Advances in Catalysis, Volume 40

ADVANCES IN CATALYSIS VOLUME 40 Advisory Board G. EWrL AerlinlDuhlrtn, Cermuny W. M. H . SACHTLER L.vun.cton. Illino...

Author: D. D. Eley | Herman Pines | Werner O. Haag

39 downloads 1031 Views 14MB 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

ADVANCES IN CATALYSIS VOLUME 40

Advisory Board G. EWrL AerlinlDuhlrtn, Cermuny

W. M. H . SACHTLER L.vun.cton. Illinois

V. B. KAZANSKY Moscow. Russia

J. M. THOMAS London, U.K .

A . OZAKI Tokyo, Jupiin

P. B. WElSZ Stute C o l l r ~ ePentqlvuniu ,

ADVANCES IN CATALYSIS VOLUME 40

Edited by D. D. ELEY The University Notringhum, England

HERMAN PINES

WERNER0. HAAG

Northwestern University Evanston, Illinois

Mobil Research and Development Corporation Princeton, New Jersey

ACADEMIC PRESS San Diego New York Boston London Sydney Tokyo Toronto

This book is printed o n acid-free paper.

@

Copyright 0 1994 by ACADEMIC PRESS, INC. All Rights Reserved. No pan o f this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher.

Academic Press, Inc.

A Division of Harcourt Brace & Company 525 B Street. Suite 1900, San Diego. California 92101-4495 United Kingdom Edition pirblished by Academic Press Limited 24-28 Oval Road, London NWI 7DX

International Standard Serial Number: 0760-0564 International Standard Book Number: 0- 12-007840-6 PRINTED IN THE UNITED STATES OF AMERICA 9 4 9 s 9 6 9 1 9x 99 Q W Y n 7 6 5

4

3

2

I

Contents CONTRIBUTORS ..................................................................................................... PREFACE............................................................................................................

vii ix

Oxidative Dehydrogenation of Light (C1 to C,) Alkanes HAROLD H . KUNC

I. I1. 111. IV. V. v 1. VII . VIII .

IX . X.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gas-Phase Reaction of Alkanes with Oxygen . . . . . . . . . . . . . . . . . . Studies of Catalytic Oxidative Dehydrogenation in the 1960s . . . . . . . . . . Oxidative Dehydrogenation of Ethane . . . . . . . . . . . . . . . . . . . . . . Oxidative Dehydrogenation of Propane . . . . . . . . . . . . . . . . . . . . . Oxidative Dehydrogenation of Butane and Cyclohexane . . . . . . . . . . . Generalized Reaction Scheme for Oxidative Dehydrogenation of Alkanes A Generic Description of the Relationship between Metal-Oxygen Bond Strength and Selectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . Selectivity Patterns for Mg3(V0&, M@V207. and ( V O ) Z P ~ O.~. . . . . . Conclusion and Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..

1

2 4 5 8 14

24 26 28 34 35

Catalysis in Coal Liquefaction ISAO MOCHIDA AND KINYA SAKANISHI

I.

I1 . Ill . 1V. V. v 1. VI1. VIII .

1x. X.

XI . XI1. XIII . XIV .

xv.

Introduction: History and Status of Coal Liquefaction . . . . . . . . . . . . . Coal Structure and Reactivity . . . . . . ............... Stages of Coal Liquefaction . . . . . . . . . . . . . . . . . . . . . . . . Coal Dissolution, Depolymerization, and Retrogressive Reactions . . . . . . . Catalysts in Liquefaction . . . . . . . . . . . . . . . . . . . . . Roles of Hydrogen Donor, Solvent Properties. and Catalyst in the Preheating and Primary Stages . . . . . . . . . . . . . . . . . . . . . . . . . . Combination of Catalyst and Solvent and Stepwise and Catalyst in the Primary Stage . . . . . . . . . ........... Catalytic Upgrading of Crude Coal Liquids in the Roles of Solvents in the Secondary Stage . . . . . . . . . . . . . . . . . . . . Life. Recovery. and Regeneration of Liquefaction Catalyst in the Primary and Secondary Stages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pretreatment of Coal and Coal Liquid . . . . . . . . . . . . . . . . . . . . . . Ionic Depolymerization of Coal . . . . . . . . . . . . . . . . . ........ Prospects for Catalysts in Coal Liquefaction . . . . . . . . . . . . . . . . . . . . Design of Multistage Coal Liquefaction with Catalysts Yet to Be Developed . . Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...................... .. V

39 41 43 45 46 51

58 62 69 10

15 77 17 19 80 80

vi

CONTENTS

Advances in Applied Electrocatalysis HAKTMUT WENDT.SVENRAUSCH. AND THOMAS BORUCINSKI

I. I1 . 111. 1V.

V. VI.

VII .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Electrocatalytically Activated, Dimensionally Stable Chlorine-Evolving Electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oxygen Evolving Anodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . Electrocatalysis of Cathodic Hydrogen Evolution . . . . . . . . . . . . . . Electrocatalysis of Cathodic Oxygen Reduction and Anodic Hydrogen Oxidation in Fuel Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Electrocatalysis in Electroogdnic Synthesis . . . . . . . . . . . . . . . . . Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

87

..... .....

91 103

.....

Ill

.....

122

. . . . . . . . . .

.....

151

168 168

Fundamental Studies of Transition-Metal Sulfide Catalytic Materials R . R . CHIANLLLI. M . DAAGE. A N D M . J . LEMUX I. 11. 111 .

IV. V. VI . VII . VIII .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Preparation of Bulk Transition-Metal Sulfides . . . . . . . . . . . . . . . . . . . Structure and Properties of TMS Catalytic Materials . . . . . . . . . . . . . . . Surfxe Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . StructureiFunction Relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . Recent Developments in the Description of the Active Promoted Phase . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

171 179 188 192 200 206 22 I 226 229

Multicomponent Bismuth Molybdate Catalyst: A Highly Functionalized Catalyst System for the Selective Oxidation of Olefin YOSHlHlKO

I. I1 . 111.

I v.

INULX

MORO-OKA AND WATARU

UEDA

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Characterization and Working Mechanism of Multicomponent Bismuth Molybdate Catalysts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stability of the Multicomponent Bismuth Molybdate Catalyst Depending on the Bulk Diflusion of Oxide Ion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

233

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

272

231 265 269 270

Contributors Numbers in purentheses indicute the p q e s on which the authors’ contributions begin

THOMASBORUCINSIU, Institut fur Chemische Technologie, Technische Hochschule Darmstadt, 0-61287 Darmstadt, Germany (87) R. R . CHIANELLI, Exxon Research and Engineering Company, Annandale, New Jersey 08801 ( 1 77) M.DAAGE, Exxon Research and Engineering Company, Annandale, New Jersey 08801 (177) HAROLDH . KUNG,Ipatieff Laboratory, Department uf Chemicul Engineering, Northwestern University, Evanston, Illinois 60208 ( 1) M. J . LEDOUX, Universitd Louis Pasteur-EHICS, Laboratoire de Chimie des Matdriaux Catalytiques, 67070 Strasbourg, France ( 1 77) ISAO MOCHIDA,Institute of Advanced Material Study, Kyushu University, Kasuga, Fukuoka 816, Japan (39) YOSHIHIKO MORO-OKA, Research Laboratory of Resources Utilization, Tokyo Institute of Technology, Yokohama 227, Japan (233) SVEN RAUSCH,Institut ,fur Chemische Technologie, Technische Hochschule Darmstadt, 0-64287 Darmstadt, Germany (87) KINYASAKANISHI, Institute of Advanced Material Study, Kyushu University, Kusugu, Fukuoka 816, Japan (39) WATARU U ~ D AResearch , Laboratory of Resources Utilization, Tokyo Institute of Technology, Yokohamu 227, Japan I (233) HAFXMUTWENDT,Institut fur Chemische Technologie, Technische Hochschule Durmstadt, 0-64287 Darmstadt, Germany (87)

‘Present address: Department of Environmental Chemistry and Engineering, Tokyo Institute of Technology, Yokohama 227, Japan.

vii

This Page Intentionally Left Blank

With Volume 39, Paul Weisz retired as our co-editor of Advances in Catalysis, a sad break for us, his colleagues for 37 years. His achievements in catalysis were recently honored by his receipt of the U S . National Medal for Technology. We would like to give him thanks for all his editorial work, and it is good to know that we may still rely on his advice as he joins the Advisory Board. At the same time, we give a hearty welcome to Dr. W. 0. Haag who joins us as a co-editor. We wish him well in his work for Advances in Catalysis, beginning with the following introductory remarks for Volume 40. D. D. ELEY H. PINES

As Drs. Eley and Pines have pointed out, readers of this 40th volume of Advances in Catalysis will notice a change in the list of editors. I would like to add my appreciation to Paul Weisz for helping to guide this publication through an astonishing growth period in the science of catalysis, driven by new techniques, methods, and catalytic materials, and by an enormous growth in industrial applications. In a very commendable article, John Meurig Thomas [Angew. Chem. 106, 963 (1994)], takes the reader on an enjoyable journey through this and earlier periods in the development of our catalytic science. Paul Weisz’s long-standing contribution to the catalytic community as a dedicated and wise editor deserves our thanks. I have reason to believe that Dr. Weisz, who has always been concerned with the relevance of science, would approve of the selections; all the articles deal with fundamental aspects of heterogeneous catalysis and their relationship to potential or currently practiced industrial applications. Two chapters deal with oxidation of aliphatic hydrocarbons. In the first article, Harold Kung summarizes the data and current understanding of the oxidative dehydrogenation of paraffins. He shows that high selectivity to the desired olefins is associated with oxide catalysts with a high heat of removal of lattice oxygen. The fifth article deals with the oxidation and ammonoxidation of propylene to acrolein and acrylonitrile, respectively. Moro-oka and Ueda show that the high activity and selectivity of new multicomponent bismuth molybdates can be understood in terms of the resulting lattice vacancies. New design principles for highly selective olefin oxidation catalysis based on modified bismuth molybdate have resulted from this study. ix

X

PREFACE

The difficult task of examining the role of catalysis in coal liquefaction has been taken on by Mochida and Sakanishi. They show the catalytic requirements in various stages of coal conversion and the many complex interactions of the catalyst with coal constituents. They also point out directions for future catalysis research needed for more economical coal liquefaction, a commendable fcature for processes requiring a long lead time. The chapter “Fundamental Studies of Transition-Metal Sulfide Catalytic Materials” by Chianelli, Daage, and Ledoux reviews current understanding of the relationship between structural and other properties of these catalysts and their catalytic activity and selectivity in hydrodesulfurization. In view of increasing environmental demands, this field has been heavily researched. The authors show how systematic studies and applications of novel methods can provide considerable understanding of these important catalysts. Two years ago, Advunces in C m l y s i s featured a chapter on chemisorbed intermediates in electrocatalysis. In this issue we follow up with a chapter by Wendt, Rausch, and Borucinski, “Advances in Applied Electrocatalysis.” The successful commercial application of electrocatalysis requires a detailed, fundamental knowledge of the many catalytic phenomena such as adsorption, diffusion, and superimposition of catalyst micro- and nanostructure on the special requirements of electrode behavior. Considerable understanding of the status and limitations of electrolysis, fuel cells, and electro-organic syntheses has been obtained and provides a sound basis for future developments. 0. HAAG WERNEK

ADVANCES IN CATALYSIS, VOLUME 40

Oxidative Dehydrogenation of Light (C, to C4) Alkanes HAROLD H. KUNG lputieff Luhorutory Department of C h e m i d Engineering Northwestern University, Evunston. Illinois 60208-3120

1.

Introduction

The low cost of light alkanes and the fact that they are generally environmentally acceptable because of their low chemical reactivity have provided incentives to use them as feedstock for chemical production. A notable example of the successful use of alkane is the production of maleic anhydride by the selective oxidation of butane instead of benzene (1). However, except for this example, no other successful processes have been reported in recent years. A potential area for alkane utilization is the conversion to unsaturated hydrocarbons. Since the current chemical industry depends heavily on the use of unsaturated hydrocarbons as starting material, if alkanes can be dehydrogenated with high yields, they could become alternate feedstock. Dehydrogenation of alkanes can be carried out thermally to produce molecular hydrogen as a by-product: CrrHzn+ 2

--+

C,,H?,z

+

H2.

(1)

The thermodynamics of this reaction for light alkanes (C, to C,), however, is such that the equilibrium greatly favors the alkane at low temperatures at atmospheric pressure. The equilibrium shifts to the side of the products as the temperature increases. The temperature for 50% conversion of alkane to the corresponding alkene ranges from about 720°C for ethane to about 600°C for propane and butane (2). In addition, the number of molecules in the product is larger than that in the reactant. Thus, operation at elevated pressures preferred in practice would shift the equilibrium conversion in the unfavorable direction. Catalytic dehydrogenation at such high temperatures has a number of disadvantages. At these temperatures, undesirable side reactions are difficult to control. The most significant side reaction is cracking of the alkane into smaller molecules. One effect of this is rapid coking of 1

Copyrighl 0 1994 by Academic Presa. Inc. All rights of rcprnluction in any h r m mcrvcd.

HAROLD H. KUNG

2

the catalyst ( 3 ) . Finally, the dehydrogenation reaction is highly endothermic, and heat must be added to sustain the reaction. In practice, the heat released by burning off the coke on the catalyst in the regeneration process is used to supply part of the heat required for reaction. An alternate method of dehydrogenation is by reaction of alkane with oxygen: C,H,,,+z

+$02

-

C,,HZn + H20.

(2)

The formation of the very stable product, water, makes this reaction very thermodynamically favorable. Thus, in principle, practically complete conversion can be obtained even at low temperatures and high pressures. This provides great advantages over the nonoxidative process based on engineering and economic considerations. The many other possible reactions between oxygen and alkenes, other desired unsaturated hydrocarbon products, as well as alkanes, place a high demand for selective catalysts. For example, reactions with oxygen could result in oxygen-containing organic products such as alcohols, ketones, aldehydes, acids, and combustion product carbon oxides. In most cases, these other reactions are much more thermodynamically favorable than the desired oxidative dehydrogenation reaction. Thus, to be able to carry out the oxidative dehydrogenation reaction with high yield is a very challenging catalytic problem. Other oxidants could be used instead of oxygen. In the 1960s, many patents and reports have appeared using bromine, sulfur, and, especially, iodine and their compounds as the oxidant. Some of these have been summarized in the treatise by Hucknall ( 4 ) . The advantages of using these oxidants is that high selectivity for dehydrogenation could be obtained. For example, 86% yield of ethene was reported in the oxidative dehydrogenation of ethane using a mixture of HCl, HzO and 0 2 over a Fe-A1 oxide catalyst ( 5 ) . However, small amounts of chlorohydrocarbons were also formed. The corrosive nature of the halogen and sulfur gases and the potential environmental concern over their use have deterred commercialization of these processes. This paper covers only oxidation using oxygen, although a few examples using nitrous oxide are also mentioned. II. Gas-Phase Reaction of Alkanes with Oxygen

Thermal reactions of light alkanes with oxygen in the combustion process have been studied extensively (6, 7). These studies were typically conducted at high temperatures-flame temperatures. The elementary reactions of the hydrocarbon species often involve reactions with atomic (H, 0) or free radical species (OH, alkyl, etc.). The initiation step is the homolytic cleavage of C-C single bonds to form alkyl radicals. The C-C bonds are the weakest bonds in an alkane molecule (Table I). The chain-propagation step

3

OXlDATIVE DEHYDROGENATION OF LIGHT ALKANES TABLE I Typical Bond Energies a Bond type

Energy (kJ mole-’)

c-c

376 420 40 1 390 361 445

Primary C-H Secondary C-H Tertiary C-H Allylic C-H Vinylic C-H

““CRC Handbook of Chemistry and Physics,” 71st ed., pp. 9-95, 1990.

involves reactions of alkane with the atomic or radical species, which abstracts a hydrogen atom from the alkane molecule to produce another alkyl radical. The activation energy and the preexponential factor of the rate constants for reactions important in the combustion of light alkanes have been reported (7). Table I1 shows the activation energy for reactions of ethane and propane that are relevant to the discussion of the catalytic oxidation reactions. TABLE I1 Activation Energies and Rate Constants of Some Gas-Phase Reactions of Ethane, Ethene, Propane, and Propene (Derived from Ref. 7) Reaction CzHs 4 CH3 + CH3 CzHs + 0 --+ CzHs + OH CzHs + OH CzHs + HzO C2H4 C2H4 C2b C2H4

-----

+0 CH, + HCO + 0 +CHzO + CH, + OH CzH, + HzO + OH CH, + CHZO

C3H8 -----* CHI + CzHs C3Hx + 0 i or n-C,H, + OH C,Hx + OH i or n-C3H7 + HzO C3Hs + H,O i or n-CsH7 + HzOz C3H8 + O2 i or n-C3H7 + HOz C3Hh C,Hs C,Hs C3H6 C3H6 C,H6 CIH, C3H6

C~HS +H C2H3 + CH3 +0 C2Hs + HCO + 0 ----j CzH4 + CHzO +0 CH, + CH3CO + OH CH, + CH3CHO + OH C,HS + HzO + OH CzHs + CHzO

E, (Kcal Mole-’)

k(sec-l or cm3 mol-sec-’) at 873 K

88.31 6.36 I .8l

2.4 X lo-’ 6.6 X 10” 3.1 x 109

1.13 5.00 1.23 0.96

1 x 10’2 2.5 x 1013 2.4 X 10” 1.2 x 10’2

84.84 3.00 0.85 18.00 41.50

1.3 X lo-’ 9 x 105 3.5 x lo* 1.6 X lo8 6 X lo1

78.00 85.80 0.00 0.00 0.60 0.00 0.00 0.00

10-5

3.5 4 8 3.5 3.5 8

x 10’2 x 10’j X lo’* x 10’’ x lot2 X 10“

4

HAROLD H. KUNG

From the activation energies and the preexponential factors, the rate constants at 873 K can be calculated. They are listed in Table 11. They show that for the gas-phase homogeneous reactions, the reactions of 0 atoms and OH radicals with ethene are very rapid and somewhat faster than their reactions with ethane. This fact would limit the maximum yield of ethene. It is well known that if the reactions of an alkane and an alkene are both first order in the hydrocarbon, then the maximum yield for the alkene of about 35% would be obtained when the rate constants, k, and k ~ for , the two reactions have equal values: alkane

101

alkene

t

degradation products.

(3)

If the rate constant kri is larger than k A , that is, the alkene reacts more rapidly than the alkane, the maximum yield would be lower. The same consideration applies to the reaction of propane. The reactions of 0 atoms and OH radicals are more rapid with propene than with propane. In fact, the differences in the rate constants are larger for the reactions of C3’s than Cz’s. This would result in a lower maximum yield for propene from propane than for ethene from ethane. Another aspect of the gas-phase homogeneous oxidation reaction at high temperatures is that the major reaction of the alkyl radicals (ethyl and propyl) is their reaction with molecular oxygen to form alkene and 0 7 H rddical: RCH2.

+0 2

--+

R’CH = CH2

+- . OzH.

(4)

Thus dehydrogenation is the primary reaction in the oxidation of alkane, and most of the degradation products are formed from secondary reactions. This has been demonstrated experimentally (8). For example, butenes and butadiene are formed with high selectivities at low conversions in the oxidation of butane.

111. Studies of Catalytic Oxidative Dehydrogenation in the 1960s

Prior to the 1960s, studies on the catalytic oxidation of alkanes emphasized their combustion. Significant activities on oxidative dehydrogenation of alkane began in the 1960s. One of the earlier reports that appeared in 1961 on oxidative dehydrogenation of pentane and 2-methylbutane (isopentane) mentioned the importance of the nature of surfaces on the conversion and selectivity ( 9 ) . However, the authors did not conduct their studies with specific catalysts. An attempt to use Group VlIl metal oxides for oxidative dehydrogenation of butane at 340-440°C did not succeed. Apparently, the reactor was not designed to minimize gas-phase homogeneous reactions,

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

5

such that the conversion of butane was about the same with and without the catalyst (10). On the other hand, cobalt molybdate was reported to greatly enhance the formation of butadiene from a mixture of butane and oxygen than from simple dehydrogenation (11). Interestingly, an Sb-Mo oxide catalyst was reported to oxidize 2-methylpropane to methacrolein with 49% selectivity at 22% conversion, and propane to acrolein with 29% selectivity at 15% conversion using a mixture of alkane, air, ammonia, and water at 508°C (12). Thus there was evidence early on that the oxidation of alkane could result in various type of product, depending on the catalyst and the hydrocarbon. The relationship among the selectivity for different products, the nature of the catalyst, and the alkane are discussed further later. Other catalysts for alkane oxidative dehydrogenation have also been reported in the patent literature. For example, it was claimed that a Na and Li phosphomolybdate produced 17% butadiene and 5% butenes at 600°C with a 1 : 1 mixture of butane and oxygen (13). IV.

Oxidative Dehydrogenation of Ethane

Unlike higher alkanes, ethane contains only primary C-H bonds, and the dehydrogenation product ethene contains only vinylic C-H bonds. As shown in Table I, these are strong bonds. Thus one would expect that, compared to other alkanes, the activation of ethane would require the highest temperature, but the reaction might be the most selective in terms of the formation of alkene. Indeed, this appears to be the case. There have been many reports on this reaction and a large number of catalysts have been studied. Figure 1 shows a summary of the performance of the better catalysts. The data points are divided into two groups according to whether the reaction temperature was below or above 600°C. In general, the selectivity for dehydrogenation was higher for data collected at the higher temperatures. These higher temperature data cluster about the line corresponding to the case of k A = kE of Eq. (3). This suggests that for these catalysts at the conditions tested, the rate constant of oxidation of ethane is comparable to the rate constant of oxidation of ethene, and a maximum yield of ethene of about 35% could be achieved. The data points of lower temperatures fall below this line. Thus their rate constants of oxidation of ethene are larger than those of ethane. The data in the figure indicate that none of the oxide catalysts known is more active for ethane activation than for ethene activation. It has been demonstrated, however, that the activity of an oxide catalyst for ethane oxidation can be preferentially increased by treating it with chlooxide catalyst was treated with ride or sulfide (14). If a Co-Zr-P-Na-K CH3C1, an ethene selectivity of 85% at 55% ethane conversion was obtained at 675"C, compared with 74% selectivity at 32% conversion on the

HAROLD H . KUNG

6

.'\

1o o p

-0

> . .

f

n

30-

2ol 10 "

0

10 20 30 40 50 60 70

Conversion % FIG. 1. Selectivity for oxidative dehydrogenation of ethane to ethene. Data taken from Table 3. (0)Reaction under or (+) above 600°C. Solid line denotes selectivity-conversion relationship for k , = k ,...

same catalyst but without the chloride treatment. Likewise, the catalyst showed 85% selectivity at 68% conversion after being treated with sulfate. At such a high reaction temperature, it is possible that chlorine of sulfur desorbed into the gas phase and promoted homogeneous dehydrogenation reaction. The use of N 2 0 instead of O2 as the oxidant has also been studied. High selectivities for ethene were obtained at low conversions (15, 16), but the reported data were not superior in terms of the dehydrogenation yield to those shown in Fig. 1. It is interesting to note that the catalysts that show good selectivities at the higher temperatures generally do not contain easily reducible metal ions, such as V, Mo, or Sb. Many of the catalysts for the lower-temperatures operation, on the other hand, contain these reducible cations. In a study using a Li-Mg oxide, it was established that gas-phase ethyl radicals could be generated by reaction of ethane with the surface at about 600°C (17). These radicals could be trapped by matrix isolation and identified by electron spin resonance spectroscopy. The dependence of ethene selectivity on the conversion of ethane for the better catalysts shown in Fig. 1 shows that the selectivity is high at low conversions and decreases as the conversion increases. This trend is consistent with a reaction pathway that consists of mostly sequential reactions [Eq. (3)]. Depending on the reaction temperature, the reaction network may involve two parallel reaction pathways shown below, which is modified from

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

7

that presented in Ref. 17. The first step in both pathways is the breaking of a C-H bond of ethane to form an ethyl species. The ethyl species could react further on the surface [Eqs. (6) and (7)] or, at sufficiently high temperatures, desorb into the gas phase to react with an oxygen molecule to form ethene [Eq. ( 5 ) ] . The contribution of the heterogeneous-homogeneous pathway increases with increasing temperature.

The surface reaction consists of two competitive pathways. Their relative rates determine selectivity. The ethyl species may undergo further dehydrogenation to form ethene [Eq. (6)] or be oxidized to ethoxide and then to acetaldehyde or acetate [Eq. (7)], and possibly to carbon oxides. The formation of ethoxide is favored at lower temperatures and in the presence of water vapor (18, 19). Other surface reactions are also possible. They are discussed later. An alternate view regarding selectivity has been proposed for Si02supported V205 (20). In this proposal, ethane could be activated in two ways. One was by dissociative adsorption across a V=O bond to form a- orJ-elimination of the ethoxide would result in acH-V-OC2Hs. etaldehyde or ethene, respectively. The other way was by dissociative adsorption across a V -0 -* bond to form * -OH and V -C2Hs. The latter species would lead to combustion products. The nature of the O(s)species for ethane activation probably varies depending on the catalyst. On reduced Mo/Si02, the surface 0- species generated by the decomposition of N 2 0 has been shown to react readily with ethane ( 1 . 3 , and 0- has also been suggested as the active species for the Li-Mg oxide catalyst (17). Oxygen vacancies or surface vanadyl groups may also be active sites. It is worth noting that catalysts containing metal cations that form stable carbonates may do so under reaction conditions. La203and Pr203 have been found to convert to La202C03and Pr202C03,respectively, under reaction conditions (21, 22). It is probable that the alkali and alkali earth cations in

8

HAROLD H. KUNC

TABLE 111 Summup Dutri of Oxidutivr 1)rhvrirogenurioii Oxide catalyst Co-Zr-P-NaLi- Na- Mg Li -Mg Li --l'i-Mn

Temp. ("C) -K

615 600 - 650 600-650 650

Li-Ti

6.50

Li -Mg

600

625 700 550

Sr-CeYb BzOj B-Si -Zn -La -- Al -Mi! MO-V -Sb M O-V -Ta

400 400

Mo--V -Nb

400 550 550 325-360

CrP04 Cr-ZrP2O7 V--P

Ethune

Ethane conv. 1%)

Ethene select. (%)

32 38 38 23 41 10 21 39 55 54 58 4 4 35 6 38 4 23 I8 25

14 85 80 84 14 92 88 74 51 64 I0 91

5x

30 24 6 15 18

25 30 500 500 390

of'

12 12 7

95

Ref. 14

66 66

67 67 17

68 69 69

41 82 58

xo 15 I1 69 6.5 60 48

12 60 55 48 40 54 46 62

70 70

70 71 71 56

21 21 72

many of the catalysts in Table Ill form carbonates if the reaction temperature is below their decomposition temperatures. Very high selectivities for ethene at high conversions could be achieved in the absence of gaseous oxygen. As mentioned before, under such conditions, the catalysts would be deactivated by coking. If the catalyst is an oxide, deactivation by reduction also occurs. The catalysts must be regenerated by treatment with oxygen in a cyclic operation ( 2 3 ) . V.

Oxidative Dehydrogenation of Propane

Propane contains one secondary carbon, and the secondary C-H bonds are weaker than the primary C-H bonds (Table I). Therefore, propane is

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

"

0

10

20

30

40

50

60

9

70

conversion %

R c i . 2. Selectivity for oxidative dehydrogenation of propane to propene. Data taken from Table 4.Solid line denotes selectivity-conversion relationship for k , = kc:.

expected to react more readily than ethane. Indeed, this has been observed (24-27). Following this argument, since the dehydrogenation product propene contains acidic allylic hydrogen and weak allylic C-H bond, the difference in reactivity between propane and propene is expected to be larger than the C2 compounds. An examination of the summary of the literature data for the better catalysts shown in Fig. 2 shows that all of the reported data fall below the line corresponding to the case of k A = k b , suggesting that propene reacts faster than propane on all of these catalysts. In the case of V-Mg oxide, the rate constant for the oxidation of propene was found to be about four times that of propane (27). The catalysts used and the temperature for the data shown in Fig. 2 are listed in Table IV. Except for one, the studies were conducted at or below SSOOC, which was substantially lower than many experiments for ethane oxidative dehydrogenation. This is because above this temperature, the contribution of homogeneous gas-phase reaction begins to be significant (see, for example, Ref. 28). The catalysts based on vanadium oxide are one of the better studied systems. A V-Mg oxide in which Mg orthovanadate (Mg*(VO&) and MgO were the only identifiable phases was a rather selective catalyst (27). Since MgO was relatively inactive in alkane activation, Mg orthovanadate was assumed to be the active component. Indeed, Mg orthovanadate prepared as a stoichiometric compound showed high selectivities for the oxidative dehydrogenation of propane (29). In this latter study, it was shown interestingly that Mg orthovanadate was the only alkali or alkali earth orthovanadate that

10

HAROLD H. KUNG TABLE IV Summary Datu ( f Oxidutive Dehydrogenation of Propune

Oxide catalyst

Temp. ("C)

V-Na-AI V-silicalite

490 400-500

V-Fe-Nd-AI MnS04

625 550

Co-Mn V-Mg"

538 540

550

B-P Bi-V Nb-promoted V-Sb-AI

550 550 550 520'

Propane conv. (%)

Propene select. (%)

21 12 23 27 30 40 4 11 32 4 12 20 29 40 63 45 32 22 10 5 8 15

80 70 65 60 70 66 79 73 55 78 62 54 43 34 38 28 38 48 60 83-87

16

Ref. 73 74

75 76

77 27

30

78 79 80 81

75

"These two samples were prepared by different methods. 'These data were for different compositions of V-Mg-0. ' NH, was added to the feed.

was selective. Ca, Sr, Ba, Li, and Cs orthovanadates were nonselective. The probable explanation is that except for MgC03, which decomposes at about 400°C, the high thermal stability of the other carbonates results in disintegration of their orthovanadate into carbonates and vanadium oxide under reaction conditions to become nonselective catalysts. Mg pyrovanadate (MgZV2O7)was also found to be a selective catalyst for this reaction (30). A later study further showed that the catalytic performance of MgZVrO7was insensitive to the method of preparation, such that there were only minor differences whether or not the oxide contained small amounts of potassium (25). In contrast, the presence of K in Mg orthovanadate degraded its selectivity noticeably. Temperature-programmed reduction with HZ and electrical conductivity characterization in the presence of propane showed that Mg pyrovanadate could be reduced by both HZ and propane faster than Mg orthovanadate containing K (31). The dependence of selectivity for propene on propane conversion for the better catalysts (Fig. 2) follows a trend similar to that for ethane oxidation.

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

11

The selectivity is high at low conversions and decreases as the conversion increases. Thus the sequential reaction of Eq. ( 3 ) is the primary reaction pathway for these catalysts, and the activation of propane is mostly by breaking of a secondary C-H bond to form a propyl species, with breaking of a primary C-H bond contributing to a minor extent. Although no direct observation of the formation of propyl species has been made, there was indirect evidence of their formation and subsequent desorption into the gas phase at sufficiently high temperatures on a V-Mg oxide catalyst. This was demonstrated by monitoring the reactions in the postcatalytic region of the reactor (28). A V-Mg oxide catalyst wafer, the composition of which was equivalent to 19 wt% Vz05 and 81 wt% MgO, was mounted into a reactor shown schematically in Fig. 3. The reactor was filled with quartz chips except for a region (void volume) immediately to one side of the catalyst wafer. The quartz chips were shown to quench homogeneous free radical reactions effectively under the reaction conditions employed. A feed mixture of propane, oxygen, and helium could be introduced from either side of the reactor by the switch of a valve. Thus reaction measurements could be made to determine the effect of the position of the void volume with respect to the wafer, that is, whether it was upstream (precatalytic configuration) or downstream (postcatalytic configuration) of the catalyst by simple switching of a valve, without making any change to the rest of the reaction system. It was expected that if there was no desorption of reactive intermediates from the catalyst that could initiate or propagate homogeneous reactions, the measured conversion and product distribution would be independent of the reactor configuration. However, if significant desorption of reactive intermediates occurred, the conversion in the postcatalytic configuration would be higher than that in the precatalytic configuration, and the product distribution would be different. Table V shows the results of this experiment. The major products were propene and carbon oxides, with small amounts of cracked products. In the

FIG. 3 . Schematic of a reactor used to monitor effect of desorbed species on the reactions in the postcatalytic region.

12

HAROLD H. KUNG

TABLE V Conversion and Product Selectivity in Propane Oxidation in the Precutalytic and Postcatalytic Conjigurations" 556°C

C3H8conversion (5%) 0, conversion (%) Carbon product sel.(%)

co co2

CH, CzH4 CzH6 "Weight of V-Mg (from Ref. 28).

570°C

Precat .

Postcat.

Precat.

Postcat.

16.7 14.9

22.5 13.6

33.2 23.7

46.2 27.1

14.9 20.8 2.6 8.9 52.8

11.0 13.4 2.9 13.1 59.6

18.0 20.8 5.1 15.2 40.9

16.3 9.8 6.2 22.8 44.9

oxide catalyst wafer, 0.021 g; propane:02:He = 6 : 12: 82 vol %

postcatalytic configuration, the conversions were higher than those in the precatalytic configuration, and the selectivities for propene were higher. The propane conversion was a combination of contributions from the catalyst wafer, the homogeneous reaction in the void volume independent of the catalyst, and the surface enhanced homogeneous-reaction due to desorption of reactive intermediates from the catalyst wafer. The contribution from the catalyst wafer should be identical for both configurations. Except for possible minor differences in the temperature profiles in the void volume, the portion of the homogeneous reaction in the void volume not induced by desorption of reactive intermediate should be quite similar also for the two configurations. Thus the different conversions observed strongly suggested desorption of reactive surface intermediates into the gas phase that enhanced the homogeneous reaction. The extent of enhancement in the conversion of propane in the postcatalytic configuration could be expressed as

E(%) =

x,,, - x,,,, x 100% X",,

(8)

where X,,, , X,,,, , and Xv,,Irespectively are propane conversion in the postcatalytic configuration, the precatalytic configuration, and in the void volume without the catalyst wafer, which could be independently determined. E was found to increase with the concentration of propane in the feed, with the weight of catalyst (Fig. 4), and with temperature (Table V). To eliminate the possibility that the enhancement was due to the thermal effect that the temperature in the void volume was being raised by the

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

13

-g 401

.

.

0.016

I

. , .

0.021

. ,

. , .

0.025

Weight of Catalyst Wafer (g)

FIG.4. Extent of enhancement ( E ) as a function of the weight of catalyst wafer at 570°C (From Ref. 28). C1/02/ He:) . ( 4/8/88; (+) 6/ 12/82; (A)8 / 16/76.

exothermic reaction on the catalyst wafer, which then increased the homogeneous reaction in that region, the same experiments were conducted using a ZnO wafer. ZnO was a nonselective oxidation catalyst, and carbon oxides accounted for 90% of the products. Any thermal effect due to reactions on the wafer should be more apparent with ZnO because of the high heat released by the combustion reactions. Indeed, the measured temperature near the wafer in the void volume was higher with the ZnO wafer than with the V-Mg oxide wafer. However, there was no measurable enhancement using ZnO, that is, E was zero within experimental uncertainties. Thus the conversion of propane in the postcatalytic configuration equaled that in the precatalytic configuration, and the thermal effect was not the cause of the enhancement. There was yet another possibility that the enhancement could be due to the fact that some stable products formed on the catalyst wafer, upon desorption into the void volume, underwent further sequential reaction with propane. If so, the enhancement would not require the immediate adjacency of the catalyst wafer and the void volume and should be observable when the catalyst and the void volume were physically separated. Such separation, however, would quench any desorbed reactive intermediates. This was tested. The wafer was separated from the void volume, and the two were separately heated to the appropriate temperatures. The result was that only a small enhancement was observed in the separated mode. This confirmed that the enhancement was due not to sequential reaction of stable products but to desorption of reactive intermediates from the catalyst surface. The small enhancement could be attributed to the higher temperature throughout the separately heated void volume in the separated mode than in the other mode.

14

HAROLD H. KUNG

Although the experiments described indicated that reactive intermediates were desorbed from the catalyst surface, the identity of these intermediates was unknown. Nonetheless, insight could be developed comparing experimental data with reactor modeling of the void volume. Using the experimental temperature profile, the product distribution and activity of the catalyst wafer, and assuming a rate of desorption of some reactive intermediates such as methyl, ethyl, propyl, and hydroxyl radicals, the product distribution at the exit of the void volume could be calculated (32). Table VI shows the results of such a calculation. It can be seen that desorption of methyl radicals would result in a much higher methane selectivity than what was observed. Similarly, desorption of ethyl radicals would result in too high an ethane selectivity. The results for desorption of either hydroxyl or propyl radicals agreed with experimental results. The results of the calculation could not distinguish hydroxyl or propyl radicals as the desorbed species, since they produced very similar product distributions. This is because they were both involved in the chain propagation reactions. Since the desorption of a hydroxyl radical required cleavage of metal-oxygen bonds in the oxide lattice, i t would be energetically less favorable. Thus it was concluded that propyl radicals were the most likely desorbed species. VI.

Oxidative Dehydrogenation of Butane and Cyclohexane

There are fewer studies of oxidative dehydrogenation of butane, and even fewer for cyclohexane than ethane or propane. The performance of the better catalysts in these two reactions are summarized in Table VII and Fig. 5. Because of the larger number of secondary carbon atoms in these molecules, they are more reactive with gaseous oxygen than the smaller alkanes. In ex-

TABLE VI Calculated Conversions cmd Product Distributions in the Postcatalytic Void Volume$)r Various Assumed Species Desorbed from V-Mg Oxide during Propane Oxiduiion ut 570°C (from Ref. 32)

Carbon products sel. (%) CIHHbasis

Experimental f(CzHs) = 30% f(CH1) = 5% f ( O H ) = 30% f(C3H7) = 23%

22.6 21.1 21.6 21.3 22.9

7.6 3. I 0.7 2.0 6.6

0.9 4.7 4.0 3.3 0

15.3 11.5 32.9 13.9 15.8

27. I 41.3 26.3 28.7 25.2

49.1 33.4 36. I 52.1 52.4

" f ( i ) is the fraction of species i produced on the catalyst that desorbed into the gas phase.

15

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES TABLE VII Surnrnury Dam qf Oxidative Deydrogenution of Butane and Cyclohexane Selectivity %

T(T)

Catalyst V-Mg-0

Cyclohexane 484

Alkane conv. (%) 9 11

V-Nd-0

484

16 21 6 10

14 22

Alkene

BD or BZ"

Total deh."

52 47 31 28 38 31 24 18

36 38 41 52 40 46 46 46

88 85 78 80 I8 71 I0

53 42 36 32 28 20 29

15 18 17 23 26 33 24

68 60 53 55 54 53 53 75 72 51 49 58 70 60 75 66 60 51

Ref. 33

33

64

Butane V-Mg-0

475 500-540 480

Ni-Mo-P-0

Mg-Ni-SO, Ni-P-0

482 538 482 565 538 538

V-K-SO, V/Si02

500 520

Mg-Ni-Sn-0

4 9.5 18 23 II 40 31 25 39 29 39 36 14 42 25 15 16 25

30 19 27

21 30 31

54 53 41

12 8 10

26

35

82

83 83 84 85 86

" Butadiene or benzene

periments using a 5-ml reactor, substantial noncatalytic oxidative dehydrogenation reactions occur for cyclohexane at about 450"C, slightly above 500°C for butane, and above 550°C for propane. The reactivity of these hydrocarbons on oxide catalysts parallel this trend (24-27, 33). In addition to the corresponding alkenes, dehydrogenation of butane and cyclohexane could result in butadiene and benzene, which are very stable conjugated unsaturated hydrocarbons. Therefore, it should be possible to attain high yields of butadiene or benzene. Indeed, the data in Table VII show that these products represent substantial fractions of the dehydrogenation products in most cases. Total yields of dehydrogenation products on these better catalysts, however, are still limited. Figure 5 shows that, similar to the oxidation of

HAROLD H. KUNG

16

0

10

20

30

40

50

Conversion %

FIG. 5. Selectivity for oxidative dehydrogenation of cyclohexane and butane. Data taken from Table 7 .

propane, these data fall below the line corresponding to kA = k ~ Indeed, . on a V-Mg oxide catalyst, it was estimated that butenes react with a rate constant four times larger than that for butane (27). Figure 5 also shows the trend of increasing selectivity with decreasing conversion, similar to those observed for ethane and propane. Thus sequential reactions [Eq. (3)] are the predominant pathway for butane and cyclohexane reactions on these catalysts (26, 33). The first step of the activation of butane and cyclohexane has been assumed to be the cleavage of a secondary C-H bond, with minor contributions from primary C-H bonds in the case of butane. This picture is supported only by indirect evidence. When the relative rates of reaction of various alkanes were compared on a V-Mg oxide and Mg2V2O7catalyst (Table VIII), it was found that alkanes with only primary carbons (ethane) reacted most slowly. Those with secondary carbons (propane, butane, and cyclohexane) reacted faster, with the rate being faster for those with more secondary carbon atoms. Finally, the alkane with one tertiary carbon (2methylpropane) reacted faster than the ones with either a single or no secondary carbon (26).From these data, it was estimated that the relative rates of reaction of a primary, secondary, and tertiary C-H bond in alkanes on the V-Mg oxide catalyst were 1 , 6, and 32, respectively (26).

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

17

TABLE VIII Relative Rates of Reaction of Alkane with Oxygen on V-Mg-0 and Mgz Vr 0,Catalysts Alkane Ethane Propane 2-Methylpropane Butane Cyclohexane

V-Mg-0 0.19 0.6

1.29 1

Mg7VZ07' 0.16 0.70 0.9 1

-6

'Reaction at 5 W C , from Refs. 26, 33. "Reaction at 540"C, from Ref. 25.

Direct evidence about the first step of activation of butane was obtained on a V-P oxide catalyst in the butane oxidation to maleic anhydride based on deuterium kinetic isotope effect (34). It was found that when a butane molecule was labeled with deuterium at the second and third carbon, a deuterium kinetic isotope effect of 2 was observed. No kinetic isotope effect was observed, however, if the deuterium label was at the first or fourth carbon. By comparing the observed and theoretical kinetic isotope effects, it was concluded that the first step of butane activation on this catalyst was the cleavage of a secondary C-H bond, and this step was the rate-limiting step. The performance of the V-Mg oxide catalyst was found to depend on its composition and the method of preparation. As to the composition, it was found that catalysts containing very small or very large amounts of vanadium were not selective. The better catalysts in terms of both activity and selectivity consisted of from about 10 to 60 wt% V,Os (3.5). Analyses of these catalysts by X-ray diffraction, Auger electron spectroscopy, and infrared spectroscopy showed that they contained only two identifiable phases: Mg orthovanadate (Mg3(V04)?)and MgO. Since MgO had low activity and poor selectivity under the reaction conditions employed, it was concluded that the active phase was Mg orthovanadate (Mg3(V04)2). Indeed, it was later shown that this compound was a selective catalyst (26). As to the method of preparation, it was found that V-Mg oxide catalysts prepared with a Mg(OH)2 precursor that was precipitated with KOH was less selective than one prepared with a MgCOl purecursor precipitated with (NH,),CO3 (25). Interestingly, unlike the butane reaction, there was no effect of preparation on the oxidative dehydrogenation of propane using the same catalysts, as mentioned earlier (25, 30). Unlike the oxidation of propane, Mg pyrovanadate was nonselective for butane (25, 26). Mg metavanadate was nonselective as well (26).

HAROLD H. KUNG

18

In addition to Mg, other orthovanadates have been studied, including Sni, Nd, Zn, Cr, Fe, Ni, Cu, and Eu (36-39). It was found that these orthovanadates showed a wide range of selectivities. Figure 6 shows the selectivity for dehydrogenation at 12.5% conversion of butane (36, 39) at 500°C. The selectivity ranged from a high of over 60% for Mg3(V0& to a low of less than 5% for Cu3(VO&. Furthermore, it could be correlated with the reduction potential of the cation: the higher the reduction potential, that is, the more easily reduced the cation, the lower the selectivity. The interpretation of the data in this figure is as follows. The structure of bonds in the lattice orthovanadate is such that there are only M-0-V (40). Since the removal of a lattice oxygen from these bonds is accompanied by reduction of the neighboring cation, the ease of removal of a lattice oxygen should depend on the reduction potential of the cation. If it is in turn related to the probability for a surface hydrocarbon intermediate to react with the lattice oxygen to form a C-0 bond and thus oxygen-containing products, then the correlation shown in the figure would be observed. This interpretation of the correlation assumes that lattice oxygen is the species that reacts with the surface intermediate. This assumption is sup-

$

Zn

40

6 P

Eu

30 -

Ni Fe

cu -3

-2

1

0

1

Reduction Potential of Cation, V FIG.6 . Dependence of selectivity for oxidative dehydrogenation of butane (to butenes and butadiene) on the reduction potential of the cations in orthovanaddtes of the formula Mi' (VO4)2 and M Z i VO,. Reactions conditions: 50O0C, butane conversion = 12.5%, butane: 0,: He = 4:8 : 88 (from Ref. 39).

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

19

ported by the pulse reaction data of butane oxidation over Mg, Zn, Fe, Ni, and Cu orthovanadates (38). After treating the orthovanadate in oxygen at 500°C and purging with He,pulses of butane were passed over the catalyst in the absence of oxygen. The selectivities for dehydrogenation to butenes and butadiene for the first pulse were found to be 68, 32, 12, 8, and 0% for Mg, Zn, Fe, Ni, and Cu orthovanadates, respectively. These values were very close to the steady-state selectivities. It is interesting to note that on Mg3(V0&, the selectivity for dehydrogenation was maintained in subsequent pulses up to the end of the experiment when an equivalent of 50% of a monolayer of lattice oxygen was consumed. The activity was decreased by only about 20%. The behavior of Ni orthovanadate with pulse number was similar. For Zn orthovanadate, the decrease in activity was more pronounced. Another method for determining the ease of removal of a lattice oxygen is by measuring the differential heats of reoxidation of a reduced oxide. Conceptually, the heat of reoxidation measures the energy released when two metal-oxygen bonds are formed in the solid together with all associated processes; the latter include structural rearrangement and changes in oxidesupport interaction. Although a quantitative measure of the heats of the associated processes is not available, it is likely that their magnitudes are smaller than the metal-oxygen bonds energy, and a meaningful qualitative trend can still be derived directly from the reoxidation heat. The differential heats of reoxidation and the reaction characteristics of a number of V205/y-A1203catalysts have been measured (41). The chemical transformation corresponding to the heat measurement in this system is ~ / 2 0 2+ V205-Jy-Al203

v 20 5 / y

-A120 s .

(9)

In these measurements, the samples were first reduced with HZto an extent roughly equivalent to the conversion of all VSf to V4+.Then they were incrementally reoxidized with doses of 0 2 . The heat released during reoxidation was measured with a microcalorimeter and the amount of O2consumed was measured volumetrically. From such measurements, the heats of reoxidation were determined for a number of samples, including two samples of 8.2 and 23.4 wt% VzO, loadings, which were equivalent to surface coverages of 2.9 and 8.2 V/nm2, respectively, as a function of the degree of reduction, (b, which was defined as the number of oxygen atoms removed per vanadium ion. It was found that the differential heats of reoxidation were function of 4, vanadium loading, and temperature. Figure 7a shows the data at 400°C for the 8.2 V/nm2 sample, and Fig. 7b shows those at 500°C for the 2.9 V/nm2 sample. For both samples, the heats were found to be high at large (b and decreased quite rapidly as the sample became close to being

20

HAROLD H . KUNG ]

-

35

- 200 I

8.2 V/nrn2

a 160-

8 Q

1 0 00 -

z / E

-r 3 -r

0

L

0

--L

1’

_-; , 0.1

0.2

0.3

0

0.4

2.9 V/nm2

0

0.1

qJ

0.2

0.9

0.4

(0 Removed per v)

FIG. 7 . Differential heat of reoxidation and selectivity for oxidative dehydrogenation of butane on V Z O S -A120z /~ samples. For the 2.9 V/nm2 sample, the selectivity was calculated for the detected gaseous products. (a) 8.2 V / n d sample, reaction at 400°C; (b) 2.9 V/nmL sample, reaction at 1180°C; (c) 8.2 V/nmL sample, reduction by CO at 530°C. butane reaction at 400°C; and (d) 2 . 9 V/nni* sample, reduction by CO at 400°C. butane reaction at 480°C. (a) and (b) are from Ref. 50; (c) and (d) and from P. J . , Andersen, Ph.D. thesis, Northwestern University , 1992.

fully reoxidized. In addition, the heat at large 4’s was somewhat higher for the sample of lower loading. These two samples were then tested for the oxidation of butane in a pulse reactor. In these studies, pulses of butane (0.05 ml at 1 atm) were passed in a He carrier over 0.05 g of 8.2 V/nm’ catalyst or 0.75 g of 2.9 V/nm2 catalyst (42).The products were collected in a trap at 77 K and later flashed into a gas chromatograph for analysis. Since no gaseous oxygen was present, the oxygen consumed by the reaction had to be originated from the lattice.

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

2o

t

0

0

80

I

/

1

A*--,:

0.1

0.2

0.3

0.5

0.4

0.8

d 2.9 V/nm2

1

0

21

0.1

(p

1

0.2

I

0.3

0

(0 Removed per V)

FIG. I . (continued)

Therefore, from the product selectivities and conversions of the butane pulses, the degrees of reduction of the catalysts could be calculated. For the 8.2 V/nm2 sample, the products observed for the pulse reaction at 400°C consisted of only dehydrogenation products (butenes and butadiene) and carbon oxides. No oxygenates were observed, and the carbon balance for each pulse was satisfied within experimental error. The selectivity for dehydrogenation is shown in Fig. 7a as a function of #. It was very low when the catalyst was in a nearly fully oxidized state, but increased rapidly when the catalyst was reduced beyond # = 0.15. It should be noted that the dependence of selectivity for dehydrogenation on Cp shown in the figure was not a result of changes in conversion of butane in the pulse since these data were for experiments of about the same conversion. The data for the 2.9 V/nm2 catalyst are shown in Fig. 7b. The activity of

22

HAROLD H . KUNG

this catalyst was lower, and the experiments had to be performed at 480°C. Over this catalyst, a fraction of the butane that reacted in the pulse remained adsorbed. The fraction decreased from 60% in the first pulse to 30 to 40% after a few pulses. To calculate the degree of reduction of the catalysts, the adsorbed material was assumed to be elemental carbon, and the hydrogen that was not accounted for in the detected products was assumed to have reacted with oxygen of the vanadia phase to form water, which subsequently desorbed. Upon reoxidation of the catalyst, the adsorbed carbon was recovered as carbon oxides. Figure 7b shows that the selectivity for dehydrogenation (based on detected products) was very low at low values of 4 on this catalyst, but increased rapidly as the catalyst was reduced. On this catalyst, small amounts of crotonaldehyde and maleic anhydride were also detected. These amounts decreased slowly with increasing 4 . The data in Figs. 7a and 7b show that there is a strong correlation between the selectivity for dehydrogenation and the heat of reoxidation of the catalyst. The selectivity is low when the heat is low and increases rapidly when the heat increases rapidly. Since the heat of reoxidation is a measure of the metal-oxygen bond strength, this observation is consistent with the model that the ease of removal of lattice oxygen is an important factor in determining selectivity for dehydrogenation versus formation of oxygen-containing products. The degrees of reduction for the data in Figs. 7a and 7b were calculated from the conversion of butane. For the data in Fig. 7b, correction was made for the carbonaceous deposit on the catalyst. To test that the correlations in these figures were not specific for reduction using butane, experiments in which the catalysts were first reduced with CO pulses to certain 4 and then butane pulses were admitted were conducted. For the 8.2 V/nm’ sample, the rate of reduction by CO was slow at 400°C. Thus the reduction was carried out at 530°C. Afterward, the temperature was lowered to 400°C for the butane pulse reaction. The results are shown in Fig. 7c. The general trend that the selectivity for dehydrogenation increases with increasing degree of reduction was again observed, although the increase in selectivity with 4 occurs at higher values of 4. This shift in the value of 4 is probably due to the higher temperature used in the reduction. It has been shown that at higher temperatures, lattice diffusion was more rapid such that there was a lower gradient of the degree of reduction between the surface and the bulk of the V20s-x particles (41).In other words, the degree of reduction at the surface was lower for the same overall extent of reduction 4. A similar experiment was performed on the 2.9 V/nm2 sample. Because of the low reaction rate, the reduction by CO had to be carried out in a flow of CO at 400°C for 1 hr. The COz formed was collected and later quantified. Afterward, butane pulses were admitted as before. The selectivity for dehy-

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

23

drogenation as a function of 4 is shown in Fig. 7d. Again, the general trend that the selectivity increased with increasing 4 was observed, and the increase was observed at values of 4 higher than that shown in Fig. 7b, similar to the case for the higher loading sample. This difference between the two methods of reduction can be attributed to the procedure used in the CO reduction, which permitted a much larger extent of surface and bulk equilibration, as explained earlier, and the errors in estimating 4 due to the carbon deposit during butane reduction. The pulse experiments using orthovanadates and V/y-A1203 catalysts showed that high selectivity for dehydrogenation of butane could be obtained by reaction of butane with lattice oxygen. This has also been demonstrated with other oxides, including Mg-Mo oxide (43, 4 4 ) . In addition to showing that the selectivity for dehydrogenation could be related to the metal-oxygen bond strength, the data in Fig. 7 also show that the selectivity was different for V/y -Also3 of different loadings. This observation was also true for the steady-state experiments. At 7-8% conversion, the 2.9 V/nm2 sample showed a butane dehydrogenation selectivity of 43%, which was substantially higher than 12% obtained with the 8.2 V/nm2 sample (42). V/y-A1203 of different loadings contained different vanadia species (45). Therefore, the variation in reoxidation heat (and thus dehydrogenation selectivity) could also be interpreted in terms of different properties of different vanadia species. In addition to V/y-AI203, it has been shown that the vanadia species in V/Si02 also depended on the loading. Vanadium oxide, when present in very low loadings on silica, does not form crystalline V2OS.Instead, a highly dispersed phase is formed (46-49). This highly dispersed phase is structurally different from crystalline V205. For example, crystalline V2OSshows peaks in a Raman spectrum at approximately 997, 703, 526, 480, 404, 304, and 284 cm-'. However, the spectrum of a 1-wt% V2OSon Si02 only shows a peak at 1040 cm-' and none of these other peaks. Peaks of crystalline V2OSbegin to appear when the vanadia loading is increased and are clearly discernable in the spectrum of a sample of 10 wt% V205loading. It has been proposed that the 1040 cm-' peak is due to the stretching vibration of the V = 0 bond in a VO, species on silica (46-49). When two V/Si02 catalysts of 1 and 10 wt% V205 were tested for butane oxidation, the data in Table IX were obtained. The data show that the lower loading sample was much more selective than the higher loading sample. This difference was not due to the effect of impurities in the support because the data were obtained on an acid-washed Davison 62 silica which contained less than 10 ppm of Na and 200 ppm of Ca. The same effect was observed on catalysts prepared with Cabosil silica, which contained no detectable impurities .

HAROLD H. KUNG

24

TABLE IX Oxidution of Butune over V, O,/SiO, Ccitulysts

SiOz"

I 1 2 2

VZOS wt %

Catalyst wt (g)

I .o 10.0 I .o 10.0

0.5 0. I 0.5 0. I

Selectivity (%y

Conv. (%) CJH,(II)

CO

CO?

I-B

t-2-B

c-2-B

BD

TD

14 2

13 2 II 2

12 0 8 0

66 9 60 10

15

19

10

27

II 16

55 IS 61

30

4

10

30 5

18

29

12

3

' I , acid-washed Davison 62; 2, Cabosii silica. "Feed: C4HI0/0dHe = 4/8/88; total flow rate, 100 ml/min; temperature; 520°C. ' I-B, I-butene; 1 - 2 4 , t-2-butene; c - 2 - 6 , (,-2-butene;BD, butadiene; TD, total dehydro-

genation. The balance was Cz and C1hydrocarbons.

VII.

Generalized Reaction Scheme for Oxidative Dehydrogenation of Alkanes

The data presented above showed that the oxidative dehydrogenation reactions of the various alkanes share many common features. Thus it is tempting to discuss selectivity for alkane oxidative dehydrogenation with a common scheme. The reaction scheme for ethane oxidation [Eqs. ( 5 ) - ( 7 ) ] provides a useful basis for such a discussion. It shows that the primary reaction of alkane oxidation can take on three different pathways depending on the reaction temperature (Scheme I). The first step in all three pathways is breaking a C-H bond, which is the rate-limiting step. The three pathways are described below.

(I) At very high temperatures, a n alkane molecule reacts on a catalyst to produce an alkyl radical, which desorbs from the surface to undergo homogeneous gas-phase reactions. (11) At lower temperatures, the alkyl species remains adsorbed on the surface. @-elimination of the alkyl species produces an alkene. The alkene molecule may readsorb to react further in the sequential reaction scheme [Eq. (3)l. Since alkenes and other unsaturated hydrocarbons (of the same number of carbon atoms) are the intermediate products in the sequence, their further reaction to forming any oxygen-containing product, such as organic acids, anhydrides, aldehydes, and carbon oxides, would lower the selectivity for dehydrogenation. Thus the selectivity for alkene is higher at lower alkane conversions, but decreases with increasing conversion. (111) Another competing pathway for the adsorbed alkyl species is the formation of a surface alkoxide, which could be further oxidized to aldehyde and carboxylates, and perhaps eventually to carbon oxides.

25

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

RCHR,,,

1

honiogeneous reaction

R’CH = CHR

RCHR

I

0

I

R’CH = CHR

K

I

R-C-R

I \0

0

Scheme I

Except for the generation of alkyl radicals on the surface, the gas-phase reactions are independent of the catalyst and therefore are not discussed further. The discussion concentrates on the surface reactions. According to their behavior, the catalysts can be classified into two groups: (i) nonselective catalysts for dehydrogenation for which the primary products include large amounts of oxygen-containing products, and

(ii) relatively selective catalysts like those shown in Tables, 111, IV, and VII for which the primary products are the alkenes, and secondary (oxygen-containing) products are formed by the sequential reaction scheme of Eq. ( 3 ) . It would be interesting to identify properties of a catalyst that determine which of these two groups the catalysts belongs (50). If the surface reactions of an alkane depicted in Scheme I is common to all catalysts, then the selectivity for dehydrogenation versus formation of oxygen-containing product would be strongly affected by the ability of the catalyst to form a C-0 bond with the surface hydrocarbon, that is, the reactivity of the oxygen species, and the number of reactive oxygen available for reaction at the active site.

26

HAROLD H. KUNG

VIII. A Generic Description of the Relationship between Metal-Oxygen Bond Strength and Selectivity

The idea that the reactivity of the oxygen at the active site is important has been proposed by others (51 -53). One can extend these concepts to illustrate the relationship among the number of oxygen atoms available at the active site, the size of the site and the surface intermediate, and the reactivity of the oxygen using the following hypothetical example. Consider two oxides each containing one of the following two type of active sites. One site consists of a M 0 4 tetrahedron (Fig. 8A) and another consists of a M 2 0 7 unit which is two corner-sharing MO, tetrahedra (Fig. 8B). The reactivity of the lattice oxygen in these sites can be represented by the heat of removal of the lattice oxygen. It is reasonable to expect that this heat increases (that is, the removal of oxygen becomes increasingly difficult) with increasing number of oxygen atoms removed from the site, that is, the degree of reduction of the oxide. This is illustrated in Fig. 9A. If the selectivity is determined by the oxide property shown in this figure, i.e., the number of oxygen removed per cation, which is the common definition of degree of reduction, the two catalysts containing these two sites would be expected to show similar selectivities since they show similar dependence of heat of reduction versus the degree of reduction of the oxide. One the other hand, consider another case that each of the two sites adsorbs one surface intermediate species. The reaction of this intermediate is

FIG. 8. Schematic drawing of the active sites: (A) V04 unit in Mg3(V04)2;(B) V20, unit in MgzV207(C) V2OXunit in (VO)rP207.Open circles are 0 ions and solid circles are V ions.

27

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

A 100-

-

h

0)

Q

80-

0.5

1

1.5

2

2.5

3

3.5

4

Deg. redn. (0 removed/metal ion)

"

1

2

3

4

5

6

7

0

Atoms 0 removed per site (molecule)

Fro. 9. Change in heat of removal of lattice oxygen with degree of reduction of the site measured as: (A) number of 0 removed per metal ion (B) number of 0 removed per site (or per adsorbed molecule). Hatched bars are for MO, sites and dotted bars are for M 2 0 7 sites (from Ref. 39).

such that it can acquire oxygen atoms up to a certain heat of reduction. In this case, as is shown in Fig. 9B, this intermediate would be able to take up a larger number of oxygen atoms from the M207 site than from the M 0 4 site. Thus the oxide with the M 2 0 7 sites would be less selective than the one with the M 0 4 sites. This example demonstrates that detailed information about the interaction of the surface intermediate with the active site is very helpful in understanding changes in selectivity patterns in oxidation catalysis. A corollary to the conclusion of the above discussion is that the number of lattice oxygen available for reaction at the active site depends not only on the atomic details of the site, but also on the rate of lattice diffusion compared with the rate of surface reaction. If the rate of lattice diffusion is slow relative to surface reaction, then the number of lattice oxygen atoms available for the reaction during the residence of an intermediate could be estimated from the atomic structure of the site, as in the examples above. On the other hand, if lattice diffusion is fast, a lattice oxygen atom will be replenished immediately after its removal. Then the behavior of the catalyst no longer depends as much on the stoichiometry of the active site, but on the degree of reduction of the active site at steady state, which would determine the heat of removal (or the reactivity) of the lattice oxygen. As to the positic-n of the surface intermediate at which the reactive oxygen reacts, it has been suggested that this is determined by the electron density of the oxygen (54, 55). A nucleophilic (high in electron density) oxygen

28

HAROLD H . KUNG

species would tend to attack the carbon atom to form C - 0 bonds, whereas an electrophilic (low in electron density) oxygen species would tend to attack the region of high electron density of the molecule (such as C = C bonds) leading to breaking of the carbon skeleton and eventually to degradation products. In addition to the nucleophilicity and electrophilicity, whether the adsorbed hydrocarbon species is situated favorably to react with the surface lattice oxygen should also be a factor (50). This is discussed further below. IX.

Selectivity Patterns for Mg,(V04)~, Mg~Vz07, and (VO)ZPZO~

Extensive work has been performed for alkane oxidation over a number of catalysts, including Mg orthovanadate (Mg3(V04)?), Mg pyrovanadate (Mg2V207),and vanadyl pyrophosphate ((Vo)zP207).It was found that the selectivities were very different on these catalysts, and they differed also greatly for different alkanes. Table X shows the typical product distributions over these three catalysts at low conversions of alkane. Since the data are for low conversions, it can be assumed that these are primary products, and secondary reactions are not important. It can be seen that the major products observed ranged from oxidative dehydrogenation products, to various oxygen-containing organic products, such as acids and anhydrides, to carbon oxides. Since the structure of these catalysts are well known, an attempt was made to understand the selectivity pattern with respect to the availability of reactive lattice oxygen (50, 56). These three catalytic systems share many common features. As was mentioned earlier, deuterium kinetic isotope studies have shown that the breaking of the first secondary C-H bond of butane is the rate-limiting step on (VO)2Pr07.By comparing the relative rates of oxidation of different alkanes (Table VIII), it could be concluded that the breaking of the first C-H bond was the rate-limiting step over both Mg3(V04)? and MgaVz07. Finally, the fact that pulse studies in the absence of gaseous oxygen over vanadyl pyrophosphate and Mg orthovanadate produced similar product distributions as in steady-state experiments suggested that lattice oxygen was involved in determining selectivity in these systems (34, 38). Although these experiments were conducted only over selected systems, the discussion assumes that the conclusions obtained are general and apply to all three catalysts for the alkanes under consideration. The active sides of these catalysts can be assumed to be VO, units in the structures. These VO, units are shown in Fig. 8. The active sites on Mg,(VO& are isolated V04 tetrahedra such that all the oxygen ions are bridged between a V and a Mg ion. For Mg2V?07,the active sites are vz07 units that can be viewed as pairs of corner-sharing V04 tetrahedra. For the

29

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

TABLE X Tvpicul Product Disiributions for Alkane Oxidation on V-Mg Oxide (Mg3(VO& MgO). Mg2 Vz 0 7 , and VPO (( V0)2P2 0,) Reactant

Reaction T ("C)

Alkane cow. (%)

Selectivity" (8)

Over VPO catalyst 305

3.8

300

8

300

I

325

7

CO CO, CzH4 C3H6 _ Ac 6 2 - 4 0 6

CO CO1 C3H6 - CzH4 - 16 10 2 I CO _ COZ_MA _ _ PA 22 15 19 44

Ar 2

& MA 1

2

6

4

Over V-Mg-O catalyst 540

5.2

so0

8.4

475

4.0

475

4.1

4x4

8.4

Over Mg2VzO7catalyst 540 475

3.2

CO COz CZH, 49 21 30

10

502

6.8

500

6.8

CO _ CO2_C4HS _ -

39 36 25 CO __ COz -C4HgC4Hh 33 33 31 2

"Ac, acetic acid; Ar, acrylic acid; M A , maleic anhydride; PA, phthalic anhydride.

VPO catalyst, they are assumed to be V208units made up of pairs of distorted edge-sharing VOs square pyramids. The assumption of these active sites, especially for the VPO catalyst, was discussed in detail in Ref. 56, in view of the fact that the most selective VPO catalyst for butane oxidation to maleic anhydride contained a slight excess of phosphorus over the stoichiometric ratio for vanadyl pyrophosphate, the phosphorus was concentrated on the surface (57-6Z), and the average vanadium valence of the catalyst under reaction conditions was about 4.1 (57, 58).

30

HAROLD H . KUNG

After the first C-H bond of an alkane is broken, a surface alkyl species is formed. There are at least two possible reactions for this alkyl species that lead to different products. Dehydrogenation products would be formed if the alkyl species reacts by breaking another C-H bond at the P-position: H M-0-M

I

+ *-CHZ-CHZ-R+M-O-M

+ CH*=CHR + *

(9)

In this equation, * represents a surface ion in the active site and is assumed to be a surface V ion. Alternatively, if the alkyl species reacts not only by breaking a C-H bond, but also by forming a C-0 bond, then an oxygen-containing organic product could be produced, such as the one shown in Eq. (10). HC-CHIR

I

0 M-0-M

+ *-CH2CH2R + *-+M

0M +

* + *-H

(10)

Since it is likely that the formation of a C-0 bond is irreversible, depending on whether the surface alkyl species reacts by forming a C - 0 bond or not, oxygen-containing (including organic products and COJ or dehydrogenation products would be formed. Thus one can view the reaction of the alkyl species as a selectivity-determining step, and the ease of removal of an oxygen atom from the lattice to form a C-0 bond with the surface intermediate should be an important factor in determining the selectivity of the reaction. This is represented by Scheme 11:

<

dehydrogenation products

alkane --+ alkyl(ad)

oxygen-containing products (oxygenates or combustion products)

SCHEMt:

11

Alternately, the selectivity-determining step could be the reaction of an adsorbed alkene formed from a surface alkyl (i.e., not by readsorption). In that case, dehydrogenation products are observed if the alkene desorbs, and oxygen-containing products are observed if it reacts further with a lattice oxygen, such as by insertion of a lattice oxygen into a C=C bond [Eq. (1211: H2C-CHR

\/ CH>=CHR

I

M-O-M-+*

0

I

--+

MOM

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

31

Then Scheme I1 is modified to Scheme 111: dehydrogenation products alkane

-+ alkyl(ad) + alkene(ad)

oxygen-containing products (oxygenates or combustion products) SCHEME I11

The discussion thus far suggests that the selectivity for oxidative dehydrogenation depends on the reactivity of lattice oxygen to form C-0 bonds with the adsorbed hydrocarbon intermediate. The data presented in Figs. 6 and 7 suggest that the reactivity of lattice oxygen is related to the strength of the metal-oxygen bond, which can also be viewed in terms of the reducibonds. For the three catalysts bility of the metal cations in the M-0-V under consideration, the structures of the active sites (Fig. 8) would suggest V bonds (Mg2V207and (VO)zP2O7)would that solids that contain V-0be nonselective and produce large amounts of oxygen-containing products, whereas those that do not (Mg3(V0&) would be selective. In an attempt to explain the selectivity pattern shown in Table X, the products were grouped into dehydrogenation or oxygen-containing products. Thus the product selectivity for various alkane oxidation on these three catalysts are simplified into the pattern shown in Table XI. When the data in this table were examined, it was apparent that although ease of removal of lattice oxygen (or reducibility of the cations) explains most of the data, it is insufficient to explain all of them. Mg,(VO& is indeed a selective dehydrogenation catalyst for most of the alkanes studied, whereas Mg2V207 and (VO)2P207produce mostly oxygen-containing products. However, contrary to expectation, propane reacts on Mg2V207with a high dehydrogenaTABLE X I SimpliJied Pattern of Produci Disiribution for Alkane Oxidation on V - M g - 0 , Mgz Vz 0 7 , and V-P-0 Dominant product" Reactant

V-Mg-0

MgzVz07

Ethane Propane 2-Methylpropane Butane Pentane Cyclohexane

ox

ox

Deh Deh Deh Deh

Deh

ox ox -

v-P-0 Deh

ox

ox

ox

ox ox

"Ox, oxygen-containing organic products and CO,; Deh, dehydrogenation products.

32

HAROLD H . KUNG

tion selectivity, whereas ethane reacts with high dehydrogenation selectivity on (VO)2P207,but with mostly combustion on Mg3(V04)2. It became apparent that another condition is needed to explain a broader range of data, in addition to the necessary condition of possessing readily removable lattice oxygen in the active site: that the formation of oxygencontaining products be enhanced if the hydrocarbon intermediate in the selectivity-determining step can be bonded to the two vanadium ions of the linked VO, units such that the hydrocarbon species is being held close to the reactive surface lattice oxygen. This latter requirement could be met only if the molecule is sufficiently large to do so. Whether this is satisfied or not can be examined by comparing the molecular size with the separation of the vanadium ions in the solid derived from crystallographic data. For the V207 unit in Mg2V207,the separation is 0.339 nm (62); for the V20s unit in (vo)2P2o7, it is 0.319 nm (63). The separation of V ions in adjacent VO, units in Mg3(V04)2on a low index plane is 0.37 nm (40). Using a value of 0.072 nm for the ionic radius of Vs+ (64) and a covalent radius of 0.07 nm for C (65), the V-C bond length is estimated to be 0.14 nm. Using this V-C bond distance, the V-V separation in an active site needed to bond with a surface 1,2-diadsorbed C2 and a 1,3-diadsorbed C3 species can be estimated to be 0.244 mm and about 0.327 nm, respectively. From these values, it can be seen that a C3 species (such as a 1,3-diadsorbed allylic species) could readily bond with the two V ions in the active site of (vo)2P207, but could do so only with difficulty with those in Mg2V207. Thus one would argue that propane would react on Mg2V207as if the active sites are isolated V04 units, like in Mg3(V04)2,whereas it would react like the large hydrocarbons on ( v o ) 2 P 2 0 7This . is indeed observed as is shown in Table XI. The manner in which a 1,3-diadsorbed Ca carbon chain is expected to bond to the two vanadium ions in the active site could explain the much larger production of acetic acid from 2-methylpropane than from propane on (vo)2P?07 (see Table X) (56). As shown in Scheme IV, when a 2methylallyl species bridges the adjacent VOs units in (VO)2P207, the methyl group in the middle carbon is not interacting with the surface. This configuration is conducive to the formation of acetate group when the two C-C bonds of the C3 unit are cleaved by reaction with the lattice oxygen. This may explain the substantial production of acetic acid observed on this catalyst. Carbon atoms in i-C, unit

0 v ion

0o

ion

33

OXIDATIVE DEHYDROGENATION OF LLGHT ALKANES

The difference between propane and 2-methylpropane on MgzVz07cannot be explained by the size of the C1 chain. It is then proposed that although with difficulties, a C7 species still has a finite probability to interact with both vanadium ions in a VzO7unit in the catalyst to lead to the formation of combustion products. This probability is twice as large for 2methylpropene (or 2-methylpropyl) than propene (or propyl), which may account for the lower selectivity for dehydrogenation for 2-methylpropane on this catalyst. The additional requirement of the size of molecule with respect to the V-V distance in the active site is perhaps the reason behind the fact that propane and butane show not only different selectivity behavior, but also different dependence of the selectivity on the reducibility of the catalyst: the selectivity for dehydrogenation in butane oxidation decreases rapidly with increasing reducibility of the catalyst (Figs. 6 and 7), but the selectivity in propane oxidation is much less dependent on it (31). The behavior of ethane is different from the other alkanes. It is the only alkane that undergoes significant dehydrogenation on the VPO catalyst, as well as the only one for which combustion is the predominant reaction on V-Mg-0. An ethyl species is too small to interact with two V ions simultaneously on any of the three catalysts. A phenomenological explanation of this behavior of ethane has been suggested (56). In this explanation, the possible reactions of ethyl, propyl, and 2-methylpropyl species were compared by statistically counting the number of various types of bonds in each species: C@Hi

H

I

I

C@H3

I

CpH3- c"- CPH7

CPH3-C" --OH,

C"H2

I *

I

I

*

In these species, a reaction in which a CP-H bond is broken would lead to dehydrogenation. However, breaking a C"-H bond or cleaving a C"-Ccp bond would lead to degradation products. The statistical probabilities of these three processes are proportional to the number of these bonds in the species, which are shown in Table XII. They show that these species TABLE XI1 Number of' Vurious Types of Bonds in an Alkyl Species Number Species Ethyl Propyl 2-Met hylpropyl

Normalized value

CO-H

C"-H

c-c

CO-H

C"-H

c-c

3 6 9

2 I

1

0

3

6 6 6

4

2

2 2 2

1

0

34

HAROLD H . KUNG

only differ in the relative number of CP-H bonds. If the C"-H and the C-C bond react with equal probability, whereas the C"H bond reacts somewhat faster, combustion would be more likely for ethane than for the other alkanes. Since the reaction conditions, especially temperatures, for the data on Mg2V207and Mg3(V0& were similar, this argument would account for the low dehydrogenation selectivity observed on these two oxides. For the VPO catalyst, for some unknown reason, which may be related to the much lower reaction temperature, the Cp-H bond breaks much more readily than the C"-H bond, and dehydrogenation becomes the dominant reaction. For the higher alkanes, the formation of 1,3-diadsorbed species is very favorable that CP- H bond breaking and dehydrogenation is not an important reaction any more. X.

Conclusion and Perspective

This paper summarizes the data and current understanding regarding the oxidative dehydrogenation reaction of alkanes. The reaction mechanism, the nature of the catalysts, and factors that determine selectivity for dehydrogenation versus formation of oxygen-containing products are discussed. From the pattern of product distribution in the oxidation of C2 to C6 alkanes obtained with supported vanadium oxide, orthovanadates of cations of different reduction potentials, and vanadates of different bonding units of VO, in the active sites, it was shown that the selectivities can be explained by the probability of the surface alkyl species (or the surface alkene formed from the alkyl) to react with a reactive surface lattice oxygen. Catalysts for which this occurs with a high probability would show low selectivities. This probability increases for vanadates that have low heats of removal of lattice oxygen, which are those that contain easily reducible cations in the active sites and for vanadates whose active sites can bind the surface alkyl species (or alkene) in a way that bring the surface intermediate close to the reactive lattice oxygen. The dependence of selectivity for dehydrogenation on the conversion of alkane shows that for the more selective catalysts known, the reaction proceeds with a sequential mechanism. The first step of the reaction is the breaking of a C-H bond of the alkane molecule, which is also the ratelimiting step. For these more selective catalysts, alkene is the primary product. Therefore, high selectivities can be obtained at low conversions. However, as the conversion increases, the selectivity decreases because of the secondary reaction of the alkene. The rate constant for the reaction of the alkene on the most selective catalyst is still about the same in magnitude as the rate constant for the activation of alkane. It is larger for the less selective catalysts. Thus the maximum yield of alkene among the catalysts known to date is still less than about 35%. To improve this yield, catalysts that react with alkene less rapidly than with alkane need to be found.

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

35

The incentive for using alkanes as a potential feedstock is because of their low prices. Among the light hydrocarbons, the price of alkane is about half the price of alkene. To make the use of alkane economically attractive compared to the use of alkene, the operation must not be more expensive than this price difference. In general, however, even if the alkane oxidation reaction proceeds with the same selectivity on a carbon basis as alkene oxidation, the reaction using alkane will consume more oxygen because it contains a larger number of hydrogen atoms. In addition, a larger gas throughput will be involved, which adds to the equipment and operating cost. Therefore, based on the current knowledge, the driving force behind the use of alkane is probably not due to its price advantage over the corresponding alkene, but due to other factors, such as environmental concerns or shortage of alkenes. This situation would change, of course, if catalytic knowledge eventually leads to the discovery of a highly selective catalyst for alkane conversion. ACKNOWLEDGMENTS The portion of this work on vanadates has been supported by the Department of Energy, Basic Energy Science, Division of Chemical Sciences and that on the heats of reoxidation by the National Science Foundation. In addition, among the students involved in the various aspects of this project, P. Michalakos acknowledges support by Battelle’s NASA Advanced Materials Center for the Commercial Development of Space, and L. Owens acknowledges fellowship support from the 3M Company and a Faculty Minority Internship from the Monsanto Company. IN PROOF NOTEADDED

After this paper was accepted for publication in November, 1992, a number of reports have appeared that deal with the subject of oxidative dehydrogenation of light alkanes. The effect of the structure of vanadia on a support has been investigated for the oxidation of butane [87] and propane [88-901. The evidence supports the concepts that the bridging oxygen in V-0-V plays an important role in the oxidation reaction [87, 901. The data also show that vanadia species of different structures on these supports have different catalytic properties, and that isolated V04 units are the most selective [91]. A number of mixed oxides have been reported to be active catalysts for oxidative dehydrogenation. They include Mg-V-Mo oxide [Y2], pyrophosphates of Zn, Cr, and Mg [93], vanadia-niobia [94], Fc in MFI-type silicate (95, 961, and Mo-V-Nb oxide [97]. It has also been reported that the use of very short contact time at high temperatures can result in high yields of dehydrogenation products [98], Also, at the high temperatures, homogeneous (vapor phase) reaction is a significant contribution to the overall reaction, and the selectivity for dehydrogenation products in the homogeneous reaction is as high as that of the best catalyst 199, 1001. The latter confirms the results of ref. 28. The role of acidity on the oxidative dehydrogenation reaction was also investigated [lol]. REFERENCES

I . Emig, G . , and Martin, F., Caraly. Today 1, 477 (1987). 2 . Brooks, B. T., Ed., “The Chemistry of Petroleum Hydrocarbons,” Reinhold, New York, 1954. 3. Satterfield, C. N., Heterogeneous Catalysis in Practice,” McGraw-Hill, New York, 1980.

36

HAROLD H. KUNG

4 . Hucknall, D. J., “Selective Oxidation of Hydrocarbons,” Academic Press, London, 1974. 5 . Magistro, A. J . , US Patent 5,087,791 (1992). 6. Hucknall, D. J., “Chemistry of Hydrocarbon Combustion,” Chapman & Hall, London, 1984. 7. Westbrook, C . K., and William, J. P., Comhust. Sci. Techno/. 37, I17 (1984). 8. Martin, R., Niclause, M., and Scacchi, G . , “Industrial and Laboratory Pyrolysis” L. F. Albright and B. L. Crynes. Eds.), p. 37. ACS Synip. Ser. No. 32, Am. Chem. Soc., Washington, DC, 1976. Y. Selwitz, C. M., and Stanmyer, J. L. Jr.. P r e p . Am. C h m . Soc. Div. Per. Chem., 6 (3B). 141 (1961). 10. Boguslavskii, E. A , . and Yukel’son, I . I . . Chem. Absrr. 64, 15639d (1966). 11. Boutry, P., Daumas. J . C., and Montarnal, R., C. R . Auld. Sci. Pork Ser. C 264, 81 (1967). 12. Netherland Patent 6,515,741 (1966); assigned to Halcon International, Inc., Chrm. Ahsrr. 65; 13547a (1966). 1 3 . McDonald. W. R., and Mclntyre, A . D., Belgium Patent 613,835 (1962); Chrm. Absrr. 58; 2368g (1963). 14. Eastman, A . D., Guillory, J. P., Cook, C. F., and Khible, J. B., US Patent 4,497,971 (19x5). 15. Ward, M. B., Lin, M . J., and Lunslord, J. H., J . Curd. 50, 306 (1977). 16. Erdohelyi, A,. and Solymosi, F., Appl. Cutul. 39, LI 1 (1988). 17. Morales, E., and Lunsford, J. H . ,J . Curd. 118, 255 (1989). 18. Memuki, M., Taouk, B., Tessier, L., Bordes, E., and Courtine, P., in Proceedings, 10th International Congress on Catalysis, Budapest, 1992” (L. Guczi, F. Solymosi, and P. Tetenyi, Eda.), Vol. A , p. 753. Institute of Isotopes of the Hungarian Academy of Science, Budapest, 1993. 1 Y . Erdohelyi. A., and Solymosi, F., J . Curd 129, 497 (1991). 20. Oyama, S. T., ./. Curd 128, 210 (1991). 21. Bernal, S., Martin, G. A., Moral, P., and Perrichon, V., Curul. L r f t . 6, 231 (1990). 22. Bernal, S . , Botana, F. J., Laachir, A., Moral, P., and Martin, G. A,, Euro. J . Solid Srurr Inorg. C h m . 28 (suppl.), 421 (1991). 2-{. See, e.g., Eastman, A. D., and Kolts, J. H . , US Patent 4,370,259 (1983); Eastman. A . D., US Patent 4,396,537 (1983). 24. Centi, G., Trifiro, F., Busca, G . , Ebner. J . , and Cleaves, J . . in Proceedings, 9th International Congress on Catalysis, Calgary, 1988” (M. J. Phillips and M. Ternan, Eds.), Vol. 4,p. 1538. Chem. Institute of Canada, Ottawa, 1988. 25. Kung, M. C., and Kung, H. H . , J . Cord. 134, 688 (1992). 26. Patel, D., Kung, M . C., and Kung, H. H., in “Proceedings, 9th International Congress on Calalysis Calgary, 19x8’’ (M. J. Philips and M. Ternan, Eds.), Vol. 4, p. 1554, Chem. Institute of Canada, Ottawa, 1988. 27. Chaar, M. A., Patel, D., and Kung, H. H . , J . Curd 109, 463 (1988). 28. Nguyen, K . T., and Kung, H. H., J . Cutul. 122, 415 (1990). 2Y. Seshan. K . , Swaan, H. M.. Smits, R . H. H., vanommen, J. G., andRoss, J. R. H., in “New Developments in Selective Oxidation” (G. Centi and F. Trifiro, Eds.), p. 505. Elsevier, Amsterdam, 1990. 30. Sam, D., Soenen, V.. and Volter, J . C., J . Cutul. 123, 417 (1990). 31. Guerrero-Ruiz, A , , Rodriguel-Ramos, I., Fierro, J. L. G., Soenen, V., Herrmann, J. M., and Volta, J. C., in “New Developments in Selective Oxidation by Heterogeneous Catalysis” (P. Ruiz and B. Delmon, Eds.), p. 203. Elsevier, Amsterdam, 1992. 32. Nguyen, K. T., and Kung, H. H., Ind. Eng. Chem. Res. 30, 352 (1991). 33. Kung, M. C., and Kung, H. H., J . Card. 128, 287 (1991).

OXIDATIVE DEHYDROGENATION OF LIGHT ALKANES

37

34. Pepera, M. A., Callahan, J. L., Desmond, M. J . , Milberger, E. C . , Blum. P. R., and Bremer, N. J., J . Am. Chem. Soc. 107, 4883 (1985). 35. Chaar, M. A., Patel., D., Kung, M . C., and Kung, H. H., J . Catul. 105, 483 (1987). 36. Owen, 0. S . , Kung, M. C., and Kung, H. H., Catal. Leat. 121,45 (1992). 37. Patel, D . , Andersen, P. J., and Kung, H. H . , J . Cutd. 125, 132 (1990). 38. Owen, 0. S . , PhD thesis, Northwestern University, 1992. 39. Owen, 0 . S . , and Kung, H. H . , J . Mol. C ~ t d .79, , 265 (1993). 40. Krishnamachari, N., and Calvo, C . , Can. J . Chetn. 49, 1629 (1970). 4 / . Andersen, P. J . , and Kung, H. H., J . Phys. Chem. 96, 31 14 (1992). 42. Andersen, P. J . , and Kung, H. H., in “Proceedings, 10th International Congress on Catalysis, Budapest, 1992” (L. Guczi, F. Solyrnosi, and P. Tetenyi, Eds.), Vol. A, p. 205. Institute of Isotopes of the Hungarian Academy of Science, Budapest, 1993. 43. Vrieland, G. E., and Murchison, C . B., US Patent 4,973,791 (1990). 44. Shapovalova, L . P., Luk’yanenko, V. P., Doroshenko, V. A , , Svintsov, N. I., Solodkaya, V. S . , Dvoretskii, M. L . , Kine?. Cutul. 26, 424 (1985). 45. Kijenski, J . , Baiker, A , , Glinski, M., Dollenmeier, P . , and Wolkaun, A , , J . Cutul. 101, l(1986). 46. Oyama, S . T., Went, G . T., Lewis, K . B., Bell, A . T., and Somorjai, G. A , , J . Phys. Chem. 93, 6786 (1989). 47. Eckert, H., and Wachs, I. E., J . Phys. Chem. 93, 6796 (1989). 48. Roozeboom, F . , Mittelmeijer-Hazeleger, M. C . , Moulijn, J. A , , Medema, J., de Beer, V. H. J., and Gellings, P. J., J . Phys. Chem. 84, 2783 (1980). 49. Deo, G., and Wachs, 1. E., J . Cural. 129, 307 (1991). 50. Michalakos, P., Kung, M. C., Jahan, I . , and Kung, H. H., Prep. Amer. Chem. Soc. LXv. Petrol Chem., 37, 1201 (1992). 51. Morooka, Y., Morikawa, Y., and Ozaki, A , , J . Cutul. 7, 23 (1967); 5, 116 (1966). 52. Sachtler, W. M. H.. Dorgelo, G . J . H., Fahrenfort, J.. and Voorhoeve, R. J . H., Rec. Truv. Chim. Pays-Bas 89, 460 (1970). 53. Callahan, J. L., and Grasselli, R . K . , AIChE J . 9, 755 (1962). 54. Bielanski, A., and Haber, J., “Oxygen in Catalysis.” Dekker, New York, 1991. 55. Haber, J . , in “Solid State Chemistry in Catalysis” R. K. Grasselli, and J. F. Brazdil, Eds.), Am. Chem. Soc. Symp. Series No. 279, p. I . Am. Chem. Soc., Washington, DC, 1985. 56. Michalakos, P., Kung, M. C., Jahan, I . , and Kung, H. H . , J . Catal. 140, 226 (1993). 57. Centi, G., Trifiro, F., Ebner, J . , and Franchetti, V. M., Chem. Rev. 88, 55 (1988). 58. Batis, N. H . , Batis, H., Ghorbel, A., Vedrine, J. C., and Volta, J. C . , J . Catul. 128, 248 (1991). 59. Wenig, R. W., and Schrader, G. L., Ind. Eng. Chem. Fundam. 25, 612 (1986). 60. Garbassi, F., Bart, J. C. J . , Tassinari, R . , Vlaic, G . , and Lagarde, P., J . Catal. 98, 317 (1986). 6 / . Morishige, H . , Tamaki, J., Teraoka, Y., Miura, N., and Yamazoe, N., J . Chem. Soc. Jpn. 113, 1983 (1989). 62. Gopal, R . , and Calvo, C., Acra. Crystallogr., Secr. B: Struct. Sci. 30, 249 1 (1974). 63. Gorbunova, Yu. E . , and Linde, S.A., Sov. Phys. Dokl. 24, 138 (1979). 64. Shannon, R . , Actu. Crystallop., Sect. A: Found. Crystallogr. 32, IS1 (1976). 65. Atkins, P. W., “Physical Chemistry,” 3rd ed., W. H. Freeman, New York, 1986. 66. Swaan, H., Toebes, A., Seshan, K., van Ommen, J. G., and Ross, J. R. H., Catul. Today 13, 629 (1992). 67. Eastman, A. D., and Kimble, 1. B., US Patent 4,450,313 (1984). 68. Velle, 0. J . , Andersen, A., and Jens, K. J . , Curd. Todtcy 6, 567 (1990). 69. Murakami, Y., Ostuska, K., Wada, Y., and Morikawa, A , , Bull. Chem. Soc. Jpn. 63, 340 (1990).

38

HAROLD H . KUNG

70. Thorsteinson, E . M., Wilson, T. P., Young, F. G., and Kasai, P. H., J . C u r d . 52, I16 (1978). 71. Han, Y., Zou, Z., Lu, H., and Hui, C., Chem. Abstr. 115; 256685 (1992). 72. Burch, R., and Swarnakar, R., Appl. Curul. 70, 129 (1991). 73. Tomezsko, E., US Patent 3,784,485 (1975). 74. Bellussi, G., Centi, G., Perathoner, S . , and Trifiro, F., Prepr. A m . Chem. Soc. Div. P e t . Chem., 37, 1242 (1975). 75. Huang, Z., Yang, J., and Zhang, G., Chem. Abstr. 110, 154918 (1989). 76. Ushkov, S., Osipova, Z., Sokolovskii, V., and Ketchik, S., Kinet. Curd. 29, 195 (1988). 77. Hardman, H. F., US Patent 4,131,631 (1978). 78. Komatsu, T., Uragami, Y., and Otsuka, K., Chem. Lerr., 1903 (1988). 79. Kim, Y. C., Ueda, W., and Morooka, Y., in “New Developments in Selective Oxidation” (G. Centi and F. Trifiro, Eds.), Studies in Surface Science and Catalysis, Vol. 5 5 , p. 491. Elsevier, Amsterdam, 1992. 80. Smits, R. H. H., Seshan, K., and Ross, J. R. H., J . Chem. Soc. Chem. Commun., 558 (1 991). XI. Andersson, A., Andersson, S . L. T., Centi, G . , Grasselli, R. K., Sanati, M., and Trifiro, F., in “Proceedings, 10th International Congress on Catalysis, Budapest, 1992” (L. Guczi, F. Solymosi, and P. Tetenyi, Eds.), Vol. A, p. 69. Institute of Isotopes of the Hungarian Academy of Science, Budapest, 1993. 82. Bertus, B. J . , US Patent 4,094,819 (1978). 83. Bertus, B. J., US patent 3,886,090; 3,886,091 (1975). 84. Ripley, D. L., US Patent 4,044,066 (1977). 85. Walker, D. W., Hogan, R. J., and Farha, F., US Patent 4,218,343 (1980). 86. Owens, L., and Kung, H. H., Prepr. Am. Chem. SOC. Div. Per. Chem. 37, 1194 (1992). 87. Owens. L. and Kung, H. H., J . Cural.. 144, 202 (1993). 88. Corma, A., Lopez Nieto, J. M., Paredes, N., Perez, M., Shen, Y., Cao, H., and Suib, S. L., New Develop. Select. Oxidation Heterog. Catal. (Stud. Surf. Sci. Catal., 72). p. 213 (1992). 89. Corma, A., Lopez Nicto, J. M., Paredes, N . , and Perez, M., Appl. Card A, 97, 159 (1993). 90. Eon, J. G., Olier, R., and Volta, J. C., 1.Cutul., 145, 318 (1994). 91. Corma. A.. Lopez Nieto, J. M., and Paredes, N., J . C a r d . , 144, 425 (1993). 92. Harding, W. U . , Kung, H. H., Kozhevnikov, V. L., and Poeppelmeier, K. R., J . Curd., 144, 597 (1993). 93. Takita Y., Kurosaki, K., Mizuhara, Y., and Ishihara, T., Chem. L e t t . , p. 2335 (1993). 94. Smit, R. H. H., Seshan, K., Leemreize, H., and Ross, J. R. H., CutaI, Today, 16, 513 (1993). 95. Uddin. M. A . , Komatsu, T., and Yashima, T., Chem. L e f t . , p. 1037 (1993). 96. Bellussi, G., Centi, G., Perathoner, S., and Trifiro, F., in “Catalytic Selective Oxidation” (ACS Symposium Series, No. 523), p. 281 (1993). 97. Desponds, O . , Keiski, R . L., and Somorjai, G . A., Card. L e t t . , 19, 17 (1993). 98. Huff, M., and Schmidt, L. D., J . Phys. Chem., 97, 11815 (1993). 99. Burch R . , and Crabb, E. M . , Appl. C u r d . A , . 100, 11 1 (1993). 100. Burch R., and Crabb, E. M., AppI. Curul. A. 97, 49 (1993). 101. Le Bars, J . , Vedrine, J. C., Auroux, A , , Trautmann. S . , and Baerns, M., Appl. Card A, 88, 179 (1992).

ADVANCES IN CATALYSIS, VOLUME 40

Catalysis in Coal Liquefaction ISAO MOCHIDA AND KINYA SAKANISHI lnstitute of Advanced Material Study Kyushu University Kasuga, Fukuoka, 816 Japan

1.

Introduction: History and Status of Coal Liquefaction

Bergius was the first to use coal liquefaction to obtain petroleum substitutes. He applied very high hydrogen pressures (-800 atm) and used inexpensive iron catalysts because they were disposed following the procedure ( I ) . During the second world war, the large-scale Bergius coal liquefaction process was extensively investigated in Germany because of the limited supply of petroleum; the cost of the fuel was ignored, although the liquefaction process was not competitive (2). Once a vast supply of petroleum was discovered, principally in the Middle East, coal liquefaction, which could not compete economically, was nearly forgotten, although basic research continued in the United States. The two recent oil crises both of which were provoked politically and economically by world reliance on oil supplies from the Middle East, created a resurgence in the study of coal liquefaction as a substitute for petroleum. However, petroleum prices are subject to change due to political factors, and today coal liquefaction must compete economically with petroleum production and refinement if it is to be a viable substitute. Although the oil crises had no direct connection to limited petroleum resources, oil shortages will occur early in the 21st century because of the present demands by developed countries and rapidly growing demands of developing countries with large populations. Coal liquefaction is expected to be a major source of liquid fuel and economical feasibility will be very important if this is to be realized. Coal liquefaction research after the second world war focused on moderating reaction conditions in terms of temperature and pressure in order to lower construction costs of the coal liquefaction facility. At present, typical conditions are 150-200 atm and 300-480°C. Moderation was achieved primarily through improvements in catalysts and donor solvents and through 39

Copyright 0 1994 by Academic Press, Inc. All rights of reproduction in any form reserved.

40

ISAO MOCHIDA AND KlNYA SAKANlSHl

new reactor configurations. Oil yield is sometimes sacrificed ( 3 ) , but the most recent processes produce fuels in higher yields than the processes developed during World War 11. Nevertheless, we have not yet developed a liquefaction process that can supply fuel at a price competitive with the present price of oil. Researchers in several countries are searching for further improvements. New catalyst and process configurations are the keys to solving the following problems. 1 . The limited activity and difficulty in recovering the catalyst force a large amount of catalyst to be wasted as it is mixed with minerals and organic residues derived from coal. High dispersion of the catalyst on the coal can significantly reduce its amount; this is rather costly and repeated use of the catalyst is not possible. 2. Large amounts of coal liquids are wasted because of the difficulty in recovering oil from solid waste. 3. The incomplete conversion of coal causes excessive amounts of organic residue to be wasted. Although the residue can be a source of fuel or hydrogen via gasification, it is far inferior to the parent coal. 4. The low yield of oil and the coalholvent ratio in slurry are important factors that influence the cost of coal-derived oil. 5. Oil products require further refinement if they are to be blended with petroleum products to meet current specifications and more stringent environmental regulations. 6. The price of hydrogen greatly influences the cost of coal-derived oils. Liquefaction processes should minimize the consumption of hydrogen, and residue should not be used for gasification. Reforming of hydrocarbon gases for hydrogen production is economical compared to gasification of liquefaction residues.

Multistage coal liquefaction has been proposed to consist of the following stages, as shown in Fig. 1 [this scheme is the basis for promising processes

12) (3) PRIMARY LHYDROgLIQUEFAC- T G E N A T I O N

TION

FIG. I .

I

I

-

T(4)

EXTRACTION 4

I

~

~

~

REFINING

Multistage coal liquefaction scheme.

-

~

~

I

N

G

I I

41

CATALYSIS IN COAL LIQUEFACTION

of current intensity ( 4 ) ] : I . coal pretreatment; 2. coal dissolution; 3 . catalytic hydrocracking; and 4. catalytic upgrading and solvent regeneration. Solid/liquid separation is important at all stages and depends on the extent of coal cleaning during pretreatment, depolymerization of coal macromolecules at the dissolution stage, and catalyst durability through the catalytic process. Hence, the catalysts and catalyses of coal liquefaction are reviewed with reference to the above stages.

II. Coal Structure and Reactivity

Coal structural studies have long been used to understand the chemical basis of coal liquefaction. However, no single structure actually exists; average structures are commonly used to provide guidance to understanding the chemical transformations involved during coal conversion. Recent studies provide quantitative images that are valuable in the design of each step of coal conversion. Shinn ( 5 ) drew models of macromolecules present in a bituminous coal based on a series of products analyzed at each step of a three-stage liquefac4 ring tion sequence, as shown in Fig. 2, where polyaromatic rings ( I units) with some naphthene rings, heterocycles, alkyl side chains, and oxygen subunits are connected by methylene, ether, thioether, and aryl-aryl linkages (sp2-sp *) to form polyaromatic-polynuclear macromolecules of variable molecular wieghts. These macromolecules include rather small molecules of a few hundred molecular weight that can be extracted by ordinary solvents. Recent studies emphasize noncovalent bonds joining constituent macromolecules through their polar groups, such as hydrogen bonds, charge-transfer interactions, cationic ion bridges, and layered stacking aromatic planes, which are responsible for low solubility, facile coking of coal components, and limited access to solid catalysts (6-8). The structure of coal is known to be dependent on rank or the degree of coal weathering and the distribution of macerals. Figure 3 illustrates Hirsch’s classic model of coal ranking ( 9 ) . Fewer and smaller aromatic rings, more alkyl- and oxygen-containing groups, and larger molecular weights are characteristic of lower rank coals. Recently, the structures proposed for lower rank coals emphasize their hydrogen bonding, charge-transfer interactions, and ion bridges between large macromolecules (10). Higher ranking coal structures emphasize large aromatic planar structures, which form stacked layers typically observed in graphite (11).The molecular size of the coal structure units decreases with increasing rank to a minimum at a bituminous coal rank of -83% carbon and then increases again to anthracite, which has a graphite-like structure. Thus, the highest solubility in

-

FIG.2 . Model structure of bituminous coal by Shim

open structure

,*".I

I+nm

89%

I

--

--- --

c

-0.8nrn

liquid structure

e / T -

\\

94% c

anthracitic structure

/

FIG. 3 . Hirsch model of coal ranks.

CATALYSIS IN COAL LIQUEFACTION

43

conventional solvents is observed for bituminous coals with this rank (-83% carbon). Recently, Iino et af. reported remarkably high solubilities of coals up to -60% in CSZ/N-methylpyrrolidonemixture, indicating strong intermolecular interactions of coal macromolecules and suggesting a limited contribution of three-dimensional covalent linkages (12, 13). The above models are representative of the active macerals, particularly vitrinite. Inert macerals, such as fusinite and micrinite, are believed to have large aromatic planar structures with fewer substituents (14) and behave similar to chars. During coal conversion, the coal structure influences both thermal and catalytic reactions. Thermal reactions of solid coals initiate the breakage of weak bonds at rates proportional to their bond dissociation energies. The radicals thus produced require stabilization by hydrogenation or addition of small molecules; otherwise the radicals couple to produce much more thermally stable bonds, which eventually leads finally to the formation of infusible and insoluble coke. Once thermal reactions begin, the coal undergoes structural changes through spontaneous bimolecular reactions between coal constituents or solvent species (15) or through catalytic reactions accelerated by added catalysts or the inherent mineral components (16). Hence, the reactivity changes with the progress of these reactions. The catalytic reactions of importance are hydrogenation, hydrocracking, heteroatom removal, acid cracking, and condensation. Catalytic activities and selectivities as well as the reactivities of coal molecules are influenced by a variety of competitive and consecutive thermal and catalytic reactions that are sensitive to various reaction conditions, such as heating rate, temperature, and hydrogen pressure. The access to or contact with the solid catalyst by the coal liquefaction intermediates is another important aspect of the catalysis involved because of the significance of interactions between solid coal or viscous coal liquids and catalysts in the initial stages of coal liquefaction. Dispersion of the catalyst on the solid coal surface or in the reaction mixture containing dissolved coal macromolecules is critical in this initial stage.

111.

Stages of Coal Liquefaction

Prior to liquefaction, coal is often washed to remove inorganic minerals and dried. This process sometimes changes the structure and assemblages of coal macromolecules, which profoundly influences the reactivity of coal as described in Section XI. In the preheater, coal with or without catalysts is rapidly heated to reaction temperature in the presence of solvent and pressurized hydrogen. Ex-

44

ISAO MOCHIDA AND KINYA SAKANlSHl

tensive decarboxylation, formation of carbonates, and dehydration take place in the preheater (17). Coal is believed to be substantially dissolved in the preheater at this stage. Rapid heating of up to several hundred degrees per minute is believed to be very essential in obtaining high oil yields and prevention of retrogressive reactions, which may take place at the same time. Catalysts are not expected to be effective in the preheater stage due to insufficient contact time. Hydrogen donor solvents play an important role in suppressing the retrogressive reactions at this stage and it is important that the capacity of the donor not be exceeded in the preheater. The amount of hydrogen consumed from solvent has been shown to be related to the heating rate. Slow heating rates allow more solvent dehydrogenation (18). It is well known that the viscosity increases very rapidly with bituminous coals that are dissolved in the solvent rapidly in the preheater. This sometimes causes problems of slurry transportation in narrow preheater tubes. The preheated coal slurry (essentially liquefied) is sent to the reactor, where thermal and catalytic cracking, hydrogenation, and hydrocracking take place. These reactions occur rather slowly because fewer reactive bonds are involved in this stage, which produces distillate range small molecules. In the earlier Bergius process, the reaction at this stage was performed under very high pressure at high temperature with disposable catalysts of low activity and was completed in a single step. Current liquefaction processes utilize two or three stages under more moderate conditions. Hydrogen donor solvents also assist in moderating the conditions required. Thus, the primary products in the first stage, together with the used solvent, are further hydrocracked and/or hydrorefined products as well as rehydrogenated solvent. Various types of feeds, distillates, nondistillable liquids free of minerals, the catalyst, preasphaltenes and unreacted coal of the first stage or whole products, including the catalyst and minerals, are charged to the second stage, depending on the liquid/solid separation procedure utilized and the durability of the catalyst in this stage. Staged heating is sometimes utilized in the first stage where the reaction temperature of each reactor is controlled separately to obtain the best oil yield with minimum formation of hydrocarbon gases and avoidance of coking (19, 20). The oil is further refined in the following stages. Such a process scheme practiced at present is called multistage liquefaction. A series of reaction temperatures is expected to improve selectively specific reactions at different temperatures in a series of consecutive reactions. Higher degrees of desulfurization and denitrogenation, longer catalyst life, less sludge formation, and higher yields of distillate are reportedly obtained by the multistage processing and refining of petroleum products (21, 22). The function of the catalysts in the various liquefaction stages are described in the following sections.

CATALYSIS IN COAL LIQUEFACTION

45

IV. Coal Dissolution, Depolymerization, and Retrogressive Reactions

The liquefaction of coal is the conversion of an ensemble of macromolecules as described above into smaller hydrocarbon molecules that are distillable. Shinn has described the changes in representative molecular structures of intermediates in the three steps of liquefaction as shown in Fig. 4 (5).The first step in the liquefaction of solid coal is the formation of liquid phase. Small molecules of the coal fuse above 350°C to form a liquid phase together with solvent (if present); some macromolecules may be dissolved in this liquid phase (fusion and dissolution mechanisms). Other molecules undergo thermal fission at their weakest bonds, such as methylene and benzylether bonds, producing fragmented radicals (23). When the radicals are capped with hydrogen from the solvent or the catalyst, they form smaller molecules that are soluble in the solvent or even fusible by themselves (first mechanism) increasing the quantity of liquid phase (24). This pyrolysis continues while the reactive bonds and stabilizing hydrogens are available. Atomic or molecular hydrogen, available in the reactor system can hydrogenate reactive sites on the aromatic rings. When the ipso-position of the strong aryl-aryl bond in the aromatic ring is hydrogenated, the bond becomes weakened and bond cleavage becomes possible via the first mechanism of depolymerization and facile stabilization (second mechanism) (25 27). Very reactive hydrogen may attack the aryl-aryl bond directly, leading to its breakage (third mechanism) (28). Aromatic rings are very stable unless they are hydrogenated to naphthenic rings, which may be thermally or catalytically cracked to open the ring (fourth mechanism) (29). Unless the fragmented radicals are stabilized, they recombine or react with other molecules, forming thermally stable bonds. Repetition of recombination reactions produces large molecules that have resistance toward depolymerization. Coking takes place when such large molecules remain at elevated temperatures for long time periods, for example, in the locations of low flow rate, such as near reactor walls, in bends of transfer lines, or on catalyst surfaces (retrogressive mechanism) (30).When radicals are trapped in the cage of coal macromolecules, such retrogressive reactions become accentuated as the radicals frequently encounter each other. The cage hinders liberation of radicals and the participation of donors. Hence, the dissolution of depolymerized coal molecules to break the cage is very important and effective in the liquefaction process (31). Strong dissociative properties of the solvent are important in minimizing macromolecular interactions of coal components or coal-derived products.

46

ISAO MOCHIDA AND KINYA SAKANISHI

Model of bituminous coal structure FIG. 4.

Coal depolymerization model of bituminous coal by Shinn.

V.

Catalysts in Liquefaction

Catalysts used in liquefaction can be classified in various ways as described below.

A.

CATALYTIC SPECIES

The most conventional catalytic material since the work of Bergius has been iron sulfide in various types. Pyrite, pyrrhotite, and various nonstoichiometric sulfides are known, and pyrrotite is postulated as the active form. Its precursors are red mud, residue of bauxite after the separation of alumina, iron ores of various sources, synthetic and natural pyrites, fine iron particles, iron dust from converters, iron sulfate, iron hydroxide, etc. (32, 33).

Short-contact time liquefaction products from bituminous coal

Two-stage products FIG.4. (conrinued)

48

ISAO MOCHIDA AND KlNYA SAKANlSHl

The next most widely used materials are Co-Mo and Ni-Mo sulfides, which have been widely used in petroleum refineries. They are usually supported on alumina of designed pore structures in which the pore diameter is usually larger than that for conventional petroleum residue (34, 35). A third type of material is the chlorides of transition metals, such as ZnClr and SnC14 (36, 37). This group of catalysts works in molten state in contrast to the solid state of the previous two groups. The corrosive nature and instability may excludes their practical application. No details are reviewed here. Ru has been used as an additive to Co-Mo and Ni-Mo (38) to improve their hydrogenation and denitrogenation activities. Hydrogen sulfide in the reaction atmosphere has been reported to accelerate liquefaction directly, in addition to controlling the extent of sulfiding of iron, Ni-Mo, and Co-Mo catalysts (39, 40). Recently, carbon black was reported to catalyze coal liquefaction (41, 42); this may initiate radical reactions of bond breakage.

B. PREPARATION Solid liquefaction catalysts have been prepared by three procedures. 1.

Fine Powder Cutulysts

Most iron catalysts are used in powdered form. Since their particle size strongly influences their activity, fine powders are preferred. Natural products are ground extensively. Magnetite for the magnetic tape is needle-like crystal of which the diameter is less than 1 pm. Recently, ultrafine powders (nanometer to tens-of-nanometer size scale) of iron oxides and sulfides have been prepared by means of vapor-phase hydrolysis of volatile compounds in a hydrogen-oxygen flame to produce nanometer-sized iron oxides (aerosol) (43, 4 4 ) ; rapid thermal decomposition of solutes (RTDS), such as Fe(N03)-’ solutions (45); laser pyrolysis of Fe(CO)s and C2H4to produce iron carbides followed by in situ sulfidation (46); precipitationkrystallization sequence from the sulfated and oxyhydroxides of iron (47); and a chemical reduction or an exchange/replacement reaction of iron salts solubilized in inverse micelles of reaction media (48). Finer powders of the iron sulfide are expected to be expensive as well as active. The cost/performance is carefully evaluated. 2. Supported Catalysts As mentioned previously, Co-Mo and Ni-Mo sulfides are usually supported on alumina. Selection of the specific alumina is conventionally stud-

CATALYSIS IN COAL LIQUEFACTION

49

ied on the basis of the pore size distribution and acidic characteristics. The supporting procedure and the amount of supported sulfides are very influential in catalyst activity. Alternative supports to alumina are the focus of current research. Titania and carbon have recently been examined as supports for iron and Ni-Mo sulfides (49, SO). Bifunctional and strong interactive roles of the support should be emphasized in addition to physical properties (51, 52). The search for additives such as phosphate and sulphate, which have been utilized for commercial CoMo and Ni-Mo base catalysts, has also been receiving much recent attention (53-S5). 3 . Highly Dispersed Catalysts on Coal

Sulfide catalysts have been dispersed directly on the coal surface. Very high dispersion on the catalyst may allow direct interactions between the catalyst and solid coal. The first application of this approach utilized molten chloride as the starting material. Later, oil- and water-soluble iron precursors were impregnated or ion exchanged onto the coal surface through the interaction with oxygen functional groups (56 -60). Recently, highly dispersed, highly active, or highly functional catalysts have been extensively investigated to reduce the amount of catalyst required for recovery and regeneration (61-68). Very fine particles of iron sulfide is one class of very promising catalysts because of lower cost and moderate activity. Presulfiding treatments for activation, ion exchange, and dispersed impregnation of catalysts or catalyst precursors are combined to enhance the catalytic activity and reduce the amount of catalyst required (69, 70). The use of highly dispersed catalysts from soluble salts of molybdenum is another approach to the reduction of catalyst amount because of their excellent activity despite their higher price. Recently, metal carbonyl compounds, such as Fe(CO)5, Ru3(CO),?, and MO(CO)~have been investigated as metal cluster catalysts. Preparation involved their deposition and decomposition on catalyst support surfaces (71 -73). It has been reported recently that highly dispersed catalyst on coal grains can accelerate the liquefaction of the coal grains without supporting catalysts (56, 60). The fine powders of the catalyst are indicated to be mobile during the liquefaction, suggesting no importance of direct interaction. Finer powders may be the key. Recoverable catalysts also offer a promising way to economize the cost of liquefaction catalysts (74, 75). Dow designed a process that utilized fine powders of MoS2 that were reported to be recoverable by hydroclone; however, specific details have not been published (76).

50

ISAO MOCHIDA AND KINYA SAKANISHI

c . CHEMICAL FUNCTIONS OF CATALYSTS Catalysts in liquefaction accelerate the reactions of hydrogenation, cracking, hydrocracking, and heteroatom removal (77-79). Because the coal is composed of a large variety of species and functionalities, intermediates should be well characterized at each step of the liquefaction reaction to clarify their adsorption and reactivity with the catalyst surface. Reactive species for hydrogenation can be radicals, olefins, and aromatic rings of various sizes with alkyl substitutents. Iron sulfide can hydrogenate olefins and radicals rather well but have low effectiveness for aromatic rings at the hydrogen pressures currently being investigated. Cracking involves aromatic dealkylation and cracking of paraffins, methylene linkages, and naphthenic rings. Aromatic dealkylation is rather easy under current liquefaction conditions (below 450°C); however, the cracking reactions are not facile. Competitive reactions of various species should be carefully considered in catalyst design. Heteroatoms, principally 0, S, and N, in aromatic rings require Ni-Mo or Co-Mo catalysts for their extensive removal to the levels experienced in petroleum refinement, where these constitutents are associated with aliphatic moieties that are easily removed thermally as well as catalytically under hydrogen pressure.

D. TYPESOF CATALYST USES Catalysts in coal liquefaction are used in moving-bed, ebulating-bed, and fixed-bed processes. Disposable iron catalysts must be used in moving beds. More expensive Co-Mo and Ni-Mo catalysts are used in either ebulating or fixed beds, and catalyst deactivation rates and ultimate lifetime are of concern (80, 8Z).In ebulating beds, a small portion of fresh catalyst is continuously fed to balance the catalyst being purged. Iron and chloride catalysts are basically disposable because they are considered to be rather cheap and difficult to recover from residual products, while Ni-Mo and Co-Mo catalysts are too expensive to be considered disposable (82). Recovery of very fine particles of MoS2 by hydroclone separation has been shown to be promising (83). Disposable catalysts added at levels similar to that of ash mineral contents significantly reduce the potential recovery of oil in both distillation and extraction. This is problematic because equal volumes of oil adhere to solid particles after separation. Slurry transportation of residues suffers from the same problem. Even if the cost of the disposable catalysts is affordable, adding 1 to 5% of the catalyst to the

CATALYSIS IN COAL LIQUEFACTION

51

coal feed produces tremendous amounts of waste oil in the commercial liquefaction plant. For example, 1800 3000 t of waste including ash is produced while liquefying 30,000 t of coal (usually 5% ash per day). If by pretreatment it is possible to reduce the mineral content before liquefaction, disposable catalyst streams will be the major waste stream to be dealt with. The recycle of the catalyst or drastic reduction of its amount is a major goal in current studies. Bottoms recycle of used catalyst and heavy residual products is one promising way to increase oil yield and to improve hydrogen efficiency. This benefit is limited because it may bring about reduction of coal concentration in the slurry, increase the slurry viscosity, and enhance retrogressive reactions, especially under lower hydrogen partial pressures or if poorer donor solvents are used (84). The amount of bottoms recycling is also limited by the amount of inorganic residues because of efficiency of reactor volume and enhanced corrosion.

-

VI.

Roles of Hydrogen Donor, Solvent Properties, and Catalyst in the Preheating and Primary Stages

In preceding sections, fundamental coal chemistry, liquefaction mechanisms, solvent and catalyst characteristics were summarized briefly. In the following three sections, the roles and improvements in solvents and catalysts in multistage liquefaction processes are reviewed in more detail on the basis of recent progress in this area.

OF HYDROGEN DONORS AND SOLVENTS IN A. FUNCTIONS COALDISSOLUTION

The solvent performs the following functions in coal liquefaction: 1. dissolution and dispersion of coal macromolecules; 2. hydrogen donation to radicals and aromatic rings; 3 . dissolution of depolymerized products; and 4. hydrogen shuttling . Hydrogen donors have long been recognized to be quite effective in dissolving coal. When donors of high boiling point are used, coal can be dissolved under normal pressure, and no gaseous hydrogen is necessary (85, 86). Tetralin has been used as a conventional model of donor (87, 88);

52

ISAO MOCHIDA A N D KINYA S A K A N l S H l

however, it is not the best model because of its lower hydrogen donor activity and poor solvent power. More effective model solvents have been formulated from model compound mixtures that combine high dispersion properties with hydrogen donor properties (76). The effectiveness of hydrogen donors is discussed in terms of reactivity for donation, content of donatable hydrogen, hydrogen shuttling ability, and inherent molecular structure stability. The liquefaction behavior of model donors has been determined by characterizing the products of their reaction with coal. In Figs. 5- 14, the products of liquefaction are described as gases (G), oils (O), asphaltenes (A), preasphaltenes (P), and residues (R). Figure 5 illustrates the liquefaction yields of Morwell coal with different hydrogen donors at 450°C (89). Tetrahydrofluoranthene(4HFI), which liberated all of the hydrogens in 5 min, suffered a deficiency of donatable hydrogens at the solventkoal weight ratio of 1 / 1 (S/C = 1). thus giving poor yields of oil and asphaltene. The spontaneous dehydrogenation of hexahydro-

ok-z--+o*

reaction timdmin) (c 1

reaction timebin) (d 1

FIG. 5. Liquefaction yields for donors of variable composition at 450°C and solventkoal weight ratio of unity. (a) I ,2,3, I0b-tetrahydrofluoranthene:4HFI;(b) 1,2,3,4,5,6,7,8-octahydroanthracene:8HAn; (c) I ,2,3,4,5,6,7,8,9,10,11, I2-dodecahydrotriphenylene:l2HTp; (d) I ,4,5,8,9, lO-hexahydrofluoranthene:6HAn. ( 0 )oil+asphaltene; (0)oil; (a) gas; ( 0 ) preasphaltene; residue.

(a)

CATALYSIS IN COAL LIQUEFACTION

53

anthracene (6HAn) took place so rapidly that coal fragment radicals were not capped. Octahydroanthracene(8HAn) and dodecahydrotriphenylene (I2HTp) provided more oil and asphaltene at S/C = 1, while significant amounts remained unreacted even after the longer reaction time of 15 min. Larger amounts of 8HAn (S/C = 1.5-3) proved less effective than the same amount of 4HF1 in increasing the oil yield as illustrated in Fig. 6. This suggests that the ability to release hydrogen is important to the liquefaction performance. 12HTp and 8HAn are not stable and thermally decompose at 450 and 480°C, respectively. Thus, the reactivity of donors, as well as the solvent/coal ratio and thermal stability is important for the production of oil and asphaltene. The liquefaction behavior of a mixed donor of 4HF1 and 8HAn was examined to determine the roles of donors of different quality and reactivity of donatable hydrogens in single-stage and consecutive two-stage liquefactions as illustrated in Figs. 7 and 8 (28). The two donors competed for the same radical fragments of coal (total S/C = 2, 4 5 T , 10 min). The yields of oil and asphaltene were 37 and 20%, respectively. These values were much the same as the average values of the separate use of each solvent. The consecutive use of 8HAn with eight donatable hydrogens of lower reactivity in the first stage (8HAn/coal = 1/1, 45OoC, 10 min) and 4HFl with four hydrogens of higher reactivity in the second stage provided higher yields of oil and asphaltene (41 and 19%, respectively) at less total consumption of hydrogen. 4HFl converted a larger amount of the heavier fraction in hexane insoluble(HI), which was obtained in the first-stage reaction of 4 5 0 W 10 min using 8HAn(XHAn/coal = l / i ) , into the lighter fractions(oi1 26%) in the second-stage liquefaction of 45OoC/10 min at 4HFL/HI = 1.5/1. Addi-

FIG. 6. Influence of solventkoal ratio on the liquefaction yields with different donors. Oil + asphaltene: A (4HFI), 0 (8HAn); oil: (4HFI), 0 (8HAn); ((3)gas; (a)preasphaltene; residue.

(a)

a

54

ISAO MOCHIDA AND KINYA SAKANlSHl

Yield(%)

0

50 .....

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

a b C

Yield(%) 50

0

1CO

..... .. .

h

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

. .. .. .. .

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

I

1

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

4

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

. ..... .. ..

......

d

. ....... .. .

e

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

...........

1CO

k

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

1

I t

f 9

FIG.7. donors. Reaction time (a-g); 10 min; (h-n); 20 min. (a, h) 4HFlicoal = 211; (b, i) 4HFI/ coal = I S / l ; (c, j) 8HAnicoal = 2/1; (d, k) SHAn/coal = 1 3 1 ; (e, 1) 4HF1/8HAn/ Coal = l/l/l;( f , m) 4HF1/8HAn/coal = I/OS/l;(g, n) 8HAn/4HFl/coal = l/0.5/1. (1-1) gas; (0) oil; (I-]) asphaltene; preasphaltene; (H)residue.

(a)

Yield(%)

50

0 1

1

1

I

I

100 I

I

I

I

I

-a

FIG. 8. Two-stage consecutive use of donors inthe liquefaction of Morwell coal at 450°C and 10 min in both stages. (a) First stage, 4HFl/coal = 1/1; Second stage; 4HF1(2 g) added (S/C = 211). (b-e) 1st stage, 8HAn/coal = 1 / 1 . Second stage: (b) 8HAn(2 g) added (SIC = 2/l); (c) 4HF1(1 g) added(S/C = 1 3 1 ) ; (d) 8HAn(l g) added(S/C = 1,511); (e) 4HF1(2 g) added(S/C = 211). (0) Gas; (0) oil (0) asphaltene; preasphaltene; ).( residue.

(m)

CATALYSIS IN COAL LIQUEFACTION

55

tional 8HAn in the second stage provided a lesser amount of oi1(13%) with larger amounts of residue and preasphaltene. Thus, two donors of different reactivities can behave cooperatively according to their reactivities in the depolymerization of coal molecules in the two-stage reaction, giving higher yields of oil and asphaltene. The mechanisms of fragment stabilization (first mechanism) and hydrogen-assisted bond fission (second and third mechanisms) may explain these results. Three kinds of polycondensed aromatic hydrocarbons, pyrene (Py), fluoranthene(Fl), and anthracene(An), were examined in combination with 4HF1 in hydrogen-transferring liquefaction of Morwell coal to define their roles as nondonor solvents as illustrated in Fig. 9 (31).The mixed solvent of 75% 4HF1 and 25% Py in liquefaction at 450°C for 10 min and SIC = 2 provided oil and oil asphaltene yields of 54 and 65%, respectively. This indicates the beneficial effects of mixed solvents for oil production compared with pure 4HF1. The efficiency of hydrogen consumption was also higher for the mixed solvent. Favorable effects of FL and 4HFL mixtures were observed at a higher solventkoal ratio of 3.5(S/C = 3.5), whereas the mixed solvent of 4HF1 and An failed to provide good results. The role of nondonor in hydrogen-transfer liquefaction should be considered when de-

+

LH FL ( Y o )

FIG. 9. Effect of nondonor solvents on the liquefaction. Oil + asphaltene: 0 (SIC = residue. (SIC = 3 ) ; oil: 0 (SIC = Z), 0(SIC = 3 ) ; ((3) gas; ( C ) ) preasphaltene; (0)

56

ISAO MOCHIDA AND KINYA SAKANISHI

signing the optimum liquefaction solvent for higher oil yields, based on the depolymerization mechanisms of coal macromolecules. F1, which is a polycondensed aromatic compound containing four rings, may aid the diffusion of the donor into coal particles and dissolve coalderived molecules as effectively as Py, but its ability to accept hydrogens from 4HF1 was found to be inferior to that of Py. The hydrogen-shuttling role of Py appears to be superior to that of F1.

OF COAL B. ROLESOF CATALYST IN THE DISSOLUTION

The reactive radicals produced through thermal breakage of bonds in coal molecules must be stabilized to prevent retrogressive reactions. Because of their high reactivity, their stabilization should be as rapid as possible. The bimolecular hydrogenation by donors is of fundamental importance in this stabilization as described above, although the radicals can undergo selfstabilization through liberation of hydrogen atoms. The radicals are produced from coal macromolecules in the solid or highly viscous state during the initial stage of their depolymerization. In such circumstances, hydrogen donor solvents play major roles in stabilizing the radicals because of their penetration into solid coal and solvent power (90). Nevertheless, hydrogens adsorbed on the solid catalyst, as well as those of the donor, can stabilize the radicals whenever they encounter the available hydrogens (91). Hence, solvents that reduce the viscosity of the liquefying system are crucially important in enhancing the mobility of radicals and donors for their efficient contact. Immobile solid catalysts are also expected to participate in radical stabilization if they are highly dispersed. Hydrogen spillover on the coal surface may also participate in radical stabilization when the catalyst is well dspersed on the coal surface (90). Another role of the catalyst in coal dissolution is generation and regeneration of donor and/or dissolving solvent through hydrogenation and/or hydrocracking reactions (92). The consumption and generation (or regeneration) of the solvents should be balanced. This requires that the catalyst activity, solvent/coal ratio, and reaction conditions are carefully adjusted. It has recently been reported that radical bond breakage may be initiated by carbon black of high surface area. The conversion of a model coal compound, 4-( 1 -naphthyl-methyl) bibenzyl was accelerated at -375°C without hydrogen pressure in either the presence or absence of hydrogen donors (41, 4 2 ) . This suggests that polarity or surface radical content of carbon black may initiate the decomposition of phenyl-methylnaphthyl linkages.

CATALYSIS IN COAL LIQUEFACTION

c.

57

HYDROGENATION AND HYDROCRACKING OF PRIMARY COALLIQUIDS

Hydrogenation and hydrocracking activity of iron catalysts has been extensively investigated using coal and model compounds (93-95). Iron catalysts can hydrogenerate olefinic unsaturated bonds, while they are known to be less active for the hydrogenation of aromatic rings compared with molybdenum-based catalysts. Hydrocracking reactions such as dealkylation and cracking of paraffins and naphthenic rings are necessary in the liquefaction process to convert heavy nondistillable products into light distillates. Iron catalysts are quite effective in dealkylation and cracking of alkyl side chains. The latter two reactions, however, hardly proceed with conventional molybdenum-based catalysts, which have higher hydrogenation activity (see Section VlII). The acidic properties of the catalyst may be important. Competitive reactivities among a variety of substractes should also be taken into account. In ebullating bed reactor, such as the H-coal process, Ni-Mo or Co-Mo alumina catalysts have been used (96). The catalyst definitely improves the oil yields by accentuating aromatic hydrocracking, achieving conversions 3%. around 95% at catalyst make-up rates of 1 Iron catalysts exhibit much lower activity for the heteroatom-removing reactions such as hydrodenitrogenation(HDN) and hydrodesulfurization(HDS) compared with Mo-based catalysts (97).

-

D. THEMECHANISMS OF RETROGRESSIVE REACTIONS AND

THEIRSUPPRESSION

Once initiated, retrogressive reactions may continue until all of the thermally produced radicals are consumed, leading to very stable condensed aromatic hydrocarbons and eventually coke, unless they are prevented (98, 99). lron catalysts do not appear to be very active catalysts for hydrogenation of aromatic rings, especially at temperatures above 450°C and at hydrogen pressures of 100 200 kgicm’ (100). Under these conditions, donors are not sufficiently regenerated once they are dehydrogenated under liquefaction conditions. Hence, the progress of liquefaction should be adjusted to prohibit retrogressive reactions or it should be terminated before the donor is completely consumed. Lowering the temperature reduces the reactivity of donors, while the thermally produced radicals readily recombine at all temperatures. Thus, retrogressive reactions are difficult to avoid. Direct cou-

-

58

ISAO MOCHIDA AND KlNYA SAKANISHI

pling, as in two-stage liquefaction, is strongly recommended so that intermediates are immediately transferred to the successive reactor for catalytic hydrogenation (76, 101). The additional supply of donor and/or very high hydrogenation activity of the catalyst is emphasized in this successive step. Production of inert asphaltenes and preasphaltenes should be avoided because the yield of distillate is inevitably reduced by their production. Retrogressive reactions can cause such production. One problem with direct coupling of process steps is that of severe catalyst deactivation due to poisons and contaminants produced in the first stage, which are all sent to the second stage. Retrogressive reactions may progress locally in reactors and transfer lines, where long residence times or poor mixing of reaction fluids takes place (102). The surface of the catalyst or minerals adsorb reactants without hydrogenation, and extensive retrogressing reactions may proceed on these surfaces (103). Again, the solvents of high hydrogen donor and dissolving abilities are keys in preventing these reactions. The design of preheater and reactor is of major importance in achieving uniformity in the degree of reaction in the homogeneous phase at all stages. VIII. Combination of Catalyst and Solvent and Stepwise Application of Donor and Catalyst in the Primary Stage

A.

PERFORMANCE OF SOLVENT IN THE PRESENCE OF

CATALYST The combined utilization of solvent and catalyst for primary coal liquefaction processes has been extensively investigated by many researchers in order to increase the distillate yield and improve the efficiency of hydrogen consumption (104). German groups insisted that the liquefaction under hightemperatures (-500°C) and high-pressure (-300 atm) conditions can provide an excellent oil yield regardless of the solvent or catalyst species (105), while other groups such as NBCL, N E D O L , and PETC examined the effectiveness of solvent in the presence of catalyst under much milder conditions (-450°C and -150 atm), indicating the cooperative role of solvent and catalyst in terms of dissolution and depolymerization of coal macromolecules, suppression of retrogressive reactions, and regeneration of the donor solvent (I06 - 108). The present authors identified an optimal mixture of solvents for catalytic liquefaction in presence of pyrite. Figure 10 shows the influences of solvent composition (4HFUPy) on the liquefaction of Morwell coal in an autoclave

59

CATALYSIS IN COAL LIQUEFACTION

A

B

80

80

-50 b?

#

v

v

5

5 a $3

2 $3 20

20

0

0

0

4HFL (70)

ion

0

I(10

4HFL (%)

FIG. 10. Influence of' solvent composition on the catalytic and noncatalytic liquefaction. (A) 3% FeS2 catalyst: (0)oil + asphaltene; (0) oil; (a)gas; ( 0 )preasphaltene; residue. (B) No catalyst: (A)oil + asphaltene; (A) oil; (A) gas; (A) preasphaltene; (A) residue.

(a)

at 45OoC, 20 min, 100 atm hydrogen pressure, and S/C = 1.5 with and without pyrite catalyst (109). A mixed solvent composed of 75% 4HF1 and 25% Py provided the highest oil and oil + asphaltene yields. Pure 4HF1 solvent increased gas yields with decreased oil and asphaltene yields compared to that with the mixed solvent. The catalyst was not as effective with solvents containing less than 75% 4HF1 as oil yields were lowered. This feature is similar to that of hydrogen-transferring liquefaction, indicating that the initial step of liquefaction under lower hydrogen pressure involves dissolution of coal by the donor as a primary route independent of the presence of catalyst. It should be noted that slow heating in an autoclave is never favorable for the donor to perform effectively in hydrogen-transfer deploymerization.

B.

STEPWISE

APPLICATION OF THE DONOR AND CATALYST IN THE PRIMARY STAGE

Two-stage liquefaction processes involving stepwise application of donors and catalyst such as EDS (Exxon donor solvent) and SCT-TSL (short contact time two-stage liquefaction) processes in the United States and LSE (liquid solvent extraction) process in the United Kingdom have been investigated as a means to producing distillable light oil directly using Co-Mo or Ni-Mo catalysts in the second stage, although no catalyst was used in the first stage.

60

ISAO MOCHIDA A N D KlNYA S A K A N l S H l

Recently, closed coupled and/or integrated two-stage liquefactions (CCITSL) have been investigated to elucidate the effects of thermal/catalytic and catalytic/thermal staging on solids buildup. Two-step liquefaction is worth examining within the primary liquefaction stage. Catalysts and donors do not always perform most efficiently under the same conditions. Hence, optimal application is achieved in consecutivc steps, where the best conditions can be selected for each step separately. In such a two-step primary liquefaction process, expensive catalysts, such as Co-Mo and Ni-Mo, are not necessarily employed. These catalysts are more appropriately used when the coal has been depolymerized to soluble products and catalyst poisons are not present in the final upgrading stages. The stages of coal conversion to final products are referred to as the primary stage, in which coal is converted to primary liquid (soluble) products in two steps and a secondary or final stage in which the primary products are upgraded to the final distillate products. Figure 1 1 illustrates the product distributions from the first stage (350°C, 20 min) without catalyst or hydrogen pressure and the successive second stage (380 or 400°C) with FeS2 catalyst in an autoclave in both stages (110). The second stage at 380 and 400°C in the presence of hydrogen converted asphaltene, preasphaltene, and residue to high oil yield but with a marked

0

50

Yield(wt%)

100

4HFL. conv.(%) 98

92

86

92

92

94

FIG. 11. Two-stage liquefaction using autoclave for both stages. L~1:Gas; 0:oil: l;?l:preasphaltene; .:residue. (a) First stage: 350°C-20 min (N2 No catalyst); (b) a + 380°C-20 min (Hz, FeSz catalyst); (c) a + 380°C-40 min (H2, FeS2 catalyst); (d) a + 400°C-20 min (Hz, FeS2 catalyst); (e) a + 40O0C-40 min (H2. FeSz catalyst); ( f ) single stage: 400°C-40 m i n (H?, FeS2 catalyst). L lasphaltene;

61

CATALYSIS IN COAL LIQUEFACTION

increase in gas. The heavier products in the first step at the lower temperature appeared difficult to catalytically upgrade without producing gas. Some retrogressive reactions are indicated to have produced less reactive asphaltenes. Figure 12 illustrates the product distributions in two-step liquefaction using a tubing bomb then an autoclave in the first and second steps, respectively, varying the temperature and time in the first step (110). The influences of time and temperature are clearly shown in Fig. 12, suggesting an optimum condition of 400°C for 10 min, which provided an oil + asphaltene yield of 81% with only 9% gas and 10% preasphaltene + residue after the second step at 400°C under 100 atm hydrogen pressure. A highertemperature (430°C) first step and short time (2 min) increased both gas and oil yields. A longer reaction time at this higher temperature accentuated this trend, decreasing the oil yield. Increasing the temperature even further at very short residence time (450"C, 0 min; heat-up time; ca. 2 min) gave far inferior results. Lower temperature (380°C) and longer time (20 min) provided a fairly high gas yield (24%) with less oil. Temperatures around 380400°C appear optimal for the catalytic step in the present scheme.

4HFL conv.(%) 84

80

69

45% 85

58

FIG. 12. Two-stage liquefaction using tube bombiautoclave for each stage. First stage: no catalyst, tube bomb, ( a l ) 380°C-20mm; (bl) 400"C-IOmin; ( c l ) 430°C-2min; ( d l ) 430°C5 min; ( e l ) 450°C-0min. Second stage: 400°C-20min, FeSz catalyst, autoclave.

62

ISAO MOCHIDA AND KlNYA SAKANISHI

VIII.

Catalytic Upgrading of Crude Coal Liquids in the Secondary Stage

The crude coal liquids produced in the primary stage must be upgraded in the second stage such that they can be further refined with products from petroleum crude into commercial products. Present coal liquid crudes consist of distillates, asphaltenes, preasphaltenes, residues, minerals, and catalysts of the primary stage. In upgrading this crude, one has several options in the choice of feed for the second stage, for example, distillate, distillate plus asphaltene, whole organic products without inorganic components, or the whole crude. Separation processes function as the interface between the primary and secondary stages. The separation process, although adding expense to the overall process, can contribute to substantial increases in valuable products. Processing easier feeds, such as light liquid distillates, may reduce the load on the catalysts while sacrificing precious products, thus wasting a considerable portion of the cost of the primary stage. Nevertheless, even the distillate from coal contains aromatic and polar organic compounds, which are certainly different from the corresponding fractions of petroleum crudes. The feed, which contains heavier polyaromatic components with large quantities of heteroatoms, unreacted coal, and inorganic solids, is very difficult to upgrade as it severely deactivates the catalyst and makes regeneration difficult. The second stage is expected to regenerate the solvent for the primary stage. Hence, the reaction configuration, reaction conditions, catalysts, and solvents must be as carefully designed in this secondary upgrading stage as those of the primary stage. A . INTERFACEOF PRIMARY AND SECONDARY STAGES-SOLID/LIQUID SEPARATIONS SolidAiquid separation is usually required at the interface of the primary and secondary stages to allow optional upgrading of the crude coal liquids of the primary liquefaction stage, by removing mineral matter, unreacted coal, heavy products, and catalysts (111, 112). Distillation, anti-solvent extraction, and centrifugation have been conventionally employed in liquefaction processes (113, 114). Direct close coupling of the primary and secondary stages has been investigated by Chevron and Wilsonville liquefaction facilities to improve liquefaction efficiency and liquid yield. A multistage liquefaction process consisting of deashing, hydrogentransfer liquefaction, catalytic depolymerization with FeS2, catalytic hydro-

CATALYSIS IN COAL LIQUEFACTION

63

racking, and hydrorefining with Ni-Mo may be an attractive alternative in designing the most efficient liquefaction process. Such a multistage liquefaction process would include completing the coal conversion to distillate in the first two stages and upgrading the distillate in a third stage (115).

B. REACTIONS AND ROLESOF CATALYSTS IN THE SECONDARY STAGE Crude coal liquids produced in the primary liquefaction stage can be further hydrotreated through hydrogenation and hydrocracking into gasoline and kerosene range distillates, from which heteroatoms can also be removed rather easily, together with petroleum streams of the same boiling range in conventional refineries. Preasphaltenes and asphaltenes in the crude are preferably depolymerized into oil selectively with minimum formation of coke and gases; however, their conversion to oil is rather difficult because of their resistance toward depolymerization. Thus, the objectives of the second stage are as follows: 1. selective upgrading of heavy liquids, especially asphaltenes and preasphaltenes to distillates 2. deep and efficient refining to remove heteroatoms, such as 0, S, and N. Asphaltenes and preasphaltenes are sent to the secondary stage after the removal of solids. Such feeds must be upgraded to distillates or at least hydrogenated such that on recycle to the primary stage, they are easier to hydrocrack. When the secondary stage is to be operated in a fixed-bed process, the primary concerns are catalyst lifetime and operability as influenced by reactor plugging. The NBCL group developed a Ni-Mo/Al*O3 catalyst modified with Ca to control the activity while limiting coking (116). The catalyst has been shown to have stable activity for 8000 hr. However, its activity for distillate production is rather low. It was also reported that Ni-Mo hydrous titanium oxide (HTO) catalysts were comparable or superior to other commercial and novel formulations tested by Amoco for second-stage upgrading of coal-derived residues (34). Higher activity for hydrocracking as well as long life are still desired. The present authors proposed the use of Ni-Mo supported on lowsurface-area clay or titania as upgrading catalysts because of their low polarities, limited micropores, and strong interactions with Ni-Mo (117). Such properties are expected to exhibit unique activity and selectivity for the

64

ISAO MOCHIDA A N D KINYA SAKANlSHl

TABLE I Their Properties

~ ' ~ i 1 t r 1 ~ v . sund t . s Some of

Commercial Support : NiO (wt%j): MoOi (wt%): Surface areah (rn'/g): Pore diameter

cat.

Cat. - A

Cat. -B"

y-Al.03

Natural clay 3 6

TiOz(Fe20\) 6 12 20

6 I2 I70

YO

(A)

30 I00 I00 -220 220 -400 400-630 630 -I I00 I 100-2 100 ~

Pore volume (mlig)

0.31 0.06 0.0 0.0 0.04 0.27

0.0 0.40 0.30 0.34 0. IS 0.12

0.03 0.01 0.03 0.04 0.04 0 . 10

In-lab. preparation. "Measured by the BET method "

heavy fractions of crude coal liquids. The authors examined the hydrotreatment of so!vent-refined coal (SRC) from Wandoan subbituminous coal using the catalyst listed in Table I ( I 1 7 ) . The products were solvent fractionated into hexane soluble (HS), hexane insoluble-benzene soluble (HI-BS), and benzene insoluble (BI) fractions. The yields of these solvent-fractionated products after hydrotreatment of SRC are plotted against the reaction time in Fig. 13. The overall activities of the catalysts were very similar to those of the commercial catalyst in spite of their lower surface areas. Both exploratory catalysts (Cat-A and Cat-B) showed similar reaction profiles, which were markedly different from those of the commercial catalyst. The BI fraction decreased over the exploratory catalysts equally as well as the over the commercial catalyst. However, the HS fraction hardly increased as long as the BI fraction was present. As the result, the HI-BS fraction increased to a maximum just before the BI fraction disappeared and then rapidly decreased to complete conversion after about 9 hr. The rate of HS formation increased correspondingly during this time. Thus, the exploratory catalysts were found to exhibit a preferential selectivity for conversion of heavier components of SRC, compared to the commercial catalyst. These results emphasize that the chemical and physical natures of the support are important in catalyst design (49).

65

CATALYSIS IN COAL LIQUEFACTION

reaction time (hr)

a

b

FIG. 13. The yields of solvent-fractionated products in the hydrotreatment of SRC vs reaction time. Catalyst: ( a ) commercial catalyst; (b) Cat. -A; (c) Cat.-B. Reaction conditions: SRC1 Cat. ratio (weight), (a) and (c) 1011; (b) 511; reaction temp, 380°C; Hz initial pressure, 100 kgcm', BI (0); HI-BS (0); HS (A);open; first run; closed; second run.

c.

EFFICIENCY OF TWO- OK THREESTEP UPGRADING THE SECONDARY STAGE

IN

Removal of heteroatoms, especially nitrogen in heterocycles, is another important consideration in catalytic upgrading. Nitrogen located in condensed aromatics is difficult to remove, since the denitrogenation proceeds only after all aromatic rings are saturated (118). The authors have identified a two-step catalytic procedure for very effective denitrogenation that uses a Ni-Mo catalyst having very high hydrogenation activity at higher temperatures ( I I Y ) . A single-step reaction at 430°C for 3 hr eliminated only 30% of the nitrogen from the heavy distillate of a Wandoan coal liquid (bp 350-500°C) as shown in Table 11. In marked contrast, a 100% denitrogenation was

66

ISAO MOCHIDA A N D KINYA SAKANlSHl

TABLE I1 Two-Stuge Hydrodenitrogenation of Coal Heavy Liquid Heavy Distillate Initial Reaction N removal Temperature Time pressure pressure Catalyst Run No. Stage ("C) (hr) (MPa) (MPa) (8) Solvent (%)

I 2 3 4

5 6 7

8 9 10

1

2 I 2 1

2 1 2 I 2 I 2 1 2 1 2 I 2

420 420 350 420 350 420 350 420 350 420 350 420 380 420 390 440 390 440

3 5 8 3 8 3 8 3 8 3 8 3 3 3 2 1 2 2

23 23 24 23 24 23 25 22 23 22 23 22 13 11 12 I1 12

10

10 12 10 12 10

12 10 12 10 12 10

6 4 5 4 5 4

11

34 55 8 60' 8 19' 23 81' 33 89' 26 loo' 22 82' 29 64" 29 83'

' I -Methylnaphthalene (solvent : distillate- I : I). "

1 -Methylnaphthalene (solvent : distillate-3

: 1).

' 1 : 4 wt : wt pyrene: I-methylnaphthalene (solvent : distillate-I : I ) . I : 4 wt : wt fluoranthene: 1-methylnaphthalene (solvent : distillate- 1 : I ) . 'Total N-removal for two stages.

achieved using two successive steps with fresh catalyst in each step [hydrogenation at 350"C, 8 hr and denitrogenation at 420"C, 3 hr (120)]. Structural analysis of polycondensed aromatic hydrocarbons in the hydrogenated coal liquid from the single-stage hydrogenation at 430°C suggested that insufficient hydrogenation of the aromatic rings including heterocyclic components was the major cause for its low degree of denitrogenation. Thus, the two-step hydrotreatment was much more effective for deep refining of coal liquids. Catalyst deactivation was also suppressed in the two-step hydrotreatment, as shown in Table 111 (21). The most effective catalysts for the secondary step possess selective activity for hydrocracking of asphaltenes, preasphaltenes, and long-chain paraffins. The two-step hydrocracking procedure enables the acceleration of all reactions through extensive hydrogenation of polar aromatics at lower temperatures (<400"C) in the first step and successive hydrocracking of hydrogenated products at higher temperatures (-420°C) in the second step, Catalysts can be separately optimized for their respective purposes in each

67

CATALYSIS IN COAL LIQUEFACTION

TABLE 111 Two-Stage Hydrotreatment of Asphaltene in Australian Brown Coal Liquid: Effects of Two-Stage Concept on Catalyst Deactivation (Changes of Catalyst Activations and Weight in the Repeated Runs")

Number of repeated run 1

2 3 380°C for 3 hr 4 5 6 7 1

2 3 430°C for 3 hr 4 5 6 7 1

Two stageh

2 3 4 5 6 7

Afa

HDN(%)

ICW(wt%)

0.20 0.20 0.20 0.20 0.20 0.20 0.20

43 38 38 38 38 37 37

8 8 8 8 9 9 9

0.17 0.16 0.16 0.16 0. I6 0. I6 0.16

70 65 62 58 58 58 58

8 9 10 II II 12 13

0.25 0.23 0.23 0.23 0.23 0.23 0.23

80 75 73 72 71 71 70

3 4 4 5 6 6 6

" A J ; , change of carbon aromaticity before and after the reaction; HDN, nitrogen removal; ICW, increment of catalyst weight (THFI on the catalyst). "390°C for 2 hr and 430°C for 2 hr.

step of the hydrotreatment. Acid zeolites have been utilized in combination with Ni-Mo alumina catalysts for paraffin cracking (121). This catalyst system can also be accomplished in a two-step procedure. Such two-step hydrotreatments are also effective for deep hydrodesulfurization of diesel fuel, heavy gas oil, and atmospheric residue, without producing fluorescent compounds or dry sludge, which are of current concern in petroleum refining, which must meet current regulations. Some results on two-step hydrodesulfurization of diesel fuel and hydrocracking of petroleum residue are summarized in Tables IV and V ( 2 2 , 122). The combined use of suitable catalysts, as well as optimization of reaction conditions at the respective step, are very important in achieving deep desulfurization without color development and higher conversion without sludge formation.

68

ISAO MOCHIDA AND KINYA SAKANlSHl

TABLE IV Two -Stuge Desrilfurizotion oj Diesel Oil Reaction condition ("C-kg/cm'-hr)

Catalyst First

Second

First

Second

Sulfur con ten t (wt%)

CoMo NiMo CoMo NiMo

CoMo NiMo NiMo CoMo

320-50- I 320-50- I 320-50- 1 320-50- I

320-50- I 320-50- I 320-50- 1 320-50- 1

0.080 0.074 0,049 <0.05

CoMo COMO CoMo CoMo CoMo

NiMo NiMo NiMo NiMo NiMo

300-50- I

320-50-1 320-50- 1 320-50- 1 300-50- 1 340-50- 1

0.074 0.05 1 0.028

340-50- 1 360-50- I

320-50- I 320-50- I

0.109 0.049

Color Transparent Transparent Transparent Light fluorescence Transpar en t Transparent Transparent Transparent Transparent

Conditions ("C-h) Catalyst

First

Second

KF-842

390-4 420-4

__

390-3

420- I

390-4 420-4 390-3

420- I

390-3 390-3 390-3 390-3

420-3 430- I 440- I 440- I

KFK- I 0

-

Recovery D Y (wt%) (wt%)

PQ')

95 75 94 98 Y1 94 Y4

94 93 9.5

"DY, 540°C distillate yield; ( ), 565°C distillate yield. " P o , product quality; g, good; vg. very good; sg, somewhat good; ng, no good.

D.

I)ESIGN OF CATAI.YSTS FOR THE SLCONDAKY STAGL,

PARTICULARLY MULTISTEP UPGRADING (123-127) Catalysts of higher activity for hydrotreating have been investigated for petroleum refinement. Improvements through novel design of supports and methods for impregnation, calcination, and sulfidation are being sought to achieve higher catalyst dispersion on the support, no sintering during use,

CATALYSIS IN COAL LIQUEFACTION

69

optimal sulfide levels, and homogeneous distributions of catalytic species. Although some significant improvements are reported, more systematic investigations of the catalyst precursors and activation procedures are still necessary. Catalysts for coal liquefaction require specific properties. Catalysts of higher hydrogenation activity, supported on nonpolar supports, such as titania, carbon, and Ca-modified alumina, are reasonable for the second stage of upgrading, because crude coal liquids contain heavy polar and/or basic polyaromatics, which tend to adsorb strongly on the catalyst surface, leading to coke formation and catalyst deactivation. High dispersion of the catalytic species on the support is very essential in this instance. The catalyst/support interactions need to be better understood. It has been reported that such interactions lead to chemical activation of the substrate (127). This is discussed in more detail in Section XIII.

IX.

Roles of Solvents in the Secondary Stage

Reaction solvents in the secondary stage of coal liquefaction play important roles in reducing catalyst deactivation by inhibiting the formation of coke from polar or polycondensed aromatic hydrocarbons, through hydrogen donation, and by removing coke precursors via dissolution from the catalyst surface (119, 120). Because of their excellent dissolving ability, fourring aromatics are very beneficial solvent components. Such components are also easily hydrogenated during catalytic upgrading and become very effective hydrogen donors as well. The authors reported an example of such solvent effects in the catalytic two-step denitrogenation of coal liquid distillates. With large quantities of added 1 -methylnaphthalene or 20% added pyrene or fluoranthene, no additional catalyst is necessary for the second step to achieve high levels of denitrogenation. The catalyst’s resistane to coking may also be improved by the solvent while maintaining high catalytic activity. Hydrogen pressure can, thus, be reduced to 110 atm in a two-step hydrotreatment, as shown in Table I1 (2f,128). The two-step hydrotreatment assisted by the proper solvent not only improves the extent of hydrocracking and heteroatom removal, but it also allows the reduction of the hydrogen pressure required for efficient upgrading . Aromatic solvents have been reported to prevent sludge formation during two-stage hydrocracking of petroleum atmospheric residue at high conversion by dissolving sludge precursors such as asphaltenes and heavy polar aromatics ( 2 2 ) . Sludges that remain on the catalyst deactivate it through coke formation.

70

ISAO MOCHIDA A N D KlNYA SAKANISHI

X.

Lifetime, Recovery, and Regeneration of Liquefaction Catalyst in the Primary and Secondary Stages

Coal and its derived liquids contain a number of catalyst poisons or poison precursors, which inevitably and severely deactivate the catalysts during the liquefaction process. The catalysts are deactivated via three major routes (129). 1. Carbonaceous substances and cokes cover the catalyst surface. 2. Mineral matter contaminates or coats the catalyst surface. 3 . Chemical changes of catalytic species occur; for example, sulfide catalysts are converted to oxides or sulfates and chlorides are converted to oxides through oxidation or hydrolysis. The difficulty in the recovery of catalysts from unreacted coal and minerals and the poor regenerability of used catalysts forces one to use disposable catalysts, especially in the primary stage. This increases the cost of coal liquefaction considerably. This section reviews the mechanism of catalyst deactivation, design of recoverable catalysts in the primary stage, and catalyst deactivation in the secondary stage. A.

MECHANISM OF CATALYST DEACTIVATION IN THE COALLIQUEFACTION

1. Deactivation by Carbonaceous Substances and Carbon (130-132) The catalysts commonly used in coal liquefaction are iron sulfide in various forms and complex sulfides of cobalt or nickel and molybdenum supported on aluminas that are acidic or at least polar. The organic substances derived from coal are mostly high-molecular-weight aromatic compounds, many of which contain oxygen and/or nitrogen functional groups. Some of them contain both basic nitrogens and acidic oxygens. These materials have low solubilities in the solvent because of their high molecular weights and strong molecular associations, even in the liquid phase. Such materials are preferentially adsorbed on the catalyst, where they are hydrogenated and hydrocracked and heteroatoms are removed. They then desorb into the liquid phase as the upgraded products of the liquefaction process. However, some of them may remain on the catalyst for extended periods of time, even if they are partially converted, because of acid-base or polar interactions with the catalyst surface or because of poor solubility in the solvent. While on the catalyst, adsorbates can be gradually converted into very heavy molecules because of insufficient hydrogen supply and/or low reactivity of

CATALYSIS IN COAL LIQUEFACTION

71

the adsorbate toward hydrogenation. Products that are difficult to desorb eventually become coke. Aromatic compounds larger than naphthalene are thermally and catalytically labile toward condensation through their oligomers and carbonaceous derivatives (133). Hence, the organic substances, which stay on the catalyst for extended periods of time, become carbonaceous, irreversibly covering the catalyst surface. Such carbonaceous substances may also trap minerals, thus increasing the volume of coating materials. The organic solids of fine size, such as inert macerals present in the starting coal, may also cover the surface or plug the pore mouth of the catalyst. However, this contribution to catalyst deactivation appears minor because their diffusion into the catalyst pores is limited. Thus, the carbon deposition on the catalyst surface principally takes place through strong substratecatalyst interactions and long residence times of high molecular weight and polar substances. On the basis of this premise, it may be concluded that the nature of the catalyst surface, the properties and reactivity of coal, and coalderived liquids and solvents are all important to the adsorption/desorption equilibrium and in hydrogenation (hydrogen transfer) of adsorbates, which influences carbon deposition and catalyst deactivation. 2. Deuctivution by Inorganic Deposits (134, 135) Two kinds of inorganic minerals are present in the coal; organic-bound ions and solid minerals such as clays and inorganic oxides, chlorides, sulfides, and sulfates. Although very fine grains of diameters less than 300 A may penetrate into the catalyst pores and precipitate by further crystal growth, most inorganic solids are too large to plug catalyst pores. Adhesion of inorganic minerals to the catalyst surface does not appear to be strong; thus, they may be excluded as being a factor in catalyst deactivation. However, such solids may be trapped with catalytic solids in the preheater, precipitate on the reactor walls, or deposit on the bottom of the reactor as scales and sludges. Precipitation of very fine solids originating from organic-bound ions that thermally decompose during the early stages of coal liquefaction appears to be a significant factor in catalyst deactivation. Basic ions, released by this decomposition, can neutralize the acidic sites of catalysts. More importantly, the carbonates, hydroxides, sulfides, and chlorides of sodium, calcium, magnesium, iron and titanium are major poisons as they crystallize on the surface of the catalyst and on the reactor walls. Such organic-bound ions can easily penetrate into catalyst pores and precipitate within the catalyst pores. This precipitation mechanism appears to be the common cause of scale and sludge formation, which accompanies other organic and inorganic solid depositions.

72

ISAO MOCHIDA A N D KlNYA SAKANlSHl

3 . Deactivation by Catalyst Transformation (136, 137)

Liquefaction catalysts, such as sulfides, lose their catalytic activity, especially hydrogenation activity when they are transformed into sulfate or oxide. Even reduction of the extent of sulfiding leads to a significant loss in catalyst activity. The crystalline form of the catalyst may also influence the catalytic activity. Thus, the level of sulfur during coal liquefaction is critical. This can be controlled by the addition of sulfur additives. Sintering, which reduces the active catalyst surface area, may also take place. However, little detailed information is presently available on this topic with respect to coal liquefaction.

B.

DESIGN OF RECOVERY AND REGENERATION OF CATALYST FOR THE PRIMARY STAGE( 1 x - 1 4 0 )

The recovery, regeneration, and repeated reuse of the active catalyst are of prime importance in substantially reducing the overall cost of coal liquefaction. The used catalysts usually remain in the bottoms products, which consist of nondistillable asphaltenes, preasphaltenes, unreacted coal, and minerals. The asphaltenes and preasphaltenes can be recycled with the catalyst in bottoms recycle processes. However, unreacted coal and minerals, if present in the recycle, dilute the catalyst and limit the amount of allowable bottoms recycle because they unnecessarily increase the slurry viscosity and corrosion problems. Hence, these useless components should be removed or at least reduced in concentration. If the catalyst is deactivated, reactivation becomes necessary before reuse. Thus, the design of means for catalyst regeneration and recycle is necessary for an effective coal liquefaction process. Several approaches to achieving these goals are discussed below. 1. Complete removal of inert macerals and minerals from the starting coal and complete conversion of reactive macerais at the expense of excess production of hydrocarbon gases. In this situation, the recycled bottoms consist of only catalyst and heavy coal liquid products. Reactor designs should attempt to avoid catalyst deactivation, providing immediate reuse of the catalyst. 2. Chemical separation of the catalyst from the bottom products. Some coals, such as Australian brown coal, consist principally of reactive macerals and contain organically bound calcium and sodium, which almost exclusively produce carbonate and chloride minerals during liquefaction. The catalyst can be separated by extracting these minerals, which exist on the catalyst surface or as precipitates in the bottoms product. 3. Physical separation of the catalyst from the bottoms product. Specitic gravity, particle size, and magnetic susceptibility of the catalyst could possibly provide the means for catalyst separation.

CATALYSIS IN COAL LIQUEFACTION

73

The first approach relies heavily on coal pretreatment. Although such pretreatment may be possible, the complete removal of inorganic minerals does not appear practical. The second approach depends heavily on obtaining very specific coals, which may exist in only limited supplies. Carbonates and chlorides have been found to be easily extracted by weak acids, such as acetic acid, without harming catalysts such as FeS or Ni-MoS, although regeneration and resulfiding may be necessary because some of the sulfur may be replaced by oxygen during liquefaction. Catalyst deactivation by minerals deposition will not be of concern for coals extracted in this way. Although the third approach is more general, proper catalyst design is essential. Dow proposed the use of a sulfide catalyst of very fine particles that can be recovered by centrifugation (76), although separation was tedious and never complete. Catalyst flotation may be possible; catalysts supported on carbon may allow facile flotation (42) and hydrophobic surfaces may also help catalysts to float in water. The present authors proposed ferromagnetic supports, which can withstand liquefaction environment and conditions by their inherent nature or by protection with a carbon coating (141). Carbon has been recognized as an excellent support for FeS and Ni-MoS as described in the previous section (50). Very fine particles of ferrite are available. Coating with carbon can be performed through precipitation of polymers or pitches by the aid of suspension agents followed by carbonization (141). Catalyst deactivation by mineral and carbon deposition should be avoided for this approach to be feasible. The removal of cations, such as sodium, calcium, magnesium, iron, and zinc from coal, by ion exchange is quite effective, as discussed earlier. If this is not done, cations can be labilized, through the decomposition of the organic functional groups to which they are bound, during the early stages of coal liquefaction, forming stable carbonates, chlorides, and oxides that grow into crystals large enough to be harmful to catalysts. Many oxide minerals present in the coal may not be strong poisons for catalysts because of their size and stability.

c.

DESIGN OF DURABLE CATALYST FOR THE PRIMARY STAGE (142, 143)

Recovered catalysts should maintain catalytic activity or at least possess regenerable sites for repeated use. Of prime importance, catalysts in the first stage of liquefaction can inhibit coke formation. Coke once produced on the catalyst is difficult to remove without burning, which inevitably converts iron sulfides to oxides and/or sulfates. In designing catalysts, low acidity or polarity may be appropriate so as not to strongly adsorb heavy, polar, or ba-

74

ISAO MOCHIDA AND KlNYA SAKANISHI

sic coal-derived intermediates. Highly porous catalysts may retain such materials until condensation takes place. But catalytic species such as iron sulfides should have high surface areas for high activity. Very small particle size or highly dispersed catalysts should be employed. Electronic interactions between catalytic species and support have been reported to increase the hydrogenation activity (52). Carbon, titania, and zirconia may have potential in this area and should be considered for further investigation (124). It is difficult to design a catalyst that can tolerate deposition of inorganic materials from organically bound sources. Such poison precursors should be removed or stabilized before the catalytic process as described earlier. Pores and pits on the catalyst surface may not trap solid catalyst poisons even if they are very fine. Again, poreless, very small particles are recommended. It is not necessary or even desirable to achieve highly upgraded coal products in this initial liquefaction step as a sufficiently depolymerized and hydrogenated product can be further upgraded in a secondary step through hydrocracking and hydrorefining.

D.

CATALYST

DEACTIVATION IN THE SECONDARY STAGE

The secondary stage is conducted either after liquid-solid separation or in a close-coupled scheme. In the former case, inorganic solids as well as heavier and less soluble portions of coal liquids, such as preasphaltenes, are basically removed from the feed to the secondary stage, although their removal is never perfect, especially very fine particles. The coal liquid after separation still contains organometallic substances, carboxylate and phenolate salts. The most stable portion of such organometallics may decompose on the catalyst surface and poison it just as was described for the first-stage catalysts. Calcium, sodium, titanium, and magnesium are suspected to be major poisons. Such poisons can be extensively removed through acid extraction and neutralization of primary products. An acidic guard bed may also work to remove these organometallic poisons (144). It should be noted that fine solid particles of low reactivity that are not removed by separation processes basically pass through fixed-bed catalyst reactors. Hence, catalyst deactivation in the secondary stage is principally due to strong adsorption of heavy polar organic materials and their coking on the catalyst surface. In the close-coupled scheme, all produts and minerals are sent to the catalytic process. Heavy deposits of carbonaceous substances are inevitable. Heavy products of preasphaltene should be converted as much as possible in the primary stage. The detailed characterization of all products, including orgnometallics, suggests ways to convert or stabilize the poison precursors in the primary stage.

CATALYSIS IN COAL LIQUEFACTION

75

The solubility of unhydrogenated aromatic hydrocarbons is sometimes limited in excessively hydrogenated or hydrocracked solvents. Overhydrogenation of the solvent produces anti-solvent characteristics and may enhance phase separation at the last minute of the coal liquefaction (145). This is believed to be the mechanism of sludge formation in hydrocracking processes. Such sludges are essentially dissolved at higher temperatures; however, some of them tend to be adsorbed on the catalyst surface and eventually become carbonaceous poisons. The adsorption, and thus deactivation, is governed by the equilibrium solubility, dissolving ability of the solvent, and the properties of the catalyst surface. As discussed above for the primary stage, hydrogen donor solvents may react with carbonaceous products on the catalyst surface in the secondary stage. Hence, high conversions of carbonaceous deposit precursors should be incorporated into the primary stage design, such that all products are distillable or are severely modified so as not to be unharmful to the catalyst of the secondary stage.

XI.

Pretreatment of Coal and Coal Liquid

As described in a previous section, coal and its liquefied products are mixtures of complex organic and inorganic materials. Their pretreatment before their liquefaction, as well as prior to catalytic hydrotreatment, profoundly influences their reactivity and behavior (27, 146, 147). Coal is conventionally washed or deashed according to the gravity differences to reduce its mineral matter and inert macerals. Such mineral reduction is very beneficial for liquefaction as it removes nonreactive substances that wear moving parts of pumps and valves and reduce the potential recovery of oil from solid residues. In addition, such solids may plug the catalyst bed. However, removal of all harmful substances may never be feasible. Different kinds of contaminants in the coal bring about their own specific problems. Hence, it is important to identify the particular species and their specific characteristics to design means for their removal and/or to render them harmless. If properly designed, the combination of donor solvents and catalysts offers the potential for almost complete conversion of coal into distillable oil, leaving no residue. Deashing pretreatments activate even some of the unreactive macerals, such as semifusinite in Australian brown and subbituminous coals, which are difficult to solubilize in liquefactions using donor solvents. Such activation by the removal of cations appears to be characteristic of Gondowanan coals, although the treatment enhances the liquefaction potential of all coals. Figure 14 shows some results of Australian brown coal (Morwell) with tetrahydrofluoranthene (4HF1) and FeSz catalyst under HZ pressure of

76

ISAO MOCHIDA A N D KINYA SAKANISHI

Flc;. 14. Effect of' decationizing pretreatment o n the liquefaction. (a I)First-stage noncatalytic hydrogen-transferred Morwell coal in the two-stage hydrotreatment (400°C- 10 min, 20 atm Nz, tube bomb and molten tin bath. rapid heating) 4HFlicoal = 3.Og/3.Og. (a21 Second-stage catalytic hydrotreated Morewell coal in the two-stage hydrotreatment (400"C-20 inin, 50 cc autoclave, slow heating).

100 kgkm' after the deionizing pretreatment (148). Complete liquefaction appears to be possible under rather milder conditions. In the coal liquefaction process, scales and sludges, formed on the reactor walls, cause severe problems, which limit long operation. The crystal growth of chlorides and carbonates appears to trigger their formation, trapping the other solids. Hence major problems may come from cations, which react with carbon dioxide or chloride ions to form insoluble crystalline solids. The intrinsic solids may not initiate the problem. Cations bound to organic functional groups in the coal-derived products are oil soluble and can diffuse through the micropores of the solid catalyst, deactivating it as described in a previous section. Such ions can also bridge coal macromolecules, preventing dissociation and lowering their solubility during liquefaction. Hence, their removal from the coal reduces the catalyst deactivation and enhances the dissolution and depolymerization of coal macromolecules. Preasphaltene, a heavy liquid product, is usually rich in polar functional groups, as they are sometimes called preasphaltol (76). Such groups still contain cations that liniit their solubility and their removal from the catalyst surface. Because these problems are associated with cations, not general mineral substances, cation removal and/or stabilization by ion exchange is beneficial. Acid leaching and pretreatments that decompose the complexes of cations and oxygen groups can also remove or stabilize these cations. Alternatively, if particles of crystalline chlorides, carbonates, or hydroxides can be grown to sufficiently large size, they will not deposit on the catalyst surface, deposit into the pores of the catalyst, or precipitate on reactor walls. Hence, such pretreatments can be carried out before the catalytic upgrading stage by the aid of donors that suppress retrogressive reactions during the pretreatment (149, 150).

CATALYSIS IN COAL LIQUEFACTION

77

Thermal pretreatment can decompose carboxylic acids or their salts prior to liquefaction. The cations are stabilized as carbonates. Acidic guard catalysts may be used to trap the ions (151). The guard catalyst can be regenerated by washing or by exchanging the deposited minerals. Coking on such guard catalysts should be carefully avoided.

XII. Ionic Depolymerization of Coal (752-759)

Coal macromolecules contain ether, ethylene, and aryl-alkyl linkages that may be cleaved ionically as described in the following equations. Ar-CH2-O-Ar Ar-CH2-CH2-Ar

+ HO-Ar Ar-CH2+ + CH-Ar.

+ Ar-CHZ+

(1 1

---*

(2)

These acid-catalyzed reactions occur at temperatures lower than those used in conventional liquefaction processes. Phenol induces acid-catalyzed solvolytic reactions at temperatures as low as 120°C. The liquefied products contain chemically bound phenolic groups, a portion of which can be recovered by cracking. Transition-metal halides have been reported to exhibit hydrogenation activity under hydrogen pressure in the temperature range of conventional liquefaction. The combination of acidic cracking and hydrogenation leads to the hydrocracking of polyaromatic rings (37), although the origin of hydrogen activation by the halides is not clear. Molten metals such as tin and zinc derived from their halides may have hydrogenation activity, especially when they arc dispersed on the coal surface. The most attractive feature of the halide catalysts is that they can liquefy coal at temperatures much lower than those of conventional liquefaction. Unfortunately, they are difficult to recover and recycle, and they are very corrosive. The halides may also react with sulfide or hydroxide, thus losing their acidity. Their acidity may also be neutralized with strong stable Nbases. In spite of their molten state, a large amount of halides is required.

XIII. Prospects for Catalysts in Coal Liquefaction

In spite of extensive efforts to improve coal liquefaction, so far, coalderived oils cannot compete with petroleum. The cost of coal liquefaction must be lowered to produce petroleum substitutes at a reasonable price. The cost of coal liquefaction is most sensitive to the capital costs of the facility. Hence, higher yields of products at the same facility or fewer and/or

78

ISAO MOCHIDA A N D KINYA SAKANISHI

less expensive facilities are targets of future studies. The prodution of hydrogen is one of the most expensive component facilities. Less hydrogen consumption by less production of hydrocarbon gases is generally the preferred means to limit hydrogen-associated costs. However, one must consider that the nature of the gasification feed greatly influences the cost of the hydrogen plant. The cost of hydrogen generation from various feeds drastically increases in the following order: natural gas, coal, coal residue, char. Coal-derived hydrocarbon gases can be steam reformed along with supplemental natural gas for hydrogen generation. Hence, the authors propose that an improved coal liquefaction process may be one in which complete conversion of coal with minimal formation of residue should be the primary goal, even if hydrocarbon gas yields increase. These hydrocarbon gases are easily converted to recover the hydrogen they consume. A liquid/solid separation step is usually required before the catalytic reaction with nondisposable catalysts. This is an expensive step. To remove it, the catalyst must be compatible with the contaminating solids. Ebulating beds can partially satisfy this condition, but the rate of catalyst replacement becomes large if the deactivation is severe. Solid coal is transported as a slurry. Hence, the solvent is necessary in any case. However, reduction of its amount is very critical to improving the productivity of the facility at reduced cost. The solvent requirement, especially after dissolution of coal, can be reduced by improved catalysts. Finally, the presence of solid catalysts in coal liquid slurries limits the recovery of oil. Thus, lower catalyst requirements are desired. All of these problems are related to the performances of the catalysts used in coal liquefaction. Very active, durable, recoverable, and regenerdble catalysts are most wanted in the primary liquefaction stage, where catalyst poisons from asphaltenes and minerals are most severe. Multifunctional catalysts should be designed by selecting supports with specific functions, such as strong but favorable interactions with catalytic species, resistance to poisons, and improved properties to allow easy recovery, while maintaining high activity. The catalytic functions, such as high hydrogenation activity under minimum pressure, high cracking activity with limited hydrogenation activity, balanced activity for hydrocracking, and selectivity for specific coal product fractions can be independently designed using consecutive multistep processes under optimal conditions in each reactor. Hydrogen pressures below 10 MPa are believed to reduce the cost of facilities significantly. Lower-temperature liquefaction processes may be the targets of future research. As discussed previously, acidic catalysts can depolymerize coal at temperatures as low as 150°C. Thus so far, rather unconventional catalysts,

CATALYSIS IN COAL LIQUEFACTION

79

which may have unacceptable costs and corrosion problems, are reported (153, 154). Biochemical and biomimetic catalysts may also be expected to

perform the reactions efficiently under milder conditions. Catalysts in liquid state are favorable, although no low-cost homogeneous catalysts yet known can hydrogenate aromatic rings. There is still a great opportunity for the discovery of new and improved coal liquefaction catalysts.

XIV.

Design of Multistage Coal Liquefaction with Catalysts Yet to Be Developed

Using known catalysts or those yet to be developed, it may be possible to design multistage coal liquefaction that can achieve complete coal conversion under hydrogen pressure around 8 MPa at coal/solvent ratios of unity, producing 80% yields of oil and leaving no organic residue to be gasified. The major portion of the hydrogen necessary for stoichiometric conversion can be produced from gaseous hydrocarbon by-products with supplemental natural gas. The removal of cations and thermal pretreatments to decompose carboxylates enhances reactivity and reduces the poisonous effects of minerals. Selected and specially prepared solvents, consisting of balanced donors and polyaromatic hydrocarbons, convert solid coal completely to liquid under 8 MPa, by virtue of their high boiling points. The remaining active minerals are transformed to nonpoisonous forms to avoid harm upgrading catalysts. Recoverable and regenerable catalysts are used to convert all coal products to distillable liquids while coproducing considerable gaseous hydrocarbon by-products. The solvent recycled to the primary step is very beneficial for this conversion if properly hydrogenated. The interface between primary and secondary stages is only flash distillation to separate distillates and solids. The recovered catalyst is regenerated and recycled. Distillable products, together with solvent and solvent range products, are upgraded in the secondary stage. This distillate is first hydrogenated extensively. Then, hydrorefining and cracking of paraffins and heavy coal products are carried out on the respective fractions, using appropriate catalysts for the respective objectives. By the aid of proper guard beds, each step of each stage can be operated with fixed-bed catalyst processes. When the primary stage leaves some asphaltenes in the product, the whole product, including the catalyst and solids, is sent to the secondary catalytic stage, using honeycomb-type catalysts of poreless supports to allow passage of all solids. Thus, flash distillation and catalyst recovery can be easily accomplished.

80

ISAO MOCHIDA A N D KINYA SAKANlSHl

XV.

Concluding Remarks

By the end of this century, the supply of petroleum will reach a maximum and then decline slowly. On the other hand, the worldwide demand for liquid fuels will remain at present levels or even increase because of rapidly growing demands of developing countries. Thus, both saving energy and finding new supplies of liquid fuels are major tasks to be developed within these 10 years. Coal liquefaction that can provide liquid fuels at the price of current petroleum (not cost but price) is one of the most important technologies that needs to be developed. The catalyst and control of its operating conditions are still key to technology for advanced coal liquefaction. The creative design of catalyst materials and reaction schemes is an important and challenging goal for the future. ACKNOWI.LWMENTS The authors are grateful to Dr. Paul B. Weisz lor the invitation to review the topic of the present article. They also acknowledge Drs. 0. Okhuma, Y . Mitarai, S . Ueda, the Ministry of Education, the Ministry of International Trade and Industry, and New Energy and Industrial Technology Development Organization for their long term support of our coal liquefaction research. The authors also express special thanks to Dr. D. D. Whitehurst and Dr. F. J. Derbyshire for their valuable comments on the manuscript. REFERENCES 1. Bergius, F., and Billiviller, J . , Germany Patent 301, 231 (1919). 2. Zielke, C. W., Struck, R. T., Evans. J. M., Costanza. C . P., and Gouin, E., Inti. Eng. Chem. Process D e s . Dev. 5, 158 (1966). 3. Lummus C o . , “The C-E Lummus Clean Fuel From Coal Process, ” 1975. 4 . Mochida, 1.. Sakanishi, K., Korai, Y., and Fujitsu, H., Fttel Process. Technol. 14, I13 ( 1986). 5 . Shinn, J . H., Fuel 63, 1187 (1984). 6. Painter, P. C . , Sobkowiak, M., and Youtheff, J., Fuel 66, 973 (1987). 7. Sasaki, M., and Sanada, Y . , J . Pet. Inst. Jpn. 34, 218 (1991). 8. Larsen, J . W., and Basker, A . J.. Energy Fuels 1, 230 (1987). 9 . Cartz, L., and Hirsch, P. B., Philos, Trans. R. Soc. London Ser. A 252, 559 (1960). 10. Hatcher, P. G . , Energy Fiiels 2, 40 (1988). 11. Wender, I., Prrpr. Am. Chem. Soc. Div. Fuel Chem. 20 (4).16 (1975). 12. Iino, M., Takanohashi, T., Ohsuga, H . , and Toda, K . , Fuel 67, 1639 (1988). 13. Wei, X . Y.,Shen, J. L., Takanohashi, T., and lino, M., Energ.y Fuels 3, 575 (1989). 14. Botto, R . E., Wilson, R . , and Winans, R. E., Energy Fuels I , 173 (1987). 15. Larsen, J. W., Lee, D., and Shawver, S . E., Fuel Process. Technol. 12, 51 (1986). 16. Mochida, I . , Yufu, A , , Sakanishi, K., Korai, Y., and Shimohara, T., J. Fuel Soc. Jpn. 65, 1020 (1986). 17. Neavel, R. C., Fuitel 55, 237 (1976). 18. Derbyshire, F. J., Davis, A . , Epstein, M., and Stansberry, P., Fuel 65, 1233 (1986). 19. Burgess, C. E., and Schobert, H. H., Fuel 70, 372 (1991).

CATALYSIS IN COAL LIQUEFACTION

81

20. Davis, A., Derhyshire, F. J., Mitchell, G. D., and Schohert, H. H., “Enhanced Coal Liquefaction by Low-Severity Catalytic Reactions. ” Report to U. S . Dept. of Energy, DOE-PC-90910-F1, p. 175. 1989. 21. Mochida, I . , Zhao, X . Z., and Sakanishi. K., Fuel 67, 1101 (1988). 22. Mochida, I., Zhao, X . Z., and Sakanishi, K . , I d . Eng. Chem. Res. 29, 234 (1990). 23. Whitehurst. D. D., Mitchell, T. O., and Farcasiu, M., “Coal Liquefaction-The Cheinistry and Technology of Thermal Processes, ” p. 207. Academic Press, New York, 1980. 24. Poutsrna, M. L., Energy Fuels 4, I 13 (1990). 25. McMillen. D. F., Malhotra, R., Hum, G. P . , and Chang, S. J., Energy Fuels 1, 193 (1987). 26. Karniya, Y., Futarnura. S. , Mizuki, T., Kajioka, M . , and Koshi, K . , Fuel Process. Techno/. 14, 79 (1986). 27. Mochida, I.. Yufu, A., Sakanishi, K., and Korai, Y . , Fuel 67, I14 (1988). 28. Mochida, I . , Takayama, A., Sakata, R . , and Sakanishi, K., Energy Fit& 4, 398 ( 1990). 29. Malhotra, R., and McMillen, D. F., Energy Fuels 4, 184 (1990). 30. Bate, K . , and Harrison, C., Fuel 71, 289 (1992). 31. Sakata, R., Takayarna, A , , Sakanishi, K., and Mochida, I., Energy Fuels 4, 585 ( 1990). 32. Montano. P. A., Boinrnannavar, A. S., and Shah. V., Fud 60, 703 (1981). 33. Suzuki, T., Yamada, H., Sears, P. L., and Watanahe, Y . , Energy Fuels 3, 707 (1989). 34. Stephens, H. P . , Dosch. R. C . , and Stohl, F. V., Ind. Eng. Chem. Prod. Res. Dev. 24, 15 (1985). 35. Derhyshire, F. J., Energy Fuels 3, 273 (1989). 36. Mobley, D. P . , and Bell, A. T., J . Cum/. 64, 494 (1980). 37. Mizurnoto, M.. Yarnashita, H., and Matsuda, S.. Ind. Eng. Chem. Prod. Res. Dev. 24, 394 (1985). 38. Hirschon, A. S . , Wilson Jr., R . B . , and Laine, R. M . , Appl. Cuful. 34, 3 1 I (1987). 39. Ogawa, T., Stenberg, V. I . , and Montano, P. A , , Fuel 63, 1660 (1984). 40. Hirschon, A. S . , and Laine, R . M . , Fuel 64, 868 (1984). 41. Rrcasiu. M.. and Smith, C., P r e p . A m . Chew. Soc. Div. Fuel Chem. 35, 404 (1990). 42. Rrcasiu, M., and Smith, C., Energy Fuels 5, 83 (1991). 43. Bacaud, R., Besson, M., and Djega-Mariadassou, G . , Prepr. A m . Chem. SOL..Div. Fuel Chem. 38 ( I ) , 1 (1993). 44. Lacroix, M.. Boutarfa, N., Cuillard, C . , Vrinat, M., and Breysse, M., J . Cutul. 120, 473 ( 1989). 45. Matson, D . W., Linehan, J. C., and Darab. J. G., Prepr. A m . Chem. SOC. Div. Fuel Chem. 38 ( I ) , 14 (1993). 46. Hager, G . T.. Bi, X . X . , Eklund, P. C., and Derbyshire, F. J., Prepr. A m . Chem. SOL.. Div. Fuel Chem. 38 ( I ) , 34 (1993). 47. Pradhan, V. R . , Hu, J., Tierney, J . W., and Wender, I . , Prepr. A m . Chem. Soc. Div. Fuel Chum. 38 (I),8 (1993). 48. Martino, A . , Wilcoxon, J. P., Sylwester, A. P., and Kawola, J. S . , Prepr. Am. Chem. Soc. Div. Fuel Chem. 38 ( I ) , 21 (1993). 49. Mochida, I.. Oishi. T., Korai, Y., and Fujitsu, H., Ind. Eng. Chem. Prod. Res. Dev. 23, 203 (1984). 50. Derbyshire, F . J., DeBeer, V. H. J., Ahotsi, G. M. K . , Scaroni, A. W., Solar, J. M., and Skrovanek, D . J., A&. Cutul. 27, 117 (1986). 51. Ehrhnrger, P., Mahajan, 0. P., and Walker, P. L., J . Cutul. 43, 61 (1976). 52. Showa, Japanese Patent 57.132547 (1982). 53. Lewis, J. M.. and Kydd, R. A , , J . Cutul. 132, 465 (1991).

82

ISAO MOCHIDA AND KlNYA SAKANISHI

54. Morales, A , , Ramirez de Agudelo, M . M., and Hernandez, F., Appl. Catal. 41, 261 (1988). 55. DeCanio, E. C.. Edwards, J. C., Scalzo, T. R., Storm, D. A., and Bruno, J. W., J . Cutul. 132, 498 (1991). 56. Cugini, A. V., Utz, B. R., Krastman, D., Hickey, R. F.. and Balsone, V., Prepr. Am. Chem. Soc. Div. Fuel Chem. 36, 91 (1991). 57. Hirschon, A. S . , and Wilson, R. B., Jr., Prepr. Am. Chem. Soc. Div. Fuel Chem. 36, 103 (1991). 58. Curtis, C. W., and Pellegrino, J. L., Energy Fuels 3, 160 (1989). 59. Snape, C . E . , Bolton, C., Dosch, R. G., and Stephens, H. P . , Energy Fuels 3, 421 ( 1 989). 60. Pradhan, V. R . , Herrick, D. E., Tierney, J. W., and Wender, I., Energy Fuels 5 , 712 (1991). 61. Suzuki, T., Yamada, O., Fujita, K., Takegami, Y., and Watanabe, Y., Fuel 63, 1706 ( 1984). 62. Suzuki, T., Ando, T., and Watanabe, Y., Energy Fuels 1, 294 (1987). 63. Suzuki, T., Yamada, H . , Yunoki, K., and Yamaguchi, H., in “Proceedings, 1991 Internation Conference on Coal Science, Newcastle upon Tyne, 1991” p. 703. 64. Holloway, P. H . , and Nelson, G. C . , Prepr. Am. Chem. Soc. Div. Fuel Chem. 26, 1352 ( 1 977). 65. Takemura, Y . , Satio, Y., Okada, K., and Koinurna, Y., Energy Fuels 3, 342 (1989). 66. Curtis, C. W . , and Cahela, D. R.. Energy Fuels 3, 168 (1989). 67. Ryan, T. A , , and Stacey, M. H., Fuvl63, 1101 (1984). 68. Mendez-Vivar, J., Campero, A , , Livage, J., and Sanchez, C., J . Non-Cyst. Solids 121, 26 (1990). 69. Naumann, A. W., Behan, A. S., and Thorsteinson, E. M . , in “Proceedings, 1982 International Conference on Chemistry and Uses of Molybdenum” (Ed: H. E. Bany, and C. N. Mitchel, Ed.), p. 313. 70. Yokoyama, S . , Narita, H., Okutani, T.. Kodaira, K., Yoshida, R . , and Maekawa, Y., J . Fuel Soc. Jpn. 62, 966 (1983). 71. Paradhan, V. R . , Herrick, D. E., Tierney, J. W., and Wender, I . , Energy Fuels 5 , 712 (1991). 72. Burgess, C . E . , and Schobert. H. H., Fuel 70, 372 (1991). 73. Cowans, B. A., Haltiwanger, R. C., and DuBois, M . R., Orgunometullics 6, 995 (1987). 74. Joseph, J. T., Fuel 70, 459 (1991). 75. Pelofsky, A. H . , Ed., “Coal Conversion Technology,” ACS Symp. Ser., Vol. 110. Am. Chem. Soc., Washington, DC, 1979. 76. Whitehurst, D. D., Ed., “Coal Liquefaction Fundamentals,” ACS Symp. Ser., Vol. 139. Am. Cheni. Soc., Washington, DC, 1980. 77. Charcosset, H . , Bacaud, R., Besson, M . , Jeunet, A , , Nickel, B . , and Oberson, M., Fuel Process. Technol. 12, 189 (1986). 78. Derbyshire, F. J., Davis, A , , Line, R., Satansberry, P. G., and Terrer, MT., Fuel Process. Technol. 12, 127 (1986). 79. Derbyshire, F . J., “Catalysis in Coal Liquefaction: New Directions for Research,” IEA Coal Research, London, 1988. 80. Chillingworth, R . S . , Hastings, K . E., and Potts, J. D., Ind. Eng. Chem. Prod. Res. Dev. 19, 34 (1980). 81. Takase, S., Ushio, M . , Oishi, Y., Morita, T., and Shioiri, T., Prepr. Am. Chem. SOL.. Div. Per. Chem. 27, 697 (1982).

CATALYSIS IN COAL LIQUEFACTION

83

82. Sebenik, R. F., and Ference, R. A , , Prepr. Am. Chem. Soc. Div. Pet. Chem. 27, 674

(1982). 83. Quarderer, G . J., and Moll, N. G., US patent 4, 102, 775 (1978). 84. Okuma, 0.. Saito, K., Kawashima, A , , Okazaki, K . , and Nakako, Y., Fuel Process. Techno/. 14, 23 (1986). 85. Mochida, I., Takarabe, A,, and Takeshita, K., Fuel 58, 17 (1979). 86. Mochida, I., Iwamoto, K., Tahara, T., Korai, Y., Fujitsu, H., and Takeshita, K., Fuel 61, 603 (1982). 87. Derbyshire, F. J., and Whitehurst, D. D., Fuel 60, 655 (1981). 88. Bockrath, B . C., Finseth, D. H., and Illig, E. G., Fuel Process. Technot. 12, 175

(1986). 89. Mochida, I . , Takayama, A., Sakata, R., and Sakanishi, K., Energy Fuels 4, 81 (1990). 90. Baldwin, R. M., and Vinciguerra, S . , Prepr. Am. Chem. SOC. Div. Fuel Chem. 27, 254

(1982). 91. Lambert, J. M., Fuel 61, 777 (1982). 92. Derbyshire, F. J., Davis, A., Schobert, H., and Stansberry, P., Prepr. Am. Chem. Soc. Fuel Chem. Div. 35 ( I ) , 51 (1990). 93. Mitchell, G . D., Davis, A,, and Derbyshire, F. J., in “Proceedings, International Conference on Coal Science, 1989,” p. 75 1. 94. Attar, A,, and Martin, J. B . , Prepr. Am. Chem. Soc. Div. Fuel Chem. 26, 73 (1981). 95. Miki, Y., and Sugimoto, G . , J . Fuel Soc. Jpn. 62, 408 (1983). 96. Comolli, A. G . , MacArthur, J. B., and Stotler, H. H . Prepr. Am. Chem. Soc. Div. Fuel Chem. 27, 104 (1982). 97. Montano, P. A,, and Granoff, B., Fuel 59, 214 (1980). 98. Montano, P. A,, Prepr. Am. Chem. Soc. Div. Fuel Chem. 28, 169 (1983). 99. Mochida, I., Moriguchi, Y., Shimohara, T., Korai, Y., Fujitsu, H., and Takeshita, K., Fuel 61, 1014 (1982). 100. Kang, C. C., and Johanson, E. S . , “Liquid Fuels from Coal” (R. T. Ellinton, Ed. ).

Academic Press, New York, 1977. 101. Stephens, H. P., Stohl, F. V., and Padrick, T. D., in “Proceedings, International Con-

ference on Coal Science, 1981,” p. 368. 102. Rosenthal, J. W., Dahlberg, A. J., Kuehler, C. W., Cash, D. R., and Freedman, W., Fuel 61, 1045 (1982). 103. Kovach, S . M., and Bennelt, J., Prepr. Am. Chem. Soc. Div. Fuel Chem. 20, 143

(1975). 104. Derbyshire, F., Davis, A , , Schobert, H . , and Stansberry, P., Prepr. Am. Chem. Soc. Div. Fuel Chem. 35 ( I ) , 51 (1990). 105. Strobel, B. O . , and Frieindrich, F., “Proceedings, International Conference on Coal

Science 1989,” p. 735. 106. Okuma, O., Yanai, S . , Tachibana, S . , Hirano, T., and Ida, T., in “Proceedings, Interna-

tional Conference on Coal Science, 1989,” p. 701. 107. Satoh. T., in “Proceedings, 7th AIST-NEDO/DOE-PETC Joint Technical Meeting, Coal

Liquefaction, 1990,” p. 27. 108. Gollakota, S. V., Vimalchand, P., Davies, 0. L., Lee, J. M., Corser, M. C., and Cantrell, C . E., in “Proceedings, 8th AIST-NEDOIDOE-PETC Joint Technical Meet-

ing, Coal Liquefaction, 1991,” p. 23. 109. Sakanishi, K., Honda, K., Sakata, R., and Mochida, I . , in “Proceedings, 5th Aus-

tralian Coal Science Conference, 1992,” p. 97. 110. Sakata, R., Sakanishi, K., and Mochida, I., in “Proceedings, International Conference on Coal Science, 1991,” p. 707.

84

ISAO MOCHIDA AND KlNYA SAKANlSHl

111. Ward, J. W., and Qader, S. A., Eds., “Hydrocracking and Hydrotreating,” ACS Symp.

Ser., Vol. 20. Am. Chem. Soc., Washington, DC, 1975. 112. Whitehurst, D. D . , Mitchell, T. O., and Farcasiu, M., “Coal Liquefaction,” Academic

Press, San Diego. 1980. 113. Sullivan, R . F., Prepr. A m . Chem. SOL*.Div. Fuel Chrm. 31 (4),280 (1986). 114. Sullivan, R . F., and Frumkin, H. A.. Am. Chrm. Soc. P r e p . Div. Fuel Chem. 31 (2),

325 (1986).

115. Sakanishi, K., J . Inst. Energy Jpn. 72, 354 (1993). 116. Masuyama, T.,Kageyama, Y., and Kawai, S. , Furl 69, 245 (l9YO). 117. Mochida, I., Sakanishi, K., Oishi, T., Korai, Y., and Fujitsu, H., Fuel Process. Techn o / . 10, 91 (1985). 118. Mochida, I., Sakanishi, K., Korai, Y., and Fujitsu, H.,Chem. Lett., 909 (1983. 119. Mochida, I . , Sakanishi, K., Oishi, T.. Korai, Y., and Fujitsu, H., i n “Proceedings. Interantional Conference on Coal Science, 19x5,” p. 63. 120. Mochida, I., Sakanishi, K., Korai, Y., and Fujitsu, H., Fuel 65, 633 (19x6). 121. Nishijima, A.. Shimada, H., Sato, T., and Yoshimura, Y., in “Proceedings, Interna-

tional Conference on Coal Science, 1985,”p. 201.

122. Sakanishi, K., Ando, M., Abe, S., and Mochida, I., J . Pet. lnsr. Jpn. 34, 553 (1991). 123. Mochida, I., Sakanishi, K., Kordi, Y., and Fujitsu, H., Fuel 65, 1090 (1986). 124. Mochida, I., Sakanishi, K., Kordi, Y., and Fujitsu, H., J . Chum. Soc. Jpn.. 1212 ( 1986).

125. Mochida, I . , and Sakanishi, K., Fuel 66, 1584 (1987). 126. Sakanishi, K . , Zhao, X . Z., Fujitsu, H., and Mochida, I., Fuel Proc~ess.Techno/. 18,

71 (1988).

127. Fujitsu. H., Ikeyama. N., and Mochida, I., J . Cutul. 100, 279 (1986). 128. Sakanishi, K . , Zhao, X . Z . , Fei, Y. Q . , Korai, Y., Fujitsu, H., and Mochida. I., Fuel Process. Technol. 20, 233 (1988). 12Y. Cable, T. L . , Massoth, F. E., and Thomas, M. G., Fuel Proc.e.rs. Technol. 10, 105

(1985).

130. Stiegel, G. J., Tisher. R. E., and Polinshi, L. M . , Ind. Eng. Chem., Prod. Res. D e v .

22, 41 I (1983). 131. Yoshimura, Y . . Shimada, H., Sato, T., Kubota, M., and Nishijima, A , , Appl. Curul. 29,

125 (1987). 132. Fleisch, T. H . , Meyers, B. L., Hall, J. B., and Ott. G . L.. J . Cutul. 86, 147 (19x4). 133. Mochida, I., Kudo. K., Fukuda, N., Takeshita, K . , andTakahashi. R . , Curhon 13, 135

(1975). 134. Bond, G. C., “Catalysis by Metals,” p. 99.Academic Press, New York, 1962. 135. Hegedus, L. L., and McCabe, R . W., “Catalyst Deactivation” (B. Delmon and G . F. Froment, Eds. ), p. 471.Elsevier, Amsterdam/New York, 1980. 136. Wynblatt, P., and Glostein, N. A.. Progr. Solid Stute Chem. 9, 21 (1975). 137. Wanke, S. E., and Flynn, P. C . , Curd. Rev. Sci. Eng. 12, 93 (1975). 138. Topsde, H., Clausen, B. S., Candic. R.. Wivel, C., and Morup, S . , J . Cutol. 68, 433 (1981). 13Y. Vogelzang, N. W., Li, C.-L., and Schuit, G. C. A,, Gates, B. C., Petrakis, L., J . Cutul. 84, 170 (19x3). 140. Ralston, D., Govek, M . , Crof, C., Jones, D. S., Totake, K., Klabunde, J. K., Woobsey, N. F., Baltisberger, R . J., and Stenverg, V. I . , Fuel Process. Rchnol. I, 143

(1977/1978).

141. Mochida. I . , Sakanishi, K., Kishino, M.. Honda. K . , Umezawa, T., and Yoon, S. H . , Prepr. Am. Chrm. Soc. Fuel Chem. Div. 38 ( I ) , 93 (1993).

CATALYSIS IN COAL LIQUEFACTION

85

142. Kageyama, Y., and Masuyama, T., in “Proceedings, International Conference on Coal Science, 1985,” p. 157. 143. Mochida, I . , Oishi, T., Korai, Y., Fujitsu, H., and Takeshita, K . , Fuel Process. Technol. 7, 109 (1983). 144. Thakur, D. S . , and Thomas, M. G., Ind. Eng. Chem.. Prod. Res. Dev. 23, 349 (1984). 145. Crynes, B. L., “Chemistry of Coal Utilization, Second Supplementary Volume” (M. A . Elliott Ed. ), p. 1991. Wiley, New York, 1981. 146. Mochida, I., Moriguchi, Y., Shimohara, T., Korai, Y., Fujitsu, H . , and Takeshita, K., Fitel 62, 471 (1983). 147. Mochida, I . , Yufu, A , , Sakanishi, K . , Zhao, X . Z . , Okuma, O., and Hirano, T., J . Fuel Soc. Jpn. 68, 244 (1989). 148. Sakanishi, K., Sakata, R., Honda, K . , and Mochida, I., Submitted for publication. 149. Mochida, I . , Sakata, R., and Sakanishi, K., Fuel 68, 306 (1989). 15’0. Sakanishi, K., Zhao, X . Z. Mochida, I., and Okuma, O., 1. Fuel Soc. Jpn. 69, 267 ( 1990). 151. Mochida, I . , Sakanishi, K . , Usuba, H., and Miura, K . , Fuel 70, 761 (1991). 152. Serio, M. A . , Solomon, P. R., Kroo, E., Bassilakis, R . , Malhotra, R . , and McMillen, D. F., Prepr. Am. Chem. Soc. Fuel Chem. Div. 35 ( I ) , 61 (1990). 153. Larsen, J. W., Pan, C.-S., and Shawver, S. , Energy Fuels 3, 557 (1989). 154. Baldwin, R. M., Kennar, D. R., Nguanprasert, O., and Miller, R. L., Fuel 70, 429 (1991). 155. Olah, G. A . , and Husain, A , , Fuel 63, 1427 (1984). 156. Olah, G. A . , Bruce, M. R., Edelson, E. H.. and Husain, A., Fuel 63, 1130 (1984). 157. Heredy, L. A , , and Neuworth, M. B., Fuel 41, 221 (1962). 158. Shabtai, J., Oblad, H. B., Kataydma, Y., and Saito, Y., Prepr. A m . Chem. Soc. Div. pet. Chem. 30 (3), 495 (1985). 159. Shabtai, J.. Skulthai, T., and Saito, Y., Prepr. Am. Chem. Soc. Div. Fuel Chern. 31 (4). 15 (1986).

This Page Intentionally Left Blank

ADVANCES IN CATALYSIS, VOLUME 40

Advances in Applied Electrocatalysis HARTMUT WENDT, SVEN RAUSCH,

AND

THOMAS BORUCINSKI

lnstitut ,fur Chrmische Technologic Technische Hochschule Durmstudt 0-64287 Durmstudt, Germany

1.

Introduction

A. WHATIs ELECTROCATALYSIS? Although the word electrocatalysis was not introduced into the language of the electrochemical community earlier than 1963 ( I ) , electrocatalysis is a topic nearly as old as electrochemical science. Electrocatalysis means the catalytic enhancement of electrochemical reaction rates which is manifested in reduced overpotentials. It is due to the catalytic acceleration of slow chemical rather than relatively fast charge transfer steps which together constitute an electrochemical reaction. Electrocatalysis in the context of electrochemical engineering science and technology does not have a comparably long history, dating back to the early seventies in the context of fuel cell development (2) and to the revolution in the chloralkali electrolysis industries brought about by Beer’s introduction (3) of dimensionally stable anodes (DSAs). Electrodes covered by electrocatalytically active coatings were introduced into the practice of electrochemical production on a large scale ( 4 , 5 ) less than 30 years ago. It should be mentioned that Beer’s introduction of DSAs was not at first saluted enthusiastically or appreciated by the practitioners in the field, but it did last some time until the dual advantages of Ru02-activated titanium anodes-dimensional stability and substantially reduced anodic overpotential-were fully realized, accepted, and exploited. Since that time, further advances have been made in the field of technical electrocatalysis, not the least because during the last 30 years fuel cell technology began to develop with ever-increasing speed quite independently from usual electrochemical engineering practice. Traditional efforts in electrochemical engineering were oriented more toward supplying robust and long-lasting equipment for electrochemical production of goods and commodities than to catalyzing electrode reactions to obtain the highest possible energy efficiency, which is the ultimate aim of fuel cell development and of electrochemical energy conversion. 87

Copyright 0 1994 by Academic Presa. Inc. All rights of reproduction in any form reserved.

88

HAFTMUT WENDT, SVEN RAUSCH, A N D THOMAS BORUCINSKI

A second important step following replacement of the chlorine evolving carbon anodes by DSAs was the introduction of catalyst-coated cathodes in chloralkali electrolysis. However, the advancement of technical electrocatalysis and its application to other processes has been sluggish. Only now is O2 evolution catalysis for 02-evolving anodes being introduced in metal electrowinning, and the replacement of cathodic H2 evolution by energy-saving electrocatalysed 0 2 reduction in choralkali electrolysis is still far from technical realization.

B. AIMOF THE ARTICLE This chapter does not intend to go into the details of the fundamentals of electrocatalysis since a number of articles and proceedings volumes on this topic have been published (6-14). The aim of this chapter is to address the problems involved in application and exploitation of electrocatalysis in technical processes, describing electrode preparation procedures, and material problems involved, and simultaneously stressing the electrochemical engineering implications. However, it should be mentioned that new ways to understand the molecular details of charge transfer at the electrode/electrolyte interface in general and electrocatalysis in particular are opening due to a variety of factors. These include the rapid development of surface spectroscopy, the application of electron synchroton X-ray absorption and microiinaging techniques (such as scanning tunnelling microscopy), and the possibility of modeling the fundamentals of electrochemical processes (namely the structure of the inner part of the double layer by ultrahigh-vacuum surface investigations). However, one cannot yet expect an immediate impact of these methods, on the development of practical catalyst systems, as they are too immature to possess predictive power. This is a familiar situation in applied catalysis, where the practitioners are always far ahead of the theorists .

c. 1.

ELECTRODE KINETICS AND ELECTROCATALYSIS

Kinetics of Interjaciul Charge Trunsfer

For an electrochemical reaction A

+ u,e-

=

B,

the rate of electrochemical conversion is determined by the applied current r = -(dNA/dt)

=

O'l/(u,F).

0, is the stoichiometric number of electrons and 8' is the current efficiency for the considered reaction.

ADVANCES IN APPLIED ELECTROCATALYSIS

89

It is usual to relate electrochemical reaction rates and currents to the unit surfice, thus obtaining surface-specific reaction rates and current densities i as a measure of electrochemical reaction rates: i =

]/A

=

(dN/dt)v,F(O')-'/A.

(2b)

It is an experimental fact that whenever mass transfer limitations are excluded, the rate of charge transfer for a given electrochemical reaction varies exponentially with the so-called overpotential q , which is the potential difference between the equilibrium potential Eo and the actual electrode potential E (7 = E - Eo). Since for the electrode reaction Eq. (1) there exists a forward and back reaction, both of which are changed by the applied overpotential in exponential fashion but in an opposite sense, one obtains as the effective total current density the difference between anodic and cathodic partial current densities according to the generalized Butler-Volmer equation: i = io (exp {+a+Fq/RT} - exp {-a-F?/RT}).

(3)

iO represents the exchange current density, which means the current density exchanged back and forth under equilibrium potential. a+ and a - represent the anodic and cathodic charge-transfer coefficients respectively, which determine the relative response of the respective partial current densities toward a change in overpotential 7 . For a given overpotential, the effective current density depends on the magnitude of the charge-transfer coefficient a as well as on the exchange current density io. If the overpotential is high enough-that is, if either -(a-Fq/RT) or (a+Fv)/RT % 1-then one of the partial current densities in the Butler-Volmer equation overrules the other: i = iClexp{a+Fq/RT}

for

afFq/RT % 1

for

-a-Fr]/RT Z=- 1.

(4)

or i = io exp{-a-Fr)/RT}

(54

From these simplified equations the so-called Tafel equation q =a

+ b log i

(5b)

is derived with

a

= -(RT/aF)lg

io

and

b

=

2.301 (RT/aF).

(54

The value b represents the Tafel slope, which for a = p = 0.5 amd 300 K amounts to 120 mV. Figure l a explains schematically in a linear and a semilogarithmic plot of current densities i versus overpotentials how charge-transfer coefficients a are determined by extrapolating the slopes of the semilogarithmic current

90 a

N -

HAFTMUT WENDT, SVEN RAUSCH, A N D THOMAS BORUCINSKI

b

3

2

400

.

300

.

9

--

200 .

E

Y

.-

h

0.

0

2

catalyzed

c

Y

100

-1 .

-2 -200

. ,,

0 -100

0

200

100

0

-1

fl I (mV)

1

2

3

log(i) (mA/cm2)

-1

0

1

2

3

log(i) (mA/cm2)

FIG. 1. (a) Extrapolation of semilogarithmic current voltage curves to the equilibrium potential, E,,, for determining the exchange current density i(,. (b) Increase of the exchange current density due to electrocatalysis. (c) Access of alternative mechanisms with decreased overpotential at high current densities due to electrocatalysis with the consequence of reduced slope of the semilogarithmic current voltage curve.

voltage curves (log i vs v) toward high cathodic and anodic current densities where either the cathodic or the anodic current alone determines i . Log io is obtained by extrapolating these slopes to the equilibrium potential (q = 0). 2 . Principles oj Electrocatalysis Simple charge-transfer reactions, like the one-electron oxidation or reduction of a dissolved solvated metal ion to a solvated ion of higher or,lower oxidation state, most commonly exhibit charge-transfer coefficients of ca. 0.5 and often possess only moderately low activation energies so that the exchange current densities io are high and may range from to-’ to 10 Akm’ [compare the Marcus theory ( 1 5 ) ] . Whenever the electrochemical reaction is more involved, that is, when it is composed of at least two charge-transfer steps with at least one intermediate chemical reaction, the reaction is likely to be kinetically hindered to a significant degree. This results in strongly re-

ADVANCES IN APPLIED ELECTROCATALYSIS

91

duced exchange current densities. Simple charge exchange, e.g., between dissolved partners of a redox system and the electrode, is usually relatively fast and therefore is only moderately sensitive to catalytic acceleration, for instance, by complex formation of the involved ions. Hindered intermediate chemical reactions in multielectron electrode reactions, however, usually being hindered by chemical kinetics, can often be strongly accelerated by means of heterogeneous catalysis at the electrode surface which affects the rate of the chemical step. 3. Electrocatalysis and Formal Electrode Kinetics

Electrocatalytic action of the electrode material may have two different consequences, both leading to reduced overpotential at higher curent densities: 1. The mechanism of the overall electrode reaction remains unchanged but accelerates, creating an increased exchange current density io (Fig. lb). 2 . The electrocatalyst accelerates the overall reaction by opening a new reaction pathway. This is usually manifested by a change in the chargetransfer coefficient, which is usually increased, leading to reduced overpotentials at higher current densities, although very often the exchange current density is lower than that in the uncatalyzed case (Fig. lc).

D.

CHEMICAL CATALYSIS AND

ELECTROCATALYSIS

Today we have an improved understanding of electrocatalysis comprising not only the mechanisms and means for electrocatalytic enhancement of chlorine and oxygen evolution (6, 7, 12, 13) but also of cathodic hydrogen generation (8, 14) and its anodic consumption and cathodic oxygen reduction. Also the role of electrocatalysis in electrochemical organic synthesis is currently understood to an extent that permits drawing technical consequences and finding practical applications. The chemistry of electrochemical reaction mechanisms is the most hampered and therefore most in need of catalytic acceleration. Therefore, we understand that electrochemical catalysis does not, in principle, differ much fundamentally and mechanistically from chemical catalysis. In addition, apart from the fact that charge-transfer rates and electrosorption equilibria do depend exponentially on electrode potential-a fact that has no comparable counterpart in chemical heterogeneous catalysis-in many cases electrocatalysis and catalysis of electrochemical and chemical oxidation or reduction processes follow very similar if not the same pathways. For instance as electrochemical hydrogen oxidation and generation is coupled to the chemical splitting of the HZmolecule or its formation from adsorbed hydrogen atoms, respectively, electrocatalysts for cathodic hydrogen evolution-

92

HARTMUT WENDT, SVEN RAUSCH, A N D THOMAS BORUClNSKl

or its oxidation-and catalysts for chemical hydrogenation are essentially the same: platinum and the transition metals of the eighth side group of the periodic table. For catalysis and electrocatalysis, the same correlation of catalytic activity and hydrogen adsorption enthalpies at the respective metal catalyst in the form of the so-called volcano curve prevails (Fig. 2). This is also understood as being due to competition of formation of adsorbed H atoms [Eq. (6a)l as a reaction intermediate, which is enhanccd by strong Me-H interaction and its desorptive electrochemical dimerization [Eqs. (6b)-(6c)] to form hydrogen molecules, which is correspondingly retarded by strong adsorption. This leads to an optimal catalytic activity of those metals that adsorb hydrogen but only to a moderate extent and are thus able to assist Hz splitting without already hindering its desorption: H?O t c

+ HJO + e -

H,$
+-

H,d

+ OH-

(64

H

Hz

+ OH-.

(6b)

The reaction sequence of the Volmer rcaction (6a) and the Heyrovsky reaction (6b) is not the only one possible; desorption of the adsorbed hydrogen may also proceed according to the so-called Tale1 reaction (6c) by chemical desorptive dimerization of two adsorbed hydrogen atoms fix which, however, the same fundamental considerations prevail (6c)

2H.,,1-+ 2H2.

Likewise, the electrochemical reactions of oxygen, which is the electrochemical anodic oxygen evolution and cathodic reduction on one hand and

-2

1

Pt

transition metals

-5

.

-8 (Pb.Cd,TI, In)

N b,Ta)

-11 100

200

300

400

500

FIG. 2. Volcano curve in the presentation of H:-exchange current densities, i,,, vs Hzadsorption enthalpies of different metals.

ADVANCES IN APPLIED ELECTROCATALYSIS

93

anodic oxidation of organics on the other hand, are catalyzed by the same metal oxides or metals respectively. These catalysts/electrocatalysts undergo redox conversion relatively easily with oxygen or are oxidized anodically to a higher valency state at potentials that are close to the oxygen equilibrium potential, in this way facilitating the release or transfer of oxygen from oxidized surface groups of the catalyst to organic substrates or leading to anodic release of molecular oxygen: Me

+ HzO

-

e-

c----f

MeOH

MeOH

-

e-

++

Me0

2MeO

+-+

Me 4-

+ H’

(74

+ H’

(7b)

0 2 .

(74

Electrocatalysis of electrochemical chlorine evolution from Ru02-coated titanium anodes rests on the same grounds according to the so-called Krasil’shchikov mechanism, which is schematically described by the reactions (8a) and (8b) for anodic chlorine evolution (9-12): MeOH

+ C1-

+-+

2MeOHCI

E.

-+

MeOHCl

+e

2MeOH t Clz.

(84 (8b)

ELECTROCATALYST PARTICLES

UTILIZATION OF POROUS

In chemical heterogeneous catalysis, it is common to use highly porous catalysts that come in particles of millimeter to centimeter size to increase the effective catalyst surface. In practical electrocatalysis, in particular applying electrocatalysis in fuel cells, it is also usual to use highly porousalthough accounting for the low diffusion coefficients in liquid electrolytes compared to gases, lo-’ cm’/sec vs 1 cm2/sec, much smaller-catalyst particles. A question of concern regarding heterogeneous catalysts used in chemical catalysis and porous electrocatalytic coatings is the problem of effective catalyst utilization. Whenever electrodes coated by porous electrocatalyst layers are used to enlarge the effective catalytic surface of the electrode by using the inner surface of the porous coating, the electrochemical engineer is confronted with mass-transfer hindrance of reactants or products into or out of the porous catalyst layer or catalyst particle, respectively. This renders retardation of the electrode kinetics by concentration polarization with the consequence of only partial utilization of the internal surface of the catalyst layer. If the thickness of the porous catalytic coating or the diameter of the porous particle does not properly match the mean diffusion length of consumed or produced species, then the inner surface of the catalyst is no longer supplied sufficiently with substrate or it becomes electrochemically inactive

94

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

0.1

10

1

100

1000

FIG. 3. Utilization of porous catalysts in dependence on the Thiele modulus &,,

due to product accumulation and concentration polarization in the pores. The chemical engineer applies the Thiele modulus as a characteristic adimensional quantity defining the degree of utilization of a porous catalyst particle, Thiele modulus

=

Qrn = L ( k / D ) ’ / ’ ,

(94

where L is the pore length, coating thickness, or radius of porous particle; k is the volume-related rate coefficient of first-order reaction; and D is the effective diffusion coefficient of the reactant in the porous matrix. Figure 3 demonstrates in a double-logarithmic plot the dependence of the catalyst utilization on %, showing that for %, = > 3 , the utilization 7 decreases according to a limiting @-‘-law. For a given overpotential, the quantity i(v)varies approximately linearly with the exchange current density io. It is shown below (Section B.3.c) that for highly porous Raney nickel coatings with io equaling roughly A/cm2, an overpotential of 100 mV and a pore length L = 1 p m , @TI, approaches or exceeds 1 and therefore deeper pores (or thicker coatings) are no longer fully utilized. With platinum metals, which are much better catalysts than nickel (iO > lo-* Akm’), a dense but very thin (d < 1 p m ) coating can be applied instead of a porous coating because the natural roughness of smooth coatings (1 p m ) already matches the effective pore depth. The definition of the electrochemical Thiele modulus [Eq. (9b)l characterizing the degree of electrocatalyst utilization is a prerequisite for properly tailoring the micromorphology of porous electrocatalytic electrode coatings and fuel cell electrodes, as it allows matching of the coating or catalyst particle dimension to the catalytic activity of the material: Electrochemical Thiele modulus =

@Th.el

= L[i(q)a/DveFc]’/*,(9b)

ADVANCES IN APPLIED ELECTROCATALYSIS

95

with i(7) the current density at overpotential, 7;u the inner electrode surface per unit volume of porous electrocatalyst; c the concentration of reac, stoichiometric number of electrons. tant in the electrolyte bulk; and t ~ the F.

FUNCTIONING, LONGEVITY, AND APPLICATION OF ELECTROCATALYST COATINGS

Electrocatalysts that are composed of highly active precious metals or oxide ceramic materials of relatively low electric conductivity (e.g., semiconducting oxides) are usually applied to metal supports, constituting the electrode proper, in the form of thin coatings amounting to catalyst loadings of only some milligrams per square centimeter or less. Long-term stability and long-term adhesion to the support of such coatings are as important as their catalytic activity. The longevity of electrocatalyst coatings might be threatened by a number of different effects. The four most important are (i) poisoning, (ii) corrosion, (iii) erosion, and (iv) loss of internal surface of porous coatings by sintering and Ostwald ripening. Poisoning can only be reliably mitigated by using carefully purified electrolytes, raw materials, and (in fuel cells) anode and cathode gases. Corrosion is most often a serious problem under nonsteady conditions; for instance, in off times, redox reactions of the catalyst surface that are due to floating and uncontrolled potential render the catalyst material thermodynamically unstable or more soluble in the electrolyte than under operation potential. For instance, cobalt in oxides of the 111 valent state (e.g., in spinels like NiCo204or C o 3 0 4 ) is insoluble in aqueous, caustic solution, but this typical oxygen-evolution catalyst becomes unstable at more cathodic potential reacting with water by release of 0 2 and forming the ll valent state, which is soluble due to formation of the complex Co(0H)i- (17). Erosion by itself and coupled to corrosion may become a serious problem whenever electrochemical gas evolution is the main or a side reaction. Erosion is also a problem if the electrode is exposed to rapidly flowing electrolyte-in particular, if the electrolyte is contaminated by dispersed solid matter. It can only be counteracted by modifying the structure and composition of the coating to increase its strength and adherance and by influencing gas bubble formation in a way that removes the physical strain due to bubble overpressure and violent two-phase flow. Most reliably, this can be achieved by displacing bubble formation off the catalyst surface in the electrolyte solution or by avoiding bubble formation in pores by reducing the mean pore radius of porous coatings below a critical value. Loss of internal surface of highly porous coatings by slow sintering can sometimes be prevented to a certain extent by stabilizing the porous structure by so-called internal or external dispersion hardening, that is, by interdispersion of a second electrocatalytically inactive, oxidic highly dispersed

96

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

powder that hinders creep or surface diffusion of the metallic catalyst niaterial, which would lead to agglomeration and surface loss. Schultze pointed out ( l K ) that whenever the enhanced catalytic activity of the catalyst is due to so-called active sites, that is, exposed crystal defects or dislocations, these sitcs will only be active long term if process& that would lead to healing or recrystallization and accordingly to deactivation have an activation energy in excess of 100 kJimol. Such high activation energies would yield at 100°C a retardation of any surface relaxation processes to effective relaxation times in excess of several years. Catalysis and electrocatalysis by active sites is usually consistent in catalytically active metals. The surfaces of oxidic redox catalysts, however, which are typically accelerating anodic oxygen or chlorine evolution or are mediating the anodic oxidation of organic substances, act catalytically as a whole not dependent on active sites. Such oxides catalyze the respective reactions with every surface metal atom exposed to the electrolyte. For this type of electrocatalysts, physically reinforcing the coatings and strengthening their adherance to the support is imperative. Application of appropriate interlayers (e.g., SnO, between RuOr and Ti-electrodes) or mixing the catalyst with inert valve metal oxides (such as TiOz, Zr02, or Ta20i) became the method of choice for catalyst strengthening.

G.

SCOPE 01: THE

ARTICI .E

The following paragraphs deal with recent advances in catalyst preparation and electrocatalyst performance, outlining the theoretical and materials science background and the technical and process engineering implications of technical electrocatalysis in the fields of:

I . Anodic chlorine evolution 2. Anodic oxygen evolution 3 . Hydrogen evolution in choralkali electrolysis and water electrolysis 4. Anodic hydrogen oxidation and cathodic oxygen reduction in different types of fuel cells 5. Electroorganic synthesis. Photoelectrochemistry in general and electrocatalysis at semiconductor electrodes in particular are not considered, since in this field too many unknowns and in general a lack of long-term performance and technical experience render the technical relevance of published data still questionable. Furthermore, the technical applicability and practical relevance of photoelectrochemistry are still disputed a great deal, and no case of this type of energy conversion has yet been technically demonstrated.

ADVANCES IN APPLIED ELECTROCATALYSIS

97

II. Electrocatalytically Activated Dimensionally Stable Chlorine-Evolving Electrodes

A.

ANODEREACTION: TECHNOLOGICAL HISTORY

Chloralkalielectrolysis was technically executed more than 100 years ago. The anodic reaction 2c1-

-

e-

+

Clz

(10)

can be and has been performed at oxidic anodes (magnetite anodes), but since the beginning of this century carbon anodes, particularly Achesongraphite anodes, have been used. Neither magnetite nor graphite catalyzes reaction (10) particularly well. Graphite is not stable but is oxidized to CO and COr and is also chlorinated to perchlorinated compounds. In 1958, Henry Beer invented the dimensionally stable anode (19. 20), which consisted of a titanium anode covered with a catalytic layer composed mainly of a mixture of TiO2 and RuOr. Due to the joint efforts of Beer and of the De Nora Company, the dimensionally stable chlorine anode became accepted everywhere and its introduction revolutionized the technology (4) of chloralkali electrolysis. In particular, the exchange of the amalgam technology by membrane technology, which initially was enforced in Japan by legislation (1976 to 1986), profited from the already existing technology of anodic Clz evolution at DSAs (21). B.

ELECTROCATALYSIS A N D SELECTIVITY OF ANODIC AT RUO~-ANODES CHLORINE EVOLUTION

Figure 4 compares the current voltage curves of anodic chlorine evolution measured at graphite and Ru02 anodes in brine solution of a concentration, which is typical for chloralkali electrolysis (300 g NaCl/dm3) (22). The Ru02 anode is at a clear advantage. Without treating the details of the electrocatalytic reaction-which have been dealt with by other authors (11, 23-25)-the mechanism is likely to be similar to that described by Krasil’shchikov (11) [Eqs. (7a)-(7c)] for anodic oxygen evolution. The same type of mechanism is also operative with anodic O2 evolution at RuOz anodes. At pH 2 (the lowest pH value maintained in chloralkali electrolysis anolytes), the equilibrium potential of the oxygen anode is more cathodic than that of the chlorine electrode (+0.96 V vs NHE compared to + I.26 V). Therefore it is of ultimate importance that RuO? catalyzes chlo-

98

:

HAHTMUT WENDT, SVEN RAUSCH, A N D THOMAS BORUCINSKI

: 1

Graphite Anode

I

1

60°C

10

100

1000

-i [mA ~ r n - ~ ] FIG. 4. Current voltage curves for anodic chlorine evolution at graphite and RuOl-coated titanium anodes.

rine evolution to a much higher extent than oxygen evolution. This is not always the case. At lower anolyte pH the oxygen potential becomes still more cathodic (60 mV/pHj. For this reason, particular care must be taken in changing the catalyst formulations with the aim to increase the catalytic activity for the C12 reaction over that for the Or reaction, especially if dilute brine (30 g/dm7 for NaOCl production) or seawater (C12 production on board for disinfection) are electrolyzed. An increase of the CI, selectivity is brought about by adding small amounts of Pt and Ir02 to Ru02. (25). Hine was one of the first scientists to publish reports of systematic investigations concerning DSA-coating preparation (26, 27).

c.

PREPARATION A N D

FORMULATION OF THE COATINGS

The method for the preparation of RuO2 coatings today is still the same as the one patented by Beer. More recent advances are discussed in ( 5 , 26 j . The metals whose oxides form the catalyst coating are dissolved in appropriate form (chlorides, organometallics) in an organic solvent of low volatility forming a so-called “paint.” Typical examples of metal compounds are RuC13, H2PtC16,CoC12 (the latter two may be added to modify the selectivity), and titanium tetrabutylate or tantalum trichloride [these might be added to improve adherance and strength of the coating by forming dispersed Ti02 or TazOs particles or mixed oxides containing TiOz (26 )].

ADVANCES IN APPLIED ELECTROCATALYSIS

D.

99

IMPROVEMENT OF ADHESION AND STRENGTH OF THE

COATINGS Long-term adhesion and resistance against erosion are of utmost importance for the long-term stability of the R u 0 2 coating. Therefore, as a first step of coating preparation, the TiO2 layer of micrometer thickness, which usually covers titanium metal and would prohibit the formation of a lowresistive and tight contact between support metal and coating, must be etched away by oxalic or hydrofluoric acid so that only a very thin oxide layer remains onto which the coating is applied. An oxidic additive to R u 0 2 that assures a good adhesion and an improved conductivity of the interlayer between the titanium metal and the coating, together with a greater physical strength, hardness, and erosion stability and simultaneously prevents the anodic oxidation of the titanium support, which would generate insulating, nonconducting TiOz layers must be supplied. Ti02 added to R u 0 2 by proper formulation of the catalyst paint, which crystallizes in the rutile lattice, forms semiconducting mixed Ru02/Ti02 phases and stabilizes mechanically and chemically the electrocatalyst without significantly affecting its catalytic activity ( 4 , 28). The Ti02 content of the coating may vary from 30 to 70 mol%. Also, Ta205is added to RuO2 as stabilizing component. I r 0 2 , a catalytically active additive, also seems to improve the wear rate (29), in particular for DSAs that are used for I2 evolution. Typical examples of high-boiling solvents are butyl alcohol, turpentine oil, or similar low volatility liquids. The “paint” is applied to the sandblasted, etched titanium electrode by brushing, dipping, or spraying. The solvent is evaporated at 100 to 200”C, and the metal compounds now covering the Ti electrode are converted to the respective oxides by firing the electrodes in air at temperatures ranging from 400°C to 600°C-but preferably 450°C (30, 31). The firing temperature must be kept below 600°C because above this temperature an undesired growth of nonconducting Ti02 interlayers on titanium is unavoidable (32). Transatti (30) has shown that a firing temperature of approximately 450°C is optimal because below this temperature oxide formation is incomplete, rendering CI-containing, less-active, and more-soluble coatings, and above 450°C the mixed oxides lose more and more residual chemically bound water and become more ordered and catalytically less active. The procedure of brushing and firing is repeated several times each time with thorough cleaning of the surface from oxidic dusts till the desired coating load of 10 to 40 rngkm’ is achieved. Quite often the coatings (one to several micrometers thick) are not closed but have microscopic cracks appearing like dried mud or filter cake.

100

H A m M U T WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI TABLE 1 1)rveloptnrnt of Catalytic. Cocifings at ICI Motid Divisiotr ( A i ~ ~ r d i ntog(33)) Formulation

Main Constituent

Commen ts

Mond Pt Ptilr metal

Electroplated Pt metal, Pt or Pt/lr from paint platinking forrnulation Pt electroplated in presencc of organic compounds RuOz/TiOz

Problems with passivation and ad herance

Improved Pt inetal

Mond 1

Mond 2 Mond 3 Mond 5ai5b

RuOL overglaze Ru02iSnOz RuO,/TiOz with addcd refractories

Nonpassive under Hg-cell condts. Coating composition depends o n cell Preferred coating in mercury cell

A report on the development of RuOz coatings is compiled in Tablc 1, which shows the different additives added to paint formulations for obtaining improved coating stability (33).

The change from nondiinensionally stable carbon anodes to dimensionally stable titanium anodes permitted dramatic innovations in cell design, operdtion conditions, and reduction of the energy consumption of chloralkali electrolysis. With metal electrodes, almost any anode design can be realized. Figure 5 depicts a typical anode for mercury cells that is immersed horizontally into the brine and is composed of the anode proper-a RuOz-coated grid or extended metal sheet-supported by a so-called x-bar, that serves as current distributor being fixed to another support structure that carries the electrode and by a thick copper rod that serves as a current collector (protected by so called titanium riser tubes). Figure 6 depicts another type of anode that is fabricated from RuOz-coated extended titanium metal. This double-sided anode is typically used in monopolar diaphragm and membrane cells in which it is fixed at the cell bottom standing upright, an expanding spring allowing adjustment of the anode-cathode distances within the cell ( 3 4 ) .

ADVANCES IN APPLIED ELECTROCATALYSIS

101

FIG. 5 . Typical Ru02-coated Ti-wire anode for chlorine evolution in the amalgam process (courtesy of Heraeus Elektroden GmbH).

So-called zero-gap membrane cells in which cathode and chlorineevolving anodes are touching the cation exchange membrane, which separates the anode from the cathode compartment, are also state of the art (3.5). Bipolar chlorine electrolyzers have also been developed (36 ), for instance, at ICI, an achievement that could only be envisaged due to the introduction of KuOr-coated titanium anodes. F. LIFETIME OF DlMENSlONALY STABLE CHLOKINE-EVOLVING ANODES The lifetime of Ru02-coated anodes has been substantially improved by intelligent coating formulations. At lower current densities (2 to 3 kA/m’), which are typical in diaphragm and membrane cells, the coatings last at least 5 years and even lifetimes of up to 10 years are reported ( 3 7 ) ,ammounting to a production of 300,000 kg CI, per square meter during the whole lifetime. In mercury cells with current densities of 10kAm2, the lifetimes amount to approximately 3 to 4 years as the wear rate of the coating increases approximately linearly with current density. As the wear rate is sup-

102

HARTMUT WENDT, SVEN RAUSCH, A N D THOMAS BORUCINSKI

FIG.6 . Ru02-coated extended metal titanium electrode li)r anodic chlorine evolution in the diaphragm and membrane process (courtesy of Heraeus Elektroden GnibH).

posed to be connected with slow oxidative dissolution of RuOz ( R u 0 4 forination), it is today believed that a substantial improvement in lifetime beyond these values may never be achieved. G . DSAS FOR

CHLORRTt AND PRODUC'rtON

HYPOCHLORITE

Hypochlorite and chlorate production rely on reacting the anodically evolved chlorine in dissolved form with cathodically generated hydroxyl ions due to homogeneous reaction kinetics. In chlorate production the formation of hypochlorite is followed by disproportionation of the hypochlorite to chloride and chlorate. 2CI 2H?O

~

2e

+ 20-

CI2 i 2 0 H 3CIO

+ CL.

dissolved

+

2OH-

--j

CI

--9

2C1

+ H?

t 0CI

t H20

+ CIO,

(anode)

( I la)

(cathode)

( 1 Ib)

(hypochlorite)

(12)

(chlorate)

(13)

ADVANCES IN APPLIED ELECTROCATALYSIS

103

For this reaction the pH value must not exceed 5 , which means that the pH value is remarkably higher than that for evolution of gaseous chlorine (pH 2) in usual chloralkali electrolysis and that the competition of anodic oxygen evolution due to the pH dependence of the oxygen equilibrium potential becomes a serious issue as the chlorine potential is approximately 200 InV more noble than the oxygen potential at pH 5 than at pH 2. Therefore, by addition of Pt, PdO, and IrOz to RuOz, the catalytic activity for anodic O2 evolution is repressed vs Clr evolution ( 2 5 ) . A Pt/Ir02 coating has particularly highy oxygen overvoltage and is recommended as a substitute for Pt/RuOz coatings (38). I -

111.

Oxygen Evolving Anodes A.

TECHNICAL PROCESSES

Oxygen evolution is the counterelectrode reaction of electrolytic hydrogen production from alkaline and in cathodic electrowinning of metals from acid aqueous solution. Examples for the latter are zinc, copper, tin, and lead electrowinning from acidic electrolytes of different composition. Water electrolysis usually is performed in alkaline solutions (30 wt% KOH), whereas the electrolytes used in metal electrowinning are strongly acidic (-1 M H2SO4). Therefore, the electrode materials and electrocatalysts are quite different for these two different types of electrolysis processes. The electrode support tor alkaline water electrolysis, which used to be mild steel, is today stainless steel or nickel for advanced water electrolyzers in which one needs materials of improved corrosion resistance. For oxygen evolution from acidic electrolytes, the traditional anode material used to be lead passivated by in siru-formed PbOz layers. These anodes are today becoming superseeded by introduction of catalyst-coated titanium anodes.

B. ELECTROCATALYSIS OF OXYGEN EVOLUTION IN ADVANCED ALKALINE WATERELECTROLYSIS 1.

Coutings Contuining Cobalt and Iron Oxides

Tseung and Jasem (39)and Trasatti (12) have discussed the catalytic activity of different transition-metal oxides in relation to the redox potential of the respective redox couples of lower-valency/higher-valency-state oxides. The result of these investigations is that cobalt oxides are unique insofar as in different oxides and mixed oxides ( C 0 3 0 4 ,Ni2Co04etc.) the redox potentials of Co(II)/Co(III) and Co(III)/Co(IV) oxides, respectively, which possess a relatively low solubility in alkaline solution as a precondition to forming stable catalyst coatings, are relatively close to the oxygen equi-

104

HARTMUT WENDT, SVEN RAUSCH, A N D THOMAS BORUCJNSKI

librium potential and therefore are able to catalyze anodic oxygen evolution according to Krasil’shchikov’s mechanism. Recently in the framework of an extended CEC-financed program on electrolytic hydrogen production (40),a series of different mixed oxides of the spinel and magetite type containing cobalt was investigated with respect to their effective catalytic activity and longevity. Figure 7 (41) compares the current voltage curves of the anodic oxygen evolution at different tempertures in 30 wt% KOH of RuO. and different cobalt-containing oxide coatings on nickel, which had been prepared by a dippinglfiring procedure siniilar to that used for preparation of RuOz-activated dimensionally stable anodes. Coatings are also reported to be prepared by plasma-spray technology (42). Also anodic formation of cobalt oxides-so-called reactive formation-from metallic cobalt or cobalt coatings are reported by which cobalt oxyhydrates are deposited at a cathode at which oxygen is being reduced (43). According to Fig. 7, RuOz is certainly the catalyst of highest catalytic activity-but it is not stable in caustic solution and dissolves due to anodic formation of RuO.{within several days. All mixed oxides containing Co irrespective of their composition show anodic overpotentials comparable to

z a

rn

700

La, $r, 600

I

5E

500

-b 0,

cn I

700 600

2

500

9 E ---

400

v

400

\

w

,COO,

v

w 300

300

d cn I

600 -

0,

z

m

600 -

I

.

w

01

1

10

100

I000

i / (mA cm *]

FIG.7 . Current voltage curves of the anodic oxygen evolution at different temperatures in 30 wt% KOH at nickel anodes with coatings containing cobalt oxides: (a) L a l , s C o I ~ 5 L ~ 0 3 , (b) Nii12,CoOsLaO3,( c ) Colo4, and, for comparison, (d) RuOz.

ADVANCES IN APPLIED ELECTROCATALYSIS

105

that of the spinel phase Co304.Since it can be shown that the mixed oxides of perovskitic as well as of magnetitic structure are not really stable under operation conditions, it is not unlikely that cobalt magnetite (Co304)is the catalyst that is really being formed on the suface of the metal oxide-coated electrodes and acts as an oxygen catalyst. This oxide, however, is not stable at potentials below approximately +0.5 V vs the reversible hydrogen electrode (see the Pourbaix diagram in Figure 8b (44)] and decomposes (Co304 + 3 0 H - + 3 H 2 0 +. 3Co(OH)1 + $02)forming soluble complexed cobalt(l1). As the dissolved cobalt is deposited at the cathode, it is completely transferred from the anode to the cathode after repeated cycling of the electrolyzer. Cobalt oxide-coated electrodes must be protected against dissolution in off times by applying a protective polarization, which, however, must be maintained at sizeable current density to assure protection. Figure 9, which compares the current voltage curve of in situ coated nickel anodes with Co304 and Fe201 coatings ( 4 3 , shows clearly that iron(II1) oxide is a sizeably better catalyst than cobalt spinel. Iron(II1) oxide (Fig. 8a) is stable down to a potential of +0.3 V. Fe203has, however, a relatively high solubility in aqueous concentrated KOH solution so that due to cathodic iron deposition, which is expected to occur around -200 mV, the catalyst is readily transferred from the anode to the cathode under normal operation conditions. This renders Fe207-aIthough a better catalyst than Co7O4and many mixed oxides containing iron oxides-inherently unstable in alkaline water electrolyzers. It is, however, found that in coatings containing iron and cobalt together both oxides are stabilized by forming mixed oxide phases of still undetermined chemical and phase composition with a lower solubility of iron and cobalt. Under these conditions, the good catalytic activity of iron oxide is preserved and the longevity of the coating is improved to a tolerable level of some thousand hours. Although during the last 10 years more than 70 publications and patents that deal with coatings of mixed transition-metal oxides and their electrocatalytic properties appeared, their use and application on technical scale for commercial water electrolysis has not been reported because of their limited long-term stability. A completely novel approach to technical electrolysis for anodic oxygen evolution from alkaline solution is the use of amorphous metals, i.e. chilled melts of nickelicobalt mixtures whose crystallization is prevented by the addition of refractory metals like Ti, Zr, B, Mo, Hf, and P (46-51). For this type of material, enhanced catalytic activity in heterogeneous catalysis of gas-phase reactions has been observed (51). These amorphous metals are shown to be more corrosion resistant than the respective crystallized alloys, and the oxides being formed at their surfaces often exhibit a higher catalytic activity than those formed on ordered alloys, as shown by Kreysa (52-54).

t

1 I

I I

106

I

I

I , -

u y

."

0

u

f ?

U 4J

1 .

m

VI

J

N

107

ADVANCES IN APPLIED ELECTROCATALYSIS

400

350

-

Smooth nickel

""'_'

150

3

10

100

1000

i / [mA cm.*] FIG.9. Current voltage curves of anodic oxygen evolution in 30 wt% KOH at 90°C from nickel and nickel anodes coated in siru by (a) Co104,(b) Fe,O, , and (c) a mixture of both.

2. In Situ Deposition of Coatings Containing Iron and Cobalt by in Situ Activation qf Nickel Anodes The low but sizeable solubility of iron and nickel oxide in caustic potash, shown by the Pourbaix diagrams in Figs. 8a and 8c, gives rise to the method of so-called in situ activation of oxygen-evolving anodes. Figure 10 depicts the evolution of the anode potential of oxygen-evolving anodes 700

0 u)

600

I, 0,

I 500

> v)

>

E w

v

400

300

200 0

100

200

300

400

FIG. 10. Demonstration of the effect of in situ activating 02-evolving nickel anodes by anodic CoiO, deposition.

108

H A W M U T WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

(0.1 A/cm', 90°C, 30 wt% KOH) with time during a batch experiment demonstrating in situ activation. Evidently the addition of only 10 mg/cm' of cobalt in the form of cobaltous nitrate, which is decomposed by generating soluble Co(0H)rand Co(0H): species, results in reduction of the anodic overpotential (+340 mV) by more than 50 mV. This is due to the anodic deposition of a thin film of electrocatalytically active cobalt spinell, Co304,according to 3 Co(OH)q

--

21,-

C0i04

+ O H + 4H10.

(14)

Addition of more cobaltous nitrate does not reduce the overpotential significantly because a surface coverage of nearly 100% has already been achieved by the initial catalyst loading (10 nig Co/cm'). Saturating the electrolyte with iron(II1) hydroxide (e.g., by addition of aqueous solutions of ferric nitrate) and simultaneously adding cobaltous salts leads to in situ formation of a mixed Fe(lII)/Co(II)/Co(III) deposit, which exhibits catalytic activity comparable to that of Fe304shown by the current voltage curve in Fig. 1 1 . Such mixed oxidic catalyst coatings are composed of very small oxide crystals, which evidently arc dissolved upon current interruption due to dissociative oxide dissolution. The transfer of dissolved metal ions to the cathode followed by cathodic deposition of the metal, however, can be completely prohibited, if the potential of the cathode due to optimal electrocatalysis of cathodic hydrogen evolution proceeds with an over-

400

I

30

100

1000

3000

i (mA/cm2) FIG. I I . Temperature dependence of current voltage curve$ of the anodic oxygen evolution from 30 wt% KOH at nickel anodes coated by mixtures of cobalt and iron oxyhydrates.

ADVANCES IN APPLIED ELECTROCATALYSIS

109

potential of less than - 100 mV, that is, more negative potential than corresponds to the equilibrium potential of the Fe/FezO1electrode in the presence of Fe201-saturated caustic potash. With excellent cathodic catalysis, Fe,Co, 0: anodes, although inherently unstable, therefore become persistent. 3 . Electrocurulysis of the Anodic Oxygen Evolution by Raney-Nickel Coutings Raney nickel is used as an effective and cheap electrocatalyst for cathodic hydrogen evolution (see Section 111.B.3). It may, however, also be used as an electrocatalyst for anodic oxygen evolution because Ni II/Ni IV oxyhydrates (Fig. 8c), which act electrocatalytically, are formed on its oxidized surface. Due to the high porosity of Raney nickel, which creates on oxidation a rough surface with roughness factors approaching 1000, such oxidized Raney-nickel coatings are catalytically more active than oxidized surfaces obtained from smooth nickel electrodes. Fig. 1 1 compares the anodic current voltage curves for O2 evolution at different temperatures at smooth, FeO, /Coo,.-covered and Raney-nickel-coated oxygen evolving anodes. The catalytic activity of the latter two coatings is quite comparable and certainly better and more persistent than that of nickel anodes coated with Co-containing mixed oxides by the spray or dippinghinter procedure. Yeager and co-workers (55) report on the electrocatalytic behavior of nickel black (NiOOH) anodes, but it is questionable whether this type of coating is stable on current interruption.

c.

CATALYST-COATEDTITANIUM ELECTRODES FOR OXYGEN EVOLUTION FROM ACIDSOLUTIONS

Traditionally, anodic oxygen evolution from acid solution-particularly from aqueous electrolytes containing sulfuric acid-has been performed at lead anodes that are passivated and stabilized against corrosion by a selfforming coating of Pb02. Oxygen evolution at PbOz anodes even at current densities not exceeding 100 mAicm’ demands overpotentials of 0.7 V or more. Therefore it is highly desirable to find an electrocatalytically active coating on Ti anodes that allows reduction of the oxygen overpotential offering at the same time an anode substrate, which, being a valve metal, is passivating and not sensitive to corrosive attack. It must be kept in mind, however, that passivation of titanium in sulfuric acid solutions is only safe and stable at lower temperatures.

110

HARTMUT WENDT, W E N RAUSCH, AND THOMAS BORUCINSKI

Scientists of ICI (28) found that in solutions containing approximately 20 g HzS04/liter at temperatures up to 40"C, titanium anodizes rapidly forming a TiOz layer of time-independent thickness of fractions of micrameters. Between SO and 70"C, continuous film growth is observed, whereas the titanium begins to dissolve actively above 70°C. Consequently, working temperatures of metal-winning and metal-plating electrolyses with anodic 0 2 evolution must be operated below 40°C if titanium anodes are intended to be used. Fortunately usually these types of electrolyses are operated at relatively low current densities allowing for relatively low process temperatures. RuOz is an excellent electrocatalyst in acid solutions for anodic oxygen evolution. It is however, unstable, being oxidized and carried away in the form of relatively volatile Ru04 ( 5 6 ) . Therefore the R u 0 2 needs to be stabilized. According to the patent literature, PtiIr oxides with added TiOz and Ta2OS/NbzOcas mechanical stabilizers are today used for anodic oxygen evolution in acid electrolytes. Still, the use of this type of anode is restricted to relative pure electrolytes, for instance, for a tin- or zinc-plating bath or for anticorrosive steel-sheet plating in the automobile industry. A more detailed study by Cominellis and Plattner (32) showed that a catalytic coating composed of lrOz and TazOs shows a lifetime optimum at 70 wt% Ir02/30 wt% TazOs. In contrast to TiO2/RuO2 mixtures, true solid solutions of the oxides are not formed. Also, exchanging tantalum for titanium as base metal improves the stability of the combination supportkoating a great deal. IrOdmixed stabilizing oxide electrode reduces the oxygen evolution overpotential by approximately 0.5 V , sparing more than 30% of the energy costs that would be consumed by using conventional PbOz (Ag) anodes. This is good reason to extend now the use of IrOr-coated anodes for plating processes (57 ) to large-scale electrowinning processes like Cu electrowinning (58) and Zn electrowinning. The membrane technology of water electrolysis, due to the high acidity of the proton-loaded, hydrated cation exchange membrane (Nafion), is essentially water electrolysis in acid solution (59). It cannot dispense with noble metal oxides. IrOz and mixtures of RuOz/Ir02 have been applied succesfully. Because of the high and expensive catalyst loading, a market for this type of electrolyzer has not yet opened up, but the technology has been demonstrated with a small number of 100-kW electrolyzer stacks, for instance, in the Solar Hydrogen Bavaria project (60). In summary i t may be stated that today RuOr/Ir02-coatedtitanium anodes mechanically stabilized by TiOrlTa20s already play a significant role in large electroplating processes operating on relatively clean acid solutions. Their use is beginning also to extend into the field of metal electrowinning processes wherever relative highly purified electrolytes are used, for instance, for cathodic zinc deposition. Whether these anodes may contribute

ADVANCES IN APPLIED ELECTROCATALYSIS

111

significantly to hydrogen and oxygen production by membrane electrolysis depends on a sizeable cost reduction in this technology.

IV.

Electrocatalysis of Cathodic Hydrogen Evolution

A . TECHNOECONOMICAL SIGNIFICANCE OF CATHODIC HYDROGEN EVOLUTION Cathodic hydrogen evolution is the counterreaction to anodic chlorine evolution. It is expected to be superseded by the energy-saving cathodic oxygen reduction (61) only after a decade of further development and only if energy prices continue to increase. The chloralkali electrolysis process is by far the most important source of electrolytically generated hydrogen because hydrogen from alkaline or membrane water electrolysis usually cannot compete with hydrogen from steam reforming followed by shift reaction and PSA purification and therefore is not performed on a large scale (62, 40). Cathodic hydrogen evolution is also of some significance as a counterelectrode reaction in anodic organoelectrosyntheses as, for instance, in the Kolbe electrosynthesis of sebacic acid (63) or the anodic oxidation of toluenes to benzaldehydes (64). In chloralkali electrolysis and conventional water electrolysis, the catholyte is strongly alkaline (cNaOH, cKOH = 30 wt%), whereas in organoelectrosynthesis it is neutral or acidic depending on whether a divided or undivided cell is used. In alkaline water electrolysis and chloralkali electrolysis, the electrolyte is less corrosive than in processes that use acidic electrolytes, and therefore the traditional cathode material was mild steel or stainless steel and was recetly exchanged in advanced electrolyzers by nickel. Since according to Fig. 8a iron is less noble than hydrogen at any pH value and since iron(II1) oxide has a finite solubility in alkaline solutions, neither the necessary longterm stability nor the demanded purity of the electrolyte (in particular with respect to poisoning of cation exchange membrances) can be expected with stainless-steel cathodes. Under the same conditions, nickel is immune and furthermore Ni(0H)r is much less soluble than Fe203.Therefore nickel became the material of choice for electrodes, current collectors, and cell bodies in advanced chloralkali electrolysis technology on the cathode side. For acidic aqueous electrolytes, the cheapest valve metal, titanium, is chosen as cathode material. The stability of this metal, however, is questionable in the presence of fluoride and complexing organic anions, so that particular care should be taken if choosing titanium for organoelectrosyn-

112

HAKTMUT WENDT, SVEN RAUSCH, A N D THOMAS BORUClNSKl

thesis processes performed in organic solvents. Since titanium as cathode material might be accessible to hydrogen enibrittlement, it must be protected by devoted coating technologies against the attack of atomar hydrogen.

B. 1.

ELECTKOCATALYST COATINGS FOR HYL)KO(;EN EVOLUTION FKOM ALKALINE SOLUTION

Technicu11.y Applied Cocrtings

Recently two surveys covered the fundamental and applied aspects of hydrogen evolution electrocatalysis (8, 16 ). Trasatti’s review (a), which is more devoted to fundamental problems, critically collects published data on the electrode kinetics of H? evolution at smooth, roughened, and Rancy Ni, interstitial compounds (carbides), sulfides, oxides, alloys, intermetallic compounds, and amorphous metals. The second review (16 ) concentrates more on the engineering aspect of electrocatalyzed hydrogen evolution. The latter review can be summarized and amended by stating that today the following catalyst coatings are technically relevant:

I . Nickel sulfide (NiS,) 2. Raney nickcl 3 . Nickel alloys containing molybdenum 4. Oxidic coatings containing platinum metal oxides, in particular RuO?, and platinum metal coatings, in particular Ru 5 . Doped nickel oxide. Due to their relatively low cost, the most frequently applied electrocatalysts are nickel sulfide and Raney nickel for alkaline conditions.

2. Nickel SulJde Coatings For a long time it has been industrial practice to activate nickel cathodcs by depositing coatings of nickel sulfide on their surface. Two different methods are used: cathodic deposition of NiS, and chemical sulfidization of nickel electrodes. The older, galvanic deposition proccss is based on cathodic deposition of nickel from baths containing sulfur donors like thiocyanate ion (65-67 ), thiourea (68-70), and thioglycol ( 7 / ) ,according to Ni”

+ .rKS + 2c~-+ NiS, + xK.

(15)

The deposit does not contain sulfur in a well-defined stoichiometry but is a mixture of different nickel sulfides, in particular NizSJ, and amorphous nickel (72). it is not really clear whether the catalytic activity is due to amorphous nickel [compare Kreysa (53)]or to Nils3. A comparison of the

ADVANCES IN APPLIED ELECTROCATALYSIS

113

effective activity of the different nickel sulfides shows that the catalytic activity of the different NiS, phases increases with their sulfur content. NiS, fi is, for instance, much more active than nickel metal and NiSz possesses the highest catalytic activity (73). Certainly the cathodic codeposition of nickel sulfides and amorphous nickel from solutions containing Ni(I1) ions and sulfur donors allows for a relatively broad variation of a number of process parameters (e.g., current density, concentration, temperature), which would be expected to bear on the composition, effective catalytic activity, and endurance of the coatings. But, although practitioners in the industry doubtlessly know optimal deposition conditions, systematic investigations of these problems have not been published yet. Nickel sulfide coatings are therodynamically unstable at the hydrogen equilibrium potential. NiSz decomposes simply by immersion in aqueous KOH. In aqueous alkaline solution at hydrogen-evolving cathodes, all NiS, phases are reduced according to NiS

+ 2r

+

Ni

+ S'--.

(16)

Microscopic and spectroscopic investigations (SEM and XPS) reveal the relatively fast change of the chemical composition of nickel sulfide coatings upon the onset of cathodic hydrogen evolution (74). Indeed, at 90°C all nickel sulfide phases are reduced to porous nickel within several days to a week's time. They lose some catalytic activity with time with an increase in overvoltage bctween 0.15 and 0.3 V after continuous operation for 1 year. It is clear that the catalyst after 1 week is already no longer nickel sulfide but some type of Raney nickel. Thus far the initial catalytic activity of the NiS, coating is of little relevance. The respective results and data are due to be published by the present authors (73). Cathodically released sulfide anions are anodically oxidized to sulfate anions, which enhance surface corrosion and stress corrosion cracking of mild and stainless steels and enhance surface corrosion even of nickel. Therefore, nickel sulfide coatings can only be used in nonpressurized electrolyzers working at relatively moderate temperatures (80°C). 3 . Runry-Nickel Coatings a. Precursor Alloys und Fabrication of Coated Cathodes. Raney nickel, which more than 30 years ago had been introduced as an anodic electrocatalyst for alkaline fuel cells by Winsel and Justi ( 7 3 , has been used for more than 20 years as a catalytic coating for hydrogen-evolving cathodes in chloralkali electrolysis and alkaline water electrolysis (76, 77). The operation conditions of the Raney-nickel anodes in fuel cells and Raney-nickel cathodes in chloralkali electrolysis are almost the same, as the electrolyte contains from 20 to 30 wt% NaOH or KOH resp. and the process tempera-

114

HAKTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

ture is between 70 and 90°C. Raney nickel is a highly porous, pyrophoric nickel, initially invented and applied for catalytic hydrogenation of organic compounds. Raney nickel is obtained from precursor alloys, e.g., from nickel alloyed with a high percentage of nonnoble metals, preferably Al and Zn, which are leached out of the alloys in caustic potash. Due to loss of the nonnoble component, the precursor alloy is converted under partial shrinking and recrystallization into a metal sponge of 20% porosity with an inner surface of 30 to 70 m’/g.The preferred precursor alloys are NiAI, and Ni2AI3and the 6 - phase of Ni/Zn, respectively. Although at least four different technologies [cold rolling, flame spraying, Zn and Al melt dipping, cathodic deposition of Ni/Zn precursor alloys (16 )] have been described, only cold rolling and cathodic deposition of precursor alloys are used for commercial production of Raney-nickel-coated cathodes. Cold rolling of the powdered precursor NiAh (10 to 20 p m particle size), mixed with granules of Mond nickel (carbonyl nickel, which is very ductile) on stainless-steel or nickel foils, creates coatings of 20 to 30% coarse porosity (Up > 10 p m ) , which firmly adhere to the supporting foil (76). Due to the excessive sheer and stress exerted during rolling, the foils are deformed and must be shaped according to demands after this process has been finished. Choosing Ni/Zn ( y - and 6-phases) as a precursor allows galvanical codeposition of nickel and zinc according to anomalous codeposition from acidic electrolytes in the form of layered coatings containing the zinc-poor aphase (25 to 50 mol%) and y - and 6-phases (78). The cathodic deposition of the NiZn precursor has the advantage of being applicable to metal electrodes of any size. The coatings are usually roughly 100 p m thick and have a relatively rough, cauliflower-like surface (Fig. I2a). During caustic leaching the Raney-nickel precursor alloy, which contains 50 mol% or more of leachable metals (A1 or Zn), the precursor shrinks and cracks that subdivide the coating penetrating through its entire extension are formed (Figs. 12b and 12c). These cracks are essential for the optimal utilization of the catalytically active coating. b. Utilization o j the cutulyst in Runey-nickel cmtings. Nickel, compared to platinum metals (e.g., Pt, proper, or Ru), which exhibit exchange current densities of the hydrogen evolution reaction in caustic electrolytes at 30°C of the order of to lo-’ A/cm2, is only a moderately active electrocatalyst with io equaling from 5 to 7. lo-‘ A/cm’ (79,80) at ambient temperature. Therefore, considering a ratio of the exchange current densities of Pt and Ni of 10’ to lo7, a naive estimation would predict comparable catalytic activities of smooth Pt-metal cathodes (see Section IV.B.6) and raney nickel cathodes, provided the catalytically active inner surface of the Raney-nickel

ADVANCES IN APPLIED ELECTROCATALYSIS

115

FIG. 12. Morphology of Raney-nickel-coated cathodes for hydrogen evolution from caustic electrolytes: (a) surface of Ni-Zn precursor coatings, (b) surface of Raney-nickel coating prepared by caustic leaching of the Zn content of the precursor, (c) cut through a Raney-nickel coating.

116

HAWMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

50 urn FIG. 12. (continued)

coating would amount to from 100 to 1000 cm2/cm2electrode surface. A statistical analysis of a number of Raney-nickel-coated cathodes produced by cathodic deposition of NiZn-precursor coatings revealed that although the catalyst surface-loadings ranged in the very narrow limits of from 4.10’ to lo4 cm’/cm’, the effective catalytic activity of different electrodes did not correlate with the inner surface of the catalyst, and moreover according to the observed reduction of overpotential, the utilization of the inner surface of the catalyst did in some cases not exceed 5% and was on the average only 10 to 15%, at current densities approaching 1 A/cm2. A theoretical study of the problem (81) reveals that this low degree of catalyst utilization is due to slow diffusive mass transport of electrogenerated dissolved hydrogen out of the nanopores of the catalyst. This phenomenon of limited catalyst utilization is not restricted to electrocatalysis but is well known in the field of heterogeneous chemical catalysis by porous catalysts, where it is treated under catalyst utilization and characterized by the so-called Thiele modulus (see Section I. F.). The problem of the utilization of a Raney-nickel pore is exemplified by Figs. 13a and 13b. Figure 13b explains schematically the increase of hydrogen concentration with pore depth. (Hydrogen bubbles cannot precipitate in nanometer pores unless the concentration of hydrogen in the electrolyte exceeds roughly 0.1 mol/liter corresponding to a H2 pressure

117

ADVANCES IN APPLIED ELECTROCATALYSIS

X

0.8

h

h

E! c

0.8 "O

-

.

5 04

E! .- 0.6 5 0.4 .-

02

0.2

O6

h

I

h

c

0 0 0

5

10

20

15

0.0 0

5

x (urn)

10

15

20

x (crm)

FIG. 13. Schematic of (a) a straight pore, (b) the concentration profile of hydrogen established in the pores of Raney-nickel coating under operation condition: (c) calculated distribution of Hz concentration, effective overpotential, and (d) current density in a pore (diameter of 2 nm).

of 100 MPa or 1000 bar.) The limiting concentration level in the pores is the concentration determined by the applied cathodic overpotential q according to Nernst's law, cx =

co exp{-2Fq/RT},

(17)

with co equaling the Hz saturation, concentration at 0.1 MPa ( 1 bar) H? pressure. As the concentration of dissolved hydrogen in the pore increases with increasing pore depth, the effective local cathodic overpotential diminishes due to concentration poliarization according to ~ ( x= ) q(0)

+ RT/W In c(x)/c(O),

(18)

[with q ( 0 ) being always negative at Hz-evolving cathodes]. Obviously the current density decreases to nil as the hydrogen concentration approaches its limited value [Eq. ( 17)]. Figure 13 depicts the calculated current density, overpotential, and hydrogen concentration distribution in an assumed ideal (straight cylindrical) Raney-nickel pore of a pore diameter of 2 nm calculated with an exchange

118

H A R M U T WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

current density of lo-‘ A/cni’-a value that is assumed to hold for nickel at 80 to 90°C. Obviously the contribution of the pore walls-according to the current density distribution-to cathodic hydrogen evolution becomes negligible beyond 10 p m pore depth so that for a perfect, undivided Raney-nickel coating of 100 p m thickness, only 7 to 8% utilization is anticipated. This is the reason why the fissures and cracks, the so-called tertiary structure of the catalyst, formed in Raney-nickel coatings by the leaching process are so important for improving its utilization. As it is clear that the higher the exchange current density, the lower the penetration depth and catalyst utilization, i t is evident that for the more active platinum metal catalysts, an improvement of catalyst performance by using highly porous coatings is neither expected nor observed, as the normal coating roughness of ? 1 p m already corresponds to the penetration depth in a nanopore (82). c. Performance and aging (fRanev-nickel coatirigs. Figure 14 compares the current voltage curves for cathodic hydrogen cvolution in 30 wt% KOH at 90°C at smooth nickel and two different Raney-nickel cathodes obtained by cathodic deposition of NiZn and cold rolling of NiAI, on Ni supports. The Raney-nickel electrodes are comparable in performance up to current densities of 100 mA/cm-2 and then deviate from each other remarkably.

9

. E

v

c

30

100

1000

-i fmA cm-*] Fic;. 14. Comparison 01 thc current-voltage curvcs of a smooth nickel cathode and two different Raney-nickel-coated cathodes posessing comparable loading and effective surface: (a) smooth nickel. Raney nickel prepared from two different precursors: (b) plasma-sprayed NiAI,, (c) Ni.4 cold rolled together with Mond nickel.

ADVANCES IN APPLIED ELECTROCATALYSIS

119

Obviously the tertiary structure of the catalyst obtained by cold rolling is optimal and is least efficient for the plasma-sprayed coating. It should be stressed that by variation of the process parameters, also galvanically generated NiZn codeposits yield optimal performance, which is comparable to that of cold-rolled coatings (83). Raney-nickel catalysts are barely sensitive to catalyst poisoning (as are Ptactivated cathodes), e.g., by iron deposition, but they deteriorate due to loss of active inner surface because of slow recrystallization-which unavoidably leads to surface losses of 50% and more over a period of 2 years. A further loss mechanism is oxidation of the highly dispersed, reactive Raney nickel by reaction with water (Ni 2 H 2 0 -+ Ni(OH)2 i02)under depolarized condition, that is, during off times in contact with the hot electrolyte after complete release of the hydrogen stored in the pores by diffusion of the dissolved gas into the electrolyte. Recrystallization may be retarded by dispersing Ti02 or other metal oxides throughout the porous structure. This measure can only be used for cold-rolled or plasma-sprayed not for cathodically deposited precursor coatings. Prevention of corrosive degradation during off times was reported recently (84, 85) by addition of hydrogen storage alloys (LaNis) and had been applied under technical conditions.

+

4.

+

Nickel Alloys Containing Molybdenum

Mahmood and co-workers report on a technically applied coating made of nickel/molybdenum (15 mol%) alloy (Mi),fabricated by initially forming an oxidic coating in a painting/firing process originally developed for DSAs. These oxidic coatings are then reduced to the metal in a Nz/Hz atmosphere followed by an in situ conditioning, that is, formation of the final surface composition under cathodic load in the electrolyzer. By this in situ formation, a small fraction of molybdenum is leached out. A catalyst is obtained with an activity that is said to be comparable to that of noble-metal catalysts at current densities below 300 mA/cm’.

5 . Coatings of Piutirzum Metal Oxides Dimensionally stable RuO2 anodes have also been used successfully for cathodic hydrogen evolution. According to patents, low Tafel slopes, low overpotential, and in particular insensitivity to iron poisoning are claimed (87-89). The oxide is thermodynamically unstable and should be reduced to ruthenium metal at hydrogen equilibrium potential. Transatti and coworkers (90) and Kijtz and Stucki (91) report apparent stability of the bulk oxide even after prolonged hydrogen evolution in alkaline and acidic electrolytes. These results cannot be confirmed by the present authors. It can,

120

HAFTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

however, be shown by cyclovoltammetry that after short exposure of the oxidic coating to hydrogen evolution under technically relevant conditions (30 wt% KOH, 80°C) the surface of the coating does not persist as RuOL but is converted to the metal, which becomes the catalyst proper although the underlying metal oxide layer remains unchanged-very likely because proton penetration into the metal oxide layer as a precondition for oxide reduction is hampered (92). I t is also the insensitivity of ruthenium metal to iron poisoning (see below) and the metallic reservoir stored in the bulk of ruthenium oxide that makes DSAs uniquely effective hydrogen-evolving cathodes. 6. Plutinurn Metul C w t i n g s

Platinum-activated iron, steel, or nickel electrodes had never been used for hydrogen evolution from alkaline electrolytes in technical electrolyscs. The main reason was certainly the high cost. Such coatings-as concerns platinum proper-also show, however, detrimental fast deactivation by iron poisoning. The iron contents of the electrolyte in conventional diaphragm chloralkali electrolyzers and old-fashioned water electrolyzers-although relatively low-is still high enough to deactivate the platinum surface by adsorption of ferrous cations and subsequent deposition of metallic iron-after sufficiently high cathodic overpotential had been obtained due to Fe(l1) adsorption (91 ). Quite different from platinum, ruthenium does not adsorb ferrous ions to a comparable extent and the catalytic activity of ruthenium coatings is not impaired by traces of dissolved iron. Ruthenium metal cannot be deposited on iron or steel but only on nickel cathodes since the performance of Ru-coated cathodes is so good that even at current densities of I Akm' the overpotential (-80 to - 100 1nV) is below the immunity potential of iron and the electrode support would corrode actively. Figures 15a and 15b depict the cyclic voltaniogratn of a ruthenium oxide and a ruthenium-coated nickel cathode together with their current voltage curves for hydrogen evolution in 30 wt% KOH in Figure 1%. The nickel electrodes are coated by a smooth ruthenium coating from the alkaline electrolyte containing coniplexed Ru(II1) ions (complexing reagent EDTA), which is deposited in a chemical rate-controlled cathodic process during hydrogen evolution (92). The deposition of the coating from complexed ruthenium allows for the formation of a compact coating of 1 p m t hickness avoiding deposition of mechanically unstable dendrites. I t is not unlikely that the in situ activation of conventional cathodes by so-called cataphoretic deposition as claimed by Nowak (93) essentially is based on this type of platinum metal deposition.

121

ADVANCES IN APPLIED ELECTROCATALYSIS

a zoo h

5

a

E

.-

200

100

100 N

0

a

E

Y

0

Y

.-

-100

-200

-100

-200 0

500

1000

1500

0

500

1000

1500

vs. RHE

3

10

100

1000

- i / (mA c m - n )

FIG. 15. Ru and RuOz-codted cathodes: cyclic voltammogram of (a) RuOz coating, (b) Rumetal cathode, (c) semilogarithmic current voltage curve of the cathodic hydrogen evolution at a Ru-coaled and a Ru02-coated cathode in 30 wt% KOH at 80°C.

7 . Active Coatings of Flame-Sprayed, Doped Nickel Oxide

Asahi Glass reports on the development and commercial use of electrodes covered by catalytically active nickel oxide coatings that are said to be fabricated by flame spraying. No detailed information of the composition, morphology, and phase content of the coating are communicated. Therefore it can only be assumed that the nickel oxide, which in the form of stoichiometric NiO is a nonconductor, is doped by some additives transforming it to a sufficiently conductive semiconductor. The electrode coating is said to deteriorate slowly by progressive reduction to metallic nickel. Lifetimes of 3 years in chloralkali electrolysis membrane cells are reported (94).

122

H A U M U T WENDT, SVEN

c.

RAUSCH, AND THOMAS BORUCINSKI

PLATINUM AND PLATINUM METALCATHODES IN MEMBRANE WATER ELECTROLYLEKS

In acidic electrolytes, only noble metals and in particular the platinurn metals persist under unpolarized conditions and can be used as electrode materials for hydrogen evolution. Due to their electronic configuration, the platinum metals exhibit particularly high hydrogen adsorbability ( W ) , which favors the electrochemical generation of hydrogen as well as its anodic oxidation, catalyzing in particular Hz-bond fission and formation. Platinum in particular is a unique electrocatalyst for HZ evolution in strongly acidic environment as prevailing in perfluorinated sulfonated polymer membranes used as solid polymer electrolyte in membrane electrolyzers that had initially been developed by General Electric and later had been developed into an established technology by BBC ( 9 6 ) . It is state of the art to deposit the platinum catalyst on the membrane surface by a diffusion process in which platinum salt solutions (from the cathode side of the membrane) and a reductant, like hydrazine, are counterdiffusing causing reductive precipitation of dispersed platinum close to and on the surface of the membrane. The BBC Membrel cell is reported to contain a catalyst load of only 0.2 mg Pt per square centimeter (97). The cathode is reported to exhibit only from 50 to 70 mV overpotential at current densities of l A/cm2 and 80°C. V. Electrocatalysis of Cathodic Oxygen Reduction and Anodic Hydrogen Oxidation in Fuel Cells A.

LOW- A N D

HIGH-TEMPERATURE FUELCELLS

According to the electrolyte and working temperature, one distinguishes the low-temperature fuel cell technologies (i) alkaline fuel cell, AFC (70 to 80°C), (ii) proton exchange membrane fuel cell, PEMFC (70 to 80°C), (iii) phosphoric acid cell, PAFC (200°C) from the high-temperature technologies, (iv) molten carbonate fuel cell, MCFC (650 to 700"C), and (v) solid oxide fuel cell, SOFC ( 1000°C). All these cells are anodically combusting hydrogen, although MCFCs and SOFCs may be supplied with methane or carbon monoxide from which by internal reforming and/or shift reaction with steam within the cell the hydrogen may be generated in situ (961-100). Anodic hydrogen oxidation and even more cathodic oxygen reduction is kinetically hampered at low temperature, so that anodic hydrogen oxidation in AFCs, PEMFCs, and PAFCs demands catalysts of highest activity, that is, platinum metals and platinum in particular. Also Raney nickel is used in

ADVANCES IN APPLIED ELECTROCATALYSIS

123

AFCs. Working temperatures of MCFCs and SOFCs, however, are high enough that the cheaper nickel metal-a much poorer catalyst than Pt-can be used. Also for cathodic oxygen reduction in low-temperature fuel cells, platinum is indispensible as a catalyst whereas the cathodic electrocatalysts in MCFCs and SOFCs are lithiated nickel oxide and lanthanum-manganese perovskite, respectively. Appleby and Foulkes in the “Fuel Cell Handbook” (101) reviewed the fundamental work as well as the technologically important publications covering electrocatalysis in fuel cells till 1989. B.

STRUCTURAL DESIGN OF GAS DIFFUSION ELECTRODES LN LOW-TEMPERATURE FUELCELLS

Low-temperature fuel cell electrodes are usually constituted of PTFEbonded highly porous catalyst particles. These catalyst particles are most often formed of Pt-doped soot agglomerates. Typically, as depicted schematically in Fig. 16A, these electrodes contain a coherent system of hydrophilic electrolyte flooded micro- and macropores and a hydrophobic, gas-filled, macropore system. The hydrophilic pores extend on the nanometer scale within the electrolyte-flooded catalyst particles the greater part of whose outer surface is covered by a thin electrolyte film and on the micrometer scale between the catalyst particles. The hydrophobic gas-conducting pore system extends on the micrometer scale throughout the working layer of the electrode between the catalyst particles and is constituted by hydrophobic PTFE fibers or partially PTFE-cladded catalyst particles. Both macropore systems are thoroughly interwoven with the hydrophilic pores establishing a good electrolytic connection between the electrode and the electrolyte-filled interelectrodic gap and the counter electrode on one hand and the hydrophobic pores being connected to the gas lumen through which the working gases are supplied to the electrodes on the other hand. Figure 16B explains the dual pore system in a Raney-nickel anode in which the catalyst particles measuring from 5 to 10 p m are much coarser than Pt-loaded soot agglomerates (d,, < 1 p m ) in PTFE-bonded carbon electrodes (102).

c. I.

KINETIC ASPECTS OF CATHODIC OXYGEN REDUCTION AT LOW TEMPERATURES (T < 250°C)

First Reduction Step

Because of its significance to fuel cell technology and air-depolarized batteries, the cathodic reduction of oxygen dissolved in aqueous electrolytes has been the subject of numerous mechanistic studies. They had been reviewed repeatedly (103-1/1), and today the mechanistic details are well

I24

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

A

1 mm

H

B

Coverinq laver

Working layer

Nickel screen

0.5 mm FIG. 16. Schematic presentation of the morphological features of gas diffusion electrodes for fuel cells of ( A ) PTFE-bonded and Pt-activated Hz anodes and 02 cathodes used for 0 2 reduction in acidic and alkaline fuel cells: (a) support, (b) hydrophobic gas diffusion layer, (c) hydrophilic electrode layer, (d) electrolyte, (e) magnified schematic of PTFE-bonded soot electrode, ( f ) adjacent hydrophobic layer, (g) microporous soot particles, (h) gas channels (mesopores), ( k ) PTFE particles, (I) flooded micro- and mesopores, (B) Schematic presentation of the morphology of PTFE-bonded Raney-nickel anodes used in alkaline fuel cells of the Siemens technology.

ADVANCES IN APPLIED ELECTROCATALYSIS

125

known and little disputed. A recently published proceedings volume of the Electrochemical Society collects important contributions that treat fundamental problems as well as the application and technology of electrocatalysis in 0: reduction (10). Cathodic Hz evolution and its reverse, the anodic HZ oxidation, do proceed essentially according to the same reaction mechanisms. Therefore at potentials close to that of the reversible hydrogen electrode, the reaction obeys microscopic reversibility . The anodic oxygen evolution and the cathodic oxygen reduction are, however, expressly hindered at low temperatures so that significant overpotentials must be applied to perform the respective reaction with measurable velocities/current densities according to distinctively different reaction mechanisms. The consequence is that microscopic reversibility is not established and oxygen reduction and anodic oxygen evolution do proceed through different reaction routes. It is therefore almost impossible to establish the equilibrium potential of the oxygen electrode in aqueous electrolytes (+ 1.23 V vs rev. hydrogen electrode) and only in ii very careful experiment of Bockris and Hug (112) has this been achieved. Usually at zero current density a lower potential of approximately 0.95 V is obtained, which is established by mixed reactions, with the main reaction being the reduction of oxygen to hydroperoxide. Although in general under all practical conditions the first one-electron transfer to oxygen is assumed to be the rate-limiting step for oxygen reduction, the one-electron reduction of oxygen to the superoxide radical anion can only be observed in aprotic media where it is a reversible step occurring at expressly high cathodic potentials (-0.35 V vs RHE):

2 . O:-Adssorption

Electrocatalyzed oxygen reduction proceeds always in the adsorbed state. Figure 17 depicts the three different modes of oxygen adsorption on a metal or metal oxide surfaces, which are supposed to be of relevance for cathodic oxygen reduction (113). So-called Griffith adsorption (I), which has been observed by Gland and co-workers (114) on Pt( 1 1 1) surfaces due to interac-

o=o

0-0

Me

Me

.1

\ /

Griffiths model FIG. 17.

0' 1 Me Pauling model

,"-",

Me

Me

Bridge model

Three different modes of adsorption of dioxygen

126

HARTMUT WENDT, W E N RAUSCH, AND THOMAS BORUCINSKI

tion of an unfilled dt orbital of the metal with r-orbital of oxygen and bridged adsorption would be expected to open the way for cathodic splitting of the dioxygen molecule, facilitating four-electron reduction of oxygen to water. The perpendicular (or bent) orientation of oxygen molecules being bonded only with one 0 atom to the catalyzing surface (Pauling model) is expected to yield mainly hydroperoxide by a two-electron reduction. Almost in any case of electrocatalyzed oxygen reduction, hydrogen peroxide is observed as a by-product if it is not the main product-as for instance with oxygen reduction in alkaline solution on graphite cathodes, on which even a technical process for hydrogen peroxide production is based ( I 15). 3 . 02-Reduction to Hydrogen Peroxide In these cases, the two-electron ECEC sequency

(0,)dd + H '

+

(OIH

).,
( Q Z H ) . , ~+~ ('- + (OrH),d

(OzH),,,

+H

4

-

(H202)d,,

Hl02

Electron tranaler

(204

Chemical reaction

(20b)

Electron transfer

(204

Chemical reaction

(20d)

is so fast that it cannot be resolved by the usual electrochemical methods available for electrode kinetic investigation, which have a time resolution o f not better than 1 ms. The four-electron reduction of oxygen may be due to fast catalytic disproportionation of hydrogen peroxide [Eq. (2 I ) ] and subsequent two-electron reduction of oxygen. This chemical disproportionation reaction is catalyzed by noble metals, including silver and in particular Pt metals (stable in alkaline and acidic solution), and by a number of transition-metal oxides as, for instance, Fe304, MnOz, Co304,and perovskites and magnetites containing NiO, FeO, MnO, which are, however, only stable in alkaline solution: 2H202

2H20

+02.

(21) Although 2e- reduction with fast, chemically catalyzed H 2 0 2 disproportionation, seems to be the mechanism of highest importance, it had been shown that hydrogen peroxide also can be electrocatalytically reduced to water according to H202

-j

+ 2~ + 2HZO + 2H20 + 2 0 H -

(22)

in a second, fast ECEC sequency, similar to (20a)-(20d) but accompanied by 0-0 splitting. In general, current density voltage curves exhibit at higher current densities Tafel slope of 120 mV. A direct four-electron reduction is rare and could, for instance, be demonstrated for the TI-modified gold electrodes with and 0, reduction catalyzed by tetraazocomplexes of transition metals (116-117).

ADVANCES IN APPLIED ELECTROCATALYSIS

127

D. MATERIAL PROBLEMS WITH RESPECTTO OXYGEN REDUCTION CATALYSTS I . Metal Catalvsts for Cathodic O x j w n Reduction As pointed out by Conway ( I I X ) , electrocatalysis of anodic evolution of oxygen as well as its cathodic reduction proceeds in most cases of practical importance not on metallic but on oxidized surfaces. A comparison of the cyclic voltamograms of gold and platinum in phosphoric acid shows that gold-the most noble of all noble metals-is anodically oxidized at potentials that exceed the oxygen equilibrium potential by approximately 100 mV, whereas platinum, like all other noble metals, become already oxidized at potentials that are 0.2 mV more cathodic than the oxygen potential. Other platinum metals, Rh, Ru, and silver, are even less noble. They do not corrode, however, at more anodic potentials because dense, passivating, and sometimes electron-conducting passivating oxide layers are formed. Schultze had pointed out that, provided such layers are nonconducting-as is PtO, it is essential that a thickness of only several atomic layers is maintained so that the electrons are able to cross this layer by tunelling ( I 19). These oxide layers are the electrocatalysts proper for oxygen reduction. (The gold surface acts elecrocatalytically in alkaline solution and induces HzOr production as does the graphite suface, but both are very poor Or-reducing catalysts in acid solution.) Most of the platinum metals, as for instance Pt, Rh, Ru, and Ir, are corrosion resistant and do not dissolve in acidic solution (Pt not even in 100% phosphoric acidic at 200°C). Silver is also a good electrocatalyst but may only be used in alkaline solution. Even there, silver cathodes and silver-containing catalysts must be protected from oxidation and dissolution of the metal by keeping them permanently, in particular in off time of the fuel cell, at a cathodic polarization of at least -200 mV vs 02-equilibrium potential.

2. Metal Oxide Elertrocotulysts As oxides of those transition metals that exhibit different valency states, such as Fe(II/III), Co(II/III>, Ni(lI/III), or Mn( II/III/IV), heterogeneously catalyze H202 disproportionation they are also expected to be catalysts for cathodic oxygen reduction. Indeed a number of investigations have been devoted to determining the catalytic activity of dift'erent transition-metal oxides [Co2Ni04, C o j 0 4 , Lao8 Sro7 Coo3 (120-122)]. All these oxides are stable only in alkaline solution they are electrocatalysts but could not be successfully worked into gas diffusion electrodes. In addition, they are easily reduced and decomposed if the cathodic overpotential exceeds -400 mV (122-124).

128

HAIYTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

Pyrochlores (general stoichiometryt: A,B2060’), whose structure is composed of a cubic A206framework of corner-shared octahedra interpenetrated by a A20’subarray of corner-shared O’Aqtetrahedra, are presently discussed as potential catalysts for cathodic oxygen reduction (125). Typical examples are Bi2RuZ07and B i r I r z 0 7or the lead pyrochlores Pbz , ~ , of the 0‘ sites and in Pb21r20h.,almost all 0’ lr20hI . In Pbz R U ~ O
ADVANCES IN APPLIED ELECTROCATALYSIS

129

Frc. 18. Face-to-face bis-porphyrino-cobalt complex catalyzing four-electron reduction of oxygen.

sis. There has been much dispute about the structures generated by pyrolysis of N d chelates on carbon. Today it seems clear that neither the N4 ring is fixed by this procedure to the carbonaceous substrate nor that the transition metal is essential for templating a catalytically active (MeN4) center on the carbon surface ( / 3 / ) .Instead it seems to be clear that pyrolysis of the N4chelates results in attaching nitrogen groups or nitrogen organogroups to the carbon surface, which are the catalytically active sites (C=N-C) proper. This is supported by the observation that heat treating the carbonaceous electrode material with ammonia NH3 or pyrolyzing a mixture of soot and nitrogen-containing polymers (like polyacrylonitrile or polyamide) produces also catalytically active soots (110). This type of N-doped soot catalyst is of particular interest for the development of advanced fuel cells. As this type of catalyst is not poisoned by carbon monoxide, it is a promising candidate for O2 cathodes in methanolconsuming fuel cells (132). In methanol-combusting cells, diffusive transport of methanol from the anode to the cathode cannot be avoided, with the consequence that the activity of Pt-activated cathodes becomes severely impaired by CO poisoning of the Pt catalyst; therefore, a CO-insensitive cathodic electrocatalyst seems to be indispensible. Yet the longevity of this type of catalyst is still in dispute (133). Summarizing the state of the art for the electrocatalytic reduction of oxygen the following statements can be made. 1 . The most important catalyst material for O2 reduction in acid as well as in alkaline solution is platinum. 2. In alkaline solution, silver is a second option, but silver must be protected against anodic oxidation and partial dissolution by safely polarizing it to at least -200 mV vs reversible oxygen electrode potential. 3 . Pyrolyzed N4 chelates and nitrogen-containing polymers on porous carbons and soots might become an option as electrocatalysts for O2

130

HARTMUT WENDT, SVEN

RAUSCH, AND THOMAS BORUCINSKI

reduction i n alkaline and acid fuel cells in the future, provided that their longevity can be proved. OXIDATION E. CATALYSTS FOR ANODICHYDROGEN The mechanism of anodic hydrogen oxidation is much simpler than that of oxygen reduction. Reaction pathways would be the reverse of the VolmerHeyrovsky or the Volmer-Tafel mechanism, that is, with anodic adsorption H r + H,,,, + Hi + e -

(234

or dissociative chemisorption Hz

2 H,J

as the initial steps. Since the electrocatalysts must be stable under open-circuit conditions for technical purposes, the materials of choice are nickel in the form of Raneynickel or carbon-supported Pt, Pd, and other platinum metals for alkaline fuel cells and carbon-supported platinum, palladium, or alloys of platinum metals for fuel cells using acid electrolytes as phosphoric acid or protonexchange membranes. Also tungsten carbide (WC) and platinized high surface WC have been proposed and used succesfully in acid solution. The mechanism of anodic hydrogen oxidation at tungsten carbide is little understood. Electrocatalysis might be due not to WC proper but to in situ generated tungsten bronce as the material in contact with aqueous electrolyte is thermodynamically unstable and should be oxidized at any electrode potential, so that the formation of surface layers of WO, are not improbable (133, 134).

F. PKOPbRTlES, PREPARATION, AND IMPROVEMENT OF EL~CTROCATALYSTS I N GASDIFFUSION ELECTRODES I . Pt -Activated Soot uncl Active Carbon From the very beginning of fuel cell development, soot and other active carbons because of their high internal surface, amounting typically to 100 m'/g, had been the most important catalyst supports for fuel cell electrodes. Platinum can be utilized on soot to a higher extent than in the form of dispersed platinum as Pt black. Carbon-supported platinum is the fuel cell catalyst of choice for the cathode as well as for the anode (135, 136). There exist three different ways to dope the carbonaceous support material with platinum or other catalytically active noble metals. 1 . Flooding of the soot with the help of wetting agents by aqueous or organic solvents containing Pt salts, drying of the flooded material, and reduction of the salts deposited within the soot particles at low temperatures by hydrogen

ADVANCES IN APPLIED ELECTROCATALYSIS

131

2. Creating acidic groups (carboxylic acids and phenol groups) on the internal soot surface by chemical oxidation with, e.g., chromic acid, fixation of cationic complexes of the catalyst (e.g., amino complexes of Pt(I1) or Pt(IV) at the inner surface of the soot particles by ion exchange, followed by chemical or electrochemical reduction of these complexes 3. Preparation of stabilized highly dispersed colloidal platinum, that is, Pt particles of from 1 to 10 nm size in aqueous solution by chemical reduction followed by adsorptive precipitation of the dispersed Pt particles on the outer and inner surface of soot particles. The ultimate goal of catalyst preparation is to obtain very small platinum particles to increase the surface to volume ratio ( S / V = 3/r for a spherical particle). Platinization by ion exchange and impregnation with colloidal platinum yield the best results in this respect (135). More recently the transformation of carbon-supported Pt colloids of approximately 1 nm diameter into Pt alloys has been reported, which seems to yield an even better catalyst, since the alloy particles-although coarser than the initial Pt particles-show improved catalytic activity and stability. It is not unlikely that from these alloys by in-situ oxidation transition-metal platinum bronzes like Ni,Ptn04 are formed, being the catalysts proper. 2 . Particle Size of Pt Microcrystals on Soot and their Effective Catalytic Activity I t is of the highest commercial importance to utilize the dispersed platinum to the highest possible degree by decreasing the particle size. However, there are limits in increasing Pt utilization by decreasing particle size due to the fact that, although in general the surface to volume ratio increases with r I , this does not hold for particular surfaces-assumed to exhibit an increased catalytic activity-for an assumed given crystal morphology. As pointed out by Kinoshita (136, 137), for Pt crystals of variable size a maximum of the mass-related activity-which is economically the figure of merit-passes through a maximum at a crystal size of 5 p m (Fig. 19). This supports, in principle, a model calculation concerning the surface to volume ratio of the 1 1 1 face and the 100 face of cubooctahedral Pt crystals depicted in the insert of Fig. 19. The surface-related catalytic activity decreases sizeably below specific surface areas of 40 m2/g of platinum, which corresponds to a particle dimension of roughly 9 nm. Obviously it makes little sense to decrease the crystal size below 1 to 2nm. As discussed recently by Stonehart (ISE-conference 1993, Berlin) it is not unlikely that the observed dccrease of catalyst utilization is due to mass transfer hindrance on a microscopic scale and is rather a question of overlapping concentration depletion spheres around particles. According to this, the most important parameter concerning catalyst grain utilization would be the crystallite distance rather than crystallite radius.

I32

H A K M U T WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI ._ d

0.5

C

3

+m

0.4

>. > ._ c

0.3

. c

0

m

73

0.2

a,

e

-m

0.1

v) v)

2

I

0.0

0

.

.

.

.

.

.

.

.

.

.

10

5

.

.

15

Crystallite size (nm)

FIG. 19. Dependence of surface-related catalytic activity of cubooctahedral (insert) tiny Pt crystals on particle diameter.

3 . Pt-Alloy CUtdy.yt.y

For more than 10 years (138-143) it has been known that by reaction of dispersed Pt on soot with nonnoble metals of Group IV B and V B dispersed alloys are formed. Treating Pt-impregnated carbons, to which salts of these metals had been added, at 900°C in inert atmosphere (Ar) leads to formation of highly dispersed Pt-Cr or Pt-V alloys. The metal salt is reduced lo the metal by the soot and forms the respective platinum alloy in siru. These alloy crystallites are reported to be highly active catalysts for phosphoric acid fuel cells. With a Pt/metal ratio from 1 : 1 to 5 : 1, V, Hf, Zr, Nb, and Ta had been tried. All these metals are nonnoble and are expected to be dissolved in phosphoric acid in the fuel cell under operation conditions. The initially used binary alloys are indeed not stable enough, and the catalyst loses its enhanced catalytic activity during several thousand hours of operation. Recently it was detected and claimed that tertiary and quarternary alloys that contain chromium are remarkably much more stable than the binary alloys, so that the aim of 40,000 hr of operation, which is the usually assumed lifetime for phosphoric acid fuel cells, can be achieved. The observed enhancement of the catalytic activity of Pt alloys is not at all understood-a common situation in catalysis research where in general the practitioners are far ahead of the theorists. Neither is the apparent stability in concentrated phosphoric acid at 200°C of these alloys understood. Bette reports that even after 9000 hr of operation P205) between 30 and 80% of at 200°C in 103% acid (H3P04 polymer the nonnoble metal still can be found undissolved in the catalyst (137). Yeager and co-workers recently showed that the Ni bronce Ni,Pt304 is an excellent and extraordinarily stable 0-reduction catalyst (Yeager ez ul., Ex-

+

ADVANCES IN APPLIED ELECTROCATALYSIS

133

tended Abstracts, 94- 1, Electrochem. SOC.,1994, Abstract 621). Therefore it is not unlikely that from Pt-alloys such bronces are formed in-situ constituting very active and stable oxidic catalysts from the alloys. This type of stabilized alloy catalyst is used today in commercial fuel cell electrodes of phosphoric acid cells and contributes significantly to the technical success of these electricity generators, which are today produced in units from 50 kW to 11 MW size.

4. Morphology and Structure o j Complete PTFE-Bonded Curhon Electrodes Figure 20 demonstrates by transmission electron microscopy the distribution of platinum microcrystals on (transparent) graphitized soot. Soot particles form agglomerates that measure around 0.1 p m in diameter. These agglomerates, after activation by dispersed platinum, are bonded and externally hydrophobized by an appropriate amount of PTFE. This bonding is achieved by mixing a PTFE emulsion (particle diameter of dis-

FIG. 20. Transmission electron microscopic picture of Pt-activated soot (courtesy of Dr. Stonehart of Stonehart Associates).

134

HAFUMUT WENDT, W E N RAUSCH, AND THOMAS BORUCINSKI

FIG. 21. Cross section through a cornmerclal Prototech electrode fabricated from a carbon fabric impregnated with PTFE- bonded, Pt-catalyzed soot particles. The layer of light color below the electrode is the Sic matrix. determined to keep the electrolyte between the electrodes.

persed PTFE approximately 0.3 to 0.5 pm) with soot, tape casting this mixture on a coarse porous carbonaceous support, and annealing the combined electrode structure at 200"C, at which temperature the PTFE begins to melt. The melting PTFE spreads and forms a thin, porous PTFE skin on the soot particles establishing the hydrophobic, gas-conducting pore system in the active layer of the electrode. Figure 21 shows the cross section of a commercial Protetech electrode. The narrow layer, which appears with a light color of 20 to 30 p m thickness, is the porous S i c matrix of the cell, whereas the porous mass between the coarse carbon fibers in the catalyst contains 30 wt% PTFE and platinized carbon on heat treated and stabilized soot with 20 wt% platinum. Bette showed by current voltage curve measurements of cathodic oxygen reduction at Pt-activated soot cathodes with Pt loadings ranging from 0.025 to 2 mg/cm' that the catalytic activity is steadily increasing with increasing Pt loading (137). But today due to improved preparation of highly dispersed alloy catalysts loadings of only 0.65 mg/cm2 became state of the art in commercial fuel cells. 5 . Pt-Cutulyst Aging

Every catalyst is aging and losing activity due to surface losses-as an example see (143). Several aging mechanisms are discussed; apart from poisoning by gaseous poisons introduced into the cell together with the working gases. The following aging mechanisms must be taken into account:

I . Anodic dissolution of Pt in the form of Pt(1I) and diffusion of the dissolved species to the anode where Pt is deposited as metal 2. Agglomeration of particles by stochastic movement on the soot particles 3. Ostwald ripening of Pt particles due to surface diffusion of Pt atoms.

ADVANCES IN APPLIED ELECTROCATALYSIS

135

Catalyst deterioration due to gas poisoning is only avoided by careful gas cleaning. Anodic oxidation followed by dissolution of Pt and transfer to the cathode is a serious cause for Pt loss. It is potential dependent and accelerates as the cathode potential increases, for instance under partial load or in off-time, when the cathode potential drifts toward the oxygen equilibrium potential. Therefore it is of utmost importance that whenever the fuel cell is switched off, the oxygen in the cathode lumen is rapidly exchanged by inert nitrogen and that the cell voltage under operation does not surmount 0.8 V. Agglomeration of Pt crystallites due to Brownian motion can really be observed and it can also be shown that, indeed, the interaction between the Pt particles and the supporting soot in the presence of the electrolyte, phosphoric acid, is weak enough to allow for relatively free movement of the Pt particles. This fast process obviously is also the reason for the nonobservability of slower surface diffusion-induced Ostwald ripening. Fortunately alloy catalysts composed of platinum and nonnoble metals seem to show a reduced tendency to agglomeration as their deterioration and activity loss is much slower than that of the pure platinum catalyst. G . SILVER CATHODES FOR OXYGEN REDUCTION IN ALKALINE ELECTROLYTE Silver is one of the best electrocatalysts for cathodic oxygen reduction in alkaline solution. As seen from the Pourbaix diagram of silver, this metal is readily oxidized to AgzO already at potentials-200 mV vs the oxygen equilibrium potential. To avoid anodic oxidation followed by dissolution of the metal it is therefore essential that in off time the silver electrode is protected against anodic oxidation by cathodic polarization. Silver has been used successfully in alkaline fuel cells and has been proposed and applied for oxygen cathodes in experimental cells for the chloralkali electrolysis process, where the exchange of oxygen reduction for the usual cathodic hydrogen evolution

2cr 1/20? + HzO

2CI-

+

1/20?

-

2r

+ 2~ + H,O

el,

anode

(244

-j

20H

cathode

(24b)

+

20H-

cell reaction.

(244

-3

+ Clz

allows saving of approximately 1 V in cell voltage. The silver catalyst is used in three different forms, namely as 1 . silver-doped carbon catalyst 2. Raney silver 3 . PTFE particles coated by a thin silver film in the form of SILFLON' cathodes.

136

HARTMUT WENDT, SVEN

RAUSCH, AND THOMAS BORUCINSKI

1000

3'

"

200

"

300

"

400

"

500

600

FIG.22. Comparison of current voltage curves of cathodic oxygen reduction in alkaline electrolyte: (a) PTFE-bonded nonactived soot cathode, (b) PTFE-bonded Pt-activated soot cathode, (c) SilHon cathodes consisting essentially of submicrometer PTFE particles covered by a 0.1-pm Ag layer.

Whereas the literature on silver-doped carbon is relatively scarce, more is known about Raney silver, as it had been applied in the alkaline fuel cell of the Siemens technology (98). As is typical for Raney metals made from metals of relatively low melting point, T,,, (Ag) = IOOO"C, Raney silver is a much coarser material than Raney nickel, T,,, = 1650°C. It exhibits a specific surface of no more than I m2/g or less. The same holds for Silflon cathodes, which with silver loadings of 0.1 gkm' of electrode area exhibit an active catalyst surface of no more than 0.24 m'/m'. These Silflon electrodes are produced by chemically depositing silver on PTFE particles of 0.5 to 1 p m size and working these tiny silver-covered particles into electrodes by calendaring or pressing. Figure 22 compares the current voltage curve of typical gas diffusion cathodes in alkaline electrolyte. Evidently t h e Silflon cathode in spite of its relatively low active surface is superior even to the PTFE-bonded Pt-activated soot cathode (144). ANODES FOR ALKALINE FUELCELLS H. RANEY-NICKEL Justi rt UI. (75) had already introduced Raney-nickel anodes for anodic hydrogen oxidation in alkaline fuel cells in the early sixties. At that time they used sintered electrodes composed of Raney-nickel particles, which are, however, too heavy and too expensive to be of use for commercial cells as too high loadings of the relatively expensive nickel are needed. Today

137

ADVANCES IN APPLIED ELECTROCATALYSIS

Raney-nickel anodes form PTFE-bonded structures whose metal loadings do not exceed 0.19 gkm’. The catalyst is produced from a mixture of NiAh and NiZAIj as a precursor that is milled. The powder fraction with diameters dp < 10 p m are used for preparation of the electrode. Mixing the catalyst either with dry PTFE and subjecting the mixture to knife milling or mixing it with aqueous Teflon emulsions yields a mixed precipitate of the Raney metal with microscopic PTFE particles and PTFE fibers that can be worked into PTFE-bonded felts by calendaring and pressing the felt into metal grids of woven metal or carbon fabric (144-146). Figure 16b depicts schematically the cross section of such PTFE-bonded Raney-nickel anodes. As compared schematically in Fig. 16a and 16b, the particle sizes of Ptactivated soot and Raney-nickel electrodes differ by at least one order of magnitude. This is not only by chance as soot agglomerates come in particles sizes of fractions of micrometers, and Raney-nickel particles are easily milled to micrometer size but not below. It is the relatively lower intrinsic catalytic activity of nickel compared to platinurn in the hydrogen oxidation A/cm’ compared to io(Pt) = A/cm2) that perkinetics (io(Ni) = mits the choice of particles for gas diffusion anode fabrication from Raney nickel larger than those from Pt-activated soot. This statement is supported by Fig. 2 3 , which shows the current voltage curve of Raney-nickel electrodes of equal catalyst loading with catalysts of different particle size ranging from 10 to 50 p m . The smaller the particle size the better the electrode works, but decreasing the particle size below 10 p m yields no further improvement. The rationale of this particle size effect is the limited utilization of too large particles resulting in too high values of the Thiele modulus (see Section 1. F.).

600

/ / /

d
E

N

400

0

2

-

d=35um

E v

.-

200

0 0

20

40

60

rl

I (mW

80

100

FIG. 23. Dependence of the anodic overpotential of hydrogen-consuming Raney-nickel anodes of equal catalyst loading on the particle size of the catalyst.

138

HAFTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

F[(;. 24. Schematic presentation of the deplection of the concentration of dissolved hydrogen in the interior of a Raney-nickel particle due to competition of diffusive mass transport and anodic oxidation.

Figure 24 depicts schematically the concentration profile of dissolved hydrogen in a porous catalyst particle brought about by interaction of diffusive mass transport of hydrogen into the porous catalyst particles and the consumption of hydrogen by anodic oxidation at the inner surface of the catalyst. For a single straight pore, the relevant differential equation reads

D6’c/6x2 = r ,

=

-i/(r2F),

(25)

where r b is the volume-related rate of H r consumption. Equation (25) can be transformed for a spherical homogeneous catalyst particle of specific catalyst surface density a [cm2/cm3]into the equation D / r 2 6 ( r ’ 6c/Sr)/6r = i u / ( 2 r F ) .

(264

Accounting for the potential and concentration dependence of the current density i one obtains (145-149) D / r 2 6 / 6 r ( r ’ 6 c / 6 r ) = ( i ( q ,c ) u / 2 F ) ,

(26b)

with i ( q , c) = i$ [c/cT exp{a+Fq/RT} - exp{-a-Fq/RT}]. cz is the deporlarizer concentration outside the pore. The solution of this differential equation is determined by the dimensionless Thiele modulus, [Eq. (9a)l which is given in this case by %t,

=

R ( i a / 2 F c=D)”’,

(27 1

ADVANCES IN APPLIED ELECTROCATALYSIS

139

where R is the radius of the particle and c, is the hydrogen concentration outside the catalyst particle. The utilization of the porous catalyst particle is given by the ratio of the actual current produced by anodic H? oxidation in the particle to the hypothetical current produced without mass-transfer limitations: utilization

=

l/Imax= 3 tanh

+Th/+~~.

(28)

According to the different exchange current densities, io, for hydrogen oxidation and hydrogen evolution on Ni and Pt, the catalytic activity of platinum is by a factor of several hundred to a thousand higher than that of nickel. Therefore, if the utilization of Raney-nickel particles below 10 p m size approaches loo%, it is clear that Pt-activated porous soot particles must be by a factor of from 10 to 30 smaller than Raney-nickel particles to achieve full utilization, that is, vanishing fuel starvation of the catalyst. This happens to be the case with soot agglomerates that are by their very nature of correct size (d, < 0.1 p m ) (150, 151).

I.

REVERSIBLE A N D IRREVERSIBLECATALYST DETERIORATION

Aging and deterioration of catalysts is as usual in electrochemical devices as in heterogeneous gas-phase reactors. These processes proceed according to very different mechanisms. Trivial is catalyst poisoning, for instance, by sulfur, dihydrogen sulfide, or other sulfur donors that affect the catalytic activity of platinum. For phosphoric acid fuel cell anodes that work on a type of synthesis gas obtained by steam reforming of methane, carbon monoxide is the most important poison at the anode side. Whereas sulfur or H2S can be kept easily below a concentration of 2 ppm by using active carbon adsorbents to purify the anode gas, CO stays after the gas has passed through a shift reactor at the percentage level. This high concentration is one of the reasons that phosphoric acid cells are operated at 200°C and above, as at these temperatures CO is adsorbed to a much lesser extent on Pt no longer impairing hydrogen oxidation even if the CO concentration rises to the level of several percent. Catalyst poisoning due to sulfur or carbon monoxide is reversible. But there are aging processes that deteriorate the catalyst irreversibly and permanently. Aging by loss of active catalyst surface due to the growth and agglomeration of Pt microcrystals due to Brownian motion is discussed in Section V.F.5. Pt redistribution and surface recovery by potential cycling of the electrodes has been shown to occur at massive Pt-metal electrodes by Arvia (152, 153). This method of restoring the catalytic activity of the Pt catalyst

140

HARTMUT WENDT, W E N

RAUSCH, AND THOMAS BORUCINSKI

I

01

s

.

100 mA c m 2 , 50%

E

Y

w

I

50

.

t (h)

FIG.25. Comparison of the deterioration with time of Raney-nickel catalyst under different operationconditions. (Jenseit, W.,Khalil, A , , Wendt, H.,J.Appl.Electrochem. 20,893(1990)).

cannot be applied and does not work with dispersed platinum microcrystals on soot in the cell. The loss of the catalytically active surface of Raney nickel due to recrystallization is a continuously progressing process that can be retarded to some extent in Raney-nickel anodes by dispersing oxide ceramic materials like ZrOz and Ti02 in the nickel matrix. More serious is anodic oxidation for some metals additionally accompanied by dissolution of the catalyst to which even platinum is subject but which is an even more serious hazard for the less noble catalysts as silver and Raney nickel. Figure 25 shows the evolution of cell voltage with time of Raney-nickel anodes that are deliberately operated at too high current densities so that the effectively applied overpotential was above the threshhold for nickel oxidation, which amounts to +80 mV vs the reversible hydrogen electrode. Evidently at a current density of 400 mA/cm2 and at 80°C the oxidation of Raney nickel proceeds within hours and at 300 mA/cm' still within a week. At lower temperature and still lower current density and anodic overpotential, respectively (50"C,100 mA/cm2, 7 = +40 mV), the observed decline of the current density with time signals physical and not chemical deterioration of the catalyst as the overpotential is not sufficient for nickel oxidation.

J. 1,

ELECTROCATALYSIS OF ANODIC METHANOL OXIDATION

Technoeconomic Significance of the Process

Fuel cells are considered promising power sources for electrotraction, provided the difficult storage problem of the fuel, hydrogen, can be solved

ADVANCES IN APPLIED ELECTROCATALYSIS

141

(154). Cryogenic and pressurized hydrogen is too expensive yet, and therefore the storage of hydrogen in the form of methanol with on-board steam reforming of methanol is one of the options being investigated by General Motors today (155). Even more advantageous would be the direct anodic oxidation of methanol in a fuel cell, because the complicated integration of the steam-reforming process into a car system could be avoided and-in principle-higher energy conversion efficiencies of the whole system could be expected.

2. Poisoning of Pt Catalyst by Oxidation Products of Methanol Although platinum strongly catalyzes methanol decomposition by chemisorption yielding adsorbed H atoms and adsorbed methoxy species (PtOCH,), the initially high anodic current density that is observed if a platinum electrode is dipped into a methanol solution decreases within seconds and minutes by orders of magnitude reaching, technically spoken, an intolerable value of only several mA/cm* at +500 mV vs rev. H2 electrode. According to UHV investigations (156), the cracking reaction of methanol at Pt surfaces yields strongly adsorbed carbon monoxide. There had been much dispute as to whether the species that blocks the electrocatalyst platinum in methanol-containing electrolytes is really carbon monoxide (157), but today the different schools seem to converge with the opinion that it is CO or CO*, some species resembling adsorbed carbon monoxide very closely in its adsorptive, optical, and electrochemical properties. 3 . Anodic Methanol Oxidation at Alloy Catalysts According to long-lasting experimental efforts, the use of alloy catalysts that contain a less noble metal whose oxide exhibits low solubilities in acid electrolytes-in particular Sn and Bi are effective in this respect-enhance the catalytic activity of platinum. The rationale of this effect has been that the oxide of the nonnoble component at close atomic distance from the Pt surface atoms supplies by “spillover” the oxygen that is necessary to oxidize the adsorbed CO species. Today research and development turn more to Ru and Ir or Rh, the more easily oxidizable platinum metals as alloying metals that seem to be at least as efficient as Bi and Sn and are certainly more stable than those in acidic environments-in particular if the anode potential becomes more anodic in cases of poor supply of fuel (158). The Pt-Ru anode exhibits a sizeable higher oxidation current for methanol and for adsorbed hydrogen than the Pt electrode, indication that a smaller part of the Pt electrode surface is blocked by CO adsorption. Still the catalytic activity is too low because the onset of the anodic peak of methanol oxidation is at a

142

HAITTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

potential of +SO0 mV vs RHE (the equilibrium potential of methanol oxidation is calculated to be almost the same as the RHE potential.) It must be stressed that there is still a fair chance of arriving at electrocatalysts that would achieve reasonable anode potentials (say +0.3 V vs RHE) at technically acceptable current densities (say 0.3 to 0.4 Akm’). The situation for the next higher alcohol, ethanol, however, is almost hopeless. Any work aimed at developing catalysts for anodic oxidation of the much more inert hydrocarbons at low temperatures will certainly be frustrated.

K.

GAS

DIFFUSION ELECTRODES IN MEMBRANE (PEM) FUELCELLS

1 . Rutionule oj‘Developing Memhrune Fuel Cells As long as fuel cells are using liquid electrolytes like phosphoric acid or concentrated caustic potash, the catalyst utilization is usually not limited by incomplete wetting of the catalyst. Provided the amount of electrolyte is sufficiently high, the hydrophilic porous particles are not only completely flooded but due to their expressed hydrophilicity are wetted externally by an electrolyte film that together with the whole electrolyte-filled hydrophilic pore system establishes the ionic contact of an electrode to the respective counterelectrode. Using proton exchange membranes as electrolytes that are quasi-solid may cause a problem with respect to the perfect wetting of the catalyst particles. In spite of this (initial) difficulty of developing solid polymer membrane fuel cells, water-swollen perfluorinated sulfonic acid polymers such as the commercial Nafion have been used for fuel cells very early since they offer the following advantages: 1. The electrode kinetics of cathodic oxygen reduction at Pt in contact with the acidic polymer is enhanced by a factor of at least 10 compared to aqueous sulfuric or phosphoric acid at temperatures of 50 to 80°C (159) because the sulfonic acid groups interact adsorptively less with Pt than SO;’ or PO: ions, leaving the greater part of the Pt surface free for adsorption of Or. 2. Transport properties of oxygen are better in the fluorinated polymer than in aqueous electrolytes. Although the diffusion coefficient is lower than in water, the oxygen solubility is largely enhanced so that the transport parameter D + cOzis higher than in water (160, 161). 3 . Although the electrolyte is strongly acidic, cell construction is less demanding as the solid electrolyte does not cause corrosion problems.

143

ADVANCES IN APPLIED ELECTROCATALYSIS

2. Improving Catalyst Utilizution by Ionomer Impregnution of Gas Diffusion Electrodes The ionic monomer that forms the proton exchange membrane (PEM) separating and ionically connecting the two gas diffusion electrodes can be dissolved in isobutyl alcohol or other organic solvents, such as isopropanol. This circumstance opens the way for improving the ionic contact between the catalyst particles of a gas diffusion electrode and the proton-conducting membrane and electrolyte. According to Raistrick (162), prefabricated gas diffusion electrodes (PTFE-bonded, Pt-activated soot, Prototech) are soaked with an alcohol solution of the monomer, and the solvent is subsequently evaporated. As the solution is wetting the electrode fairly well the active electrode layer is evenly impregnated by the ionomer. The impregnated electrode is subsequently glued to the membrane by hot pressing. If the face of the gas diffusion electrode is further sputtered with only 0.05 mg/cm2 of platinum, the electrode performance of the fuel cell is improved further (163), (164). Figure 26 compares for equal Pt coatings of 0.35 mg/cm’ the performance of ionomer-impregnated 02-reducinggas diffusion electrodes incorporated into an H d 0 2 PEM fuel cell for (a) 20% Pt on 0.050 mg/cm2 sputtered platinum. Instead carbon, (b) 20% Pt on carbon of sputtering, the catalyst may also be applied in a controlled manner to the face of the electrode by galvanic deposition (165, 166). Careful analysis shows that the degree of ionomer filling of the electrode is critical for cata-

+

1100

5E

900

v

a,

0

-0a > c

700

500

0

300 0

500

1000

i

1500

2000

/ (mA cm-2)

Fici. 26. Comparison of current voltage curves of membrane fuel cells using impregnated PTFE-bonded electrodes (a) prepared from soot activated by 20 wt% Pt and (b) with additional Pt sputtering of the contacting surfacc.

144

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

lyst utilization ( 1 6 7 ) . Recently, Gottesfeld and associates have shown that producing electrodes of I to several micrometers thick by printing a “paint” composed of 30 wt 5% Pt on soot dispersed in a Nafion-isobutanol solution allows to keep the Pt loading of these electrodes below I mg cm-’. (S. Gottesfeld and associates, 185th Meeting of the Electrochem. Soc., Abstracts 94-1, Abstract No. 607).

3 . Electrocutulysis in Regenerurivc. Fuel Cells So-called regenerativc fuel cells are devices that can be used for electrolytic water splitting as well as for reconverting the chemical energy stored in hydrogen and oxygen into electricity. This would be desired for space applications as well as for dispersed application whenever medium and long-term storage of high-quality energy (electricity from hydroelectric, wind, or solar devices) is the aim. Fuel cells can only be made to become regenerative if the following are true: 1 . The gas diffusion electrodedgas-evolving electrodes dispense at least on the side of the oxygen anode with activated carbon, because at oxygen evolution potentials the carbon is subject to rapid oxidation. 2. The oxygen cathode-for which platinum catalyst due to its outstanding structural and catalytic capability is the rule-is not used as an oxygen evolution anode in the electrolyzer operation mode because oxidation of Pt and fast catalyst deterioration would be the consequence. Therefore an oxygen cathode based on a platinum catalyst must operate as a Hz-evolving cathode in the regenerative mode. 3. The oxygen anode (in the regenerativc mode) as a metal oxide is also able to work in the oxidic form as hydrogen anode (in the fuel cell mode).

According to these demands, Ledjeff and colleague have designed a regenerative membrane cell that is supplied on one side with dispersed platinum catalyst as an Or-reducing and Hr-oxidizing catalyst and on the other side with a mixture of Ru02/IrOz as a cathodic oxygen-evolving and anodic hydrogen-consuming catalyst (168, 169). This type of cell is said to have approximately 50% efficiency per cycle. L.

IN HIGH-TEMPERATURE FUEL ELECTROCATALYSIS C~LLS

1 . Electrode Reactions and Eleclrode Muterials at High Temperatures

An increase in process temperature changes the reaction rates and hence the overall kinetics of more evolved electrochemical reactions that are composed of charge transfer and chemical steps to a sizable extent. For hydro-

ADVANCES IN APPLIED ELECTROCATALYSIS

145

gen evolution from aqueous solutions, for instance, effective activation energies of from 40 to 60 kJ/mol are reported (41). Increasing the temperature from ambient temperature to 600°C would be expected to increase the rate of such reactions by a factor of at least 2 X lo4. Therefore the electrochemical evolution and oxidation of hydrogen, which exhibits at room temperature a nickel exchange current densities of the order of lo-' A/cm2, would proceed at 600°C with io = lo-' A/cm2 even at high current densities (1 Aicm') with almost vanishing charge transfer overpotential. Also the much slower oxygen electrochemistry with somewhat higher activation energies would become relatively fast at 600°C. Therefore, in high-temperature fuel cells the demand for highly efficient-and expensive-noble metal catalysts simply does not exist. Finding appropriate electrode materials for high-temperature fuel cells is rather a question of finding materials that are resistant against chemical attack of the working gas and an agressive electrolyte, keep dimensionally stable over long periods of time (40,000 hr), and are not subject to creeping or shrinking by progressive sintering. As highly dispersed materials, only oxide ceramics and refractory metals with melting points above 1800°C would be expected to persist. Nickel with a melting point of 1400"C, which is the preferred anode metal of both high-temperature fuel cell technologies (molten carbonate and solid oxide cells), is not stable above 450°C in the form of the highly porous Raney nickel. At temperatures above 600"C, highly dispersed nickel sinters to granules of micrometer size within a fraction of an hour. Even with this size the particles tend to grow and lose specific surface. The nickel sponge anodes of high-temperature fuel cells must be stabilized against aging by sintering and creep. Therefore, the effective catalytic activity of nickel anodes in high-temperature fuel cells is rather more a question of establishing and stabilizing a relative coarse but long lasting dispersed structure than improving the intrinsic catalytic activity of the material. Cathode materials of both high-temperature cells are composed of chemically and morphologically relatively stable oxide ceramics. For molten carbonate cells, lithiated NiO is used and the cathode of oxide ceramic cells usually is made of porous LaMn03. 2. Electrode Kinetics and Electrocatnlysis in Molten Carbonute Fuel Cells a. Anodic Hydrogen Oxidation. As one would assume, the electrode kinetics of the electrochemical hydrogen reaction is fast but its dependence on hydrogen, carbon dioxide, and water vapor pressure is more complex than naively anticipated. As found by Ang and Sommels and Selman and his colleagues (1713,the rate equation reads

I46

HAWMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

The reaction order 0.5 of hydrogen is understood in terms of a fast esrablished dissociative adsorption equilibrium at vanishing coverage, whereas the formal reaction orders of carbon dioxide and water vapor are less easily understood. Interesting enough, copper, which at ambient temperature does not catalyze hydrogen oxidation, catalyzes it almost as well as does nickel at 650°C since the exchange current densities are of the same order of magnitude (/7/), As hydrogen in the molten electrolyte at 600°C possesses a solubility comparable to that in water close to ambient temperature and the electrode particles are coarse (of micrometer size as shown in Fig. 27a), it is essential to establish a morphology and structure of the porous anode with larger gasconducting pores that must not be flooded and also provide a thin electrolyte film covering the spongelike metal structure. However, nickel is not well wetted by the melt as it exhibits a wetting angle of approximately 30". Dispersing oxide ceramic materials as A1203, which forms in contact with the Li,COJK2COI melt lithium aluminate, LiAlO2, or TiO;?, which forms Li2T:03 (both being almost insoluble in the melt), one induces an improved wetting behavior of the sintered nickel sponge. One achieves improved resistance to creep and sintering of the metal by the wetting of the nickel sponge's surface and decreasing the surface tension, which is the driving force for surface loss by sinter densification (Fig. 27b)(/72). An investigation of the effective kinetic law of the anodic hydrogen oxidation in the ccll shows a kinetic law that differs from Eq. (29). The latter law had been obtained by procedures that eliminate mass-transfer limitations. Figure 28 shows the result of the kinetic in-cell investigation. For three anodic overpotentials there are plotted the obtained current densities versus gas composition, which simulates different degrees of hydrogen conversion. The gas composition changes from pure hydrogen (left) to H20/C02 = 1 / 1 mixtures. As predicted by Eq. (30), not only for complete conversion but also for vanishing conversion the current approaches zero as neither vapor nor carbon dioxide is present. The detailed analysis yields the rate equation with u + b = 0.37. Whereas the sum a + b is sufficiently close to that obtained by Selman, the reaction order of hydrogen is close to unity instead of 0.5. This is a clear indication that the anodic reaction of hydrogen is almost mass-transfer controlled in the cell; that is, film diffusion of hydrogen decisively influences the rate of anodic hydrogen conversion (I 72). The observed anodic current voltage curves (Fig. 29) exhibits nonetheless only quite low overpotentials due to concentration depletion. It is comparable to the voltage generated by passage of the anode currents through the electrolyte film.

ADVANCES IN APPLIED ELECTROCATALYSIS

147

FIG. 27. Morphology of the nickel sponge forming the MCFC anode, containing (a) only nickel and (b) nickel and 10 wt% TiO2.

148

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

500 400

u

h

N

E

0

300

? E

v

.-

200 100 0 0.0

0.2

0.4

0.6

0.8

1.0

FIG. 28. Kinetic in-cell investigation of the anode reaction of the MCFC establishing an effective H2-reaction order of close to one.

b. Cathodic Oxygen Reduction. The cathodic reduction of oxygen in carbonate melts involves carbon dioxide also, since carbonate ions are formed cathodically, decreasing the steady-state concentration of oxygen anions at the cathode surface and depolarizing the cathode: $0, + COz

+ 2e- -+Cot-.

250 MCFC-anode

200 150

100 50

0

0

100

200

300

400

i / [mA cm-*] FIG. 29. Current voltage curve of MCFC anodes: (a) nickel sponge anode, (b) dispersionimpregnated nickel sponge (Li2Ti03).

ADVANCES IN APPLIED ELECTROCATALYSIS

149

The carbonate anion migrates from the cathode to the anode, serving as a shuttle for the oxygen dianion which is released as water vapor at the anode after combining with protons. The cathode today consists of a porous layer of lithiated nickel oxide Li,Nil-,O, which, being a p-type semiconductor of high conductivity, provides the necessary electronic conductivity and an internal cathode surface, which is catalytically active for dissociative reduction of 0 2 species. Oxygen dissolves in the melt by reaction with carbonate, forming hyperoxide and peroxide anions: 2CO;2CO:

+ 20,

2

2COZ + 30;

+ O2 2 2COZ + 20:-

hyperoxide

(324

peroxide.

(32b)

These equilibria are responsible for the relatively complicated electrode kinetics. Kinetic data, which are in situ measured, are inconsistent with a simple rate law. Appleby and Nicholson (173, 174) report the rate laws obtained at flag electrodes that are consistent with the reduction of hyperoxide and peroxide anions, respectively. Diminishing the high cathodic overpotential is of highest technical relevance. The question of which part of the cathodic overpotential is to be attributed to true kinetic hindrance and which part must be attributed to poor conduction of the network of NiO-crystals which are forming the porous cathode and to mass transport hindrance of dissolved oxygen species and carbon dioxide in the porous cathode matrix is not yet answered. Therefore a clear indication of the importance of the electrocatalytic action of the cathode matrix cannot yet be given.

M. ELECTROCATALYSIS IN SOLID OXIDE FUELCELLS 1.

Electrodes and Electrode Structure

Figure 30 shows the enlarged cross section of a solid oxide fuel cell composed of a 50- to 100-pm-thick ZrOz membrane, a relatively thick (-100 p m ) porous cathode, and a porous Ni/ZrOz cermet anode. a. The anode. The anodes and cathodes of solid oxide fuel cells exhibit quite different morphologies. The anode structure is characterized by a finely dispersed cermet composed of yttria-stabilized zirconia (YSZ) and nickel. The electrode reaction [Eq. (33)] takes place near the phase boundary of the Ni/YSZ grains. There the actively working anode is not nickel but the surface of the zirconia. The coherent particulate nickel matrix

150

H A R M U T WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKl

electrolyte oxygen

J

anode

/@

/

J

J

water vapour

J

\hydrogen \

oxygen ions

FIG. 30. Cut through a solid oxide fuel cell

serves mainly as a current collector and is not the anode proper. The main purpose of using a porous cermet is to stabilize a high specific surface area for the interface ZrOdfuel gas where the equilibrium (02-)dcj, Ni/ZrOz

+ Hz = HzO + 2r-

Ni/Zr02

(33)

is established. As the mobility of electrons in the yttria-doped zirconia surface is relatively high, whereas their concentrations is relatively low, the mean migration length for electrons in zirconia must be kept very short and the volume-specific interfacial area of metal/solid electrolyte interface (Ni/ZrOz) has to be high enough to avoid charge-transfer limitations. Therefore, the morphology of the SOFC anode resembles that of lowtemperature fuel cells with liquid electrolytes-although the limiting factors for electrode performance are very different (solubility and diffusivity of gases in the case of low-temperature fuel cells, low minority charge carrier concentrations in zirconia for SOFCs). The cermet anode is therefore composed of small zirconia particles of 0.5 to I p m diameter and 5- to 10 p m nickel grains, the sinall dimensions of the zirconia grains rendering short transport distances for electrons in the ceramics (175). b. The cuthode. The SOFC cathode is usually composed of mixed oxides [La(Sr)Mn03+,in particular] that possess good oxygen ion mobilities together with high electron conductivities approaching metallic conductivity. Such electrode materials could, in principle, work as flat, twodimensional films. However, their limited electronic conductivity demands an enhanced not too small thickness of the electrode layer to provide a reasonably low lateral resistivity of the electrode for current collection (175).

151

ADVANCES IN APPLIED ELECTROCATALYSIS

This is the only reason why SOFC cathodes are also formed as porous layers, but they are composed of one singular oxidic phase only-usually LaMnO?. The cathode thickness for cells of approx. 1 cm length is about 200 p m with a mean porosity of 40% and a mean particle and pore diameter of 5 p m (176). Current densities in the cathode are mainly determined by the respective value of oxide anion conductivity compared to the electronic conductivity ( K ; - and K ; , coupled to each other in Wagner diffusion). Equation (34) describes the current density limit for coupled transport of oxygen anions and electrons (f 77): where d l . d M n 0 3 is the particle diameter of the electrode material. Therefore, the thickness of such closed cathode layers or coatings, d , must be kept below a value of several tens of micrometers. Coatings of this thickness, however, would provide too low lateral electronic conductivities for current collection along the cell surface. Therefore, a porous cathode is constructed with approx. 50% porosity and at least 200 p m thickness, and with pore and particle size of approx. 10 p m . The main part of this porous layer serves simply as current collector, and only the lowest part, i.e., those crystallites that contact the zirconia electrolyte, act as a cathode surface across distances of the order 10 p m (178).

VI.

Electrocatalysis in Electroorganic Synthesis

A.

INTRODUCTION INTO 'THE

FIELDOF ELECTROORGANIC SYNTHESIS

Electroorganic synthesis deals with conversion of organic compounds into useful products by anodic oxidation or cathodic reduction. Today there exist literally thousands of published examples of electrosynthesis reactions but only a very small number-certainly not more than several tens-are really exploited commercially, the best known example being the cathodic hydrodimerization of acrylonitrile to adipodinitrile, a precursor to hexamethylene diamine, which is the aminoconstituent of nylon 6,6 (f 79): 2H2C

=

CH-CN

+ 2e- + 2H'

(35)

+ CN(CH~)JCN.

A second, less well-known example is the anodic conversion of toluenes in methanol as solvent to benzaldehyde-dimethoxy acetals (180,f8f): RC,,H4CHI + 2CH30H + RChH4CH(OCH3)2+4Ht

+ 4e

.

(36)

152

HARTMUT WENDT, SVEN RAUSCH, A N D THOMAS BORUCINSKI

I . Mediuted Electrochemicd Conversions of Orgunic Subst rutes The anodic and cathodic conversion of organic compounds might be mediated by regenerable redox systems. In these cases of mediated electrochemical conversion one partner of a redox couple undergoes oxidation or reduction respectively at the electrode and reacts with the substrate in the bulk of the solution shuttling the electronic charge from the electrode to the substrate. Examples for anodic mediators are the redox couples CnO;. / Cr3+,Mn'+/Mn'+, and Ce"tjCe7', and examples of cathodic mediators are Ti-"/Ti4+ and V3'/Vs+. If the anode or cathode surface is coated with an insoluble redox system that is able to oxidize or reduce organic substrates by a heterogenous reaction, one may call this type of mediated conversion "heterogeneously catalyzed" anodic oxidation or cathodic reduction, respectively. Examples are oxidation at lead electrodes by PbOz (182) or by MnO' coatings (183) and as a particular very recent innovation, Cr(V1) species tixed to Ti02 coatings on Ti electrodes by the spray sinter technique (184). With this type of electrochemical conversion, the organic substrate interacts chemically with the redox mediator contained in the coating. For instance alcohols that are to be oxidized by Cr(V1) species are initially bound to this species as a chromate ester-a mechanism usually known as the Wiberg mechanism and operative in homogeneous chromate oxidation of alcohols (185) before they undergo charge exchange and oxidation by the mediator. I t is this chemical interaction that justifies the expression of heterogeneously catalyzed oxidative or reductive conversion.

2. Direct Anodic urid Cuthodic Electrochemiid Conversions of'Orgunic Substrates Many anodic and cathodic conversions are not mediated by soluble redox couples or by redox coatings but are essentially initiated by a one-electron charge transfer between the electrode and the organic molecule-generating radical ions or radicals as first, reactive intermediates: A

+

pp =

A-r

H' + e - = B .

C=C?+e

D

=

D.

+ e-

radical anion

(374

radical

(37b)

radical cation

(37c)

radical.

(37d)

These intermediates, reactive as they are, may react with different competing reactants yielding different reaction products. Very reactive and shortlived intermediates, which do not live long enough to leave the electrode surface by diffusion and intermediates which are adsorbed at the electrode

ADVANCES IN APPLIED ELECTROCATALYSIS

153

surface, do react in truly heterogeneous reactions. As the electrode surface according to its very nature-e.g., metallic vs oxidic, lipophilic vs lyophilic-adsorbs different reactant molecules to a different degree, it may be anticipated and it is observed that the main products obtained by direct electrochemical conversion of the same molecule may be quite different at different electrode materials. In this respect carbon electrodes-composed mainly of a polycrystalline matrix of “Acheson graphite”-play an important role. Their surface is nonpolar and favours the adsorption of nonpolar organic molecules like olefines or arenes. Adsorption enhances their surface concentrations and enhances the reaction rate of adsorbed radicals and radical ions with the more strongly adsorbed nonpolar molecules compared to their rate of reaction with polar molecules, which are less adsorbed. Thus hydrophobic carbon electrodes catalyze for instance, addition reactions of radicals and radical ions to unsaturated compounds versus their solvolytic reactions involving the polar solvent molecules, whose adsorption on the other hand would be favored at hydrophilic electrodes.

B . ANODIC OXIDATIONS MEDIATED BY REDOX COATINGS 3 . Comparison of the Rates of Homogeneous and Heterogeneous Redox Reactions Since the work of Murray and Anson and many other authors (186, 187, and citations therein) it is state of the art to attach or anchor almost any homogeneously acting redox system to the surface of a chemically modified electrode. Although generally such electrodes are not stable enough for technical purposes they have been used for laboratory investigations. Concerning surface-modified electrodes supplied at their surface with a redox system at a surface coverage of unity, corresponding to surface concentrations of approximately 2 X lo-’‘’ mol cm-’, Saveant and co-workers (188) established criteria for the technical competivity of heterogeneously vs homogeneously mediated ox idations /reductions. Although rate-limited current densities are diminished by attaching the redox system to the electrode, heterogeneously catalyzed redox conversions are justified because of two technically important circumstances: 1 . The necessity to recover and recycle a homogeneously dissolved medi-

ator may be dispensed with, and 2. the kinetics of competing reaction paths may be favorably influenced and the selectivity of a heterogeneously catalyzed electrochemical conversion may thus be improved due to the very nature of surface reactions vs homogeneous reactions.

154

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

2. Elrc~lmcatalytic.Oxidations by Oxides qf MultiplyV d w t Met& Solidkolid redox couples, as defined by metal oxides of metals exhibiting different valencies (e.g. Pb0dPbS04, M n 0 2 / M n 0 0 H , NiOOHINiO), are used as oxidants in preparative organic syntheses and may be used as heterogeneous mediators for the anodic oxidation of organics, provided: 1 . the respective redox potential is positive enough, 2. the anodic reoxidation of the respective lower valent oxide is fast, 3 . the frequent chemical reduction/anodic reoxidation does not lead to corrosive deterioration, and 4. the heterogeneous oxidation of the organic substrate is fast enough to guarantee technically acceptable current densities.

Tdble I1 collects some metal oxide couples and their redox potentials together with the usual base metal. Burke discussed recently that for easily oxidizable substances, such as alcohols and amines, the redox potentials of Table I1 might be irrelevant, as incipient formation of oxidic layers at lower oxidation potentials and not the bulk oxide may already cause the observed oxidative conversion (189). Nonetheless, for substances that are less easily oxidized for instance, olelines or arenes, the given redox potentials are certainly relevant with respect to the rate of their mediated oxidation.

TABLE I1 MrtcrliMc.trrl-Oxidt~Couples Cupable off'ncting us H e t ~ r o g ~ n c ~ oMcvkuors ~r,s $>r lhc~O.ridutiori of Orgunic. Sichstnrtrs Dis,so/vc~din A yucwus Solution Kedox couplc"

Base metal

R,/V" vs RHE

cu>oicuo MnOliMnrOl c o o i c o3 0 , AgZOiAg NiOOHiNiO

cu

0.74

Ti Co, Fe

0.94

PhOJPhSOj

Cr(V!)iCr(tl!)

Ag

Ni Ph TiiTiO,

I .04 1.18 I .29 I .69" I .7

" Recalculated with reference to the reversible hydrogen electrode potential from D. Dohos, Electrochemical Data, Elsevier, Amsterdam. 1975. ''~l,,2so =j I vdl/dn14,

ADVANCES IN APPLIED ELECTROCATALYSIS

155

TABLE I11 Heterogeneous Rate Data (cmlsec).for Oxidation of Amines and Alcohols at Dqyerent AnodeJ (25"C, I M Aqueous KOH) (190).

Substrate Electrode: electrode potential ( E vs NHE) Methylamine Methanol n-Propylamine n -Propano1 Isopropanol

Nickel 0.6 1.1 7 4.2 5.2 4

Silver 0.85

x 10-4 x lWb x X

Copper 0.7

1 . 9x 1 0 - 4 4.5 x 8.2 x 2 x 10-6

<

x 10-6

10-8

x x x x 1,1 x

2.8 6.2 1.1 2.4

10-5 lo-' 10-4 10-5

10-5

Cobalt 0.6 1 . 8 x 10-5 < lo-" 5.5 x 10-6 < lo-* < lo-"

Table 111 collects and compares the rate date for the oxidation of some aliphatic amines and alcohols on nickel, silver, copper, and cobalt oxide anodes. Homologous alcohols always react 10 times slower than amines. The reaction rates on silver and on cobalt are sizeably slower than those on copper and nickel, so that silver and cobalt could be ruled out for technically applied anodic oxidation (190). Kuhn and colleagues (191) investigated the kinetics of the heterogeneously catalyzed benzene oxidation at Pb/Pb02 electrodes in sulfuric acid. This reaction was worked out on a semitechnical scale for quinone/ hydroquinone production (192): PbS04 3Pb02

+ 2Hz0

+ C<,Ht,+ 3HrS04

+ H2S04 + 2 C

--j

PbOz

-j

3PbS0,

+ C,H402 + 4Hz0.

(384 (38b)

As shown in Figs. 31a, and 31b, the current density is critical in determining the current yield of the reaction sequency (38a)-(38b). Raising the current density above 10 mA/cm2 leads to increasing current efficiency losses as more, and more oxygen is evolved as the rate of the chemical surface reaction, Eq. (38b), can no longer follow the faster rate of charge transfer. The heterogeneously catalyzed Mn02-mediated oxidation of diacetone-L sorbose to diacetone-2keto-~sorbic acid, the latter being a precursor to vitamin C, at nickel anodes and based on the chemical oxidation of the substrate by NiOOH is of technical relevance. The limiting current density in 1 M KOH solution is under operation conditions only 10 A/cm2 leading to relatively poor space-time yields. Robertson and Ibl showed that acceptable space-time yields can by obtained by using thin layer cells of "Swiss roll" type (193, 1 9 4 , which leads to an efficient compression of the cell width to fractions of a millimeter.

156

H A m M U T WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

h

N

10 1

5

emulsion

a

. E

v

.-

no benzene

1 ;

0.1

b

100 80

s

60

c

40

h

v

20 0 1000 1 1 0 0 1200 1300 1400 1 5 0 0 1 6 0 0

E

I (mv)

FIG.31, (a) Current voltage curve of the PbOz anode in electrolytes without and saturated with benzene and (b) potential dependence of current efficiency of benzene oxidation.

Beck and Schulz have been working since 1984 on titanium anodes on whose surface chromate is anchored by the usual baking process, which produces dimensionally stable anodes for chloralkali elektrolysis [e.g., (185, 196, 197)l. Figure 32 explains schematically the oxidation of the Cr(I1I) species to Cr(V1) (CrO,) and its conversion with the substrate. The figure stresses also that the chromium coatings deteriorate because the Cr(V1) species dissolves slowly forming soluble chromic acid (197). Still, these coatings are far from ideal and are not yet sufficiently stable for technical application. Their reoxidation is strongly hindered, necessitating a relatively high overpotential of several hundreds of millivolts and their stability is insufficient, with chromium bleeding out as chromate. Yet the results obtained by the method that allows fixing with an easily applicable method a soluble redox mediator to an oxidic anode is encouraging. The an-

ADVANCES IN APPLIED ELECTROCATALYSIS

157

0

(HCrO;)

TiAnode

Oxide-

layer

Electrolyte

FIG. 3 2 . Schematic of the redox reaction of chromium species attached to the surface of titanium anodes

odic oxidation of isopropanol to acetone has been shown to proceed with 100% selectivity and current yield-although with very low current densities (0.5) mA/cm’) (198).

c.

ELECTROCATALYTIC HYDROGENATION AND

ELECTROCATALYZELI MEDIATED REDUCTION Commercial catalytic hydrogenations of unsaturated compounds use Raney nickel or-less commonly-Pt catalyst supported on active carbon. Electrocatalytic hydrogenation can be performed at platinized platinum or other platinum-metal electrodes. Adsorbed hydrogen atoms are the active reactant in catalytic as well as in electrocatalytic hydrogenation. In general cathodic electrocatalytic hydrogenation and catalytic hydrogenations are not supposed to differ with respect to mechanism, yield, and selectivity (199, 200). Since the adsorbed hydrogen as well as the adsorbed substrate participates in the reaction and the adsorption equilibrium coefficients of both reactants vary independently and often in adverse manner with the electrode potential, the effective hydrogenation rate may change not-as expected-exponentially with increasing cathodic polarization. Fujikawa and Kita, elucidating the kinetics of electrocatalytic hydrogenation of ethylene at platinum cathodes showed that the surface coverage of Hdd increases from 20 to almost 100% in changing the electrode potential from +0.2 to 0 V vs the reversible hydrogen electrode. Simultaneously, ethyl radical coverage decreases from almost 100 to less than 10% so that the rate of electrocatalytic ethylene hydrogenation with C2Hs H -+ C2Hbas the rate-determining

+

158

HARTMUT WENDT, W E N RAUSCH, A N D THOMAS BORUClNSKl

step depending on the product of the surface concentrations of H a d and C?Hs,,rshows a maximum at +O. 15V vs RHE (See Fig. 33) (201). Beck has shown (200) that electrocatalytic hydrogenations compare favorably with catalytic hydrogenation. The latter being H2 mass transfer hampered are by a factor of from 5 to 50 slower than electrocatalytic hydrogenations as the hydrogen is generated at the site where it is consumed, so that mass transfer for hydrogen is irrelevant with the electrocatalytic reaction. Use of Kaney nickel-coated cathodes for electrocatalytic hydrogenation would promise to take advantage of the enhanced hydrogen concentration (and surface concentration) in the nanopores of the catalyst. As discussed in the context of cathodic hydrogen evolution and utilization of nanoporous catalyst coatings (Section IV.B.3) it is possible to achieve hydrogen concentrations in these narrow proces that are by a factor of up to 1000 higher than the equilibrium concentration at 1 bar hydrogen. Very little systematic work has been conducted with Raney-nickel cathodes. Junghans reported on electrocatalytic hydrogenation of steroids, and he stresses that electrohydrogenation with Raney nickel yields higher selectivities (truns- vs cis- hydrogenation) than the usual catalytic hydrogenations (202, 20.3). Pletcher ct ul. (204) and Lassard and co-workers (205) have investigated the influence of the nature of the electrode material on rate, current yields. and selectivities of the electrocatalytic hydrogenation of acetophenone, phenanthrene, and other unsaturated compounds. They showed that substances that are inherently difficult to hydrogenate (for instance, styrene and benzonitrile) cannot be electrocatalytically hydrogenated at Pt electrodes at which hydrogen evolution is strongly catalyzed; therefore, the activity of the

1.2 1.o

0.8 L

a ”

0.6 0.4

0.2 0.0 50

100

150

200

250

E vs. R H E / (mV) FIG.3 3 . Dependence of surtace coverage of (a) ethyl radicals CLH5., (b) hydrogen atoms, and (c) ratc of ethylene hydrogenation at Pt cathodcs.

ADVANCES IN APPLIED ELECTROCATALYSIS

159

gas is low, corresponding to approximately 1 bar effective pressure or less. These substances can, however, be hydrogenated at Raney nickel cathodes, because the effective activity of hydrogen is increased by several orders of magnitude in the nanopores. Recently, Lasia and co-workers performed some kinetic studies, which however are not yet conclusive (206) with respect to rates and mechanisms. It is not known whether these advantages of electrocatalytic hydrogenation are used commercially. The counterpart of anodically electrocatalyzed oxidation by redox oxides, namely the cathodic reduction of organic substrates by surface-coupled redox system with sufficiently negative redox potential, is almost unknown. Beck reports that specially prepared Ti02 coating on Ti -electrodes can be reduced cathodically and that the electrogenerated Ti(1II) and Ti(I1) species do in fact reduce nitrobenzene to aniline (207). Still details are not yet known and a sound judgement on the technical feasibility of this reaction mode cannot yet be given. SURFACE AS MEDIUM CATALYZING D. THEELECTRODE CHEMICAL REACTIONSOF ELECTROGENERATED REACTIVE ORGANIC INTERMEDIATES One of the present authors has investigated the importance of the nature of the electrode/electrolyte interface for the yields and selectivities of some anodic electrosynthesis reactions. A series of four successive reviews reports on the gathered information and improved understanding of the chemical kinetics of reactive intermediates generated at the interface carbon electrodeinonaqueous solvent (208 -212) and citations of detailed investigations therein. The electrosorption of reactive intermediates and of organic molecules at this interface is generally weak. due to physical adsorption. Nonetheless, in particular if the reactive intermediates are so reactive that they do not survive for much longer than lo-’ sec and therefore cannot escape frotn the electrode surface, the chemical composition of the adsorbate layer being different from that of the bulk electrolyte composition influences the course of consecutive reactions and their yields and selectivities decisively. I.

The Electrode Surface as Mediutar and Catalyst for Chemicul Reactions of Electrogenerated Radicals and Radical Ions

As discussed elsewhere (212), the one-electron oxidation cir reduction of, e.g., unsaturated hydrocarbons or other electroactive organic compounds and ions [Eqs. (37a)-(37d)] creates radical cations, radical anions, or radicals with an energy content, which surmounts that of the original substrate

160

HARTMUT WENDT, SVEN KAUSCH, AND THOMAS RORUCINSKI

by at least 100 kJ/mol, provided that their oxidation or reduction potential amounts to more than + I V or - 1 V resp. vs aqueous SCE. Correspondingly, the primarily generated oxidation or reduction products of organic substrates are so reactive that they do not need any significant further catalytic activation, e.g., by chemisorption at the electrode surface looked at as heterogeneous chemical catalyst. Nonetheless, a number of electroorganic synthesis reactions are known whose outcome i.e., whose yield and selectivity, is decisively determined by the nature of the electrode so that heterogeneous acceleration of at least one of several competitive reactions of the electrogenerated reactive intermediates might be anticipated. A famous case is the Kolbe reaction, which is essentially the anodic dimerization of alkyl radicals that are generated at platinum anodes by anodic oxidation of the anions of carboxylic acids: K-CHZ-COO.

+

R-CHI-COO

'

+

R-CH?

ZR-CH? '

+

K-CH2-CHL-R.

- '6

R-CHZ-COO

'

+ COL

(3%) (3%) (3%)

If these anions are oxidized at carbon anodes instead of Pt anodes, the main product is not the Kolbe dimer but an ester of the original carboxylic acid. This reaction (Hofer-Moest) is explained by the inherent instability of C radicals at highly anodic potentials, which are necessary for the anodic oxidation of carboxylate anions. At these potentials, the C radicals are oxidized to carbonium ions that react with carboxylate anions forming esters: KCH2' RCHi

-

+ RCH?C(O)O

+

--j

RCHi

RCH?-O(O)CCHjR

(404 (4W

Thermodynamically, the latter reaction is to be expected, as C radicals exhibit an oxidation potential that is at least 1 V more negative than the oxidation potential of carboxylate anions (213, 214 1. It is therefore intriguing to understand what is the particular role of the platinum/electrolyte interface in the Kolbe synthesis favoring that reaction path-Eqs. (39a)-(39c)-which is thermodynamically disfavored and unlikely to occur. A closely related reaction whose kinetics are easier to investigate with conventional electrode kinetic methods is the anodically initiated addition of N3 radicals to olefins, discovered by Schafer and Alazrak (215). The consecutive reactions, which follow the initial generation of the reactive intermediate, an NJ radical, are somewhat slower than that of the Koibe radicals, so that their rate influences the shape and potential of the current voltage curves which can be evaluated in terms of reaction rates and rate laws.

ADVANCES IN APPLIED ELECTROCATALYSIS

161

According to detailed electrode kinetic investigations, this reaction starts with the anodic generation of azide radicals from azide anions ( 2 1 6 ) . N1 - e

(4 1 a)

-+N3..

Addition of these radicals to olefinic double bonds Ni . t R-CH=CH2

+ (R-CH-CHzN3)

.,

(4 1b)

e.g. to styrene, competes with annihilation, of the radical-radical dimerization followed by release of molecular nitrogen: 2N3.

-+

Ng

+ 3N2.

(4 1c)

The consecutive reactions of the addition radical (R-CH-CHlR). follow the same pattern as the reaction of the Kolbe radicals. At Pt anodes they undergo radical-radical reactions forming either dimers (41d) or bisubstituted monomers (41e), 2(R-CH-CH?N3) . -+ ((N3)(R)CHCH;)2 (41d) (R-CH-CH2N3).

+ N3

+ RCH(N,)-CHZN3

(414

R-CH'-CHzNj

(4 1 f )

and at carbon anodes they are oxidized (R-CH-CHZN~)

-

F

-+

to carbenium cations [Eq. (41f)], which solvolyze-e.g., react with the solvent methanol [Eq. (4lg)l-or react with other nucleophiles [Eq. (41 h)] present in the solvent, e.g., with azide anions.

+ HOMe R-CH(OMe)-CH2N2 + H ' (418) R-CHt-CH2N3 + Nu- + R-CH(Nu)-CHrN3, (4 1h) where Nu = Ny,OR-, CHsCN, etc. It has been shown that at Pt anodes, R-CH'-CH,N1

---f

radicals, such as N3 * or CH3CO0 -, are obviously strongly adsorbed and simultaneously stabilized against further oxidation. At such chemically modified Pt-electrode surfaces, consecutive chemical reactions proceed comparatively slowly, which shows up in a relatively large steady-state concentration of Nl radicals in the immediate vicinity of the anode, which gives rise to an anodic shift of the current voltage curve accordingly. The reason is that adsorption of larger molecules and radicals as styrene or the azidoethyphenyl radical, the intermediate generated by addition of azide radicals to styrene, is repressed by strongly adsorbed and deactivated N3. radicals. As a consequence, the further oxidation of these strongly chemisorbed radicals at the anode is inhibited. Therefore physisorbed radicals desorb and their further reactions proceed as homogeneous reactions in the diffusion layer close to the electrodes, provided that Pt anodes are used (2 17).

162

HAWMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

TABLE IV Physisorption Equilibria of Styrene and Azide Radicals

Electrode Pt Fe304 Glassy carbon Graph. carbon

Krd (styrene)* ( M )

K,,(Nd(M)

I 2

10

I

9

If, however, carbon anodes are used, the initial radicals RCHCH,N, are only weakly adsorbed and not stabilized in the adsorbed state. The adsorption of the intermediate radicals that are formed in consecutive steps is predominant and the adsorbed radicals are readily oxidized. The consecutive reactions of the carbenium cations then proceed according to heterogeneous kinetics at the surface of the carbon anode (218, 219). This is the explanation of the cathodic shift of the current voltage curves at glassy and graphitic carbon anodes vs Pt anodes, which is to be understood in terms of lowered radical concentration due to faster consecutive oxidation, which is strongly retarded at Pt anodes. As a quantitative result of these investigations, Table 1V presents data of the adsorption equilibrium coefficient for physisorption of styrene as well as for azide radicals as obtained from evaluating the anode potential of the azide anion oxidation and prevailing strong chemisorption of azide radicals at platinum, magnetite, glassy carbon, and graphitic carbon anodes from acetonitrile. It is obvious that at graphitic carbons the physisorption of styrene as well as of azide radicals is much stronger than at the three other anode materials. Thus contrary to the naive understanding, Pt does not catalyze but prohibits the thermodynamically favored oxidation, allowing for homogenous dimerization of the Kolbe radicals and similar radicals due to blocking the anode by a layer of adsorption-stabilized species, whereas C anodes allow for physisorption of radicals with subsequent immediate anodic formation of carbenium cations, which constitutes a completely different reaction path. So the Kolbe syntheses may be named a noncatalyzed rather than an electrocatalyzed reaction and the Hofer-Moest reaction (41g)-(41 h), although in most cases undesired, is heterogeneously catalyzed.

2. Heterogeneous Cutulysis in the Anodic Oxidation of Olefins and Bisarene Suijides Contrary to the noncatalytic Kolbe synthesis or other anodically initiated radical reactions at Pt anodes, the anodic functionalization and anodic coupling of vinyl compounds by direct anodic oxidation of olefins in alcohols as

163

ADVANCES IN APPLIED ELECTROCATALYSIS

solvents (MeOH in particular) is heterogeneously catalyzed in its decisive consecutive chemical steps due to physisorption of the reactive intermediate -a carboradical cation. In particular, this effect can be observed at graphitic carbon anodes (220). The reaction starts with the anodic oxidation of an adsorbed olefin molecule to the respective radical cation (R-CH=CH?)ad

-

e-

+

(R-CH=CHZ)

(42a)

+ad,

which is also adsorbed, followed either by nucleophilic addition of this radical cation to an unoxidized olefin, (R-CHSH2) 1

+ (R-CH=CH2)ad

+R

CH . -CH2-CH2-CHi

R

(42b)

which paves the way to anodic dimerization:

+ 2HOR

RCH . -CH,-CH2-CH+R

- e- - 2H+ + (RCH(OR)-CH,-),.

(4%')

The alternative reaction path that leads to the bifunctionalized monomer is the solvolys is of the initial radical cation (R-CHSHz):

+ HOR

+

RCH

*

-

CH20R

+ H+,

(424

followed by immediate further oxidation and hydrolysis (219, 220).

+ HOR + RCH(OR)-CH,OR + H+.

RCH . -CH20R - e -

(42~')

The kinetic circumstances have been elucidated in a series of papers by one of the present authors and his associates (221-225). As the current voltage curve is typically irreversible or close to irreversible, its analysis reveals only details concerning the first charge-transfer step [Eq. (42a)], which has been shown to follow physisorption of the olefin. The consecutive reactions have been investigated by the method of "preparative" voltammetry, namely by analyzing the product mixture, which is essentially composed of the anodic monomer and dimer independence of working potential or current density resp. Correlating the respective current and mass yields of monomers and dimers with current density allows the kinetic laws governing the selectivities at different anode materials to be obtained. At the mass-transfer limited current density (i/i~ = 1) always the dimer yield vanishes and it approaches its respective highest value, which depends-but not linearly-on the styrene concentration at vanishing current densities. This result can be formalized by assuming the coupling reaction [Eq. (42b)l and the solvolysis [Eq. (42c)l of the primary radical cation as a heterogeneous reaction of adsorbed reactants. Under this condition, one obtains the expression for the dimer yield, yield Dllncr =

k,d

K::'

*

c""' r;=o. (,$$'. K : d .

~ o1-0 lef

+ k"J',

(43)

164

H A H T M U T WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

TABLE V Anodic Styrene Oxidution in Merhunol

Anodc

Relative adsorption coefficients for physiosorbed styrene" K,ld(electr.X)/K,,(C)

Relative dimer yield," c,,ync.nc (Mol/dm')

I 0.2 0.1 0.005'

Graphitic carbon Fe 0, PbOz PI

I 10 5.10-'

2.10

'

"Dimer yield at carbon anodes with I M styrene approaches 100'%. "Adsolute surface concentration of styrene on carbon anodcs in equilibrium with I M styrcne is much lower than saturation concentration. ' Surface concentration of physisorbed styrene on Pt anodes is very low because of strong chemisorption of styrene and the solvent, methanol, which leads to complete coverage of the Pt surface with chemisorbed species.

where kad is the heterogeneous rate coefficient of second order for the addition reactional (42b), k,,,l, is the heterogeneous pseudo-first-order rate coefficient for the solvolysis reaction (42c), K$' is the adsorption equilibrium coefficient for the olefin, and c% is the concentration of the olefin at the electrode surface. This may be lower than the olefin bulk concentration due to mass transfer limitations. Table V collects the dimer yields that represent the essential quantity of' the preparative data of anodic styrene oxidation. The table also contains the adsorption equilibrium coefficients K%' for styrene obtained by evaluating current voltage curves. Again, the particular physisorptive properties of graphitic carbon for unsaturated molecules are stressed by the data of Table V. A comparable positive influence as experienced with anodic dimerization of olefins, styrene in particular, exerts carbon anodes on the anodic formation of trisarene sulfonium cations by anodic coupling of diarylsulfide radical cations to arenes (226-228). Ar-S-Ar-r Ar-S-' Ar

+ ArH + ((Ar), SArH)'

(44b)

+ H'

(444

- e--H+

(44d)

((Ar)2SArH)++. (A& S . ArJS + (Ar), S'

Ar-S'-Ar

(444

= ArS'Ar

+ Ar-S-Ar

+

Ar-S+Ar

I

Ar-S -A1

(444

ADVANCES IN APPLIED ELECTROCATALYSIS

165

Since arenes are adsorbed much better on carbon than on platinum, it is possible to avoid undesired anodic self-coupling (44e) of bisarene sulfides at carbon anodes but not at Pt anodes. These two reaction types are anodic examples of the heterogeneous catalytic enhancement of the rates of sufficiently reactive intermediates, brought about by increasing the local concentration of a reaction partner due to preferential physisorption of the respective moiety. 3 . Cathodic Pinucolisation qf Aliphatic Aldehydes and KivoneA

Cathodic pinacolisation of 0x0-compounds [Eq . (45a)l has been studied widely-the work has recently been reviewed (229). Some observations point toward the key step of the reacton, the radical dimerisation [Eq. (45c)], to be a heterogeneous reaction (230): 2 R , Rz CO

Rl RJ CO

+ 2e + 2 H'

+r +

Hi

+

+-+

2 Ri Rz ( O H ) C .

-j

(Rl Rr C(OH))z

(454

R , R , (OH) C .

(434

(Rl R>C(OH))Z

(45c)

The cathodic pinacolisation of 2- and 4-acetylpyridine, which had been investigated by one of the present authors (231-233), offered the chance for a complete kinetic analysis as the respective current voltage curves are of reversible character. They allow for evaluation of the kinetics of consecutive reactions, and one can show that at low pH reaction, Eq. (45c) is only possible if strong surfactants are absent. Such surfactants, by occupying the electrode surface, displace ketyl radicals, RIR2(0H)C., from the electrode surface because the latter are relatively weakly adsorbed and cannot compete with strong surfactants in adsorption. Ketyl radicals dissolved in aqueous or organic solvents of low pH are protonated in a fast almost diffusion-controlled reaction. After protonation they are further immediately reduced to form the monomeric carbinol instead of the hydrodimer-the pinacol: R I Rl(OH)R.

+ Hi + r

-j

R I Rz(0H)RH.

(454

The particular enhancement of a radical dimerisation reaction, Eqs. (45a) and (45b), is observed on hydrophobic metals like Hg, Pb, and Sn at potentials close to their respective point of zero charge where electrosorption is strongest and diminishes at more negative and positive potential. If the reduction potential of the 0x0-compound, which in accordance with the equilibrium reaction (45b) is pH dependent (60 mV per pH unit), is shifted by more than 100 mV away from the point of zero charge, then the ketyl radicals are no longer adsorbed strongly enough and they are displaced from the

166

HARTMUT WENDT, W E N RAUSCH, AND THOMAS BORUCINSKI

surface by solvent molecules so that only benzhydrol instead of pinacol is produced according to homogeneous kinetics, i.e., protonation vs dimerization leading to alcohol formation. Adding strong surfactants to the clectrolyte has the same effect even at potential of zero charge. As they displace the ketyl radicals from the cathode surface, benzhydrol is the main product, not pinacol. This type of reaction exemplifies that preferential adsorption of reactive intermediates catalyzes reactions with reaction orders of the intermediate of greater than I because adsorption of the reactant provides under otherwise comparable kinetic conditions a locally enhanced (surface) concentration of the intermediate. Model calculation show that at moderate current densities of from 0.01 to 0. l Akm' the concentration of the intermediate is enhanced due to adsorption by a factor of 10 to 30 leading to a rate enhancement of a bimolecular reaction vs a monomolecular one by a factor of 100 to 1000 (219). Action of Nonrtwctunl Electrosorbed 4. Elt~c,trac.utulytic. Spec.irs-Elec~trocatulysis of the Second Kind

If as surface active agent with the reduction of 2- and 4-acetylpyridine the optically active proton-donating alcaloids strychninium or brucinium are adsorbed at the cathode, relatively high yields of optically active alcohol (benzhydrol) are obtained (234, 235). A detailed study (23f-233) reveals that most probably the protonated, adsorbed alcaloid offers the chird template for the inductive synthesis. I t participates as a proton donor in a fast preestablished 2e-/2H+ equilibrium followed by a final desorptive step that establishes and stabilkes the acquired molecular configuration. Figure 34 describes the reaction model-albeit somewhat naively and oversimplified.

H' PyrCOCH,

'HPyrCOCH,

25%

75%

I FIG. 34. Schematic presentation of the reaction scheme for heterogeneous electrocatalysis of the second kind: catalytic protonation of carbonyl-radical anions by an electrosorbed proton donating surfactant (STR. strychninium).

ADVANCES IN APPLIED ELECTROCATALYSIS

167

The model had been substantiated by measuring the potential dependent electrosorption isotherms of all species involved that show that the protonated alcaloids are those species that are by far most strongly adsorbed, whereas the acetyl pyridines are least strongly adsorbed, especially at lower pH relative to the electrosynthesis, which is performed at pH 4 to 4.5. The optically inductive reduction of 2- and 4-acetyl pyridine to the optically active carbinols demands the formation of a dense but not too densely packed surface layer of adsorbed protonated alcaloid, which still allows for insertion of the 0x0-compound or the ketyl radical, respectively. Performing the reaction with too high alcaloid concentrations leads to compaction on the adsorbate layer, the ketyl radical is squeezed out, and optimal induction is no longer observed. One could call this type of electrocatalysis, which is due to the catalytic action of adsorbed species, "electrocatalysis of the second kind." Most remarkably the selectivity and commercial success of the Monsanto processthe hydrodimerisation of arylonitrile to adipodinitrile2CH2=CH-CN

.f

2e- - 2Hf -+ (NC-CH2-CN)2,

(46a)

is founded on electrocatalysis of the second kind. The hydrodimerisation of acrylonitrile can only effectively compete with the undesired reduction of acrylonitrile to propionitrile, CH?=CH-CN

+ 2 ~ +- 2H'

+

(CH~-CHL-CN)?,

(46b)

if strongly surfactive tetraalkyl ammonium cations are present in the electrolyte. Guidelli and associates showed (236 ) by electrode impedance measurements that acrylonitrile is coadsorbed, into an adsorbate layer of these cations, CHLZCH-CN (CHz=CH -CN),4,,, (46c) --j

and that it and any intermediate formed by cathodic reduction of acrylonitrile experience an environment that is far less protic than the bulk of the electrolyte, which is mainly water. If the acrylonitrile surface concentration in the adsorbate layer is sufficiently enhanced, due to sufficiently high acrylonitrile concentrations in the bulk of the electrolyte, the Michael addition [Eq. (46d)], of the primarily formed radical anion of the coadsorbed acrylonitrile can compete effectively with (retarded) protonation [Eq. (46e)l. Eventually further reduction and protonation [Eq. (46f)l yields the desired hydrodimer, an important electrochemically produced commodity: (CH:=CH -CN),d

+ (CH=CH-CN),d

(CHz=CH-CN): NC-CH-CH?-CH?-C-CN

+

NC-CH-CHL-CHZ-CH -CN

+ 2H' + e - + CH?=CH2-CN

+ e - + 2H'

+

(NC-CH2-CHz)2.

(464 (46e) (46f)

168

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

VII.

Summary

Electrocatalysis is, in the majority of cases, due to the chemical catalysis of the chemical steps in an electrochemical multi-electron reaction composed of a sequence of charge transfers and chemical reactions. Two factors determine the effective catalytic activity of a technical electrocatalysts: its chemical nature, which decisively determines its absorptive and fundamental catalytic properties; and its morphology, which determines mainly its utilization. A third issue of practical importance is long-term stability, for which catalytic properties and utilization must occasionally be sacrificed. Due to the very nature of electrochemical processes, which apply mainly to reactions of relatively large positive Gibbs energies, the variance in electrode reactions of technical interest is small, comprising mostly the electrochemistry of hydrogen, oxygen, and chlorine. This small variance is the reason for the high standard achieved in technical electrocatalysis during the past three decades. Today, fuel cell catalysts and electrodes are produced commercially under strict quality control measures. This achievement was made possible because a thorough understanding of the role of optimized micro- and nanomorphologies of electrodes and their catalytic effectivities had been established, and this knowledge had been evaluated and applied to optimized dedicated production procedures. Electrocatalysis of electroorganic synthesis reactions remains a challenge to the electrochemist and electrochemical engineer. Commercial electroorganic syntheses are still scarce and the field of electrochemical technology is developing slowly. But continuous efforts could open up an interesting new field of electrochemical synthesis of commercial interest in which electrocatalysis might play an important role.

RIFERENCES

I . (a) Grubb, W. T., “Low Temperature Hydrocarbons 17: Annual Power Sources Conference, Atlantic City, 1963” [Cited by J. 0’ M. Bockris in (2a)l; (b) W. 1’. Crubb, Ncitrtrc, (London) 198, 883 (1963).

2. (a) Bockris. J. 0’ M., and Minevski, Z. S. I n t . Hydroget! Energy 17,423 (1992). (b) Kordesch, K . , “Brennstoffbatterien.” Springer-Verlag. Wien New York, 1984. 3. Beer, J . B., J . Ekctrochem. So<. 127, 303c (19x0). 4 . de Nora, V., “Proceedings, Symposium on Performance of Electrodes for Industrial Electrochemical Processes.” (Hine, F.. Fenton, J. M., Tilak, B. U., and Lisius, J. B., Eds.), Vol. PV-89-10, p. I . Electrochem. Soc., Pennington, 1989. 5 . Hine, F.. Fenton, J. M., Tilak, B. U., and Lisius, J. B., Eds., ”Proceedings, Syrnposiuni on Performance of Electrodes for Industrial Electrochemical Processes,” vol. PV-Xg- 10, Electrochem. Soc., Pennington, 1989. 6 . Trasatti, S., E/ec/rochirn. Actu 32, 369 (1987). 7. Trdsatti, s., Ekctrochim. Actu 29, 1503 (1984). 8 . Trdsatti, S., in “Advances in Electrochemical Science and Technology” (Gerischer, H., and Tobias, W., €cis.), Vol. 2 , p. I . VCH Verlagsgesellschaft, Weinhein], 1990.

ADVANCES IN APPLIED ELECTROCATALYSIS

169

9. Mclntyre. J. E. D., Weaver, M. J.. Yeager, B. E., Eds., “Proceedings, Symposium ‘The Chemistry and Physics of Electrocatalysis’,” Vol. PV-84- 12. Electrochem. Soc., Pennington 1984. 10. Sherson, D., Tryk, D., Daroux. M., and Xing, X., Eds., “Proceedings, Syposium ‘Structural Effects in Electrocatalysis and Oxygen Electrochemistry’,” Vol. PV-92- 1 1. Electrochem. Soc., Pennington, 1992. 1 1 . Krasilshchikov, A. I . , Zh. Fiz. Khim. 37, 531 (1963). 12. Trasatti, S . , in “Electrochemical Hydrogen Technologies” (H. Wendt, Ed.), p. 104. Elsevier, Amsterdam, 1990. 13. Wiesener, K., and Ohms, D., in “Electrochemical Hydrogen Technologies” (H. Wendt, Ed.), p. 63. Elsevier, Amsterdam, 1990. 14. Wendt, H., Ed., “Electrochemical Hydrogen Technologies.” Elsevier, Amsterdam, 1990. 1.5. R. A. Marcus, 1. Chem. Phys. 24, 966 (1956); 43, 679 (1965). 16. Wendt, H. and Plzak, V., in “Electrochemical Hydrogen Technologies,” p. 15. Elsevier, Amsterdam, 1990. 17. Hofmann, H. and Wendt, H.. in “Proceedings, Symposium ‘The Chemistry and Physics of Electrocatalysis’ ” (Mclntyre. J. E. D.. Weaver, M. J.. and Yeager, B. E.. Eds.). Vol. PV-84- 12, p. 537. Electrochem. Soc., Pennington 1984. 18. Schultze, J. W., in “Elektrochemische Stoffgewinnudg” ( Kreysa. G., Ed.), Drchemu Monogruphie. Vol. 125, p. 27. 1992. 19. H. B. Beer Magneto Chemie N.V., British Patent No. 855, 107 (1958). 20. H. B. Beer British Patent No. 6490, 67 (1967). 21. Yoshida, M., and Noaki, Y., in “Proceedings, Symposium on Performance of Electrodes for Idustrial Electrochemical Processes” (Hine, F., Fenton, J. M., Tilak, B. U., and Lisius, J. B., Eds.), Vol. PV-89-10, p. 15. Electrochem. Soc., Pennington. 1989. 22. Hine, F., “Electrode Processes and Electrochemical Engineering,” p. 147. Plenum Press, New York London, 1985. 23. Krishtalik, L. J., Elec,rochim. Actu 16, 329 (1981). 24. (a) Harrison, J. A , , in “Modern Chloralkali Technology” (Jackson, C., Ed.), Vol. 2, p. 246. Ellis Horwood, Chichester, 1983; (b) Janssen, L. J. J., in “Modern Chloralkali Technology” (Jackson. C., Ed.), Vol. 2, p. 271. Ellis Horwood, Chichester, 1983. 25. Hirakata, K . , Itakura, N . , Shibuya, M., and Ishi, G . , in “Proceedings, Symposium on Performance of Electrodes for Industrial Electrochemical Processes” (Itine, F.. Fentor, J. M., Tilak, B. U., and Lisius. J. B., Eds.), Val. PV-89-10, p. 161. Electrochem. Soc., Pennington 1989. 26. Jackaon, C., Ed., “Modern Chloralkali Technology,” Vol 2. Ellis Horwood, Chichester, 1983. 27. (a) Hine, F., and Yasuda, M., J . Electrochem. Soc. 124, 500 (1977). (b) Hine, F., Yasuda, M., Noda, T., Yoshida, T., and Okuda, J., J . Elecfrochem. Soc 126, 1439 (1979). 28. Brooks, W. N., Denton, D. A , and Sammes, N. M., in “Proceedings, Symposium on Performance of Electrodes lor Industrial Electrochemical Processes” (Hine, F., Fenton, J . M., Tilak, B. U . , and Lisius. J. B., Eds.), Vol. PV-89-10, p. 39. Electrochem. Soc., Pennington, 1989. 29. Hansen, H. J., Krenk, C. W., Koster, T. P. M., and Macker, A,, in “Elektrochemische Stof~gewinnung-Grundlagen und Verfahrenstechnik” (Kreysa, G., Ed.), Dechenia Monographien, Vol. 125, p. 413, 1992. 311. Trasatti. S . , Ekclrochim. Actu 29, 1503 (1984). 31. Kotowski. S . , Parr, J., “Titanium Anodes Used in Electrochemical Processes, Chemical Plants and Processes,” reprinted by Herdeus Elektrochemie, Firmenschrift, March 1992. 32. Cominellis, G . , and Plattner, E., in “Proceedings, Symposium on Performance of Electrodes for Industrial Electrochemical Processes” (Hine, F . , Fenton, .I.M., Tilak, B. U . , and Lisiu, J. B., Eds.), Vol. PV-89-10. p. 229. Electrochem. Soc., Pennington, 198Y.

170

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

33. Denton, D. A., DeSouza. J. T., Entwisle, J. H., Lee, D . . and Wilson, H. G . . in “Modern Chloralkali Technology” (Jackson, C . , Ed.), Vol. 2, p. 215. Ellis Horwood. Chichester, 1983. 34. Schmittinger, P., in “Electrochemical Hydrogen Technologies” (Wendt, H . , Ed.), p. 261. Elsevier, Amsterdam, 1990. 35. Morris, G . J., and Wilson, R . D., in “Proceedings, Symposium on Performance of Electrodes for Industrial Electrochemical Processes” (Hine, F . , Fenton. J . M., Tilak, B. U., and Lisius, J . B., Eds.), Vol. PV-89-10, p. 147. Electrochem. Soc., Pennington, 1989. 3 6 . Brook, W. N. V., C h m . Bri. 22, 1015 (1986). 37. Shiinanune, T.. Kawanche, J., Nakamatsu, S . , Nishiki. Y., and Hayashi. R . . in “Proceedings, Symposiun on Performance of Electrodes for Industrial Electrochemical Processes” (Hine, F.. Fenton, J. M., Tilak, B. U . , and Lisius, J. B., Eds.), Vol. PV-8910, p. 69. Electrochem. Soc., Pennington, 1989. 3H. Muranaga, T.. Kanaya, Y.. and Yokota, N.. in ”Proceedings, Symposium on Performance of Electrodes for Industrial Electrochemical Processes” (Hine, F., Fenton, J. M . , Tilak. B. U.. and Lisius. J. B.. Eds.). Vol. PV-89-10. p. 77. Electrochem. Soc., Pennington, 1989. 39. Tseung, A. C. C . , and Jasem. S . , Electrochem Actu 22, 31 (1977). 40. Wendt, H., and Imarisio, G., J . AppI. Elektrochem. 18, I (1988). 4 I . Wendt, H., and Plzak, V., Eii,c.troc/zim. Actu 28, 27 (1983). 42. Schnurnberger. W., Henne, R . , and Von Bradke, M . , Adv. Hvdrogen Energy. Hwhugivi Energy Prog. 2, 933 (1984). 43. Tseung, C. Q., Jiang, S . P., and Tseung. A . C. C.. J . Elec/ro~hem.SOC. 139, 1.53.5 ( 1992). 4 4 . Ba1e.j. J . , and Divisek. J . , KfA Julich. Germany, personal communication. 45. Schmidt, T h . , and Wendt, H., “Electrokatalyse durch in-situ Formierung von anodischen Electrokatalysatoren;” poster exhibited at Jahrestagung der Fachgruppe Angewandte Electrochemie, Strassburg, France, October 1992. 46. Cocke. D. L., J . Meids 38, 70 ( 1986). 47. Yoon, C., and Cocke, D. L . , J . Non-Cost. Solids 79, 217 (1986). 48. Yokoyama, A . , KomiydnTd, K., h u e , H., Masumoto, T., and Kiwura, H. M.. J. Cutcrl. 68, 355 (1981). 4Y. Shibata, M . , Masunioto. T., in “Preparation of Catalysts IV” (Delmon, B.. Grange, P . , Jacobs, P. A , , and Poncelet, G.. Eds.) p. 353. Elsevier, Ainsterdam, 1987. SO. Baiker, A . , Furuduy Disc.. Chern. 87, 239 (1989). 51. Brower, W., Matyjaszczyk, M . , Pettit. T., and Smith, G . , Nuturr (London) 301, 497 (1983). 5 2 . Kreysa, G . , Goniez, J . , Baro, A , , and Arvia, A. J . , J . EIectrounul. Chrrn. 265. 67 (1989). 53. Kreysa. G . . Hakansson. B., Ekdunge. P., Elecrrochirn. ACIU33, 1351 (198XI -54. Ekdunge, P., Thulin, H., and Kreysa, G . , Chem. Ing. E c h o / . 61, 638 (1989). 55. G r o a I. A . . Prakash, 1.. and Yeager, E. B., in “Proceedings, Symposium on Performace of Electrodes for Industrial Electrochemical Processes” (Hine, F., Fenton. J . M., Tilak, B. U . , and Lisius. J. B.. Eds.), Vol. PV-89-IO. p. 205. Electrochem. Soc.. Pennington, 1989. 56. Sedlak, J. M., Lawrence. R . J . , and Enos, J . G . , 1n1. J . Hvdrogen Energy 6 , 1.59 ( 1981) . 57. Kotowski. S . , and Busse, B., in “Proceedings, Symposium on Performance of Electrodes for Industrial Electrochemical Processes” (Hine, F., Fenton. J. M . . Tilak, B. U., and Lisius. J. B., Eds.), Vol. PV-89-10, p. 245. Electrochem. Soc., Pennington, 1989. 58. Fcrron. C. G . . and Duby, P. F., in “Proceedings, Symposium on Performance of Electrodes for Industrial Electrochemical Processes” (Hine. F., Fenton, J. M., Tilak, B. U . , and Lisius, J . B.. Eds.). Vol. PV-89-10, p. 259. Electrochein. Soc., Pennington, 19x9.

ADVANCES IN APPLIED ELECTROCATALYSIS

171

59. Scherer, G . G., Derontay, M., Oberlin, R., and Stucki, S . , in “Technische Elektrolysen” (D. Behrens, Ed.), Dechemu Monogruphie, Vol. 98, p. 407, 1985. 60. Winter, C. J., and Fuchs, M., f n t . J . Hydrogen Energy 16, 723 (1991). 61. Simmrock, K. H., in “Electrochemical Cell Design” (White, R. E.. Ed.), p. 72. Plenum Press, New York/London, 1984. 62. Wendt, H., Ed., “Electrochemical Hydrogen Technologies,” p. 204. Elsevier, Amsterdam, 1990. 63. Eberson, L., and Utley, J. H. P., in “Organic Electrochemistry” (Baizer M., and Lund, H., Eds.). Dekker, New YorkiBasel, 1983. 64. Kreh, R. P., Spotnitz, R. M., and Lundquist, .I. T., “Electrosynthesis” (Weinberg, N. L., Ed.), Chap. 10, p. 187. The Electrosynthesis Company lnc., E. Amherst, N Y , 1990. 65. Toyo Soda Manuf. Co. Ltd., Japanese Patent JP 6017, 069, Chem. Abstr. 102, 194045 (1985). 66. Yamakawa, K . , Tsubakino, H., Akiyoshi, K . , and Yoshimoto, K . , “Ext. Abstract, Electrochem. Soc. Meeting,” p. 1633, 1987. 67. Mori, T.. Fukuda, K . , and Ameniya, K . , “Ext. Abstract, Electrochem. Soc. Meeting,” p. 812, 1987. 68. Harang, H., and Widding, K . , Norsk Hydro; German Offen. DE 2, 087, 627, Chem. Abs t r . 89, 188077b (1983). 69. Gosei, T., Chem. Ind. Co. Ltd., Japanese Patent 5967, 380 Chem. Ahsrr. 101, 62654d ( 1984). 70. Gosei, T., Chem. Ind. Co. Ltd., Japanese Patent JP 6092, 496, Chem. Abstr. 103, 94993w (1983). 71. Nidola, A., and Shira, R., Int. Hydrogen Energy 11, 449 (1986). 72. Vandenborre, H., and Vermeiren, P., Elerrrochim. Actu 29, 297 (1994). 73. Borucinski, Th., Doctoral thesis, Darmstadt, 1993. 74. Brontin, P., Ehrhardt, J . J., Pentenaro, A , , and Gras, J. M., J . M i c w s r . Spectrosc. Electron. 9 57 (1984). 75. Justi, E., Pilkuhn, M., Scheibe, W., and Winsel. A., Abhdlg. Murk. N u t . Klusse 8 (1959). 76. Miiller, J. M., Lohrberg, K., and Wiillenweber, H., Chem. Ing. Technol. 25, 453 (1980); Germany Offcn. DE 2,829,901, Chem. Abstr. 92, 171626~(1980). 77. Liederbach, T. A , , Greenberg, A. M . , and Thomas, V. H., in “Chlorine Cells in Modern Alkali Technology” (M. 0. Coulter, Ed.), p. 145. Ellis Horwood, Chichester, 1980. 78. Rausch, S., and Wendt, H., J . Appl. Elecrrochem. 22, 1031 (1992). 79. Trdsatti, s.,J . Electround. Chem. 33, 351 (1971); 39, 161 (1972). 80. Krishtalik, L. I., Furuduy Disc. Chem. Soc. 74, 205 (1982). 81. Rausch, S . , and Wendt, H . , J . Appl. Elektrochem. 22, 1025 (1992). 82. Heidrich, H-J., Muller, L., and Podlovchenko, B. 1.; J . Appl. Electrochem. 20, 686 (1990). 83. Lohrberg, K . , and Kohl, P., Electrochim. Act 29, 1557 (1984). 84. Hall, D. E., and Shepard, V. R., f n r . J . Hydrogen Energy 9, 1005 (1984). 85. (a) Asahi Glass Corp. Ltd., Japanese Patent JP 60 67, 687, Chem. Abstr. 103, 131,261d (1985); (b) Yoshida, N., Yoshitake, M., Endoh, E., and Morimoto, T., in “Proceedings, Symposium on Performance of Electrodes for Industrial Electrochemical Processes” (Hine, F., Fenton, J. M., Tilak, B. U . , and Lisius, J. B., Eds.), Vol. PV-89-10, p. 125. Electrochem. Soc., Pennington, 1989. 86. Brown, D. E., Fogarty, P. O., Mahmood, M. N., and Turner, A. K . ; in “Modern Chloralkali Technology” (Jackson, C., Ed.), Vol. 2, p. 233. Ellis Horwood, Chichester, 1982. 87. Cairns, J. F., Denton, D. A,, and Izasdt, P. A,; European Patent Appl. EP 129 374 (1984). 88. Nidola. A., PCT, International Appl. WO 86, 790, Chem. Abstr. 105, 122074 (1986). 8Y. Debrodt, H., European Patent Appl., E. P. 120 088, Chem. Abstr. 102, 122,040 (1984).

172

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

YO. Galizzioli, D . , Tandardini. F., and Trasatti, S., 1.Appl. Electrochem. 5, 203 (1975). Y I . Kotz, E. R . , and Stucki, S . , 1.Appl. Elecrrochem. 17, I190 (1987). 92. Schmidt, Th., Doctoral thesis, Technische Hochschule Darmstadt, 1994, German Patent

4232958.-2 (1.10.1992). Y3. Einbock, P.. Nadler, H. G., and Nowak, W., Chem. Ing. Techno/ 61, 557 (1989). Y4. Yoshida. M., and Naaki, Y.. ill “Proceedings, Symposium on Performance of Electrodes

for Industrial Electrochemical Processes” (Hine, F., Fenton, J. M., Tilak, B. U . , and Lisius. J . B., Eds.), Vol. PV-89-10. p. 15. Electrochem. Soc., Pennington, 1989. 95. Balandin, A . A , , Adv. Card. 19, I (1969). 96. Divisek, J . , in “Electrochemical Hydrogen Technologies” (Wendt, H . , Ed.), p. 137. Elsevier. Amsterdam, 1990. 97. Sievert. D., March, A , , and Stucki, S . in “Proceedings, Electrochem. Soc., Spring Meeting, Philadelphia, 1987. 08. Winsel, A . , and Richter, G. J., in “Electrochemical Hydrogen Technologies” (Wendt. H., Ed.), p. 381. Elsevier, Amsterdam, 1990. 99. George, T. J . , and Mayfield, M. J., “Fuel Cells, Technology Status Report,” DOE/ MTC-9010268, DE 90 00 9686 PAFC-program p. 8. US Department of Energy. Washington, DC. 100. Appleby, A . J . . and Selnian, J . R . , in “Electrochemical Hydrogen Technologies” (H. Wendt, Ed.), p. 42.5. Elsevier, Amsterdam, 1990. 101. Appleby, A. J., and Foulkes, F. R . , ”Fuel Cell Handbook.” Van Nostrand-Reinhold, New York, 1989. 102. Wendt, H., and Rohland, B., Krrnrechnik 56, 161 (1991). 103. Coehn, A . , and Osaka, Y., Z . Anorg. Chem. 34, 86 (1903). 104. Bowden, F. B., Proc. R . Soc. A 126, 107 (1929). 10.5. Hoare, J . P . , Pro(.. R. Soc. A 142, 628 (1993). 106. Hoare, J. P . , “The Electrochemistry of Oxygen.” Interscience, New York, 1968. 107. Hoare, J. P . , it1 “Encyclopedia of Electrochemisty of the Elements” (Bard, A. J . , Ed.), Vol. 2, p. 19 I . Dekker, New York, 1974. 108. Damjanovic, A . , in “Modern Aspects o f Electrochemistry” (O’M Bockris, J.. and Conway, B., Eds.), Vol. 5 , p. 369. Butterworths. London, 1969. IOY. Appleby, A . J . , in “Modern Aspects o f Electrochemistry” (Conway, B. E . , and O’M Bockris, J., Eds.), Vol. 9. p. 369. New York, 1974. 110. O‘M. Bockris, J . , in “Modern Aspects of Electrochemistry” (O’M Bockris, J . , and Conway, B. E., Eds.). Vol. I , p. 180. London, 1954. 11 I . Wiesener, K., and Ohms, I)., in “Electrochemical Hydrogen Technologies” (Wendt, H., Ed.), p. 63. Elsevier, Amsterdam, 1990. 112. O’M Bockris, J . , Hug, A. K. M. S., Proc,. R. Soc. London A 237, 277 (1956). I/.<.Tanata, A . , Kananiura, K., Adzic, R . , Cahan, B., and Yeager, E . , in “Proceedings, EIecrrochem. Soc. Meeting 90- I, Montreal, 1990,” p. 969. 114. Gland, J. L., Sertini, B. A., and Fischer, B. G . , Surf. Sci. 95, 587 (1980). 1 1 5 . Hillrichs, E., and Lohrberg, K . , in “Elektrochemische Stoffgewinnung” (Kreysa, G . , Ed.,) Dechema Monograph Vol. 125, p. 233, 1992. 116. Mclntyre, J. D.E.. and Peck, W. F., Jr., in “Proceedings, Symposium ‘Electrode processes 1979,”’ (Bruckenstein, s., Mclntyre, J. D. E . , Miller, B., and Yeager, E., Eds.), Vol. PV 80-3, p. 322. “Electrochem. Soc., Princeton, NJ, 1980. / / 7 . Collmann, .I.P., Marrocco, M . , Denisevich, P . , Koval, C . , and Anson, S . C . ; J . E k f r o a n d . Chetn. 101, I17 (1979). 118. Conway, B. E., and Liu. T. C., in “Proceedings, Symposium ‘Structural Effects in Electrocatalysis and Oxygen Electrochemistry’ (Sherson. D . , Tryk. 0 . . Daroux, M., and Xing. X., Eds.). Vol. PV-92-1 I , p. 395. Electrochem. Soc.. Pennington, 1992. ”

ADVANCES IN APPLIED ELECTROCATALYSIS

173

Schultze, J. W., and Haja, M . , Z. Phys. Chem. N . F. 104, 73 (1977). Tseung, A , , Bobbs. B., and Jantram, A , , Elecrrochim. Actu 15, 473 (1970). Goldstein, J. R., and Tseung, A , , J . Phys. Chrm, 76, 3646 (1972). Kadish, K. V., in “Handbook on Fuel Cell Technology” (Berger, E., Ed.), p. 361. Prenticc-Hall, Englewood Cliffs, NJ. 1968. 123. Wendt, H., unpublished results. 124. Rao, C . N., Golapakrishnan, J . , Vidyasagar, K . , Gangule, A . K . , Mainanan, A . , and Ganapdthi, L., 1.Mut. Res. 1, 280 (1986). 125. Karlson, G . , J . Power Sources 10, 319 (1983). 126. Goodenough, J. B., and Monoharan, R . , in “Proceedings, Symposium ‘Structural Effects in Electrocatalysis and Oxygen Electrochemistry”’ (Sherson, D., Tryk, D., Daroux, M., and Xing, X . , Eds.), Vol. PV-92- I I , p. 5 2 3 . Electrochem. Soc., Pennington. 1992. 127. Horowitz, H. S . , Longo, J. M . , and Horowitz, H. H . , J . Electrochem. Soc. 130, 1951 (1983). 128. Meadowcraft, D. B., Nuture (London) 226, 874 (1970). 12Y. Collrnan, J. P., Denisowitsch, P., Konai, Y., Marocco, M., Covol, C., and Anson, F., J . Am. Chem. Soc. 103, 6027 (1980). 130. Duraut, R . R . , J r . , Bencosme, C. R., Collmann, J. P., and Anson, F., J . Am. Chern. Soc. 105, 2710 (1983). 131. Tanaka, A., Gupta, S. L., Tryk, G . , Fierro, C., Yeager, E. B., and Sherson, D. A , , in “Proceedings, Symposium ‘Structural Effects in Electrocatalysis and Oxygen Electrocheniistry’” (Sherson, D . , Tryk, D . , Daroux, M . , and Xing, X . , Eds), Vol. PV-92-1 I , p. 555. Electrochem. Soc., Pennington, 1992. 132. Vielstich, W., Rer. Bunsenges. Phys. Chem. 94, 1030 (1990). 133. Wiesener, K . , Schneider, W., and Miibius. A , , J . Electrochem. Soc. 136, 377 (1989). 134. Miibius, A . , and Wiesener, K . , Bcr. Runsenges. Phys. Chern. 94, 1039 (1990). 135. Kinoshita, K., “Carbon-Electrochemical and Physicochemical Properties.” Wiley , New York, 1987. 136. Kinoshita, K., and Stonehart, P . , in “Modern Aspects of Electrochemistry” (O’M. Bockris, J., and Conway, B. F., Eds.) Vol. 12, p. 183. Plenum Press, New York, 1988. 137. Stonehard, P., Kinoshita, K., and Bette, J. A . S., E/ectrocutulysis 79, 275 (1976). 1-38. Bette, J. A. S . , in “Proceedings, Symposium ‘Structural Effects in Electrocatalysis and Oxygen Electrochemistry’” (Sherson, D., Tryk, D., Daroux, M . , and Xing, X., Eds.), Vol. PV-92-1 I , p. 573. Electrochem. Soc., Pennington, 1992. 139. Stonehart, P., Rer. Bunsenges. Phys. Chern. 94, 913 (1990). 140. Ito, T., and Katoh; K., N. E. Chem. Cat. Corp., European Patent 90-104453 (1990). 141. Jalan, V. M . , and Landsmann, D. A , , US Patent 5 , 186,110 (1980). 142. Jalan, V. M., Giner Inc., Eruopean Patent Appl. 165024 (1985). 143. Kush, A . , Christensen, L., and Maru, H. C.. in “Program and Abstracts, 1988 Fuel Cell Seminar, October 23-26, 1988, Long Beach, CA,” p. 173. Courtesy Associates, Washington DC, 1988. 144. Jenseit, W., Khalil, A., and Wendt, H . , J . Appl. Electrochem. 20, 893 (1990) 145. Gulzow, E . , Schnurnberger, W., “Degradation von PTFE-gebundenen Elektroden,” in ”Elektrochemie in Energie und Umweltschutz” (Winsel, A , , Ed.), Dechemu Monogr.. 124, p. 675, (1991). 146. Jenseit, W., Biihme, O., Leidich, F. U., and Wendt, H . , Electrochim. Aetu 38, 21 15 (1993). 147. Mund, K . , Richter, G., and Sturm, F. U., J . Elrcrrochem. Soc. 124, 1 (1977). 148. Mund, K . , and Sturm, F. U., Electrochim. Actu 20, 463 (1975). 119. 120. 121. 122.

174

HAIYTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

149. Rousar, R., Micka, K., and Kimla, A . , “Electrochemical Engineering,” Part I I , p. 109.

Elsevier, AinsterdamiOxfordiNew York/Tokyo, 1986. 150. Wendt. H., and Rohland. B., in “Electrochemical Cell Design and Optimization Procedures” (Kreya, G . , Ed), Dechemu Monogruph Vol. 123, p. 77, 1991. 151. Celiker, H., Saleh, M. A . , Gultekin, S. , and Sakri, A. S . , J. E/ec,trochcm. S w . 138, 1671 (1991). 152. Cervino, R. M., Triaca, W. E., and Arvia, A. J.. J . E/ec~troanu/.Chem. 182, 5 I ( 1985). 15.3. Cervino. R. M., Triaca, W. E., and Arvia, A. J.. J . E/ectrochein. Sot,. 132, 266 (1985). 154. Peschka, W., “Fliissiger Wasserstoff als Energietrager,” p. 54. Springer-Verlag. WieniNew York, 1984. 1.55. Sutton, R. D . , and Vanderhorgh, N. E., in “Program and Abstracts, I992 Fuel Cell Seminar, November 29-December 2, 1992, Tuczon AZ,” p. 201. Courtesy Associates, Washington, DC. 156. Kishakhevariam, N., and Stuve, E. M., in “Proceedings Symposium ‘Structural Effects in Electrocatalysis and Oxygen Electrochemistry’ ” (Sherson, D., Tryk, D., Daroux, M . , and Xing, X . , Eds.), Vol. PV-92-1 I , p. 40. Electrochem. Soc., Pennington, 1992. 157. Leger, J. M . , Lamy, C., Her. Bunsenges. Phys. Chem. 94, 1021 (1990). 158. Hammet. A.. Weeks. S. A . , Kennedy, B. J . , Troughton, G., and Christensen, P. A . , Ber. Bunsenges. Phys. Chum. 94, 1014 (1990). 159. Srinivasan, S . , Partasarakhty, A., Velev, O., Ferreira, A. C., Mukherjee, S. M., and Applehy, A. J . , in “Proceedings, Symposium ‘Structural Effects in Electrocatalysis and Oxygen Electrochemistry’” (Sherson, D., Tryk, D., Daroux, M., and Xing, X., Eds.), Vol. PV-92-1 I , p. 474. Electrochem. Soc., Pennington, 1992. 160. Iwasita, T., Nort, F. C . , and Vielstich, W., Ber. Bunsenges. Phvs. Chcvn. 94, 1030 (1990). 161. Yeager, E., Razak, M . , Gervasio, D . , Razak, A , , and Tryk. G.; in “Proceedings, Symposium ‘Structural Effects in Electrocatalysis and Oxygen Electrochemistry’ (Sherson. D . , Tryk, D., Daroux, M., and Xing, X.. Eds.), Vol PV-92-1 I , p. 440. Electrochem. Soc., Pennington, 1992. 162. Raistrick, F. D.; US Patent 5.876. I 15 (1989). 163. Uribe, F. A . , Wilson, M. S., E. Springer, Th. E.. and Gottesfeld, S . , in “Proceedings, Symposium ‘Structural EKects in Electrocatalysis and Oxygen Electrocheniistry’ (Sherson, D . , Tryk, D.. Daroux, M . , and Xing, X.. Eds.). Vol. PV-92-1 I , p. 494. Electrochem. Soc., Pennington, 1992. 164. Wilson, M. S.. Gottesfeld, S. , J . Appl. Electrochem. 22, 1 (1992). 165. Taylor, E. J . , Vilambi, N. R. C . , Anderson, E. B., and Donahme, K.. iii “Proceedings, Symposium ‘Structural Effects in Electrocatalysis and Oxygen Electrochemistry’ (Sherson. D., ’Iryk, D., Daroux, M . , and Xing. X., Eds.), Vol. PV-92-1 I , p. 540. Electrochem. Soc., Pennington, 1992. 166. Taylor, E. J . , Anderson, E. B., and Lilambis, N. R . K.. J . Electrochem. Soc., EIPctrochem. Soc. L e i / . 139, ( 5 ) . L 45 (1992). 167. Poltarewski, 2.. Staiti, P., Aldorucci, V., Wiezorek, W., and Giordano, N., J . E&c.trochem. Soc. 139, 761 (1992). 168. Ahn, J., Ledjeff, K.; Gcrrnan Patent P 4027655.4 (1990). 16Y. Lejeff, K.. Heinzel, A . , Mahlendorf, F., and Peinicke, V., in “Elektrochemische Energiegewinnung” (Kreysa, G., and Jiittner, K., Eds.), Dechema Monograph Vol. 12X p. 103, 1993. 170. (a) Ang., P. G . P., and Sommels, A . F., J . E/cctrochern. Soc. 127, 1279 (1980). (b) Lu, S . H., and Selinan, J. R . , J . E/ectmchem. Soc. 131, 1279 (1980). ”

”

ADVANCES IN APPLIED ELECTROCATALYSIS

175

Selman, R., 1.Electrochem. Soc. 136, 1058 (1989). Leidich, F. U., Bohme, O., and Wendt, H.. Int. J . Hydrogen Energy 19, 349 (1994). Appleby, A. J . , and Nicholson, S . B., J . Elecirounul. Chem. 53, 105 (1974). Appleby, A. J . , and Nicholson, S. B., 1.Elecirounul. Chem. 112, 71 (1980). Middleton, P. H . , Seiersten, M. E., and Steel, B. C. H.; in “Proceedings, 1st International Symposium on SOFC” (Singhal, S. C., Ed.), Vol. PV-89-1 l , p. 90. Electrochem. Soc., Pennington, 1989. 176. Koc, R., Anderson, H. H., Howard, S. A., and Spartin, D. M.; in “Proceedings, 1st International Symposium on S O F C (Singhal, S. C., Ed.), Vot. PV-89-1I , p. 220. Electrochem. Soc., Pennington, 1989. 177. Belzner, A., Gur, T. M. and Huggins, R. A , , in “Proceedings, 1st international Symposium on SOFC” (Singhal, S. C., Ed.), Vol. PV-98-11, p. 200. Electrochem. Soc., Pennington, 1989. 178. Rohland, B., “Beitrage zur effektiven energetischen und stofflichen Nutzung von Oxidationsprozessen mit Luftsauerstoff,” Habilitationsschrift, Universitat Greifswald, 1984. 179. Danly, D. E., in “Organic Electrochemistry” (M. Baizer and H. Lund, Eds.), p. 959. Dekker, New York, 1983. 180. Wendt, H . , and Bitterlich, S., Electrochim. Actu 37, 1951 (1992). 181. Wendt, H . , Bitterlich, S . , Lodowich, F., and Liu, Z.; ElecrrocAirn. Actu 37, 1959 (1992). 182. Rickert, H., in “Electrodes of Conductive Metallic Oxides,” (Trdsatti, S . , Ed.), p. 138 Part A. Elsevier, Arnsterdam/Oxford/New York, 1980. 183. Trasatti, S . , and Lodi, G., in “Electrodes of Conductive Metallic Oxides” Part A, (Trasatti, S . , Ed.), Part A, p. 301. Elsevier, AmsterdamlOxfordiNew York, 1980. 184. Beck, F., and Schulz, H., J . Appl. Elrctrochern. 17, 914 (1987). 185. Wiberg, K . B., “Oxidation in Organic Chemistry.” Academic Press, New York, 1965. 186. Murray, R. W., Ace. Chem. Res. 13, 135 (1980). 187. Anson, F. C., Saveant, J . M., and Shigehara, K., J . Electrounul. Chem. 145, 423 (1983). 188. Andrieux. C. P., Dumas Bouchait, J . M., and Saveant, J . M., J . Electround. Chem. 123, 271 (1981). 18Y. Burke, L. D., and Lee, B. H., J . E k t r o c h c m . Soc. 138, 2496 (1991). 190. Fleischmann, M., Korinek, K., and Pletscher, D., J . Chem. Soc., Perkin Trans. 2. 1396 (1972). 191. Clarke, I . S . , Ehigamusoe, R. S . , and Kuhn, A. T., J . Electrounul. Chem. 70, 333 (1976). 192. Fremery, M.. and Schwarzlose, G . , Chem. Ing. Technol. 46, 635 (1974). 193. Robertson, P. M . , and Schwager, F., and Ibl, N., J . Electrounul. Chem. 65, 883 (1975). 194. Robertson. P. M., Berg, P., Reimann, H., Schleich, K., and Seiler, P., J . Elecrrochem. Soc. 130, 591 (1983). 195. Beck, F., and Schulz, H., Ber. Bunsenges. Phys. Chem. 88, 155 (1984). 196. Beck, F., and Schulz, H., Electrochim. Actu 29, 1569 (1984). 197. Beck, F., and Schulz, H., J . Electrounul. Chem. 229, 339 (1987). 198. Schulz. H., and Beck, F., Angew. Chem. 97, 1047 (1983). 199. Beck, F., and Gerischer, H., 2. Elektrochem. 65, 504 (1961). 200. Beck, F., Chem. Ing. Techno/. 48, 1097 (1976). 201. Fukijama, K . , and Kita, H., J . Chem. Soc. Furuduy Trans. 75, 2638 (1979). 202. Junghans, K., Chern. Ber. 107, 3191 (1973). 171. 172. 173. 174. 175.

176

HARTMUT WENDT, SVEN RAUSCH, AND THOMAS BORUCINSKI

203. Junghans, K., Chem. Eer. 109, 395 (1976). 204. Casadei, M. Antonietta, Pletcher, D., Electmc~him.A ~ w .33, 177 (19x8). 205. Maqhvadi, B . , Chapuzet, J . M.. Lessard, L., E/ec~trochim.Acvu, 38, 1377 (1993). 206. Amoy, C., Lasia, A . Lessard, J . , J . Electrochem Soc. 41, 975 (1994). 207. Beck. F., and Gabriel, W., A n p w . Chem. Int. Ed. 24, 771 (1985). 208. Kiister, K., Riernenschneider, P., and Wendt, H . , lsrcwl J . Chrm. 18, 141 (1979). 2UO. Wendt, H., Angew. Chem. I n t . Ed. 21, 256 (1982). 210. Wendt, H. Electrochim. Acta 29, IS13 (19x4). 211. Wendt, H . , J . Mol. Catul. 38, 98 (19x6). 212. Kiister, K., and Wendt, H., in “Comprehensive Treatise of Electrochemistry” (O’M. Bockris, J., Conway, B. E., Yeager, E., and White, R . E., Eds.), Vol. 2, p. 251. Plenum Press, New York London. 213. Steckhan, E., Angiw. Chem. 98, 681 (1986). 214. Eberson, L., A d v . fhys. Org. Chem. 18, 79 (19x2). 215. Schifer, H., and Alazrak, A , , Angew. Chrm. 80, 4x5 (1968). 216. Wendt, H., and Plzak, V., Ber. Hunsenges. Phvs. Chem. 83, 481 (1979). 217. Wendt, H.. and Plzak, V., J . Electround. Chem. 154, 13 (1983). 218. Plzak, V., and Wcndt, H . , J . Elehiround. Chem. 154, 29 (1983). 21Y. Wendt, H., and Plzak. V.. J . E / ~ ~ c ~ t r o u Chem. n u / . 180, 85 (1984). 220. Schifer, H., and Steckhan, E., Angew. Chem. 81, 532 (1969). 221. Katz, M., Riemenschneider, P.. and Wendt, H., Ekwtrochim. Actcr 17, 1595 (1972). 222. Wendt, H., Chem. l n g . Techno/. 45, 1303 (1973). 223. Katz, M., Saygin, Oe., and Wendt, H . , Electrochim. Aciu 19, 193 (1974). 224. Plzak, V., Schneider, H . , and Wendt, H., Rer. Bunsengus. Phvs. C h m . 78, 1373 (1974). 225. Katz. M., and Wendt, H., Elecrroc~hitn.Aciu 21, 215 (1976). 226. Hoffelner, H., Yorgiyadi, S . , and Wendt, H . , J . E/ec.trounir/. Chem. 66, 13%( I975 j . 227. Wendt, H . , and Hoffelner, H . , Elrctrochirn. Actu 28, 1453 (1983). 228. Wendt, H.. and Hofielner, H . , Electrochim. Actu 28, 1465 (19x3). 22Y. Feokstistov, L. G., and Lund, H . . in “Saturated Carbonyl Compounds in Organic Electrochemistry” (Baizer, M . , and Lund, H . , Eds.), p. 315. Dekker. New York/Baxl, 1983. 230. Flazquez, M . , Rodriguez-Mellado, M . . and Ruiz. J. J . . W w t r o i ~ h i m Acrtr . 30, IS27 (1985). 231. Kiister, K . , and Wendt, H., J . E i e ~ t r ~ ( ~ iChcm. i u l . 138, 209 (1982). 232. Kiister, K., Wendt, H . . and Lebouc, A,, J . Eli~ctrounal.C h m . 157, 89 (1983). 233. Kiister, K . , Wendt, H . , and Lebouc, A . , J . Ele~~trocinul. Chem. 157, 1 13 ( 19x3). 234. Kariv, E . , Ferni, H. A . . and Gileadi, E . , Elwtrochim. A m 18, 433 ( l 9 7 3 j . 235. Kapilov, J . , Kariv, E., and Miller, L. L . , J . A m . Chem. Soe~.99, 3450 (1977). 236. Piccardi, G.. Nucci. L., Pergola. F., and Guidelli, R . , J . Electroanul. Chrm. 164, 145 (1984).

ADVANCES IN CATALYSIS, VOLUME 40

Fundamental Studies of Transition-Metal Sulfide Catalytic Materials R. R. CHIANELLI AND M. DAAGE Exxon Research and Engineering Company Annandale, New Jersey 08801

AND

M. J. LEDOUX Universiti Louis Pusteur- EHICS Laborutoire de Chimie des Muthriaux Cutalytiques Strasbourg, France

1.

introduction

In the 15 years since Massoth reviewed ( I ) the field of molybdenumbased hydrotreating catalysts, considerable progress in understanding the basis for their function and activity has been made. We now know that the activity and selectivity of these catalysts originates in the sulfided molybdenum phase. At the time of the Massoth article, it was not clear what role the alumina support played in the activity of these catalysts. Since the availability of probes such as EXAFS (extended X-ray fine structure), it has become clear that the alumina’s primary function is to disperse and stabilize the MoS2-based hydrotreating catalyst. It is also well established that the alumina strongly interacts with the MoS2 to do this. Less well known is that in anchoring and dispersing the M0S2, the alumina consumes some of the active sites making supported MoS2 catalysts less active per site than the corresponding unsupported catalyst (2). The alumina also serves the function of making the catalyst less expensive by diluting the metal. Because of the enormous commercial success of these catalysts, for over 50 years much of the scientific effort in understanding these catalysts was directed at the supported versions. This resulted in a

177

Copyright 0 1994 by Academic Press, Inc. All rights of reproduction in any form reserved.

178

R. R . CHIANELLI, M . DAAGE, AND M. J . LEDOUX

massive number of papers which numbered in the thousands by the time Weisser and Landa wrote their classic book “Sulfide Catalysts, Their Properties and Applications” ( 3 ) and many have followed since. But because the MoSz on these supported catalysts is highly disordered and because many of the catalysts were studied under conditions that were not related to the catalytic environment under which they operate, most of these studies shed little new light on the fundamental nature of the transition metal sulfide (TMS) catalysts. The TMS catalysts have been reviewed recently by Prins et ul. ( 4 ) , which brings the picture up to 1989. In this article Prins et al., while attesting to the wealth of new knowledge added to the field since the Massoth article, also outlines questions that are still open today. Some of these questions are: 1 . How does the catalyst function at the surface in the presence of a reacting molecule? 2. What is the location and the structure of the Co and Mo at the edges of the MoS2- and WSz-layered catalysts? 3. What is the function of Co and Ni promoters, and how do they operate? In this article we report on the recent progress in understanding these and other questions and comment on areas requiring more understanding and future research. We concentrate primarily, although not entirely, on the current understanding of the unsupported, pure TMS catalysts in their stable catalytic states in the belief that further understanding will emerge by studying these simplified systems. It is also interesting to note that as more activity is required of these catalysts, the metal content is increasing and the commercial catalysts are containing more unsupported MoS2. Supported studies that extend in a clear manner the knowledge gained from studies of unsupported catalysts are also included. The primary emphasis is on studies that have provided knowledge of the TMS beyond the point reviewed by Massoth and Prins. This report also points to the need for even simpler model systems for further progress to occur. It should also be recognized that the TMS catalysts have more or less been taken for granted by the petroleum refining industry. They have traditionally been called upon to meet environmental specifications for liquid transportation and heating fuels, and they have consistently been able to meet this demand. But in doing so they have added little in value to the fuel as determined by historical economic evaluation methods. Thus, they have not always received the attention that they deserve from the scientific community. However, in an age of increasingly stringent environmental regulations and concerns, the TMS should have an even more important future, especially as product costs begin to include environmental factors.

TRANSITION-MmAL SULFIDE CATALYTIC MATERIALS

179

II. Background TMS catalysts have played a crucial role for many years in the petroleum and chemical industry. Early in this century, it was discovered that sulfurcontaining compounds poisoned many metallic catalysts but that transition metals such as Mo and W, while converting to sulfides during use, retained their ability to hydrogenate aromatic compounds. Thus, one of the first uses of TMS catalysts was for the hydrogenation of coal liquids. Developed in Germany for this purpose, the TMS was later applied to petroleum and tar upgrading as well. Originally used as unsupported catalysts, the aluminasupported catalysts first appeared just before World War I1 and emerged shortly after as the modern petroleum processing catalyst used today in virtually every refinery in the world. TMS catalysts fall into a special category due to their exceptional resistance to poisons. In fact, the presence of sulfur compounds, the most common poison of metallic and oxide catalysts, does not decrease their catalytic activity, but is needed to maintain high activity. Sulfide catalysts are also very resistant to carbon deposition, which is illustrated by their use for converting residual oils. Arsenic, as well as nickel and vanadium contained in heavy petroleum fractions, are some of the few substances that cause significant deactivation, and this only occurs by physical blockage of pore structure in supported catalysts. The major reactions catalyzed by TMS are: hydrogenation of olefins, ketones, and aromatics; hydrodesulfurization (HDS); hydrodenitrogenation (HDN); hydrodemetallation (HDM); dealkylation; and ring opening of aromatics. But the TMS catalysts have many other uses as well, including: reforming, isomerization of paraffins, dehydrogenation of alcohols, FischerTropsch and alcohol synthesis, hydration of olefins, amination, mercaptan and thiophene synthesis, and direct coal liquefaction. More esoteric applications can also be found in polymer synthesis and in inorganic synthesis (synthesis of silanes and metal hydrides). A more comprehensive list of the TMS-catalyzed reactions is found in ( 3 ) . A. HISTORY OF THE PROMOTION EFFECT Because of the importance of the promotion effect and because many of the central questions surrounding TMS catalysis are about promotion, it is valuable to review a history of the effect. The first reference to a catalyst based on molybdenum and cobalt sulfides capable of desulfurizing coal oils in the presence of hydrogen was a patent from I. G. Farben Industrie dated May 24, 1928 (5). Before this, M. Pier and his team at BASF (1924-1925)

180

R. R. CHIANELLI, M.

DAAGE, AND M. J. LEDOUX

had taken advantage of their great experience with catalysts and reactions under high pressure drawn from the ammonia synthesis invented by F. Haber and demonstrated the high activity of molybdenum sulfide for the hydrogenation (liquefaction) of coal. The results were never published. It was in 1933 that Pease and Keithon (6) at the University of Princeton published the first results showing the activity of a mixture of cobalt and molybdenum oxides and sulfides for the hydrodesulfurization of a mixture of benzene and thiophene. The irreproducibility of their results already indicated the extreme complexity of the nature of this catalytic system, highly dependent on the conditions of preparation and pretreatment and the experimental conditions. This marked the beginning of a long controversy about the nature of the active phase of these catalysts. The brief entry of an academic team into this industrial world was not to be repeated until the beginning of the 1970s. Therefore, it is only through the scarce publications from the closed world of petroleum companys’ research laboratories that it is possible to follow the evolution of the cobalt-molybdenum system applied to HDS and to observe the birth of the concept of synergy. In 1943, A . C. Byrns et al. (7), of Union Oil of California published the first study showing under semi-industrial conditions the relative activities of Moos and COO and the mechanical mixture of these two oxides, which they compared to CoMo04 supported on bentonite. These authors demonstrated that a mixture of molybdenum and cobalt in their oxidic state should be chemically associated in order to be very active, while the simple mechanical mixture only showed the additive activities of the individual oxides. However, these authors mainly emphasized the behavior of these catalysts under different industrial conditions and reduced their discussion of catalyst structure and characterization to a few lines of speculation. In 1959, H. Beuther et al. (8) of Gulf Oil Company published the first systematic study of the HDS activity of CoMo and NiMo supported on a h mina as a function of the atomic ratio Co(Ni)/Mo. As a result, they showed what they called a “promoter effect” of the cobalt (or nickel) on the molybdenum for atomic ratios Co/Mo = 0.3 and Ni/Mo = 0.6. This publication was preceded by several patents proposing similar atomic ratios for cobalt by Union Oil of California (1948) (9) and Shell Oil Company (1954) (10) and for nickel by Union Oil of California (1954)(11). Figure I shows a typical activity curve of NiMo/A1203catalysts as a function of the value of the atomic ratio Ni/Mo ( 1 2 ) . B. EARLYMODELSFOR THE PROMOTION EFFECT An explosion of publications in the 1970s followed the scarcity of articles of the preceding decades. This phenomenon was directly linked to the first

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

18 1

100

50

0 0.0

0.2

0.4

0.6

0.8

1.0

FIG. 1. Typical activity curve of NiMo/A1203catalysts as a function of the nickel concentration ( 1 2 ) .

oil crisis and the realization by all the consumer countries of the necessity to upgrade dirtier crudes now that the lighter “sweet” crudes were becoming rarer and more expensive. Refinement of the heavy crude fractions, especially desulfurization, became imperative to make up the deficit in light fractions used in transportation and the petrochemical industry. These heavy fractions had previously been either burned in industrial boilers or asphalted. The first research group to propose a description of the structure of CoMo catalysts was led by Schuit and Gates (13). This group introduced the socalled “monolayer” model directly derived from the physical studies of CoMo oxide precursors supported on y-alumina carried out by J. T. Richardson (14) (Richardson first proposed the existence of a special Co/Mo entity.) In this model the upper or first layer contained only sulfur atoms, each bonded to a molybdenum atom of the second layer (below the first one), these molybdenum atoms being bonded to two oxygen atoms also located in this second layer. When a sulfur atom was removed by reduction (HZflow) of Mo5+to Mo3+,a vacancy was formed at the surface and became the preferential adsorption site of a sulfur atom in the organic gas phase. The presence of cobalt incorporated into underlying layers of the alumina

182

R. R. CHIANELLI, M. DAAGE, AND M. J . LEDOUX

support pushed up to the second layer aluminum ions (Al”) in tetrahedral sites, resulting in each atom of sulfur from the first layer now being bonded to two molybdenum atoms. A sulfur vacancy now liberated two molybdenum atoms instead of one for the adsorption site, which would explain the higher activity of this catalyst. This elegant model was, however, to be rapidly discarded by Schuit’s team itself, which observed that the sulfur stoichiometry of the molybdenum and cobalt compound corresponded to a mixture of MoS2 and Co& instead of the MoS, stoichiometry required by the monolayer model (15).Today this model is of historical interest only, but at the time it was quite influential. It also points out the dangers of focusing too heavily on a complex solid-state model. At the same time, Voorhoeve and Stuiver, (16) and Farragher and Cossee (17) from Shell proposed a new model derived from the structure of sulfides called intercalated solids. In these type of compounds, the Group VIII transition metals of the first row of the periodic table (Fe, Co, Ni) can penetrate the Van der Waals layers of the lamellar tungsten sulfide to occupy the octahedral symmetrical voids (Fig. 2). However, for MoS2 in the presence of Co or Ni, these intercalated compounds could only be prepared at temperatures greater than 800°C resulting in surface areas too low for catalytic measurements. Furthermore, it was difficult to prove that intercalates existed under catalytic conditions, and the authors of the model had to make the assumption that only the sites located on the edges of the MoS2 crystallites were occupied by the Co2+and Ni2+ ions under catalytic conditions. They called this model the “pseudointercalation” model to incorporate this idea. Voorhoeve et af.did not study HDS directly but rather the hydrogenation of benzene and cyclohexene in the presence of CS2 were studied. It was found that Ni strongly promoted the hydrogenation of benzene but only weakly the hydrogenation of cyclohexene. From other evidence, they concluded that the active sites were unsaturated W3+ with different degrees of

0

+

0

S

0

Mo

+

Octahedral site Tetrahedral site

FIG. 2. Pseudointercalation model of Voorhoeve and Stuiver (16) and Farragher and Cossee (17).

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

183

unsaturation for binding of benzene and cyclohexene. They called these edge sites (doubly unsaturated binding cyclohexene) and corner sites (triply unsaturated binding benzene). They also concluded on the basis of electron spin resonance (ESR) evidence that the W3+ sites were increased by the pseudo-intercalation of Ni through charge transfer. This idea means that the active site is the same in the unpromoted and promoted system with the number of sites increasing via indirect electron transfer to the layers. However, they did not explain why the reaction of cyclohexene was so weakly promoted. Also, their assignment of the ESR signal to W3+is questionable. Nevertheless, these papers represent an elegant introduction of interdisciplinary science into the field of TMS catalysis, and they form the basis for many studies that followed. At about the same time, Delmon and co-workers (18) at Louvain-laNeuve University proposed a simple model that explained their experimental observations. Since they often found that the unsupported catalysts they prepared had the phases MoS2 and cogs8 present, the model was based on the interaction of these phases. It was also based on the fact that at that time no unique Co/Mo/S phase had been identified in any catalyst nor were they able to find one in their studies. Most importantly, promotion was demonstrated in bulk catalysts in the absence of any support, thus excluding its role in the origin of the effect. It was also observed that the synergetic effect could be produced in a mechanical mixture of MoSz and c09s8. This work provided a key step in reducing the number of components in a very complex system from three to two. The Louvain group proposed that the MoS2 and Cogss acted together by being in close contact, from which came the term “contact synergy” (Fig. 3). The HDS reaction took place at the interface between the two sulfides with each phase “helping”. The authors also proposed a reaction mechanism, which they called “remote control mechanism,” where the

FIG.3 . Contact synergy model of Delmon and co-workers (18).

184

R . R . CHIANELLI, M . DAAGE, A N D M. J . LEDOUX

sulfided organic molecule was adsorbed on MoS2 to be desulfurized by dissociated migrating hydrogen atoms coming from COVSX active sites (Fig. 4). The remote control mechanism can be dismissed on the grounds that both CosS8 and MoS2 are active as HDS catalysts as described below. Although the contact synergy model was heavily criticized as being overly simplistic, it does explain macro-features of the promoted systems. It also clearly points to the importance of the unsupported catalyst as containing the essential ingredients for promotion. It is our view that both the Voorhoeve and Delmon models have features that are valuable, and depending on the location of the particular catalyst of interest in the Co/Mo/S phase diagram, one or the other may be a more useful description as discussed below. For example, a pseudo-intercalated MoS2 phase may occur at the interface between MoS2 and CoVSx. C 4 + H2S

cogsg

FIG.4. Remote control model of Delmon and co-workers (18).

The concept of a mixed phase containing both cobalt and molybdenum responsible for the synergy was considered seriously, but it was only the first publication of Topsere and his group that presented the first physical proof of a specific Co environment. The introduction of very small amounts of Co into bulk MoS2 allowed this group to observe a specific emission Mossbauer spectroscopy (EMS) signal from a new Co site different from the usual band characteristic of Co& (19). A systematic study of CoMo catalysts with this new technique showed the existence of three different Co compounds in the catalyst (Fig. 5): cobalt contained in alumina support as aluminate (type l), cobalt contained in Cogs8(type 2), and a third cobalt species (type 3 ) associated with the cobalt found in small amounts in the MoS2. Topsere et al. (19) located the cobalt inside or on the edges of MoS2 crystallites and called the type 3 Co the CoMoS phase. In addition, using EXAFS analysis, this group showed importantly that the local atomic structure of MoS2 was preserved in the sulfided catalyst (20). They also applied many other physical characterization techniques to the analysis of this system to confirm the existence of the Co/Mo/S phase: 0 2 and NO chemisorption, XPS, and transmission electron microscopy (TEM), with marginal results because of the difficulty caused by the intrinsic disorder present in the MoS2, as discussed below. This disorder makes

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

1

Alumina

0

185

oo

Mossbauer

..

FIG. 5 . CoMoS phase of Topsoe et al. (19).

unambiguous interpretation of adsorption, EXAFS, and other physical techniques very difficult. Only TEM on MoS2 crystallites, containing a small amount of Co, showed convincingly that the Co atoms were located on the edges of the MoS2 platelets ( 2 1 ) , leading to a model very close to the pseudo-intercalation model with similar limitations. Topsoe's group was able to evaluate the amount of the new CoMoS phase contained in different catalysts with increasing loading of Co by using EMS, identifying the type 3 Co that correlates linearly to the HDS activity of these catalysts (22). However, most of the measurements were made on fresh catalysts and the correlation needs to be better established on catalysts that are stabilized in a catalytic environment to show conclusively that the CoMoS phase measured by Mossbauer is directly related to the promotion site and not a precursor that leads to the active site. This is further complicated by the practical difficulties encountered in applying EMS. For example, the sample must be prepared with 57C0metal to observe the effect. If these results did not give any information regarding the local structure of the active phase, they did provide strong evidence in favor of the existence of a specific active phase involving Co and Mo and were the first to do so. A group of French researchers from the University of Lille and the French Petroleum Institut (IFP) rationalized the location and the number of Mo, Co, and S atoms by modeling (23). This allowed them to find a very

186

R. R. CHIANELLI, M . DAAGE, A N D M . J . LEDOUX

good correlation between Mo atoms at the edge of the particles and the HDS activity for catalysts containing only MoS2; but no new information on the role of Co could be extracted from this study. In all the early models described above, only geometric and crystallographic aspects of the catalyst were tentatively described; very little physical evidence regarding the electronic or chemical origin of the promotion effect was presented by these groups to explain why the pseudo-intercalation phase, the CoMoS, or the frontier between Co9Ss/MoSz,was more active than the MoS2 or Co9Ss taken separately. In an XPS study of these CoMo catalysts only Topsoe’s group found a shift that they attributed to an electron transfer from Co to Mo synergy, but no specific electronic mechanism was proposed ( 2 4 ) . Only in the case of Voorhoeve’s papers was the suggestion made that pseudo-intercalation of Ni in WS2 resulted in charge transfer to the W increasing the number of W3+ atoms and a presumption made that increasing the number of W3+atoms increased activity. This group also made the argument that the pseudo-intercalation sites should now be isostructural with those that are in ReS2 and that a compound with these properties was known (CooSMoS2), which might serve as a promotion model. However, Topsoe’s papers state that this phase has the wrong Mossbauer signature and is inactive in catalytic measurements. In this paper we argue that the catalytic measurements described were made on catalysts that were too low in surface area and that the entire series Co,MoS2 (0 < x < 0.5) needs to be investigated to fully describe the relevant promoter phase. C.

IMMISCIBLEPHASESAND SYMMETRICAL SYNERGY

One very interesting but relatively unknown paper is that by Phillips and Fote (25). The authors take the view that the promoted systems may be viewed as mixtures of two immiscible phases (c09ss and MoS2) that interact by forming surface phases or surface complexes. We can use this concept to describe a picture that contains all of the above models and describes all possible mixtures of Co/Mo (Ni/Mo) in catalyst systems. This is illustrated in Fig. 6 by a hypothetical phase diagram of Co/Mo/S catalytic materials. First, it should be noted that Fig. 6 is the phase diagram of a hypothetical Co/Mo/S existing below about 550°C. Below this temperature no ternary phases have been prepared in the Co/Mo/S system. All current catalyst preparation techniques that produce stable sulfide catalysts with sufficient surface area for catalytic activity produce materials in the phase region indicated. The left-hand portion of the phase diagram is the region most commonly accessed by conventional catalyst preparations. In this region as Co is added

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS Co on MoSz Interface Region

187

Mo on Cogs8 Interface Region

\ MoS2

+ Cogs8

Coexistence HD S Activity

MoS2 + 50:50 Co + M o Ratio

FIG. 6. Phase diagram for hypothetical Co/Mo/S catalytic material illustrating symmetrical synergy.

to the system, the promotion effect increases until a maximum is reached. No C O ~ SisSdetected because the Co/Mo/S surface phase is being produced. The boundary of this region is determined by the MoS2 “edge” area available as is discussed below. Thus, higher-edge area catalysts will require more Co to reach the activity maximum, which will be higher. Such is the case for supported catalysts. Lower-edge area catalysts will require a lower amount of Co to reach a lower maximum. At some point as more Co is will begin to phase separate, and we enter a region where both added, COSSS phases coexist. In the idealized model presented, this occurs at the promotion maximum, but in real catalysts the situation may be more complex with an equilibrium existing between the surface Co/Mo/S phase and the c09s8. On the right-hand portion of the phase diagram, a Mo/Co/S surface phase exists that is exactly analogous to the CoiMolS phase with Mo promoting Cod%. That c O 9 s s can be promoted just like MoS2 has been overlooked in the literature. This can be seen from the work of the Delmon group cited above and from our own unpublished results. The activity of the promotion peak in the right-hand region is usually lower because it is difficult to prepare Co9Ss of high surface area. The activity of the resulting catalyst that occurs in the region in the middle of the phase diagram will be the resultant of the mixture of the two surface-enriched phases. This can be quite complicated because each phase affects nucleation and surface area of the other phase during preparation. We term this behavior “symmetrical synergy,” and similar arguments apply to the other promoted systems. This behavior is

188

R . R . CHIANELLI, M. DAAGE, AND M. J . LEDOUX

MO/CO/S

Surface

\

Co/Mo/S

/

Surface Phase

FIG. 7. Schematic microstructure of hypothetical unsupported Co/Mo/S catalysts.

illustrated in Fig. 7. The proposed Co/Mo/S surface-phase structures are described in more detail elsewhere in this article. But the Mo/Co/S surfacephase structure has never before been described. The existence of this “symmetry” in the promotion behavior must be taken into consideration in any model and strongly suggests an electronic origin of promotion since any structural model proposed cannot be too specific to one of the participating bulk structures. Most of the popular early models of promotion fail to take this into consideration, but may be valid for the Mo-rich end of the promotion system.

HI.

Preparation of Bulk Transition-Metal Sulfides

To further the understanding of the TMS, it is necessary to synthesize catalysts supported and unsupported, using methods that offer a good degree of control over their properties and stoichiometry. In the following section, we

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

189

review some of the preparations that formed the basis for the studies that follow in other sections. TMS catalysts are generally prepared by direct sulfiding of a metal salt (oxide, halide, etc.) or by decomposition of a sulfur-containing precursor. Mixed metal sulfides, for example, promoted catalysts, can also be obtained by direct synthesis or by subsequent impregnation of the promoter metal onto the pure sulfide. The resulting precursor is generally dried and then converted to a sulfide catalyst by thermal treatment, according to specific conditions. This last step is often the most critical and is specific for preparing catalysts for targeted applications. A.

BINARYSULFIDES

Even though most binary sulfides can be obtained by direct sulfidation of oxides and chlorides in the presence of hydrogen sulfide or a mixture of 1015% hydrogen sulfide in hydrogen, low-temperature precipitation is often preferred. A general synthetic method has been reported by Chianelli for the preparation of Group IVB, VB, and VIB binary sulfides (26). Lowtemperature metathetical reactions are convenient and allow good control over parameters such as particle size and composition. The general equation is MX,

+ n / 2 A2S --+

MS,,Z

+ n AX,

where M is the transition-metal ion, X the salt anion (Cl-, carboxylate, . . . ), and A an alkali-like cation (Li', Na', NH;, . . . ). The materials obtained by this route are generally amorphous and have high surface area. Poorly crystalline to crystalline solids are then obtained by thermal treatment under flowing inert gas, hydrogen, or H2SIH2. Another common synthesis of group VIB sulfides consists in the preparaand (NH4)2WS4.These salts are easily tion of thiosalts such as (NH4)2M~S4 obtained by bubbling hydrogen sulfide in a solution of ammonium molybdate or tungstate in ammonium hydroxide at room temperature. The thiosalts are then converted to the corresponding sulfides by thermal decomposition as described above. Similarly, tetraalkylammonium thiosalts have also been synthesized for Mo, W, and Re and converted to sulfides with high surface areas as described below (27). Group VIII sulfides are obtained by the low-temperature precipitation described above. However, a preferred synthesis involves the direct sulfidation with hydrogen sulfide of ammonium hexachlorometallates. A good example is found in the synthesis of RuS2 and RhzSs (28). However, synthetic techniques and the solid-state chemistry of the Group VIII sulfides are not as well developed as those for the Group VII sulfides, and much research needs

190

R . R. CHIANELLI, M. DAAGE, A N D M . J. LEDOUX

to be done to provide better catalytic materials that have higher surface areas and are supportable.

B. MIXED-METAL SULFIDES Mixed-metal sulfides are generally prepared by one of the four following methods: (1) comaceration, (2) homogeneous sulfide precipitation, (3) mixed-metal salts or organometallic clusters decomposition, and (4) impregnation of a binary sulfide. Most of these techniques were developed for the synthesis of Co/Mo or NiiMo catalysts but are applicable to other systems. The comaceration method consists of reacting freshly prepared oxides with a solution of ammonium sulfide at moderate temperature (-70°C). The slurry is continuously stirred until evaporation to dryness. The catalysts are obtained by thermal treatment under vacuum (18). The homogeneous sulfide precipitation (HSP) corresponds to a low-temperature coprecipitation. In that particular case, two or more salts are dissolved prior to the addition of the sulfiding agent. The synthesis of unsupported Co-Mo sulfide provides a good example of the HSP method: A solution of cobalt nitrate and ammonium heptamolybdate is poured into a hot solution of ammonium sulfide. The slurry is then evaporated to dryness (29). More recently, thermal decomposition of molecular metal/sulfur complexes was studied, either starting from salts where both the anion and the cation contains one of the metals or starting from molecular complexes in which the two metals are part of the same molecular sulfur-ligated cluster at the outset. Co(en)sMoS4 (30) and Co/Mo cube (31) are examples of such precursors. These precursors can then be decomposed under different conditions such as inert atmosphere, hydrogen, H2S/H2, or in situ. A given promoter to Mo ratio may be obtained by preparing the mixed double salts, for example, Nil-,Zn,, where x can be varied from zero to one in the above formula. These preparations are particularly useful for controlling and optimizing the properties of the TMS as described below. The second metal, for example, the promoter, may also be added by subsequent impregnation of binary sulfide. When a nonreactive promoter precursor, for example, metal nitrate, is used it is necessary to resulfide the impregnated sulfide in order to decompose the precursor. Another variation of this method consists in using reactive promoter precursors that will react with the surface of the binary sulfide. In this case, further treatment of the catalyst may not be required. Good precursors include metal carbonyls and metal alkyls (32, 33). The precursor decomposition approach been most widely applied to the MoS2-based systems. However, it has also been extended to the mixed noble-metal sulfides by Breysse and co-workers (34) at Lyon following the work of Passaretti et a1 (35).

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

191

Finally, it should be mentioned that highly crystalline material and/or doped single crystals can be obtained from the elements by vapor transport techniques (36) or by impregnation of the highly crystalline binary sulfide (37). These materials serve as important models for higher-surface-area materials and often yield characterization information that is clear compared to more complicated catalysts.

c.

PRECURSOR CONTROL OF ACTIVITY AND SELECTIVITY

The use of precursor synthesis techniques as described above is driven by the fact that decomposition of the precursor material into the catalyst often results in catalysts that have activity or selectivity superior to that of preferred products. The amine thiomolybdate decomposition described above results in MoS2 catalysts with surface areas exceeding several hundred square meters per gram (27). The HDS activity increases correspondingly and unpromoted catalysts have activities approaching those of promoted systems. Similarly, precursors that combine promoter and Mo can upon decomposition yield very high surface areas with very high activities. A series of promoter-containing precursors was recently studied by Chianelli et al. (38). This series contained three different Co/Mo precursor compounds, and the catalysts prepared from them were compared with a standard catalyst prepared in a conventional manner. The best catalyst was almost three times as active as a standard commercial Co/Mo catalyst. The activity per Mo of this catalyst was almost equal to the activity of a supported promoted Mo-based catalyst. This very high activity is a result of the production of a very high-surface-area catalyst through the precursor route. An unsupported MoSz of this type represents the maximum volumetric activity possible because no amount of volume is "wasted" with inactive support material. The Co in the three precursors was all chemically bound in different ways, but not enough information is available to say what the relation is between precursor and catalyst structure. The precursor route to active catalyst is hindered, however, by the cost of preparing the catalyst. Organic components, synthesis, can be costly, as can disposal of waste products produced during decomposition. Also, to achieve the very high activities described above requires the use of three to four times the amount of active metals as contained in the typical commercial catalyst. Nevertheless, catalysts prepared by the precursor methods provide interesting insight into important aspects of catalyst activity and selectivity. Furthermore, future processes that take advantage of the high volumetric efficiencies of these catalytic materials may be designed in the future.

R . K. CHIANELLI, M. DAAGE, A N D M . J . LEDOUX

192

IV. Structure and Properties of TMS Catalytic Materials

The catalytic activity of TMS catalysts is related to defects in their crystal lattice that occur at the surface. However, the properties and stability of these defects are determined by the bulk atomic and eiectronic structure. I t is therefore particularly important to know which phases are stable in a catalytic environment and what are their crystal structures. This knowledge was incomplete when the Massoth article was written ( I ) . The subsequent development of this knowledge formed the basis on which new understanding of the fundamental origins of the catalytic effects in the TMS was developed. A.

THESTABLE TMS CATALYTIC PHASES A N D THEIR STRUCTURES

Sulfide catalysts operate at high temperature (typically 200-400°C) and under reducing conditions, and significant restructuring can occur under catalytic conditions. In situ characterization of TMS materials is generally difficult and limited because of the conditions under which they are used (high-temperature, high-pressure, and liquid-phase reactions). Therefore, i t is necessary to characterize the catalyst that has stabilized under catalytic conditions as well as the fresh catalyst. The phases that are stable under catalytic conditions are the “stable states,” and it is from these that information regarding the relation between the electronic/geometric structure and the catalytic activity/selectivity can be obtained. Unfortunately, too many papers on the TMS ignore this point. Information obtained on fresh catalysts can be highly misleading but is useful in determining how properties of the stable states arise (surface area, degree of crystallinity, etc.). After HDS at 400°C and 35P kPa Hz, the stable phases of the binary TMS are: TiS2, VS,, Cr&, MnS, F e S l - , , Co&, Ni2S3, Fibs2, MoSz, RuS, Tc&, Rh2S3, PdS,, TaS2, WS2, Re&, OsS,, IrS,, PtS (Fig. 8) (39). The identification and composition of these phases is obtained by using conventional techniques such as X-ray diffraction and elemental analysis. However, good catalysts are often poorly crystalline, which results in broad Xray diffraction peaks making absolute identification difficult (FeS, , for example). The interpretation of these diffraction patterns can be somewhat complex, and particular care should be taken as described below. Additionally, in some cases the phase may vary with catalytic conditions. For example PdS is usually found, but under some cnnditions Pd4S is found. This type of change results from differing sulfur partial pressures in different reactor studies. Thus, a definition of the stable TMS states should always contain some statement regarding sulfur activity in the reactor.

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

Layered TMS Rci.

8.

Amorphous TMS

193

isotropic TMS

The stable states of TMS catalytic materials in the HDS environment ( 3 9 ) .

For the binary phases, there are two main classes of TMS structures: Isorropic sulfides dominate the Group VIII sulfides. Many different structures exist in this group, such as pyrite (RuS2), pentlandites (Ni3SZ),and others. The coordination of the metals in this group by sulfur is generally , sixfold, octahedrally arranged, but some special structures, such as COSSX can also contain mixed sixfold and fourfold tetrahedral coordination. Some structures, such as PdS and PtS, also contain metals in distorted sixfold coordination. Layered sufjdes predominate in Groups IV to VII except for MnS, which is isotropic. The main structural feature of this class is a strong chemical anisotropy that greatly affects their catalytic properties. In the molybdenite structure (MoS2), the metal is in trigonal prismatic coordination. Other layered sulfides are octahedrally coordinated or distorted octahedrally coordinated (Re&) with the metal surrounded by six sulfur atoms. It should also be noted that some TMS sulfides have structures that fall in between isotropic and layered sulfides. Rh2S3 is an example and others are completely amorphous, such as IrS, and OsS,. When a second or third metal is present, the identification of the phases becomes more complex. In general, poorly crystalline mixed-metal materials, such as Co/Mo/S, are not stable as intercalated bulk phases under catalytic conditions and are phase separated (Co& + MoS2). The X-ray diffraction pattern of the bulk material is characteristic of a mixture of the two corresponding binary phases with no evidence of a mixed phase (see Fig. 7). For materials containing at least one element crystallizing in a iayered phase, a few well-defined ternary compounds do exist and correspond to intercalates such as CoMozSd, which is isostructural with Re& (40).

I94

R . R.

CHIANELLI, M .

DAAGE, AND M . I. LEDOUX

However, because they require high-temperature synthesis, they cannot be prepared in a dispersed form and their catalytic activity is low. When both components belong to the isotropic class, solid solutions, crystallizing in a cubic pyrite structure, may be obtained as observed for C O , - ~ R U . ~Col-,Rh,S2, S~, and Rhl-,Ru,S2 (35).However, under HDS conditions, these ternary compounds also separate into their binary component phases (for example, CosSx and RuS2). Recently, Chevrel phases have also been studied for HDS (41). Again, their high-temperature synthesis is the major road block in preparing active catalysts for useful application. However, it has recently been reported that Chevrel phases have been prepared from inorganic precursors at low temperature, and it remains to be seen if this provides a route to active non-phase-separated catalysts (42). Nevertheless, the Chevrel phases are the only known ternary phases that appear to exhibit high activity and remain stable without undergoing phase separation. However, characterization of these materials is not well enough advanced to unambiguously demonstrate this point.

B.

MORPHOLOGY, PARTICLE SIZE, AND SURFACE AREA

The morphology, particle size, and surface area of TMS are generally obtained by conventional methods such as transmission electron microscopy, X-ray diffraction line broadening, and nitrogen, BET (Brunauer-EmmettTeller) measurements. The isotropic class of TMS compounds presents no major difficulties in this respect and is well behaved (43). However, major difficulties are encountered when characterizing TMS materials of the anisotropic class. The basal plane surface corresponds to a close-packed surface of sulfur atoms bounded to three metal atoms. This particularly stable environment of the sulfur results in weak Van der Waals interaction between the layers. Because of this, the basal plane exhibits a low reactivity and leads to the ability of MoS2 to act as a very effective lubricant. In contrast, the edge plane is highly reactive, and accurate determination of the amount present relative to the basal plane area is of primary importance for developing catalytic structure/function relationships in anisotropic systems. To further complicate the situation, MoS2 produced under conditions leading to good catalytic properties is in a poorly crystalline and highly disordered state best described as the “rag structure.” A quick look at an electron micrograph rapidly convinces the researcher of the difficulty in estimating the particle size of such a material (Fig. 9) (44). X-ray diffraction may be used, but the line broadening will depend upon the bendingfolding of the layer and will not be representative of the particle size as it is in the isotropic materials. When applying the Debye-Scherrer line-broadening equation to the 002 diffraction peak, a reasonable value of the coherence

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

195

Pic;. 9. MoSr in the "rag" structure (44). (a) Large rag approximately 2000 A across and several layers thick. (b) Small rags approximately 6 A in width consisting of single layers.

196

R . R . CHIANELLI, M . DAAGE, A N D M . J . LEDOUX

length of the crystallites is obtained for the stacking of the layers along the c axis, agreeing within a factor of 2 with that estimated from TEM. However, the same analysis applied to the 100 or I10 peaks results in values of the coherence length along the c( axis of between 7 0 and 100 A, whereas the real dimension can be as large as several micrometers. This illustrates a key problem with materials like MoS2. Measuring the dispersion of these materials remains a difficulty; therefore, many catalytic correlations appearing in the literature do not yield reliable information regarding these systems because of this problem. EXAFS measurements have also been used, particularly in supported materials, and this technique was the first to show unambiguously that MoS2 was the stable state on supported catalysts. However, when used to estimate the average coordination numbers of the metal atoms in highly anisotropic materials, and thus the particle size of the layered catalyst disorder, again renders the results ambiguous, and significant discrepancies arise. The area under the Mo-Mo peak in the transform of the EXAFS is used to estimate the particle size under the assumption that the coordination number will lower as the particle size decreases, because Mo atoms on the edge of the crystallite will have lower coordination numbers, therefore reducing the overall average coordination number of the sample. The assumption, while valid, is not the only way that the average coordination number can be reduced. Disorder will also reduce the coordination number as measured by EXAFS. For example, the attenuation of the Mo-Mo peak in a sample of MoS2 prepared at 900°C using a catalytic preparative technique is 50% of that observed for highly crystalline materials prepared at the same temperature. Therefore, the corresponding coordination number suggests that the particle sizes are much smaller than they in fact are as estimated from TEM studies. The misuse of EXAFS data in determining particle size in sulfide catalysts is widespread and, thus, the resulting dispersion estimates have lead to considerable confusion. BET measurements of the anisotropic TMS present a similar problem. Selected preparations can yield surface areas up to 400 m2/g for unsupported catalysts. At this level the MoS2 has surface area equal to the surface area of supported MoSr , which is conceptually “removed from the support.” After catalytic reaction, most poorly crystalline TMS sulfides have surface areas in the 10 to 100 m’/g range. But most of the surface area of these catalysts is associated with the basal plane of the layer, which does not contribute to catalytic activity. The surface area contribution of the basal plane may be illustrated by calculating the surface area of a catalyst consisting of infinite layers of MoS2. Such a catalyst would have a surface area of 327 m2/g, but would be completely inactive for HDS because no active edge planes were present. The basal plane surface area for any MoS2 catalyst is inversely proportional to the stacking of the layers and is equal to 327/n, where n is the

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

197

number of stacked layers. Using this equation, an estimate of the basal plane contribution to surface area for a given catalyst can always be made. As of this writing, there is no simple way to measure precisely the edge plane area. The problem of edge area/basal area has been addressed through the use of O2 chemisorption and is discussed in detail later in the section on the surface characterization (45). One exception to this is found when directly estimating from scanning electron microscopy (SEM) the edge surface area of microsingle crystals that are regular in shape and contain little disorder (46). Using this method, the intrinsic activity of the edge plane has been measured and used to standardize a model reaction, which can then directly be used to measure edge area. Similar difficulties exist for mixed-metal systems when anisotropic structures are present and are further complicated by the presence of the second phase. The second metal, generally a first-row transition metal such as Co, affects the morphology by increasing the stacking of the MoS2 layers giving another indication of the interaction of the Co with the layers. Pratt et a / . reported the results of extensive studies on unsupported Ni/Mo/S and Ni/Mo/O catalysts (47, 48). In these studies they pointed to two modes of interaction of MoS2 with a second phase. The first mode of interaction involves the interaction of the basal plane with the second phase. In this mode the MoS2 crystallites lie flat on the second phase and, in some cases, completely wraps the second phase in several continuous layers (Fig. 10). Second phases that show interaction in this mode include NiO, Moo3, and T i 0 2 . In the second mode of interaction, the MoS2 crystallites are anchored through their edge planes. Phases that interact in this mode include A l z 0 3 and pyrrhotite (Fig. 1 I ) (49). It has also been shown that the orientation of the MoSz can change from one mode of interaction to another mode by heat-

~

Interface Region

FIG. 10. Modes of MoS2 interaction with second phases

198

R . R . CHIANELLI, M . DAAGE, A N D M . J . LEDOUX

FIG. I I . Single layers of MoSz on pyrrhotite crystallite ( 4 9 ) . Pyrrhotite lattice lines are 5 . 8 . k apart.

ing. As the temperature was raised on the MoS2 supported on a second phase (AhO?), MoS2 crystallites that were laying flat stand up, reflecting the stronger mode of binding through the edge planes (50). It can also be observed from the schematic of the interactions shown in Fig. 10 that interaction zones that provide the possibility for sharing sulfur atoms by the two different phases exist. The structure of these interaction zones are crucial to understanding the interaction of sulfides with supports and promoter phases. The basal plane interaction with a second phase can be expected to be weak, but charge transfer might be expected if metal atoms are exposed on the second phase. Such an interaction would be similar to an intercalation interaction. The edge interaction will be much stronger, and here a transition zone with a possible epitaxial relationship between MoS2 and the second phase is expected. It is in this zone that we can expect to find the surface phases as described above (for example, the CoMoS phase). But in the cases described here, the surface phase becomes a line phase at the boundary between the two bulk phases. It is our belief that the detailed study of these phases represents a key area for future research in TMS catalysis.

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

199

C. METALOXIDATION STATEAND STRUCTURAL ENVIRONMENT Massoth, when discussing the oxidation state of the TMS catalysts, concentrated on the typical commercial supported catalyst (1). Because of this, the article reflected a very confused picture with heavy emphasis on the supported and reduced state of the oxide catalyst. The emphasis was placed here because it was still believed at the time that the support was fundamentally crucial to the activity of the catalyst. Today we know that the role of the support is to disperse the catalyst and that the sulfided state of the catalyst is responsible for the stable activity. Massoth reported that at that time the state of the sulfided catalyst was very unclear. By the time Prins et a1.(4) wrote their article, it was clear that the stable operating states of Mo/Co and related systems were as the sulfides. It is therefore essential to understand the oxidation state of the bulk sulfide and how this affects the oxidation state of the surface defects. The oxidation state of the metals in the bulk sulfides corresponds to what is expected for a given metal in a sulfur environment. Generally the lattice sulfur is tightly bound to several metal atoms (usually three) and is very stable. Consequently, narrow valence distributions are observed. For the early layered TMS, the oxidation state is 4 + . In the later isotropic TMS, the valence is usually 2+ or 4+. For example, in RuSr the oxidation state is 2+ because of the pyrite structure. In Rh2S3the oxidation state is 3 + . In other TMS inthe stable catalytic state such as amorphous OsS, it is not clear what the oxidation state is although formally it would be 2 + . When other metal oxidation states are detected, their concentration is low and they are generally associated with the presence of defects and/or sulfur vacancies at the surface. For example, an ESR signal, attributed to defects such as Mo(II1) and/or Mo(V), can easily be observed in poorly crystalline MoS2. However, magnetic susceptibility measurements provide evidence that spin coupling due to cation pairing exists, suggesting that only a fraction of the defects are detected by ESR (51). Both magnetic susceptibility and ESR correlate to the HDS of dibenzothiophene (DBT) indicating that surface defects are being measured (41). These surface defects are on the order 1 X 10” spins for a typical catalyst, but absolute determination of the number and oxidation state of these defects has been hard to unambiguously determine. In contrast with some oxide catalysts where the lattice oxygen has been identified as a critical component of the catalytic behavior, the bulk atoms of the sulfides do not participate directly in the reaction. Therefore, the bulk structure and oxidation state should be considered as a structural and “electronic” support for the surface structure that is catalytically active. It should be noted that careful application of resonance techniques to the problem of active site oxidation state will be an extremely useful area of research in the future.

200

R. R . CHlANELLl,

M. DAAGE, A N D M .

J. LEDOUX

EXAFS and particularly XANES (X-ray Absorption Near Edge Structure) are used for determining the coordination number of the metals and their local environment. The average local structure is obtained by comparing the spectral features of the catalyst with the spectra of well-defined model compounds. Octahedral, trigonal prismatic, tetrahedral, and square pyramidal environments have been reported for the metal atoms ( 4 ) . However, there are some limitations associated with the EXAFS technique as described above. In amorphous or poorly crystalline materials, the environment of the metal is generally disordered, and the coordination numbers could be lower than expected. This effect is larger for highly anisotropic sulfides because of the folding of the layers. As mentioned above for MoS2, the coordination number is then unreliable for estimating particle sizes. XANES is not affected by disorder and gives different information than EXAFS. Information obtained from XANES reflects the oxidation state of the metal and in some case can be quite definitive.

V.

Surface Composition

Catalytic events such as adsorption, breaking, and forming of bonds are obviously associated with the surface of the solid. Any information regarding the composition of the surface is therefore essential in providing a good understanding of the catalyst. Multiple approaches are generally used, especially chemisorption and spectroscopic methods.

A.

CHEMISORPTION AND MOLECULAR PROBES

Chemisorption has been applied to numerous catalytic systems including the TMS catalysts, and valuable information on the active surface area or catalytic sites densities has been obtained. However, it is important to keep in mind that the interpretation of the results is subject to the assumption that the stoichiometry of the chemisorption is known. It is well known that on metal surfaces, dispersion is calculated using one hydrogen atom per metal atom. Consequently, dispersion higher than 100% is not unusual for highly dispersed catalysts with particle sizes below 15 A, reflecting the existence of metal atoms associated with more than one hydrogen. This assumption is also made for TMS catalysts. A good example for sulfides is provided in the chemisorption of oxygen on the edge surface of MoSz . If site densities are calculated using one oxygen atom per molybdenum on the edge, the densities are inexact, particu-

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

20 1

larly when the chemisorption is measured at room temperature, because 15 oxygen atoms per edge molybdenum are consumed. This clearly shows that oxygen chemisorption is corrosive because the oxygen attacks more than just the edge atoms as discussed below. However, it is well known that a linear relationship between the site density and the oxygen chemisorption does exist. This is due to the topotactic nature of the oxidation of MoS2 in which sulfur atoms are replaced with oxygen atoms deep into the edges while retaining the approximate morphology of the crystallites (34). Thus, oxygen chemisorption is proportional to the edge area but does not directly measure the actual number of sites at the edge. A similar relationship can be observed with promoted MoS2. Each family of catalysts has its own linear correlation, which cannot be compared to each other directly because of the corrosivity problem. More recently, lowtemperature oxygen chemisorption has been claimed to be more reliable, but it also lacks a well-determined stoichiometry (52). Oxygen chemisorption has also been applied to tungsten and rhenium sulfides, as well as promoted molybdenum and tungsten sulfides. In the isotropic class, it has been applied only to ruthenium sulfide, in which case it gives approximately the same result as a BET measurement due to the isotropic nature of this sulfide (41). Two other molecules can be used for the characterization of the active surface of TMS catalysts (53, 54). CO has been rarely used, whereas NO has been extensively applied to the CoMoS system. Both molecules present an additional advantage in the sense that they are easily detected by infrared spectroscopy, providing some information on which metal atom they bind. However, it has been shown that under UHV conditions NO can be decomposed at 130 K , suggesting that the interpretation of the data might be complex and that partial oxidation of the surface is possible (55). Even though chemisorption techniques can lead to confusing interpretations regarding the site density, it remains the most convenient way to screen a family of commercial catalysts, limiting therefore the cost and time needed for the reactor tests. It should also be noted that Topsge and Topsde used the adsorption of NO combined with infrared spectroscopy (IR) to conclude that the Co was at the edge of MoS2 and not in an intercalated position (54). They based this on the fact that the NO adsorbed on Co has a unique IR spectrum. As they added Co to the system, they saw an increase in the IR absorption peak due to NO adsorption on Co, which occurs in the region from 1850 to 1785 cm-I. They concluded that this could only be the case if the Co was on the edge of the MoS2 and not intercalated. However, our own studies with intercalated TMS show that MoS2 has sufficient transparency in the IR region in which NO adsorbs and that NO adsorbed on Co would be detected several hundred layers into the bulk. Thus, this study could not distinguish between Co on the edge of MoS2 and intercalated Co with this method.

202

R . R . CHIANELLI, M . DAAGE, A N D M . J . LEDOUX

B. HYDROGEN CHEMISORPTION AND PSEUDOINTERCALATION One form of chemisorption has been particularly useful in generating new insight into the nature of catalytic MoS2 but has been largely ignored. This is chemisorption of hydrogen in combination with neutron scattering studies as performed by Wright et al. (56).This paper reports the high-temperature (300°C) adsorption of hydrogen on MoS2 and the study of this adsorbed hydrogen by neutron scattering techniques. First the authors show that the surface area for hydrogen adsorption is from 3.5 to 7.5 times greater than its nitrogen BET surface area. This striking result indicates that hydrogen penetrates deeply into bulk MoS2 under catalytic conditions. The only possible explanation of this phenomenon is intercalation of the type H,MoS2. In one case a value of x = 0.37 was measured indicating that approximately onethird of the bulk material was intercalated. They also reported that while the rate of chemisorption increased in the presence of Co, the total uptake was unaffected. Furthermore, neutron scattering studies directly observed the chemisorbed hydrogen through the appearance of a new peak at 640 cm-' in the low-frequency neutron vibration spectra of MoS2. This type of spectra is roughly equivalent to an 1 R spectrum but much more sensitive to weakly bound hydrogen. The authors also observed a slight contraction of the MoSz lattice using low-angle neutron scattering data. They concluded that the hydrogen was bound to sulfur atoms in the van der Waals gap. This study raises several interesting questions but appears to establish unambiguously that intercalation of hydrogen is important under catalytic conditions. This fact has generally been missed by catalytic scientists working on TMS. This study further demonstrates the danger of drawing conclusions from data obtained at room temperature and low pressure in this field. C. SURFACE CHARACTERIZATION USINGSPECTROSCOPIC TECHNIQUES

The elemental composition and the oxidation states of surfaces are most frequently determined by X-ray photon spectroscopy (XPS), UV photon spectroscopy (UPS), Auger spectroscopy, and high-resolution electron loss spectroscopy (HREELS). Several limitations can be encountered when using these techniques. In situ characterization cannot be achieved since ultrahigh vacuum is needed. Because of the sulfides' high sensitivity to air exposure, the use of external cells for high temperature and sulfiding is mandatory for obtaining good data. With few exceptions (molybdenite and pyrite), sulfide single crystals are not commercially available and must be

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

203

synthetically grown by vapor transport. The resulting crystals are small ( 1 to 10 mm diameter), which renders the analysis more difficult. When considering sulfides from the isotropic class, the small crystal size is the major hurdle. With the exception of RuS2, almost none of the isotropic sulfide catalyst surface has been characterized. This is the consequence of the almost exclusive use of Mo- or W-based catalysts in the petroleum industry. Most surface studies, which are limited for the above described reasons, have been carried on MoSl and its promoted analogs. In contrast to isotropic sulfides, crystals of layered sulfides are easier to grow but are very thin. This has led to numerous studies on the structure of the basal plane, including atomic resolution imaging by scanning tunneling microscopy (STM) (Fig. 12) (57). Most importantly, Salmeron et al. showed that the basal plane was relatively inert by physical adsorption studies of thiophene at low pressure (58). Spatially resolved electron energy-loss spectroscopy

F I G . 12. STM image of ReS2 basal plane (57). Atomic spacing corresponds to actual ReSz basal plane spacing.

204

K. K. CHIANELLI, M . DAAGE, AND M . J . LEDOUX

(SREELS) has also been achieved on MoSz platelets, showing an enhancement of surface plasmon modes at corners and edges (59). The surface plasmon at the edges of the crystallites presumably is associated with the active sites. These studies emphasized that further characterization should be focused on the edge surfaces in MoSz and other related systems and this remains the major challenge to the surface scientist. For example, the complexity of the problem is shown by noting that in MoS2 the edge surface structure depends on the stacking sequence of the layers, and different surfaces would be obtained for the rhombohedra1 (3R) and hexagonal (2H) structures of the crystal. However, each layer will be terminated with one of two types of Mo atoms that correspond to the (iOl0) and (1070) planes in 2H MoS2 (Fig. 13). Surface techniques can therefore be applied on synthetic edge surfaces. A good example is found in the XPS and UPS study of edges created by lithography of a single crystal of MoS2. In particular, it was found that the Mo atoms on the edges have a lower oxidation state (less than 4) than that of the bulk (60). Another surface science approach consists in cleaning an edge surface of a natural crystal. These surfaces are generally thicker than synthetic crystals, but do contain some metallic contaminants. A HREELS study of CO adsorbed on a reduced edge surface has been carried out by using this approach. Temperature-programmed desorption (TPD) and Temperature programmed reduction (TPR) are also useful and provide some additional information on the reducibility of the different types of sulfur atoms

‘Terminal S

0Bridged S FIG. 13.

0Bulks

Edge structure of MoS2

=

Mo

205

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

and their ability to form anionic vacancies (61). For example, it is believed that in MoS2, terminal sulfurs are removed below 200°C, whereas bridging sulfurs react between 200 and 500°C (62). Above 500"C, the triply bound sulfur becomes unstable and the crystal structure collapses. Also, the defects located on the edge planes have well-defined optical characteristics and a midgap optical absorption similar to that observed for other semiconductors, providing a direct correlation between the electronic structure of MoS2 and its catalytic properties. The infrared optical absorption is determined by using photothermal deflection spectroscopy (PDS) (46). This is a calorimetric technique that directly measures the deflection of a laser beam. It is insensitive to optical scattering and thus ideal for powders. The defects that are catalytically active would be expected to have energy levels lying between the conduction and valence bands of MoS2 and thus absorb photons with below band gap energies. In fact, a set of X a calculations, which modeled different types of sulfur vacancies (Fig. 14) that could occur at the edge surface, showed that allowed optical transitions fall into the observed energy ranges below 1.2 eV (63). These results suggest that the sulfur vacancies are responsible for the optical absorption measured for the edge planes. This technique is particularly useful since it is the only spectroscopy that allows the correlation of the catalytic properties for a wide range of MoS2 dispersions (single crystal to poorly crystalline powders). Note, however, that it cannot be used for conductors such as cobalt promoted MoS2. NMR, Mossbauer, and EXAFS can also be used as a surface-sensitive technique (64-66, 18). They are particularly useful for char1

c

.-c0

0-i-

F

1

0

u)

n

a

large platelets

0.ol-

0.001

I

I

I

I

I

I

206

R . R . CHIANELLI, M. DAAGE, A N D M . J . LEDOUX

acterizing the structure of the promoter atoms of a layered TMS catalyst, since they are located at or near the edge surface. Surface sulfur structure can also be analyzed by using selenium as a tracer atom that can be characterized by solid-state NMR (67). VI.

Structure/Function Relationships

Considerable progress has been realized in characterizing the TMS materials. Much effort has been put into trying to understand the fundamental basis for their catalytic activity and selectivity. Numerous relationships that highlight the particular importance of the electronic structure, the crystallographic structure, and the sulfur vacancies have been reported in the literature. A. IMPORTANCE OF THE ELECTRONIC STRUCTURE Periodic effects, which describe the ability of a TMS to catalyze a given reaction, form the underpinning for any fundamental understanding related to the electronic structure. First, measured for the desulfurization of dibenzothiophene, a typical “volcano” plot between the activity and the periodic position emerged (Fig. 15) (39).Group VIII TMS such as Ru, Rh, Os, and Ir were the most active. Similar trends were later observed for hydrogenation reactions and HDN (68-71), as well as for the hydrotreating of a heavy gas oil (72). These trends are of fundamental importance because they emphasize the importance of the 4d and 5d electrons and the metal sulfur bond strength in catalysts. Several correlations to the HDS activity were described in the original paper (metal-sulfur bond strength, optimum heat of formation of the sulfide, and Pauling percentage d-character). That the heat of formation of the sulfides are also linearly correlated to heat of adsorption of sulfur on the transition metals has been shown by Bernard et al. (73). A theoretical foundation for understanding these correlations is found in the calculated bulk electronic structures of the first- and second-row TMS. The electronic environment of the metal surrounded by six sulfur atoms in an octahedral configuration was calculated, using the hypotheses that all the sulfides could be represented by this symmetry as an approximation. There are several electronic factors that appear to be related to catalytic activity: the orbital occupation of the HOMO (Highest Occupied Molecular Orbital), the degree of covalency of the metal-sulfur bond, and the metal-sulfur bond strength. These factors were incorporated into an activity parameter ( A 2 ) , which correlates well with the periodic trends (Fig. 16) (74, 75). This parameter is equal to the product of the number of electrons contained in the

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

207

HDS activity

lo1

lo’

’

IV

v

VI

VII

VIII

IB

FIG. 15. Periodic trends for HDS (39).

HOMO of the molecular sulfide by a factor B , which “measures” the nature (symmetry and strength) of the metal-sulfur bond. It is now possible to understand why the heat of formation was an insufficient parameter, because it did not contain the same “quantities” as this B parameter. In the case of manganese, the agreement between the “measured” and the “calculated’ is excellent. To be a good catalyst, the sulfide should have a HOMO filled with as many electrons as possible and be of r28 symmetry instead of e , .

208

R . R . CHIANELLl, M . DAAGE, A N D M . J . LEDOUX

1000 -

1100

900 800 L Q)

U

2 a

z

:1 0 l 8

-

700 600 500

-

h

110

-

U

*

.CI

.C1

U

u

400 -

Q

2 300 a u ; 200 U

u

;1 0 1 6

100 0

I

I

1

I

I

I

A shortcoming of the above approach lies in the fact that the HDS activity of bulk sulfides was measured, the real number of accessible active sites was ignored, and only the apparent activity instead of the turnover number per site was obtained. To overcome this difficulty, Ledoux et aE. (76) have carried out the same study as Chianelli but dispersed the sulfides at very low metal loading on an activated charcoal and checked that they reached the maximum dispersion. In this case, the turnover number per site could be quite safely deduced from the apparent specific reaction rate. By measuring the HDS activity for thiophene at very low conversion to fulfill kinetic requirements, Ledoux found a better agreement between his measurements of activity and the calculated values obtained by Harris and Chianelli than they themselves had obtained. For instance, the maximum activity in the periodical trend was observed on rhodium instead of ruthenium for the second periodical line and on iridium instead of osmium for the third line (Fig. 17). Another study performed at the same time by Prins and co-workers (77) on charcoal was less convincing because of the high metal loading and the too high reaction conversions, but it showed a very similar periodic

209

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS 1000

100

5

.-u

. I

i,

U

m

a,

2

. c) I

m

I

a, k

10

g

3:

1

1

I

I

I

I

I

0.1

FIG. 17. Periodic trends on carbon supports (76)

trend. Other similar trends have been seen in hydrogenation reactions (78) and in real feeds, and it is now well established that these trends are the most fundamental aspect of catalysis by TMS. Ledoux et al. (76) further improved the theory described above by removing the assumption made in the earlier work to simplify calculation that each sulfide should be considered in an octahedral symmetry. The early TMS have coordination of 6, but for the latter TMS (i.e., Pd, Pt, Ni) the stable form requires a coordination of 4.The situation is even more complex for Ir, Rh, and C o , where the two coordinations of 4 and 6 coexist in the same structure. This could be the reason for the poor fit between the calculated values and the experimental data for Co, Ni, Pt, and Pd. The original calculation showed that the HOMO of the molybdenum sulfide contains two

210

R . R . CHIANELLI, M . DAAGE, AND M . J . LEDOUX

electrons in a t2Rsymmetry while the HOMO of the cobalt sulfide in octahedral symmetry contains only one electron in e, symmetry. However, eight cobalt atoms in nine per unit cell of the stable cobalt sulfide Cogs8have tetrahedral symmetry. Using the same calculation but introducing a tetrahedral symmetry for the cobalt sulfide leads to a HOMO containing three electrons in a t?#symmetry, which better explains the HDS activity of the cobalt sulfide. Ledoux et al. (76) proposed that the electronic configuration of the HOMO was a necessary condition but not a sufficient one. The structure of the sulfide should be taken into account. All the active sulfides not only exhibit an adequate electronic configuration, but also have the ability to adsorb sulfur reversibly in over stoichiometry, either with an amorphous phase [MoS2/MoS7, WS2/ WS?, Re& /Re2S7, M(Ru or Os)Sr/M(Ru or Os)S2+ ,I or with a crystalline phase at a higher coordination number [PdS(coord4)lPdSz(coord8), PtS(coord4)/P tS2(coord6)]. The introduction of this structure requirement adds an additional important factor to the electronic theory, because it avoids the formation of a sulfur vacancy to adsorb the gas-phase sulfur, a vacancy that destroys the electronic structure of the HOMO. This also explains the very high HDS activity of Co, Rh, and Ir in their respective periodic lines because they are perfect catalysts in terms of the Sabatier definition (81). Coordination 4 has the right HOMO for HDS and is the most stable form, but, naturally, coexists with coordination 6, which can be temporarily adopted to adsorb the gas-phase sulfur without perturbing the electronic population of the HOMO in creating a sulfur vacancy. This concept of metal lability to sulfur adsorption/desorption, coordination flexibility, and oxidation state requires more work, but will be crucial to further progress in understanding the TMS catalytic materials. The effect of the metal-sulfur bond energy is also supported by the trend between activity and heat of formation of the binary sulfide. The highactivity catalysts have a heat of formation ranging from 30 to 50 kcal/mol. The metal-sulfur bond should be strong enough for the sulfur-containing molecule to bind to the metal, but weak enough for H2S to desorb. Of course, this scheme was an oversimplified one not adequately fitting all TMS catalysts such as MnS, which, as the authors pointed out, did not contain the crucial 4d and 5d electrons. The calculated trends previously mentioned above did a better job, but it was this simple correlation that first lead to a unified view of synergy as described below. Recently, Norskov et ul. (79) extended the theoretical investigation of the TMS using ab initio calculations. They emphasize the binding energy of sulfur to the transition metal as the important variable in explaining the periodic trends and the promoted systems in keeping with the original work. They emphasize the importance of the sulfur chemisorption energy and the sulfur-induced distortions of the metal lattice. This work represents a con-

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

21 1

tinuing effort by several groups-to understand these trends in a more rigorous manner. That the sulfur chemisorption energy should be related to the heat of formation was pointed out above. B.

AN ELECTRONIC HYPOTHESIS FOR SYNERGY

Based on the above and the optimum heat of formation idea, Chianelli et al. were the first in 1983 to propose a real “chemical” explanation for the synergy (80). In this model of reaction, the desulfurization of an organic molecule was considered equivalent to the sulfurization of the metal used as catalyst. The metals for which the sulfide has a high heat of formation are too strongly bonded to the sulfur, are poisoned by the sulfur coming from the gas phase, and cannot regenerate the sulfur vacancy, which in turn is the catalytic site. The metals of which the sulfides have a low heat of formation are too weakly bonded to the sulfur, the adsorption of the organic molecule by its sulfur atom is too weak, and the reacting molecule rarely bonds. Only metals of which the sulfides have an intermediate heat of formation, strong enough to lead to a significant adsorption of the organic molecule and weak enough to release H2S after desulfurization, will be good catalysts. This is exactly the definition of a good catalyst given by P. Sabatier at the beginning of the century (81). By averaging the heat of formation of the sulfides of each of the following couples CoMo, NiMo, Cow, and NiW, all couples presenting the phenomenon of synergy, one finds a medium value between 30 and 50 kcal/mol. It is then only necessary to imagine that the labile sulfur atom, or the corresponding vacancy, is shared by one Mo or W atom on one hand and one Co or Ni atom on the other hand. If the Sabatier principle, applied to the heats of formation, is acceptable for the monometallic sulfides, its extension to the bimetallic systems or couples is less easy, because even if a strong interaction between cobalt and molybdenum for instance exists, it is unlikely that it can be evaluated by the simple arithmetic average of the heats of formation of the two sulfides taken individually. Nevertheless, the simple scheme did for the first time give a common origin to all pairs of known promoters and pairs such as Fe/Mo, which were not promoters. In this sense it generalized the earlier ideas of electron transfer proposed by Voorhoeve to include all relevant catalyst systems. It also suggested that promotion occurred through mimicking of noble-metal sulfide electronic structure, and the term “pseudobinary synergic pairs” was coined. Presumably, the important electronic interaction occurred at the interface between the two-phase-separated sulfide phase. This idea is reminiscent of “synergy by contact.” The promotional effect of first-row transition metal on the activity of molybdenum-based catalysts has also been correlated with a similar activity

212

R . R. CHIANELLI, M . DAAGE, AND M . J . LEDOUX

parameter suggesting that the interaction between the Mo 4d electrons and the 3d electrons of the second metal is required (82). This interaction has been theoretically described in terms of the existence of an electronic transfer between the two metals at a site existing near the surface or interface of the two relevant phases as described above (Fig. 18). For example, Co and Ni act as promoters by increasing the number of electrons in the HOMO from two in MoSz, to three in CoMo and to four in NiMO systems, respectively. In the catalytic tests, the HDS activity of the pairs V-Mo, Cr-Mo, Mn-Mo, Fe-Mo, Co-Mo, Ni-Mo, Cu-Mo, and Zn-Mo were measured, showing that V, Mn, Fe, and Zn were neutral toward the molybdenum sulfide activity, while Cu and to a lesser extent Cr were poisoning MoS2, and Co and Ni were promoting the activity of MoS2. The calculation used the same conceptual approach as that for the monometallic sulfides but applied to a molecular entity containing one Mo atom, nine S atoms, and one transition-metal M (M = V, Cr, Mn, Fe, Co, Ni, Cu, or Zn). In octahedral coordination, the calculation showed that Co and Ni had the property to transfer electrons to Mo while, inversely, Cu and Cr had the property to attract electrons from Mo. The other metals did not transfer or attract electrons to or from Mo. Thus, the presence of Co or Ni should increase the number of electrons in the HOMO of MoS2 and in consequence increases its HDS activity as described above. The correlation between measured activity and calculated parameter is illustrated in Fig. 18. This electronic explanation of the synergy between Co or Ni and Mo does not give any direct information on the structure of the active phase except that cobalt or nickel atoms should be bonded to molybdenum atoms through bridging sulfur atoms according to the molecular model used by the authors. The precise structure of this “pseudobinary cluster” is as yet unknown, but

800

600

400

200

V

Cr

Mn

Fe

Cu

Ni

Cu

Zn

FIG. 18. Calculated trends vs experiment for the synergic pairs (80).

213

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

is consistent with several of the models previously described. For further progress to be made, a deeper understanding of the effect of structure is required, and this is the subject of the next section. C.

IMPOWANCEOF CRYSTAL STRUCTURE

As mentioned in the preceding sections, the TMS catalysts belong to two major classes defined by the isotropic/anisotropic character of their crystal structures. Evidence of this important distinction is found in the existence or lack of correlation with the surface area, as illustrated in Fig. 19 for RuS2 and M0S2. The isotropic nature of RuS2 is demonstrated in the linear correlation between activity and surface area. RuS2 is an example of a well-behaved catalyst system unlike M0S2, where the anisotropy plays a large role and BET surface area does not correlate linearly with activity. However, linear relationships between the activity for DBT HDS and the oxygen chemisorption are observed for both materials (Fig. 20). This indicates that the chemisorption is at least proportional to the density of catalytic sites, but cannot be used to measure their concentration without a precise knowledge of the stoichiometry of the chemisorption process. For RuS2, the stoichiometry can be estimated from the slope of the linear correlation found between the oxygen chemisorption and the surface area. A value around 0.5 oxygen atom per surface Ru atom is found. This suggests 40

I

I

I

I

300

I

0 250

h

150

5

k

0

=e 0

100

g

i/i

t

10

v

50

0

0 0

20

40

60

80

100

Surface Area (rn2/g)

FIG. 19. Surface area vs HDS activity for MoS2 (84).

120

214

R . R . CHIANELLI, M . DAAGE, A N D M . J . LEDOUX

0

so

100

Oxygen Adsorbed (pmoldg)

FIG. 20. O2 chemisorption on RuS2/MoS2vs HDS Activity (84).

that a corrosive process is not likely to be associated with the chemisorption. In the case of M o S r , such estimation of the stoichiometry is more difficult to obtain. Because it was shown that the oxygen accumulates on the edges of the crystallites, the knowledge of the edge surface area becomes critical. The only simple way to obtain such information is to geometrically estimate the edge surface area of well-defined microcrystals from SEM micrographs. By measuring the catalytic activity of microcrystals having an average dimension of 1.7 pm, the turnover frequency for the HDS of DBT has been determined to be 0.079 molecule/site/s ( 4 4 ) . Since it has been shown that the optical defects measured by PDS are of the same nature regardless of the crystallinity of the sample, this turnover frequency is in fact applicable to any MoS2 catalysts. Consequently, a direct correlation between the edge Mo density and the chemisorption of oxygen has been established (Fig. 21). The slope of the linear relationship shows that the chemisorption at room temperature is corrosive and that 15 oxygen atoms per edge Mo are consumed. Numerous other correlations with activity have been reported for MoS2. These include measurements such as the ESR spin densities, magnetic susceptibility, and other chemisorptions (as referenced above). The promotion of bulk binary sulfide is limited exclusively to the promotion of molybdenum and tungsten by the first-row transition metal. The effect of the structural promotion (creating more of the same sites) is always coupled to the electronic promotion (creating more active sites). One ap-

215

TRANSITION-METAL SlJLFlDE CATALYTIC MATERIALS

1

I

l

1

2

3

0 (I

4

Site Density (ld’sitedg)

FIG. 21. 0, adsorbed edge atoms vs site density (84).

proach consists of estimating the promotion factor obtained for different morphologies and/or structures of the Mo/S systems that will convert to the catalytically active Co/Mo/S system. For example, when different MoSz microcrystals and poorly crystalline powders are promoted by reaction with cobalt carbonyl, the promotion factor for the HDS of DBT varies between 3.3 and 16 (37). Once corrected for the contribution of the electronic factor, the remaining promotion factor lies between 0 and 4.3. The highest values have always been obtained for materials where significant morphological changes that can be observed by TEM occur upon promotion. It is believed that weak bonds that prevail in the highly disordered materials are being broken, resulting in the formation of smaller particles, hence more catalytic sites. That is the effect responsible for the wide range of promoter to molybdenum ratio observed in the literature. Values as low as 0.1 and as high as 0.8 have been claimed, and numerous relationships have been established for different reactions. The same study (37) does provide the clearest evidence for electronic promotion without confusion of edge area change. The microcrystallites investigated were hexagonal platelets of average diameter 1.7 p m corresponding to an edge site density of 6. I X 10” sites/g. After promotion with Co carbony1 and sulfiding, no change could be detected in the edge area. When the activity was measured for HDS of DBT, the relative rate changed from 4.8 to 16.1, a factor of 3.4. There was no change in the selectivity, as the hydrogenation rate of these crystallites was not detectable. It could not be

216

K. K. CHIANELLI, M . LIAAGb, A N D M . J . LEIIOUX

determined in these studies whether this was the maximum factor obtainable because the final Co/Mo ration was uncertain. Future research involving microcrystallites with well-defined edge areas should give a clear picture of the electronic promotion factor. It is clear that the close association of the Co with molybdenum is critical. As previously described, good correlations between the amount of Co present in the MoSz-like “CoMoS” phase as measured by Mossbauer emission spectroscopy have been established ( 1 9 ) . In this phase, the cobalt is thought to be located on the edge of a small slab of MoS2 allowing the formation of the Co/Mo pair required for the electronic promotion. However, it remains unknown how many Co per edge Mo are needed for maximum catalytic activity. We previously noted that the CoMoS phase generally coexists with the binary phases and that its amount and dispersion is highly dependent upon the mode of preparation. D.

ROLE OF CRYSTAL STRUCTURE O N

REACTION

SELECTIVITY The above discussion centers on the total HDS activity. However, equally important but little studied is the question of selectivity of the HDS reaction and the physical basis for its control by the wlfide catalyst. The theme of two sites on MoS2 is well established throughout much of the literature cited in this article (see Voorhoeve et al. ( 1 6 ) for example). One site is usually described as a sulfur removal site and the other as a hydrogen site. Tanaka and Okuhura (83) in an elegant experiment showed that by cutting a single crystal into several pieces the selectivity of the reaction investigated could be modified. The pathway for the HDS of DBT is shown in Fig. 22. The reaction has two primary products biphenyl (BP) and tetrahydrodibenzothiophene (HDBT). In one of our investigations it was noted that on the small microcrystals of MoS,, BP was the only product observed in the HDS of DBT and that no hydrogenation to HDBT occurred (84). This is in stark contrast to normally encountered poorly crystalline MoS, , which exhibits both products in varying amounts. In these materials the formation of HDBT is always observed, leading by desulfurization to significatlt amounts of cyclohexylbenzene (CHB). The microcrystals of MoS2 could be induced to form HDBT as a product through an etching procedure using either HCI or 0,. Furthermore, when HCI was used as an etching agent, the optical absorption of edge surfaces of the microcrystals as measured by PDS increases. Since [he optical defect density has been correlated with the HDS activity, the change observed reflects a change in the catalytic performance of such microcrystals. TEM

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

217

FIG.22. Reaction pathway for the HDS of DBT (84).

studies on related samples revealed that the etching process "roughened' the edges creating steps in the previously smooth well-ordered samples, strongly suggesting that the hydrogenation reaction occurs on the steps and that PDS measures both sites. These hydrogenation sites, we call "rim" sites for reasons described below, are not related to the chemical treatment used for the etching (influence of residual chlorine for example), but rather to the morphological changes that introduce steps into the edge surfaces (84). Poorly crystalline MoSz powders were prepared by annealing of amorphous MoS3 in a sulfiding atmosphere, leading to the formation of a structure previously termed the rag structure (44). By varying the preparation conditions, such as the annealing temperature, one can vary the stack height (average number of MoS2 layers in sample) as well as the radial dimension of the layers. Higher temperature increases the crystalline order along the c axis and thus the stack height ( h ) as determined by narrowing of the 002 Xray diffraction. The series of poorly crystalline MoS2 catalysts prepared in this manner cover a large range of stack heights (3 < layers < 1000) and they were tested with the probe DBT reaction. The rate constants for the hydrogenation reaction (kHDBT)and the formawere obtained by using a Langmuir-Hinshelwood kition of biphenyl (hp) netic model to describe the reaction pathway illustrated in Fig. 22. Both rate constants decrease as MoSz becomes more crystalline as the average radial dimension of the layers also increases, causing the total edge area to drop. However, the formation of the HDBT product is more dramatically affected than the BP production. For the former, a 36-fold variation is observed, whereas for the latter an 8-fold variation occurs as the crystallinity increases. The ratio S = kHueT/& (S represents selectivity) is linearly correlated with the line width of the 002 reflection peak (Fig. 23), the reciprocal of the stacking height ( h ) . This result is unique in catalysis, a simple physi-

218

K . K . CHIANEILI, M . DAAGE, A N D M . J .

LEDOUX

layers

3

n

m

.

2

Y

I-

m

0

I Y

1

0

0

2

1

3

degrees

A002 FIG.

23. Selectivity vs X-ray difhdction parameters ( 8 4 ) .

cal parameter (X-ray line broadening) correlated to the selectivity of the reaction. We have developed a two-site model for this result. Both sites are at the edge of MoSz and the total amount of sites (and thus the total activity) is determined by the total edge area in keeping with the accepted picture of MoS2 (85).However, the two sites are distinguished by their location on the edge. One site exists at the edge of exterior layers with adjacent basal planes exposed to the reacting environment. These are the “rim sites” on which both hydrogenation and desulfurization occur. The second site occurs on interior layers that have no exposed basal plane surfaces. These are the “edge sites” and only the desulfurization reaction occurs on this site. The relative proportion of rim and edge sites can then be calculated by using a simple physical model to describe the morphology of the MoSz particle as illustrated in Fig. 24. This model assumes that the catalyst particles are made of discs n layers thick with a diameter d. In this case, the ratio of rim sites to total sites can be deduced from the expression r

r

+e

-

2rrd 2 rrdn n’

-

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

219

Diameter

Sites

=

rim

0edge EZZl basal

-1

stacking height

FIG.24. The rimledge model (84).

where r is the number of rim sites and e the number of edge sites. It is important to note that the ratio of sites does not depend upon the particle diameter, but only on the stacking of the layers. Thus, the total activity is determined by the diameter of the disc and the selectivity by the number of layers. The number of layers is determined from X-ray diffraction and the diameter of the disc is determined as previously described. In real catalysts, the average diameter of the disc corresponds to measuring the total edge area with all of the difficulties discussed. The model predicts that catalysts with a predominance of single unstacked layers will have a greater selectivity to hydrogenated products than those that with a predominance of stacked layers. In a previous section we described single layers of MoS2 supported on small crystallites of pyrrhotite (FeS,-,). Catalysts of this type were tested on aromatic distillate feeds and exhibited stable activity for over 500 hr at temperatures of up to 355°C in the absence of any conventional support material (49). This attests to the stability of the MoS2/FeSI-, interaction, which is of the sulfide-sulfide-type interactions described above. These catalysts showed unusual selectivity toward HDN in the presence of sulfur. HDS was relatively low. Conventional thought held that HDS was always easier than HDN and therefore a catalyst with high HDN and low HDS is unusual. However, in keeping with the rim/edge discussion above it is expected that hydrogenation will be favored on single layers and therefore HDN. Thus, this catalyst is an example of the rim/edge idea confirmed in a real feed system. Complete understanding of the catalytic properties of MoS2 requires more knowledge of the edge planes that terminate the anisotropic layers and are the location of catalytically active sites. Further progress in developing this

220

K . K . CHIANELLI, M. DAAGE, A N D M. J . L.EI)OUX

understanding has been hindered by the absence of absolute knowledge of MoS2 dispersion leading to models that greatly underestimate crystallite sizes. However, we do know that the bulk electronic structure of MoS? is directly related to the catalytically active surface states. The catalytic states, created by edge termination or more precisely sulfur vacancies, lie just above the dz2 band near the Fermi level. Optical and electron spectroscopic measurements on geometrically well-defined catalysts yield an HDS edge site turnover frequency that does not suffer from an ambiguity in dispersion. This turnover number now permits the determination of the dispersion of all unpromoted MoS2. It remains for future reasearch to determine the precise structure and chemisty of these sites supported by the bulk MoS2. The catalytic selectivity of MoS2 for the HDS and hydrogenation reactions is understood in terms of the existence of two different sites. Their relative concentration is directly dependent upon the morphology of the particle and is controlled by the height of the layer stack. Hydrogenation reactions occur exclusively on “rim” sites whereas both rim and edge sites are responsible for the desulfurization activity. At this writing, the origin of this effect is not understood; however, steric interactions are strongly suspected. Therefore, the effect may not be seen in small molecules such as thiophene and may be even more dramatic in molecules that are larger than DBT. A deeper understanding of and control over the selectivity as suggested by the current results can have a profound effect on future processes, which will require greater selectivity to meet environmental demands.

E. ROLEOF SULFUR VACANCIES The formation of sulfur vacancies is also related to the electronic effect because of the relatively narrow range of metal-sulfur bond strength required for high catalytic activity. Consequently, there is a general belief that sulfur vacancies are critical. Note, however, that recently some organometallic studies suggest that the catalytic reaction may involve the second, rather than the first, coordination shell of the Mo atom. Given the high temperature and the reducing atmosphere under which the catalytic reactions are generally performed on TMS, there is little doubt that the vacancies exists in significant amounts. Several structure/activity relationships strongly support this idea. The presence of optical defects that absorb below the band gap has been associated with sulfur-deficient molybdenum atoms and linearly correlates with the HDS of DBT. Another example, even though on a supported catalyst, clearly indicates that the removal of sulfur is critical for the hydrogenation and isomerization of 1,3-pentadiene (62). Finally, the inhibition effect of H2S is well known and is found for many reactions. As described above, the addition of promoter also results in the

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

22 I

modification of the sulfur species on the surface. A sulfur shared by a Co and a Mo is more likely to be removed than a sulfur shared by two Mo atoms. This is in fact observed in the reactivity of small Co/Mo/S clusters. F. RULEOF CARBON Reczntly, the entirc surface picture of the active sites regarding the role of sulfur vacancies has been called into question. This has come about because evidence is building that carbon plays an important role in the surface properties of active sulfided catalysts. Early evidence for this can be found in the patents of Chianelli and Pecoraro (86). They found that a RuS2 catalyst that had been stabilized in a catalytic environment had a composition, RuSz xC,. This composition corresponded to the 100-A catalyst particle having its surface sulfur stripped and replaced with carbon. This behavior was also exhibited by the layered TMS (27). Prins et 01. review the emerging evidence that the role of carbon is significant. The Mo single-crystal work of Bussell and Somorjai (87) is particularly noteworthy in this respect. The role of carbon is an important area for future study. However, the early results strongly suggest that a surface model for the HDS reaction on sulfided catalysts remains to be developed and that all current models require at least some modification to include the role of carbon.

VII.

Recent Developments in the Description of the Active Promoted Phase

Electron microscopy has failed to give a clear picture of the CoMoS or NiMoS phase supported on alumina because of interference from the alumina support itself and because of the high dispersion of the sulfide. However, Pratt and Sanders (88) for bulk NiMo sulfide and Vrinat and De Mourgues (89) for bulk CoMo sulfide were able to produce pictures from electron microscopes, showing a high dispersive effect of cobalt and nickel added to molybdenum sulfide. With very high-resolution transmission electron microscopy, Yacaman et al. (90)and Topspre et al. (19) were even able to localize cobalt on the edge of bulk MoSr . The Yacaman study also clearly showed the existence of a pseudointercalated phase in Co-doped MoS2 single crystals. The same result was also found by Angulo et al. (32) using scanning Auger spectroscopy on single crystals of MoS2 decorated by cobalt. Many groups (91, 92) have used EXAFS technique to obtain a more detailed structure. This technique showed that most oi the molybdenum atoms were, in terms of number of neighbor sulfur atoms and Mo-S distances, in a M o S A k e structure. Attempts to measure the dispersion or the

222

K. R . CHIANELLI, M . DAACE, A N D M . J . LELIOIJX

size of these M o S A k e structures are highly questionable because of disorder as described above. Ledoux et al. (64) introduced ”Co metal solid NMR to this problem. CO&Bcontains two different cobalt sites, one octahedral, and eight tetrahedral per crystal unit (Fig. 25). They can be easily distinguished by NMR because of their environmental and electronic properties. The octahedral site has a spherical symmetry in terms of neighbor cobalt atoms; it is thus placed in a homogeneous electric field. The spin of the cobalt nucleus is 7/2, and one expects a quadrupolar splitting into seven bands in an electrical gradient field. In the homogeneous field, the octahedral site must show a single band. The tetrahedral sites are positioned in a nonspherical symmetry, inside an electric field gradient, and must show even bands. In addition, the octahedral site is less “conducting” than the tetrahedral site because of the distance between the cobalt atoms. One can expect a much longer relaxation time T I for the former site. These bands appear in the ratio of 8 : I in wellcrystallized Co9Sx. On Co and CoMo catalysts supported on silica or carbon, Ledoux et al. (64) found two Co sites similar to the two sites existing in Co9Ss at high metal loading at the appropriate ratio for Co9Sx. At lower loadings, a new band appeared over a broad background. They assigned the latter to a distorted tetrahedral site and the former to a site that is close to octahedral with a short relaxation time, which, for this reason, were called “rapid octahedral.” A structure that explained the different observations made in catalysts studied was proposed. The structure localized the distorted rapid octahedral cobalt or “trigonal prismatic” cobalt, on the edge of MoS2, sharing four sulfur atoms with molybdenum atoms on one side, and sharing other sulfur atoms with the “tetrahedral distorted” cobalt on the other side. This structure is precisely the type of interfacial structure or surface structure required

0 Sulfur 0

FIG.25.

Structure of Co& ( 6 4 )

Cobalt

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

223

by arguments made in previous sections. We may also observe that the distorted rapid octahedral structure could also correspond to the distorted octahedral coordination in ReSz. These tetrahedral distorted cobalt atoms can be observed by NMR as a pure phase on carbon supports in the absence of molybdenum and are thus stable; these probably correspond to the Co sites observed by Topsgie’s group using Mossbauer spectroscopy because CrajC et al. (93) found a similar Mossbauer doublet for both cobalt in CoMo catalysts and pure cobalt sulfide on carbon support. They are also active for HDS and confirm the findings of Prins and co-workers (94) and Ledoux (96). These different structures are in full agreement with the XANES experiments performed by Prins and co-workers (95) and Ledoux (96). These structures also led Ledoux et al. to an incorrect interpretation of the synergy effect (64). On poorly dispersed catalysts supported on silica or in bulk form, their presence and activity are large enough to explain the increase in activity when cobalt is added to molybdenum, but on well-dispersed catalysts; i.e., on alumina or carbon support this interpretation is shown to be incorrect if the activity is carefully measured. In a second publication, Ledoux et al. (97) were able to measure the activity of catalysts, highly dispersed on carbon support, and at the same time to observe the size and shape of the active particles by high-resolution transmission electron microscopy, since the carbon support is almost transparent to electrons. On the most highly dispersed form of MoS2, the average size of the particles was 100 8, with one eighth of the molybdenum atoms located on the edges. By adding to the same catalyst the exact amount of cobalt or nickel to reach maximum activity, they observed two kinds of particles: circular with an average diameter of 36 8, and chain like, 150 8, long, and 14 8, wide. From these observations and from data described above, two structures were proposed (Figs. 26 and 27). In this model the “prismatic” cobalt atoms bind and stabilize the unsaturated sulfur atoms on the edge of small MoSz crystallites; the stable distorted tetrahedral cobalt atoms terminate the construction and stabilize the octahedral cobalt atoms. In the circular particles, one observed that 12 molybdenum atoms were stabilized by 6 cobalt atoms (3 prismatic and 3 tetrahedral) while in the chain-like particles, 2 cobalt atoms (1 prismatic and 1 tetrahedral) were enough to stabilize 4 molybdenum atoms. In each case, the ratio Co/Mo equals 1/2 and corresponds to the ratio observed at the maximum of synergy in these studies. When HDS activity was measured on these catalysts, an activity seven times higher than that of the MoS2 of 100-8,catalyst was found. If it is assumed that only the edge atoms are good sites for HDS catalysis, then the role of cobalt or nickel would be to stabilize a high dispersion of MoSz. The electronic transfer from octahedral cobalt to molybdenum calculated by Chianelli or observed by Topsgie and co-workers by XPS would have a double

224

I<. K. (’HIANELLI, M . I>AAGb,, ANI) M . J . I . t D O U X

0 00 00 0 0

$=36A 0

Distorted tetrahedral Co Rapid prismatic Co

0 Mo

0 s FIG.26. Small particle of promoted Co/Mo model (Y7).

effect. First it would provide more electrons in the molybdenum’s HOMO even after the formation of the sulfur vacancy which is accompanied by the loss of one electron in this HOMO and, second, it would stabilize the bond between cobalt and molybdenum atoms via the sulfur atoms. It should be pointed out, however, that the interpretation of these electron micrographs are still open to question. For example, many reports of “chain-like” structures appear in the literature; however, it seems that in many cases these chains are “turned up edges,” which are attached to the bulk layer. Interpretation of the two-dimensional TEM images, which are

f-

15OA

FIG. 27. Chain-like particle of Co/Mo (97).

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

225

projections of three-dimensional structures, is quite difficult and remains a future challenge. However, this does not affect the local chemical model described above. In addition the above models require highly sulfur-rich catalysts that are generally not observed. Very recently, Topsae and co-workers (98) for CoMo catalysts and Prins and Louwers (99) for NiMo catalysts performed EXAFS on the Co and Ni edges, respectively. The two groups reached the same conclusion and localized the Co(Ni) atoms on the edge of MoS2 crystallites in a very odd symmetry with five coordination. Each of these Co(Ni) atoms shares four sulfur atoms with molybdenum is a square plane with the fifth sulfur atom sticking out in the opposite perpendicular direction. As EXAFS gives an average situation and can only discriminate with large errors between different sites present together in the catalyst, the coordination five could be the result of averaging coordinations four and six as proposed in the model extrapolated from NMR analyses. The lack of precision due to the technique mainly affects the calculated number of neighbors, while distance are generally fairly accurate. Thus, the 5gC0NMR studies discussed above seem to give the best look at the unique Co and no such tool currently exists for Ni. As has been discussed above, knowledge of the dispersion of the catalyst is extremely difficult to obtain, especially in the promoted systems. This is the main obstacle to clear differentiation between the promoter models proposed and remains so today. Chemisorption techniques to count the active sites present on the sulfided surfaces have had limited success. O2 chemisorption has been associated with edge vacancies (active sites) on pure MoS2 (100).However, the same authors showed that it was impossible to correlate activity and amount of O2 chemisorbed on Co- or Ni-promoted molybdenum sulfide (101). The use of other test molecules was disputable, particularly NO, which can strongly modify the structure of the surface during the measurement (102). The introduction by Ledoux et al. (67) of "Se NMR applied to the catalysts' study gives further insight into this difficult problem. First they showed that 5yC0 NMR spectra in sulfided and selenided CoMo catalysts were almost identical, meaning that the CoMoS and CoMoSe had probably the same microstructure, justifying the substitution of sulfur by selenium. They next showed from the NMR analyses of MoSe2 bulk, of MoSe2 progressively dispersed on silica support, and finally of CoMoSe catalyst that two sorts of Se atoms clearly existed in terms of their magnetic properties: Se atoms with long relaxation time (TI > 2 sec) and those with short relaxation time (TI < 10 msec) bound to highly magnetic Mo atoms. As these more magnetic atoms are located at the edge of the particles because only these atoms bear anionic vacancies, a direct correlation between short T I selenium and dispersion was found. In addition, they directly confirmed that the addition of cobalt strongly increases the number of short TI selenium, thus the dispersion of MoS2.

226

K. R . CHIANELLI, M . DAAGE, A N D M . J . LEDOUX

In the second part of the study the same authors reacted thiophene on MoSe2 and CoMoSe silica-supported catalysts and took "Se NMR spectra before and after reaction. If the edge selenium atoms (i.e., short relaxation time) were the active mobile atoms, forming vacancies, they should be progressively replaced by sulfur atoms during the course of the reaction and disappear selectively from the NMR spectra. This was observed on two families of catalysts. This work provides for the first time direct proof of the active site location on the edge of the promoted MoSZ. One further comment should be made regarding the CoMoS phase described above and measured by the EMS technique introduced by Topsoe et al. (19). In a previous section we described a stable unsupported catalyst that contained single layers of MoSz in contact with crystallites of pyrrhotite (49). The precursor to this catalyst was completely amorphous and had an Fe Mossbauer spectrum virtually identical to the reported Co EMS spectrum of the CoMoS phase. This spectrum was pure FeMoS with no other phases present. The amorphous precursor had an approximate stoichiometry of FeS/MoS2, but because of its completely amorphous nature its structure is as yet unknown. During the catalytic run, it converted completely to the phase-separated catalyst described above with no detectable FeMoS component. This suggests that the CoMoS phase is a similar amorphous precursor, which is unstable under extended high-pressure catalytic runs. Its presence predicts the activity of the final stabilized catalyst but is not the active promoted phase. It predicts the activity because its presence leads to the required Co/Mo/S surface phase. More work needs to be done to prove this point but the evidence is very suggestive, and examples of stabilized catalysts that contain the CoMoS phase are needed.

VIII. Conclusion In this article we have reviewed the current understanding of the relation between the properties of the TMS and how they catalyze hydrotreating reactions. We have emphasized recent developments as well. There follows a short summary of what appears to be established and what is yet to be understood. It is now well established that the TMS are unique class of catalysts that are able to perform numerous hydrogenation and hydrogenolysis reactions in the presence of sulfur. In fact, they require sulfur for activity maintenance. The catalytic activity and selectivity of the TMS arises from the electronic and structural properties of the sulfides themselves. Support effects are secondary, improving sulfide dispersion and reducing metal cost in commercial catalysts. Fundamental effects can only be elucidated by studying TMS catalysts in their fully sulfided and catalytically stabilized states. Studies are of-

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

227

ten simplified by studying the unsupported TMS, which in the past have been commercial catalysts with extraordinarily stable catalytic activity. The study of supported catalysts can also be useful if the experiments are carefully designed and the results are compared consistently with those from unsupported systems. Study of the unsupported TMS has established their stable sulfide states under hydrotreating conditions for all active elements in the periodic table. The TMS catalysts can generally be divided into isotropic and anisotropic sulfides from a structural point of view. However, many sulfides are highly disordered, even amorphous, and some have large nonstoichiometries that may vary under different catalytic conditions. Nevertheless, measurements of catalytic activity across the periodic table using supported and unsupported catalysts show smooth volcano plots with maxima occurring in the Group VIII TMS for the second and third transition series. In the first transition series, the catalysts are also more active in Group VIII but significantly less active than the second and third transition series. These results establish the most fundamental properties of the TMS. The catalytic activity of the TMS is optimized when the maximum number of 4d and 5d electrons are present in a suljided stable state. Theoretical studies have extended this result to understanding the importance of electronic structure in stabilizing the 4d and 5d electrons in frontier orbitals, ?r-bonding, metalsulfur bond strength, and optimum heat of formation of the sulfide. Better theoretical studies are needed, particularly involving bound organic intermediates. The above understanding takes on additional significance when applied to understanding the promoted or “synergic” effects. The relation of these commercially important systems to the simple binary systems can be understood as arising from the same electronic origin. The catalytic activity of the promoted (and poisoned) TMS is optimized when the maximum number of 4d and 5d electrons are stabilized at the surface of the catalyst. This is accomplished by sharing sulfur atoms at or near the surface of the promoted phase or at the interface between two metals or by electron transfer between the metals that form a synergic pair. For example, COYSS and MoS2 can establish such a state at the interface between them (synergy by contact). This state can also be established at the surface of MoS2 (Co-promoted MoSd or at the surface of Cog& (Mo-promoted Cog Ss). This is called “symmetrical synergy” and all promoted catalyst systems are located by referring to a binary immiscible phase diagram appropriate .for the system. Although many studies and models have been proposed for the structure of Co(Ni)/Mo( W) entities that share sulfurs or transfer electrons at or near the surface, the exact picture is still uncertain. This is a result of the difficulty in measuring the dispersion of the layered TMS. All models depend for their proof on this knowledge. 57C0 Mossbauer and j’Co NMR

228

K. K. CHlANELLl, M . IIAAGE, AND M . J . LEDOUX

have given the best evidence that a unique Co/Mo species exists and is responsible for the promotion effect. However, the techniques are not yet widely used and none currently exists for Ni. It seems clear that pseudointercalation models and edge decoration models are beginning to describe the appropriate states, but details are still lacking because of the dispersion and disorder problem inherent in the layered TMS. Lack of knowledge of the layered TMS dispersion also leads to an ambiguity in describing the electronic role of the promoter. Does the promoter directly lead to a more active ‘)seudobinary” site or does the electronic interaction stabilize more of the same sites? The evidence, particularly the existence of symmetrical synergy, seems to argue for the direct electronic effect, but more evidence is needed. It is also now well established that the basal plane is inert and that reaction takes plucc ut the edge or edge-like defects on the luyered TMS. However, disorder has hindered the application of classical surface science studies to elucidating the structure and properties of these edges. It is known that maximizing the edge optimizes the activity of the catalyst and the effect of the promoter. Only PDS and 77SeNMR reliably measure edge dispersion, but these techniques are not yet widely available. However, PDS has permitted the calibration of model compound reactions, which can then directly measure dispersion, activity, and selectivity effects. It is a major conclusion of this report that such normalized model compound reactions are the best currently available methods of giving unambiguous information regarding the dispersion of the layered TMS. Where model compound studies that have been calibrated with physical studies are done, new fundamental information has resulted. A prime example of this is the understanding of the dependence of the selectivity of the DBT reaction on MoS2 physical structure. The ratio of hydrogenated to desulfurized products on MoSz depends on the degree of stacking of the MoS2 layers. This intriguing and fundamental result gives a foretaste of future progress in this field. Future work should include: 1. Model compound studies that are calibrated with physical studies and

connected to real feed studies. 2. Wider use of resonance and optical techniques that are not hindered by the presence of disorder. 3. Theoretical studies that deal with extended solids and interacting molecules. 4. Solid-state investigation of immiscible phase diagrams and the effect of surface area and carbon on them. 5 . High-resolution TEM and related techniques to establish local atomic structure in disordered systems.

TRANSITION-METAL SULFIDE CATALYTIC MATERLALS

229

We conclude this report by emphasizing that the role of carbon is the most neglected and potentially the most important factor not yet icvestigated in this system. Lack of understanding regarding the role of carbon hinders the development of an accurate picture of the role of the interacting molecules. We believe that by concentrating on this and the research items listed above, a new era of understanding of the TMS catalysts will emerge. This understanding is crucial to the continued commercial success of the TMS as environmental factors surrounding the use of hydrocarbon base takes center stage. ACKNOWLEDGMENTS The authors are extremely grateful to Joni Phillips and Margaret L. Lipani for assistance in preparing this manuscript.

REFERENCES I . Massoth, F. E., Adv. Cutal. 27, 265 (1978). 2 . Tauster, S. J., personal communication. 3. Weisser, O., and Landa, S . , “Sulphide Catalysts, Their Properties and Applications,” Pergamon, New York, 1973. 4. Prins, R., De Beer, V. H. J., and Somorjai, G. A,, Cutal. Rev.-Sci Eng. 31 (12), 1 (1989). 5. Farben, 1. G., Brit. Pat. 315439 (1928). 6. Pease, R. N., and Keithon, W. B. Jr., Ind. Eng. Chem. 25, 1012 (1933). 7 . Byrns, A. C., Bradley, W. E., and Lee, M. E., Ind. Eng. Chem. 35, 1160 (1943). 8 . Beuther, H., Flinn, R. A., and McKinley, J. B., Ind. Eng. Chem. 51, 1349 (1959) 9. Huffman, H. C., Union Oil of Cal., U.S. Pat. 2437533 (1948). 10. Engel, W. F., and Hoog, H., Shell, U.S. Pat. 2697683 (1954). 11. Hendricks, G. W., Union Oil of Cal., U.S. Pat. 2687381 (1954). 12. Ledoux, M. J., Hantzer, S . , and Guille. J., Bull. Soc. Chim. Belge 96, 855 (1987). 13. Schuit, G. C. A., and Gates, B. C., AIChE J. 19, 417 (1973). 14. Richardson, J. T., lnd. Eng. Chem. Fundam. 3, 154 (1964). 15. Gates, B. C., Katzer, J. R., and Schuit, G. C. A., “Chemistry of Catalytic Processes,” p. 415. McGraw, New York, 1979: De Beer, V. H. J., Van Sin Fiet, T. H. M., Van Der Steen, G. H. A. M., Zwaga, A. C., and Schuit, G. C. A,, J . Catal. 35, 297 (1974). 16. Voorhoeve, R. J. H., and Stuiver J. C. M., J . Catal23, 228 and 236 (1971); Voorhoeve, R. J. H., J . Catal. 23, 243 (1971). 17. Farragher, A. L., and Cossee, P., in “Proceedings, 5th International Congress on Catalysis, Palm Beach, 1972” (J. W. Hightower, Ed.), p. 1301 (1973). North-Holland, A m sterdam, 1973. 18. Hagenbach, G . , Courty, P., and Delmon, B., J. Cutal. 23,295 (1971); 31, 264 (1973). 19. Topsee, H., Clausen, B. S., Candia, R., Wivel, C., and Morup, S., J., Curd 68, 433 (1981). 20. Candia, R., Clausen, B. S., Bartholdy, J., Topsbe, N. Y., Langeler, B., and Topsee, H., in “Proceedings, 8th International Congress on Catalysis, Berlin, 1984, Vol. 2, p. 375. Dechema, Frankfurt-an-Main, 1984.

230

K. K . CHIANELLI, M . DAAGE, A N D M . J . LEDOUX

21. Sorensen, O . , Clausen, B. S., Candia, R., and Topsge, H., Appl. Curd. 13, 363 (1985). 22. Candia, R . , Clausen, B. S., and Topsge, H., J. Cutul. 77, 564 (1982). 23. Kasztelan, S . , Tolhouat, H., Grimblot, I., and Bonnelle, J . P., Appl. Curd. 13, 127 (1984). 24. Alstrup, I., Chorkendorf, I., Candia, R., Clausen, €3. S., and Topsee, H., J. C u r d 77, 397 (1982). 25. Phillips, R. W. and Fote, A. A , , 1. Curd. 41, 168 (1976). 26. Chianelli, R. R., f n t . Rev. Phys. Chem. 2, 127 (1982). 27. Seiver, R. L. and Chianelli, R. R., U.S. Patent 4,430,443 (1984). 28. Passaretti, J. D., Collins, R. C., Wold, A., Pecoraro. T. A , , and Chianelli, R. R., Muter. Res. Bull. 14, 1967 (1978). 29. Candia, R . , Clausen B. S., and Topsge H.,Bull. Soc. Chim. Belg. 90, 1225 (1981). 30. Ho, T. C., Young, A. C., Jacobson, A. J . , and Chianelli, R. R., U.S. Patent 4,591,429 (1986). 3 1 . Halbert, T. R . , Cohen, S. A., and Stiefel, E. I., Orgunomezullics 4 1689 (1985). 32. Angulo, M., Mauge, F., Duchet, J. C., and Lavalley, J . C., Bull. SOC.Chim. Belg. 96, 925 (1987). 33. Burmistrov, V . A , , Startsev, A. N., and Yermakov, Y. I., React. Kinet, Curd. Lett. 24, 365 (1984). 34. Vrinat M., Lacroix, M., Breysse, M., Bellalavi, A., Mosorin, L., and Roubin, M., in “Proceedings, 9th International Congress on Catalysis, Calgary, 1988” (M. J. Phillips and M. Ternan, Eds.), p. 88. Chem Institute of Canada, Ottawa, 1988. 35. Passaretti, J. D., Chianelli, R. R., Wold, A , , Dwight K., and Covino, J., J . Solid Stute Chem. 64, 365 (1986). 36. Chianelli, R. R., Ruppert, A. F., Behal, S . K., Kear, B. H., Wold, A., and Kershaw, R., J. Cutuly. 92, 56 (1985). 37. Halbert, T. R., Ho, T. C., Stiefel, E. I., Chianelli, R. R., and Daage, M., J. Cutul. 130, I I6 (1991). 38. Chianelli, R. R., Daage, M., Halbert, T. R., Ho, T. C., and Stiefel, E. I., in “Proceedings, Symposium on Chemistry of the Transition Metal Sulfides in Heterogeneous Catalysis,” ACS Boston Meeting, p. 227. Am. Chem. Soc., Washington, DC, 1990. 39. Pecoraro, T . A., and Chianelli, R. R., J. Curul. 67, 430 (1981). 40. Gullevic, P. J. Acta Crystullogr B . 30, I 1 1 (1974). 41. McCarty, K. F., Anderegg, J. W., and Schrader, G. L., J. Curul. 93, 375 (1985). 42. Rao, C. N . R., in “Proceedings, 3rd International Conference on Advanced Materials, Tokyo, 1993.” 43. Chianelli, R. R., Cutul. -Sci. Eng. 26, 361 (1984). 44. Chianelli, R. R., Prestridge, E. B., Pecoraro, T. A,, and DeNeufville, J . P., Science 203, 1105 (1979). 45. Tauster, S. J . , Pecoraro, T. A,, and Chianelli, R. R., J. Carol. 63, 515 (1980). 46. Roxlo, C. B., Daage, M., Ruppert, A. F., and Chianelli, R. R., J . Cutul. 100, 176 (1986). 47. Pratt, K. C., Sanders, I. V., and Christov, J. Catul. 124, 416 (1990). 48. Pratt, K . C., and Sanders, J. V., J. Catul. 67, 331 (1981). 4Y. Ho, T. C., Jacobson, A. J . , Chianelli, R. R., and Lund, C. R. F., J . Cutul. 138, 351 (1992). 50. Hayden T. F., and Dumesic, J. A , , J. Cutul. 103, 366 (1987). 51. Johnston, D. C., Silbernagel, B. G . , Daage, M., and Chianelli R. R . , in “Abstracts, 189th ACS National Meeting, Miami, 1985,” Paper PETR 76. 5 2 . Zmierczak, W., Muralidhar, G . , and Massoth, F. E., J . Cutul. 77, 432 (1982).

TRANSITION-METAL SULFIDE CATALYTIC MATERIALS

23 I

53. Bachelier, J., Duchet, J. C., and Cornet, D., Bull. SOC. Chim. Belg. 93, 743 (1984). 54. Tops~~e, N.-Y., and Topsoe, H., J. Catul. 84, 386 (1983). 55. Shuxian, Z., Hall W. K.,Ertl, G., and Knozinger, H., J. Curul. 100, 167 (1987). 56. Wright, C. J., Sampson, C., Fraser, D., Moyes, R. B., Wells, P. B., and Riekel, C., J. Chem. Soc. Furuday Trans, I 76, 1585 (1980). 57. Kelty S. P., Chianelli, R. R., Ren J., and Whangbo, M.-H., J . Am. Chem. Soc., in

press. 58. Salmerun, M., Somurjai, G. A., Wold. A,, Chianeili, R. R . , and Liang, K . S., Chem. Pbys. Lett. 90, 105 (1983). 59. Disko, M. M., Treacy, M. M. J., Rice, S. B., Chianelli, R. R., Gland, J. B., Halbert, T. H., and Ruppert, A. F. Ultrumicruscopy 23, 313 (1987). 60. Roxlo, C. B., Deckman, H. W., Gland, J., Cameron, S. D., and Chianelli, R. R., Science 235, 1629 (1987). 61. Mangnus, P., J., Ph.D. Dissertation, Univ. of Amsterdam, Netherlands, 1991. 62. Kasztelan, S . , Jalowiecki, L., Wambeke, A,, Grimblot, J., and Bonnelle, J. P., Bull. SOC. Chim. Belg. 96, 1003 (1987). 63. Horsley, J. A,, personal communication. 64. Ledoux,M. J., Michaux, O., Agostini, G., and Panissod, P., J., Cutul. 96, 189 (1985). 65. Clausen, B. S., Niemann, W., Zeuthen, X., and Topsge, H., Prepr. Div. Petrol. Chem. 35, 208 (1990). 66. Bouwers, S. P. A., Prins, R., Prepr. Div. Petrol. Chem. 35, 211 (1990). 67. Ledoux, M. J., Segura, Y . , Panissod, P., Prepr. Div. Petrol. Chem. 35, 217 (1990). 68. Daage, M., and Chianelli, R. R., unpublished results. 69. Ledoux, M. J., and Djellouli, B., J . Cutuf. 115, 580 (1989). 70. Eijibouts, S., DeBeer, V. H. J., and Prins, R., J. Cutul. 109, 217 (1988). 71. Lacroix, M., Boutarga, N., Guillard, C., Vrinat, M., and Breysse, M., J. Cutuf. 120, 473 (1989). 72. Ternan, M., J. Catal. 104, 256 (1986). 73. Bernard. J., Oudar, 3.. Barbouth, J., Margot, E., and Berthier, Surf. Sci. 88, L35, ( 1979). 74. Harris, S., and Chianelli, R. R., Chem. Phys. Lett. 101, 603 (1983). 75. Harris, S., and Chianelli, R. R., f. Curul. 86, 400 (1984). 76. Ledoux, M. J., Michaux O., Agostini, G., and Panissod, P., J. Curul. 102,275 (1986). 77. Vissers, J. P. R., Groot, C. K., Van Oers, E. M., De Beer, V. H. J., and Prins, R., Bull. SOC. Chim Belge. 93, 813 (1984). 78. Bonnelle, J. P., personal communication. 79. N@rskuv,J. K., Clausen, B. S., and Tfipsoe, H. Curd. Lett. 13, 1 (1992). 80. Chianelli, R. R., Pecoraro, T. A., Halbert, T. R., Pan, W.-H., and Stiefel, E. I., J. Cutul. 86, 226 (1984). 81. Sabatier, P., Ber. Deutsch Chem. G e s . 44, 2001 (1911). 82. Harris, S . , Chianelli, R. R., J. Cutal. 98, 17 (1986). 83. Tanaka, K. and Okuhara, T., in “Proceedings, 3rd International Conference on the Chemistry and Uses of Molybdenum, Climax Molybdenum,” p. 170 (1979). 84. Daage, M. and Chianelli, R. R., in “Preprints. 10th International Congress on Catalysis, 1993.” 85. Daage, M., and Chianelli, R. R., J . Cutul., in press (1994). 86. Chianelli, R. R., and Pecoraro, T. A., Exxon, U.S. Pat. 4288422 (1981). 87. Bussell, M. E., and Somorjai, G. A,, J. Cural. 106, 93 (1987). 88. Pratt, K. C., and Sanders, J. V., in “Proceedings, 7th International Corgress on Catalysis, Tokyo, 1980’ (T. Seiyama and K. Tanabe, Eds.), p. 1420. Elsevier, Amsterdan, 1981. 8Y. Vrinat, M. L., and De Mourgues, L., Appl. Catul. 5, 43 (1983).

232

K . R. CHlANELLl, M . DAAGE, A N D M . J . LEDOUX

90. Yacaman, M. J . , Chianelli, R. R., and Gland J . L., Paper E6.4, MRS Fall Meeting, Boston, MA, (1984). 91. Bouwens, S . N. A . M., Koningsberger, D. C., De Beer, V. H. J . , and Prins, R . , Bull. Soc. Chim. Belg. 96, 951 (1987). 92. Chiu, N. S . , Bauer, S. H., and Johnson, M. F. L., J . Caial. 98, 32 (1986). 93. Crajk, M. W. J . , Ph.D. Dissertation, T. U. Delft, 1992. Craje, M. W. I., De Beer, V. H. J . , and Van Der Kraan, A. M., Bull. Soc. Chim. Belg. 100, 9.53 (1991). 94. Duchet, J. C . , Van Ders, E. M . , De Beer, V. H. J . , and Prins, R., J . Cuiul. 80, 386 (1983). 95. Houwens, S . N. A. M . , Koningsberger, D. C., De Beer, V. H. J . , and Prins, R., Cutul. /err. 1, 55 (1988). 96. Ledoux, M. J . , Cuial, L m . 1, 429 (1988). 97. Ledoux, M. J . , Maire, G., Hantzer, S . , and Michaux, 0.. in “Proceedings, 9th International Congress on Catalysis, Calgary, 1988” (M. J. Phillips and M. Ternan, Eds.), p. 74. Chem Institute of Canada, Ottawa 1988. 98. Clausen B. S . , Niemann, W., Zeuthen, P., and Topsfie, H., P r e p . Div. Per. Am. Chem. SIX. 35, 208 ( 1990). 99. Louwers, S. P. A., and Prins, R., Prep-. D k . Pet. A m . Chem. Soc. 35, 21 I (1990). 100. Bachelier, J . , Tiliette, M. J . , Duchet, J. C., and Cornet, D., J . Catut. 76, 300 (1982). 101. Bachelier, J . , Duchet, J . C., and Cornet, D., J . Caial. 87, 283 (1984). 102. Zhuang, S . , Hall, W. K., Ertl, G., and Knozinger, H . , J . Cutul. 100, 167 (1986). /03. Riaz, U., Curnow, 0. J . , and Curtis, M. D., J . Am. Chem. Soc. 116,4357 (1994).

NOTEADDEDIN PROOF. During the final preparation of this manuscript the work of Riaz et al. (103) came to our attention. This work provides what may be the best model to date of a Co/Mo cluster which effects the HDS of sulfur bearing molecules.

ADVANCES IN CATALYSIS, VOLUME 40

Multicomponent Bismuth Molybdate Catalyst: A Highly Functionalized Catalyst System for the Selective Oxidation of Olefin YOSHIHIKO MORO-OKA AND WATARU UEDA* Research Laboratory of Resources Utilizution Tokyo Institute of Technology Yokohama 227, Japan

1.

Introduction

Catalytic oxidation and ammoxidation of lower olefins to produce a ,punsaturated aldehyde or nitrile are widely industrialized as the fundamental unit process of petrochemistry. Propylene is oxidized to acrolein, most of which is further oxidized to acrylic acid. Recently, the reaction was extended to isobutylene to form methacrylic acid via methacrolein. Ammoxidation of propylene to produce acrylonitrile has also grown into a worldwide industry. During the history of a half century from the first discovery of the reaction ( I ) and 35 years after the industrialization ( 2 - 4 , these catalytic reactions, so-called allylic oxidations of lower olefins (Table I), have been improved year by year. Drastic changes have been introduced to the catalyst composition and preparation as well as to the reaction process. As a result, the total yield of acrylic acid from propylene reaches more than 90% under industrial conditions and the single pass yield of acrylonitrile also exceeds 80% in the commercial plants. The practical catalysts employed in the commercial plants consist of complicated multicomponent metal oxide systems including bismuth molybdate or iron antimonate as the main component. These modern catalyst systems show much higher activity and selectivity * Present address: Department of Environmental Chemistry and Engineering, Tokyo Institute of Technology, 4259 Nagatsuta-cho, Midori-ku, Yokohama 227, JAPAN

233 Copyright 0 1994 by Academic Press, Inc. All rights of reproduction in any form reserved.

234

YOSHIHIKO MORO-OKA AND WATARU UEDA

A

O,, N H ,

?L

3'

MC- Bi-Mo-0 MC -Fe-Sh-0 0,.AcOH

A

Pd/C

. .

A

C

A

N

OAc

-

HIO. -AcOH

OH

('Multiconiponent bismuth molybdate catalyst Heteropoly compound.

' IndustrialiLed.

compared to those of the simple bismuth molybdate catalyst, which was used mainly in the early stage of the industry. The reaction mechanism of the allylic oxidation of propylene has been most extensively investigated on the simple catalyst systems, especially bismuth molybdate (5-8). The first step of the propylene oxidation is the same irrespective of the presence or absence of ammonia. The rate-determining abstraction of an a-methyl hydrogen to form a r-ally1 intermediate has been well established in both the oxidation and the ammoxidation from the isotope k ~1.82) and the isotopic distributions of oxygen and nitrogen effect ( k ~ / = insertion products from either ally1 or vinyl D-labeled propylenes (9-15). The subsequent steps in the conversion of the allylic intermediate to acrolein or acrylonitrile were mainly investigated by the SOH10 group, giving a detailed reaction mechanism (15-20). It was proposed that molybdenumdioxo-groups are the propylene chemisorption sites and are bridged to bismuth-oxygen groups that bring about the rate-determining a-hydrogen abstraction. Ammonia is probably activated as an imido (Mo=NH)species. Ammoxidation needs higher temperatures (400 480°C) than simple oxidation (320 360°C), which suggests that ammonia blocks propylene chemisorption sites irreversibly at lower temperatures and reversibly at elevated temperatures (18, 19). At lower propylene partial pressure, the ra-

-

-

235

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

tio of acrylonitrile formation to acrolein depends linearly on the ammonia to propylene ratio. This behavior was interpreted to mean that only one ammonia molecule is adsorbed at sites permitting nitrogen insertion in a given propylene molecule. At higher propylene pressure, the product ratio is a linear function of [NH3I2/[C3H6], corresponding to two ammonia molecules activated at the nitrogen-insertion site per acrylonitrile formed (20). Normal- and D-labeled allyl alcohol and allyl amine as probes for the reaction intermediates suggest. that .rr-ally1 species are reversibly converted to u-nitrogen-ally1 and oxygen-ally1 species, in both ammoxidation and oxidation. The higher increase in selectivity with increasing temperature in ammoxidation is due to the higher activation energy for the second hydrogen abstraction in the a-nitrogen-ally1 complex compared to that in the oxygen-ally1 species. This second hydrogen abstraction is the slow step in the conversion of the a-ally1 species to selective (amm)oxidation products (16-20). The overall mechanisms that form acrolein and acrylonitrile on bismuth molybdate are suggested as shown in Fig. 1 and 2 (19,20).Identification of active oxide ions in a bismuth molybdate was also investigated with "0, kinetic

I

0.

I FIG. I . Mechanism of selective oxidation of propylene to acrolein over bismuth molybdate catalyst by Burrington er al. (19).

236

YOSHlHIKO MORO-OKA A N D WATARU UEDA

FIG.2. Mechanism of selective amrnoxidation of propylene to acrylonitrile over bismuth molybdate catalyst by Burrington er al. ( l Y ) .

experiments, and in situ Raman spectroscopy ( 2 1 ) . The conclusions concerning active sites and intermediate species are supported by quantum chemical calculations (22) and homogeneous models (23-25). In spite of the accumulated mechanistic investigations, it still seems difficult to explain why multicomponent bismuth molybdate catalysts show much better performances in both the oxidation and the ammoxidation of propylene and isobutylene. The catalytic activity has been increased almost 100 times compared to the simple binary oxide catalysts to result in the lowering of the reaction temperatures 60 80°C. The selectivities to the partially oxidized products have been also improved remarkably, corresponding to the improvements of the catalyst composition and reaction conditions. The reaction mechanism shown in Figs. 1 and 2 have been partly examined on the multicomponent bismuth molybdate catalysts. However, there has been no evidence to suggest different mechanisms on the multicomponent bismuth molybdate catalysts. Our recent studies have been undertaken to clarify the working mechanism of the multicomponent bismuth molybdate systems. Apart from the re-

-

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

237

action pathways illustrated in Figs. 1 and 2, the oxidation of lower olefin is also characterized by a deep reduction/reoxidation cycle according to the Mars and van Krevelen mechanism (26). By using an I8O2 tracer, it was demonstrated that bulk lattice oxygen plays a major role in the reaction on both the simple bismuth molybdate (27-33) and the multicomponent catalyst systems (30, 32 -42). The role of additives in the multicomponent bismuth molybdate system is closely connected to that in the redox cycle (36, 38-43). The multicomponent bismuth molybdate catalyst is usually composed of many kinds of different phases of composite metal oxides (35, 41 -51, 77, 77u). In addition, each catalyst particle has no uniform bulk composition. Volatile oxides, such as bismuth molybdates and molybdenum trioxide, are located in the surface layer of the catalyst particle, and transition-metal molybdates are concentrated in the core (41 -45). The structure of the catalyst system mentioned above is also closely related to the catalyst performance (40-42, 46, 52 -55). We discuss these two issues in light of new information gained from the use of the '*O tracer technique and surface and structural analysis. It is concluded that the multicomponent bismuth molybdate catalyst is a complicated but highly functionalized catalyst system in which rapid bulk migration of oxide ion plays an important role in the catalysis. Migration of activated oxygen through the bulk diffusion makes it possible to supply active oxygen species to the reaction site, and effective collaboration between the oxygen activation site and the reaction site enhances both catalytic activity and selectivity extensively. The proposed working mechanism is also effective in the enhancement of the catalyst stability. This is discussed in the last part of the article.

II. Characterization and Working Mechanism of Multicomponent Bismuth Molybdate Catalysts

A . CLASSIFICATION OF MULTICOMPONENT BISMUTH MOLYBDATE CATALYSTS 1 . History of Allylic Oxidation Catalysts Catalytic oxidation of propylene to acrolein was first discovered by the Shell group in 1948 on CuzO catalyst ( I ) . Both oxidation and ammoxidation were industrialized by the epoch-making discovery of bismuth molybdate catalyst by SOH10 (2-4). The bismuth molybdate catalyst was first reported in the form of a heteropoly compound supported on S O ? ; , BirP,Mo12052/Si02 having Keggin structure but it was not the sole active species for the reactions. Several kinds of binary oxides between molybdenum trioxide and bismuth oxide have been known, as shown in the phase

238

YOSHIHIKO MORO-OKA AND WATARU UEDA

diagram ( 5 . 5 ~ )It. was soon found that any binary oxide between bismuth and molybdenum, Bi2(Mo04)3 (a-phase, distorted scheelite structure), Bi2M0209@-phase), or Bi2Mo06(7-phase, koechlinite structure) shows a considerable activity for either oxidation or ammoxidation of propylene and oxidative dehydrogenation of n-butene. Significant discussion on the comparison of the catalytic activity of each phase took place, but different conclusions were reported depending on the type of the reaction and its conditions (30, 56-59). Matsuura et af. reported that Bi2M0209is most active for the oxidative dehydrogenation of 1-butene to butadiene, but the surface layer of any catalyst having the composition between a-phase (Bi/Mo = f ) and y-phase (Bi/Mo = 2) shows the same Bi/Mo ratio as 1.0 by XPS analysis irrespective of the bulk composition (60). Synergy effects in the catalytic properties of bismuth molybdate were also proposed (61). In the 196Os, a number of binary oxides, including molybdenum, tellurium, and antimony, were found to be active for the reactions and some of them were actually used in commercial reactors. Typical commercial catalysts are Fe-Sb-0 by Nitto Chemical Ind. Co. (62-64) and U-Sb-0 by SOH10 (65-67), and the former is still industrially used for the ammoxidation of propylene after repeated improvements. Several investigations were reported for the iron-antimony (68-72) and antimony-uranium oxide catalysts (73-75), but more investigations were directed at the bismuth molybdate catalysts. The accumulated investigations for these simple binary oxide catalysts are summarized in the preceding reviews (5-8). The first important modification of bismuth molybdate catalyst was done by Knapsak (76). It was found that the replacement of half the amount of the bismuth component by iron in Bi9PlM012052 increased the catalytic activity for the ammoxidation of propylene extensively. It is noteworthy that the Knapsak catalyst, Fe45Bi45PI MoI2Ox,is neither a heteropoly compound nor a single composite oxide. The Bi-Fe-Mo-0 system consists of several different composite oxides including bismuth molybdate, iron molybdate, and Bi3FeMo2OI2(77, 77u). The improved catalytic performance depends mainly on the multiphase structure of the catalyst. Following the improvement by Knapsak, the research group of Nippon Kayaku found that the modification of bismuth molybdate by divalent transition-metal cations such as Co2+and Ni2+ with iron (Fe2+and Fe3+)also enhanced the catalytic activity and selectivity significantly in the oxidation of propylene to acrolein (78-80). This information was immediately spread over the world, and a great number of modified bismuth molybdate catalysts were claimed in patents. Within a few years, bismuth molybdate catalysts composed of multielements and multiphases drove out all simple or binary oxide catalysts but Fe-Sb-0 from the market of allylic oxidation and ammoxidation catalysts (Table 11). The improvement of bismuth molybdate catalyst by the addition of various kinds of metal elements has been continued after the establishment of

TABLE I1 Typical Reaction Conditions for the Oxidation and Ammoxidation of Propylene on the Simple and Multicomponent Bismuth Molybdate Catalyst" Catalyst Conditions C; conc.

Oz/Ci

w a w

Bi-Mo-0 (Bi-P-Mo-OiSiOZ) MC-Bi-Mo-0 a

Amrnoxidation

Oxidation

Temp. ("C)

Conv. Sel. to Yield of Contact time (sec)

(%)

6-10

1.2-1.8 410-460

1-7

57

5-8

1.2-1.8 290-350

1-7

95-99

AL, acrolein; A N acrylonitrile Including acrylic acid.

AL(%)b AL(%)b

71 92-%

Conv. Sel. to Yield of

Conditions C; conc. Oz/Ci NH&i

40

8-12

1-2

90-95

6-10

1.7-2

1-1.5

Temp. Contact ("C) time(sec) 450-500

3-10

1.05-1.2 430-480

3-8

(%)

AN(%) AN(%)

50-65 97-99 80-84

79-83

240

YOSHIHIKO MORO-OKA AND WATARU UEDA

TABLE 111 Crystul Structure of Divalent-Metal Molvbdute (81)

Metal: Ionic radius (A): ~~

Stable Structure

Ni 0.69

Co 0.72

Fe 0.74

(Mg) 0.66

Mn 0.80

Cd 0.97

Ca 0.99

Pb 1.20

~~

*

t--

a-CoMo04 type (monoclinic) M(I1): 6-coordination M" : 6-coordination

CaW0,t ype (scheelie, tetragonal) M(I1). 8 -coordination M b i : 4-coordination

a-MnMo04 type M(11):6-coordination Mb': 4-coordination

the concept of multicomponent bismuth molybdate catalyst in the early 1970s. The stream of improvement is classified in two ways depending on the different catalyst structure resulting from the different ionic radii of added metal cations, especially divalent and trivalent ones that form corresponding metal molybdates in the catalyst system. The crystal structure of metal molybdate of divalent cation changes according to its ionic radius as summarized in Table 111 (81). Metal cations having the smallest ionic radii such as Ni2+ (0.69 A), CO" (0.72 A) and Fe2+ (0.74 A), form cu-CoMo04type crystal (82), whereas Mn2+ (0.80 A), Mg2+ (0.66 A), and sometimes Co2' and Fe2+ assume the cu-MnMoO4-type structure (83). The scheelitetype structure (84) is most stable for the metal cations having larger ionic radii, i.e., Cd2+ (0.97 A), Ca2+ (0.99 A), and Pb2+ (1.20 A). Metal molybdates of trivalent cation having ionic radii larger than 0.9 A, such as Ce3+ (1.034 A) and La3+ (1.016 A), also assume the scheelite structure. One typical way to improve the catalyst system was directed at the multicomponent bismuth molybdate catalyst having scheelite structure (83, where metal cations other than molybdenum and bismuth usually have ionic radii larger than 0.9 A. It is important that the a-phase of bismuth molybdate has a distorted scheelite structure. Thus, metal molybdates of third and fourth metal elements having scheelite structure easily form mixed-metal scheelite crystals or solid solution with the a-phase of bismuth molybdates. Thus, the catalyst structure of the scheelite-type multicomponent bismuth molybdate is rather simple and composed of a single phase or double phases including many lattice vacancies. On the other hand, another type of multicomponent bismuth molybdate is composed mainly of the metal cation additives having ionic radii smaller than 0.8 A. Different from the scheelitetype multicomponent bismuth molybdates, the latter catalyst system is never composed of a simple phase but is made up of many kinds of different crys-

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

24 1

tals. These two types of multicomponent bismuth molybdate catalyst have been designed and improved by independent principles in the different streams of investigation. 2. Multicomponent Bismuth Molybdate Catalysts with Scheelite Structure The multicomponent bismuth molybdate catalyst with scheelite structure was first developed by the researchers of Du Pont (86-88), and the results are summarized by Sleight in a review (85). The term scheelite derives from the mineral CaW04 and its general composition is expressed as AMX4. Many different A(A+-A4+) and M(M2+-M8+) cations have been found in the scheelite structure and the A or the M site can be occupied by more than one cation. As a result, a number of metal oxides with scheelite or distorted scheelite structure have been reported. It is noteworthy that Bi3+ has a relatively larger ionic radius (0.96 A) and a-phase of bismuth molybdate, Bi2(M00~)~ also has a distorted scheelite structure. Bismuth cation dissolves readily in metal molybdate such as Ce2(Mo04)3(96) and PbMoO4 ( 9 9 , where divalent or trivalent metal cations have ionic radii larger than 0.9 A. Such divalent or trivalent cations also dissolve in the a-phase of bismuth molybdate and the system yields a single-phase material or two single-phase materials. The most important point in designing a scheelite-type catalyst is making lattice vacancies in the structure (89-96). Both molybdenum and bismuth are essential elements, and several types of scheelite having lattice vacancies were reported as excellent catalysts for the allylic oxidation. Crystal chemistry and catalytic performance of the Pbl--7xBi2xq5xMo04 system were extensively investigated by Aykan et al. (89). The concentration of cation vacancy is increased with increasing x in PbI-3,Bi2,+xMo04, and considerable amounts of distorted A cation vacancies are stabilized in the system. The oxidation and ammoxidation of propylene and the oxidative dehydrogenation of 1-butene were examined by changing x in the catalyst system. It was demonstrated that catalytic activity as well as selectivity increased in every reaction with the increase in the degree of substitution (see Figs. 3 and 4). This tendency was generally found in all scheelite-type multicomponent bismuth molybdate catalysts. The importance of the disordered vacancies was also pointed out by Sleight on the basis of the fact that Bi2(Mo04)3does not show the highest catalytic activity in spite of its higher concentration of bismuth as well as ordered cation vacancy. The role of the defect is suggested to be the stabilization of H+ in the rate-determining dissociative adsorption of propylene to form a .rr-ally1 intermediate (85). The increase of x in the system also accelerates the bulk migration of the lattice oxide ion, which is involved exclusively in the catalytic reaction. The

242

YOSHIHIKO MORO-OKA AND WATARU UEDA

70

60

50

25

40

20

30

15

20

10

10

5

RC3

RC4

.(

PbM004

X in

Pb,-3,Bi2x+x (Moo41

Biz(Mo04)3

FIG.3. Dependencies of catalytic activities for olefin oxidation on the catalyst composition of the B i , - , / , B i ~ ~ . M o Ostystem , (78).

importance of high mobility of oxide ion was demonstrated in the Bil-31x&3VI-xMox04system using an "0 tracer (38, 39). Investigations into the scheelite-type catalyst gave much valuable information on the reaction mechanisms of the allylic oxidations of olefin and catalyst design. However, in spite of their high specific activity and selectivity, catalyst systems with scheelite structure have disappeared from the cornmercia1 plants for the oxidation and ammoxidation of propylene. This may be attributable to their moderate catalytic activity owing to lower specific surface area compared to the multicomponent bismuth molybdate catalyst having multiphase structure. 5.

Multicomponent m m u m Morybaate Latarysts Multiphase Structure

witn

Bismuth molybdate catalysts activated by the metal cations with ionic radii smaller than 0.8 (Ni", Co2+,Fez+,Mg2+,and/or Mn2+ with Fe3+) are never composed of a single phase, such as scheelite structure, and many kinds of metal molybdate, including various phases of bismuth molybdate,

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

243

1

loo

FIG. 4. Dependencies of propylene and ammonia consumption and product yields on the composition of the Bii- ,/,BiZ,&,MoO4system at 440°C (90).

are found in their catalyst particles. In addition, some components are concentrated in the bulk of the particle and others are mainly located in the surface layers. The structure is seriously affected not only by its composition and atomic ratio of each component but also by the preparation method and conditions. The complexity mentioned above makes it very difficult to study the catalyst system systematically. However, the excellent catalytic performance of the multicomponent system is based mainly on the complicated catalyst structure. The catalyst system originated from the Knapsak catalyst (76) for the ammoxidation and catalysts found in Nippon Kayaku (78-80) for simple oxidation. A number of catalyst systems have been indicated in the patents in the past 25 years, and some of them are used practically in the industrial production. Strictly speaking, almost all catalyst systems may be designed and prepared on the same principle irrespective of their different compositions. The catalyst system is generally expressed as shown in Fig. 5. The first four elements are essential and consist of a fundamental structure of the catalyst system, and the other elements are added for the enhancement of the catalyst life and mechanical strength and minor improvement of the catalytic activity and selectivity.

244

YOSHIHIKO MORO-OKA AND WATARU UEDA Atomic %

MOW Bim

50-55 3-7

MU

30-35

Mrn

8-1 5

Fe, Cr, Al, especially Fe

MI X

small

K, Na, Cs, TI, ...

Co, Ni,

Fe, Mg, Mn;-

Sb, Nb, V, W, Te,

Y

p,

...

B

FIG. 5. Composition of multicomponent bismuth molybdate catalyst having multiphase structure.

Molybdenum comprises usually 50% or a little more of the total metallic elements. Most of molybdenum atoms form (M004)~- anion and make metal molybdates with other metallic elements. Sometimes a little more than the stoichiometric amount of molybdenum to form metal molybdate is included, forming free molybdenum trioxide. Since small amounts of molybdenum are sublimed continuously from the catalyst system under the working conditions, free molybdenum trioxide is important in supplying the molybdenum element to the active catalyst system, especially in the industrial catalyst system. In contrast, bismuth occupies a smaller proportion, forming bismuth molybdates for the active site of the reaction, and too much bismuth decreases catalytic activity somewhat. The roles of alkali metal and two other additives are very complicated. Unfortunately, few reports refer to these elements, except patents. In this article, discussion is directed only at the fundamental structure of the multicomponent bismuth molybdate catalyst system with multiphase in the following paragraphs. TABLE 1V Chururterizutiun of Typicul Tri- and Tetruromponent Bismuth Mulybdutr Catalysts (4 I ) Catalyst Bi2M0301Z COMO04 Mo12Bi I ColI 0, Mo12Bil CogNi30, Mo12BiICogMg,0, Mo12Bi CogFe, 0, Mo12BiosCo8Fe30, Mo12BioICosFe30, Mo12Co~Fe30, Mo 12Bi Mg8Fe30, Mo12BiI Co8Cr30, MoI2BilCogAl3Or

Phases detected by XRD

Surface area(mz/g)

a-Bi2(Mo04),

p -COM004 p -CoMo04, ~ x - B i ~ ( M o 0 ~ ) ~ p-CoMoO4, a-Bi2(Mo0&, y-Bi2MoOs p-CoMo04, a-CoMo04 ( ~ - B i ~ ( M o 0 ~ ) ~ P-CoMo04, a-Bi2(Mo0&, Fe2(Mo04)3,FeMo04 p-CoMo04, ~ x - B i ~ ( M o 0Fe2(Mo04)1 ~)~. p-CoMo04, Fe2(Mo04)3 p -CoMo04, Fez(Moo4), , FeMo04 p-CoMo04, a-Bi2(Mo04)>,Fe2(Mo04),, FeMo04 p -CoMo04, ( ~ - B i ~ ( M o 0Cr2(Mo04h ~)~, p-CoMo04, a-Bi2(Mo04)3,A12(Mo04)3

1.8 14.3

7.9 11.7 10.6 7.1 5.9 7.8 10.2 5.9 5.8 8.5

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

245

l3. CATALYTIC PERFORMANCE DEPENDING ON THE CATALYST COMPOSITION Choosing divalent and trivalent cations and determining the composition is the most important in designing the multicomponent bismuth molybdate catalyst system. Catalytic activities of typical tri- and tetracomponent bismuth molybdate catalysts having multiphase structure were reported for the oxidation of propylene to form acrolein (35,36, 40-43,97, 98). A typical example of the activity test is shown in Fig. 6. Summarizing the results shown in Fig. 6 and reported previously (30, 43, 4 4 , the following trends are generally found. 1 . The specific activity of pure bismuth molybdate, a-phase Bir(MoO& or y-phase Bi2Mo06, is fairly high. However, owing to its low surface area, the activity per unit weight of the catalyst is not so prominent. 2. The addition of divalent metal cation, M(I1) with ionic radius smaller than 0.8 A (Ni2+, Co2+, Fez+, Mg2+, Mn2+, etc.), to the pure bismuth molybdate increases the specific surface area of the catalyst system, but the specific activity of the tricomponent system, Mo-Bi-M(I1)-0 never exceeds that of pure bismuth molybdate. 3. The replacement of a part of the third component M(I1) in the tricomponent system by a trivalent metal cation, M(III), with ionic radius smaller

M o2Bll ~ C0eNi30x Mo12BilCoeMg30x

3 6 0 380 4 0 0 4 2 0 4 4 0 4 6 0 4 8 0 5 0 0

Reaction temp. ("C) FIG. 6 . Catalytic activity of multicomponent bismuth molybdate catalysts in the oxidation of propylene (41).

246

YOSHIHIKO MORO-OKA A N D WATARU UEDA

than 0.8 A increases the specific activity as well as the surface area of the catalyst system. Consequently, the tetracomponent system, Mo-Bi-M(I1)M(II1)-0, exhibits the highest catalytic activity in all tri- and tetracomponent bismuth molybdate systems. Although chromium and aluminum are effective to some extent, iron increases the specific activity remarkably while keeping excellent selectivity to the partially oxidized product. 4. On the contrary, replacement of M(I1) by another divalent metal cation, M(II), is ineffective for improving the specific activity of the MoBi-M(I1)-0 system. 5. Iron exhibits both divalent and trivalent cations. Fez+ and Fe3+ are interchangeable under the reaction conditions (97, 98) and consequently, iron is able to play both M(I1) and M(II1) as shown by the Knapsak catalyst (76). 6. Unlike Fe" for M(III), many kinds of divalent cations are suitable for M(II), and sometimes the mixtures of divalent cations are employed effectively in the Mo-Bi-M(I1)-M(II1)-0 systems. 7. When the added metal cation, M(I1) and/or M(III), has an ionic radius larger than 0.9 A and forms a molybdate with a scheelite structure, the catalyst system shows a performance quite different than that mentioned above for the scheelite-type catalysts. In conclusion, to make an excellent catalyst system, it is important to activate bismuth molybdate by both the divalent and trivalent metal cations with ionic radii smaller than 0.8 A at the same time. A part of the increasing activity of the Mo-Bi-M(I1)-M(II1)-0 system compared to the pure bismuth molybdate comes from the increase in surface area and the remains arise from the increase in specific activity. Semiquantitative evaluations of tri- and tetracomponent bismuth molybdates are listed in Table V in comparison with simple bismuth molybdate catalyst. TABLE V Semiquantitative Evaluations of Tri- and Tetracomponent Bismuth Molybdate Catalystsfor the Oxidation of Propylene Oxidation of propylene to acrolein Surface area

(m2/s)

Catalyst

2-3 1-2

BizMo3O,2 Bi2M0209

- 15

CoMoOi

10 446-

Mo-Bi-Co-O Mo-Bi-Co-Mg-0 Mo-Bi-Co-Fe-0 ~~~

8 10

14

Relative activity"

Specific activity"

1

I

1.3

1.2 0.01 0.2 0.1 5

0.1 0.9

0.8 20

~

"Based on the activity of the BizMo3012 catalyst.

Selectivity to acrolein(%) 90 90

20 70 70 95

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

247

C. STRUCTURAL ANALYSIS The structure of the multicomponent bismuth molybdate depends not only on the composition but also on the preparation method. Many different kinds of binary or ternary metal oxides are found in the catalyst system by X-ray analysis. Bi2(Mo04)3,a- and P-CoMo04, Fe2(Mo0&, FeMo04, and free Moo3 are detected mainly in the typical tetracomponent system, MoBi-Co-Fe-0 (42 -46).Bi2M03012is a major composite oxide between bismuth and molybdenum in most cases but Bi2Mo06 is also found in the catalyst system containing Ni2+ as the M(I1) component (46). Some ternary composite oxides, such as Bi3FeMo2OI2(93) or BiFeMo04 (99, loo), were reported depending on the composition and preparation method, although the selectivity to partially oxidized product on the ternary composite oxide is not so high (59). Surface analyses were investigated mainly by using XPS (Fig. 7). It was clearly indicated that many composite oxides found by XRD are located unhomogeneously in the catalyst particle. Molybdenum and bismuth are undoubtedly concentrated in the surface layer of the catalyst particle and divalent and trivalent metal cations are found in the bulk of the catalyst. As a result, it is clear that bismuth molybdates, especially its a-phase, is located on the surface of each particle, and metal molybdates of divalent and trivalent cations are situated in the bulk of the catalyst.

Fe rn 0

1

2

3

4

Concentration of 13i in the Catalyst ( 70)

FIG. 7 . Surface concentration of each metal element in the MolzBio-,Co8Fe30,catalyst determined by XPS analysis (41).

248

YOSHIHIKO MORO-OKA AND WATARU UEDA

A particle model for the multicomponent bismuth molybdate catalyst was first proposed by Wolfs ef uf. (44, 45) (Fig. 8). It was suggested that each catalyst particle consists of a core of M(II)Moo4 and M(II1)2(Mo04)3with a thin shell of bismuth molybdates. Several revised models have been proposed on the basis of the physicochemical analyses. Matsuura revised Wolfs’ model and suggested that the catalyst surface is covered not only by bismuth molybdates but also by free MOO? (46). On the other hand, Margolis and her co-workers proposed another model for a multicomponent bismuth molybdate with the composition MolzBiosMg6Feo with small amounts of alkali metal, where the catalyst contains three regularly crystallized phases, p-MgMo04, Bi2(Mo04)1, and Moo3 individually (53, 5 4 ) . Fez(Moo& is located at the boundaries between individual crystallites forming a defective garnet structure, and the redox between Fe” and Fe” increases the probability of electron transfer and active oxygen diffusion enhancing catalytic activity. Strictly speaking, it is difficult to conclude which model is most reasonable. However, summing up the results obtained by the surface analyses, it is sure at least that bismuth molybdates are concentrated on the surface of the catalyst particle. Our investigations for Mo-Bi-Co*+-Fe’+-O also support the conclusion mentioned above, and the core-shell structure proposed by Wolfs et af. may be essentially reasonable. However, since small amounts of divalent and trivalent metal cations are observed in the surface layers, the shell structure may be incompletely constructed. The epitaxial effect has been assumed on the condensation of bismuth molybdates on the divalent and trivalent metal molybdates on the basis of the fact that the y-phase of bismuth molybdate is mainly formed on NiMo04 but the a-phase is predominant on other divalent and trivalent metal molybdates ( 4 6 ) . The

2700 A FIG. 8 . Particle model of Mo-Bi-Co-Fe-0

by Wolfs et al. (44).

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

249

surface concentrations of cobalt and iron are not decreased with increasing concentration of bismuth from 0 to 10% in the catalyst system, Mo-Bi-CoFe-0, as shown in Fig. 7 (41). In spite of no direct evidence, it seems likely that the condensation of bismuth molybdates takes place exclusively on the specific faces of M(II)Mo04 and M(III)z(MoO~)~. D. REACTION KINETICS Oxidation of propylene to form acrolein depends on the first order of propylene and is independent of oxygen on multicomponent bismuth molybdate catalysts under the usual reaction conditions. The observed kinetics is the same with simple bismuth molybdates and suggests that the oxidation of propylene proceeds via the similar reaction scheme reported for simple molybdates, the slow step being the abstraction of allylic hydrogen (9-15, 19, 20). However, the reaction sometimes depends on the partial pressure of oxygen under lower temperature and lower oxygen pressure (41, 42). Arrhenius plots of the reaction show rather complicated apparent activation energies depending on the type of multicomponent bismuth molybdate catalysts and reaction temperature (41, 42). Different inclination of Arrhenius plots were obtained depending on the temperature range on a typical tetracomponent bismuth molybdate catalyst, MoIZBiICo8Fe3Or,as shown in Fig. 9. The apparent activation energy at temperatures higher than 390°C is 13 kcal/mol and surely lower than that of bismuth molybdate, 19 kcal/mol. However, a much higher value, 35 kcal/mol, was obtained at a lower tem-

-.-

Mo12Bi,Ml10,, M = -71

I

I

E

-e a 0

OgFe3

-9

m

I = 1.30

1.40

in(103

1.50

~-1)

FIG. 9. Arrhenius plots of the oxidation of propylene over various multicomponent bismuth molybdate catalysts.

250

YOSHIHIKO MORO-OKA AND WATARU UEDA

perature range. Generally speaking, inclination of Arrhenius plots obtained on the heterogeneous catalyst is influenced by many factors other than the real energy gap between the activated complex and the starting materials, for example, by variations of reaction kinetics, catalyst structure, and the number of active site with reaction temperature. In fact, an important component of M(II)Mo04 has a- and p-structure, depending on the temperature. @-Form is stable at higher temperature and more effective as the promoter for bismuth molybdates (97, 98). Sometimes the transformation between the a - and P-type was observed during the catalysis under the reaction conditions. At this stage, it seems to be difficult that we draw some conclusions on the working mechanism of multicomponent bismuth molybdate catalysts from the results of kinetics.

INVESTIGATIONS: INVOLVEMENTOF E. "0TRACER LATTICE OXIDE ION IN THE REACTION The involvement of lattice oxide ion in several kinds of oxide catalysts in the surface reaction has been implicitly perceived through the rate equation based on the Mars and van Krevelen mechanism (26). The phenomenon was clearly recognized in the early investigations by Keulks (27) and Wragg et al. (28, 29) using "0 tracer over simple bismuth molybdate catalyst. The involvement of lattice oxide ion in the surface reaction is inevitably accompanied by the bulk migration of oxide ion in the catalyst particle. A considerably high rate of oxide ion migration was observed in every phase of bismuth molybdate, and the order was reported as Bi2Mo06 > Bi~Mo209> Bi2Mo1OI2(43, 30). Lattice oxide ion is also involved in the reaction on the multicomponent bismuth molybdate catalyst. The migration of oxide ion in the multicomponent system is rather slow compared to that of the pure bismuth molybdates ( 4 3 , and the participation of lattice oxide ion in the reaction was never correlated with the excellent activity of the multicomponent catalyst system in the early investigations. Incorporation of lattice oxide ion into the oxidized products has been systematically investigated in the oxidation of propylene to acrolein using 99.1% " 0 2 tracer on a series of tri- and tetracomponent bismuth molybdate catalysts in our laboratory (35, 36, 40-42). CH2=CH -CH?

'SO2

Mu,,,Lomponent Ri-Mo-%

CH2=CH-CH1'0

or CHI=CH-CH180.

cataly5t

Slow heterogeneous exchange reactions between gaseous oxygen or oxidized products with catalyst oxides were observed but they were not quick to exert any serious effect on the evaluation of the participation of lattice oxide ion in the oxidation of propylene. I6O atom from the bulk of the catalyst was

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

25 1

found equally in both acrolein and carbon dioxide in the initial stage of the reaction in every catalyst tested, but the degree of the participation of lattice oxide ion differed significantly depending on the composition of the catalyst system. The most prominent difference was found between the results obtained on the tetracomponent systems, including both divalent and trivalent metal cations, Mo-Bi-M(1I)-M(II1)-0, and those obtained on other catalyst systems having no trivalent metal cations. In the case of the tricomponent system without trivalent metal cations, MoI~BiICollOx, the incorporation of I60?a;tice at the initial stage of the reaction terminates within a relatively short time and "0 concentration in the products reaches the same value as that of gaseous oxygen. The multicomponent catalysts include crystalline bismuth molybdates covering the surface of catalyst particles. Since the migration of oxide ion in any kind of bismuth molybdate is fairly rapid, it is easily estimated that most of the original oxide ion, I6O2-, in the bismuth molybdate phase of the catalyst would be recovered in the oxidation products after the reaction. Thus, the total amount of I6O incorporated into the oxidized products is compared with the amount of oxide ion in the bismuth molybdate phase of the catalyst. The arrow mark in Fig. 10 indicates the position where the total amount of I6O found in the reaction products equals the amount of oxide ion in the bismuth molybdates phase. The incorporation of lattice oxide ion practically termi-

B

m

Mo12BilCo110X o Mo12BilCoBFe30X0

5 j /

r

or0

o;

o;

IlO

160

C o n s u m e d Oxygen (layer) FIG. 10. concentration in acrolein as a function of consumed oxygen in the oxidation of propylene with ' * 0 2 and I6O catalyst. The arrows show the point where consumption of '60~at,lcs reaches the same amount of oxide ions as the Bi2(Mo0& phase in the catalyst system (40).

252

YOSHIHIKO MOKO-OKA AND WATARU UEDA

nates at the arrow point as shown in Fig. 9 for the MoI2BilCoIIOA. All multicomponent bismuth molybdate catalysts without trivalent metal cations gave the same results as summarized in Fig. 1 la, and it is clear that only the lattice oxide ion in the bismuth molybdate phase is active and can participate in the oxidation of propylene. On the other hand, different results were obtained on the tetracomponent catalyst system having trivalent metal cations. The "0 concencatalyst is plotted against tration of acrolein produced on a Mo12BiICoRFe~Ox normalized reaction time in Fig. 10 and compared with the results obtained on the Mo12BilCol,0x catalyst. The incorporation of I6O does not terminate within the reaction time and continues beyond the arrow mark on the Mo12BiICo8Fe30, catalyst including Fe3'. Almost the same time course of the 160incorporation was obtained on a MoL2C08Fe40x catalyst having no bismuth component, whereas the selectivity to acrolein decreased remark-

a

Mo,,Bi,Co,Ni,O, I

Mo12BilCo8Mg30x

I

Mo,,BilNi,,Ox Mo,zBi,Mg,,Ox Mo,,Bi Col7OX Mo, zBi,Co, Mo,Bi,Co,O, 0

5

17

Amount of "0 (104g-atom)

b

Mo,,Bi

Mg,Fe30,

Mo,,BilCo,A130x Mol,BilCo8Cr30, M o , ,BiO.,C osFe30x Mo,,Bio,,CoeFe,Ox

xzxII2

Mo12BilCosFe30,'

___1

Mo,,Bi,Co,Fe,O, 0

5

17

Amount of "0 (10-4g-atorn)

FIG.11. Comparison of the amount of i60j& incorporated into the oxidation products over various multicomponent bismuth molybdate catalysts. Open columns, amount of whole i60~aII,ce in the catalyst; shaded columns, amount of i60ul,,, in the BizMo30i2phase; solid columns, total amount of I6O incorporated into the oxidation products; *, oxygen conversion was 80% and the others were 60%. (a) Tricomponent system, Mo-Bi-M(I1)-0, and tetracomponent system, Mo-Bi-M(I1)-M'(I1)-) without M(II1). (b) Tetracomponent system, MoBi-M(I1)-M(II1)-0 (41).

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

25 3

ably in the absence of bismuth. The total amounts of 160incorporated into the reaction products on the tetracomponent catalyst systems with trivalent metal cations are also compared with the amounts of the lattice oxide ions in the corresponding catalyst systems in Fig. 1 lb. It is evident that the lattice oxide ions not only in the bismuth molybdate pliase but also in other transition-metal molybdates are active in the tetracomponent system, including trivalent-metal cations, and can take part in the reaction. Interestingly, the replacement of a part of Co2+ in the MolzBilColrOx by another divalent cation, NiZi or Mg”, does not improve the degree of participation of lattice oxide ion at all and only the lattice oxide ion in the bismuth molybdate phase is active in the Mo12BiICoxNilO, or Mo12RilCo8Mg30, catalyst. It is noteworthy that the situation mentioned above corresponds exactly to the catalytic activity of the tri- and tetracomponent bismuth molybdate catalysts.

F. OXIDATION OF PROPYLENE ON THE MODEL CATALYSTS A number of articles report that both divalent- and trivalent-metal molybdates employed as the activator for the bismuth molybdate catalyst show very poor activity in forming partially oxidized products of olefin. The structural analyses and activity investigations suggest that the divalent- and trivalent-metal molybdates play a kind of support for the active component, bismuth molybdate. Thus, Bi2(Mo04)3catalysts supported on P-CoMo04 or P - C O ~ ~ ~ were ~ ~prepared F ~ ~as/ the~ model ~ M catalyst ~ ~ ~( 4 2 , 5 2 ) . Transmission electron microscopy (TEM) and X-ray emission analysis elucidated that the BL(MoO~)~ crystal grew with an increasing amount of loaded bismuth and molybdenum after covering the surface of the support with a thin layer of Bi2(Mo04)3.Most of the iron existed as Fe2+ but a part of it was found by ESR to be Fe3+.Recently, the solid solution between CoMo04 and FeMo04 was investigated by Ponceblanc ef al. (101, 102). Two continuous solid solutions exist between the two polymorphic forms of CoMo04 and FeMo04 ( a , p). Fe3+ was also shown to be present in the solid solutions where it substitutes for Fe2+ and Co2+.Catalytic activity and selectivity of the model catalysts for the oxidation of propylene to acrolein were reported by varying the loading amount of bismuth molybdate as shown in Figs. 12 and 13 (52). The drastic promotion effect of the iron component is again observed in the results. The support molybdate, CoMo04 or Co11/1~Fe1/12M00x, never shows higher activity in forming acrolein without bismuth and gives mainly carbon oxides under the reaction conditions. In the case of BL(Mo04)3 supported on Co11~12Fel/l~MoOx, small amounts of Bi2(M~04)3 depress the formation of carbon oxides extensively, and the catalytic activity to form acrolein increases remarkably with an increasing When the loading amount of B ~ * ( M O Oreaches ~ ) ~ the amount of Bi2(M~Oq)3.

254

YOSHlHlKO MORO-OKA AND WATARU UEDA

0 0.01

0.1 "0.3

0.05

Catalyst / Support (molar ratio) FIG. 12. The catalytic activity forming acrolein per unit weight of the supported bismuth molybdate catalysts (52). (0) Biz Mo30IdCoMo04 ; (0)B2 Mo3012/CoI 2 FellI MOO,.

amount that can cover the surface of C o I 1 / ~ 2 F e ~ / ~ ~by M omonolayer, Ox the specific activity of the catalyst system is almost comparable to that of the pure BL(MOO~)~. Since the support has a larger specific surface area than pure BL(MoO~)~, the Biz(MoO~)~/CO I 12Fe zMoO, system shows much higher activity per unit weight of catalyst than the pure Bi*(MoO&. The specific activity of the Bi2(Mo04)3/C~I l/12Fe1/12M00x increases further with increasing the loading amount of Bi2(Mo04)3, but too much loading decreases somewhat the catalytic activity. On the other hand, the activity of

0 0.01

0.05

I

I+

0.1 0.3

Catalyst / Support (molar ratio)

FIG.13. The catalytic activity forming acrolein per unit surface area of the supported bismuth molybdate catalysts (52). (0) Bi2Mo3OI2/CoMoO4;(a)Bi2Mo3OI2/CoI I/ 12Fe I/ I 2 Moo,.

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

0 0.01

0.05

255

0.1 0 3

Catalyst / Support (molar ratio) FIG. 14. The selectivity forming acrolein in the oxidation of urouvlene over the supported bismuth molybdate- catalysts (52). (0) BizMo3012;CoMo04; (0) Bi?Mol 0 d C o I I , I2 Fel,12M00x.

B~~(MoO~)~/COM increases O O ~ more slowly with the loading amount of Bi2(M00~)~ and never exceeds that of the pure Bi2(Mo04)3.The improvement of the selectivity to acrolein is also poor as shown in Fig. 14. The same synergy effect between bismuth molybdates and mixed iron and cobalt molybdates on the mechanical mixture of both particles was reported by Millet et al. (98). However, it was also found that the surface of mixed iron and cobalt molybdate particle was changed during catalysis and a thin layer of bismuth molybdate was formed on the surface of mixed iron and cobalt molybdates after the reaction. It is doubtful that pure mechanical mixture shows the synergy effect for propylene oxidation, and it seems likely that propylene was mainly oxidized on the thin layer of bismuth molybdates formed on the mixed iron and cobalt molybdate in the experiment reported by Millet et al. (98). The prominent difference in the catalytic activity depending on the type of the support molybdate is reflected in the apparent activation energy (Fig. 15). The apparent activation energy of the Bi2(Mo04)3/CoMo04system varies considerably with the loading amount of Bi2(Mo04)3and finally coincides with that of the BL(MoO~)~ at higher loadings. The facts mentioned above show clearly that the oxidation of propylene proceeds on pure Bi2(Mo04)3in the Bi2(Mo04)dCoMo04system and CoMo04 serves as a simple support for the active component. On the other hand, the apparent system is little afactivation energy of the Bi2(Mo04)3/Co11/~2Fe1~12M00x fected by the loading amount of the bismuth molybdate and is lower than that of the pure Bi2(Mo04)3by 3 kcal/mol. This implies that the oxidation of propylene mainly proceeds on the thin layer of bismuth molybdate con-

25 6

YOSHIHIKO MOKO-OKA AND WATAKU UEDA

1 EY15$. a,

.

;"

_y

0.1

0.3

Catalyst / Support (molar ratio) FIG. 15. The apparent activation energy for the oxidation of propylene over the supported bismuth molybdate catalysts (52). (0) Bi2Mo3Ol2/CoMoO4; ( 0 ) Biz Mo,Olz/Col ,/ , Z Fel/12Mo0,.

densed on the Coll/lZFe1/12M00x surface and the Co11/12Fe1/12M00A is not a simple support to enlarge the surface area of Bi2(Mo04)3and gives some influence on the active site of the propylene oxidation. Investigation using the tracer technique was also reported for the model catalysts mentioned above (42, 52). Three kinds of oxygen, i.e., oxide ion in BiZ(Mo04)1roxide ion in Co11/12Fe1/~2M001, and molecular oxygen, are available as the source of the oxygen atom incorporated into the oxidized products in the oxidation of propylene on the Bi2(Mo04)4CoI1 / 1 2 F e l / l ~ M o 0 x catalyst with molecular oxygen. To clarify what kind of oxygen is first involved in the reaction, the reaction was carried out in the reaction system where each oxygen source was labeled by I8O. The typical result obtained is on the I*O-labeled catalyst, Bi2(Mol604)4Col1/12Fe1/12M0180~, with 1602 shown in Fig. 16, where "0 concentration in acrolein is followed by the reaction time. The I*O concentration of acrolein starts at a very low level at the initial stage of the reaction. It increases with the reaction time, reaches a maximum, and then decreases. The same experiments were done by changing the labeled sources of oxygen. The results suggest that the incorporation of the lattice oxide ion into the reaction products is not attributable to the simple oxygen exchange reaction between surface active oxygen species and bulk oxide ion in the catalyst. It is also clear that the oxide ion in the Bi2(Mo0& phase was involved in the reaction first and then the lattice oxide ion in Co11~12Fe1/12M00x was involved and then in gaseous oxygen. The observed results are reasonably interpreted by the assumption that the activation of oxygen and the reaction of propylene take place on the different sites, and the oxygen species activated on the separate site from the re-

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

0

5

10

257

15

Reaction time (min) FIG. 16. “0 concentration of acrolein with the reaction time in the oxidation of propylene on ‘80-labeledcatalyst, Bi2Mo~6012/Col l / l ~ F e 1 ~ ~ 2 M 0(40). 180x

action site migrates by the bulk diffusion through Co11/12Fe1/12M00.rand Biz(MoO&. The principle is based on a kind of spillover and is illustrated simply in Fig. 17. Delmon and his co-workers (104-109) also suggested in their “remote control theory” that oxidation and reduction occur at different sites. They classified metal oxides into two groups, of which one is more effective for the activation of oxygen and another is suitable for the selective conversion of olefin to oxygenated products. It was proposed that an excellent catalyst system may be designed by combining two groups where activated oxygen species is transferred from one oxide to another. Although no definite evidence has been reported for the component and structure of the oxygen activation site of the multicomponent bismuth molybdate catalyst system, it may

A

FIG. 17. Scheme of the active oxygen migration through bulk diffusion in Bi2Mo?O12/ Coll/lZFel~lZMoOr.

25 8

YOSHIHIKO MORO-OKA AND WATARU UEDA

be possible that cobalt or iron cation serves to activate molecular oxygen because cobalt or iron oxides are more active for the oxygen equilibrium reaction (110) or the complete oxidation of hydrocarbons (111, 112) than bismuth or molybdenum oxide. As suggested in the tracer experiments, the activation sites of propylene and oxygen seem to be spatially separated, active oxygen diffusing from one site to another. It is pointed out that this migration is possible only when the catalyst system contains both divalent cation and trivalent cation, especially Fey+, simultaneously. The phenomenon is closely associated with the en2 works. hancement of the catalytic activity as demonstrated in the 180tracer Matsuura suggested that part of the divalent cation in M(II)Mo04 may be replaced by M(III), making cation vacancies when both M(I1) and M(II1) have almost the same ionic radii (46). This was confirmed by Ponceblanc et al. in the CoMo04-FeMo04 solid solution system with some Fe3+ (101, 102): M(II)MoO,

+ 2~ M(II1)

-

M(II)I.~,M(III)~~~M + o3~O M(I1) ~

where 4 is the cation vacancy. The replacement of M(II1) in M(III)2(Mo04)3 by M(I1) may also possibly form anion vacancies. The importance of doping M(II1) in M(II)Mo04 was clearly demonstrated by Sleight et al. in the scheelite-type multicomponent oxide catalysts (85-91). Acceleration of the diffusion of lattice oxide ion by lattice vacancies was also reported in the defective scheelite-type oxide, Bil-x/3~3/xV1-xMox04. The principle of the doping effect established for the scheelite-type oxide is also applicsble to the multicomponent bismuth molybdate catalyst, Mo-Bi-M(I1)-M(II1)-0. Accordingly, the migration of oxygen and the participation of lattice oxide ion in the molybdates of M(I1) and M(II1) into the reaction are attributable to the doping effect of M(II1) in M(II)Mo04 and M(I1) in M(III)2(Mo04)3 and to the resulting acceleration of oxide ion migration by lattice vacancies. The principle seems to explain reasonably most results reported for the catalytic performance of multicomponent bismuth molybdate catalysts. G . WORKING MECHANISM OF MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

Significant progress has been made in explaining the prominent catalytic performances of the multicomponent bismuth molybdate catalysts. Summing up the progress, we have seen the following: (1) Molybdenum and bismuth are indispensable elements, forming the aphase of bismuth molybdate, which is located mainly on the surface of the catalyst particle and constitutes the reaction site of propylene. (2) Both divalent and trivalent cations having ionic radii smaller than 0.8 A are essential for the activation of bismuth molybdate. They are con-

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

259

centrated in the bulk of the catalyst particle, forming the corresponding metal molybdates, M(II)Mo04 and M(III)2(Mo04)3.Co’+, Ni’+, Mn’+ , Mg2+,Fe2+,or their mixture is frequently employed as a divalent cation and Fe3+ as a trivalent one, but the employed cation can be replaced by another one without decreasing higher activity and selectivity. (3) Bulk oxide ion is preferentially involved in the reaction. The activation sites of propylene and oxygen are spatially separated, with active oxygen migrating from one site to another through the bulk diffusion of oxide ion. It is highly probable that lattice vacancies formed by the doping of M(II1) in M(II)Moo4 and M(I1) in M(III)2(Mo04)3accelerate the bulk diffusion of oxide ion. (4)The lower activation energy of the multicomponent bismuth molybdate catalyst suggests that some structural or electronic modification of the active component, Bi2(Mo0&, is given by M(I1) and M(II1) molybdates, which also contribute to the enlargement of the surface area of bismuth molybdate. The catalytic performance and working mechanism of multicomponent bismuth molybdate catalysts are closely associated with their structures. As mentioned in the preceding sections, several structural and functional models have already been proposed by different groups. Matsuura proposed an active site model as shown in Fig. 18, suggesting that Bi2M0209is surrounded by Moo3 as an active species (46). Lattice vacancies prepared by the doping of Fe3+ in M(II)Mo04 play an important role in increasing reducibility of the active component oxide (43). The structural and functional models proposed by Margolis and co-workers (53)and Krylov et al. (54)

FIG. 18.

Active surface model for Mo-Bi-Co-Fe-0

by Matsuura (46).

260

YOSHIHIKO MORO-OKA AND WATARU UEDA

\ FIG. 19. Active structure model for Mo-Bi-Co-Fe-0

by Krylov er

(11.

(54).

are shown in Fig. 19. These researchers also proposed cooperation between different phases in the M 0 , 2 B i " . ~ C o ~ F eand ~ , ~suggested ~ 0 ~ ~ that the oxygen is activated on p-FeMo04 stabilized by CoMo04 and the vacancy-containing F ~ * ( M o O ~phase ) ~ transfers active oxygen to the active center on BL(MoO~)~, which is responsible for adsorption and activation of olefin. These models may explain the experimental results to some extent, but it seems to be difficult to cover all results accumulated. Free Moo3 is often observed on the surface of the catalyst particle depending on the ratio of molybdenum to other elements but the role of Moo3 in the catalytic cycle is still obscure. Ferrous and ferric molybdates, FeMo04 and Fez(Mo04h,play an essential role in activating molecular oxygen and supplying active oxygen species to bismuth molybdate in Krylov's model. However, the presence of iron in the catalyst system is also effective when it is doped completely in the M(II)Mo04 matrix without forming any FeMo04 or Fe2(Mo04)3crystals (41, 42, 97, 98). In 1985, Matsuura and co-workers proposed another model for the multicomponent bismuth molybdate catalyst on the basis of our tracer experiments and it is shown in Fig. 20 (113). The active species of bismuth molybdate, Bi2Mo3OI2,is located on the divalent-metal molybdate, M(II)Mo04, crystal in the form of a thin layer. It is probably grown on the specific faces of the M(II)Mo04 crystal, i.e., the (010) and/or (100) face of CoMo04. Parts of M(I1) sites in M(II)Moo4 are replaced by Fe3+,forming lattice vacancies. Molecular oxygen is activated on the different face of the M(II)Mo04crystal, for example, (001) of CoMo04, and active oxygen is transported through the bulk diffusion of oxide ion to the thin layer of

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

26 1

C o M o 0 4 liiiving lattice vacancies by the rcplacenicnt of a part of co2+ 11y ~ c : ’ + . FIG.

20. Active structure model for Mo-Bi-Co-Fe-0

revised by Matsuura et al. (113).

Bi2Mo3OI2. The proposed model seems to be applicable for all kinds of multicomponent bismuth molybdate catalysts except those having a scheelite structure thus far. The detailed working mechanism of the multicomponent bismuth molybdate catalyst still leaves scope for discussion. The reaction kinetics of the propylene oxidation on the multicomponent bismuth molybdate catalyst depends on the first order of propylene and is independent of oxygen, the slow step being the abstraction of an allylic hydrogen in a manner similar to that of pure bismuth molybdate. Thus, the excellent catalytic activity of multicomponent bismuth molybdate catalysts should be attributed to either acceleration of the slow step of the reaction, i.e., the abstraction of the allylic hydrogen of propylene, or the increasing number of active reaction sites of the catalyst. The latter is partly true because the surface area of bismuth molybdate is enlarged by supporting the divalent- or trivalent-metal molybdate. However, specific activity is also enhanced by the synergy effect between bismuth molybdates and divalent- with trivalent-metal molybdates. The observed lower apparent activation energy of the multicomponent bismuth molybdate (41, 4 2 , 52) may suggest some electronic and structural modifications of bismuth molybdates by the sublayer of mixed divalent- and

262

YOSHIHIKO MORO-OKA A N D WATARU UEDA -2 I

"

0

20

40

60

Fe atom 70

FIG. 21. Variation of the logarithm of electrical conductivity of mixed iron and cobalt molybdates as a function of total iron content (97).

trivalent-metal molybdates. Ponceblanc et al. (97) suggested that electronic conductivity of mixed iron and cobalt molybdates was increased remarkably with increasing iron content (Fig. 21). This might stabilize the dissociatively adsorbed state of propylene on the multicomponent bismuth molybdate catalyst and lower the activation energy of the slow step (97). Although the detailed stabilization scheme of the dissociative adsorbed state of propylene has not been known, this is surely a possible case. Epitaxial growth of the thin layer of bismuth molybdate on the M(II)Mo04 surface is also expected and may lower the activation energy of the abstraction of allylic hydrogen. The situation, however, seems to be not so simple. FeMo04 includes inevitably a small amount of Fe3+ and shows higher electrical conductivity than mixed cobalt and iron molybdate (97), but the mixture of FeMo04 with bismuth molybdates shows rather moderate synergy effects for the propylene oxidation compared with cobalt and iron mixed molybdates (98). The apparent activation energy obtained on the multicomponent bismuth molybdate catalyst is also not simple. Sometimes an Arrhenius plot of the propylene oxidation shows two different inclinations depending on the temperature. At lower reaction temperatures, the apparent activation energy of Mo-Bi-Co-Fe-0 is rather high compared to that of Bi2Mo3012. It has been well accepted that the apparent activation energy on the heterogeneous catalyst is influenced by many factors and does not reflect the real energy gap between an activated complex and starting materials. In addition, the catalytic activity of another kind of multicomponent bismuth molybdate increases with the concentration of dopant while keeping the activation energy constant (38). The direct correlation between the lower apparent activation

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

263

energy and the function of the active site for propylene oxidation is easily understandable, whereas the possibility that the number of active reaction sites is increased cannot be rejected. At this stage, it is not clear whether almost all molybdenum and bismuth atoms located on the surface may be involved in the reaction or only some. Volta et al. (114-116) demonstrated that the catalytic activity of Moo3 for olefin oxidation differs significantly depending on each face of the single crystal. Of further interest, Hayakawa et al. (117, 118) found that the activity order of metal oxides for propylene oxidation changes remarkably when active oxygen is supplied by the supporting system. The catalytic activities of Moo3, Bi203, and bismuth molybdates for propylene oxidation were observed in the order Bi2M0209(/3)> Bi2Mo0,(y) 9- Bi2M03012(a)> MOO,

- Bi203

under the usual catalytic conditions. Whereas a different order Moo3 > Bi2M0209(/3) > BiZMo3012(a) > Bi2Mo06(y)9- Bi203

was found under the conditions where oxide catalysts were supported on an oxygen transport membrane, Zr02 doped by Y 2 0 3 , and the reaction was carried out without gaseous oxygen by supplying oxide ion through the membrane (Fig. 22). It is well known that MooBis not catalytically active for olefin oxidation without promoters. Surprisingly, the highest activity was observed on Moos under the conditions mentioned above. The results by Hayakawa et al. (117, 118) suggest the possibility that even free MOO?and less active parts of bismuth molybdates may take part in the reaction in the

Bi / (Mo + B i )

FIG. 22. Catalytic activities of Moo3, BizOj, and bismuth molybdates for the oxidation of propylene to acrolein. (0)Under usual catalytic conditions. (0) In the absence of gaseous oxygen. Active oxygen is supplied through an oxygen transfer membrane. Data were taken from (IIK).

264

YOSHIHIKO MORO-OKA AND WATARU UEDA

multicomponent bismuth molybdate catalyst. If this is the case, the observed activation energy of the multicomponent bismuth molybdate catalyst including both M(I1) and M(II1) may also be changed while enhancing catalytic activity. This possibility was strongly suggested by the authors on the basis of tracer experiments. When a thin layer of bismuth molybdate is supported on the core of divalent-metal molybdate having satisfactory concentration of lattice vacancies by the doping of trivalent cation, active oxygen species required for the reaction would be easily supplied to the active site for oxygen through the bulk diffusion of oxide ion. This is the case for Mo-Bi-M(I1)M(II1)-0, where lattice oxide ions in the bismuth molybdate phase as well as in other transition-metal molybdates can participate in the reaction. On the contrary, the migration of active oxygen to the reaction site of propylene is not satisfactory when bismuth molybdate is supported on the inert carrier such as pure CoMo04 or silica gel, and most parts of the surface sites do not work well under the reaction conditions. It may be suggested that this is the main reason why the tricomponent system, Mo-Bi-M(I1)-0, and tetracomponent system, Mo-Bi-M(I1)-M’(I1)-0 including no trivalent cation, do not show higher catalytic activity in spite of their relatively high surface area. The concept proposed by us is pictured simply in Fig. 23. The level of water represents the chemical potential of active oxygen involved in the oxidation of propylene, and the vessels connected on the tank are involved in two kinds of active sites that activate molecular oxygen to atomic species and oxidize propylene to acrolein. If active sites expressed by vessels are isolated from each other, each site must do everything by itself to convert propylene to acrolein. This situation is less convenient than the preparation of the active catalyst system. When active species of oxygen can migrate rapidly through the bulk diffusion of oxide ion as shown in Fig. 23, equal-

Fm. 23. A water tank model for the concept of rnulticomponent bismuth molybdate catalyst.

MULTJCOMPONENTBISMUTH MOLYBDATE CATALYST

265

ization of the chemical potential of active oxygen is attained any time on every reaction site for propylene during the catalysis. In this case, collaboration between different kinds of active site proceeds quite efficiently and one site activates molecular oxygen exclusively while another site mainly consumes it for the reaction. At this stage, it is still difficult to determine whether the conclusion is appropriate for the fundamental part of the multicomponent bismuth molybdate catalyst. Unfortunately, we have no available information on the number of active reaction sites on the catalyst system. In the heterogeneous catalysis, apparent activation energy does not necessarily correspond to the real energy barrier of the elementary slow step of the reaction. Multicomponent bismuth molybdate catalyst has been established industrially, whereas only parts of the fundamental structure and working mechanism have been elucidated. In addition, important roles of alkali metals and other additives such as lanthanides remain unknown. Apparently, further investigations should be done to clarify the complete working mechanism of the multicomponent bismuth molybdate catalyst. 111. Stability of the Multicomponent Bismuth Molybdate Catalyst Depending on the Bulk Diffusion of Oxide ton

In the preceding section, we explained that the bulk diffusion of oxide ion plays an important role in the enhancement of the catalytic activity of the multicomponent bismuth molybdate systems. Here, another important role of the oxide ion migration in increasing the stability of the catalyst system is introduced. The catalyst system Bil-x/3VI-xMox04,having a scheelite structure, was first reported by Bierlein and Sleight in 1975 (119). They found that V5+ion in BiV04 can be replaced by Mo6+ until x = 0.45, forming cation vacancies without changing its original structure. As often reported in the systematic studies of scheelite catalysts by Sleight et a1 (85, 89-94), catalytic activity forming both acrolein and acrylonitrile in the oxidation or ammoxidation of propylene increases drastically with increasing concentration of cation vacancies in Bil-,/,V,-,Mo,O4. The selectivity of the reaction is also improved with the degree of substitution to some extent. The kinetic parameters of propylene oxidation are independent of the catalyst composition in spite of the drastic change in the specific activity. The reaction is first order with respect to propylene and independent of oxygen pressure, and the apparent activation energy is 19 0.5 kcal/mol for every catalyst (38). The mobility of lattice oxide ion under the working conditions was examined in the oxidation of propylene with I8O2 for a series of

*

266

YOSHIHIKO MORO-OKA AND WATARU UEDA

Bil-x/3V I - r M ~ x 0(38). 4 The participation of the lattice oxide ion into the reaction is quite prominent in this catalyst system, but the degree of the participation depends on the catalyst composition. The most prominent particiwhere pation of the lattice oxide ion was observed on Bio the complete mixing of all lattice oxide ions in the catalyst with surface active species of oxygen occurred. Experimentally, 93% of 1601atrse in the virginal oxide catalyst was found in the reaction products. Recently, effectiveness of lattice oxide ion for the selective conversion of olefin was further confirmed by Hayakawa et al. (120) using Bil-x/3VI-*Mox04catalyst supported on the oxygen transport membrane. The mobility of lattice oxide ion was evaluated by the complete mixing volume of lattice oxide ion according to the method proposed by Keulks and co-workers (30). By assuming the I8Oconcentration of lattice oxide ions in a dilution volume of catalyst is equal to that in oxidized products in the reaction at a given time, the complete mixing volume can be calculated from the results of I8O2 tracer experiments by the following:

"0%in products = 1 - exp(-A/V). I8O% in gaseous oxygen In this equation ( I ) , A is the total amount of molecular oxygen consumed in the reaction until a given time and V is the complete mixing volume. Equation (1) is rewritten as ln[l-

1

'*O% in products = -A/V. ''0%in gaseous oxygen

The complete mixing volume calculated by Eq. (2) is a hypothetical volume of catalyst oxide, where lattice oxide ions are under complete equilibrium with surface active species of oxygen. Accordingly, the larger the volume, the higher the mobility of oxide ion migration. The fraction of the complete mixing volume to the total mount of lattice oxide ions in the catalyst system is presented in Fig. 24 for each catalyst. The mobility of oxide ion increases with increasing x until x reaches 0.45, and this variation is closely compatible with the specific activity of the catalyst. Sleight (85) suggested that the increase of the specific activity with increasing x came from the higher proton acceptability by increasing cation vacancies. However, as mentioned above, the results may also be attributable to the increase of active sites. The same value of activation energy in spite of different catalytic activity and increasing mobility of lattice oxide ion with x are rather favorable for the latter interpretation. Clear experimental evidence of the enhancement of catalyst life by the rapid migration of lattice oxide ion was demonstrated by our investigations (39). The stability of the tricomponent catalyst system Bil-x/3VI -.MoxOs was examined by the reduction and reoxidation of the samples using XRD

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

BiV04

267

Bi2MogO 12

X in B ~ ~ . x / ~ V ~ - X M O X O ~ FIG. 24. Comparison of the specific activity of the Bi,-3,xV,-,Mo,04 with the fraction of lattice oxide ions involved in the oxidation of propylene to the total lattice oxide ions in the catalyst (38).

and temperature programmed reoxidation (TPRO) technique. Catalyst XRD patterns at the fresh state and after the reduction to 6% are presented in Fig. 25. As shown in the figure, free bismuth metal was detected after the reduction in the catalysts having lower molybdenum contents, especially in BiV04. The relative XRD intensity of liberated bismuth metal decreased with increasing content of molybdenum and no free bismuth metal was detected in Bio.ssVo.ssMoo.4504, which showed the highest mobility of lattice oxide ion. On the contrary, all peaks of fresh catalysts having higher molybdenum concentrations were broadened after the reduction. It is shown clearly that the reduction was localized in the vicinity of the surface but the degree was quite deep in the catalysts having lower molybdenum concentrations. On the contrary, the reduction spread to the whole oxide, but the catalysts kept their original structures in spite of some distortion in the case of higher concentrations of molybdenum. After the TPRO up to 450°C, all catalysts recovered their original patterns recorded before the reduction, but several unidentified new peaks were detected in the case of BiV04. When the redox cycle was repeated many times, unidentified peaks in BiV04 grew and finally the catalyst was decomposed completely. The structural deformation of the catalysts having higher molybdenum concentrations would be milder owing to its higher mobility of oxide ion and the catalysts are scarcely changed after the repeated redox cycles.

268

YOSHIHIKO MORO-OKA A N D WATARU UEDA

0

Metallic Bi Unidentified peak

O!!

20

40

60

28 (deg)

FIG. 25. XRD patterns of the Bi,-,,,V,-,Mo,O, (39). ( I ) At fresh state; (2) after reduction in hydrogen (6%);(3) after TPRO up to 230°C; (4) after TPRO up to 450°C; ( 5 ) after four reduction (450°C. in HI) and reoxidation (TPRO, up to 450°C) cycles. (a) BiVOa; (b) 91" 97 V, I Moo (c) Bio 93 Vo 79 Moo I O4; (4 Bio 9,VO73 MOO2 7 0 4 ; (e) Bi,, 8 ~ V ~ s r M o ~ 4 5 0 4 .

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

BiV04

Bi0.85V0.55M00.4504

(Rigid and unstable)

(Elastic and stable)

Surface of catalyst is reduced severely

Catalyst is reduced homogeneously

0

269

FIG.26. Concept for the stabilization of multicomponent metal oxide catalyst.

The catalytic oxidation reaction generally obeys the Mars and van Krevelen mechanism based on the redox cycle. In addition, oxidation catalysts are often used in the fluidized reactor, where catalysts are exposed to both oxidizing and reducing atmosphere. Thus, the rapid migration of oxide ion is quite important in preventing the deactivation resulting from the decomposition of the active component. This concept is illustrated in Fig. 26. The structural deformation of the catalyst generally results in a change of catalytic properties. The decomposition of active components, the formation of different metal oxides, the growth of larger crystals to decrease surface area, the transformation from amorphous state to crystal, the liberation of active elements, and the loss of a particular element due to vaporization result in the deactivation of the catalyst. To diminish the deactivation, it is important to suppress the structural deformation. The migration of oxide ion is mainly accelerated by lattice vacancies. Knowledge of how to introduce such vacancies into the catalyst by the substitution of part of a component metal cation for another one having a different valence is widely realized for the preparation of excellent industrial oxidation catalysts. IV.

Conclusion

Some progress has been made in explaining the splendid catalytic performance of multicomponent bismuth molybdates that are used widely for the industrial oxidations and ammoxidations of lower olefin. We have seen that the catalytic activity and selectivity are greatly enhanced by the multifunctionalization of the catalyst systems. Many functions newly introduced are

270

YOSHlHlKO MORO-OKA AND WATARU UEDA

deeply associated with lattice vacancies formed by the introductions of third and fourth elements. The design of the excellent oxidation catalyst depends seriously on the method of selecting these additives by considering their valencies, electronegativities, and ionic radii. The rapid migration of oxide ion and electron transfer are also important in enhancing the catalyst stability. Thus, the appropriate introduction of additional elements into the catalyst system makes it more flexible and durable under the working conditions. Even if the description of the fundamental structure and working mechanism of multicomponent bismuth molybdate catalysts is accepted, there still remains much to be investigated for the complete understanding of their catalytic performances. The important role of alkali or rare-earth metal in the catalyst system is still unknown. Industrial researchers are still making a big effort to increase the yield of target product 1 or 2% more. The multicomponent bismuth molybdate catalysts have provided a number of important principles with which to design excellent oxidation catalysts. These principles may be applicable further to many other catalytic oxidations. The structural and mechanistic investigations of practical catalyst systems will be more and more important in solving expected accumulated problems.

REFERENCES 1 . Hearne, G . W., and Adams, M., U.S. Patent 2,451,485, Shell Development Co. (1 948).

2 . Idol, J . D., U.S. Patent 2,904,580, Standard Oil Co., Ohio (1959). 3. Veatch, F., Callahan, J . L., Milberger, F. C., and Formane, R. W., in “Proceedings, 2nd International Congress on Catalysis, Paris, 1960,” Vol. 2, p. 2647. Technip, Paris, 1961. 4 . Veatch, F., Callahan, J. L., Idol, J. D. Jr., and Millberger, E. C., Hydrocarbon Process, Pet. Refiner 41, I87 (1962). 5 . Sampson, R. J., and Schooter, D., in “Oxidation and Combustion Reviews” (C. F. H. Tipper, Ed. ) Vol. 1, p. 223. Elsevier, Amsterdam, 1965. 6. Voge, H. H., and Adams, C. R., Adv. Cutul. 17, 151 (1967). 7 . Keulks, G. W., Krenzke, L. D., and Notterman, T. M., Adv. Cutul. 27, 183 (1978). 8. Grasselli, R. K., and Burrington, J. D., Adv. Curd 30, 133 (1981). 9. Adams, C. R., and Jennings, T. J., J. Cutul. 2, 63 (1963). 10. Adams, C. R., and Jennings, T. J., J. Cutul. 3, 549 (1964). 1 1 . Voge. H. H., Wagner, C. D., and Stevenson, D. P., J. Cutul. 2, 58 (1963). 12. McCain, C. C., Cough, G., and Godin, G. W., Nature (London) 198, 989 (1963). 13. Sixma, F. L. J., Duynstee, E. F. J . , and Hennekens, J. L. J. P . , Rec. Truv. Chim. 82, 901 (1963). 14. Sachtler, W. M. H., Rec. Truv. Chim. 82, 243 (1963). 15. Dozono, T., Thomas, D. W., and Wise, H., J. Chem. Soc., Faraduy Truns I , 69, 620 (1973). 16. Burrington, J. D., and Grasselli, R. K., J. Catal. 59, 79 (1979). 17. Burrington, J . D., Kartisek, C. T., and Grasselli, R. K., J. Cutul. 63, 235 (1980).

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST 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.

27 1

Burrington, J. D., Kartisek, C. T., and Grasselli, R. K., J. Cutul. 75, 225 (1982). Burrington, J. D., Kartisek, C. T., and Grasselli, R. K., J. Cutul. 81, 489 (1983). Burrington, J. D., Kartisek, C. T., and Grasselli, R. K., J. Cutal. 87, 363 (1984). Glaeser, L. C., Brazdil, J. F., Hazel, M. A,, Menieic, M., and Grasselli, R. K., J . Chem. Soc., Faruduy Trans. I , 81, 2903 (1985). Allison, J. N., Goddard W. A , , 111, in “Solid State Chemistry in Catalysis” (R. K. Grasselli and J. F. Brazdil, Eds. ) ACS Symp. Ser., p. 279. Am. Chem. SOC.,Washingon, DC, 1985. Chan, D. M. T., Fultz, W. C., Nugent, A. W., Roe, D. C., and Tulip, T. H., J. Am. Chem. Soc. 107, 251 (1985). Chan, D. M. T., Fultz, W. C., Nugent, W. A., Roe, D. C., and Tulip, T. H., in “Molecular Approaches to Surface Catalysis” J. M. Basset, Ed. ) p. 187. Kluwer, Budapest, 1988. Maatta, E. A., and Du, Y. H., J . Am. Chem. Soc. 110, 8249 (1988). Mars, P . , and van Krevelen, D. W., Chem. Eng. Sci. (Spec. Suppl. ) 3 , 41 (1954). Keulks, G. W . ,J. Curul. 19, 232 (1970). Wragg, R. D., Ashmore, P. G., and Hockey, J. A,, J. Cutul. 22, 49 (1971). Wragg, R. D., Ashmore, P. G., and Hockey, J. A., J. Cutul. 28, 337 (1973). Howfs, E. V., Monnier, J. R., and Keulks, G. W., J . Cutul. 57, 331 (1979). Keulks, G. W., and Krenzke, L. D., in “Proceedings, 6th International Congress on Catalysis. London, 1976’ (G. C. Bond, P. B. Wells, and F. C. Tompkins, Eds. ) Vol. 2, p. 806. The Chemical Society, London, 1977. Krenzke, L. D., and Keulks, G. W., J . Catal. 61, 316 (1980). Ueda, W., Moro-oka, Y., and Ikawa, T., J. Chem. SOC., Farady Trans I , 78, 495 (1982). Moro-oka, Y., Ueda, W., Tanaka, S., and Ikawa, T., in “Proceedings, 7th International Congress on Catalysis, Tokyo, 1980’ Eds. ), (T. Seiyama, and K. Tanabe,) Part B, p. 1086. Kodansha, Tokyo/Elsevier, Amsterdam, 1981. Ueda, W., Moro-oka, Y., and Ikawa, T., J . Cutal. 70, 409 (1981). Ueda, W., Moro-oka, Y., Ikawa, T., and Matsuura, I., Chem. Lett., 1365 (1982). Ueda, W., Moro-oka, Y., and Ikawa, T., J. Catul. 88, 214 (1984). Udea, W., Asakawa, K . , Chen C. -L., Moro-oka, Y., and Ikawa, T., J. Catal. 101,360 (1986). Ueda, W., Chen, C. -L., Asakawa, K., Moro-oka, Y., and Ikawa, T., J. Cural. 101, 369 (1986). Moro-oka, Y., He, D.-H., and Ueda, W., Stud. Surf. Sci., Catul. 67, 57 (1991). Udea, W., He, D.-H., Miura, N., and Moro-oka, Y., submitted for publication. He, D.-H., Ueda, W., and Moro-oka, Y., submitted for publication. Brazdil, J. F., Suresh, D. D., and Grasselli, R. K., J . Cutal. 66, 347 (1980). Wolfs, M. W. J., and Batist, Ph. A , , J. Cutul. 32, 25 (1974). Matsuura, I., and Wolfs, M. W. J., J. Cutul. 37, 174 (1975). Matsuura, I., in “Proceedings, 7th International Congress on Catalysis, Tokyo, 1980’ (T. Seiyama and K . Tanabe, Eds. ) Part B, p. 1009. Kodansha, Tokyo/Elsevier, Amsterdam, 1981. Prasada Rao, T. S. R., and Menon, P. G., J . Cutal. 51, 64 (1978). Ooij, W. J. V., and Muizebelt, W. J., in “Proceedings, International Vac. Congress and 3rd International Conference on Solid Surfaces, Vienna,” p. 839, 1977. Uemura, S., Ohdan, K., and Asada, H., in “Proceedings, 5th Soviet-Japan Catalysis Seminar,” p. 60,1979. Krylov, 0. V., Kiner. Katal. 25, 955 (1984). Prasada Rao, T. S. R., and Krishnamurthy, K. R., J. Cutul. 95, 209 (1985).

272

YOSHIHIKO MORO-OKA AND WATARU UEDA

52. He, D. -H., Ueda, W., and Moro-oka, Y., Curul. Left. 12, 35 (1992). 53. Vrob’eva, G . A., Rozentuller, B. V., Maksimov, Yu. V., Kutyrev, M. Yu., and MargOliS, L. y d . , J . CUtUl. 71, 405 (1981). 54. Krylov, 0. V., Maksimov, Yu. V., and Margolis, L. Ya., J. Cutul. 95, 293 (1985). 55. Matsuura, I., Stud. Surf. Sci. Cutul. 44, 1I 1 (1989). 55u. Egashira, M., Matsuo, K., Kagawa, T., Seiyama, T., J. Curul. 58,409 (1979). 56. Aykan, K., I. Cutul. 12, 281 ( 1968). 57. Agagusinova, M., Poladvo, F. M., Vassel, M., and Adzhamov, K. Yu., Azerb. Khirn. Zh. 4,27 ( 1 977). 58. Batist, Ph. A., Lippens, B. C., and Schuit, G. C. A,, .I Cutul. . 5, 55 (1966). 59. Daniels, C., and Keulks, G. W., J. Curul. 29, 475 (1973). 60. Matsuura, I., Schut, R., and Hirakawa, K., J. Curul. 63, 152 (1980). 61. Carson, D., Coudurier, G., Forissier, M., Vedrine, J. C . , Laarif, A , , and Theobald, F., J . Chem. Soc., Furuduy Trans. I , 78, 1921 (1983). 62. Miyamoto, K . , Aritomi, L., Shimizu, N., and Yoshino, T., Japanese Kokai JP

38/19111, Nitto Chem. Ind. Co. (1963). 63. Yoshino, Y., Saito, S., Ishikura, J., Sasaki, T., and Sofugawa, N., Japanese Kokai JP

46/2804, Nitto Chem. lnd. Co. (1971). 64. Sasaki, T., Nakamura, Y., Moritani, K., Morii, A., and Saito, S., Shokubui 14, 191P

(1972). 65. Callahan, J. L., and Gertisser, B., U.S. Patent 3,198,750, Standard Oil Co., Ohio

(1965). 66. Callahan, J. L., and Gertisser, B., U.S. Patent 3,308,151, Standard Oil Co., Ohio

(1967). 67. Callahan, J . L., and Knipple, W. R., U.S. Patent 3,341,151, Standard Oil Co., Ohio ( 1967).

Fattore, V, Fuhrman, Z. A , , Manara, G., and Notari, B., J. Curd. 37, 223 (1975). Sala, F., and Trifiro, F., J . Cuful. 41, 1 (1976). Aso, I., Furukawa, S., Yamazoe, N., and Seiyama, T., 1.Cutul. 64, 29 (1980). Yamazoe, N., Aso, I., Anamoto, T., and Seiyama, T., in “Proceedings, 7th International Congress on Catalysis, Tokyo” (T. Seiyama and K. Tanabe, Eds. ), Part B, p. 1239. Kodansha, Tokyo/Elsevier, Amsterdam, 1981. 72. Burriesci, N., Garbassi, F., Petrera, M., and Petrini, G., J. Chem. SOC. Furuduy Trans. I , 78, 817 (1982). 73. Grasselli, R . K., and Callahan, J . L., J., Cutul. 14, 93 (1969). 74. Grasselli, R . K., Suresh, D. D., and Knox, K., 1. Cuful. 18, 356 (1970). 75. Grasselli, R. K., and Suresh, D. D., J . Cutul. 25, 273 (1972). 76. Dutch Patent 1,125,901, Knapsack A. -G. (1962); German Patent 1,243,175 (1967). 77. Nasa, S., and Shimizu, S . , J. Curul. 104, 164 (1987). 77u. Annenkova, B., Alkhazov, T. G., and Belen’kii, M. S., Kinet. Kurd 10, 1076 (1969). 78. Takenaka, S., and Yamaguchi, G., Japanese Kokai JP 43/002,324, Nippon Kayaku Co. (1968). 79. Takenaka, S., and Yamaguchi, G.. Japanese Kokai JP 44/005,855, Nippon Kayaku Co. (1969). 80. Takenaka, S., and Yamaguchi, G., Japanese Kokai JP431006.264, Nippon Kayaku Co. (1969). 81. Matsuura, I., Shokuhui (Cutu/ysr) 21, 409 (1979). 82. Smith, G. W., and Ibers, J. A , , Actu Crystullogr. 19, 269 (1965). 83. Abrahams, S. C., and Reddy, J. M., J. Chem. Phys. 43, 2533 (1965). 84. Leciejewicz, J . , Z. Kristullogr. 121, 158 (1965). 85. Sleight, A. W., in “Advanced Materials in Catalysis” (J. J. Barton and R. L. Garten, Ed.) p. 181. Academic Press, New York, 1977.

68. 69. 70. 71.

MULTICOMPONENT BISMUTH MOLYBDATE CATALYST

273

86. Aykan, K., Rogers, D. B., and Sleight, A. W., U.S. Patent 3,806,470, E. I. Du Pont de Nemours Co. ( 1974). 87. Aykan, K., Rogers, D. B., and Sleight, A. W., U.S. Patent 3,843,553, E. I. Du Pont de Nemours Co. (1975). 88. Aykan, K., Rogers, D. B., and Sleight, A. W., US. Patent 3,843,554, E. I. Du Pont de Nemours Co. (1975). 89. Aykan, K., Sleight, A. W., and Rogers, D. B., J. Curd 29, 185 (1973) 90. Aykan, K., Halvorson, D., Sleight, A. W., and Rogers D. B., J . Curd. 35,401 (1974). 91. Linn, W. J., and Seight, A. W., J. Curd 41, 134 (1976). 92. Sleight, A. W., and Linn, W. J., Ann. N. Y. Acad. Sci. 272, 22 (1976). 93. Sleight, A. W., and Jeitschko, W., Muter. Res. Bult. 9, 951 (1974). 94. Sleight, A. W., Aykan, K., and Rogers, D. B., J. Solid Stare Chem. 13, 231 (1975). 95. Brazdil, J. F., Gleaser, L. C., and Grasselli, R. K., J . Cutul. 81, 142 (1983). 96. Brazdil, J. F., and Grasselli, R. K., J. Card. 79, 104 (1983). 97. Ponceblanc, H., Millet, J. M. M., Coudurier, G., Herrmann, J. M., and Verdine, J. C . , J . Cutul. 142, 373 (1993). 98. Millet, J. M. M., Ponceblanc, H., Coudurier, G., Herrmann, J. M., and Vedrine, J. C., J . Cutul. 142, 381 (1993). 99. Notermann, T., Keulks, G. W., Skliarov, A , , Maximov, Yu, Margolis, L. Ya, and Krylov, 0. V., J. Curd. 39, 286 (1975). 100. LoJacono, M., Notermann, T., and Keulks, G. W., J. Curd. 40, 19 (1975). 101. Ponceblanc, H., Millet, J. M. M., Coudurier, G., Legendre, O., and Vedrine, V., J. Phys. Chem. 96, 9462 (1992). 102. Ponceblanc, H., Millet, J . M. M., Thomas, G., Herrmann, J. M., and Vedrine, V., J. Phys. Chem. 96, 9466 (1992). 103. Daniels, C., and Keulks, G. W., J . , Curd 29, 475 (1973). 104. Delmon, B., and Ruiz, P., Cutul. Today 1, 1 (1987). 105. Ruiz, P., Zhou, M., Remy, T., Machej, F., Aoun, F., and Delmon, B., Cutul Today, 1, 181 (1981). 106. Ruiz, P., and Delmon, B., Cutul. Today 3, 199 (1988). 107. Qui. F. Y., and Weng, L. T., Ruiz, P., and Delmon, B., Appl. Curd. 47, 115 (1989). 108. Qui, F. Y., Weng, L. T., Sham, E., Ruiz, P., and Delmon, B., Appl. Cutul. 51, 235 (1989). 109. Weng, L. T., Ma, S. Y., Ruiz, P., Delmon, B., 1.Mol. Curd. 61, 99 (1990). 110. Boreskov, G. K., Adv. Curd. 15, 285 (1966). 111. Moro-oka, Y., and Ozaki, A., J. Cutul. 5, 116 (1966). 112. Moro-oka, Y., Morikawa, Y., and Ozaki, A., J . Cutul. 7, 23 (1967). 113. Hashiba, H., Kanesaka, M., and Matsuura, I., Shokubui (Catalyst) 27, 434 (1985). 114. Volta, J. C . , Forissien, M., Theobald, F., and Pharn, T. P., Discuss. Furuduy SOC. 72, 275 (1981). 115. Volta J. C . , Synthetic Muter. 4, 319 (1982). 116. Volta, J. C., Tatibouet, J. M., Phichitkul, C., and Germain, J. E., in “Proceedings, 8th International Congress on Catalysis, Berlin, 1984,” Vol. 4, p. 451, Dechema, Frankfurtam-Main, 1984. 117. Hayakawa, Y., Tsunoda, T., Orita, H., Kameyama, T., Takahashi, H., Takehira, K., and Fukuda, K., J. Chem. SOC. Chem. Commun., 961 (1986). 118. Hayakawa, Y., Tsunoda, T., Orita, H., Kameyama, T., Takahashi, H., Fukuda, K., and Takehira, K., J. Chem. SOC. Chem. Commun., 870 (1987). 119. Bierlein, J. D., and Sleight, A. W., Solid State Commun. 16, 69 (1975). 120. Hayakawa, Y., Tsunoda, T., Orita, H., Kameyama, T., Takahashi, H., Ueda, M., Fukuda, K., and Takehira, K., J. Chem SOC. Chem Commun. 1593 (1988).

This Page Intentionally Left Blank

Index

A

coatings containing cobalt and iron oxides, 103-105 electrocatalysis by Raney-nickel coatings, 109 in situ disposition of coatings, 106- I09 catalyst-coated titanium electrodes, acid solutions, 109- 1 11 technical processes, 103 PbOz, 155-156 oxygen evolution, 109-1 10 react ion, history, 97 solid oxide fuel cells, 149-150 Aryl-aryl bond, breakage, 4.5 Asphaltenes, depolymerization, 63 Azide radicals, physisorption, 162

Acetylpyridine, cathodic pinacolisation, 165 Aging, catalyst, 134-135 Alcohols, oxidation, 155 Aldehydes, cathodic pinacolisation, 165-166 Alkaline electrolyte, oxygen reduction, silver cathodes, 135-136 Alkaline fuel cells, Raney-nickel anodes, 136-1 39 Alkanes, light, oxidative dehydrogenation, 1-35 butane, 14-24 catalytic classification, 25 1960s studies, 4-5 cyclohexane, 14-24 ethane, 5-8 gas-phase reaction with oxygen, 2-4 generalized reaction scheme, 24-25 molecular hydrogen as by-product, 1-2 propane, 8-14 relationship between metal-oxygen bond strength and selectivity, 26-28 relative rates of reaction, 17 thermodynamics, 1-2 water as product 2, 2 Alloy catalysts, anodic methanol oxidation at, 141-142 Alumina support, role, 177- 178 Arnines, oxidation, 155 Anodes dimensionally stable cell design using, 100-101 chlordte and hypochlori te production, 102- I03 chlorine-evolving, lifetime, 101- 102 oxygen evolving, 103- I 1 I advanced alkaline water electrolysis

B Bisarene sulfides, anodic oxidation, 162-165 Bismuth molybdate catalyst, multicomponent , 233 -270 active oxygen migration through bulk diffusion, 256-257 active structure model, 259-261 active surface model, 259 allylic oxidation catalysts, history, 237-241 apparent activation energy, 255-256, 261-262 catalytic activities, 263 classification, 237-244 electrical conductivity as function of iron content, 261-262 lattice oxide ion involvement, 250-253 metal addition, 238, 240 metal molybdatc crystal structure, 240

275

276

INDEX

Bismuth molybdate catalyst (continued) model propylene oxidation, 253-258 with multiphase structure, 242-244 performance and composition, 245 -246 propylene ammoxidation to acrylonitrile, 235-236 oxidation to acrolein, 235 reaction kinetics, 249 - 250 with scheelite structure, 241-242 specific activity, 266-267 stability, 265-269 structural analysis, 247-249 surface concentration of metal elements, 247 tr i - and tet racornponent characterization, 244 semiquantitative evaluations, 246 water tank model, 264-265 working mechanism, 258-265 XRD patterns, 267-268 Butane, oxidative dehydrogenation, 14-24 CO reduction, 22-23 heat of reoxidation, 19-22 orthovanadate catalysts, 18-19 over V/SiOz catalysts, 23-24 product selectivity, 15- 16 reaction rates on vanadium oxide catalysts, 16-17 Butler-Volmer equation, 89

C Carbon catalyst deactivation, 70-7 I role in transition-metal sulfide catalysts, 22 1 Carbonaceous substances, catalyst deactivation, 70-71 Carbon electrodes, PTFE-bonded, morphology and structure, 133-134 Carbon monoxide, reduction by, 22-23 Catalysts fine powder, preparation, 49 highly dispersed, preparation, 49 recovery and reuse, 72-73 supported, preparation, 48-49 transformation, catalyst deactivation, 72 Cathodes platinum and platinum metal, membrane water electrolyzers, I22

Raney-nickel coated, 113-1 15 Ru02-coated, 120-121 silver, 135-136 solid oxide fuel cells, 150-151 C-C bond, energies, 3 C -H bond breaking alkane oxidation, 28, 30 butane and cyclohexane activation, 16 energies, 3 formation in alkane oxidation, 30 Na-Chelates, catalysts for cathodic oxygen reduction, 128- 130 Chemical catalysis, 91 -93 porous electrocatalyst particles, 93-95 Chemisorption, transition-metal sulfide catalysts, 200-20 1 hydrogen, 202 Chlorate, production, dimensionally stable anodes, 102- 103 Chloride, catalysts, SO Chlorine anodic evolution, at RuOz anodes, 97-98 dimensionally stable anodes evolving, 101-102 Coal deashing pretreatments, 75-76 Hirsch model of ranks, 41-42 ionic depolymerization, 77 pretreatment , 75-77 reactivity, 43 solvent-refined, hydrotreatment, 64- 65 structure, 41-43 Coal liquefaction, 39-80 Bergius process, 39 catalysis chemical functions, 50 materials, 46, 48 preparation, 48-49 types, 50-51 catalyst deactivation mechanism, 70-72 in secondary stage, 74-75 catalytic upgrading of crude coal liquids, 62- 69 catalyst design, 68-69 interface of primary and secondary stages, 62-63 reactions and roles of catalysts, 63-65 two- or three-step upgrading efficiency, 6.5 - 68

277

INDEX design durable catalyst for primary stage, 73-74 multistage with catalysts, 79 recovery and regeneration of catalyst for primary stage, 72-73 dissolution catalyst role, 56 hydrogen donor and solvent functions, 51-56 dissolution, depolymerization, and retrogressive reactions, 45-47 donor and catalyst, stepwise application, 59-61 history and status, 39-41 hydrogenation and hydrocracking, primary coal liquids, 57 multistage, 40-41 nondonor solvent effect, 55-56 prospects, 77-79 reaction conditions, research on moderating, 39-40 retrogressive reaction mechanisms, 57-58 solvent performance in presence of catalyst, 58-59 role in secondary stage, 69 stages, 43-44 two-stage liquefaction, 60-61 Coal liquid, pretreatment, 75-77 Coatings active, flame-sprayed, doped nickel oxide, 121 containing cobalt and iron oxides, 103-105 in siru deposition, 107-109 electrocatalyst, functioning, longevity, and application, 95-96 nickel alloys containing molybdenum, 119 nickel sulfide, 112-113 platinum metal, 120-121 platinum metal oxides, 119- 120 Raney-nickel, 113-1 10 redox, anodic oxidation mediation, 153-157 RuOZ development, 100 improvement of adhesion and strength, 99-100 preparation and formulation, 98 technically applied, 112

Cobalt coatings containing, in situ deposition, 107-109 Pourbaix diagram, 106 Cobalt magnetite coatings, 105 Cobalt oxides, coatings containing, 103-105 Cold rolling, Raney-nickel-coated cathodes, 114 CoMo catalysts, 181 Contact synergy model, 183 Cogss, structure, 222 Cracking, 50 Cyclohexane, oxidative dehydrogenation, 14-24 other orthovanadate catalysts, 18- 19 product selectivity, 15- 16 reaction rates on vanadium oxide catalysts, 16-17

D Dehydrogenation, oxidative, see Alkanes, light, oxidative dehydrogenation Depolymerization, coal, 45-47 alphaltenes, 63 ionic, 77 Deterioration, of catalysts, 139-140 Dioxygen, absorption modes, 125-126 Dissolution, coal, 45 catalyst role, 56 hydrogen donor and solvent functions, 51-56

E Electrocatalysis, 87- 168, see also Electroorganic synthesis; Fuel cells anodic methanol oxidation, 140- 142 cathodic hydrogen evolution, 11 1- 122 active coatings of flame-sprayed, doped nickel oxide, 121 nickel alloys containing molybdenum, 119 nickel sulfide coatings, 112- 113 platinum metal coatings, 120-121 platinum metal oxides coatings, 119-120

278

INDEX

Electrocatalysis (continued) platinum and platinum metal cathodes in membrane water electrolyzers, 122 Raney-nickel coatings, 113-1 19 technically applied coatings, I12 technoeconomical significance, 111-112 definition, 87-88 electrode kinetics, 88-91 principles, 90-91 of the second kind, 166-167 Electrocatalyst, see also specific catalysts coatings, functioning, longevity, and application, 95-96 particles, porous, utilization, 93-95 Electrochemical catalysis, 9 1-93 Electrodes electrocatalytically activated dimensionally stable chlorine-evolving, 97- 103 anode reaction, 97 cell design using DSAs, 100-101 coating preparation and hnulation, 98 DSAs for chlorate and hypochloritc production, 102-103 improvement of adhesion and strength of coatings, 99-100 lifetime, 101-102 selectivity at RuOz-anodes, 97-98 gas diffusion catalyst aging, 134- 135 complete PTFE-bonded carbon electrodes, 133- 134 ionomer impregnation, 143 low-temperature fuel cells, 123 in membrane fuel cells, 142-144 Pt-activated soot and active carbon, 130-131 Pt-alloy catalysts, 132- 133 Pt microcrystal particle size on soot. 13 I - 132 kinetics and electrocatalysis, 88-91 surface, as mediator and catalyst, 159-162 titanium, catalyst-coated, oxygen evolution from acid solutions, 109-1 I 1 Electrolysis, alkaline water, 103-109 coatings containing cobalt and iron oxides, 103- 105

electrocatalysis by Raney-nickel coatings. 109 in siru deposition of coatings, 106-109 Electroorganic synthesis, 15 I - 167 anodic oxidations mediated by redox coatings, 153- 157 homogeneous and heterogeneous redox reaction rates, 153 by oxides of multiply-valent metals, 154-157 cathodic pinacolisation of aliphatic aldehydes and ketones, 165- 166 direct anodic and cathodic conversions of organic substrates, 152- 153 electrocatalytic action of nonreactant electrosorbed species, 166- 167 clcctrocatalytic hydrogenation, 157- 159 electrocatalytic reduction, IS9 electrode surface as mediator and catalybt, 159-162 heterogeneous catalysis, anodic oxidation of olefins and bisarene sulfides, 162-165 Kolbe reaction, 160-161 mediated conversions of organic substrates, 152 Electrosorbed species, nonreac tant , electrocatalytic action, 166- 167 Ethane gas-phase reactions, activation energies and rate constants, 3-4 oxidative dehydrogenation, 5 -X behavior, 33 catalysts with metal cations, 7-8 pathways and temperature, 7 product selectivity, 6-7 Ethene, gas-phase reactions, activation energies and rate constants, 3-4

F FezOl coatings. 105 Fuel cells alkaline, Raney-nickel anodes, 136- I39 anodic hydrogen oxidation catalysts, 130 catalyst aging, 134- 135 cathodic oxygen reduction at low temperatures, kinetic aspects, 123, 125- I26 high-temperature, 122- 123, 144-149

279

INDEX electrode reactions, 144-145 molten carbonate cells, 145- 149 low-temperature, 122- 123 gas diffusion electrodes, 123- 124 membrane development rationale, 142 gas diffusion electrodes, 142- 144 metal catalysis, cathodic oxygen reduction, 127 metal oxide electrocatalysts, 127-128 molten carbonate, 145- 149 anodic hydrogen oxidation, 145- 148 cathodic oxygen reduction, 148-140 N4-chelates, 128-130 Pt-activated soot and active carbon catalysts, 130-131 Pt-alloy catalysts, 132-133 PTFE-bonded carbon electrodes, 133- 134 Pt niicrocrystal particle size on soot, 131-132 regenerative, 144 reversible and irreversible catalyst deterioration, 139- 140 silver cathodes, 135- 136 solid oxide. 149-151

liquefaction behavior, 52-55 Hydrotreating, catalysts of higher activity, 68 - 69 Hypochlorite, production, dimensionally stable anodes, 102- 103

I Inorganic deposits, catalyst deactivation, 7 1 Interfacial charge transfer, kinetics, 88-90 Iron catalysis, 50 catalysts, coal liquefaction, 57-58 coatings containing, in situ deposition, 107-109 Pourbaix diagram, 106 Iron oxides, coatings containing, 103-105

K Ketones, cathodic pinacolisation, 165- I66 Kolbe reaction, 160-161

M H Heat of reoxidation, 19-22 Heyrovsky reaction, 92 Hydrocracking primary coal liquids, 57 procedure, 66-67 Hydrodenitrogenation. coal heavy liquid heavy distillate, 65 -66 Hydrodesulfurization, diesel fuel, heavy gas oil, and atmospheric residue, 67-68 Hydrogen anodic oxidation, 130 molten carbonate fuel cells, 145-148 cathodic evolution, see Electrocatalysis, cathodic hydrogen e vol ut ion chemisorption, transition-metal sulfide catalysts, 202 Hydrogenation electrocatalytic. 157- 159 primary coal liquids, 57 Hydrogen donors functions in coal dissolution, 51 -56

Membrane water electrolyzers, platinum and platinum metal cathodes, 122 Metal catalysis, cathodic oxygen reduction, 127 Metal oxides electrocatalysts, 127- 128 multiply-valent metals, electrocatalytic oxidations, 154- 157 Metal-oxygen bond, strength and selectivity, oxidative dehydrogenation of alkanes, 26-28 Methanol, anodic oxidation, 140-142 Mg orthovanadate alkane oxidation active sites, 26, 28-29 molecule size requirement, 32-33 product distributions, 28-29, 31 selectivity patterns, 28-34 vanadium ion separation, 32 oxidative dehydrogenation butane and cyclohexane, 17, 19 propane, 9-10

280

INDEX

Mg pyrovanadate alkane oxidation active sites, 26, 28-29 molecule size requirement, 32-33 product distributions, 28-29, 3 1 selectivity patterns, 28-34 vanadium ion separation, 32 oxidative dehydrogenation butane and cyclohexane, 17 propane, 10 Molybdate, divalent-metal, crystal structure, 240 MoS2 catalytic properties, 219-220 edge structure, 204 optical spectra, 205

N

Nernst’s law, I17 Nickel alloys containing molybdcnuin, coatings, 1 I9 anodes, in situ activation, 107-109 Pourbaix diagram, 106 Nickel oxide, flame-sprayed, active coatings, 121 Nickel sulfide, coatings, 112-1 13 NiMo/A1201catalysts, activity curve, I8 I Ni-Mo hydrous titanium oxide catalysts, in secondary liquefaction stage, 63-64 Nitrogen, removal, in catalytic liquefaction upgrading, 65-66 Nuclear magnetic resonance,77Se, transition-metal sulfide catalysts, 225-226

0 Olefins anodic oxidation, 102- 165 catalytic ox idat ion and amnioxidat ion, 233 Organic substates direct anodic and cathodic electrochemical conversions, 152-153 mediated electrochemical conversions, 152

Oxidation allylic history of catalysts, 237-241 lower olefins, 233-234 anodic mediated by redox coatings, 153- I57 olefins and bisarene sulfides, 162- 165 styrene, 164 propylene, on model multicomponent bismuth molybdate catalyst, 253-258 Oxide ion bulk diffusion, multicomponent bismuth molybdate catalyst stability dependence, 265-269 lattice, involvement in multicomponent bismuth molybdate catalyst reactions, 250-253 Oxides active sites, 26 of multiply-valent metals, electrocatalytic oxidations, 154- 157 Oxygen absorption modes, 125- 126 active migrat ion, mult icomponrn t bismuth molybdate catalyst, 256-257 cathodic reduction, molten carbonatc lucl cells, 148-149 gas-phase reaction with alkancs, 2-4

P PbOz anode, 155 - I56 oxygen evolution, 109- I10 Petroleum, see Transition-metal sulfide catalysts Pinacolisation, cathodic, aliphatic aldehydes and ketones, 165-166 Platinum-alloy catalysts, 132- 133 Platinum catalyst, poisoning by methanol oxidation products, 14 I Platinum metal catalysts, cathodic oxygen reduction, 127 coatings, 120- 12 1 Platinum metal oxides, coatings, 119- 120 Preasphaltenes, 76 depolymerization, 63 Precursor synthesis techniques, transition-metal sulfide catalysts, 19 I Propane gas-phase reactions, activation energies and rate constants, 3-4

28 1

INDEX oxidative dehydrogenation, 8- 14 enhancement extent postcatalytic configuration, 12-13 product distributions, 14 product selectivity, 9-12 Propene, gas-phase reactions, activation energies and rate constants, 3-4 Propylene activation sites, 258 allylic oxidation, 234 ammoxidation, reaction conditions, 238 -239 oxidation Arrhenius plots, 249 multicomponent bismuth molybdate catalyst activity, 245 on model multicomponent bismuth molybdate catalyst, 253-258 reaction conditions, 238-239 selective, to acrolein, 235 selective ammoxidation, to acrylonitrile, 235-236 Proton exchange membrane, gas diffusion electrodes, 142- 144 Pseudointercalation model, 182 transition-metal sulfide catalysts, 202 Pyrochlores, cathodic oxygen reduction catalyst, 128

coatings development, 100 improvement of adhesion and strength, 99-100 preparation and formulation, 98 stabilizing, 110

s Silver cathodes, oxygen reduction in alkaline electrolyte, 135- 136 Solid/liquid separations, coal, 62- 63 Solvents, coal liquefaction coal dissolution role, 51-56 performance in presence of catalyst, 58-59 role in secondary stage, 69 Spectroscopy, transition-metal sulfide catalysts, 202-206 Styrene anodic oxidation in methanol, 164 physisorption, 162 Sulfur, vacancies, role in transition-metal sulfide catalysts, 220-221 Synergy, electronic hypothesis, 21 1-213

T R Raney-nickel anodes, alkaline fuel cells, 136-139 coatings, 113- I19 anodic oxygen evolution electrocatalysis, 109 catalyst utilization, 114, 116- 118 precursor alloys and fabrication of cathodes, 1 13- 1 I5 Reduction, electrocatalytic, 159 Remote control model, 183-184 Reoxidation, heat of, 19-22 Rimledge model, 218-219 RuO~ anodes anodic chlorine evolution selectivity, 97-98 platinum metal oxide coatings, 119-120

Tafel equation, 89 Tafel reaction, 92 Thiele modulus, 94, 138 Transition-metal sulfide catalysts, 177-229 carbon role, 221 ColMo, chain-like particle, 223-224 Co/Mo/S catalytic materials, 186-187 unsupported, microstructure, 187- 188 COMOSphase, 184-185 contact synergy model, 183 couss, 222 crystal structure importance, 213 -2 16 reaction selectivity role, 216-220 edge and corner sites, 183 electronic structure importance, 206-21 I HDS activity, 206-209 reaction pathway, 216-217 HOMO, 209-210

282

INDEX

Transition-metal sulfide catalysts (continued) immiscible phases and symmetrical synergy, 186-188 isotropic sulfides, 193 layered sulfides. 193 metal oxidation state and structural environment, 199-200 morphology, particle size, and surface area, 194- 198 MoS~ crystallites, 197- 198 interaction with second phases, 197- I98 local atomic structure, 184-185 0 2

adsorbed edge atoms versus site density, 214-215 chemisorption on RuS2/MoSZ,213-214 preparation, 188- 19I binary sulfides, 189- 190 mixed-metal sulfides, 190- 19 1 precursor control of activity and selectivity, 191 promoted CoiMo model, 223-224 promotion effect early models, 180- 186 history, 179- 180 pseudointercalation model, 182 “rag” structure, 194- 195 recent developments, 221 -226 remote control model, 183-184

resistance to poisons, 179 rim/edge model, 2 18-2 I9 selectivity versus X-ray diffraction parameters, 217-218 ”Se NMR, 225-226 stable catalytic phases and structures, I 92- 194 sulfur vacancy role, 220-221 surface area versus HDS activity tor MoS2, 213 surface composition, 200-206 chernisorption and molecular probes. 200201 hydrogen chemisorption and pseudointercalation, 20 I spectroscopic techniques, 201 -206 synergy, electronic hypothesis, 2 I 1-21 3

V Vanadyl pyrophosphate, alkane oxidation active sites, 26, 28-29 molecule size requirement, 32-33 product distributions, 28-29, 3 I selectivity patterns, 28-34 vanadium ion separation, 32 Volcano curve, 92 Volmer reaction, 92

This Page Intentionally Left Blank

ADVANCES IN IMMUNOLOGY VOLUME 40, Volume 40

Read more

Advances in Catalysis, Volume 38

Read more

Advances in Catalysis, Volume 27

Read more

Advances in Catalysis, Volume 52

Read more

Advances in Catalysis, Volume 48

Read more

Advances in Catalysis, Volume 34

Read more

Advances in Catalysis, Volume 46

Read more

Advances in Catalysis, Volume 29

Read more

Advances in Catalysis, Volume 32

Read more

Advances in Catalysis, Volume 37

Read more

Advances in Catalysis, Volume 42

Read more

Advances in Agronomy, Volume 40

Read more

Advances in Catalysis, Volume 28

Read more

Advances in Catalysis, Volume 33

Read more

Advances in Catalysis, Volume 24

Read more

Advances in Catalysis, Volume 23

Read more

Advances in Catalysis, Volume 23

Read more

Advances in Catalysis, Volume 31

Read more

Advances in Catalysis, Volume 50

Read more

Advances in Chromatography, Volume 40

Read more

Advances in Catalysis, Volume 26

Read more

Advances in Catalysis, Volume 34

Read more

Advances in Catalysis, Volume 36

Read more

Advances in Catalysis, Volume 30

Read more

Advances in Catalysis, Volume 41

Read more

ADVANCES IN CATALYSIS VOLUME 5, Volume 5

Read more

ADVANCES IN CATALYSIS VOLUME 11, Volume 11

Read more

ADVANCES IN CATALYSIS VOLUME 39, Volume 39

Read more

ADVANCES IN CATALYSIS VOLUME 16, Volume 16

Read more

ADVANCES IN CATALYSIS VOLUME 2, Volume 2

Read more

Recommend Documents

ADVANCES IN IMMUNOLOGY VOLUME 40, Volume 40

ADVANCES IN Immunology VOLUME 40 This Page Intentionally Left Blank ADVANCES IN Immunology EDITED BY FRANK J. DI...

Advances in Catalysis, Volume 38

ADVANCES IN CATALYSIS VOLUME 38 Advisory Board M. BOUDART Stanford, California V. B. KAZANSKY Moscow, Russia G. A....

Advances in Catalysis, Volume 27

ADVANCES IN CATALYSIS VOLUME 27 Advisory Board G. K . BORESKOV Novosibirsk, U . S . S . R . P. H. EMMETT Portland Or...

Advances in Catalysis, Volume 52

ADVISORY BOARD M. CHE Paris, France ´S A. CORMA CANO Valencia, Spain D.D. ELEY Nottingham, England G. ERTL Berlin/Dah...

Advances in Catalysis, Volume 48

ADVANCES IN CATALYSIS VOLUME 48 Advisory Board M. CHE D.D. ELEY G. ERTL Paris, France Nottingham, England Berlin...

Advances in Catalysis, Volume 34

ADVANCES IN CATALYSIS VOLUME 34 Advisory Board M. BOUDART Stunford, California V. B. KAZANSKY Moscow, U.S.S.R. G. ...

Advances in Catalysis, Volume 46

ADVANCES IN CATALYSIS VOLUME 46 Advisory Board D. D. ELEY G. ERTL V. B. KAZANSKY Nottingham, England Berlin/Dahle...

Advances in Catalysis, Volume 29

ADVANCES IN CATALYSIS VOLUME 29 Advisory Board G. K. BORESKOV Novosibirsk, U . S . S . R . M. BOUDART Stanford, Cal...

Advances in Catalysis, Volume 32

ADVANCES IN CATALYSIS VOLUME 32 Advisory Board M. BOUDART Stanford, California M. CALVIN Berkeley, California V. B...

Advances in Catalysis, Volume 37

ADVANCES IN CATALYSIS VOLUME 37 Advisory Board M. BOUDART Stanford, California V. B. KAZANSKY Moscow, U . S . S . ...