, aromatic hydrocarbons formed in a step-wise series of condensation/dehydrogenation reactions from the hydrocarbons in the feed, and (ii) pyrolitic carbon formed from olefins in the gas phase, encapsulating the catalyst pellet. Formation of elemental carbon or coke is detrimental in several ways. First, it decreases the catalyst activity by blocking active sites. This causes the reforming reaction rates in the reactor to decline over time as the coke accumulates. Second, carbon formation may further cause mechanical breakage of the catalyst particles, which may result in increasing pressure drop, and ultimately reactor plugging. Preventing carbon deposition on the catalyst is one of the most challenging problems in the reforming of liquid fuels. The following sections describe the thermodynamics of the formation of both elemental carbon and coke, and provide a review of systematic studies on these reactions in the reforming of liquid fuel compounds.

2.1.1 Elemental Carbon-Forming. For liquid fuels, formation of elemental carbon takes place principally (but not exclusively) via the following reactions:30,110 2 CO 2 C þ CO2

(6)

CnHm - n C þ m/2 H2

(7)

CO þ H2 2 C þ H2O

(8)

The ‘Boudouard reaction’ (6) produces what is known as ‘‘Boudouard carbon’’ and is favored at lower temperatures. Reaction (7) is due to thermal cracking and is favored at higher temperatures. Reaction (8) is the reversible gasification of elemental carbon, and is favored (as written) at higher temperatures. Reactions (6) and (8) are reversible, whereas reaction (7) is irreversible for n 4 1. The morphology of the carbon on the surface can assume several forms: a two-dimensional film or so-called ‘‘whisker’’ carbon, which is formed when the carbon dissolves in the supported metal catalyst, diffuses through the metal, and forms a growing filament that lifts the metal from the catalyst surface. Whisker carbon is typically associated with Ni-based catalysts because carbon is soluble in Ni at reforming conditions. Whisker carbon tends to form at higher temperatures, low steam to hydrocarbon ratios and higher aromatic content of the feeds. This type of carbon formation may be minimized by the use of precious metals as catalysts, because these metals do not dissolve carbon. On a nickel surface, the whisker mechanism can be controlled by sulfur passivation.

192

Catalysis, 2006, 19, 184–254

2.1.1.1 Thermodynamics. Reaction conditions and catalyst surface properties can control carbon formation on the catalyst surface. The most important reaction conditions that affect carbon formation are temperature and feed compositions (steam/carbon (S/C) and oxygen/carbon (O/C) ratios). Table 2 shows the thermodynamics of the three major carbon-forming reactions (6)–(8), at 6001C, 1 atm, using n-C16 and i-C8 as surrogates of diesel and gasoline, respectively. These results show that each of the carbon-forming reactions is thermodynamically favorable at typical reforming temperatures. The reactions of n-C16 and i-C8 are irreversible at these conditions. It can be seen that the thermodynamic driving force for carbon formation decreases as temperature increases. Carbon formation from the Boudouard reaction is thermodynamically favored at lower temperatures because this reaction is exothermic. This kind of carbon formation usually dominates at the reactor inlet (or feed lines) where the temperature is lower. However, higher temperatures favor the cracking reaction (7). Therefore it is often desirable to conduct the hydrocarbon reforming at an intermediate temperature where the thermodynamic driving force for carbon formation is minimal. There are few reported analyses of the thermodynamics of carbon deposition in the ATR of liquid fuels. Though typically not stated in these analyses, the calculations were presumably carried out using the thermodynamic properties of elemental carbon (e.g., as formed in reactions (6)–(8) above), rather than any ‘‘coke’’ species (which consist of a wide range of polynuclear aromatic compounds with quite different thermodynamic properties). This is an important difference, since the results apply only to elemental carbon, not ‘‘coke deposition’’ in general. There is also a thermodynamic driving force for the formation of elemental carbon for the ATR reaction, when both steam and oxygen are present in the feed. Consider the formation of elemental carbon as follows (this stoichiometry is based on thermal neutrality, DH6001C ¼ 0 kJ/mol, of the ATR of n-C16): C16H34(g) þ 4O2(g) þ 7H2O(g) ¼ 4C(s) þ 9CO(g) þ 24H2(g) þ 3CO2(g)

(9)

DG6001C ¼ 2525 kJ/mol; DH6001C ¼ 0 kJ/mol Figure 4 shows the effect of temperature on carbon formation at equilibrium, starting with a stoichiometric mixture of n-C16, O2 and H2O. At each temperature, the formation of elemental carbon by reaction (9) is favorable, but Table 2

Thermodynamics of Elemental Carbon-Forming Reactions at 6001C, 1 atm (Calculations using HSC Chemistry 4.0153)

Reaction

DHr,6001C, kJ/mol

DGr,6001C, kJ/mol

Keq,6001C

2 CO 2 C þ CO2 n-C16H34 - 16 C þ 17 H2 i-C8H18 - 8 C þ 9 H2 CO þ H2 2 C þ H2O

172 þ453 þ262 136

17.6 1034 515 10.6

11.3 7.43 1061 6.17 1030 4.31

Catalysis, 2006, 19, 184–254

193

Figure 4 Equilibrium composition versus temperature; initial composition 1 mol nC16(g), 4 mol O2(g), and 7 mol H2O(g); 1 atm (Calculations using HSC Chemistry 4.0153)

decreases slightly with temperature above about 5001C. However, there is a significant amount of elemental carbon at equilibrium for all temperatures shown. Added steam would be expected to prevent carbon formation on the catalyst surface by the reverse of reaction (8). Increased steam should also shift more CO towards CO2 via the WGS reaction and, hence minimize carbon formation due to the Boudouard reaction (6). However, the effect of steam on the equilibrium formation of elemental carbon is insignificant until impractically high levels of steam are added, as shown in Figure 5. These thermodynamic results have been qualitatively confirmed in other studies. For example, Ahmed and Krumpelt71 have analyzed the thermodynamics of carbon formation for these reforming reactions using i-C8 (Figure 6). For the ATR reaction analogous to reaction (9), the minimum temperature needed to prevent carbon formation at an oxygen/carbon (O/C) ratio of 1 is 1,0801C, well above the operating temperature of typical reformers. Increasing the O/C ratio from 1 to 2 decreases this to 5751C. As expected, the higher the O/ C ratio, the lower the minimum temperature required to prevent (thermodynamically) the formation of carbon. This shows quantitatively the importance of O/C ratio in deactivation of these catalysts. Comparison of the POX and SR results shows that for a given O/C ratio, less carbon is formed at equilibrium if the oxygen is supplied as steam rather than as O2(g). 2.1.2 Coke Formation. As discussed above, ‘‘coke’’ refers to both polycyclic, condensed polymeric hydrocarbons formed during the reaction of

194

Catalysis, 2006, 19, 184–254

Figure 5 Equilibrium composition versus temperature; initial composition 1 mol n-C16; 1 mol O2(g); 1, 10, and 50 mol H2O(g); 1 atm. See reaction (9) (calculations using HSC Chemistry 4.0153)

Figure 6 Reactor temperatures required to prevent thermodynamically the formation of elemental carbon in reforming of i-C8, from thermodynamic calculations (Reprinted from Ahmed et al.,71 copyright (2001), with permission from Elsevier)

hydrocarbons on metal catalysts, and pyrolytic carbon compounds formed directly from olefins in the gas phase. The formation of coke is typically thought to proceed through the following sequence:73,110,114,115 CnHm ¼4 olefins ¼4 polymers ¼4 coke ¼4 graphitic carbon Note, however, that this sequence is not universal; that is, carbon can be formed by the direct decomposition of CnHm as shown in reaction (7). This reaction sequence dominates at low temperatures, low S/C ratios, higher aromatic content in the feeds, and with high boiling feedstocks. The risk of

195

Catalysis, 2006, 19, 184–254

polymeric carbon formation increases when the rate of hydrocarbon adsorption becomes higher than the gasification rate of the adsorbed species. For Nibased catalysts, there is a temperature above which carbon formation takes place by whisker mechanism at a given S/C ratio, and a temperature below which the catalyst is deactivated by polymeric carbon formation. Coke formation by pyrolysis increases strongly with increasing temperature. Therefore, there is a limited range of temperatures in which the reforming of higher hydrocarbons can be carried out under relatively carbon-free conditions. But it is also important to note that coke can be formed at conditions that are thermodynamically carbon-free.73 This may be caused by local microscopic conditions that differ from the bulk measured conditions (e.g., local temperatures differ from the bulk, or non-ideal mixing).114 2.1.2.1 Thermodynamics. Formation of coke is thermodynamically favorable even at POX conditions. Consider the formation of two compounds that approximate the structure of coke, anthracene (C14H10, a 3-ring polycyclic aromatic compound) and naphthacene (C18H12, a 4-ring polycyclic aromatic). These compounds can be formed from n-C16 by the following two reactions: n–C16H34(g) þ O2(g) ¼ C14H10(s) þ 2CO(g) þ 12H2(g) n–C16H34(g) þ 7/2O2(g) ¼

1 2C18H12(s)

þ 7CO(g) þ 14H2(g)

(10) (11)

Table 3 shows that both reactions are highly favorable and irreversible at reforming temperatures. Literature studies confirm this result for other liquid fuels. For example, Docter et al. have calculated the regions of coke formation for ATR and POX of a synthetic gasoline mixture at various S/C ratios (Figure 7).73 For this calculation, a ‘‘synthetic’’ gasoline is used, consisting of 35% n-C6, 25% hexene, and 40% xylene. As expected, the results show that as the oxygen concentration (‘‘air ratio’’) increases, the thermodynamic driving force for coke formation decreases. Increasing the S/R ratio from zero (i.e., POX) to 0.7 also decreases coke formation. Coke is not thermodynamically favored at air ratios above 0.3, which corresponds to temperatures above B8501C. 2.1.3 Reaction Studies. There have been a significant number of studies on the deactivation of catalysts in the reforming of liquid fuels, especially the formation of elemental carbon and coke. Collectively, these studies show that the Table 3

Thermodynamics of Coke-Forming Reactions at 6001C (Calculations using HSC Chemistry 4.0153)

Reaction n-C16H34(g) þ O2(g) ¼ C14H10(s) þ 2CO(g) þ 12H2(g) n-C16H34(g) þ 7/2O2(g) ¼ 1/2C18H12(s) þ 7CO(g) þ 14H2(g)

DHr,6001C, DkJ/mol

DGr,6001C, kJ/mol

Keq,6001C

þ397

830

4.6 1049

305

1963

2.9 10117

196

Catalysis, 2006, 19, 184–254

Figure 7 Thermodynamic calculation of product composition and coke formation for a synthetic gasoline mixture (35% n-C6, 25% hexene, and 40% xylenes) for ATR (S/C ¼ 0.7) and POX (S/C ¼ 0); Air ratio ¼ (nair/nfuel)actual/(nair/nfuel)stoichiometric (Note that temperature is a dependent variable in this figure, fixed by air/fuel ratio) (Reprinted from Docter and Lamm,73 copyright (1999), with permission from Elsevier)

presence of aromatics and sulfur in the fuel affect the formation of these compounds, and that the structure of the catalyst can greatly minimize deactivation. 2.1.3.1 Effect of Aromatics on Formation of Coke and Carbon. Conventional liquid fuels contain widely differing levels of aromatics; gasoline usually contains more than diesel. The studies show that these compounds cause more rapid deactivation than linear alkanes alone. A related result is that the steady state conversion of diesel fractions is also reduced in the presence of aromatics—i.e., these compounds act as kinetic inhibitors, limiting the production of H2. An example of this is illustrated by Palm et al.116 who studied the ATR of a mixture of n-C13–C19 alkanes. The effect of adding two aromatics typical of those found in diesel (tetrahydronaphthalene (THN) and decahydronaphthalene (DHN)) resulted in decreased conversion over a wide range of conditions. Figure 8 is typical of the results of this study. Another effect of aromatics is increased carbon formation, which has long been recognized as the primary means of catalyst deactivation in the ATR of hydrocarbons. Using carbon-forming reactions (6)–(8), an equilibrium line for carbon formation as a function of O2/C and S/C ratios can be calculated. Figure 9 shows the results of this calculation for n-C14, along with the experimental results for two Ni-based commercial catalysts.30

Catalysis, 2006, 19, 184–254

197

Figure 8 Effect of the addition of THN on conversion of n-C13-C19 mixture (S/C ¼ 2.20) (Reprinted from Palm et al.,116 copyright (2002), with permission from Elsevier)

Experimentally observed carbon formation followed the general trend predicted by the equilibrium calculations as a function of S/C and O2/C ratios. There are differences among the catalysts shown in Figure 9. The ‘‘ICI-46-4’’ catalyst contains potassium, which is added to conventional steam-methane reforming catalysts to suppress carbon formation. Carbon forms on a conventional catalyst (presumably without potassium) at conditions further from the equilibrium curve than for the ICI-46-4 catalyst, confirming the carbon-limiting effect of potassium. However, other significant differences in these catalysts, such as Ni loading, make it impossible to attribute the carbon-suppressing activity to potassium alone. Similar experiments in this study using benzene and naphthalene as reactants show that these aromatics (commonly found in diesel) result in carbon formation at conditions even further removed from equilibrium (Figure 10). Similar results were obtained by Houseman et al.29 in the ATR of No. 2 fuel oil (in this study, the equilibrium calculations were based on data for graphitic carbon). They reported that both aromatic and olefinic compounds are more prone to form carbon than aliphatic compounds at conditions thermodynamically removed from those that favor carbon formation. SEM, TGA, and XRD analysis of the ICI Ni-based catalysts showed that carbon tended to form in downstream sections of the reactor, after the O2 was consumed. There were apparently two types of carbon. The first type apparently formed within the catalyst pores, causing physical breakage of the particles. The second type corresponded to whisker carbon with Ni particles at the ends (discussed earlier)—a deactivation pattern similar to that observed in conventional methane SR (also Ni-based) catalysts.117 ATR of pure aromatics might be expected to lead to rapid deactivation. However, Springmann et al.67 did not observe deactivation in ATR of toluene at 600–8001C, S/C ¼ 1.5–2.4, although SR of toluene at these same conditions caused rapid deactivation.

198

Catalysis, 2006, 19, 184–254

Figure 9 Carbon formation on two commercial Ni-based catalysts in the ATR of n-C14; open symbols—carbon free; closed symbols—carbon formation (Reprinted from FlytzaniStephanopoulos and Voecks,30 copyright (1983), with permission from Elsevier)

Carbon formation is also different for diesel and gasoline. The long chain hydrocarbons present in diesel or kerosene fuel are more difficult to reform than the shorter chain hydrocarbons present in gasoline, while aromatics in gasoline hinder the overall reaction rate.68,118–121 An example is found in the results of Ming et al.68 who showed that SR of n-C16 required a higher steam/ carbon ratio to avoid coke formation than i-C8. The cetane number of the feed had little effect on carbon formation. Carbon formation can often be attributed to fuel pyrolysis that takes place when the diesel fuel is vaporized. This is considerably decreased when the steam content in feed increases.

Catalysis, 2006, 19, 184–254

199

Figure 10 Carbon formation on a commercial Ni-based catalyst in the ATR of benzene and benzene/naphthalene feeds; open symbols—carbon free; closed symbols—carbon formation (Reprinted from Flytzani-Stephanopoulos and Voecks,30 copyright (1983), with permission from Elsevier)

2.1.3.2 Effect of Sulfur. Small amounts of sulfur in the feed may actually minimize coke formation,122,123 even though sulfur deactivates the metal component of the catalyst. Several studies suggest that, at least for Ni-based catalysts, sulfur occupies fourfold sites at low coverages, leaving open sites where sulfur is not adsorbed. Apparently, these open sites are of a critical size that is active for SR with minimal coke formation: SR involves ensembles with 3–4 Ni atoms, while carbon formation requires 6–7 atoms.111,124 Trimm122 found that the critical ensemble size was generated at a fractional sulfur coverage of 0.7–0.8, corresponding to H2S/H2 ratios in the feed greater than 7.5 107.

200

Catalysis, 2006, 19, 184–254

2.1.3.3. Effect of Metal Ensemble Structure. Recent theoretical calculations and experimental studies have confirmed the validity of the effect of ensemble size on carbon formation. Besenbacher et al.125 found theoretical evidence from density functional theory (DFT) and experimental data to show that gold alloyed into the surface layer of a Ni catalyst decreases graphitic carbon formation, while maintaining methane reforming activity. This can be explained by DFT calculations that gold increases the energy barrier for the dissociation of methane. It would be expected that these results could be applied to SR of higher hydrocarbons as well, since the surface species leading to carbon formation may be common to both.122 However, carbon formation also depends on the stability of the adsorbed carbon atoms. The less stable the adsorbed carbon, the more likely it is to react with adsorbed oxygen to form CO and hence, lower the carbon coverage on the catalyst surface. The most stable adsorption site on the pure Ni surface is the threefold hexagonal closedpacked site. DFT results show that if a threefold site is adjacent to a gold atom the adsorbed carbon is completely destabilized; even the threefold sites that are next nearest neighbors to the gold atoms are considerably destabilized. Certain step sites in Ni catalysts are appreciably more reactive than the closed-packed facet sites in Ni for SR and carbon formation reactions.126 Promoters such as alkali, gold, and sulfur preferentially bind to the step edges of Ni, making them effective in the suppression of carbon formation during SR over Ni catalysts.

2.1.3.4 Effect of Oxygen Conductivity in the Support. The effect of oxygen conductivity in the support on carbon deposition has been investigated for reactions closely related to the reforming of liquid fuels, such as methane reforming.127,128 A recent review by Salazar-Villalpando et al.129 shows that the rate of carbon deposition on nominally similar catalysts (primarily Ni-based) varies by several orders of magnitude, depending on the support. Ceria and zirconia were found to be effective, with a mixed ceria-zirconia support being superior to either in one case.130 A temperature-dependent mechanism is postulated involving dissociative adsorption of oxygen on the metal and spillover to the support.131 It is not clear if this mechanism would be applicable under conditions present in the reforming of liquid fuels. There are few reports that deal specifically with the role of oxygen-conducting supports in the reforming of liquid fuels. Krause et al.132 show that Rh supported on oxide ion-conducting gadolinium- and samarium-doped ceria are stable for 1,000 h in the reforming of a ‘‘benchmark’’ fuel (74 wt% i-C8, 20 wt% xylenes, 5 wt% methylcyclopentane, 1 wt% pentene) at 7001C (Figure 11). A supported Pt catalyst, tested at similar conditions, showed lower H2 yield with comparable CO/CO2 yields. The addition of these dopants improves the oxide ion conductivity and increases the number density of surface oxygen vacancies,133 lessening carbon deposition. Clearly, these types of materials are of interest in studying the mechanisms and type of deactivation that take place in the reforming of liquid fuels.

Catalysis, 2006, 19, 184–254

201

Figure 11 Yields of H2, CO, CO2, and CH4 from reforming of benchmark gasoline fuel catalyzed by Rh- or Pt-CGO supported on a cordierite monolith (Conditions: O/C ¼ 0.88, S/C ¼ 1.6, GHSV ¼ 9,000 h1)132

2.1.3.5 Effect of Noble Metal Addition to Ni-Based Catalysts. Unmodified Ni-based catalysts tend to deactivate rapidly due to both carbon deposition and sintering. As suggested by the thermodynamic analysis above, carbon formation can be significantly limited by the addition of steam, which can react with carbon by the reverse of reaction (8). Figure 12 shows typical results for such a catalyst, which contained 5.4 mg carbon per gram of catalyst after 54 h on stream.134 Analysis of the catalyst showed that the Ni particles had severely sintered. The addition of Pd to this catalyst reduced carbon deposition to 1.6 mg carbon/ g cat at the same time on stream. SEM also showed no detectable sintering of this Pd-modified catalyst. Carbon formation is also observed on noble metal catalysts, with substantially different behavior on Pt versus Pd. For example, Figure 13 shows that Pt deactivated much more rapidly than Pd in the ATR of n-C8.135 At similar conditions, but at 5 mol H2O/mol C, a bimetallic Ni–Pd/alumina catalyst did not deactivate over this same 100 h period. However, the

202

Catalysis, 2006, 19, 184–254

Figure 12 Deactivation of 5% Ni/g-alumina in the ATR of n-C8 at 7501C, 1,000 h1, O/ C ¼ 0.25 (K S/C ¼ 5 and m S/C ¼ 3) (Reprinted from Zhang et al.,134 copyright (2003), with permission from Elsevier)

Figure 13 Deactivation of Pt/alumina (K) and Pd/alumina (J) catalysts in the ATR of n-C8, 7001C, O/C ¼ 0.25, S/C ¼ 3 (Reprinted from Yanhui and Diyong,135 copyright (2001), with permission from Elsevier)

conversion was essentially complete at this condition, so some deactivation may have taken place in the upstream part of the reactor. The authors state that the Pt catalyst deactivates due to the presence of CO and water, stating that the Pd does not form ‘‘carbon metal’’ while Pt does. The deactivation was not further studied, nor was the catalyst examined after the reaction to compare its properties with those of the fresh catalyst.

Catalysis, 2006, 19, 184–254

203

2.1.3.6 Effect of Alkali Addition to the Catalyst. Coking takes place either by reaction on the metal surface or by cracking on the support material and in the gas phase. The acidic nature of the support material results in the higher cracking of hydrocarbons. As an example, Rostrup-Nielsen136 studied coke formation on Ni catalysts for SR of hydrocarbons in a thermogravimetric system. Coking was significantly retarded by neutralizing acidic sites by adding alkali or magnesia to the support material. Also, alkali has a promoting effect on the reaction between coke and steam. It was also concluded that the coking rate strongly depends on the unsaturated character of the hydrocarbon. For example, SR of ethylene yielded 17,500 mg/min coking rate, whereas at the same reaction conditions the coking rates with benzene, cyclohexane, and n-C7 were 532, 64, and 135 mg/min, respectively. 2.2 Sulfur Poisoning. – Many conventional liquid hydrocarbon feeds contain large quantities of sulfur. For example, military diesel may contain as much as 3,000 ppm. Currently in the European Union the sulfur content in diesel is limited to 350 ppm, which was reduced to 50 ppm for both diesel and gasoline in 2005. Even with the pending limits on the sulfur content for U.S. fuels, deactivation by sulfur will be an ongoing concern. Studies of the deactivation of ATR catalysts show that the sulfur present in conventional fuels is responsible for rapid deactivation of both Ni-based and noble metal catalysts. At some conditions, sulfur appears to selectively poison the sites responsible for the SR reaction(s). 2.2.1 Thermodynamics. At higher temperatures, Ni and other metal catalysts are believed to be less vulnerable to sulfur poisoning, since their sulfides are thermodynamically less stable. Figure 14 shows the change in DG with temperature for the following reaction. Metal þ H2S - Metal-S þ H2

(12)

Although Figure 14 shows that the sulfides become less thermodynamically stable with increasing temperature (at least according to reaction (12)), PdS is stable up to about 7501C, and PtS is stable up to about 9501C. For Ni, the sulfide is stable well above 1,0001C. [Note that this does not take into account the effect of other gases, such as CO or H2O that may affect the stability of the sulfides.] This suggests why reforming of heavier feedstocks is conducted at higher temperatures, at which sulfur is not as poisonous to the catalyst. 2.2.2 Reaction Studies. Despite the large body of work on the effect of sulfur on Ni-based and precious metal catalysts for reforming of methane, there is relatively little work on sulfur deactivation in the reforming of liquid fuels. In most reported studies, model sulfur compounds are added to the liquid fuel to study their effect on deactivation. Sulfur in the fuel usually reacts to form hydrogen sulfide in the reformer, which is then removed downstream. In one such study, Palm et al.116 describe the ATR of mixtures of n-C13–C19

204

Catalysis, 2006, 19, 184–254

Figure 14 Effect of temperature on DG for sulfide formation for Ni, Pd, and Pt, according to reaction (12) (Calculations using HSC Chemistry 4.0153)

hydrocarbons that approximate the composition of diesel fuel. Changes in activity were measured for 100 h. The catalyst is described only as a ‘‘proprietary noble metal catalyst’’. The addition of 11 and 30 ppmw of benzothiophene to the inlet stream (which already contained B1 ppmw sulfur in the nC13–C19 feed) resulted in significant and only partially reversible deactivation (see Figure 15). The authors believe that sulfur selectively poisons the SR reaction (Reaction (2)), rather than the oxidation reactions (Reactions (1) and (4)). However, because the catalyst is neither described nor analyzed before or after the reaction, the cause(s) and extent of the deactivation are not clear. A similar study reports the results of adding 100 ppm thiophene to i-C8.137 As in the Palm et al. study,116 the catalyst is not described; rather, it is identified only as a ‘‘commercial naphtha reforming catalyst,’’ presumably Pt-based. In their reactor, the reformate from the ATR step passes through separate high and low temperature shift reactors before being analyzed. Thus, it was not possible to determine the effect of sulfur on the reforming step alone, nor was any post-reaction characterization of the catalyst reported, for example to determine coke or sulfur content. Figure 16 shows the observed deactivation, as measured by a decrease in H2 and CO concentrations. This decrease of H2 and CO, coupled with an increase in methane concentration, is attributed to deactivation of the shift catalysts and increased methanation reaction. However, this does not explain the increase in CO2 concentration, and it is difficult to envision that the addition of sulfur would increase the methanation activity. The reformate produced in the ATR step was not analyzed, making it difficult to interpret these results.

Catalysis, 2006, 19, 184–254

205

Figure 15 Effect of the addition of benzothiophene on conversion of n-C13–C19 mixture (O/C ¼ 0.86, S/C ¼ 2.20) (Reprinted from Palm et al.,116 copyright (2002), with permission from Elsevier)

Figure 16 Deactivation of a naphtha reforming catalyst in the reforming of i-C8; feed 100 ppm thiophene; 700 1C, 8,776 h1, S/C ¼ 3, O/C ¼ 1 (Reprinted from Moon et al.,137 copyright (2001), with permission from Elsevier)

2.2.2.1 Noble Metal Catalysts. Sulfur-containing liquid fuels can be successfully reformed using precious metal catalysts, typically Pt, Pd, Rh, and Ru, which appear to have greater resistance to sulfur poisoning than Ni-based catalysts. For example, Kopasz et al.78 reported that sulfur in the feed actually improved H2 production during ATR over Pt-containing catalysts, but reversibly poisoned Ni-containing catalysts. They also reported very little degradation of the Pt catalysts after extended operation (41,000 h) with sulfur containing feed (50 ppm sulfur in a benchmark gasoline fuel).

206

Catalysis, 2006, 19, 184–254

Encouraging results have been reported for other noble metal-based catalysts, though the catalyst composition is not provided in some cases. Researchers at Engelhard Corporation138,139 reported reforming a No. 2 fuel oil containing 1,200 ppm sulfur in a two-stage autothermal process. A preheated stream of steam, air, and No. 2 fuel oil (S/C ¼ 2.57 and O/C ¼ 0.82) was introduced into the first catalytic oxidation zone, which was comprised of Pt group metal catalysts dispersed on a lanthana-baria stabilized alumina washcoat. The conversion of O2 was complete, and a sufficient amount of the hydrocarbon was oxidized to raise the temperature high enough for SR to take place. The first stage effluent was then introduced into a second catalyst zone, which contains a Pt group metal SR catalyst. The hydrocarbon conversion was greater than 96% with a maximum H2 composition of 63% (N2 free basis) in the product gas stream. Similarly, a hydrocarbon feed consisting of JP-4 produced a H2 composition of 62% (N2 free basis), with complete conversion of the fuel. Similar results were achieved over a Rh/alumina monolith catalyst140 using catalytic POX for the reforming of a simulated JP-8 military feed containing 500 ppm of sulfur (as benzothiophene or dibenzothiophene). Stable performance for over 500 h with complete conversion of the hydrocarbons to syngas at 1,0501C, B0.5 s contact time, and LHSV of about 0.5 h1 was reported. At this high temperature, carbon formation was not reported and the sulfur exited as hydrogen sulfide. 2.2.2.2 Ni-based Catalysts. Other than the work of Kopasz et al. (discussed above), there are few systematic studies of the effect of sulfur on Ni-based catalysts used to reform liquid fuels. In one study, there is evidence of differences in sulfur tolerance between Ni-based catalysts and a comparable aluminate without a metal. Minet et al.141,142 developed a hybrid fuel processor for sulfur-containing hydrocarbon distillates, such as No. 2 fuel oil. The hybrid fuel processor consisted of a regenerative-tube high temperature steam reformer (HTSR) with an autothermal reformer, both using sulfur resistant catalysts. They utilized two different catalysts for the reforming process: calcium aluminate containing a high loading of calcium, and a Ni-based CaO–alumina catalyst. The regenerative HTSR contained calcium aluminate at the front-end, and Ni–CaO at the back-end, whereas the autothermal reformer contained only Ni–CaO. Tests were carried out at relatively high temperatures, typically more than 9001C. The calcium aluminate showed little carbon formation, as expected, but also low activity, with a high methanation selectivity. The Ni–CaO/ alumina catalyst was resistant to sulfur at temperatures above 9001C, and to carbon formation above about 9601C. No extended runs were carried out. 3

Steam Reforming

Steam reforming of hydrocarbons is one of the oldest and most widely used processes in the chemical industry for producing H2. In this process, hydrocarbons react with steam in the presence of a supported metal catalyst at elevated temperatures to generate primarily H2 and CO (reaction (2)). This

Catalysis, 2006, 19, 184–254

207

reaction step is typically followed by a series of WGS reactions (high then low temperature) (reaction (3)) to produce additional H2. Higher yields of H2 are obtained from catalytic SR than from other reforming reactions, since steam also contributes to H2 production. Brown102 compared seven common fuels as H2 sources for fuel cells in automotive applications: natural gas, methanol, ethanol, gasoline, jet fuel, diesel fuel, and H2 itself. This study showed that the gases produced by SR of these fuels contain B70–80% H2, whereas those obtained by POX contain B35–40% H2. Higher H2 content in the product gases results in higher efficiencies of fuel cells. Industrially, the SR reaction is carried out at high temperatures (600– 1,0001C) and high pressures (15–35 atm) in the presence of Ni-based catalysts.69,143 Cobalt and precious metals are also active for SR, but are more expensive.144 Other non-metallic catalysts have also been reported, but their activity is very low.111 The reaction of steam and higher hydrocarbons takes place by irreversible adsorption of the hydrocarbon onto the catalyst surface, followed by cleavage of terminal C–C bonds one by one until the hydrocarbon is converted into C1 components.145 Although this mechanism is general, the reaction rate of individual hydrocarbons over a given catalyst generally increases with carbon number. Higher hydrocarbons are also more prone to thermal cracking. The thermal or catalytic cracking of the higher hydrocarbons produces olefins or aromatics, which are precursors of carbon formation. It is not the intent of this work to discuss the details of industrial SR. Our intention is to present recent studies about SR of liquid hydrocarbons to produce H2 for fuel cell applications. The reader is therefore referred elsewhere for detailed discussions of industrial-based SR catalysts, mechanisms, and kinetics.96,98,111,146–150 3.1 Thermodynamics. – Several researchers have studied the thermodynamics of SR of higher hydrocarbons.116,151,152 A key conclusion from those studies is that high reactor temperatures and S/C ratios are required to preclude carbon formation during the SR process. Thermodynamic calculations presented here are based on Gibbs free energy minimization and were carried out using HSC Chemistry.153 The equilibrium amount of each species that is formed is normalized on the basis of one mole of n-C16, a model compound for diesel fuel, fed to the reactor. Carbon formation is a function of both the S/C ratio and reforming temperature. Figure 17 shows the minimum amount of S/C ratio thermodynamically required for carbon-free SR of n-C16 at a given temperature. Carbon-free operation of n-C16 is thermodynamically possible above the curve. Higher temperatures and S/C ratios inhibit carbon formation. 3.1.1 Effect of Steam to Carbon Ratio. Figure 18 shows the effect of S/C ratios on the product composition from SR of n-C16 at 8001C. At 8001C, the SR of n-C16 is thermodynamically possible without any carbon formation at a S/C

208

Catalysis, 2006, 19, 184–254

Figure 17 Carbon formation line for n-C16 SR (Calculations using HSC Chemistry 4.0153)

Figure 18 Effect of S/C ratio on n-C16 SR: products from 1 mole of n-C16 at 8001C (Calculations using HSC Chemistry 4.0153)

ratio greater than 1.1. High S/C ratios produce higher H2 yields, but require significant heat input that can make the process economically unfavorable. Therefore, in practice a S/C ratio between 2 and 3 is typically used for the SR of liquid hydrocarbons. 3.1.2 Effects of Temperature and Pressure. SR of n-C16 is shown as an example to illustrate the effects of temperature and pressure on the SR thermodynamics. The major reactions involved during SR are: C16H34 þ 16H2O - 16CO þ 33H2,

DH298 ¼ þ2474 kJ

(13)

16CO þ 16H2O 2 16CO2 þ 16H2,

DH298 ¼ 658 kJ

(14)

Therefore, the overall reaction can be written by combining reactions (13) and (15), C16H34 þ 32H2O - 16CO2 þ 49H2,

DH298 ¼ þ1816 kJ

(15)

Catalysis, 2006, 19, 184–254

209

Reaction (13) is highly endothermic, and reaction (14) is the mildly exothermic WGS reaction. Therefore, the overall SR reaction (15) is highly endothermic and H2 production is favored at higher temperatures. Figure 19 shows the effect of temperature on the equilibrium product composition from the SR of n-C16 at a S/C ratio of 2.5. At a given S/C ratio, H2 yields generally increase with temperature, but the reverse of the WGS reaction dominates at elevated temperatures (47751C). Consequently, the yield of H2 decreases with increasing temperatures in that temperature range, allowing a practical operating temperature range of 700 to 8001C for hydrocarbon SR. Due to the greater number of moles of product produced, versus moles of reactants consumed (i.e., 65 vs. 33 as in SR of n-C16), the SR reaction is favored at low pressures, as shown in Figure 20. However, industrial SR is carried out at high pressures (i.e., 15–35 atm) because much of the H2 produced is supplied in ammonia and methanol plants where higher pressures facilitate better heat recovery and result in compression energy savings.111

Figure 19 Effect of temperature on n-C16 SR: products from 1 mole of n-C16 at a S/C ratio of 2.5 (Calculations using HSC Chemistry 4.0153)

Figure 20 Effect of pressure on n-C16 SR: products from 1 mole of n-C16 at a S/C ratio of 2.5 and 8001C (Calculations using HSC Chemistry 4.0153)

210

Catalysis, 2006, 19, 184–254

3.2 Catalysts. – Group VIII metals, conventional base metal catalysts (Ni, Co, and Fe) as well as noble metal catalysts (Pt, Ru, Rh, Pd) are active for the SR reaction. These are usually dispersed on various oxide supports. g-Alumina is widely used but a-alumina, magnesium aluminate, calcium aluminate, ceria, magnesia, pervoskites, and zirconia are also used as support materials.111 The following sections discuss the base metal and noble metal catalysts in detail, focusing on liquid hydrocarbon SR for fuel cell applications. 3.2.1 Nickel and Other Base Metal Catalysts. Supported Ni is widely utilized as a catalyst for the industrial SR of hydrocarbons. The type of feedstocks and reaction conditions used for SR determine the choice of support, promoter, and loading of Ni. Typically, 15–25% nickel oxide loading is used in commercial SR catalysts. These supports must have high crush strength and stability so they can sustain severe reaction conditions. Alkali oxides such as K2O are used to minimize carbon formation on the Ni catalyst. The alkali may evaporate at elevated reaction temperatures; however, its loss can be controlled by adding acidic components, such as silica.111 Oxides of alkali earth metals, such as magnesia or calcia are also added to the support to neutralize highly acidic sites, which are mainly responsible for the carbon forming reactions. Compositions of various commercial Ni-based SR catalysts are listed in Table 4. Since SR is a highly energy-intensive process, heat must be transferred through the reactor wall and through the catalyst bed. Efficient radial heat transfer in the steam reformer helps to achieve a uniform reactor temperature zone, minimizing carbon formation in the catalyst bed associated with hot spots. Flytzani-Stephanopoulos and Voecks70 found honeycomb monolithic supports promising for SR catalysts in improving radial heat transfer in the reactor. Additionally, monoliths can provide faster response during transients in fuel cell applications due to their better heat transfer properties. They performed SR of n-C6, a gasoline surrogate, with a Ni catalyst supported over a conventional pellet support and two monolithic supports, a ceramic monolith of Ni impregnated on g-alumina washcoated cordierite, and a metal monolith of Ni supported on g-alumina washcoated Kanthal (Fe–Cr–Al alloy). A hybrid monolith comprised of metal monolith at the top, and ceramic monolith at the bottom provided higher H2 yields and n–C6 conversions compared to pellets, at a S/C ratio of 2.5, temperature of 9271C, and space velocity of 4,000 h1. Bimetallic Ni-based catalysts were also studied for SR of higher hydrocarbons in order to avoid the carbon formation and sulfur poisoning problems of conventional Ni catalysts.158–160 Murata et al.158 prepared a series of bimetallic catalysts by adding alkali and alkaline-earth metals to Ni catalyst supported on zirconia and alumina for SR of i-C8 and methylcyclohexane (MCH). The performance of various bimetallic catalysts for SR of i-C8 and MCH are summarized in Figures 21a and 21b, respectively. It was reported that the stability of Ni/ZrO2 is considerably improved by the addition of alkaline-earth metals (M), particularly strontium, to the catalyst with an M/Ni ratio of 0.5 by

211

Catalysis, 2006, 19, 184–254

Table 4

Composition of Some Industrial Steam Reforming Catalysts (NG ¼ natural gas, HC ¼ hydrocarbon, PR ¼ prereforming, LPG ¼ liquefied petroleum gas, SEC ¼ secondary reforming)

Supplier Haldor Topsoe154

Commercial Name R-67-7H AR-301 RKNGR

RK201

Johnson Matthey155,b

RK202 23-4 57-3 57-4 25-4 46-3Q

46-5Q

Sud-Chemie156

Unicat Catalysts157

46-6Q G-90 C11-9 G-91 C11-NK C14 NG-600-6H NG-605 NG-610-6H NG-610L NGR-612P3H/6H NGR-615-K NG-620-7H NGPR-1 NGR-600x

Feedstock

NiO wt%

Promoter, wt% Carrier

SiO2

NG to naphtha NG to light HC (PR) NG to Naphtha (PR) NG & LPG

412

–

MgAl2O4

o0.2

430%

2–5a

MgAl2O4

–

25%

–

MgO, Al2O3 (11 wt%)

–

415

o0.2

Naphtha NG & light HC NG & light HC NG & light HC Light HC Naphtha

415 18

K2O, MgAl2O4 40.3 K2O, 41 MgAl2O4 – Al2O3

16

–

CaAl2O4

o0.15

18

–

CaAl2O4

o0.1

18 23

o0.1 15

Light HC & Naphtha Naphtha NG NG NG/LPG Naphtha (SEC)

20

K2O, 1.8 CaAl2O4 K2O, 7; CaAl2O4c Na2O, o0.5 K2O, 1.8 CaAl2O4 – –

CaAl2O4 CaAl12O19 a-Al2O3 K2O, 1.6 CaK2Al22O34 K2O, 8.5 CaAl2O4 – a-Al2O3 La2O3, Al2O3 1–1.5 – Al2O3 – CaAl2O4 La2O3, 5.7 Al2O3 Al2O3 2d

0.5 – – – – – o0.1

K2O, 1.5–8 – Traced Cr2O7, 4

0.1– 0.2 – o0.05

16 14 18 18 25 12 8–10

Naphtha

414 415 12 15–17

Naphtha

15-20

(PR)

13–15 25 5

Al2O3 Al2O3 MgAl2O4 Al2O3

o0.2 o0.1

0.15

– o0.2 o0.2 0.2

a

Unspecified sintering inhibitors and promoters; b Johnson Matthey catalysts’ commercial names carry a prefix of KATALCO; c Support is promoted by zirconia (amount not specified); d Unspecified promoters.

weight. The NiSr/ZrO2 catalyst gave a steady H2 production for 100 h from SR of MCH without signs of deactivation. This catalyst was also effective for the SR of a surrogate gasoline fuel (30 vol% n-C6, 30 vol% i-C8, 35 vol% MCH, and 5 vol% benzene). The addition of Sr also led to reduced carbon formation:

212

Catalysis, 2006, 19, 184–254

48 wt% carbon on Ni/ZrO2, versus 11 wt% on NiSr/ZrO2 estimated by TGA. The formation of mixed oxide species such as SrZrO2 (cubic), SrZrO2 (tetragonal), and SrNiO2 (orthorhombic) was confirmed using XRD; reduced carbon formation and decreased sintering of Ni was attributed to those mixed oxide species present in the catalyst. In other studies by the same group,159,160 reforming of gasoline surrogates (75% MCH and 25% toluene) was conducted in a single step by integrating SR and WGS reactions over bifunctional catalysts, such as Ni–Re/Al2O3 and Ni– Mo/Al2O3. The WGS reaction is thermodynamically favored at low temperatures, whereas SR requires relatively high temperatures. Therefore, a dual catalyst bed separated by quartz wool at two different temperatures was used to optimize H2 formation by integrating SR and the WGS reaction. A high CO conversion of 83% at a hydrocarbon conversion of 69% was obtained on the two beds of Ni–Re/Al2O3 catalysts with a S/C ratio of 3.5 and bed temperatures of 5801C and 4801C, respectively.159 No catalyst deactivation was observed on

Figure 21 Performance of various bimetallic catalysts during SR (a) fuel: i-C8, 5501C for 5 h and (b) fuel: MCH at 7001C for 5 h (RUA: Ru/Al2O3) (Reprinted with permission from Murata et al.,158 copyright (2004) American Chemical Society)

Catalysis, 2006, 19, 184–254

213

either catalyst, Ni–Mo/Al2O3 or Ni–Re/Al2O3, for SR of sulfur-free fuels over 200 h of operation, but the catalyst activity declined sharply during SR of fuel containing 5–10 ppm sulfur. The sulfur tolerance of both catalysts was significantly improved by addition of 10–20 wt% ZSM-5. The role of ZSM-5 was to promote the cracking of complex organic sulfur compounds, such as thiophene, into elemental sulfur, which adsorbed in the channels of ZSM-5, instead of binding with metal catalysts, and because sulfur can be removed from the surface of the catalyst. In addition to Ni catalysts, Lee and Park161 explored some unconventional catalysts, such as limestone, dolomite, and iron ore, in a fluidized bed reactor to carry out SR of kerosene and bunker oil. H2 yields from SR of bunker oil over various catalysts (temperature ¼ 8001C, bed height ¼ 10 cm, superficial gas velocity ¼ 20 cm/sec, and S/C ¼ 1.6) were: sand (20%), iron ore (29%), commercial Ni catalyst (89%), limestone (93%), and dolomite (76%). Limestone as a SR catalyst looked very promising, but H2 yields over a limestone catalyst decreased over time due to elutriation of fines during the reaction. A fluidized-bed reactor was advantageous for reforming of higher hydrocarbons, due to its ability to replace coked catalyst with fresh catalyst during operation. 3.2.2 Noble Metal Catalysts. As discussed in Section 2 (Deactivation), coke formation and sulfur poisoning during hydrocarbon SR are key problems with Ni-based catalysts. They may be controlled by picking proper reforming conditions, such as the S/C ratio, or by using carbon-resistant catalysts. However, a higher S/C ratio makes the reforming process economically unfeasible. Because of their ability to inhibit coke deposition, noble metal-based catalysts have been reported to be more effective and have been proposed to replace conventional Ni-based catalysts for SR in fuel cell applications. Ru-based catalysts have been used for SR of higher hydrocarbon fuels to mitigate the catalyst deactivation caused by coking and sulfur poisoning.35,39,162,163 Suzuki et al.35 have carried out a long-term study (8,000 h) of SR of desulfurized kerosene (C10H22 with o0.1 ppm sulfur) at 8001C (S/C ¼ 3.5) using a highly dispersed Ru/CeO2–Al2O3 catalyst (2 wt% Ru, 20 wt% CeO2, balance Al2O3), reporting that the sulfur resistance was significantly improved through the addition of CeO2. The conversion of kerosene was maintained at approximately 97–99% and the composition of H2 in the product stream was approximately 70% over the duration of the run. When the same catalyst was used for SR of kerosene containing 51 ppm sulfur, the conversion of kerosene decreased to 85.5% after 25 h on stream, whereas that on Ru/Al2O3 (without CeO2) was 60% after 25 h. Suzuki’s group also evaluated lanthanum oxide supported precious metals (Rh, Ru, and Ir) catalysts for SR of n-C12—a surrogate of kerosene—at a temperature range of 500–8001C.162 At SR conditions (S/C ¼ 3, 6001C), Rh and Ru catalysts produced higher hydrocarbon conversions (B95%) and H2 yield (4100%) than those with Ir (B60% and B80%, respectively). However, at autothermal conditions (S/C ¼ 1, O/C ¼ 1, 6001C), the conversion of n-C12 reached 95% with 60% H2 yield over the Ir catalyst.

214

Catalysis, 2006, 19, 184–254

Ceria-supported Pt and Pd catalyst formulations have also attracted interest for SR because of the oxygen-conducting properties of ceria. Wang and Gorte164 reported that the ceria-supported Pt (1 wt%) and Pd (1 wt%) catalysts are more active for SR of liquid fuels than the alumina-supported metal catalysts. There was, however, no distinct effect of the choice of Pt or Pd on the reaction. The primary role of ceria is its ability to dissociate water and then transfer the oxygen produced by that reaction to the supported metal. The transfer of oxygen to the metal helps to maintain a carbon-free surface. Spectroscopic results show that ceria is covered by a carbonate layer under SR conditions, which could limit the rate at which ceria can be reoxidized. Higher reaction rates may be possible if the carbonate layer formation is inhibited. Mixed oxides of ceria and zirconia were recommended for further investigation for SR, since mixed oxides may hinder carbonate formation. Ceria-supported metal catalysts are also active WGS catalysts and are thus advantageous for fuel cell applications, since most of the CO can be converted during the main reforming reaction. 4

Partial Oxidation

Catalytic POX is an attractive option for H2 and CO production from liquid hydrocarbon fuels when compact or mobile fuel processing systems are desirable. This process has the advantage of inherently rapid reforming kinetics with rapid light-off characteristics. Since there is no need to feed water as in autothermal or SR, POX reactors are easily integrated into transportation– based, onboard fuel reforming systems. Two major issues that distinguish higher hydrocarbon POX from methane POX are the propensity to coke the catalyst, and the presence of residual organic sulfur compounds, such as dibenzothiophene, that are not easily removed through hydrodesulfurization. In industry, the POX of heavy liquid petroleum feedstocks is carried out without the use of a catalyst at temperatures between 1,250 and 1,5001C and pressures between 25 and 80 atm.165 In smaller systems, the introduction of a catalyst is necessary for operational stability, and to reduce the operating and light-off temperatures. The reactions occurring in the POX of liquid hydrocarbons are extremely complex. However, the overall catalytic POX reaction can be described by the simple reaction (1). This is followed by the WGS reaction (reaction (3)) and the methanation reaction. The associated temperature rise from reaction (1) also results in the fragmentation of the heavier molecular weight hydrocarbons. These fragments react with the steam and CO2 produced by reactions (2) and (5) to form CO and H2. To avoid excessive fragmentation of the feed, which could result in the formation of unsaturated hydrocarbons and carbon deposits on the catalyst, feed preheat temperatures for liquid fuels range from ambient to just above their boiling point. Reaction temperatures are typically in the range of 700 to 9001C. If organic sulfur is present in the feed, then the reactor is typically operated at higher temperatures, where metal sulfides are less stable. The O2 in

Catalysis, 2006, 19, 184–254

215

the feed initially converts organic sulfur compounds in the feed into SO2; however, SO2 may later be converted into H2S. 4.1 Thermodynamics. – Thermodynamic calculations on the reformate product distribution and carbon formation characteristics for n-C16, a model diesel fuel compound, were examined over the temperature range 500 to 9001C and over an O/C ratio of 1 to 1.5 using HSC Chemistry.153 The equilibrium amount of each species formed was normalized on the basis of one mole of n-C16 fed to the reactor. 4.1.1 Effect of the Oxygen to Carbon Ratio. The effects that the O/C ratio has on carbon formation and product distribution for n-C16 POX are shown in Figure 22. Conversion of n-C16 is complete over the range of O/C ratios studied. Thermodynamically, carbon formation is avoidable with O/C ratios greater than 1.1, with high H2 and CO yields near this ratio. At higher O/C ratios, both the propensity for carbon formation and the methane selectivity decrease. The selectivity toward combustion products, CO2 and H2O, also increases with a corresponding decline in the selectivity toward H2 and CO. 4.1.2 Effect of Temperature. Figure 23 shows the effect of temperature on n-C16 POX over a temperature range of 500 to 1,0001C, at an O/C of 1.2 and a pressure of 1 atm. At all temperatures, n-C16 conversion was complete. H2 and CO selectivities increase from 500 to 7501C, after which they remain essentially constant. Carbon formation is avoided by operating above 7501C. From 500 to 7501C, selectivities toward H2O and CO2 diminish. Selectivity toward CH4 also diminishes from 500 to 8251C, where it essentially reaches zero. 4.2 Catalysts. – The catalysts used in the POX process are typically noble metals of Pt, Ru, Pd, or Rh and transition metals such as Ni dispersed on an appropriate support.

Figure 22 Product distribution for the POX of n-C16 at 8001C and 1 atm (Calculations using HSC Chemistry 4.0153)

216

Catalysis, 2006, 19, 184–254

Figure 23 Product distribution for the POX of n-C16 at an O/C ratio of 1.2 and pressure of 1 atm (Calculations using HSC Chemistry 4.0153)

4.2.1 Nickel Catalysts. Ni-based catalysts are typically modified with various promoters to limit coke formation, often by neutralizing acid sites thought to be responsible for coking. n-Heptane POX was studied over NiLiLa/g-alumina by Ran et al.166 The catalyst was tested over the temperature range of 700 and 8501C at an O/C ¼ 1 and a GHSV of 38,000 cm3/g/h. The Li–La promoter was added to both disperse the Ni over the surface of the catalyst, and to titrate active Lewis acid sites on the catalyst surface. This catalyst converted 100% of the n–C7 in the feed with 93% H2 selectivity over a period of 4 h. Similar results were obtained from n-C7 POX when the catalyst LiLaNiO/g-Al2O3 was combined with a dense oxygen permeation membrane Ba0.5Sr0.5Co0.8Fe0.2O3 that can supply pure O2 for the reaction.167 The H2 selectivity of Ni/g-alumina, also tested in the study, dropped to 30% after 4 h.166 Another approach to limit coke formation is to incorporate Ni into a cokeresistant lattice. n–C14 POX was studied over LaNiAl11O19 by Gardner et al.168 Carbon formation resistance was achieved by dispersal of the Ni directly into the lanthanum hexaaluminate structure. The temperature programmed reduction profile of LaNiAl11O19 in H2 exhibited a single broad reduction peak for Ni21 to Ni0 centered at 9961C, indicating that the substitution of Ni into the hexaaluminate structure imparted a high degree of reduction stability. The catalyst was tested isothermally at 8501C, O/C ¼ 1.2 and a GHSV ¼ 10,000 cm3/g/h. The average H2 selectivity obtained over 24 h was 66.2% and CO was 60%. The H2/CO ratio remained unchanged over this time period at 1.18. 4.2.2 Noble Metal Catalysts. Rh-based catalysts have been investigated on different supports, resulting in different H2 and CO yields. Gasoline and naphtha POX over a supported Rh catalyst were reported by Fujitani et al.169 For g-alumina supported Rh catalyst, maximum yields of 96% of both H2 and CO were reported with 0.2 wt% Rh loading at 7001C, an air equivalence ratio of 0.41, and a liquid hourly space velocity (LHSV) of 2 h1. A 0.05 wt% Rh supported on zirconia yielded 98% H2 and 85% CO at 7251C, an

Catalysis, 2006, 19, 184–254

217

equivalence ratio of 0.41, and a LHSV of 2 h1. However, 0.1 wt% Rh loaded on a honeycomb structure of a-alumina-magnesia gave the highest yield of H2 and CO (both 98%) at 8201C, air equivalence ratio of 0.41, and a LHSV of 2 h1. Furthermore, carbon deposition was not observed with these supported Rh catalysts. Tanaka et al.170 reported that gasoline POX over Rh, Pt, and Pt-Rh is promoted by alkali (Li) and alkaline earth metals (Ba, Ca, K) supported on magnesium aluminate spinel. The catalysts were tested isothermally at 8001C at an air to fuel ratio of 5.1 and GHSV 50,000 h1. Li, Mg and MgLi promoters were added to Pt supported on MgAl2O4 spinel. All catalysts produced similar H2 and CO reformate concentrations of 23 and 25 vol%, respectively. There was a discernable difference in the carbon deposition. The unpromoted Pt catalyst showed carbon levels of 0.02 wt% carbon, where the alkali and alkaline earth promoted Pt catalysts had carbon levels of 0.01 wt%. In another study by the same group,170 K, Ca and CaK promoters were added to Rh supported on MgAl2O4 spinel. Both modified and unmodified catalysts produced similar H2 and CO reformate concentrations of 23 and 25 vol%, respectively. The different modifiers did affect carbon production on the catalysts. The unpromoted Rh catalyst showed carbon levels of 0.03 wt% carbon, where the RhK catalyst had 0.02 wt% carbon, the RhCa 0.015 wt%, and the RhCaK only 0.01 wt%. Bimetallic Pt–Rh with Li, Ba and Li–Ba modifiers supported on MgAl2O4 spinel were also examined for H2 and CO yields and carbon formation resistance, with similar results to previous work for the yields. The unpromoted Pt–Rh catalyst showed carbon levels of 0.01 wt% carbon, where the promoted Pt–Rh catalysts all showed reduced coking levels of 0.005 wt%. The POX reaction has also been studied using reactors with very short contact times. Cyclohexane, n-C6, n-C8, n-C10, n-C16, i-C8, toluene, naphthalene, and gasoline POX has been studied over Rh-based monolithic catalysts at millisecond contact times.171–174 Several factors affect the conversion and selectivity of these fuels. The mean cell density, typically defined as pores per inch (ppi) for foam materials [e.g., 40 ppi corresponds to a mean cell diameter of B0.6 mm], significantly affected the syngas selectivities, but the gas space velocity did not. Table 5 shows the effect that pore density has on CO and H2 selectivity. The higher density 80 ppi washcoated alumina monolith containing 5 wt% Rh catalyst proved to be the best for POX. Higher conversions of fuels (495%) and syngas yields 490% were obtained even at very high space velocities (3,000,000 h1). The POX of commercial grade gasoline yielded 75% selectivity to CO, but the sulfur and metal contaminants present in gasoline poisoned the catalyst over time. POX of a 75% i-C8 þ25% toluene blend was conducted to simulate the aromatics contained in fuels. The lower conversion of toluene compared to i-C8 was attributed to the absence of secondary or tertiary C–H bonds that facilitate adsorption and reaction. The transient lightoff experiments showed that the steady state temperature can be attained starting from room temperature in less than 5 seconds in these millisecond catalytic POX reactors; homogeneous combustion rapidly heats the catalyst.

218

Catalysis, 2006, 19, 184–254

Table 5

Effect of Catalyst Pore Density on CO and H2 Selectivities171,a Selectivity (%)

Rh catalyst density (ppi)

Pore diameter (mm)

Surface area (m2/g)

Back-face temp (1C)

H2

CO

CO2

20 45 80 80 with washcoat

B800 B350 B200

0.2–0.4 0.2–0.4 0.2–0.4

930 910 843

70 74 84

75 78 83

11 10 9

B100–150

2–5

795

93

90

9

a i-C8-air feed at O/C ratio of 1, space velocity of 730,000 h1, and feed temperature of 2501C. Conversions of both i-C8 and O2 were nearly 100% for all fuels.

Krummenacher et al.74 have reported syngas selectivities greater than 80% with greater than 99% conversion of hydrocarbons from the catalytic POX of n-C10 and n-C16 over a Rh-coated monolith at 5 to 25 ms contact times. POX of a high grade diesel fuel (10 ppm sulfur, 8% aromatics, 90% alkanes) produced syngas at greater than 98% fuel conversion. Maximum selectivities of H2 and CO observed were 70% and 80%, respectively, at an O/C ratio of 1.4 and 25 ms contact time. 5

Autothermal Reforming

ATR is an extremely complex process. The overall reaction is made up of a series of sub-reactions including total oxidation (which almost inevitably takes place first in the reactor), POX, SR, thermal and catalytic cracking, dehydrogenation, WGS, methanation, carbon formation/ deposition, and others. Add to this the fact that an operating system may always be in a transient mode, and the complexity of the reformer becomes apparent. The discussion here is limited to catalysts for the overall reforming reaction of liquid fuels, i.e., reaction (16). Issues such as reactor design, heat integration, and transient behavior are not discussed, although they are critical to the design of a practical reformer. The main individual reactions that take place in the reformer (e.g., reactions (1), (2) and (5)) will be considered separately from the overall autothermal reaction for two reasons. First, in ATR the reactor can be considered as two plug-flow reactors in series: (1) a very fast POX reaction occurs at the top of the catalyst bed and utilizes a small portion of the bed; and (2) a slow SR utilizes the remainder of the reactor bed. Therefore, an optimal ATR catalyst must have excellent SR catalytic properties. Second, there may be situations in which liquid fuels are reformed using only these individual reactions; e.g., diesel fuel may be reformed using only SR (reaction (2)) or only by POX (reaction (1)). 5.1 Thermodynamics. – The thermodynamics of ATR have been reviewed by a number of researchers.71,73,152,175–181 Collectively, these papers provide guidance on the optimal operating temperature and the effects of steam and air concentrations on reformate composition. The effects of pressure on H2

219

Catalysis, 2006, 19, 184–254

production do not appear to be significant.180 The effects of some of the more important parameters on ATR thermodynamics are discussed below in detail. 5.1.1 Effect of Air/Fuel Ratio. ATR28–31,72,182 is the reaction of a fuel with proportions of steam and air/O2 that results in a thermally neutral overall reaction. To illustrate the overall stoichiometry, consider the reaction of a liquid hydrocarbon with air to produce only H2 and CO2 in the reformate: Cn Hm þ lðO2 þ 3:76N2 Þ þ ð2n 2lÞH2 O m ! nCO2 þ 2n 2l þ ð16Þ H2 þ 3:76l N2 2 where l is the air to fuel ratio, which can be used to control the amount of heat generated and thus the reactor temperature. Therefore, 8 0 > steam reforming > > < 1 DHf ;diesel DHf ;CO2 nþ2 thermoneutral DHf ;H2 O ð17Þ l¼ > n partial oxidation > > :n þ m combustion 4 An important advantage of ATR is direct thermal integration of the heat generated by POX with the catalyst bed, which significantly simplifies start-up procedures and reduces transients during load changes.183 The air to fuel ratio, l, is adjusted to make ATR thermoneutral. The heat generated in POX compensates for the heat required by the SR, so that the overall heat of reaction becomes zero. For l ¼ 0, reaction (16) is reduced to the SR, which is highly endothermic. The reaction becomes less endothermic as the value of l increases and becomes thermoneutral at a value of l which depends on the specific hydrocarbon fuel. For higher values of l, the reaction becomes highly exothermic. The value of l for thermal neutrality is 0.23 for methanol,40 0.37 for i-C8,40 and 0.35 for n-C16. Figure 24 shows the effect of O/C and S/C ratios on the H2 and CO produced under equilibrium conditions from ATR of 1 mole of n-C16 at 8001C. Increasing the O/C ratio decreases H2 and CO production, while increasing the S/C ratio increases H2 and decreases CO formation. 5.1.2 Effect of Temperature. To illustrate the effects of temperature on product composition from ATR, consider the reforming of n-C16. The reactions that take place in the reformer can be summarized by the following: C16 H34 þ

49 O2 ! 16CO2 þ 17H2 O 2

ð18Þ

C16H34 þ 16H2O - 16CO þ 33H2

(19)

CO þ H2O 2 CO2 þ H2

(20)

1 CO þ O2 $ CO2 2

ð21Þ

220

Catalysis, 2006, 19, 184–254

1 H2 þ O2 ! H2 O 2 CO þ 3H2 2 CH4 þ H2O

ð22Þ (23)

For n-C16, reaction (16) becomes thermoneutral when l ¼ 0.353, and the reaction can be written as: C16H34 þ 20.70H2O þ 5.65O2 - 16.00CO2 þ 37.70H2

(24)

An equilibrium analysis of the composition of the reformate as a function of temperature, at 1 atm absolute pressure, is shown in Figure 25. This analysis shows the following: (i) At all temperatures from 200 to 1,0001C, neither O2 nor n-C16 is present at equilibrium, as expected.

Figure 24 The effect of O/C and S/C ratios on the H2 and CO formation from ATR of 1 mole of n-C16 at 8001C (’ S/C ¼ 1, E S/C ¼ 1.5, m S/C ¼ 2, and K S/C ¼ 2.5) (Calculations using HSC Chemistry 4.0153)

Figure 25 Effect of temperature on ATR: products from 1 mole of n-C16 at thermoneutral conditions (Calculations using HSC Chemistry 4.0153)

Catalysis, 2006, 19, 184–254

221

(ii) Above B4001C, methane decreases, reaching essentially zero at B7001C. (iii) The maximum H2 concentration (at B8001C) is about 43% (47% on a dry basis), at which the corresponding CO concentration is B36% (39% dry basis). (iv) Above B8001C, the reverse shift reaction results in H2 loss. Other reactions may also occur. These include carbon formation, hydrocracking or thermal cracking, dehydrocyclization of paraffins to naphthenes, and dehydrogenation of naphthenes to aromatics. These have been discussed in the deactivation of reforming catalysts, in Section 2. 5.2 Catalysts. – The catalysts that have been studied for ATR of liquid fuels can be divided into base and noble metal materials. The following summary will show that neither type of catalyst has been optimized for ATR of liquid fuels. The studies reported to date have largely been carried out using Ni-based methane-SR catalysts, simple (or proprietary) noble metal formulations. Much remains to be learned about catalysts for this reaction. 5.2.1 Base Metal Catalysts. – 5.2.1.1 Nickel-based Catalysts. Studies on Ni-based catalysts at conditions of interest here are limited to simple Ni/ alumina, and commercial methane-SR catalysts. For example, Yanhui and Diyong135 studied the ATR of n-C8, used as a surrogate for gasoline fuel over Ni/Al2O3 (with Ni loadings from 1–10%). Conversion and selectivity increased with Ni content and temperature, although there was little difference for Ni content greater than 5%. Maximum conversion was B80% (at 7001C), with a corresponding selectivity (to CO þ H2) of 60%. The concentrations of H2 and CO are not given. Similar results were reported by Zhang et al.134,184 using a Ni/g-alumina catalyst prepared by conventional impregnation methods. Using commercial Ni-based methane-SR catalysts obtained from vendors, Flytzani-Stephanopoulos and Voecks30,185 studied the ATR of n-C14. The catalysts in this study varied significantly in Ni content, level of promoter, type of support, and physical properties, making direct comparisons difficult. In addition, their reactor bed used three different catalysts in contiguous reactor zones to promote different reactions from the inlet to the outlet. They showed that POX takes place at the bed entrance, followed by SR. They used a low activity catalyst (7% NiO/93% ZrO2) in the first zone to avoid an abrupt increase in reactor temperature that may cause carbon formation. Catalysts with different Ni loadings were used in the POX reaction zone (middle zone). A catalyst with lesser Ni content (11% NiO/77% Al2O3/12% CaO) was utilized for the feed containing aromatics to alleviate the large heat release from the oxidation of aromatics. For n-C14 reforming, a catalyst with higher Ni (21% NiO/14% SiO2/29% Al2O3/7% K2O/13% MgO/13% CaO/3% Fe2O3) content was used. A highly active SR catalyst (31% NiO/60% Al2O3/7% MgO/1% SiO2) was used in the bottom of the reactor. Results also showed that benzene is actually formed within the bed from n-C14. H2 and CO yields were B15 and 10

222

Catalysis, 2006, 19, 184–254

mol/mol n-C14, respectively, at an air-to-carbon ratio of 1.8, S/C ratio of 0.6, preheat temperature of 6001C, and space velocity of 22,000 h1. Trace levels of methane, ethane, and ethylene were formed as well. The product composition was within 10% of equilibrium predictions, with the exception of hydrocarbon conversions, which were lower than expected. In another study, Ni-based catalysts doped with small amounts of transition metals (Ni (11.8 mol%)/M (2.9 mol%)/MgO/Al2O3, where M ¼ Fe, Co, Mo) showed high H2 selectivities for ATR of i-C8 compared to a Ni/MgO/Al2O3 catalyst (see Table 6).186,187 Results suggest that the addition of a small amount of transition metal facilitates the WGS reaction, which results in higher H2 selectivities. These catalysts also exhibited stable performance over extended periods with sulfur-free fuels. There were no phase changes observed after the reforming reaction. However, the presence of sulfur in the fuel (100 ppm sulfur in i-C8) caused a slight decline in the catalyst activity over time. In another study, Moon et al.188 used a commercial methane reforming catalyst (NiO/ SiO2/MgAl2O4) to investigate the effect of fuel components and sulfur impurities found in gasoline on the carbon formation. It was found that carbon formation occurs at lower temperatures (o6401C) and is promoted by sulfur impurities in fuels. Hence, it was recommended to maintain an operating temperature of 47701C to minimize carbon formation due to sulfur poisoning in the ATR of gasoline. 5.2.1.2 Perovskites. Recently, researchers at Argonne National Laboratory (ANL) proposed the use of oxide ion-conducting perovskites as ATR catalysts of liquid hydrocarbons.132,189,190 Typically, perovskites are of the form ABO3, where A is a lanthanide element, and B is a first row transition metal. The A and/or B components may be doped with other transition metals to enhance the stability and performance of perovskite-type materials. Undoped perovskites, LaCoO3 and LaNiO3, produced the highest yields of H2, but they were not stable at the autothermal conditions, and decomposed into La2O3 and metallic Co and Ni, respectively. However, replacing most of the Ni with Cr (LaCr0.9Ni0.1O3) improved the structural stability with a minor reduction in H2 production, compared to LaNiO3. Mawdsley et al.189,190 investigated the effects of replacing La with Sr and Ni with Mn, Co, Cr, or Fe. They reported that La0.8Sr0.2Cr0.9Ni0.1O3 provided a stable H2 production during ATR of gasoline Table 6

Product Distribution (vol%) from the ATR of i-C8 Using Ni/M/MgO/ Al2O3 (M ¼ Fe, Co, and Mo) (T ¼ 7001 C, space velocity ¼ 8,776 h1, S/C ¼ 3, and O/C ¼ 1)187

M

H2

CO

CO2

CH4

a

– Feb Cob Mob

47 63 64 62

44 12 12 12

3 23 23 23

6 1 1 2

a

Ni content ¼ 14.7 %;

b

Ni content ¼ 11.8% and M content ¼ 2.9%.

Catalysis, 2006, 19, 184–254

223

for 31 h. However, perovskites are not sulfur resistant, and were poisoned at even 5 ppm sulfur levels in fuel (see Figure 26). Recently, Qi et al.191 used Ce to partially substitute La in LaNiO3 perovskite, enhancing its activity and stability for gasoline ATR. They observed a pure perovskite phase when La was substituted by 10% Ce. However, both NiO and CeO2 started segregating from the main perovskite phase of La1xCexNiO3 as x increased above 0.1 and eventually no perovskite phase was observed at x ¼ 0.5. Yields of CO þ H2 increased with increasing x up to 0.2, but further increasing x resulted in lower yields of CO þ H2 (see Table 7). Ce-substituted perovskites showed no deactivation during ATR of n-C8 for 220 h, while performance of LaNiO3 declined considerably after 8 h of operation (see Table 7). Like Wang and Gorte,164 they also attributed the improved performance of Ce-substituted perovskites to the enhanced oxygen mobility in ceria-based catalysts. However, the Ce-substituted perovskites were not sulfur tolerant, which is consistent with the ANL study.190 5.2.2 Noble Metal Catalysts. Noble metal-based catalysts have been widely used in reforming reactions, and are logical choices for ATR. Results of reaction studies generally suggest that these catalysts are comparable in activity to Ni-based catalysts, but they appear to be somewhat more resistant to deactivation. 5.2.2.1 Platinum and Palladium Based Catalysts. Researchers at ANL have also developed an ATR catalyst formulation comprised of a transition metal element supported on an oxide ion-conducting substrate, such as ceria,

Figure 26 The effect of 5 and 50 ppm sulfur on H2 production from ATR of a gasoline benchmark fuel over perovskite catalysts (La0.8Sr0.2Cr0.9Ni0.1O3, La0.8Sr0.2Mn0.9Ni0.1O3, and La0.8Sr0.2Fe0.9Ni0.1O3)190

224

Table 7

Catalysis, 2006, 19, 184–254

Effect of Partial Substitution of La with Ce on Perovskite (La1xCexNiO3) Activity (T ¼ 8001C, O/C ¼ 0.76, S/C ¼ 2.0, and GHSV ¼ 8,000 h1)191 Yield of (COþH2), mol/mol C

n-C8 conv. (%)

x

Initial

8 hr

Initial

8 hr

0 0.1 0.2 0.5

2.34 2.30 2.36 2.12

2.20 2.30 2.36 2.12

100 100 100 98

98 100 100 98

lanthana gallia, or zirconia.77,132,133,190,192–197 The oxide-ion conducting support is doped with a small amount of gadolinium (Gd) or samarium (Sm) to improve its oxide ion conductivity by increasing the number of surface oxygen vacancies.196 The Sm–Gd doped ceria (Ce0.8Sm0.15Gd0.05O2) was evaluated for its catalytic activity for ATR (O/C ¼ 0.92, S/C ¼ 1.14, and SV ¼B3000 h1) of i-C8.77,193,197 The i-C8 conversions and H2 selectivities at different temperatures over various transition metals are shown in Figures 27a and 27b, respectively. The base metals (i.e., Fe, Co, Ni) have activities comparable to noble metal catalysts, but noble metal catalysts demonstrated higher H2 selectivities. This was attributed to the high surface area of the noble metal catalysts.193 Although the authors also suggested that high-surface area base metal catalysts can be prepared using alternative synthesis methods since most of these base metals are more miscible with CeO2 than noble metals. However, noble metal catalysts are more active for H2 dissociation and spillover on CeO2 than Co or Ni which can account for their better performance rather than any surface area effects. Furthermore, they evaluated several supports while using Pt in order to investigate the effect of the oxide ion-conducting supports on the reforming reaction. As shown in Figure 28, the Sm–Gd-doped ceria showed highest yields of H2 and the least hydrocarbon slip among all oxide ion-conducting supports employed for ATR.195 Also, Pt supported on Gd-doped ceria demonstrated activity and stability for 1,000 h on stream for a surrogate diesel fuel containing 50 ppm of sulfur. To minimize diffusional resistances and improve catalyst stability, the ANL group developed monolithic and spiral microchannel configurations.195 The ANL catalyst (not identified, but presumably a Pt supported on Gddoped ceria) was also successfully used for ATR of diesel fuel.31,198 Tests of three different types of diesel fuels (n-C16, low-sulfur diesel, and regular diesel) showed complete conversion of hydrocarbons at 8001C. The diesel surrogate n-C16 yielded 60% H2 on a dry, N2-free basis at 8001C, whereas the other two diesel fuels required higher temperatures (48501C) to yield similar levels of H2 in the product gases. Similar or improved H2 yields from diesel ATR were observed with a microchannel monolith catalyst, compared with extruded pellets in a fixed-bed reactor.31 Moon et al.137 studied the ATR of i-C8 over a ‘‘commercial naphtha reforming catalyst’’ that is not further identified (presumably a Pt-based

Catalysis, 2006, 19, 184–254

225

Figure 27 ATR of i-C8 as a function of temperature for various transition metals supported on the Sm-Gd doped ceria substrate (a) i-C8 conversion and (b) H2 selectivity (Reprinted from Krumpelt et al.,77 copyright (2002), with permission from Elsevier)

Figure 28 Product composition from ATR of i-C8 over Pt- supported on various oxide ion substrates195

226

Catalysis, 2006, 19, 184–254

catalyst). Unfortunately, in their experimental system, the reformate from the ATR step passes through a high and low temperature shift reactor before being analyzed. Thus, it is not possible to determine the product spectrum for the ATR step alone, except for one of their tests. In that test, they reported an exit product composition from the ATR section of the reactor (which they call ‘‘POX’’) that contains 30% air, in addition to H2, CO, CO2 and trace levels of CH4. A comparison of Al2O3-supported Pd and Pt showed that their activity is comparable to or greater than Ni catalysts up to 6001C, but they were less active at higher temperatures.135 Interestingly, bimetallic Ni–Pd catalyst had much greater activity than either the Pd/Al2O3 or Ni/Al2O3 (Figure 29a) though the difference in reaction conditions among the studies does not allow a direct comparison of the results. Moreover, the metal loadings of the catalysts were not provided. However, this result is in agreement with Zhang et al.134,184 who studied a Ni–Pd bimetallic catalyst and compared it to a Ni catalyst on an identical alumina support. They observed a similar improvement in activity and stability for the Ni–Pd catalyst, but the difference was not as dramatic as shown in Figure 29a. As expected, selectivity (to CO þ H2) depended on the O/C and S/C ratios, but was independent of temperature (Figure 29b). However, the Ni– Pd bimetallic catalyst showed a better performance than the Ni–Pt or individual metals over 500 h on stream for ATR of n-C8.134,184 A direct comparison of Al2O3-supported Pt, Pd, and Ru suggests that Ru is the most active metal for diesel reforming, at least on this support. Berry et al.199 studied diesel reforming at a temperature range of 750 to 8501C and GHSVs of 25,000 to 200,000 h1. Activity increased in the order: Pd o Pt o Ru. Complete conversion of diesel was obtained at 8501C and space velocity of 50,000 h1 from the ATR of diesel over a g-alumina supported Ru catalyst. Encouraging results have been reported on other noble metal-based bimetallic catalysts. Researchers at Engelhard Corporation138,139,200 reported reforming of a No. 2 fuel oil containing 1,200 ppm sulfur in a dual-bed autothermal process. A preheated stream of steam, air, and No. 2 fuel oil (S/ C ¼ 2.57 and O/C ¼ 0.82) was introduced into the first catalytic oxidation zone, which was comprised of Pt group metal catalysts (Pt/Pd in equal portions by weight) dispersed on a lanthana-baria stabilized alumina washcoat. The conversion of O2 was complete, giving a temperature high enough for SR. The first stage effluent was then introduced into a second catalyst zone which contains a Pt group metal SR catalyst (Pt–Rh/Al2O3). The hydrocarbon conversion was greater than 96% with a maximum H2 composition of 63% (N2 free basis) in the product gas stream. Over the same catalyst-bed configuration, a JP-4 hydrocarbon produced a H2 composition of 62% (N2 free basis) with a complete conversion of the fuel.106 Similar improvement in Pt-based catalyst activity by the addition of a second metal has been shown for Pd–Pt/alumina and Ni–Pt/ceria.201–203 High H2 yields from ATR of diesel, compared to monometallic catalysts, were shown at S/C ¼ 3, O/C ¼ 1, preheat temperature ¼ 4001C, and space velocity ¼ 17,000 h1 (Figure 30). Based on TPR and XPS studies, the superior performance of the

Catalysis, 2006, 19, 184–254

227

Figure 29 ATR of n-C8 (a) Activity comparison of reforming catalysts; J Ni–Pd, K Ni, and m Pd (note: Ni-Pd catalysts tested under ‘‘non-reduction’’ conditions; Ni and Pd under ‘‘reduction’’ conditions) and (b) Selectivity of Ni-Pd catalysts; D O/C ¼ 0.25 and S/C ¼ 5; K O/C ¼ 0.25 and S/C ¼ 3; } O/C ¼ 0.50 and S/C ¼ 3 (Reprinted from Yanhui and Diyong,135 copyright (2001), with permission from Elsevier)

bimetallic catalysts was attributed to a strong metal-metal interaction in the bimetallic sample. The order of impregnation had no affect on the performance of Pt–Pd catalysts (Figure 30a), but interestingly, the impregnation order in a Pt–Ni bimetallic catalyst significantly affects selectivity for H2 production (Figure 30b). Higher H2 yields were always observed when the Pt was impregnated second. Also, the Pt–Ni/ceria catalyst showed better sulfur resistance capabilities over 50 h in the presence of a sulfur-laden JP-8 fuel. 5.2.2.2 Rhodium-Based Catalysts. Krause et al.132 showed that Rh and Ni— which are better SR catalysts than Pt—supported on ceria-doped gadolinium

228

Catalysis, 2006, 19, 184–254

Figure 30 H2 yields from ATR of diesel (a) over alumina supported Pd and/or Pt catalysts (b) over ceria supported Pt and/or Ni catalysts (Reprinted from Cheekatamarla and Lane,201 copyright (2003), with permission from authors)

oxides (CGO), yielded a higher H2 production than Pt supported on the same substrate for ATR of i-C8 (Figure 31). This is consistent with the results of Berry et al.,199 who also observed higher yields of H2 from diesel ATR over Ru supported on alumina, which is also considered a good SR catalyst. The other advantage of Rh over Pt as a reforming catalyst is that Rh metal has a lesser tendency to form hydroxyl radicals—which lead to the formation of water—from oxidation of surface hydrogen atoms than Pt.204,205 Consequently, association of surface hydrogen atoms is a more favorable reaction over the Rh surface. The high activity of Rh compared to conventional Ni-based catalysts may also lead to a lower operating temperature of the reformer, eliminating highand low-temperature shift reactors206 and minimizing the O/C. At 5501C (O/C ¼ 1, S/C ¼ 3.0, and GHSV ¼ 179,290 h1), Newson et al. obtained a H2 yield of

Catalysis, 2006, 19, 184–254

229

Figure 31 Yields of H2, CO, CO2, and CH4 (mole/mole of i-C8) from ATR of i-C8 over Rh-, Ni-, or Pt-CGO supported on a cordierite monolith (Conditions: 7001C, O2/C ¼ 0.5, S/C ¼ 1.2, and GHSV ¼ 11,000 h1)132

68% with 84% conversion of i-C8 feed over a proprietary Rh catalyst on mixed oxide supports. A lower conversion of hydrocarbons (73%) was observed at similar conditions when the feed contained B22% of gasoline-range aromatics. Rh on a calcia-impregnated alumina support (0.5 wt% Rh on 15 wt% calcia, balance alumina) was also found to limit carbon formation from ATR of No. 2 fuel oil at a low O/C ratio (and so low temperature) of 0.72.207,208 Therefore, pyrolytic carbon formation, which occurs at high temperatures, and consequently the O/C ratio can be minimized by using the Rh catalyst. Greater than 98% conversion of No. 2 fuel oil was obtained at an O/C ratio of 0.72 and space velocity of 12 lbs fuel/ft3 reactor-h. Weiland et al.209 observed that a small amount of Pt metal present in the Rhbased catalyst could significantly improve the catalyst activity for ATR of gasoline range fuels. They claimed that the role of Pt is to enhance oxidation activity, whereas Rh provides high SR activity. The Rh–Pt/alumina catalyst used in the study was supported on monolithic honeycombs and had a Rh to Pt ratio of 3–10 by weight. The geometry (metal monolith, ceramic monolith, or ceramic foam) of the support did not affect the product composition.210 The differences in reactions at different reactor positions was studied by Springmann et al.67 who reported product compositions for ATR of model compounds as a function of reactor length in a metal monolith coated with a proprietary noble metal containing Rh. As expected, the oxidation reactions take place at the reactor inlet, followed by the SR, shift, and methanation reactions. Figure 32 shows the product concentration profiles for a 1-hexene feed, which are typical results for all the fuels tested. These results show that steam, formed from the oxidation reactions, reaches a maximum shortly after the reactor inlet, after which it is consumed in the shift and reforming reactions. H2, CO and CO2 concentrations increase with reactor length and temperature. In this reactor, shift equilibrium is not reached, and the increase in CO with distance from the inlet is the net result of the shift and SR reactions. Methane is

230

Catalysis, 2006, 19, 184–254

Figure 32 Product profiles for 1-hexene ATR as a function of reactor length, S/C ¼ 2.3, P ¼ 3 atm; T ¼ 600–6501C, (nair/nfuel)actual/(nair/nfuel)stoichiometric ¼ l ¼ 0.32 (Reprinted from Springmann et al.,67 copyright (2002), with permission from Elsevier)

also formed (not shown), either due to hydrogenolysis or thermal decomposition of the 1-hexene. Recently, the use of Rh supported on washcoated alumina monoliths has attracted interest for ATR of higher hydrocarbons.177,211,212 Reyes et al.177 carried out ATR of n-C6 in monolithic catalysts containing Rh as an active component. A maximum H2 yield of 170% was obtained from the reforming of n-C6 at an O/C ratio of 1, a S/C of 1, preheat temperature of 7001C, and GHSV of 68,000 h1. Brandmair et al.211 also carried out ATR of n-C6 over Rh supported on ceramic monoliths at similar conditions, and reported that the Rh catalyst provided better performance over time.

Catalysis, 2006, 19, 184–254

231

5.3 Exhaust Gas Reforming. – The process of reforming of automotive fuels by direct contact with engine exhaust gases is known as exhaust gas reforming. The fuel gas generated from the exhaust gas reforming can then be fed to the combustion engine, along with conventional liquid fuel. The combustion of H2rich gas can reduce the number of hot spots in the engine, which are major sources of NOx formation.213 Therefore, a combustion engine running with such a fuel gas can increase the overall energy efficiency and decrease emissions, compared to standard combustion engines. The probable sequence of reactions in exhaust gas reforming is illustrated below using i-C8 as a model compound for gasoline. (1) Combustion of fuel in engine: C8H18 þ 12.5b(O2 þ 3.76N2) - 8CO2 þ 9H2O þ 12.5(b1)O2 þ 47bN2 (25) Where b is the excess air ratio used for the combustion. For gasoline and diesel engines, b41. (2) Reforming of fuel using exhaust from step (1): C8H18 þ x[8CO2 þ 9H2O þ 12.5(b–1)O2 þ 47bN2] - 8(1 þ x)CO2 þ 9(1 þ x)H2 þ 47bxN2

(26)

where x is the fraction of exhaust gas required to reform 1 mole of the fuel. Ideally, x can be chosen in such a way that the inlet O2 is fully consumed in the reaction. Therefore, the value of x can be calculated from the following equation (for b 4 1.3), 12.5bx 8(x þ 1) ¼ 0.

(27)

(3) Combustion of reformed fuel obtained in step (2): 8(1 þ x)CO2 þ 9(1 þ x)H2 þ 47bxN2 þ 12.5b(O2 þ 3.76N2) - (1 þ x)(8CO2 þ 9H2O þ 12.5(b 1)O2 þ 47bN2)

(28)

Reaction (26) shows that exhaust gas reforming is a combination of SR and POX and, thus, falls in a subcategory of ATR. The extent of SR and POX reactions during exhaust gas reforming is dependent on the exhaust gas temperature and composition (i.e., steam and O2).214 The exhaust gas compositions depend on the type of engine—diesel or gasoline—and the engine operating conditions. Table 8 shows the properties of an exhaust gas stream from a diesel engine at four operating conditions. One of the initial attempts of exhaust gas reforming, as well as onboard H2 generation, was reported by Newkirk and Abel.215 Their process of hightemperature, non-catalytic SR of gasoline resulted in carbon formation in the reformer; however, their objective, to reduce emissions by feeding H2 to the gasoline engine, was achieved. Recently, the exhaust gas reforming process has been investigated extensively at the University of Birmingham, UK. Initial studies of exhaust gas reforming were conducted to power gasoline engines,216,217 however, the catalyst was not identified; presumably it was a Ni-based SR catalyst, since very low space

232

Catalysis, 2006, 19, 184–254

Table 8

Exhaust Gas Temperature and Composition (%vol., dry basis) at Different Diesel Engine Operating Conditions213

Engine speed (rpm) IMEPa (bar) EGRb þ H2 (%) Exhaust temperature (1C) CO (%) H2 (%) CO2 (%) O2 (%) HC (ppm) NOx (ppm) Excess air used, l a

Condition 1

Condition 2

Condition 3

Condition 4

1500 4.6 0 250 0.01 0 3.8 13.9 18 460 3.0

1500 6.1 0 360 0.01 0 5.7 12.2 10 670 2.2

2000 6.6 0 430 0.01 0 6.1 12.3 16 650 2.0

1500 6.1 15 374 0.01 0 5.3 11.8 14 435 1.9

Indicated Mean Effective Difference (IMEP);

b

Exhaust Gas Recirculation (EGR).

velocities (B1,000 h1) were used. Low conversions of gasoline, along with low yields of H2 and CO were observed. Tsolakis et al.213,214,218,219 studied catalytic exhaust gas reforming of diesel fuel for compression ignition engines using low loadings of precious metal supported on a metal oxide catalyst in either a packed bed or coated on a 900 cpsi ceramic monolith. Approximately 20% H2 content was obtained from exhaust gas reforming of an ultra-low-sulfur diesel fuel at an engine exhaust temperature of 4601C and a GHSV of 90,000 h1.214 At an exhaust temperature of 2901C over the same catalyst, adding water to the reforming reactor produced up to 15% more H2 content compared to operation without water.213,218 The addition of water, however, can result in an overall efficiency reduction of the engine-reactor system.218

6

Pyrolysis/Cracking

The decomposition of hydrocarbons can be used to produce H2 by adding heat at appropriate temperatures to dissociate the fuels.66 The cracking reaction of hydrocarbon fuels may be represented by reaction (7). This reaction is endothermic, but can occur at relatively low temperatures in the presence of an appropriate catalyst.66 Several hydrocarbon byproducts are also expected from the decomposition, ranging from methane to aromatics, depending upon the temperature and catalyst. Resulting carbon formation on the catalyst makes this process rather complicated for use with onboard fuel cell applications. However, the catalyst can be regenerated by passing air through the reactor at 46001C to burn off the carbon. The overall process is simplified, since no CO clean up or steam generation is required. In an early investigation on fuel cell applications, Callahan studied catalyzed pyrolysis of various hydrocarbon fuels (combat gasoline, Sunoco 190 gasoline, JP-4 jet fuel, and diesel fuel) for onboard H2 generation.220 Ni supported on

Catalysis, 2006, 19, 184–254

233

alumina or zirconia showed good performance, but degraded severely over time at the elevated temperatures (920 to 1,2001C) necessary for cracking reactions. However, a mixed bed consisting of pure Ni and alumina rings showed both performance and stability at the severe reaction conditions that were employed. With this mixed catalyst bed, hydrocarbon conversions greater than 95% were obtained, along with product streams consisting of more than 90% of H2. The data indicated that two reactors with internal volumes of 675 cm3 each could produce about 0.34 m3/h of H2 from 0.23 kg of a typical hydrocarbon fuel, such as gasoline. According to Callahan, this would be sufficient to run a 500-W acid electrolyte fuel cell stack. Otsuka et al.221 used a Ni/fumed silica catalyst at a relatively low temperature (5001C) for catalytic thermal decomposition of gasoline range hydrocarbons to produce H2. They observed higher H2 selectivities (490%) from different gasoline range alkanes (3-methylpentane, n-C6, cyclohexane, and nC8). However, the supported Ni catalyst deactivated rapidly during the cracking reaction. The number of carbon atoms deposited per Ni atom after complete catalyst deactivation was in the range of 550–1,050, depending on the starting alkanes. They also claimed to recover the catalyst activity fully by CO2 treatment of the deactivated catalyst at 7001C. Recently, Takenaka et al.222 studied a series of base metal catalysts supported on various ceramic oxides for catalytic cracking of kerosene fuel. Yields of H2 and methane from a model kerosene fuel (52 wt% n-C12, 27 wt% diethylbenzene and 21 wt% t-butylcyclohexane) over various base metals at 6001C are shown in Figure 33. Ni/TiO2 showed the highest catalytic activity for the cracking reaction of kerosene fuel, and also maintained a better performance for the kerosene feed that contained benzothiophene. However, the catalytic performance of the

Figure 33 Yields of H2 and CH4 from decomposition of a model kerosene fuel over various catalysts at 6001C (Reprinted with permission from Takenaka et al.,222 copyright (2004) American Chemical Society)

234

Catalysis, 2006, 19, 184–254

Ni/TiO2 catalyst was significantly affected by the type of hydrocarbon used. In particular, aromatic compounds caused a significant deterioration in the catalyst activity, but the addition of Zn metal into Ni/TiO2 catalysts improved the catalytic performance for aromatics cracking reactions. Carbon-based catalysts can provide advantages over metal catalysts for pyrolysis of liquid hydrocarbons, since there is no need for carbon separation from the catalyst surface. Muradov223 used an activated carbon (coconut) catalyst to produce H2 from the pyrolysis of diesel and gasoline. Initially (15– 20 mins), a higher H2 production rate was observed from pyrolysis of diesel and gasoline at 7501C. Afterwards, the H2 production rate dropped over a period of 1 h and reached the steady-state rate (see Figure 34). On the other hand, CH4 and C21 production rates increased before reaching the steady state. It was suggested that the carbon buildup on the surface of the original catalyst can be continuously removed from the reactor, for example, using a fluidized bed reactor.

Figure 34 Catalytic pyrolysis of gasoline (a) and diesel fuel (b) over activated carbon (coconut) at 7501C (Reprinted with permission from Muradov,223 copyright (1998) American Chemical Society)

Catalysis, 2006, 19, 184–254

7

235

Plasma-Assisted Reforming

Plasmas can be classified as either thermal or non-thermal.224 Thermal plasma is a highly energetic state of matter, characterized by thermal equilibrium between the three components of the plasma: electrons, ions, and neutrals. However, it requires high-energy input to achieve high temperatures. Researchers at MIT used a non-catalytic thermal plasma technology to produce H2 from liquid hydrocarbons.175,225–228 Non-catalytic processes are beyond the scope of this work, and will not be discussed. In non-thermal plasmas, also known as non-equilibrium plasmas, there is a significant difference in temperature between electrons and ions/neutrals. Nonthermal plasmas can initiate a chemical reaction even at relatively low temperatures by generating free radicals (i.e., H, O, .OH, CH3, etc), which propagate the reaction. 7.1 Non-Thermal Plasma. – Non-thermal plasmas combined with a catalyst have been studied by Biniwale et al.229 for the reforming of i-C8. The bimetallic catalysts, Ni–Mn, Ni–W, and Rh–Ce supported on alumina mesh and wrapped over a quartz tube were used in a plasma-catalytic reactor. A tungsten wire coil was used to provide heat to the catalyst mesh. The catalyst mesh and a point on the tungsten wire served as electrodes for plasma discharge generation, whereas a heating coil made of tungsten served as a ground electrode. At 6001C over a Ni–Mn catalyst, the H2 production rate increased by 20–120%, depending on the injection frequency of the spray-pulsed injection of the fuel. However, the effects of the plasma on carbon formation or hydrocarbon conversion were not discussed in the article. Sobacchi et al.230 investigated a non-thermal plasma system for pretreatment and post-treatment of products for a catalytic reforming reactor to enhance H2 production. Pretreatment of the feed boosted the concentration of reactive species to the reactor, thereby enhancing the kinetics of H2 production. Plasma post-treatment was carried out to further reform any hydrocarbon slip through the reactor. They employed ANL’s reforming catalyst, presumably Pt-based, in the catalytic reactor at S/C of 0.8 and O/C of 0.7. Figure 35 shows the H2 yields at different reactor temperatures for various experimental configurations. Effects of plasma processing (pre or post) were significant at low temperatures. The higher yields of H2 with the pretreatment configuration suggest that the catalyst performs more effectively on the intermediate oxidized species generated by the plasma pretreatment, rather than on the untreated feed.

8

Supercritical Reforming

A supercritical water POX technique can also be used to produce H2 for fuel cells. The solubility of supercritical water is more like high-pressure steam than water. Therefore, supercritical water can extract hydrocarbons or sulfur species of low volatility from catalyst pores in situ during a heterogeneous catalytic

236

Catalysis, 2006, 19, 184–254

Figure 35 H2 yields (moles of H2/mole of i-C8) at different reactor temperatures for various experimental configurations used in non-thermal plasma reforming of i-C8. Maximum H2 yield from i-C8 is nine as shown by dashed line (Reprinted from Sobacchi et al.,230 copyright (2002), with permission from Elsevier)

reaction. This in situ extraction may result in increased pore accessibility, enhanced catalyst stability to coking and sulfur poisoning, and better product selectivity.231,232 However, the technology is still at the academic level and further exploration of H2 production for fuel cell applications is needed. Recently, the supercritical reforming technique has received some interest for H2 production from liquid fuels.233–235 However, an optimal catalyst for supercritical reforming has yet to be established. Pinkwart et al.234 used four commercial SR catalysts with varying Ni content to reform diesel fuel in supercritical water. Highest yields of H2 were obtained from n-C10 reforming with higher Ni-content catalysts at 5501C, 250 atm, a S/C of 9.6, and a 40 s residence time. Interestingly, Taylor et al.233 evaluated a tubular reactor wall as the catalyst (constructed from Inconel 625 a Ni/Cr alloy) for reforming methanol and diesel at supercritical conditions. Methanol reforming with the Inconel reactor yielded near-equilibrium compositions; however, diesel reforming led to large amounts of black residue in the reactor. It was claimed that the black residue resulted from incomplete mixing of the diesel and water. Watanabe et al.235 investigated reforming of n-C16 in supercritical water (4001C and 400 atm) with basic catalysts ZrO2 and NaOH. As shown in Figure 36, yields of H2 were four times better with NaOH, and 1.5 times better with ZrO2, compared to reaction without a catalyst. The basic catalysts enhance the decomposition of intermediate species (i.e., aldehydes and ketones) to CO, as well as catalyze the WGS reaction. High CO production, coupled with enhanced WGS reaction using the base catalysts, resulted in high H2 yields.

Catalysis, 2006, 19, 184–254

237

Figure 36 H2 yield from reforming of n-C16 in supercritical water with and without catalyst (Reprinted from Watanabe et al.,235 copyright (2003), with permission from Elsevier)

9

Prereforming

Because aromatics are more prone to coking during reforming, their presence can lower the yields of syngas over time due to catalyst deactivation. Furthermore, aromatics may hinder the overall reforming rate by occupying catalytic active sites for a longer time, due to a p-complexation between the d-electrons of the metal and p-electrons of the aromatics. Complex sulfur compounds, particularly in diesel and jet fuels, may poison the catalyst and fuel cell electrodes. However, conversion of such complex hydrocarbons into a light gas mixture before the reformer can minimize the effects of these compounds. Typically, the prereforming process is performed in an adiabatic fixed-bed reactor upstream of the main reformer.145–146,236 In the pre-reformer, higher hydrocarbons are converted into a mixture of CO, CO2, H2, CH4, and light hydrocarbons via a series of reactions: Steam reforming: CnHm þ H2O - CO þ H2 þ CH4 þ Cn 0 Hm 0 , DH298 4 0

(29)

Methanation reaction: CO þ 3H2 2 CH4 þ H2O, DH298 o 0

(30)

Water Gas Shift reaction: CO þ H2O 2 CO2 þ H2, DH298 o 0

(31)

Irreversible and endothermic SR (reaction (29)) is followed by two equilibrium driven exothermic reactions, methanation (reaction (30)) and WGS (reaction (31)). The prereforming is carried out at a relatively low temperature range, 400 to 5501C,237 with the overall reaction being close to autothermal. Prereforming provides the following advantages:145,146,236–238 Traces of sulfur can be eliminated prior to entering the main reformer, due to the lower operating temperatures of the prereformer, which favors the deposition of sulfur on the Ni-based pre-reforming catalyst.

238

Catalysis, 2006, 19, 184–254

The lifetime of the reformer, as well as downstream shift catalysts, is extended due to the complete sulfur removal. The probability of carbon formation on the reformer catalyst is significantly decreased since no higher hydrocarbons are present in the feed. The main reformer can be operated at a lower S/C ratio without any catalyst deactivation, thereby reducing its size and heat duty. The prereformer allows for greater feed temperatures to the main reformer, thereby reducing the size of the latter. Typically, a Ni-based SR catalyst is used in the prereforming process.238 The catalyst must be highly active, since the reaction occurs at low temperatures. Alkali promoters, such as those used in a SR catalyst, are also used to reduce carbon formation. A wide variety of fuels such as diesel,141,142,236 jet fuel,236,239 and NATO F-7636,238 have been reformed in an adiabatic prereformer. Christensen236 used a high surface area Ni catalyst supported on magnesium oxide for adiabatic prereforming of diesel and jet fuels. The prereforming was conducted at 4801C, 25 atm, and a S/C ratio of 2.4 for an extended period (diesel for 1,020 h and jet fuel 495 h). Catalyst deactivation during diesel fuel reforming was significantly higher than that of jet fuel reforming, due to both carbon formation and sulfur poisoning, whereas jet fuel reforming suffered only from sulfur poisoning. Pfefferle239 carried out prereforming of JP-4 fuel, feeding a mixture of fuel, H2, and steam with Pt and Rh supported on alumina containing 6% silica at a temperature below 7001C. The presence of H2 in the feed allows prereforming to occur at low temperatures and, subsequently, a S/C ratio as low as 1 can be used without substantial catalyst deactivation. More than 90% hydrocarbon conversion to methane, H2, CO, and CO2 was observed in prereformer without any catalyst deactivation. Minet et al.141,142 used a combination of calcium aluminate that contained a high loading of calcium (a silica and Ni-free SR catalyst) at the front-end, and a Ni-based CaO–alumina catalyst at the back-end of the reactor for prereforming of No. 2 fuel oil. High S/C ratios (3.9–6.7), as well as high temperatures (940– 1,0001C) were utilized in this study. The effluent from the prereformer was fed to an autothermal reformer containing a Ni-based CaO-alumina catalyst. Up to 60% of hydrocarbons were converted into CO and CO2 in the prereformer without any catalyst deactivation due to carbon formation. Pt-based catalysts have also been explored. For instance, Trimm et al.240 carried out ATR of a surrogate gasoline mixture in a two-bed reactor system consisting of a Pt supported on ceria oxidation catalyst, followed by a Ni-based SR commercial catalyst. The gasoline was prereformed over the Pt/ceria in order to reduce the deactivation of the SR catalyst. Prereforming was initiated at room temperature with an O/C ratio of 1.2, however, the bed temperature reached 5801C during the reaction. Subsequent SR of the effluent over a commercial SR catalyst gave 70% conversion of gasoline and a H2 selectivity of about 70% at a S/C ratio between 2 and 3.4. They reported no changes over several cycles of the reaction.

Catalysis, 2006, 19, 184–254

239

Catalytic cracking as a prereforming step was utilized by Campbell et al.241 for JP-8 fuel. Two different types of catalysts were studied in a packed reactor: Mn supported on g-Al2O3, and mixed MFI and BEA acidic zeolites. Both catalysts gave similar conversion levels over a range of temperatures and space velocities. The primary compounds formed from the JP-8 fuel catalyst cracking were determined to be H2, CH4, C2H4, C3H6, and benzene. Each type of catalyst gave greater than 80% cracking conversion at liquid hour space velocity (LHSV) of 5.5 h1 and o2501C, while a temperature greater than 5201C was required for LHSV of 44 h1. The light sulfur species present in the cracked gas product stream can be removed by adsorption. Non-volatiles and higher aromatics—which are difficult to convert into lighter hydrocarbons in the catalytic cracking process—can be removed by gas/liquid separation. Thus, the light product gas stream can be steam reformed into H2-rich gas without any catalyst deactivation due to sulfur or coke precursors. More recently, cool flame POX has gained momentum for atomization and vaporization in reforming liquid fuels.94,242–244 At temperatures as low as 1201C, fuel-air mixtures can react chemically and produce very weak flames called cool flames. Unlike conventional flames, cool flames generate very little heat, CO, formaldehyde, and other oxygenated compounds.245 This technique is limited to catalytic POX and consists of two isolated and thermally independent reaction chambers. In the first reactor, processing of the complex fuels via a cool flame leads to reaction products comprised of small molecules, such as lower molecular weight alkenes and oxygenated compounds, including aldehydes, ketones, and alcohols. These small molecules are sent to the second reactor for reforming to produce syngas. Oxygenated compounds formed in the first reactor are more easily reformed in the second reactor, and are thought to form less carbon in the reformer than the corresponding hydrocarbons (e.g., ethanol would form less carbon in the reformer than ethane).

10

Kinetics

A fundamental understanding of the reaction kinetics is essential in process development, scale-up, and design. Determination of a detailed reaction model for liquid hydrocarbon reforming is very complicated because liquid hydrocarbons are complex mixtures of hundreds of components, and each one undergoes several different reactions. For example, in addition to the reactions associated with reforming, other reactions may take place: WGS, carbon formation, methanation, hydrocracking, dehydrocyclization, dehydrogenation, ring opening, hydrogenation, etc. Also, the activity of the catalyst changes rapidly during the reaction. For such complex reactions, the experimental rate data are fit into power law or even first-order rate expressions for simplification. Unfortunately, these tend to be limited to a specific catalyst, fuel composition, and operating conditions. It would be desirable to develop predictive models to account for variations in these parameters, but meager information is available on the kinetics of liquid

240

Catalysis, 2006, 19, 184–254

hydrocarbon reforming in order to produce H2. Literature is limited mostly to kinetic studies of SR of single paraffinic components. For SR of higher hydrocarbons, Rostrup-Nielsen111,149 and Tottrup246 postulated a Langmuir-Hinshelwood-Houghen-Watson (LHHW) kinetic model. It was assumed that the hydrocarbon chemisorbs on a dual catalytic site, followed by successive a-scission of the C–C bond. The resulting C1 species react with adsorbed steam to form H2 and CO. The expressions were fit to data for SR of n-C7 on a Ni/MgO catalyst at 5001C; the overall rate expression is:111,246 pC7 H16 240 exp 8150 T r¼h ð32Þ i2 pH2 pH2 O 1 þ 25:2pC7 H16 pH þ 0:08 pH2 2O This rate expression was reduced to a power law-type model, r ¼ k 0 eð

Þ p0:2

8150 T

0:2 0:4 C7 H16 pH2 O pH2 2

ð33Þ

Where r is the rate of reaction in kmol/m (Ni)/h and the partial pressures, p, are in MPa. The reaction order with respect to steam was very sensitive to temperature, decreasing from 0.6 at 4501C, to 0.2 at 5501C. The steamsurface interactions weaken with increasing temperature, hence the reaction order with respect to steam approaches zero due to weak adsorption at high temperatures. Recently, Praharso et al.247 also developed a Langmuir-Hinshelwood type of kinetic model for the SR kinetics of i-C8 over a Ni-based catalyst. In their model, it was assumed that both the hydrocarbon and steam dissociatively chemisorb on two different dual sites on the catalyst surface. The bimolecular surface reaction between dissociated adsorbed species was proposed as the ratedetermining step. The following generalized rate expression was proposed: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi krxn piso pH2 O pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi r¼ ð34Þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð1 þ Kiso piso Þð1 þ KH2 O pH2 O Þ where, Kiso and KH2O are the adsorption constants for i-C8 and steam, respectively. A regression of the rate data over a temperature range of 310– 3501C produced the following power law rate expression: r ¼ k 0 eð

Þ p0:2 p0:5

5300 T

iso H2 O

ð35Þ

A very low reaction order, with respect to i-C8, suggests a strong adsorption of i-C8 molecules on the Ni surface. Rostrup-Nielsen149 also noticed a retarding effect of aromatics and higher hydrocarbons on the reaction rate. Due to the presence of p electrons, aromatics can strongly adsorb onto the catalyst surface and cause the reaction order to approach zero. Recently, Pacheco et al.178 developed and validated a pseudo-homogeneous mathematical model for ATR of i-C8 and the subsequent WGS reaction, based on the reaction kinetics and intraparticle mass transfer resistance. They regressed kinetic expressions from the literature for POX and SR to determine the kinetic parameters from their i-C8 ATR experimental data using Pt on ceria.

241

Catalysis, 2006, 19, 184–254

Table 9 shows the rate expressions used by Pacheco et al. in the reformer modeling. The parameters of these rate expressions are summarized in Table 10. Berry et al.80 also developed a simple power law type rate expression for diesel ATR as shown: EA

a b CH Cc rHC ¼ k0 e RT CHC 2 O o2

ð36Þ

The kinetic parameters for this diesel ATR rate expression are presented in Table 11; they are highly catalyst dependent. The ATR rate shows a nonmonotonic dependence with respect to O2 varying from positive to negative order. The low reaction orders with respect to O2 suggest that the rate of diesel ATR is almost independent of O2. Water strongly inhibits the reaction rate, as shown by its negative reaction order. Minet et al.141,142 studied the ATR kinetics of a sulfur-containing petroleum middle distillate No. 2 fuel oil over a calcium aluminate catalyst system. They developed a single lumped kinetic model, based on bulk hydrocarbons in the reactor feed. It was assumed that the hydrocarbon SR reaction was the ratelimiting step. In the combustion section of the ATR process, it was assumed that Table 9

Rate Equations Used in the Reforming Model of Pacheco et al.178

Reaction

Expression for ri

C8H18 þ 12.5O2 ) 8CO2 þ 9H2O

r1 ¼ k1 PiC8 PO2 r2 ¼ Pk2:52 ð1þK

C8H18 þ 8H2O 3 8CO þ 17H2

CO þ H2O 3 CO2 þ H2

CO CO

2

T: 800–9001C and Catalyst: Ni/Al2O3248;

2

b

H2

H2

iC8 iC8

H2 O H2 O

(b ) (a ) (b ) (b )

H2

T: 500–7501C and Catalyst: Ni/MgAl2O3249,250

Regression of Kinetics Parameters for the Rate Equations of Table 9178

Parameter

Pre-exponential factor 2

k1 (mol/(gcat s bar )) k2 (mol bar0.5/gcat s)) k3 (mol/gcat s bar2)) k4 (mol bar0.5 /(gcat s)) k5 (mol/(gcat s bar)) KH2O (dimensionless) a

2

P2 P2H 2 r3 ¼ k3 PiC8 PCO2 1 K3 PCO iC8 PCO2 2 PiC8 PH O P4H PCO2 =K4 2 2 r4 ¼ Pk3:54 ð1þKCO PCO þKH2 PH2 þKiC8 PiC8 þKH2 O PH2 O =PH2 Þ2 H2 PCO PH2 O PH2 PCO2 =K5 r5 ¼ PkH5 2 ð1þK P þK P þK P þK P =P Þ

C8H18 þ 16H2O ) 8CO2 þ 25H2

Table 10

PiC8 PH2 O P3H PCO =K1

CO PCO þKH2 PH2 þKiC8 PiC8 þKH2 O PH2 O =PH2 Þ

H2

C8H18 þ 8CO2 3 16CO þ 9H2

a

(a )

2.58 2.61 2.78 1.52 15.5 1.57

Heat of adsorption of water (DHH2O).

8

10 109 105 107

104

Activation energy (kJ/mol) 166.0 240.1 23.7 243.9 67.1 88.7a

242

Table 11

Catalysis, 2006, 19, 184–254

Kinetic Parameters for Diesel ATR from Three Different Catalysts

Catalyst

Pre-exponential factora

Activation energyb

a

b

c

Pt/g-Al2O3 Pd/g-Al2O3 Ru/g-Al2O3

18.4 44.7 0.02

70 69 41

0.87 0.74 0.50

1.02 1.07 1.39

0.13 0.10 0.03

a

Units: (L/mol)a1b1c1/s;

b

kJ/mol.

evolved H2 was also oxidizing to produce heat. Thus, the expression developed was zeroth order in O2, first order in hydrocarbons, and second order in steam. The activation energy was determined to be 82 kJ/mol. Later, Edelman et al.251 developed a more fundamental model for No. 2 fuel oil ATR kinetics. This model assumes homogeneous pyrolysis of unstable hydrocarbons resulting in propane, ethane, acetylene, ethylene, and methane fragments. These more stable hydrocarbon fragments then undergo both catalytic POX and catalytic SR, based on pseudo-homogeneous assumptions and elementary behavior. Collectively, these studies indicate that although the kinetics of reforming of pure paraffins may be satisfactorily represented using LHHW types of models, such models are difficult to apply to real fuels, such as diesel or gasoline. For such complex systems, it may be more practical to use pseudo-homogeneous kinetics. For example, hydrocarbon fuel components can be lumped in groups with similar properties and kinetic behaviors, e.g., paraffins, naphthenes, and aromatics. Consequently, the corresponding elementary reactions are also grouped into lumped reactions. However, the levels of simplification must be carefully evaluated to make them consistent with the final aim of the kinetic model. 10.1 Reactivity of Hydrocarbons. – Each homologous series in a liquid fuel can exhibit different kinetics upon reforming under similar reaction conditions. For example, aromatic compounds are the most difficult to reform and require higher temperatures and lower space velocities. Aromatics also contribute significantly to carbon formation, compared to paraffins and naphthenes. At the same reaction conditions, the H2 production rates are typically in the order: aromatics { naphthenesoparaffins.80,81 The relative reactivities of various higher hydrocarbons are summarized in Table 12. Differences in the relative reactivities of individual components in a fuel mixture as well as the specific reforming reaction must be considered in studying important roles in the reforming of diesel or similar complex fuels. Shekhawat et al.81 studied a binary fuel intended to simulate diesel (i.e., n-C14 and 1-methylnaphthalene), to understand the combinatorial effects of feed components. They observed that the overall yields from this binary mixture were not simply additive of those from individual fuel components. The relative reactivity of one fuel component considerably affects the conversion pattern of others, as well as the overall product distribution. The greater the difference in reactivity of binary components, the greater the effect on reforming. For example, aromatics are less reactive than paraffins, hence highly reactive

243

Catalysis, 2006, 19, 184–254

Table 12

Relative Reactivities of Hydrocarbons

Reference Bridger

150

Reforming type Catalyst SR

Ni/MgO/ 5001C, 30 Al2O3 atm

Kopasz et al.78 ATR

ANL catalyst

Berry et al.80

Pt/gAl2O3

ATR

Krummenacher POX et al.74 Moon et al.137

ATR

Reaction conditions

6001C, GHSV ¼ 15,000 h1 8501C, GHSV ¼ 50,000 h1

Relative reactivities Cyclohexane 4 trimethylbutane 4 n-C4, n-C10 4 n-C7 4 C2 4 benzene 4 CH4 i-C8 4 n-C8 4 methylcyclohexane 4 toluene 4 trimethylbenzene n-C14 4 decalin c 1-methyl naphthalene

n-C16 4 n-C10 4 n-C6 Rhcoated monolith Ni/ 7001C, SV i-C8 4 toluene MgAl2O4 ¼ 8,770 h1

paraffins would consume the available O2 in POX and ATR reactions. Therefore, the conversion of highly reactive fuel components proceeds toward completion and produces combustion products. Oxygen is not available for the less reactive component; consequently, the less reactive components predominantly undergo pyrolysis reactions. Side reactions specific to one component play an important role in the reforming of a mixture. For example, aromatics are more prone to coking upon reforming, so their presence in a mixture can lower syngas yields over time due to catalyst deactivation. Also, the catalyst surface-component interactions may play an important role in the reforming of a mixture. For example, aromatics have an abundance of p-electrons, so they may occupy active sites for a longer duration, due to p-complexation between d-orbitals of the metal and p-electrons. Hence there will not be enough reactive sites available for the desired reaction to occur. 11

Concluding Remarks

There is still a significant amount of research and development work left to be done before an effective and reliable H2 generator from liquid fuels can be finalized for the fuel cell market. However, several private sector companies, government laboratories, and universities are involved in extensive research and development activities to overcome various problems associated with fuel reforming catalysts in order to eventually them cost-effective. Based on this literature review, the following issues regarding catalyst development for reforming of liquid fuels have been identified. Among three predominant modes of reforming, POX is more prone to coking, but SR is more difficult to operate and requires more equipment. ATR is a good compromise because of anode-recycle potential. A portion of

244

Catalysis, 2006, 19, 184–254

the fuel cell anode exhaust can be recycled to the reformer to provide an onboard source of water. Deactivation is due primarily to two mechanisms: formation of carboncontaining deposits and sulfur poisoning. Carbon deposition may be minimized by the addition of alkali metals, optimization of metal cluster size, and use of oxygen ion-conducting supports. Sulfur poisoning is usually irreversible and there are few reports of catalysts that are tolerant of sulfur levels typical of commercial fuels. Conventional Ni-based catalysts still dominate in SR applications; however, ceria-supported noble metal catalysts have also attracted interest recently. The study of Rh for both POX and ATR has increased since Rh is in general more active for reforming and is less prone to form carbon. H2 and CO selectivities in Rh-based catalysts have been shown to be affected by catalyst geometry. This indicates that feed mixing and mass transfer can play an important role. Solid oxide catalysts such as hexaaluminates and perovskites, in which an active metal catalyst is incorporated into a coke-resistant lattice, are effective for liquid hydrocarbon reforming due to their thermal stability over a broad-range of temperature. However, sulfur tolerance of those materials has yet to be demonstrated. Oxide-ion conducting supports such as ceria, doped ceria (with Sm or Gd), or perovskites are found to be effective for reducing carbon formation on the catalyst during reforming of liquid hydrocarbons. Novel reforming techniques such as supercritical reforming, plasma-assisted reforming, and cool flame POX have also attracted attention for syngas production for fuel cell applications. Fixed bed reactors still predominate for fuel processing. However, fixed beds are susceptible to vibrational and mechanical attrition. Recently, monolithic reactors, either metallic or ceramic, have attracted interest for reforming processes since they offer higher available active surface areas and better thermal conductivity than conventional fixed beds. Low-pressure drop and robustness of the structure are major advantages of monolithic reactors. Detailed knowledge of the reaction mechanisms and pathways of the reforming system can lead to optimization of reaction conditions, and catalyst design. Unfortunately, very meager information is available on the kinetics of liquid hydrocarbon reforming. Literature is limited mostly to kinetic studies of SR of single paraffinic components.

References 1. M. Williams, ‘National Energy Technology Laboratory/Strategic Center for Natural Gas’ Secretary’s WeeklyReport, August 15, 2002. 2. R.J. Farrauto, in ‘Studies in Surface Science and Catalysis (Science and Technology in Catalysis 2002)’, eds. M. Anpo, M. Onaka and H.Yamashita, Elsevier Sci., 2003, vol. 145, pp. 21–29.

Catalysis, 2006, 19, 184–254

245

3. MTU CFC Solutions GmbH (2005), http://www.mtu-friedrichshafen.com/cfc/en/ down/pics/broschuere.pdf, accessed on Feb 13, 2005. 4. M.Y. El-Sharkh, A. Rahman, M.S. Alam, P.C. Byrne, A.A. Sakla and T. Thomas, J. Power Sources, 2004, 138, 199. 5. A. Ferguson and V.I. Ugursal, J. Power Sources, 2004, 137, 30. 6. A.S. Patil, T.G. Dubois, N. Sifer, E. Bostic, K. Gardner, M. Quah and C. Bolton, J. Power Sources., 2004, 136, 220. 7. M.L. Wald, Scientific American, 2004, 290(5), 66. 8. S.R. Ovshinsky, in ‘Materials and Technology for Hydrogen Economy’, ed. G.-A. Nazri, M. Nazri, R. Young and P. Chen, MRS Symposium Proceedings, Material Research Society, Warrendale, PA, 2004, vol. 801, pp. 3–14. 9. L. Barreto, A. Makihira and K. Riahi, Int. J. Hydrogen Energy, 2003, 28, 267. 10. M.G. Schultz, T. Diehl, G.P. Brasseur and W. Zittel, Science, 2003, 302(5645), 624. 11. J.P. Breen and J.R.H. Ross, Catal. Today, 1999, 51, 521. 12. S. Velu, K. Suzuki, M.P. Kapoor, F. Ohashi and T. Osaki, Appl. Catal. A: Gen., 2001, 213, 47. 13. S. Cavallaro, Energy Fuels, 2000, 14, 1195. 14. W. Mitchell, in ‘Fuel Cells for Transportation TOPTEC’, Proc. Society of Automotive Engineers, Arlington VA, 1996. 15. R. Bellows, in ‘International Fuel Cells’, Preprints of Symposia, ACS Division of Fuel Chemistry, South Windsor, CT, 2001, vol. 46(2), pp. 650–651. 16. U.S. Department of Energy (2003), Hydrogen Research, http://www.nrel.gov/lab/ pao/hydrogen.html, accessed Feb. 18, 2003. 17. U.S. Department of Energy (2005), Facts and Figures about Hydrogen, http:// www.hydrogen.energy.gov/facts_figures.html, accessed on Feb. 13, 2005. 18. J.O. Bockris, Int. J. Hydrogen Energy, 1999, 24, 1. 19. S.S. Bharadwaj and L.D. Schmidt, Fuel Proc. Tech., 1995, 42, 109. 20. J. Devitt, J. Xu, C. Lukaniuk, G. Price and M. Staite, in ‘Solid Oxide Fuel Cells VIII’, ed. S.C. Singhal and M. Dokiya, Proc. from the ECS Meeting, Paris, France, 2003, vol. 7, pp. 1276–1285. 21. Z. Zhan, B.D. Madsen, J. Liu and S.A. Barnett, in ‘Solid Oxide Fuel Cells VIII’, ed. S.C. Singhal and M. Dokiya, Proc. from the ECS Meeting, Paris, France, 2003, vol. 7, pp. 1286–1294. 22. X. Wang and R.J. Gorte, Catal. Lett., 2001, 73, 15. 23. B. Lindstrom, L.J. Pettersson and P.G. Menon, Appl. Catal. A: Gen., 2002, 234, 111. 24. J.P. Breen, R. Burch and H.M. Coleman, Appl. Catal. B: Env., 2002, 39, 65. 25. T. Mizuno and T. Nakajima, J. Chem. Eng. Japan, 2002, 35(5), 485. 26. V.V. Galvita, G.L. Semin, V.D. Belyaev, T.M. Yurieva and V.A. Sobyanin, Appl. Catal. A: Gen., 2001, 216, 85. 27. K. Takeishi and H. Suzuki, Appl. Catal. A: Gen., 2004, 260, 111. 28. C. Pereira, R. Wilkenhoener, S. Ahmed and M. Krumpelt, AIChE Spring Meeting, Atlanta, GA, 2000. 29. J. Houseman, G. Voecks and R. Shah, Electric Power Research Institute Report EM-1126, Project 1041-2, Jul. 1979. 30. M. Flytzani-Stephanopoulos and G.E. Voecks, Int. J. Hydrogen Energy, 1983, 8, 539. 31. C. Pereira, J.-M. Bae, S. Ahmed and M. Krumpelt, U.S. DOE Hydrogen Program Technical Review, San Ramon, CA, 2000. 32. D. Edlund, W. Pledger, T. Snider and C. Renn, Fuel Cell Seminar, Miami, FL, 2003.

246

Catalysis, 2006, 19, 184–254

33. S.W. Scoles and M.A. Perna, Fuel Cell Seminar, Portland, OR, 2000, pp. 538–541. 34. M. Perna, M. Scotto, M. Kantak, J. Budge and B. Ryan, Fuel Cell Seminar, Miami, FL, 2003. 35. T. Suzuki, H. Iwanami and T. Yoshinari, Int. J. Hydrogen Energy, 2000, 25, 119. 36. G. Steinfeld, R. Sanderson, H. Ghezel-Ayagh, S. Abens and M.C. Cervi, AIChE Spring Meeting, Atlanta, GA, 2000. 37. N. Minh, A. Anumakonda, B. Chung, R. Doshi, J. Ferrall, G. Lear, K. Montgomery, E. Ong, L. Schipper and J. Yamanis, Fuel Cell Bull., 1999, 2(6), 9. 38. S. Ahmed, J.P. Kopasz, B.J. Russell and H.L. Tomlinson, Proc. 3rd Inter. Fuel Cell Conf., Nagoya, Japan, 1999. 39. T. Fukunaga, H. Katsuno, H. Matsumoto, O. Takahashi and Y. Akai, Catal. Today, 2003, 84, 197. 40. J.P. Kopasz, R. Wilkenhoener, S. Ahmed, J.D. Carter and M. Krumpelt, U.S. DOE Report ANL/CMT/CP-98970, 1999. 41. L.J. Pettersson and R. Westerholm, Int. J. Hydrogen Energy, 2001, 26, 243. 42. C. Bernay, M. Marchand and M. Cassir, J. Power Sources, 2002, 108, 139. 43. C.-J. Brodrick, T. E. Lipman, M. Farshchi, N. P. Lutsey, H.A. Dwyer, D. Sperling, S.W. Gouse III, D.B. Harris and F.G. King Jr, Transportation Res. Part D., 2002, 7, 303. 44. K.S. Jeong and B.S. Oh, J. Power Sources, 2002, 105, 58. 45. J.M. Ogden, Int. J. Hydrogen Energy, 1999, 24, 709. 46. J.M. Ogden, Physics Today, Apr. 2002. 47. S.S. Penner, A.J. Appleby, B.S. Baker, J.L. Bates, L.B. Buss, W.J. Dollard, P.J. Farris, E.A. Gillis, J.A. Gunsher, A. Khandkar, M. Krumpelt, J.B. O’Sullivan, G. Runte, R.F. Savinell, J.R. Selmen, D.A. Shores and P. Tarman, Prog. Energy Combust. Sci., 1995, 21, 145. 48. W.P. Teagan, J. Bentley and B. Barnett, J. Power Sources, 1998, 71, 80. 49. A. Bauen and D. Hart, J. Power Sources, 2000, 86, 482. 50. A. Bauen, D. Hart and A. Chase, Int. J. Hydrogen Energy, 2003, 28, 695. 51. M.W. Melaina, Int. J. Hydrogen Energy, 2003, 28, 743. 52. A.A. Evers, Int. J. Hydrogen Energy, 2003, 28, 725. 53. B. Eliasson and U. Bossel, Fuel Cell World, Lucerne, Switzerland, Jul. 2002. 54. A.K. Shukla, C.L. Jackson and K. Scott, Bull. Mater. Sci., 2003, 26(2), 207. 55. B.D. McNicol, D.A.J. Rand and K.R. Williams, J. Power Sources, 2001, 100, 47. 56. M. Conte, A. Iacobazzi, M. Ronchetti and R. Vellone, J. Power Sources, 2001, 100, 171. 57. G. Cacciola, V. Antonucci and S. Freni, J. Power Sources, 2001, 100, 67. 58. G.J.K. Acres, J. Power Sources, 2001, 100, 60. 59. R.S. Cherry, Int. J. Hydrogen Energy, 2004, 29, 125. 60. N.A. Darwish, N. Hilal, G. Versteeg and B. Heesink, Fuel, 2004, 83, 409. 61. C. Koroneos, A. Dompros, G. Roumbas and N. Moussiopoulos, Int. J. Hydrogen Energy, 2004, 29, 1443. 62. R.K. Ahluwalia, X. Wang, A. Rousseau and R. Kumar, J. Power Sources, 2004, 130, 192. 63. N.H. Afgan and M.G. Carvalho, Int. J. Hydrogen Energy, 2004, 29, 1327. 64. P. Devlin, U.S. DOE Hydrogen and Fuel Cells Annual Program/Lab R&D Review, Golden, CO, 2002. 65. D.A. Berry, R. James, T.H. Gardner and D. Shekhawat, First Int. Conf. on Fuel Cell Sci. Eng. and Tech., Rochester, NY, Apr. 2003, pp. 425–431. 66. Y. Jamal and M.L. Wyszynski, Int. J. Hydrogen Energy, 1994, 19, 557.

Catalysis, 2006, 19, 184–254

247

67. S. Springmann, G. Friedrich, M. Himmen, M. Sommer and G. Eigenberger, Appl. Catal. A: Gen., 2002, 235, 101. 68. Q. Ming, T. Healey, L. Allen and P. Irving, Catal. Today, 2002, 77, 51. 69. U.S. DOE Report DOE/CE/50343-2, May 1994, Arthur D. Little, Inc. 70. M. Flytzani-Stephanopoulos and G.E. Voecks, U.S. DOE Report DE-AI03-78ET11326, Oct. 1980. 71. S. Ahmed and M. Krumpelt, Int. J. Hydrogen Energy, 2001, 26, 291. 72. M. Flytzani-Stephanopoulos and G.E. Voecks, U.S. DOE Report DOE/ET-113261, 1980. 73. A. Docter and A. Lamm, J. Power Sources, 1999, 84, 194. 74. J.J. Krummenacher, K.N. West and L.D. Schmidt, J. Catal., 2003, 215, 332. 75. A.F. Ghenciu, Cur. Opinion Solid State Mat. Sci., 2002, 6, 389. 76. P. Davis, U.S. DOE Hydrogen and Fuel Cells Annual Program/Lab R&D Review, Golden, CO, 2002. 77. M. Krumpelt, T.R. Krause, J.D. Carter, J.P. Kopasz and S. Ahmed, Catal. Today, 2002, 77, 3. 78. J.P. Kopasz, D. Applegate, L. Ruscic, S. Ahmed and M. Krumpelt, Fuel Cell Seminar, Portland, OR, 2000, pp. 284–287. 79. J.P. Kopasz, D. Applegate, L. Miller and C. Rossignol, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2003, pp. IV-138-IV-141. 80. D.A. Berry, D. Shekhawat and T.H. Gardner, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2003, pp. IV-147IV-152. 81. D. Shekhawat, D.A. Berry and T.H. Gardner, U.S. DOE Office of Fossil Energy, Fuel Cell Program Annual Report, 2004, pp. 141-146. 82. J.P. Kopasz, D. Applegate, L. Miller and S. Ahmed, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2002, pp. 106–110. 83. J.P. Kopasz, L. Miller and D. Applegate, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2004, pp. 481–484. 84. J.P. Kopasz, D. Applegate, L. Miller, S. Ahmed and M. Krumpelt, U.S. DOE Hydrogen and Fuel Cells Annual Program/Lab R&D Review, Golden, CO, 2002. 85. A.M. O’Connor and J.R.H. Ross, Catal. Today, 1998, 46, 203. 86. E. Ruckenstein and Y.H. Hu, Appl. Catal. A: Gen., 1999, 183, 85. 87. M. Prettre, C. Eichner and M. Perrin, Trans. Faraday Soc., 1946, 43, 335. 88. W.J.M. Vermeiren, E. Blomsma and P.A. Jacobs, Catal. Today, 1992, 13, 427. 89. A.P.E. York, T. Xiao and M.L.H. Green, Topics Catal., 2003, 22, 345. 90. K. Aasberg-Petersen, J.-H.B. Hansen, T.S. Christensen, I. Dybkjaer, P.S. Christensen, C.S. Nielsen, S.E.L.W. Madsen and J.R. Rostrup-Nielsen, Appl. Catal. A: Gen., 2001, 221, 379. 91. J.M. Ogden, M.M. Steinbugler and T.G. Kreutz, J. Power Sources, 1999, 79, 143. 92. C. Song, Catal. Today, 2002, 77, 17. 93. J.N. Armor, Appl. Catal. A: Gen., 1999, 176, 159. 94. A. Naidja, C.R. Krishna, T. Butcher and D. Mahajan, Prog. Energy Combust. Sci., 2003, 29, 155. 95. D.L. Trimm and Z.I. Onsan, Catal.Rev., 2001, 43, 31. 96. J.R. Rostrup-Nielsen, J.-H.B. Hansen and L.M. Aparicio, Sekiyu Gakkaishi, 1997, 40(5), 366. 97. L. Carrette, K.A. Friedrich and U. Stimming, CHEMPHYSCHEM, 2000, 1, 162.

248

Catalysis, 2006, 19, 184–254

98. J.R. Rostrup-Nielsen, J. Sehested and J.K. Norskov, in ‘Advances in Catalysis’, ed. B.C. Gates and H. Knozinger, Academic Press, London, 2002, vol. 47, pp. 65–139. 99. J.D. Holladay, Y. Wang and E. Jones, Chem. Rev., 2004, 104, 4767. 100. S. Ahmed, R. Kumar and M. Krumpelt, Fuel Cell Bull., 1999, 2(12), 4. 101. F.R. Kalhammer, P.R. Prokopius, V.P. Roan and G.E. Voecks, ‘Status and Prospects of Fuel Cells as Automobile Engines’, Report prepared for State of California Air Resources Board, Sacramento, CA, 1998. 102. L.F. Brown, Int. J. Hydrogen Energy, 2001, 26, 381. 103. A.K. Avci, Z.I. Onsan and D.L. Trimm, Appl. Catal. A: Gen., 2001, 216, 243. 104. S.L. Douvartzides, F.A. Coutelieris, A.K. Demin and P.E. Tsiakaras, AIChE J., 2003, 49(1), 248. 105. M. Wang, J. Power Sources, 2002, 112, 307. 106. S. Dunn, Int. J. Hydrogen Energy, 2002, 27, 235. 107. G. Erdmann, Int. J. Hydrogen Energy, 2003, 28, 685. 108. C.E. Borroni-Bird, J. Power Sources, 1996, 61, 33. 109. K. Joon, J. Power Sources, 1998, 71, 12. 110. R.J. Farrauto and C.H. Bartholomew (eds.), in ‘Fundamentals of Industrial Catalytic Processes’, Blackie Academic and Professional, 1997, p. 274. 111. J.R. Rostrup-Nielsen, in ‘Catalysis, Science, and Technology’, ed. J.R. Anderson and M. Boudart, Springer-Verlag, Berlin, 1984, vol. 5, pp. 1–118. 112. C.H. Bartholomew, Catal. Rev. Sci. Eng., 1982, 24, 67. 113. C.H. Bartholomew and J.B. Butt (eds.), in ‘Catalysts Deactivation 1991’, Elsevier Science Publishers, B.V., Amsterdam, 1991. 114. C.N. Satterfield (ed.), in ‘Heterogeneous Catalysis in Practice’, McGraw-Hill, New York, 1980, pp. 141–143. 115. J. Hagen and S. Hawkins (eds.), in ‘Industrial Catalysis: A Practical Approach’, Wiley VCH, 1999, p. 186. 116. C. Palm, P. Cremer, R. Peters and D. Stolten, J. Power Sources, 2002, 106, 231. 117. J.R. Rostrup-Nielsen, in ‘Catalyst Deactivation 2001’, ed. J.J. Spivey, G.W. Roberts and B.H. Davis, Elsevier, 2001, vol. 139, pp. 1–12. 118. R. Borup, L. Perry, M. Inbody and J. Tafoya, Prepr. Pap.-Am. Chem. Soc., Div. Fuel Chem., 2002, 47(2), 547. 119. R. Borup, M.A. Inbody, J.I. Tafoya, W.J. Vigil and D. Guidry, Prepr. Pap. -Am. Chem. Soc., Div. Fuel Chem, 2003, 48(2), 740. 120. R.L. Borup, M.A. Inbody, T.A. Semelsberger, J.I. Tafoya and D.R. Guidry, Catal. Today, 2005, 99, 263. 121. R.L. Borup, M. Inbody, T. Semelsberger, L. Perry and J. Parkinson, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2002, pp. 342–347. 122. D.L. Trimm, Catal. Today, 1997, 37, 233. 123. H.C. Dibbern, P. Olesen, J.R. Rostrup-Nielsen, P.B. Tottrup and N.R. Udengaard, Hydrocarbon Processing, Jan. 1986, pp. 71–74. 124. J.R. Rostrup-Nielsen, in ‘Stud. Surf. Sci. Catal’, ed. C.H. Bartholomew and J.B. Butt, Elsevier, 1991, vol. 68. 125. F. Besenbacher, I. Chorkendorff, B.S. Clausen, B. Hammer, A.M. Molenbroek, J.K. Norskov and I. Stensgaard, Science, 1998, 279, 1913. 126. H.S. Bengaard, J.K. Norskov, J. Sehested, B.S. Clausen, L.P. Nielsen, A.M. Molenbroek and J.R. Rostrup-Nielsen, J. Catal., 2002, 209, 365. 127. T.J. Huang, Abstracts of Papers, 228th ACS National Meeting, Philadelphia, PA, Aug. 2004.

Catalysis, 2006, 19, 184–254

249

128. S. Hamakawa, S. Yoshino, J. Nakamura, Y. Liu, A. Tsyganok, K. Suzuki, K. Murata and T. Hayakawa, Electrochem. Solid State Lett., 2001, 4(10), pp. D9-D11. 129. M. Salazar-Villalpando, D.A. Berry, D. Shekhawat, T.H. Gardner and I. Celik, 2nd ASME Fuel Cell Sci. Eng. and Tech. Conf., 2004, pp. 681–690. 130. C. Descorme, Y. Madier, D. Duprez and T. Birchem, in ‘Studies in Surface Science and Catalysis (12th International Congress on Catalysis)’, ed. A. Corma, F.V. Melo, S. Mendioroz and J.L.G. Fierro, Elsevier Science, 2000, vol. 130, p. 347. 131. C. Descorme and D. Duprez, Appl. Catal. A: Gen., 2000, 202, 231. 132. T. Krause, J. Mawdsley, C. Rossignol, J. Kopasz, D. Applegate, M. Ferrandon, J.D. Carter and M. Krumpelt, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2002, pp. 332–336. 133. T. Krause, J. Mawdsley, C. Rossignol, J.D. Carter, J. Kopasz and M. Krumpelt, Fuel Cell Seminar, Palm Springs, CA, 2002. 134. J. Zhang, Y. Wang, R. Ma and D. Wu, Appl. Catal. A: Gen., 2003, 243, 251. 135. W. Yanhui and W. Diyong, Int. J. Hydrogen Energy, 2001, 26, 795. 136. J.R. Rostrup-Nielsen, J. Catal., 1974, 33, 184. 137. D.J. Moon, K. Sreekumar, S.D. Lee, B.G. Lee and H.S. Kim, Appl. Catal. A: Gen., 2001, 215, 1. 138. H.-S. Hwang, R.M. Heck and R.M. Yarrington, Engelhard Corp., US Patent 4522894, 1985. 139. R.M. Heck and P. Flanagan, Engelhard Corp., US Patent 4844837, 1989. 140. A. Anumakonda, J. Yamanis and J. Ferrall, Allied Signal, Inc., US Patent 6221280, 2001. 141. R.G. Minet, D. Warren, T. Tomita, K. Hirokawa, K. Osaki and T. Shinjo, Electric Power Research Institute Report EM-1487, Project 1041-1, Aug. 1980. 142. R.G. Minet, D. Warren and J. Bett, Electric Power Research Institute Report EM2096, Project 1041-1, Oct. 1981. 143. J.R. Rostrup-Nielsen and T. Rostrup-Nielsen, CATTECH, 2002, 6(4), 150. 144. J.R. Rostrup-Nielsen, Phys. Chem. Chem. Phys., 2001, 3, 283. 145. F. Joensen and J.R. Rostrup-Nielsen, J. Power Sources, 2002, 105, 195. 146. J.R. Rostrup-Nielsen, T.S. Christensen and I. Dybkjaer, in ‘Studies in Surface Science and Catalysis (Recent Advances in Basic and Applied Aspects of Industrial Catalysis’, ed. T.S.R.P. Rao and G.M. Dhar, Elsevier Science, 1998, vol. 113, pp. 81–95. 147. D.E. Ridler and M.V. Twigg, in ‘Catalysis Handbook’, ed. M.V. Twigg, Manson Publishing Ltd, London, UK, 1996, 2nd Edn., vol. 1, Ch. 5, pp. 225–282. 148. J.R. Rostrup-Nielsen, Catal. Today, 1993, 18, 305. 149. J.R. Rostrup-Nielsen, J. Catal., 1973, 31, 173. 150. G.W. Bridger, Catalysis, 1980, Vol. 3, pp. 39–69. 151. A.E. Lutz, R.W. Bradshaw, J.O. Keller and D.E. Witmer, Int. J. Hydrogen Energy, 2003, 28, 159. 152. K. Ahmed and K. Foger, in ‘Solid Oxide Fuel Cells VIII’, ed. S.C. Singhal and M. Dokiya, Proc. from the ECS Meeting, Paris, France, 2003, vol. 7, pp. 1240–1249. 153. A. Roine, HSC Chemistry Ver. 4.0., Outokumpu Research Oy, Pori, Finland, 1999. 154. Haldor Topsoe A/S (2005), Haldor Topsoe Hydrogen and Synthesis Catalysts, http://www.haldortopsoe.com/site.nsf/all/BBNN-5PFGYX?OpenDocument; accessed on Feb. 13, 2005. 155. Johnson Matthey (2005), Johnson Matthey Catalysts, Steam Reforming, http:// www.synetix.com/refineries/hydrogen-steamreforming.htm; accessed on Feb. 13, 2005.

250

Catalysis, 2006, 19, 184–254

156. Product Bulletin from Sud-Chemie (2005), ‘‘Catalysts for ammonia, methanol, hydrogen, carbon monoxide, reducing gases’’. 157. Unicat Catalyst Technologies, Inc. (2005), Steam Reforming Catalysts, http:// www.unicatcatalyst.com/SYNGAS.htm; accessed on Feb. 13, 2005. 158. K. Murata, L. Wang, M. Saito, M. Inaba, I. Takahara and N. Mimura, Energy Fuels, 2004, 18, 122. 159. L. Wang, K. Murata and M. Inaba, Ind. Eng. Chem. Res., 2004, 43, 3228. 160. L. Wang, K. Murata and M. Inaba, Appl. Cat. B: Env., 2004, 48, 243. 161. J.K. Lee and D. Park, Korean J. Chem. Eng., 1998, 15(6), 658. 162. T. Ando, N.-O. Ikenaga, K. Nakagawa and T. Suzuki, J. Japan Petro. Inst., 2002, 45(6), 409. 163. A. Segawa, T. Matsumoto, Y. Sato, K. Ishizuki, M. Adachi and Y. Kobori, Fuel Cell Seminar, Miami, FL, 2003. 164. X. Wang and R.J. Gorte, Appl. Catal. A: Gen., 2002, 224, 209. 165. H. Gunardson (ed.), in ‘Industrial Gases in Petrochemical Processing’, Marcel Dekker, Inc., New York, 1998, p. 69. 166. R. Ran, G. Xiong, S. Sheng, W. Yang, N. Stroh and H. Brunner, Catal. Lett., 2003, 88, 55. 167. W. Zhu, G. Xiong, W. Han and W. Yang, Catal. Today, 2004, 93–95, 257. 168. T.H. Gardner, D. Shekhawat and D.A. Berry, AIChE Fall Meeting, Austin, TX, 2004. 169. Y. Fujitani and H. Muraki, Kabushiki Kaisha Toyota Chuo Kenkyusho, Japan, US Patent 4087259, 1978. 170. M. Tanaka, M. Takemura and N. Ito, Nippon Soken, Inc., Japan, US Patent 4088608, 1979. 171. R.P. O’Connor, E.J. Klein and L.D. Schmidt, Catal. Lett., 2000, 70, 99. 172. L.D. Schmidt, E.J. Klein, C.A. Leclerc, J.J. Krummenacher and K.N. West, Chem. Eng. Sci., 2003, 58, 1037. 173. R.P. O’Connor, E.J. Klein, D. Henning and L.D. Schmidt, Appl. Catal. A: Gen., 2003, 238, 29. 174. R. Subramanian, G.J. Panuccio, J.J. Krummenacher, I.C. Lee and L.D. Schmidt, Chem. Eng. Sci., 2004, 59, 5501. 175. A.E. Lutz, R.W. Bradshaw, L. Bromberg and A. Rabinovich, Int. J. Hydrogen Energy., 2004, 29, 809. 176. B.F. Hagh, J. Power Sources, 2004, 130, 85. 177. S.C. Reyes, J.H. Sinfelt and J.S. Feeley, Ind. Eng. Chem. Res., 2003, 42, 1588. 178. M. Pacheco, J. Sira and J. Kopasz, Appl. Catal. A: Gen., 2003, 250, 161. 179. T. Rampe, A. Heinzel and B. Vogel, J. Power Sources, 2000, 86, 536. 180. T.A. Semelsberger, L.F. Brown, R.L. Borup and M.A. Inbody, Int. J. Hydrogen Energy, 2004, 29, 1047. 181. A. Ersoz, H. Olgun, S. Ozdogan, C. Gungor, F. Akgun and M. Tiris, J. Power Sources, 2003, 118, 384. 182. L. Ma, D.L. Trimm and C. Jiang, Appl. Catal. A: Gen., 1996, 138, 275. 183. C.L. Verrill, L.J. Chaney, K.E. Kneidel, R.A. McIlroy and R.M. Privette, McDermott Technology Inc., US Patent 5938800, 1999. 184. J. Zhang, Y. Wang, R. Ma and D. Wu, Korean J. Chem. Eng., 2003, 20(2), 288. 185. M. Flytzani-Stephanopoulos and G.E. Voecks, Energy Progress, 1981, 1(1–4), 52. 186. D.J. Moon and J.W. Ryu, Catal. Lett., 2003, 89, 207. 187. D.J. Moon, J.W. Ryu, S.D. Lee, B.G. Lee and B.S. Ahn, Appl. Catal. A: Gen., 2004, 272, 53.

Catalysis, 2006, 19, 184–254

251

188. D.J. Moon, J.W. Ryu, S.D. Lee and B.S. Ahn, Korean J. Chem. Eng., 2002, 19(6), 921. 189. J. Mawdsley, T. Krause, J. Kopasz, D.-J. Liu, S. Ahmed, J. Ralph, L. Miller and H.-K. Liao, Fuel Cell Seminar, Miami, FL, 2003. 190. J. Mawdsley, M. Ferrandon, C. Rossignol, J. Ralph, L. Miller, J. Kopasz and T. Krause, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2003, pp. IV-83-IV-87. 191. A. Qi, S. Wang, G. Fu, C. Ni and D. Wu, Appl. Catal. A: Gen., 2005, 281, 233. 192. M. Krumpelt, S. Ahmed, R. Kumar and R. Doshi, University of Chicago, US Patent 6110861, 2000. 193. M. Krumpelt, J.D. Carter, R. Wilkenhoener, S.H.D. Lee, J.-M. Bae and S. Ahmed, Fuel Cell Seminar, Portland, OR, 2000, pp. 542–545. 194. T. Krause, D. Carter, J.-M. Bae, J. Kopasz, S. Ahmed and M. Krumpelt, Proc. North American Catalysis Society, Toronto, 2001. 195. M. Krumpelt, R. Wilkenhoener, J.D. Carter, J.-M. Bae, J. Kopasz, T. Krause and S. Ahmed, U.S. DOE Annual Progress Report: Transportation Fuel Cell Power Systems, 2000, pp. 65–70. 196. M. Krumpelt, T. Krause, J. D. Carter, J. Mawdsley, J.-M. Bae, S. Ahmed and C. Rossignol, U.S. DOE Annual Progress Report: Transportation Fuel Cell Power Systems, 2001, pp. 90-95. 197. T. Krause, M. Ferrandon, J. Mawdsley and J. Ralph, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2004, pp. 467–471. 198. C. Pereira, R. Wilkenhoener, S. Ahmed and M. Krumpelt, U.S. DOE Report ANL/ CMT/CP-99684, 1999. 199. D.A. Berry, D. Shekhawat, T.H. Gardner and W. Rogers, U.S. DOE Progress Report for Hydrogen, Fuel Cells, and Infrastructure Technologies Program, 2002, pp. 337–341. 200. D.O. Ham, P.F. Lewis, G.W. Lord, R.M. Yarrington and H.S. Hwang, U.S. DOE Report DE-AC21-79MC-12734, Feb 1982. 201. P.K. Cheekatamarla and A.M. Lane, AIChE Spring Meeting, New Orleans, LA, 2004. 202. P.K. Cheekatamarla and A.M. Lane, Prepr. Pap.-Am. Chem. Soc., Div. Fuel Chem., 2004, 49(1), 139. 203. P.K. Cheekatamarla and A.M. Lane, Int. J. Hydrogen Energy, 2005, 30, 1277. 204. W.R. Williams, C.M. Marks and L.D. Schmidt, J. Phys. Chem., 1992, 96, 5922. 205. M.P.Z. Mallen, W.R. Williams and L.D. Schmidt, J. Phys. Chem., 1993, 97, 625. 206. E. Newson and T.B. Truong, Int. J. Hydrogen Energy, 2003, 28, 1379. 207. W.E. Houghtby, R.F. Buswell, J.A.S. Bett, R.R. Lesieur, A.P. Meyer, J.L. Preston and H.J. Setzer, Electric Power Research Institute Report EM-1701, Project 1041-4, Feb. 1981. 208. H.J. Setzer, R.R. Lesieur, S. Karavolis and W.G. Wnuck, United Technologies Corp., U.S. Patent 4473543, 1984. 209. S. Wieland, F. Baumann and R. Ahlborn, US Patent Application, US20020009408A1, 2002. 210. S. Wieland, F. Baumann and K.-A. Starz, Fuel Cell Seminar, Portland, OR, 2000, pp. 309–312. 211. M. Brandmair, H. Apfel, J. Find, C. Cremers, U. Stimming and J. Lercher, Fuel Cell Seminar, Miami, FL, 2003. 212. L.D. Schmidt, Presentation at National Energy Technology Laboratory, Dec. 4, 2004.

252

Catalysis, 2006, 19, 184–254

213. 214. 215. 216.

A. Tsolakis, A. Megaritis and M.L. Wyszynski, Fuel, 2004, 83, 1837. A. Tsolakis, A. Megaritis and M.L. Wyszynski, Energy Fuels, 2003, 17, 1464. M.S. Newkirk and J.L. Abel, SAE Transactions Paper No. 720670, Vol. 81, 1972. M.R. Jones and M.L. Wyszynski, in ‘Vehicle Thermal Management Systems’, Proc. SAE – IMECHE Conf., Columbus, OH, 1993, SAE Paper No. 931096. Y. Jamal, T. Wagner and M.L. Wyszynski, Int. J. Hydrogen Energy, 1996, 21(6), 507. A. Tsolakis and A. Megaritis, Int. J. Hydrogen Energy, 2004, 29, 1409. A. Tsolakis, A. Megaritis and S.E. Golunski, Energy Fuels, 2005, 19, 744. M.A. Callahan, in ‘Electrocatalysis to Fuel Cells’, ed. G. Sandstede, Univ. Washington Press, Seattle, WA, 1972, pp. 189–204. K. Otsuka, Y. Shigeta and S. Takenaka, Int. J. Hydrogen Energy, 2002, 27, 11. S. Takenaka, Y. Kobayashi and K. Otsuka, Energy Fuels, 2004, 18, 1775. N.Z. Muradov, Energy Fuels, 1998, 12, 41. G.S. Selwyn, H.W. Herrmann, J. Park and I. Henins, Contrib. Plasma Phys., 2001, 6, 610. L. Bromberg, D.R. Cohn, A. Rabinovich and J. Heywood, Diesel Engine Emission Reduction Workshop, San Diego, CA, Aug, 2000. L. Bromberg, D.R. Cohn, A. Rabinovich, N. Alexeev, J.B. Green Jr., N. Domingo, J.M.E. Storey, R.M. Wagner and J.S. Armfield, Government/Industry Meeting, Washington, DC, Jun. 2000. H. Jabs, S. Shandy, D. Westerheim, A. Cisar and Z. Minevski, Fuel Cell Seminar, Miami, FL, 2003. L. Bromberg, A. Rabinovich, N. Alexeev and D.R. Cohn, ACS National Meeting, Anaheim, CA, Mar. 1999. R.B. Biniwale, A. Mizuno and M. Ichikawa, Appl. Catal. A: Gen., 2004, 276, 169. M.G. Sobacchi, A.V. Saveliev, A.A. Fridman, L.A. Kennedy, S. Ahmed and T. Krause, Int. J. Hydrogen Energy, 2002, 27, 635. B. Subramaniam, Appl. Catal. A: Gen., 2001, 212, 199. P.E. Savage, Catal. Today, 2000, 62, 167. J.D. Taylor, C.M. Herdman, B.C. Wu, K. Wally and S.F. Rice, Int. J. Hydrogen Energy, 2003, 28, 1171. K. Pinkwart, T. Bayha, W. Lutter and M. Krausa, J. Power Sources, 2004, 136, 211. M. Watanabe, H. Inomata, M. Osada, T. Sato, T. Adschiri and K. Arai, Fuel, 2003, 82, 545. T.S. Christensen, Appl. Catal. A: Gen., 1996, 138, 285. I. Dybkjaer, Fuel Proc. Tech., 1995, 42, 85. S. Katikaneni, C. Yuh, S. Abens and M. Farooque, Catal. Today, 2002, 77, 99. W.C. Pfefferle, Engelhard Industries, Inc., US Patent 3481722, 1969. D.L. Trimm, A.A. Adesina, Praharso and N.W. Cant, Catal. Today, 2004, 93–95, 17. T.J. Campbell, A.H. Shaaban, F.H. Holcomb, R. Salavani and M.J. Binder, J. Power Sources, 2004, 129, 81. L. Hartmann, K. Lucka and H. Kohne, J. Power Sources, 2003, 118, 286. L. Hartmann, C. Mengel, K. Lucka and H. Kohne, in ‘Soild Oxide Fuel Cells VIII’, ed. S.C. Singhal and M. Dokiya, Proc. from the ECS Meeting, Paris, France, 2003, vol. 7, pp. 1250–1259. J.M. da Silva, I. Hermann, C. Mengel, K. Lucka and H. Kohne, AIChE J., 2004, 50(5), 1042.

217. 218. 219. 220. 221. 222. 223. 224. 225. 226.

227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 237. 238. 239. 240. 241. 242. 243.

244.

Catalysis, 2006, 19, 184–254

253

245. H. Pearlman and R.M. Chapek, NASA, Cool Flames and Autoignition in Microgravity, http://www.grc.nasa.gov/WWW/RT1999/6000/6711wu.html, accessed on May 23, 2005. 246. P.B. Tottrup, Appl. Catal., 1982, 4, 377. 247. Praharso, A.A. Adesina, D.L. Trimm and N.W. Cant, Chem. Eng. J., 2004, 99, 131. 248. W. Jin, X. Gu, S. Li, P. Huang, N. Xu and J. Shi, Chem. Eng. Sci., 2000, 55, 2617. 249. J. Xu and G. Froment, AIChE J., 1989, 35, 88. 250. J. Xu and G. Froment, AIChE J., 1989, 35, 97. 251. R.B. Edelman, R.C. Farmer and T.S. Wang, US DOE Report DE-AC21-79MC 12735, Aug. 1982.

Catalysis: Vol. 19 (Specialist Periodical Reports)

Catalysis: Vol 17 (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. 19 (Specialist Periodical Reports)

Catalysis: Vol 17 (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