Advances in Catalysis, Volume 37

ADVANCES IN CATALYSIS VOLUME 37 Advisory Board M. BOUDART Stanford, California V. B. KAZANSKY Moscow, U . S . S . ...

22 downloads 1034 Views 15MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

ADVANCES IN CATALYSIS

VOLUME 37

Advisory Board

M. BOUDART Stanford, California

V. B. KAZANSKY Moscow, U . S . S . R .

G. A. SOMORJAI Berkeley, California

G. ERTL BerlinlDahlem, F.R.G.

A. OZAKI Tokyo, Japan

W. 0. HAAC Princeton, New Jersey

W. M. H. SACHTLER Evanston, Illinois

J . M. THOMAS London, U . K .

ADVANCES IN CATALYSIS VOLUME 37

Edited by D. D. ELEY

HERMAN PINES

The University Nottingham, England

Northwestern University Euanston, Illinois

PAULB. WEISZ University of Pennsylvania Philadelphia, Pennsylvania

ACADEMIC PRESS, INC. Harcourt Brace Jovanovich, Publishers

San Diego New York Boston London Sydney Tokyo Toronto

This book is printed on acid-free paper.

@

Copyright 0 1990 By Academic Press, Inc. All Rights Reserved. No part of 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. San Diego, California 92101 United Kingdom Edition published by Academic Press Limited 24-28 Oval Road. London NWI 7DX

Library of Congress Catalog Card Number:

49-7755

ISBN 0-12-007837-6 (alk. paper)

Printed in the United States of America 9 0 9 1 9 2 9 3 9 8 7 6 5 4

3

2

1

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

vii ix

Spontaneous Monolayer Dispersion of Oxides and Salts onto Surfaces of Supports: Applications to Heterogeneous Catalysis YOU-CHANG X I E 1.

IS.

111.

IV . V.

AND YOU-QI

TANG

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . -1 Phenomenon and Nature of Monolayer Dispersion . . . . . . . . . . . . . 2 Effects of Monolayer Dispersion . . . . . . . . . . . . . . . . . . . . . . 19 Applications to Heterogeneous Catalysis . . . . . . . . . . . . . . . . . . 34 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

Catalysis by Supported. Unsupported. and Electron-Deficient Palladium ZBIGNIEW KARPIIQSKI 1. I1 . 111. IV . V.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Catalysis over Unsupported Palladium . . . . . . . . . . . . . . . . . . . Catalysis by Electron-Deficient Palladium . . . . . . . . . . . . . . . . . Catalysis over Supported Palladium . . . . . . . . . . . . . . . . . . . . . Conclusions-Suggestions for Future Work . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45 47 61 77 93 94

The Bond-Order Conservation Approach to Chemisorption and Heterogeneous Catalysis: Applications and Implications EVGENY SHUSTOROVICH 1. 11. 111. IV . V. VI . VII .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . BOC-MP Formalism: A Brief Reminder . . . . . . . . . . . . . . . . . . Basic BOC-MP Applications . . . . . . . . . . . . . . . . . . . . . . . . Mechanisms of Catalytic Heterogeneous Reactions . . . . . . . . . . . . Comparisons with Other Theoretical Techniques . . . . . . . . . . . . . . The BOC-MP Model: Comments and Summary . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V

101

103 117 134 151

155 156 156

vi

CONTENTS

Solid Superacids KAZUSHl ARATA 1.

I1 . 111.

IV . V. VI . v11. VIII . IX .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Liquid Superacids Supported on Solids . . . . . . . . . . . . . . . . . . . Superacid-Intercalated Graphites . . . . . . . . . . . . . . . . . . . . . . Aluminum Halide-Metal Salt Mixtures . . . . . . . . . . . . . . . . . . . Nafion-H (Perfluorinated Resin Sulfonic Acid) . . . . . . . . . . . . . . . Sulfate-Supported Metal Oxides . . . . . . . . . . . . . . . . . . . . . . . Superacids by Metal Oxides . . . . . . . . . . . . . . . . . . . . . . . . . Aluminum Halides Supported on Alumina . . . . . . . . . . . . . . . . . Conclusion and Prospects . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

165 . (68 171 172 173 177 201 205 205 206

Oscillatory Catalytic Reactions at Single Crystal Surfaces G . ERTL I. I1 . 111. 1V .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Catalytic Oxidation of Carbon Monoxide on Pt(100) and Pt(l10) Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Other Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

213 215 260 271 273

Role of Sulfur in Catalytic Hydrogenation Reactions J . BARLIIER. E . LAMY.PITARA.P. MARECOT.J . P . BOITIAUX.J . COSYNS.A N D F . VERNA 1.

I1 . 111.

IV . V. VI .

INDEX

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sulfur Adsorption on Metals . . . . . . . . . . . . . . . . . . . . . . . . Sulfur Effect of Adsorption of the Reactants . . . . . . . . . . . . . . . . Effect of Sulfur Adsorption on the Catalytic Activity . . . . . . . . . . . Effect of Sulfur Adsorption on Catalytic Selectivity . . . . . . . . . . . . Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

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

279 280 294 300 308 312 315 319

Contributors Numbers in parentheses indicate the pages on which the authors’ contributions begin.

Department of Science, Hokkaido University of Education, Hakodate, Japan 040 (165) J . B A R B I E R , UniversitP de Poitiers, Laboratoire de Cutulyse en Chimie Organique, C N R S U . R . A . 350, 86022 Poitiers, France (279) J. P. B O I T I A U X , Institut Fruncais du PPtrole, 92506 Ruril Malmaison, France (279) J. COSYNS, Institut Francais du PPtrole, 92506 Rued Malmaison, France (279) G . E R T L , Fritz-Haber-Institut der Max-Planck-Gesellschafr, D-1000 Berlin 33, Federal Republic of Germany (213) ZBIGNIEW K A R P I N S K I , Institute of Physical Chemistry, Polish Academy of Sciences, Warsuw, Polund (45) E. LAMY-PITARA, UniversitP de Poitiers, Laboratoire de Catalyse en Chimie Organique, CNRS U . R . A . 350, 86022 Poitiers, France (279) P. MARECOT, UniversitP de Poitiers, Laboratoire de Catalyse en Chimie Organique, CNRS U . R. A . 350, 86022 Poitiers, France (279) EVGENYSHUSTOROVICH, Corporate Research Laboratories, Eastmun Kodak Company, Rochester, New York 14650 (101) You-QI T A N G , Laboratory for Structure of Matter, Institute of Physical Chemistry, Peking University, Beijing, China ( 1 ) F. VERNA,Institut Francais du PPtrole, 92506 Rueil Malmaison, France (279) YOU-CHANG XIE,Laboratoryf o r Structure ofMatter, Institute ofphysical Chemistry, Peking University, Beijing, China ( 1 ) K A z u s H I ARATA,

vii

This Page Intentionally Left Blank

Preface Catalytic science embraces a variety of fields and traditionally each volume of Advances in Catalysis articles covers a wide subject of interest. The current volume is not an exception. It contains articles originating from six countries on three different continents. The first article by Xie and Tang provides a review of “Spontaneous Monolayer Dispersion of Oxides and Salt onto Surfaces of Supports: Applications to Heterogeneous Catalysis.” Even though the versatility of palladium in hydrogenation reactions is recognized, the explanation of its catalytic properties is still far from being satisfactory. The chapter by Z . Karpifiski gives a comprehensive survey of “Catalysis by Supported, Unsupported, and Electron-Deficient Palladium. ’’ The chapter by Evgeny Shustorovich attempts to interpret phenomena occurring in heterogeneous catalysis based on bond-order conservation potential to chemisorption. Novel organic syntheses that are not possible in usual acidic media can be accomplished by superacids. The review by K. Arata on “Solid Superacids” summarizes recent research on synthesis of superacids and their catalytic actions. Oscillatory kinetics in heterogeneous catalysis, first reported about two decades ago, stimulated extensive study of this interesting phenomenon. G. Ertl gives us an indepth review of this subject in the article “Oscillatory Catalytic Reactions at Single Crystal Surfaces. ” Sulfur is generally considered to be a poison of hydrogenation reactions. However, in commercial hydrogenation and dehydrogenation reactions, sulfur is also used as a modifier, and in some cases as activator of catalytic hydrogenation reactions. Barbier and co-authors review the “Role of Sulfur in Catalytic Hydrogenation Reactions.”

HERMANPINES

ix

This Page Intentionally Left Blank

ADVANCES IN CATALYSIS. VOLUME 37

Spontaneous Monolayer Dispersion of Oxides and Salts onto Surfaces of Supports: Applications to Heterogeneous Catalysis YOU-CHANG XIE

AND

YOU-QI TANG

Laboratory for Structure of Mutter Institute of Phy.\icul Chemistry Peking Uniuerrity Beijing, Chinu

1.

Introduction

The forms of active components present in heterogeneous catalysts are of importance to catalysis. A supported catalyst usually consists of an active component dispersed on a support with a highly specific surface. According to current opinions ( I ) , an active component dispersed on a support may end up in one of three forms: ( I ) it may retain its chemical identity as a separate crystalline o r amorphous phase, ( 2 ) it may form a new stoichiometric compound with the support or additive, or (3) it may dissolve in the support to give a solid solution. Examples of these forms are readily available from the literature. In the investigation of a few catalysts, some authors have suggested that the active components may be present as a monolayer. For example, Russell and Stokes (2) have concluded that the catalyst Mo0,ly-Al,O, , made by impregnating y-alumina with ammonium molybdate solution followed by calcination, may form a monomolecular layer of MOO, on the surface of alumina, because the catalytic activity is linearly related to the molybdenum content up to a limit corresponding to full coverage. Lipsch and Schuit ( 3 ) have found that heating a mixture of MOO, and alumina causes an increase in intensity of the MOO, absorption bands of reflectance spectra, thus suggesting that MOO, spreads to form a monolayer. This phenomenon has also been found by Massoth ( 4 ) and Giordano et al. ( 5 ) . Buiten ( 6 ) has described a method to chemisorb gaseous MoO,(OH), at 1 Copyright 0 1990 hy Academic Press. Inc. All rights of reproduction in any form reserved.

2

YOU-CHANG XIE A N D YOU-QI TANG

600°C on SnO, in order to get a complete monolayer. Mars and co-workers (7) prepared a monolayer of MOO, on A1203,Cr,O,, TiO,, CeO,, SO,, and ZrOz by using this method and extending it to adsorption of molybdate ions from acid solution. But there were disagreements with the monolayer model (1, 8), because compounds such as AI,(Mn0,)3, formed by the active component with the support, had been found in some preparations. In 1979 we suggested a monolayer model for a high-efficiency polyethylene catalyst in which TiCI, forms a monolayer on the surface of a magnesium chloride support (9). We have also reported that an oxychlorination catalyst, CuCl,/y-Al,O, , is of the monolayer type (10). After a systematic investigation of supported catalysts, we have found that monolayer dispersion is a common or even ubiquitious form rather than a rare occurrence (11-14). A great many oxides and salts can disperse spontaneously onto the surface of supports to form a monolayer or submonolayer, because in these cases the monolayer is a thermodynamically stable form. This article provides a review of various aspects of the phenomenon of spontaneous monolayer dispersion, namely, its nature, effects, and applications. It is based in the main on the work that has been carried out in our laboratory in the last 16 years. Relevant data and results from other laboratories have been included for discussion. II. Phenomenon and Nature of Monolayer Dispersion

A.

SOMECLUESTO MONOLAYER DISPERSION

In the 1970s we began to be interested in the phase composition of many commercial catalysts that are based on supports with highly specific surfaces; we found that oxides or salts, as active components in these catalysts, even though present in considerable quantity, did not show up in the X-ray diffraction (XRD) patterns. For example, Fig. 1 shows XRD patterns for some of these catalysts. The content of transition metal oxides or salts in these catalysts is of the order of 10% by weight, but they still fail to give peaks in XRD patterns. The patterns are merely those of amorphous supports. In general, it has been taken for granted that the disappearance of XRD peaks of an active component in a catalyst is due to the fact that the active component is present in such a small quantity as to evade detection by XRD. For example, in a supported noble metal catalyst, such as a platinum-reforming catalyst, Ptly-Al,O,, the metal content, of the order of O.l%, is too low to be detectable by XRD. However, this is not the case for the catalyst of our present interest. For the catalysts referred to in Fig.

MONOLAYER DISPERSION OF OXIDES AND SALTS

3

- f

C

- b

/ a

20

30

40

20‘

FIG. 1. XRD patterns of some catalysts. (a) HDS catalyst, 10% Moo3 and 3% COO/ y-AI2O3.(a’) 1% crystalline MOO, added to a. (b) Methanation catalyst, 15% NiO/y-Al,O,. (c) Oxychlorination catalyst, 10% CuCI2/y-Al2O3.(d and e) Two vinyl acetate catalysts: (d) 10% ZnO/silica gel and (e) 30% ZnAc2/active carbon. (f) Vinyl chloride catalyst, 10% HgC12/active carbon. (f’) 1% (by weight) crystalline HgC12 added to f.

1, the content of the active component is about two orders of magnitude higher. Normally, for crystalline transition metal compounds, a content of the order of 1% by weight is sufficient to give rise to sharp peaks in the XRD pattern. Patterns a’ and f ’ in Fig. 1 provide a convincing illustration of this point. A comparison of XRD patterns a’ and a or f and f ’ in Fig. 1 raises the question as to what has become of the active component, the quantity of which is about 10% by weight in a catalyst. First of all, the active components no longer exist in the crystalline state. It is also unlikely that they have become an amorphous mass. Later on we shall provide further evidence to verify that the active component in each case has not formed stoichiometric compound with the support nor has dissolved in the support to form solid solution. It is then reasonable to think that the active component has dispersed as a monolayer on the surface of the support. In view of the high surface area of the supports and the heavy metals in the active components of these catalysts, a monolayer may still amount to a considerable percentage, usually more than 10% by weight, of an active

4

YOU-CHANG XIE A N D YOU-QI TANG

component in the catalyst. This point has been checked by a simple calculation on the basis of the size of a “molecule” of the active component and the specific surface of the support. According to a simple model based on the assumption that the anions of oxide or salt form a close-packed monolayer on the surface of the support and the cations occupy the interstices left over by anions, one can figure out the close-packed monolayer capacity for oxide or salt on a unit area of the support. We estimate it at 0.10 g/100 m2 or higher for various active components (see later, Table 11). The specific surface of the support is about 200 m2/g for y-Al,O, , 300 m2/g for silica gel, and 1000 m2/g for active carbon. Although each of the catalysts in Fig. 1 contains aconsiderable amount of active component, its content is still lower than that estimated on the basis of a close-packed monolayer. Therefore, the monolayer dispersion in many of these catalysts does not correspond to the full coverage of the support surface, and more precisely is known as submonolayer dispersion.

B. THE SYSTEM Mo0,/y-AI,O3 A mixture of 0.10 g MOO, and 1.0 g y-Al,O, with a BrunauerEmmet-Teller (BET)-specific surface of 178 m2/g gives in Fig. 2 an XRD pattern b’ carrying sharp peaks of crystalline MOO,. After the mixture is heated for 24 h at 450”C, a temperature well below the melting point (795°C) of MOO,, the peaks of crystalline MOO, disappear and pattern b’ changes into b, a pattern resembling pattern a of y-Al,O, . The crystalline phase of MOO, has vanished but the support remains unchanged. The experiments indicate that MOO, is well retained by the sample and does not escape from the sample, because it does not matter whether we have heated the mixture in an open o r a sealed glass tube. It is also very unlikely that the crystalline phase of Moo, has been transformed into an amorphous one after the heat treatment. So it seems to us that the only reasonable answer is that MOO, has dispersed as a monolayer onto the surface of the support. However, when the content of MOO, in a mixture with y-A120, exceeds a critical amount, the peaks of crystalline MOO, do not disappear but are markedly reduced after the heat treatment. The XRD pattern c ‘ in Fig. 2 belongs to a mixture containing 0.30 g of Moo, and 1 .O g of y-Al,O,. After this mixture has been heated at 450°C for 24 h, pattern c‘ changes into c , in which the peaks of crystalline MOO, do not disappear but recede markedly. It indicates that some residual crystalline MOO, remains. Prolonged heating does not cause the sample to change its XRD pattern. It suggests

MONOLAYER DISPERSION OF OXIDES A N D SALTS

cc-cc

10

- b

-

I

20

5

4

1

I

30

40

- 0

20*

FIG. 2. XRD patterns of MoO3/y-Al2O,.(a) y-Alz03of a specific surface (178 ~ n ’ / g ) ~ . (b‘) Mixture of 0.10 g Mo03/g y-Al,03. (b) Sample b’ after a heat treatment at 450°C for 24 h. (c’)Mixture of 0.30 g Mo03/gy-Alz03.(c) Sample c‘ after a heat treatment at 450°C for 24 h.

that heating at this temperature only helps MOO, to disperse as monolayer onto the surface of y-Al,O,, but not to diffuse into or react with the bulk. From Fig. 2 it appears that there is a critical dispersion capacity of MOO, on the surface of y - A l , O , , between 0.10 and 0.30 g/g y-Al,O, . When the MOO, content in the mixture exceeds this critical dispersion capacity, there will be a residual crystalline phase of MOO, (after a heat treatment).

C. XRD QUANTITATIVE PHASEANALYSIS The amount of residual crystalline MOO, can be determined by XRD quantitative phase analysis. Liu e t al. (14) have achieved this by using a reference method (15). They use an inner standard, namely, a-Al,O,, added to MoO,/y-Al,O, samples, and measure the peak area for reflections 110 and 021 of MOO, and 113 of a-Al,O,. The peak intensity ratio IMoo,l Zy.AI,O, is reasonably assumed to be proportional to the ratio of the content of crystalline MOO, to that of a-Al,O,, as IM~O,/I~-AI~O~ = k x XM~O,/&-AI~O,

6

YOU-CHANG XIE A N D YOU-QI TANG

Total MOO, g/g y-Al2O3

FIG.3. Residual amount of Moo3versus total amount of Moo3in Mo0,/y-A120, samples prepared on y-A1203of a specific surface (178 rn?/g) by heating at 450°C for 24 h.

where I and X stand for XRD intensity and weight percentage, respectively, and k is a proportionality constant determined from a sample of known phase composition. With known content of a-Al,O,, the weight percentage of crystalline MOO, can be derived from the intensity ratio ~ M o O , ~ ~ r Y - A l ,*O ,

D. THETHRESHOLD Using the method described above, Liu rt ul. (14) have determined the residual amount of crystalline MOO, for a number of Mo0,/y-A120, samples after a heat treatment at 450°C for 24 h, and have obtained a plot of the residual amount of crystalline MOO, versus the total amount of crystalline MOO, before the heat treatment, as shown in Fig. 3. From the plot one can see a threshold at 0.22 g MoO,/g y-Al,O, , corresponding to the critical dispersion capacity. When the content of Moo, in the sample is below this threshold value, no crystalline MOO, can be detected. If the MOO, content exceeds this threshold, the residual amount of MOO, increases with the total amount of MOO,, as is shown by the straight line in the plot. This straight line does not go through the origin but gives an intercept corresponding to the utmost dispersion capacity. In an ideal case the slope of this line is unity. We have also observed that the threshold value or the utmost dispersion capacity of MOO, on y-Al,O, is almost linearly proportional to the specific surface of the latter. In one series of our samples, the specific surface of y-Al,O is 178 m2/g, and the threshold value is 0.22 g MoO,/g y-Al,O,,

MONOLAYER DISPERSION OF OXIDES A N D SALTS

7

FIG.4. Close-packed monolayer model for MOO,

corresponding to a dispersion capacity of 0.12 g/100 m2of y-A120, surface. In other words, a “molecule” of MOO, occupies 20 A2. The results obtained by Fransen et af. (7) from their preparative work by means of adsorption in gas and solution are 17 and 22 A2/“molecule” of MOO,, respectively. Liu et al. (16) have also used y-alumina made from different hydrates (such as boehmite, bayerite, and gibbrite) or prepared by different procedures, and obtained the same dispersion capacity of 0.12 g MoO,/IOO m2 of y-Al,O, surface. We have also tried a lower temperature (35OOC) and a longer heating time, and obtained the same threshold. As long as the temperature is not too high and the time is long enough, the same dispersion capacity is obtained. The time required depends on the temperature used in heat treatment and will be shortened by thoroughness of mixing. If the samples are heated at a temperature over 5OO0C, the compound AI,(MoO,), is formed and can be detected by XRD (14, 16). In this case there will be, of course, no threshold. If a temperature below 350°C is applied, the dispersion will take an unduly long time to reach the equilibrium state.

E. CLOSE-PACKED MONOLAYER MODEL One can estimate the utmost monolayer capacity by using the simple close-packed monolayer model. Assuming that 02-ions from MOO, form a close-packed layer on the surface of y-A120, and that the Mo6+ ions occupy the interstices formed by 02-ions, as is shown in Fig. 4, we can get, by taking 1.4 A for the radius of 02-ion (17), a close-packed monolayer capacity of 0.12 g Mo0,/100 m2, which is in good agreement with the dispersion capacity from the plot in Fig. 3 based on the data from XRD phase analysis. To recapitulate we point out that MOO, disperses spontaneously onto the surface of y-A120, and forms a close-packed monolayer.

8

YOU-CHANG XIE A N D YOU-QI TANG

- b' ma

15

20

25

30

35

28"

FIG. 5. XRD patterns of CuCl2/y-AI2O3.(a) y-A1203of specific surface (170 m2/g). (b) Mixture of 0.10 g CuC1, and 1 .0 g y-AlzO,. (b') Sample b after a heat treatment at 350°C for 12 h. (c) Mixture of 0.30 g CuC1, and 1 .0 g y-A120,. (c') Sample c after a heat treatment at 350°C for 12 h.

F. OTHERSYSTEMS The phenomenon of monolayer dispersion observed in the system MoO,/y-Al,O, is by no means restricted, but is instead rather widespread. We have prepared two samples by mixing, respectively, 0.10 and 0.30 g CuCI, with 1.O g y-Al,O, of a specific surface of 170 m2/g. Figure 5 shows XRD patterns for y-Al,O, (a) as well as for our two samples (b and c), which were obtained after a heat treatment for 24 h at 350"C, a temperature well below the melting point of CuCI, (498°C). By comparing pattern c' with b', we can estimate that the threshold value is between 0.10 and 0.30 g CuClJg y-Al20,. Based on the results from the XRD phase analysis we have obtained a plot, shown in Fig. 6, and have derived the threshold value 0.13 g CuC12/ g y-Al,O,, corresponding to a dispersion capacity of 0.077 g CuC1,/100 m2 of y-A1,03 surface. By taking 1.80 A as the radius of CI- ion (17), one can estimate the close-packed monolayer capacity at 0.10 g CuC1,/100 m2 of y-Al,O, surface, which suggests that CuCI, forms a submonolayer covering about 77% of the y-Al,O, surface. We have extended our investigation to a great many systems of oxides and salts on supports with highly specific surfaces (11-14, 18-21). They all display the phenomenon of spontaneous monolayer dispersion. In Table I these systems are given along with the temperature and the period of time for a suitable heat treatment. In Fig. 7 we have collected XRD patterns for various systems of salts/

9

MONOLAYER DISPERSION OF OXIDES A N D SALTS

Total CuCI, g / g T - A I 2 4

FIG.6. Residual amount of crystalline CuClz versus total amount of CuC12 in CuCI2/ y-Al,O, samples after a heat treatment at 350°C for 24 h. The specific surface of y-Alz03is

170 m'lg.

TABLE 1 Systems Displaying Spontaneous Monolayer Dispersion ~~~

Support y-A1203

Compound

MOO, CUCl2 CUCl

vzo,

S O , gel

Active C

TiOz

NaCl NaF KCI NaBr Nal LiCl FeCll . 6 H 2 0 NiCI, . 6 H 2 0 C O C I ~' 6H20 Fe(NO,), . 9 H 2 0 Ni(NO& . 6 H 2 0 Co(NO?)Z ' 6Hz0 MOO? HgCh ZnAcl HgClz ZnAcz MOO,

Melting point ("C) 195 498 422 690 80 1 980 176 155 65 1 613 31 86 41 57 < 100 795 216 242 216 242 795

Heat treatment

"C

Hours

350-450 350 350-400 650 400-550 500 500 500 400 400 10 70 70 30 30 30 450 I00 170 25 105 400

24 24 2 34 54-24 24 24 24 34 34 78 78 78 89 89 13

14 1 18

24 22 24

Reference 11-14, 16

I I , 19 I I , 20 I1 11, 18 11 11 11

I1 II I1 11

II II I1 II 23 II II 11 11

23

10

YOU-CHANG XIE AND YOU-QI T A N G

b'

FIG. 7. XRD patterns for several systems of salt/y-A1203before (a-f) and after (a'-f') heat treatment according to the conditions stated in Table I. Added to 1.0 g y-Al,O, of a specific surface (170 m2/g) are (a) 0.11 g KCI, (b) 0.05 g NaCI, (c) 0.03 g LEI, (d) 0.032 g NaF, (e) 0.06 g NaBr, and (f) 0.10 g NaI.

y-Al,O, before and after heat treatment according to the conditions stated in Table I. Figure 8 shows XRD patterns for ZnAc,, MOO,, and HgCI, supported on silica gel. Some salts that melt at low temperatures can disperse spontaneously onto the surface of a suitable support even at room temperature. Worthy of special mention is the behavior of HgCI, mixed with active carbon. This low-melting-point salt disperses at 30°C onto a support having a very specific surface, namely, active carbon, at a noticeable rate, as is shown in Fig. 9. Hydrated nitrates and chlorides can also disperse spontaneously at ambient or mild temperatures onto the surface of y-Al,O,, as is evidenced in Fig. 10. We may notice a general trend in Table 1 that a lower melting point compound usually disperses at a lower temperature. But this is not always the case. Substances such as NaCI, NaF, and Moo3 can disperse on y-Al,O, at a temperature about 400°C below their melting points. And

MONOLAYER DISPERSION OF OXIDES A N D SALTS

10

15

20

25

11

30 2%"

FIG. 8. XRD patterns for systems of ZnACz, Mooi, and HgCI2 on silica gel before (a-c) and after (a'-c') heat treatment according to the conditions stated in Table I. Added to I g silica gel of a specific surface (300 m?/g) are (a) 0.40 g ZnAcz. (b) 0.12 g MOO,, (c) 0.10 g HgCI,, and (d) silica gel.

some other substances, such as ZnAc, and V,O,, disperse, respectively, on silica gel and y-Al,O, only at a temperature close to their melting points. Plots for these systems, similar to one shown in Fig. 3 , have been obtained on the basis of the data from XRD quantitative phase analyses, but they are omitted here. Each plot contains a threshold dispersion capacity. The dispersion capacities so derived in our work are listed in Table I1 (Section 11,G). We can see that within the limits of experimental error the dispersion capacities are either equal to or lower than the respective closepacked monolayer capacities. So we come to the conclusion that these compounds disperse spontaneously onto the surface of the support to form submonolayers more often than monolayers.

12

YOU-CHANG XIE A N D YOU-QI TANG

I

20

25

30

28"

FIG. 9. XRD patterns of HgClz on active carbon of a specific surface (1000 rn'/g). (a) Active carbon. (b) Fresh mixture of 0.13 g HgCI7/g active carbon. (b') Sample b after holding at 2S"C for I h. (b') Sample b after holding at 25°C for 24 h.

G. THENATURE OF SPONTANEOUS MONOLAYER DISPERSION

So far we have elaborated on the phenomenon of spontaneous dispersion of oxide or salt as monolayer or submonolayer onto the surface of supports with highly specific surfaces. At this point one may well inquire into the nature of this interesting phenomenon. It is only natural for us to think about, first of all, the questions concerning the origin of spontaneity, the dispersion of oxide or salt as monolayer or submonolayer instead of multilayer, and the ubiquity of the phenomenon. Monolayer dispersion is a spontaneous process. Thermodynamics would require that a spontaneous process should proceed with diminishing free enthalpy G or AG < 0. Normally, a process that disperses a substance in a crystalline state as a monolayer or submonolayer, if not as amultilayer, onto the surface of a support would gain in entropy. If this process is energetically not so unfavorable as to reverse its trend, the free enthalpy would decrease and so occurs the spontaneity. Otherwise, the process of a crystalline substance dispersing as monolayer onto the surface of a support would not proceed at all. In the case of an ionic compound such as oxide or salt dispersed on the surface of y-Al,O, or Ti02, the surface bond between the monolayer and the surface of support is usually strong enough to make the entropy effect

MONOLAYER DISPERSION OF OXIDES A N D SALTS

A

h

A

13

a a'

b'b

A

c'

15

20

25

30

35

z 8"

FIG. 10. XRD patterns of some nitrate and chloride hydrates on y-AlzO, of a specific surface (170 m2/g). (a) Fresh mixture of0.21 g Fe(NO,), . 9H20/g y-AlzO,. (a') Sample a after holding at 30°C for 89 h. (b) Fresh mixture of 0.16 g CO(NO,)~. 6H20/gy-Al,O,. (b') Sample b after holding at 30°C for 73 h. (c) Fresh mixture of 0.16 g Ni(NO,), . 6H20/g y-Al?O,. (c') Sample c after holding at 30°C for 89 h. (d) Mixture of 0.18 g CoCI, . 6H20/g y-Al,O,. (d') Sample d kept at 70°C for 78 h. (e) Mixture of 0.18 g NiClz. 6H20/g y-AI2O3.(e') Sample e kept at 70°C for 78 h.

a determinative factor. This accounts for the widespread occurrence of monolayer or submonolayer dispersion in these systems. As is shown in Fig. 1 1 , molecular compounds such as naphthalene, sulfur, borneol, and succinic acid can also disperse spontaneously onto the surface of -y-A1203.But so far we have never succeeded in dispersing metals such as Sn, Bi, Pb, and Zn onto the surface of y-Al,O,. Relevant XRD patterns are shown in Fig. 12. In these cases the dispersion process is expected to be energetically too unfavorable to proceed. With reference to the values of coverage (EIC) given in Table 11, we would not attach much significance to their minor deviation from unity. However, a drop in EIC of one to two orders of magnitude becomes observable when a support with a highly specific surface is used, and would amount to a configurational entropy that would loosen the energetic requirement for the process of monolayer dispersion accordingly. At pres-

14

YOU-CHANG X l E A N D YOU-QI TANG

TABLE 11 Monolayer Dispersion Capacities for Various Systems Dispersion capacity (g/100 m2) Compound

Close-packed estimate ( C )

Experimental value ( E )

Coverage (EIC)

MOO, CUCI, CUCI NaCl KCI CSCl La20y CdO" HgO" MgO" CUO" W0;l NiO"

0.12 0.10 0.15 0.085 0.11 0.25 0.27 0.3 I 0.53 0.10 0.19 0.21 0.18

On yA1203 0.12 0.077 0.095 0.064 0.056 0.085 0.28 0.20 0.067 0.070 0.041 0.21 0.12

NiO"

0.18

MOO, W0;l ZnO" NiO"

0.12 0.21 0.20 0.18

MOO, NiO"

0.12 0.18

0.12 0.11

I .o 0.61

HgC12

0.21

On active carbon 0.02

0.10

1 .o 0.77 0.59 0.75 0.51 0.34 1 .o 0.65 0.13 0.70 0.30

Reference 11-14, 16 1 1 , 19 II,20 18

0.66

34 22 22 33 32 24 32-34

0.50

33

0.23 0.09

23 25 37

1 .o

On T ~ A I ~ O ,

0.090 On silica gel

0.028 0.018 0.22 0.002 On Ti02

1.1 0.01

23

' Samples prepared by impregnation method.

ent no detailed information in connection with the surface structure of these systems has been made available. But one may well imagine a checkerboard pattern of cations and anions on the surface of supports, similar to that in an ionic solid. Now we can talk about the role played by the heat treatment and the mechanism of monolayer dispersion. If the process is to proceed to its

MONOLAYER DISPERSION OF OXIDES A N D SALTS

15

d'

A

c

d

c'

f

-

b

b'

-

A

15

n

20

-

-a

A

25

30

a'

35 28"

FIG. 1 I . XRD patterns of some molecular compounds on y-Alz03of a specific surface (170 mz/g). (a) Mixture of 0.10 g naphthalene/g y-A1203.(a') Sample a after a heat treatment at 40°C for 60 h. (b) Mixture of 0.10 g sulfur/g y-Alz03.(b') Sample b after a heat treatment at 100°Cfor 16 h. (c) Mixture of 0.10 g borneol/g y-AlzO,. (c') Sample c after a heat treatment at 40°C for 24 h. (d) Mixture of succinic acid/g y-Alz03. (d') Sample d after heat treatment at 120°C for 60 h.

limit, which is set by thermodynamics or characterized by an equilibrium state, sufficient time is always needed. Suitable temperature is also desirable. The process would not proceed fast enough at a temperature too low to overcome kinetic resistance. However, the temperature cannot be unduly high either, because at high temperatures oxide or salt may react with the bulk of the support or the support may lose its highly specific surface. For instance, a suitable temperature for the system MoO,ly-Al,O, is between 350 and 450°C. Below 300"C,the dispersion process would take an unduly long time. At 550°C or higher, MOO, would react with the bulk of y-A120, to form Al,(MoO,), and the specific surface of the y-Al,O, support would also begin to diminish. As a rule, the smaller the particles, the more thorough the mixing, then the more readily this process proceeds and the less time it takes to reach its limit. Factors facilitating diffusion would certainly help to alleviate the kinetic resistance. In this special process, ion pairs or molecules move from the surface of the crystalline substance to that of the support with the highly specific surface, either through the vapor phase or directly. However, we have found that direct migration across particles seems to be far more important. In fact, the surface is a much more fluid medium

16

YOU-CHANG XIE A N D YOU-QI TANG

d'

d C'

n c

b'

b

a'

a I

25

30

35

40

2 ea

FIG. 12. XRD patterns of some low-melting-point metals supported on y-Al,O, of a specific surface (170 m2/g). (a) Mixture of 0.10 g S d g y-Al,O,. (a') Sample a after a heat treatment at 350°C for 7 h. (b) Mixture of 0.15 g Bi/g y-AI2O3. (b') Sample b after a heat treatment at 350" for 7 h. (c) Mixture of 0.10 g Pb/g y-A120,. (c') Sample c after a heat treatment at 350°C for 7 h. (d) Mixture of 0.10 g Zn/g y-A1203.(d') Sample d after a heat treatment at 440°C for 6 h. Sn, Bi, Pb, and Zn melt at 232,271, 327, and 419"C, respectively.

than is the bulk of a solid. The melting temperature T , of a solid is its melting point on the absolute temperature scale. In terms of this temperature, ions start to migrate at about OST, in the bulk, but at about 0.3T, on the surface of a solid. In each case, the respective temperature (OST, or 0.3T,) is referred to as the Tammann temperature (26). The phenomenon of monolayer dispersion described above may well be attributable to solid/solid adsorption. However, for the phenomenon of monolayer dispersion, we wonder whether the analogy between liquid/ solid and solidisolid is so close as to justify borrowing a term such as wetting (27-31).

H. IMPREGNATIONMETHOD Nickel oxide, NiO, has a high melting point, namely, 1900°C. We cannot find a temperature high enough to disperse NiO onto the surface of a y-Al,O, support as a monolayer, but also low enough to keep the surface area of y-Al,O, intact. It is also noteworthy that NiO and y-Al,O, form a spinel phase, NiAI,O,, at 700°C or higher ( I ) .

MONOLAYER DISPERSION OF OXIDES A N D SALTS

1

45

1

1

55

1

,

65

.

1

75

1

I

85

17

I

28"

FIG. 13. XRD patterns of NiO supported on y-Al,O, of a specific surface (205 d i g ) . Samples are prepared by impregnating y-A1203in a solution of Ni(N03)2,then by drying and calcining at 450°C for 2 h. (a) y-Alz03. (b) 0. I5 g NiO/g y-AlzOi. (c) 0.40 g NiOig y-AlzOl. (d) 0.61 g NiO/g y-Al2O3.

We have devised an impregnation method to prepare samples of NiO dispersed as monolayer on the surface of a y-Al,O, support (32-34). Such a sample can be obtained by impregnating y-Al,O, in a solution of Ni(NO,),, followed by drying and calcining the impregnated support. The calcination proceeds at 450°C for 2 h. Different loading of NiO can be achieved by varying the concentration of Ni(NO,), in the solution. XRD patterns for samples of NiO supported on y-Al,O, are shown in Fig. 13. The threshold value derived on the basis of the data from XRD quantitative phase analyses is 0.12 g Ni0/100 m2 of y-Al,O, surface, as shown in Table 11. We have prepared systems such as La,O,ly-Al,O, (34), MgOly-Al,O, (33), CuOly-Al,O, (32, 35), CdOly-Al,O, (22), HgOly-Al,O, (22), WO,/ y-Al,O, (24), WO,/silica gel (25),NiO/silica gel, ZnOlsilica gel, and NiO/ TiO, by the impregnation method. The preparative work was always guided by XRD patterns. Dispersion capacities were derived on the basis of the data from XRD quantitative phase analyses (Table 11). The impregnation method is also applicable to many systems mentioned previously. For example, samples of MOO, dispersed as monolayer on y-Al,O, can be prepared either by subjecting the mixture to a heat treatment at about 400°C or by impregnating the support in a solution of (NH,),Mo,O,, then by subjecting the impregnated support to a calcination at 400°C. The dispersion capacities derived for samples from the two different methods are in good agreement. However, the impregnation method left smaller particles of residual crystalline oxide or salt that give rise to broader peaks in the XRD pattern. O N INTERNAL SURFACE OF ZEOLITE I. DISPERSION

Zeolites are known to possess a tremendous amount of internal surface. It is only natural for us to extend our work to this special class of materials.

18

YOU-CHANG XIE AND YOU-QI TANG

20

30

40

20”

FIG.14. XRD patterns of NaCI/NaY zeolite. (a) Nay. (b) Mixture of 0.25 g NaWg Nay. (b’) Sample b after a heat treatment at 550°C for 24 h. ( c ) Mixture of 0.50 g NaCl/g N a y . ( c ‘ ) Sample c after a heat treatment at 550°C for 24 h.

We are pleased that the approach and the methodology developed for supports such as y-Al,O, are also applicable to zeolites (35).For example, when a mixture of NaCl (mp 8OlOC) and a zeolite, namely, Nay , is subjected to a heat treatment at 550°C for 24 h, the XRD peaks of NaCl will disappear or recede depending on whether the loading of NaCl is lower or higher than the threshold value, as is shown in Fig. 14. The threshold value derived from a plot of the residual amount of crystalline NaCl versus the total amount of NaCl in each sample, as shown in Fig. 15, is 0.39 g NaCl/g Nay. This value corresponds to a dispersion capacity of 10.5 NaCl “molecules”/sodalite cage in the NaY zeolite. In view of the fact that a sodalite cage can accommodate only one NaCl “molecule,” most of the NaCl “molecules” are dispersed on the wall of the supercages. On the basis of the close-packed monolayer capacity of NaCl (0.085 g/100 m2) taken from Table I1 and the BET surface area of NaY zeolite (800 m2/g), we estimate the utmost monolayer capacity at 0.68 g/g Na y, which is reasonably higher than the experimental value of 0.39 g NaCllg N a y , because in our calculation we have neglected heterogeneity of the internal surface of the zeolite. We have tried other oxides or salts and succeeded in dispersing most of them spontaneously on the internal surface of zeolite by employing a

MONOLAYER DISPERSION OF OXIDES AND SALTS

19

Total NaCl g / g NaY

FIG. 15. Relationship between the residual amount of crystalline NaCl and the total amount of NaCl in NaCUNaY samples prepared by heating the respective mixtures at 550°C for 24 h.

suitable heat treatment. Table 111 gives the qualitative results we have obtained. The dispersion capacities in terms of “molecules”/supercage as well as gramfgram zeolite are given in Table IV. For high-melting-point oxides such as LaOC1, ZnO, and CuO, the impregnation method was used. Rabo and co-workers (36) used almost the same impregnation method to prepare the adducts formed by salts with Y-zeolite. However, they added one more step using water to wash away all the salt “molecules” in supercages, and got adducts with only one “molecule” of salt left in each sodalite cage. 111.

Effects of Monolayer Dispersion

Oxides or salts in a monolayer state and in their crystalline state behave differently in many respects. Effects of monolayer dispersion show up in spectra as well as in the properties of the oxides and salts. A N D AUGER ELECTRON SPECTROSCOPY A. X-RAYPHOTOELECTRON

Because X-ray photoelectron spectroscopy (XPS) is a surface-sensitive technique, oxide or salt dispersed on the surface of a support as monolayer will give an XPS signal much stronger than that given by the respective mixture of oxide or salt and the support. This prediction has been borne out well by XPS studies on systems such as MoOJTiO,, Mo03/y-Al,0,, Mo0,/Si02, WO,/yAI,O, , WO,/SiO,, ZnO/SiO,, and CuClly-Al,O, by Gui et al. (21, 24, 25, 37, 38).

20

YOU-CHANG X I E A N D YOU-QI TANG

TABLE 111 Spontuncoiis Dispcwion of Oxides und Salts onlo Internul Sirrfuce of Zeolites Heal treatment Zeolite NaY

13X

5A

Compound NaCl CUCl Sb?O, CUClZ NiCI, . 6 H z 0 Ni(NO,), . 6 H I 0 CUCl Hglz NiCI? . hH,O KH5C,OdU CuCl CuBr

Hdz 4A

3A

ZSM-5

KHSC,Od" NaCl CUCI CuBr HglI CUCI CuBr Nal Sb@,

Melting point ("C) 80 I 422 636

498 57 422 259 422 504 259 80 I 422 504 259 422 504 65 I 656

"C

Hours

550 350 450 350 15 30 350 35 30

24 24 24 24 24

230

350 350 130 230 550 350 350 I30 350 350 400 500

6

24 48 24 24 24 24 28 24 24 24 24 28 24 24 24 2

Potassium hydrogen phthalate.

Curves a and b in Fig. 16 represent, respectively, the XPS peak intensity ratios of ZMo3d/lTi2p as a function of the content of MOO, in its mixtures with TiO, of a specific surface 60 m2/gbefore and after a 24-h heat treatment at 400°C. After the heat treatment the XPS peak intensity ratio becomes much higher as is shown by curve b. Obviously, the heat treatment has caused MOO, to disperse as monolayer and to give a stronger XPS signal. The turning point at 0.073 g MoO,/g TiO, in curve b corresponds to a dispersion capacity of 0.12 g MoO,/I00 m2 of the support surface, which is in good agreement with the XRD value in Table 11. Samples of MoO,/TiO, prepared by an impregnation method using a solution of (NH,),Mo,O, have been studied recently by Quincy et al. (39). They obtained a similar plot, which allows us to obtain the same monolayer dispersion capacity, namely, 0.12 g MoO,/lOO m2 of TiO, surface.

MONOLAYER DISPERSION OF OXIDES A N D SALTS

21

TABLE IV Dispersion Capacilies for OxidelZeolite or SulrIZeolite

Dispersion capacities Zeolite

Compound

NaY

g/g Zeolite

''Molecules ' 'lsupercage

0.39 0.52 0.93 0.41 0.22 0.06 0.41 0.30 0.55 0.20

10.6 8.4 5.1 3.4 4.3 1.2 1.9 8.7 9.5

NaCl CuCl LaOCI" ZnO" CUO" CuCl NaCl CuCl

13X 4A

ZSM-5

Sb203

-

" Samples prepared by impregnation method. Impregnation, respectively, in solutions of LaCI, , Zn(N0,)2, and CU(NO,)~ followed by calcination at 550"C, respectively, for 24, 4 and 4 h.

In the system ZnO/SiO, (37), the binding energy of Zn2p3,, for monolayer-dispersed ZnO (E,, = 919.4 eV) has been found to be about 1.2 eV higher than that for the crystalline ZnO (& = 918.2 eV). In CuClly-Al,O, (38),the binding energy of Cu'Lp,,, for monolayer-dispersed CuCl (Eb = 934-935 eV) is about 3 eV higher than that for crystalline CuCl (Eb = 931.8 eV). Samples with lower loading of CuCl give higher values of binding energy. The shape of the valence band or Cu3d spectra is quite

0.40

cu"

.-

t-

Y

'

-r5

0.20 -a 0.10

0.20

0.30

0.40

versus the content of MOO, in its mixtures FIG.16. XPS peak intensity ratio IMo3dllTi2p with a TiO, support of a specific surface (60 m21g)before (a) and after (b) a heat treatment at 400°C for 24 h.

22

YOU-CHANG XIE AND YOU-QI TANG

0

10

20

8.E. I e v 1

FIG. 17. XPS spectra of valence band Cu3d. (a) Crystalline CuCI; (b) 0.12 g CuCl monolayer dispersed on I .0 g y-Al,O, of a specific surface (343 d i g ) ; B.E., binding energy.

different from that of crystalline CuCl, as is shown in Fig. 17. The Auger electron spectroscopy (AES) spectra of monolayer-dispersed and crystalline CuCl are also distinctly different (38).

B. STATICSECONDARY IONMASSSPECTROSCOPY Static secondary ion mass spectroscopy (SSIMS) uses a very low primary ion current to bombard a relatively large area of the surface and detects the secondary ions ejected from it. This is indeed a surface-sensitive technique. Benninghoven (40, 41) has suggested that if we find an exponential time dependence of a secondary ion current generated from a surface, we can conclude that the “parent” is only present in the first layer or monolayer. Guo et al. (38) and Huang et al. (42) have studied SSIMS for the systems NiOly-AI20,, Mo03/y-A120,,P,05/y-A1203,and CuCl/y-A1203by using a fixed Ar’ ion beam to bombard a spot on the surface of samples. When the concentrations of NiO, Moo3, P,O,, and CuCl are lower than their respective monolayer dispersion capacities, their signals decline exponentially with time. Figure 18 shows the SSIMS results for CuCl that is monolayer dispersed on the surface of y-A1203.

MONOLAYER DISPERSION OF OXIDES A N D SALTS

I

2 thin)

23

3

FIG. 18. From a sample of 0.12 g CuCl monolayer dispersed on 1 g y-AI,O, of a specific surface (343 m2/g);SSIMS signal is found to decline exponentially with time. Art ion beam: 1 keV and 0.6 nA.

The signal I,, decreases exponentially with time, as is verified by the , versus time in Fig. 18. straight line of In Z

C. ION-SCATTERING SPECTROSCOPY Low-energy ion-scattering spectroscopy (ISS) is a sensitive technique for the outmost atomic layer. Knozinger and co-workers have studied the systems MoOJy-Al,O, (43)and MoO,/TiO, (44) by JSS. Figure 19 shows the ISS intensity ratios ZMo/ZAi as a function of He+ ion fluence (bombardment time) for a mixture of 7.6% Mo0,ly-Al,O, heated at 450°C in a moist oxygen stream for various time periods. For a mechanical mixture, the independent distribution of Mo and A1 makes the intensity ratio ZMo/ZA, almost unaffected by the fluence. The steep rise of the initial intensity ratio IMoIIAI after a prolonged calcination indicates that more MOO, has dispersed onto the surface of y-Al,O,. However, after calcining the sample for more than 5 h , the initial intensity ratio remains unchanged. This indicates that an utmost limit of dispersion has been reached. During the ion bombardment, the surface atoms sputter off and the intensity ratio ZMolIA, decreases. Similar results were reported for the same sample heated in a dry oxygen stream, and also for the system MoO,/TiO, heated under 450°C either in a moist or dry oxygen stream. All these ISS studies are consistent with the spontaneous monolayer dispersion model.

24

YOU-CHANG XIE A N D YOU-QI TANG

1.0~-

Ob

5 10 15 FLUENCE 1x10'' H i / c m 2 l

3

FIG.19. ISS M o and Al peak intensity ratios as a function of He* ion fluence. A mixture of 7.6 wt% Mo0,ly-AI20, was subjected to a heat treatment at 450°C in a moist oxygen stream for 0 h (a), 2 h (b), and 5 h (c), respectively.

D. RAMANSPECTROSCOPY Monolayer-dispersed oxide or salt should give a Raman spectrum different from its crystalline counterpart. This statement has been confirmed by recent work in several laboratories (39, 45-50). Knozinger and co-workers (46) have reported that a mixture of MOO, and y-A120, gives the Raman spectrum shown in Fig. 20a, but if the mixture is heated at 450°C in air for 30 h, the spectrum changes (Fig. 20b). This change has been interpreted as conversion of Moo3 into surface poly(mo1ybdate). Similar results (45-50) have been observed for the systems MoO,/TiO,, MoO,/SiO,, WO3/y-AI2O3,V20S/yA120,,and VzOs/ TiO,, for example. It is noteworthy that water vapor can promote the dispersion of MOO,, WO,, and V2OSonto the surface of the supports and influence their Raman spectra markedly (45, 46, 48, 49). Systems Mo03/y-A120,, Mo0,/Si02, and MoO,/TiO, have been subjected to a quantitative Raman study in our laboratory by Zhao et af. (50). They introduce KNO, into the samples as an internal reference. By measuring the 820-cm-' peak of crystalline MOO, and 1050-cm-' peak of KNO,, they were able to get the Raman intensity ratios Z820/1,0so and to derive the content of crystalline MOO,. Figure 21 shows the content of the crystalline MOO, as a function of the total content of MOO, in the system MoO,/TiO,. The threshold value in Fig. 21 corresponds to 0.115 g MOO,/ 100 m2 surface of TiO,. It is in good agreement with the XRD value of the monolayer capacity in Table 11. We have done similar work on the systems

MONOLAYER DISPERSION OF OXIDES AND SALTS

25

FIG.20. Raman spectra: (a) Mechanical mixture of 9 wt% Mo03/y-A1203; (b) the same mixture after a heat treatment at 450°C for 30 h in an open crucible.

FIG.21. Content of crystalline MOO, derived from quantitativeRaman study as afunction of total content of MOO, in the system MoO,/TiO,.

26

YOU-CHANG XIE A N D YOU-QI T A N G

210

400

600 Wavelength(nm1

800

FIG.22. U V diffuse reflection spectra of the system CuCl/y-A1203.(a) Mixture of 13 wt% CuCl/y-A120,. (b) The same mixture after a heat treatment at 350°C for 4 h. (c) Crystalline CUCI.

MoOJy-AI,O, and MoO,/SiO, (50). Hercules and co-workers have reported a similar result for the system MoOJTiO, (39).

E.

U V DIFFUSE REFLECTANCE SPECTROSCOPY

The UV diffuse reflectance spectrum discerns between the monolayerdispersed and the crystalline states of oxide or salt. For example, Gui et al. (20) have reported that a mixture of 13 wt% CuClly-Al,O, gives a U V diffuse reflectance spectrum with a band edge of crystalline CuCl at 390 nm, as is shown in Fig. 22a and c. After heating the mixture at 350°C for 4 h, we can see an obvious change in the U V diffuse reflectance spectrum: the band edge of CuCl has shifted to 360 nm and its absorption intensity has increased enormously because of the multiple reflection effect between light and dispersed CuCI. A similar result has also been observed for the system NiO/y-AI,O, (51) by Liu et al. F. EXTENDED X-RAY ABSORPTION FINESTRUCTURE SPECTROSCOPY Because a monolayer-dispersed oxide or salt on a support with a highly specific surface usually is present in considerable quantity, the sensitivity of extended X-ray absorption fine structure (EXAFS) analysis is good

27

MONOLAYER DISPERSION OF OXIDES AND SALTS

FIG.23. Radial distribution around Ni. (a) Crystalline NiO. Two monolayer-dispersed samples: (b) 0.10 g NiOig y-AI2O, ; (c) 0.20 g NiOig y-A120,. Monolayer dispersion capacity is 0.26 g NiOig y-A1203.See text for discussion of peaks.

enough to give us useful surface-structural information for the monolayerdispersed phase.

1. NiOIy-Al,O, An EXAFS study on the system NiOly-Al,O, (52) has been reported by Jin et al. Figure 23 shows the radial distribution function around Ni obtained by Fourier transform from EXAFS spectra of two samples in the system NiOly-Al,O, and a sample of crystalline NiO. In the sample of crystalline NiO, the coordination numbers of Ni-0 and Ni-Ni are 6 and 12, respectively, and their respective distances are 2.09 and 2.96 Taking phase shift into account, we assign peak 1 in Fig. 23 to Ni-0 and the highest peak 2 to Ni-Ni. We note in passing that the concentrations of NiO in samples b and c in Fig. 23 are still below the monolayer dispersion capacity stated in Table 11, and in these two samples the Ni-Ni peak (peak 2) is lower than the Ni-0 peak (peak l), and is incomparably lower than the Ni-Ni peak in sample a. A quantitative estimation of the Ni-Ni peak in b and c gives, respectively, for the coordination number of Ni-Ni, values of 2 or less and 3.5, in contrast to the value of 12 for the crystalline

A.

28

YOU-CHANG XIE A N D YOU-QI T A N G

I

FIG.24. Radial distribution functions around Cu. (a) CuCI,; (b) CuO; (c) 0.064 g CuC1,/ g y-Al,O, . Specific surface of y-A1203is 170 m2/g.

NiO. These results provide additional evidence that NiO in samples b and c has, indeed, dispersed as monolayer onto the surface of y-Al,O,. It is also worth notice that the coordination number of Ni-Ni decreases markedly with the content of NiO in the NiOly-Al,O, samples. It may well suggest that the monolayer-dispersed NiO distributes randomly and does not form islands or patches on the surface of y-Al,O,. A quantitative estimation of the Ni-0 peak (Fig. 23, peak I , in samples b and c) gives a value of 4-5 for the coordination number of Ni-0, in contrast to the value of 6 in crystalline NiO. These findings fit the monolayer dispersion model very well. The Ni-0 distances in samples b and c derived from EXAFS are very close to that in crystalline NiO. 2. CuCl2ly-Al203

The EXAFS for the system CuC1,ly-Al,O, (53) has been worked out by Cai et al. Figure 24 shows the radial distribution functions around Cu in crystalline CuCl, and CuO, and in a monolayer-dispersed CuC1,ly-Al,O,. The main peak in a is assigned to the Cu-Cl peak of crystalline CuCI,. The main peak in c is not as sharp as the Cu-CI and Cu-0 peaks, respec-

MONOLAYER DISPERSION OF OXIDES A N D SALTS

29

...

-3

-2

-I

0

I

2

3

Velocity i m m / s e c l FIG.25. Mossbauer spectra of FeCI,. (a) FeC1, in the crystalline state. (b) Monolayerdispersed FeCI, in the mixture 0.15 g FeCl,/g y-AlzO3 after a heat treatment at 180°C for 16 h. (c) Mixture 0.62 g FeCl,/g y-Al2O3after heat treatment. Specific surface of y-A1203is 233 m2/g.

tively, in a and b, and occupies an intermediate position. This suggests that CuCI, in the sample CuC1,l-y-Al,O, has dispersed onto the surface of y-Al,O,, and Cu2+ ions are coordinated by 02-as well as C1- ions. In fact, the broad peak results from an overlap of the Cu-Cl and Cu-0 peaks. Cai has also reported the observation that this broad peak shifts toward the Cu-0 peak as the content of CuCl, in the sample decreases.

G. MOSSBAUER SPECTRA We have studied the effects of monolayer dispersion of Mossbauer spectra for the system FeC1,ly-Al,O, (537). Figure 25a shows the Mossbauer spectrum of FeCI,. It contains only one line: no quadrupole should be expected when the Fe3+ions have a spherically symmetrical d5electron distribution and also a symmetrical coordination by six C1- ions. However, if FeCl, is mixed with y-A1,03 and heated at 180°C for several hours, its Mossbauer spectrum changes markedly. Figure 25b is a spectrum of

30

YOU-CHANG XIE A N D YOU-Q1 TANG

...:,.#",-

+. ....._.\., .- ......

,,,.I

\. ..

*.(,%

.........%. .....,.-...

.........

dL+'.*

-6

-3

s .

.,.w*

....... ...;............ ,,.-.:.:'. ..:,..,...... .': ..

,::;:, '

-9

.. %..,*.,,. '.:.'......' -.

'. .izh.-.*:.. . :

\..._I.-

..

r.

0

3

6

.b

9

Velocity1 mmisecl

FIG.26. Mossbauer spectra of Fe20,. (a) cu-Fe203in the crystalline state. (b) Dispersed Fe203 in 0.05 g Fe,O,/g y-A1203.(c) 0.29 g Fe203/gy-A1203. Samples are prepared by impregnation method. Specific surface of y-A1203is 233 m2/g.

0.15 g FeCl,/g y-Al,O, after a heat treatment at 180°C for 16 h. It displays a quadrupole-split double line. The monolayer-dispersed FeCl, provides a new environment for the Fe3+ ions with an electric field gradient that gives rise to a quadrupole splitting. Figure 25c is a spectrum of the sample containing excessive FeCl,, namely, 0.62 g FeCl,/g y-Al2O3.It displays the doublet as well as the singlet line, and the latter obviously belongs to the residual crystalline FeCl,. Figure 26 shows the Mossbauer spectra of Fe,O,. Fe3+ cations (spectrum a) in crystalline a-Fe,O, give a ferromagnetic spectrum consisting of six lines due to the magnetic hyperfine splitting. The quadrupole-split double line in spectrum b may well belong to the monolayer-dispersed Fe,O, instead of "superparamagnetic fine particles" (536)in the sample, 0.05 g Fe,O,/g y-Al,O,. The sample in spectrum c, 0.29 g Fe,O,/g y-Al,O,, gives a Mossbauer spectrum. It is the superposition of a ferromagnetic spectrum of six lines and a quadrupole-split double line, respectively, due to the crystalline and monolayer-dispersed Fe,O,. We can derive from the relative intensities of these two sets of lines the monolayer capacity of 0.12 g Fe,O,/g y-Al,O,, in good agreement with the XRD value ( 5 3 ~ ) .

H. TRANSMISSION ELECTRON MICROSCOPY A N D HIGH-ENERGY ELECTRON DIFFRACTION A support carrying a monolayer or submonolayer of oxide or salt should give a transmission electron microscopy (TEM) micrograph and a high-

MONOLAYER DISPERSION OF OXIDES AND SALTS

31

energy electron diffraction (HEED) pattern of the support. This has been verified by Zhang et al. (32). Figure 27b is a TEM micrograph of a NiO/ y-Al,O, sample with the NiO content below its monolayer capacity. We cannot distinguish it from the TEM micrograph of the support y-Al,O, shown in a. Figure 27b' is a HEED pattern for the same sample of NiO/ y-Al,O,. It gives only diffuse diffraction rings indistinguishable from the HEED Gattern of the amorphous support y-Al,O, shown in a'. However, once the content of NiO in the sample NiOly-Al,O, exceeds the monolayer capacity, crystalline NiO particles and their diffraction spots will show up, respectively, in TEM micrographs and in HEED patterns, as shown in Fig. 27c and c'.

I. DIFFERENTIAL THERMAL ANALYSIS The monolayer-dispersed oxide or salt exhibits no peak at its melting point in the differential thermal analysis (DTA) pattern of the sample. Figure 28a shows a DTA pattern of a CuCl,/y-Al,O, sample without a peak at the melting point of CuCI, (498"C), because the CuCl, content of the sample is still below the monolayer dispersion capacity. However, b and c do exhibit peaks at 498°C because of the residual crystalline CuCl,, which is present either due to low specificity of the surface of the support or to the high content of CuCl,.

J.

SURFACE ACIDITY

Zhao et al. (54) have found that an acidic oxide dispersed on the surface of a support often causes the acidity of the surface to increase and attains the highest value as the content of the oxide reaches the dispersion capacity. Figure 29 shows the relation between the acidity of the system MOO,/ SiO, and the loading of MOO, in the sample. The turning point in the plot corresponds to monolayer dispersion threshold obtained previously by XRD. Before this turning point is reached, the mole ratio between acid sites and MOO, is high and ranges from 0.5 to 1. It strongly suggests that MOO, is dispersed as monolayer. Other systems, such as MoO,/y-Al,O, (54), MoO,/TiO,, P,O,/y-Al,O, ( 5 3 , WO,/SiO,, and WO,/TiO, (56),have also been studied. Each of them gives a similar plot, with the turning point corresponding to the monolayer dispersion capacity. It seems to us that the surface acidity for these systems should unmistakably originate from the monolayer-dispersed oxides on the surface of the supports.

32

YOU-CHANG XIE A N D YOU-QI T A N G

FIG.27. TEM micrographs and HEED patterns: (a) and (a’) y-A1203;(b) and (b’) 0.15 g NiO/g y-A1203; (c) and (c‘)0.30 g NiO/g y-A120,. Samples are prepared by the impregnation method. Their monolayer dispersion capacity is 0.25 g NiO/g y-Al,03.

MONOLAYER DISPERSION OF OXIDES A N D SALTS

33

0

G b

"C

500

I20

FIG.28. DTA patterns of the system CUC~~/AI,O, : (a) 0.10 g CuC12/gy-Al,03 ; (b) 0.10 g CuCl,/g a-AI203 ; (c) 0.40 g CuCl,/g y-Al,03.

K. ADSORPTION

It is a well-known fact that compounds such as CuCl and AgCI, either in solution or in a crystalline state, can form m-complexes with alkene or carbon monoxide. We then speculate on the opportunity for monolayerdispersed salt or oxide of copper or silver to adsorb alkene or carbon monoxide by forming surface 7~ complexes. Gui et al. (20) have used the temperature-programmed desorption (TPD) technique to study the adsorption of ethylene on CuClly-Al,O, at room temperature. They have found that after a heat treatment at 350°C for 4 h, the mixture of CuCl and y-Al,O, displays an increase of three orders of magnitude in the adsorption capacity for ethylene. Figure 30 shows the ethylene adsorption capacity as a function of CuCl content in the CuClly-Al,O, samples. The ethylene adsorption capacity increases with the content of CuCI and reaches the highest value at the point near the monolayer dispersion threshold. Similar plots have been reported by Zhao et al. (57) in the investigation of ethylene adsorption on CuOly-Al,O, and by Duan et al.

I

I

I

10

20

I

30

Mo0,Wt %

FIG.29. Relation between acidity and MOO,content in Mo03/Si02.Monolayer dispersion capacity is 0.12 g MoO,/g SiO,.

34

YOU-CHANG XIE A N D YOU-QI T A N G

FIG.30. Ethylene adsorption as a function of CuCl content in the system CuCl/y-AlzOl.

(58) studying CO adsorption on CuOly-Al,O,. Systems such as CuBrlyA1,0,, AgClly-AI,O,, Ag,O/y-Al2O3, CuCl/SiO,, and CuCllzeolites can also adsorb ethylene and carbon monoxide efficiently (12). All these results are in accord with the predictions derived from the monolayer dispersion model.

IV. Applications to Heterogeneous Catalysis The active components of many commercial supported heterogeneous catalysts are oxides or salts. Even for many metal catalysts, the precursors of metallic particles are also oxides or salts in some dispersed form. Hence the preparation of heterogeneous catalysts is deeply concerned in one way or another about the dispersion of oxides or salts on support surfaces. Furthermore, promoters or additives added to heterogeneous catalyst systems are also oxides or salts. Therefore, the spontaneous monolayer dispersion of oxides or salts on supports with highly specific surfaces as a widespread phenomenon will find extensive application in heterogeneous catalysis. Examples illustrative of this viewpoint are cited in the following sections.

A . PREPARATION OF HIGHLY ACTIVE MONOLAYER-DISPERSED CATALYSTS Rendering the active component into the monolayer-dispersed state is an important measure to undertake to enhance the activity of a catalyst. A striking example is the highly active catalyst system for

MONOLAYER DISPERSION OF OXIDES A N D SALTS

35

polymerization of ethylene and propylene. The main active component of the early Ziegler-Natta catalyst, invented in the 19SOs, was TiCl, crystallites. A new type of polyethylene and polypropylene catalyst based upon monolayer-dispersed TiCI3 on a MgCl, support was developed in the 1970s (9, 60). By employing this new catalyst system, the polymerization activity can be multiplied by several orders of magnitude. Hence, the catalyst residues in the polymer products are reduced to such an extent as to dispense with the deashing procedure and significant cost reduction becomes realizable. As was mentioned before, spontaneous monolayer dispersion of oxides and salts occurs so readily that the preparative method for many commercial oxide or salt catalysts can be advantageously improved and simplified. In the preparation of a hydrodesulfurization catalyst it has been discovered that the catalyst obtained by mixing MOO, with y-Al,O, and then calcining at 450°C has the same activity as the one made by the impregnation method (12). Similar results have been observed in the preparation of an oxychlorination catalyst (CuCl,/y-Al,O,) (12) and a catalyst (ZnAcz/active carbon) used for the synthesis of vinyl acetate from acetylene (59). In both cases each catalyst, which was prepared by mixing the active component with the support and calcining at an appropriate temperature, has nearly the same catalytic property as the one made using an impregnation method. It was interesting to note that even at ambient temperature monolayer dispersion of the active component sometimes occurred. For example, the catalyst HgClJactive carbon, used for the synthesis of vinyl chloride from acetylene, can be made by mixing HgCI, with active carbon and then simply leaving the mixture at ambient temperature for several hours. In short, a highly active monolayer-dispersed catalyst can be prepared by calcining a mechanical mixture of the active component and the support, if the melting point of the former is not too high. A dry method of this kind is highly recommended, its advantages over conventional wet methods being simplicity and economy. Monolayer dispersion capacities of the oxides and salts are extremely useful for determining the correct recipes for preparing catalysts. In commercial catalysts, for example, the hydrodesulfurization catalyst Mo0,iy-Al,O,, the oxychlorination catalyst CuClJy-Al,O,, the HgCI,/ active carbon catalyst for synthesis of vinyl chloride from acetylene, and the ZnAc,/active carbon catalyst for synthesis of vinyl acetate from acetylene, the contents of the active components are all near or less than their respective monolayer dispersion capacities. In fact, if the content of the active component exceeds its monolayer dispersion

36

YOU-CHANG XIE A N D YOU-QI TANG

FIG.31. Effect of La203 content in Ni/Laz03/y-Al,03catalysts on their XRD peak 11 1 of Ni. (a) 0.10 g Ni/g y-Alz03.(b) 0.10 g NU0.05 g La203/gy-A1203.(c) 0.10 g NU0.30 g La203/ g ?-A1203.

threshold, the surplus will exist in a crystalline state and thus cannot bring its function into full play.

B. MONOLAYER-DISPERSED OXIDESAND SALTS AS SURFACE MODIFIERS Oxides or salts may play the part of promoters or additives in a heterogeneous catalyst. Their function in various catalyst systems can vary widely and is too complicated to have been adequately elucidated so far. However, we have found that they often operate as surface modifiers. For example, adding a small amount of rare earth oxide such as La,O, to the methanation catalyst Nily-Al,O, can significantly increase its activity and thermal stability (34). We have proved that La,O, is monolayer dispersed on the y-alumina support and that the surface of the support is modified by the La,03 monolayer. This is related to the improved performance of this catalyst. Figure 31 shows the effect of La,O, content in the catalyst Ni/La,O,/yA1203on the XRD peak 11 1 of Ni metal. Obviously, the peak is leveled down or the size of the Ni crystallites decreases markedly as the addition of La203to 1.0 g of y-Al,O, increases from 0.05 to 0.30 g. This catalyst was prepared by impregnating y-Al,O, with a Ni(NO,), and La(NO,), solution. By drying and calcinating the impregnated support we then obtained a system of monolayer-dispersed NiO and La,O, on

MONOLAYER DISPERSION OF OXIDES AND SALTS

37

#

43

, 44

45

46

47

20'

FIG.32. XRD peak 11 1 of Ni. (a) 0.15 g NiOig y-Al,O,. (b) 0.60 g NiOig y-Alz03 after reduction. The monolayer dispersion capacity is 0.25 g NiOig y-AI,O,.

y-Al,O,. This system was reduced selectively with H, to give the methanation catalyst Ni/La20,/y-AI,0,. In the presence of La,O, on the surface of y-Al,03, NiO would be reduced to Ni crystallites of smaller size and higher thermal stability, and the improved performance of the methanation catalyst may well be due to the modified surface. Similar effects of rare earth oxides have been observed on the catalyst Ptly-Al,O, (61).

C . PREPARATION OF SUPPORTED METALPARTICLES FROM MONOLAYER-DISPERSED OXIDE As was stated previously, metal cannot disperse as a monolayer onto catalyst supports. However, oxide precursors of metals can monolayer disperse on supports, and supported metal particles can be prepared from the monolayer-dispersed oxide by reduction. Preparation of the methanation catalyst Nily-Al,O, proceeds from NiO/ y-Al,O, by reduction, and the latter is prepared by the impregnation method discussed in Section 11. Different loading of NiO can be achieved by varying the concentration of Ni(NO,), in the solution. Zhang et al. (32) studied Ni crystallites prepared from NiOly-Al,O, of different NiO loading. The sizes of Ni crystallites from samples a and b in Fig. 32, 0.15 g NiO/g y-Al,O, and 0.60 g NiO/g y-Al,O,, are distinctly different. In sample a all NiO is monolayer dispersed and gives very small particles of Ni after reduction. However, in sample b, more than half of the NiO is in the crystalline state and the crystalline NiO produces Ni crystallites that are characterized by a distinct XRD peak. Figure 32 shows the XRD peak 111 of Ni crystallites. The result is consistent with electron microscope observations.

38

YOU-CHANG XIE A N D YOU-QI TANG

300

400

500

600

t('Ci

FIG.33. The sample contains 0.70 g NiOig y-Alz03.(a) The TG curve recording the loss of weight in temperature-programmed reduction. (b) The DTG curve.

It has also been observed by Zhang et al. that NiO in the monolayerdispersed state is much more difficult to reduce than is NiO in the crystalline state. Figure 33 shows the thermogravimetry (TG) and differential thermogravimetry (DTG) curves recorded in temperature-programmed reduction of a sample containing 0.70 g NiO/g y-Al,O,. In this sample, about one-third of NiO is monolayer dispersed and the other portion is in the crystalline state. They behave differently during the process of reduction. The weight loss in the range 300-390°C on the TG and the distinct peak at 337°C on the DTG curve are due to the reduction of NiO in the crystalline state. The second weight loss shown on the TG in the high-temperature range 420-550°C and the leveled peak in this range on the DTG curve can be ascribed to the reduction of the monolayer-dispersed portion of NiO. It is worth mentioning that spontaneous monolayer dispersion is also a very useful scientific basis underlying the process of regeneration of deactivated metal catalysts. Supported metal catalysts may sinter during use at elevated temperatures. Sintering will cause the metal catalyst to lose initial activity, and in order to recover it one has to find an effective way to redisperse the metal on the catalyst support. Applying what we have learned from our studies on spontaneous monolayer dispersion to

MONOLAYER DISPERSION OF OXIDES A N D SALTS

39

this problem, we suggest first to oxidize the metal to oxide, then to disperse the latter as monolayer onto the support surface, and finally to reduce the monolayer-dispersed oxide into highly dispersed metal particles. It was reported by Yao et al. (62) that for the catalyst Ptly-Al,O, of low Pt content, redispersion of Pt metal on the support surface can be effected by heating the catalyst in an oxygen atmosphere at 500°C and then by reducing the oxide in H, at 300°C. Most likely, the Pt metal is oxidized to PtO, and the latter is spontaneously dispersed on the surface of y-Al,O,. Then the reduction of the monolayer-dispersed PtO, in H, leads to the formation of small supported Pt particles. The catalyst Irly-Al,O, can also have its Ir metal redispersed when it is heated in oxygen (63, 64). Redispersion through an oxidation-reduction cycle as described previously is, indeed, an effective way to regenerate supported metal catalysts that have been deactivated because of sintering, and the underlying principle is spontaneous monolayer dispersion.

D. MODIFICATION OF INTERNAL SURFACE OF ZEOLITES Zeolites possess an enormous internal surface and a system of pores and channels. Many relevant properties of zeolites such as acidity, composition, area of the internal surface, and geometry of the pore and channel system can be modified by dispersing oxides or salts on the zeolite internal surface. Such modification would certainly cause the catalytic behavior of the zeolite to alter. Modification of zeolites by means of monolayer dispersion of oxides or salts is different from that by ion exchange. By ion exchange, only cations are introduced into the zeolites, but in the case of monolayer dispersion anions disperse along with cations onto the internal surface. Monolayer dispersion of oxides and salts proceeds more readily than ion exchange. It can be done efficiently by both dry and wet methods. Lee et al. (65) reported that the selectivity in connection with the formation of p-xylene from methanol and toluene can be improved significantly by adding Sb203to the HZSM-5 zeolite catalyst. Without this additive, the xylene produced in the catalytic reaction of methanol and toluene over HZSM-5 at 400°C is an equilibrium mixture containing 23.1% p-xylene. However, the para selectivity approaches 100% if the reaction proceeds over the modified HZSM-5 catalyst prepared by calcining a mixture of HZSM-5 AND Sb,O, in air at 500°C for 2 h. Figure 34 shows how the para selectivity of the methylation of toluene depends on the content of Sb203in the catalyst. The para selectivity reaches the maximum as the content of Sb,O, in the catalyst approaches

40

YOU-CHANG XIE A N D YOU-QI TANG

FIG.34. Para selectivity in toluene methylation with methanol (a) and amount of undispersed or crystalline antimony oxide in terms of XRD intensity (b) versus the total amount of Sb,03 added to HZSM-5 zeolite.

its monolayer threshold. This improvement in selectivity can be ascribed to the geometrical constraint imposed by the narrower pores of the modified HZSM-5, which makes it sterically inhibitive to form m- and o-xylene molecules. Recently Mikae et af. (66) showed that NaY zeolite impregnated with NaCl solution can be used as a para selective catalyst for chlorination of chlorobenzene, which is indeed another example of zeolite modified advantageously by means of monolayer dispersion of salts.

V. Concluding Remarks Spontaneous monolayer dispersion of compounds on supports is a widespread phenomenon and displays many unique effects. The principles involved have applications not only to heterogeneous catalysis, but also to materials science and other related fields. The theoretical and practical aspects of this phenomenon appear to offer prospects that should not be overlooked. Studies in connection with this phenomenon are continuing in our laboratory. ACKNOWLEDGMENTS

The authors acknowledge China's National Natural Science Foundation for generous support of this work, which is a part of the major project "Structural Chemistry and Molecular Design." We are grateful to all our colleagues who have contributed toward a better understanding of the phenomenon of spontaneous monolayer dispersion. Thankful acknowledg-

MONOLAYER DISPERSION OF OXIDES AND SALTS

41

ments are due to Professors Zi Gao and Lin-lin Guei for their timely help in one way or another. We are also grateful to Ge Yang, Qiang Xu, Xian-ping Xu, and Xiao-ming Pan for their help in preparing the manuscript. The senior author wishes to dedicate this review t o the memory of Zi-qing Huang and Ying Fu, the late professors of physical chemistry, Peking University. REFERENCES 1 . Pott, G. T., and Stork, W. H. J., in “Preparation of Catalysts” (B. Delmon, P. A.

Jacobs, and G. Poncelet, eds.), p. 537. Elsevier, Amsterdam, 1976. 2. Russell, A. S., and Stokes, J. J., Jr., Ind. Eng. Chem. 38, 1071 (1946). 3. Lipsch, J. M. J. G., and Schuit, G. C. A., J . Catal. 15, 174 (1969). 4. Massoth, F. E., J . C a r d 30, 204 (1973). 5. Giordano, N., Bart, J. C. J., Vaghi, A., Castellan, A., and Martinotti, G., J . Catal. 36, 81 (1975). 6. Buiten, J., J . Catal. 10, 188 (1968). 7. Sonnemans, J., and Mars, P., J. Catal. 31,209 (1973); Fransen, T., Van Berge, P. C., and Mars, P., in “Preparation of Catalysts” (B. Delmon, P. A. Jacobs andG. Poncelet, eds.), p. 405. Elsevier, Amsterdam, 1976. 8. Stork, W. H. J., Coolagem, J. G . F., and Pott, C. T., J. Catal. 32, 497 (1974). 9. Xie, Y. C., Gui, L. L., Liu, W. Q., Bu, N. Y., and Tang, Y. Q., Sci. Sin. (Chin. Ed.) p. 665 (1979); Sci. Sin. (Engl. Ed.) 22, 1045 (1979). 10. Xie, Y. C., Zhang, H. X., and Wang, R. H., Sci. Sin. (Chin. Ed.) p. 337 (1980); Sci. Sin. (Engl. Ed.) 23, 980 (1980). 11. Xie, Y. C., Yang, N. F., Liu, Y. J., and Tang, Y. Q., Sci. Sin., Ser. B (Chin. Ed.) p. 673 (1982); Sci. Sin., Ser. B (Engl. Ed.) 26, 337 (1983). 12. Xie, Y. C., Gui, L. L., Liu, Y. J., Zhao, B. Y., Yang, N. F., Zhang, Y. F., Guo, Q. L., Duan, L. Y., Huang, H. Z., Cai, X. H., and Tang, Y. Q., Proc. I n t . Congr. Catal., 8th 5, 147 (1984). 13. Xie, Y. C., Gui, L. L., Liu, Y. J., Zhang, Y. F., Zhao, B. Y., Yang, N. F., Guo, Q. L., Duan, L . Y., Huang, H. Z., Cai, X. H., and Tang, Y. Q.. in “Adsorption and Catalysis on Oxide Surface” (M. Che and G. C. Bond, eds.), p. 139. Elsevier, Amsterdam, 1985. 14. Liu, Y. J., Xie, Y. C., Ming, J., Liu, J., and Tang, Y. Q . , J . Catal. (Chinu) 3, 262 (1982). 15. Chung, F. H., J. Appl. Crystalfogr. 7 , 526 (1974). 16. Liu, Y. J., Xie, Y. C., Li, C., Zou, Z. Y., and Tang, Y. Q., J. Catal. (China) 5 , 234 ( 1984). 17. Wells, A. F., “Structural Inorganic Chemistry,” 4th Ed., p. 259. Oxford Univ. Press, London and New York, 1975. 18. Yang, J. Y., Yan, H. G., and Liu, J. B., Acta Pet. Sin. (Pet. Process. Sect.) (China) 3, 51 (1987). 19. Xie, Y. C., Cai, X. H., Gui, L. L., and Tang, Y. Q., Acta Phys. Chim. Sin. (China) 2, 519 (1986). 20. Gui, L. L., Guo, Q. L., Xie, Y. C., and Tang, Y. Q., Sci. Sin., Ser. B (Chin. Ed.) p. 1 (1984); Sci. Sin., Ser. B (Engl. Ed.) 27, 445 (1984). 21. Gui, L. L., Liu, Y. J., Guo, Q. L., Huang, H. Z., and Tang, Y. Q., Sci. Sin., Ser. B (Chin. Ed.) p. 509 (1985); Sci. Sin., Ser. B (Engl. Ed.) 28, 1233 (1985). 22. Liu, Y. J., Dong, B. H., Xie, Y. C., and Tang, Y. Q., Acta Phys.-Chim. Sin. (China) 2, 470 (1986).

42

YOU-CHANG X I E A N D YOU-Ql TANG

23. Liu, Y. J., Xie, Y. C., Xie, G., and Tang, Y. Q., J . C u t d . (Chinu) 6, I01 (1985). 24. Liu, Y. J., Zhao, M., Guo, Q. L., Gui, L. L., Xie, Y . C., and Tang, Y. Q..Actu Chiin. Sin. (Chin. Ed.) 43, 728 (1985); Actu Chirn. Sin. (Engl. Ed.) p. 313 (1985). 25. Liu, Y. J., Wu, J . P., Guo, Q. L . , Gui, L. L., and Tang, Y . Q.,J . Curul. (Chinu) 8, 14 (1987). 26. Thewlis, J., “Encyclopaedic Dictionary of Physics,” Vol. 7. p. 171, Pergamon, Oxford, 1962. 27. Haber, J., in “Surface Properties and Catalysis by NonMetals” (J. P . Bonnelle. B. Delmon, and E. Derouane, eds.), p. 1 . Reidel, Dordrecht, Netherlands, 1983. 28. Haber, J., f l 4 W Appl. Chem. 56, 1663 (1984). 29. Haber, J., Proc. In/. Congr. Carul., 8th 5, 85 (1984). 30. Haber, J . , Machej, T . , and Czeppe T., Surf. Sci. 151, 301 (1985). 31. Haber, J., Proc. I n / . Symp. Reoct. Solids, I O r h , 1984 Part A, p. 14 (1986). 32. Zhang, Y. F., Xie, Y. C . , Zhang, Y., Zhang, D. L., and Tang, Y . Q., Sci. Sin., Sur. B (Chin. Ed.) p. 805 (1986). 33. Zhang, Y. F., Xie, Y. C., Xiao, N . H., Han, W . , andTang, Y. Q.,Petrochem. Tec,hnol. (Chinu) 14, 141 (1985). 34. Xie, Y . C., Qian, M. X . , and Tang, Y. Q.,Sci. Sin., S e r . B (Chin. Ed.) p. 788 (1983); Sci. Sin., Ser. B (Engl. Ed.) 27, 549 (1984). 35. Xie, Y . C., Zhang, 0. W., Yang, Ge, Xu. X. P., and Tang, Y . Q., Proc. ChinoJ p n . - U . S . Sytnp. Cutul., 3th, Xiamen, Chinu B-03 (1987). 36. Rabo, J. A.. Poulsma, M. L., and Skeels, G. W., Pruc. In/.Con$ Zeolite p. 98 (1352). 37. Zhao, B. Y., Wu, N. Z., Gui, L. L . , Zhang, L., Bai, N. B., Xie, Y. C., and Tang, Y. Q . , Sci. Sin., Ser. B (Chin. Ed.) P. 281 (1985); Sci. Sin., Ser. B (Engl. Ed.) 29, 579 (1986). 38. Guo, Q. L.. Huang, H. Z., Gui, L. L., Xie, Y. C . , and Tang, Y. Q., Acru Phys-Chim. Sin. (Chinu) 3, 389 (1987). 3Y. Quincy, R. B . , Houalla, M . , and Hercules, D. M . , J . C u r d . 106, 85 (1987). 40. Benninghoven, A , , Surf. Sci. 35, 427 (1973). 41. Benninghoven, A , , Surf. Sci. 53, 596 (1975). 42. Huang, H. Z., Zhao, B. Y., Guo, Q. L., Gui, L. L., and Tang, Y. Q . , J . Curd. (Chinci) 8, 151 (1967). 43. Margraf, R., Leyrer, J . , Knozinger, H . , andTaglauer, E . , Surf: Sci. 189/190,842(1987). 44. Margraf, R . , Leyrer, J., Taglauer, E., and Knozinger, H., Reacr. Kine/. Card. Leu. 35, 261 (1987). 45. Knozinger, H., personal communication. 46. Leyrer, J., Zaki, M. I., and Knozinger, H., J . P h y s . Chem. 90, 4775 (1986). 47. Stampfl, S. R., Chen, Y., Durnesic, J. A . , Niu, C., and Hill, C. G . , Jr., J . C ar d. 105, 445 (1987). 48. Stencel, J. M., Makovsky, L. E.. Sarkus, T. A , , De Vries, J., Thomas, R., and Moulun, J . A . , J . Cural. 90, 314 (1984). 49. Payen, E., Kasztelan, S . , Grimblot, J . , and Bonnelle, J. P., J. Raman Spectrosc. 17, 233 (1986). 50. Zhao, B . Y . , Xu, Q . , Xie, Y . C., and Yang, X. C., Cham. J . Chitz. Uniu. 11,54 (1990). 51. Liu, Y. J., Yang, J. P., Gui, L. L., and Tang, Y. Q., Acru Pet. Sin. (Per. Process. Sect.) 3, 40 (1987). 52. Jin, X. L., Cai, X. H., Ge, Z. H., Xie, Y. C., and Tang, Y. Q., Actu Phys.-Chem. Sin. (Chinu) 5, 206 ( I 989). 53. Cai, X . H., Xie, Y. C., and Tang, Y. Q.,f r o c . EXAFS Symp., Chinu, 1987. 530. Xie, y. C., Xu, X . P., Wu, G. B., and Tang, Y. Q., Proc. Chin. Congr. Cutul., 4th. Tiunjin I-E-29 (1988).

MONOLAYER DlSPERSlON OF OXIDES A N D SALTS

43

5%. Yashioka, T., Koezuka, J., and Ikoma, H., J . Cutul. 16, 264 (1970). 54. Zhao, B . Y . , Kang, Z. J., Li, C., Xie, Y. C.. and Tang, Y. Q., J . C u d . (Chinu)6 , 219 (1985). 55. Zhao, B. Y., et ul., this lab., unpublished results. 56. Yamaguchi, T., Tanaka, Y., and Tanabe, K., J . Cutul. 65, 442 (1980). 57. Zhao, B. Y., Zhang, Y. F., Duan, L. Y . , Xie, Y. C., and Tang, Y. Q., J. Cutul. (Chinu) 3, 101 (1982). 58. Duan, L. Y., Du, Z. M., Wang, X . G., and Xie, Y. C., Nut. Gas Chem. Techno/. (China) 10(6), 7 (1985). 59. Chen, S., Li, G. Y., Wang, Y. Q., and Yu, B. L . , J . C u d . (Chinu)7, 155 (1986). 60. Bu, N. Y., Xie, Y. C., Bai, C. L., Sun, M. F., andTang, Y. Q., Chem. J . Chin. Univ. 3, 542 (1982). 61. Yang, J. Y., and Swartz, W. E., Spectrosc. Lett. 17, 331 (1984). 62. Yao, H. C., Sieg. M., and Plummer, H. K., Jr., J . Catul. 59, 365 (1979). 63. Fiedorow, R. M . J., Chahar, B. S., and Wanke, S. E., J . C u r d 51, 193 (1978). 64. Mcvicker, G. B., Garten, R. L., and Baker, R. T. K., J . C a r d . 54, 129 (1978). 65. Lee, G. Y., and Zhao, J. C . , Petrochem. Techno/. (China) 16, 266 (1987). 66. Mikae, T., Sekizawa, K., Hironaka, T., Nakano, M., Fujii, S., and Tsutsumt, Y., New Deu. Zeolite Sci. Techno(., Proc. Int. Zeolite Conf., 7th p. 747 (1986).

This Page Intentionally Left Blank

ADVANCES IN CATALYSIS, VOLUME 31

Catalysis by Supported, Unsupported, and Electron-Deficient Palladium ZBIGNIEW KARPINSKI Institute of Physical Chemistry Polish Academy of Sciences 01-224 Wursuw. Poland

1.

Introduction

Among transition metals that are commonly used as heterogeneous catalysts, palladium occupies a special place. Being widely used for many years by organic chemists as a catalyst for selective hydrogenation of multiple C-C bonds and selective reduction of functional groups, it has been the subject of several compilations and guides ( 1 4 ) . Even though the versatility of palladium in hydrogenation reactions is generally recognized by chemists, the explanation of its “magic” catalytic properties, however, is still very far from being satisfactory ( 5 ) . In the process of elucidation of catalytic properties of metals, increasingly more is expected from the application of surface science methods, such as low-energy electron diffraction (LEED), ultraviolet photoelectron spectroscopy (UPS), X-ray photoelectron spectroscopy (XPS), extended X-ray absorption fine structure spectroscopy (EXAFS), and Auger electron spectroscopy (AES). It should be stressed that surface studies in relation to heterogeneous catalysis have gained importance during the past years. However, a considerable progress in utilization of these methods depends not only on their wider use (they are rather expensive) but, first of all, on the advancement in the development of surface science methods for tackling such difficult targets as very complex surfaces of metals dispersed over carriers. Application of models, such as clean surfaces of single crystals of metals pretreated under carefully controlled conditions, undoubtedly brings us nearer to knowledge about the situation on a metal surface on an atomic scale. On the other hand, the currently insufficient state of development of sophisticated techniques (theory and experiment) rather rarely allows us 45 Copyright 0 1990 by Academic Press, Inc. All rights of reproduction in any form reserved.

46

ZBIGNIEW KARPINSKI

to probe successfully surfaces of complex metallsupport catalysts (6-8). Insulating granular carriers are still not well suited for sophisticated surface investigations. The origin of the frequently observed chemical shifts in the binding energy of core levels in the case of very small particles dispersed over an insulating support is still debated (initial-state effect versus final-state effect; see discussion in Section 111,A).Another difficulty that may arise during interpretation of XPS or EXAFS spectra of lowconcentration-metal loaded, supported catalysts is the problem of detectability. Beard and Ross (9) analyze this problem for the case of a 1-wt% Pt/TiO, catalyst, which might form a Pt-Ti complex as the result of reduction in H, at 400-500°C. In principle, XPS is capable of determining the large changes in valency of the Ti species (from + 4 to near zero valency; large chemical shifts in core levels), and the Ti K-edge EXAFS shift is also pronounced. However, because the fraction of all Ti atoms undergoing this change is small, the signal from them is either too low to be detected [by Ti(2p); XPS] or is lost in the background of the signal from the majority of the Ti atoms present (in Ti K-edge EXAFS). Respective electronic changes for Pt are small (XPS) and Pt L-edge EXAFS is dominated by Pt-Pt coordination because Ti is a weaker scatterer of electrons compared with Pt. The use of models (e.g., Pt and TiO, phases in an “inert” matrix) solves the problem of detectability, but direct applicahility to real metallsupport catalysts may be lost, On the other hand, chemical methods of characterization of supported metals, such as conveniently chosen catalytic reactions or chemisorptions, may still play the role of a useful probe. In the present review article we shall show such an approach to the study of Pd catalysts. As a result of consistent investigation of a reaction course andlor chemisorption on intentionally modified Pd catalysts, starting from simple flat surfaces of unsupported metal, through a gradual introduction of some morphological changes on the metal surface, the observed variations are hoped to be attributed to these morphological variations. Therefore, an interpretation of the catalytic behavior of a supported Pd catalyst [also in the state of strong metal-support interactions (SMSI)’ ( l o ) ]would be accomplished in terms of separable factors. An obvious condition is that any selected probe reaction must be sensitive to changes in the metal surface. Having in consideration the problem of chemical probing of Pd/support systems, it is relevant to know the catalytic properties of unsupported palladium. Three types of reactions will be considered in this article: I In this work the term SMSI is not restricted only to the cases when the reduced support creeps over the metal particle, so that the latter becomes partially covered with metal cations and anions of the support (e.g., Ti’+ and 0 2 -from TiO?), but confines also the cases when the interaction leads to the formation of intermetallic compounds ( e g , Pd3Si).

CATALYSIS BY PALLADIUM

47

reactions of unsaturated hydrocarbons with H, , conversions of alkanes, and reaction of CO with H,. The knowledge about reaction mechanisms is recalled only in such cases when it is important for further treatment. During discussion on catalytic properties of Pd h p p o rt systems we shall consider only these problems, which result from the chemical probing. In particular, we shall assume an attitude toward the problem of structure sensitivity of reactions catalyzed by palladium, palladium and support interactions, and evolution of palladium surfaces in the course of catalytic reactions. During the last 15 years scientific literature has brought forward substantial evidence on the existence of electron-deficient palladium in carefully reduced Pd/support (the support is alumina or silica-alumina) catalysts. The importance of this fact for the course of some catalytic reactions has been either suggested or more or less strongly established. We shall refer to the existence of electropositive Pd species in the situations well documented by the results of physical techniques such as XPS or electron spin resonance (ESR). Although the undertaking of writing a review on electron-deficient Pd in supported catalysts is justified for its own sake, we present a view that this knowledge may be essential in the discussion of various aspects of the kinetic results obtained for supported Pd catalysts. This will help us in the interpretation of structure sensitivity of some chemical reactions catalyzed by palladium, in chemical probing, and in the interpretation of evolution of palladium catalysts during a reaction. 11.

Catalysis over Unsupported Palladium

The exceptional catalytic behavior of platinum (versus other Group VIII metals in alkane rearrangement reactions) has stimulated a great number of studies involving various model forms of platinum catalysts, such as flat, stepped, and kinked surfaces of Pt (11-16). These investigations identified different active centers on the Pt surface. On the contrary, although the versatility of palladium in catalytic hydrogenation is widely appreciated, the scientific literature does not have many examples of results of kinetic studies using Pd single crystals. A noteworthy exception is the reaction of trimerization of acetylene to benzene, thoroughly investigated in the 1980s by three groups of scholars (17-19). Attaching here more attention to the reactions with participation of H, , we note that only very recently single crystals of Pd have been employed as catalysts (20, 21). Therefore, in the discussion concerning catalysis ovbr unsupported palladium, we shall not confine our review only to catalytic results obtained on well-characterized surfaces of Pd single crystals, We shall in-

48

ZBIGNIEW KARPINSKI

clude also the available data on palladium films and powders, especially when their texture was determined directly (e.g., by electron or X-ray diffraction) or indirectly (e.g., from experiments using other probes, such as CO adsorption). In several cases a surface modification of palladium was realized by adopting some specific specimen pretreatment, such as an annealing of initially rough surfaces, a low-temperature treatment in hydrogen to produce the P-PdH phase and subsequent decomposition of this phase, or an oxidation/reduction cycle. It has been demonstrated that all these pretreatments produce definite changes in surface topography, so they can be used as convenient means of modification of Pd samples. In other words, certain procedures can be employed as suitable recipes. Of course, a reference to the origins of such recipes is obligatory, especially if one does not apply any physical method for determining surface changes. It is obvious that further experiments with real single-crystal surfaces of Pd, at carefully controlled conditions, are needed for verification of the results obtained with Pd films and powders. For that reason, the following discussion should be regarded as a tentative one.

A. REACTIONS OF UNSATURATED HYDROCARBONS WITH H, First results concerning hydrogenation of 1,3-butadiene on Pd(ll1) and (110) single crystals were published by Massardier et al. (20) in 1988. The more open (110) face was approximately eight times more active than the (111) one. The selectivity for the half-hydrogenation (butene formation) was 100%on the two samples. Upon exposing to hydrogen at -1 atm, an irreversible reconstruction of Pd single crystals (due to the formation of a P-PdH phase) occurred with the production of macroscopic defects. Such samples exhibited somewhat higher activities presumably because of an increase of the roughness factor (not determined). Initially very poor selectivity (50%) of a highly dispersed Pd/SiO, catalyst increased up to 90% after some hours of the reaction. At the same time, the activity decreased to the level characteristic of the Pd(l1 I ) sample. The authors attribute this activity decline to the deactivation of very active Pd sites with a very low coordination number. These sites are plentiful on the fresh catalyst surface, but they become deactivated in consequence of hydrogenolysis of reacting hydrocarbon species in the reaction course. Less active plane atoms maintain their activity. The explanation given above recalls earlier speculations presented by Ledoux et al. (22, 23). The authors assigned a different role to various surface centers on a palladium film in the reactions of isomerization, hydrogenation, and deuterium exchange of butenes. The authors defined

49

CATALYSIS BY PALLADIUM

active sites A, B, and C on the surface of palladium (Fig. 1). These sites are identified with the normal low-index surface atoms (A sites), or with some defects arising in a rough surface: edge atoms (B sites) and corner or adatoms (C sites). Any modification of the Pd surface alters the relative amounts of sites A, B, and C. The authors proposed that A sites (low Miller-index plane atoms) with a single free valency available are responsible for direct cis-trans isomerization, without hydrogen incorporation. Isomerization according to the Horiuti-Polanyi mechanism proceeds on B sites. Both kinds of sites (A and B) exist on Pd films annealed at 470°C. On the other hand, C sites, plentiful on a rough Pd film condensed at O"C, are responsible for the isomerization via .rr-allylic species. Figure 1 also presents the reactions catalyzed on A, B, and C sites. Metal-catalyzed hydrogenation of olefins proceeds via the HoriutiPolanyi mechanism ( 2 4 , generally accepted since the 1930s (Scheme 1). A recent Fourier transform infrared (FTIR) study of the adsorption and hydrogenation of ethylene on 3% Pt/A1,0, showed that both forms of the adspecies suggested in step (b) in Scheme 1 [ap-diadsorbed ethane (25) and .rr-bonded ethylene (26)] occur and react rapidly with H, (27). Slowly reacting residues, such as ethylidyne (*=t-CH,) or related ethylidene (*=CH-CH,) species, were also suggested to take an active part in the hydrogenation (28), but there is strong evidence that this is not the case, at least for Pd catalysts (29).Therefore, the role of more strongly adsorbed adcarbene and/or adcarbyne species is seen mainly as self-poisoning (30). It should be mentioned here that Pd is a very poor metal catalyst in forming aa-diadsorbed (like ethylidene) or aaa-triadsorbed (like eth ylidyne) species in the presence of excess hydrogen (see later). In the case of chemisorption of acetylene on Pd( 11 l), high-resolution electron energy loss spectroscopy (HREELS) data suggest that ethylidyne coexists with vinylidene (*=C=CH,) at this surface (31). In their ultraH2

+ 2*

___*

H2C=CH2

-

2H*

+ 2*

H2C-CH2

*I

1*

and/or H2C=CH2

.1

ap-diadsorbed ethane a-bonded ethylene

H2C-CHz

1 1

+H *I

H3C-CH2

+H I

I

-1

,CH, H2C

-

C2H6

+ 2*

+ 2*

SCHEME I.

(a) (b)

Sites A :

Sites B

S i t e s C:

P

w + D M - D

- y- pt

P

H-M-U

ti-M-D

P w+DM-D

P

DM-O

FIG. 1 . Geometry of different active sites on the Pd surface and reactions of olefins catalyzed by these sites. (Taken from Ref. 22.)

CATALYSIS BY PAL LA D I U M

51

high-vacuum (UHV) study of the hydrogenation of acetylene on the Pd( 1 1 1) surface, Tysoe et ul. (32) come to the conclusion that vinylidene may be the precursor to ethylene formation. On the other hand, ethylidyne was suggested to be involved in the direct hydrogenation of acetylene to ethane (33, 34). Recently Hub and Touroude (35)showed that during hydrogenation of I-butyne, the Pd surface is covered by strongly but reversibly adsorbed 1 -butyne molecules (adcarbenes or adcarbynes), which may lead to direct hydrogenation. These species are in equilibrium with vinylic adsorbed species, which play a role of a reactive intermediate to butenes. Maier et a f .( 5 , 2 1 )found very pronounced surface structure sensitivity in the hydrogenation of hex-2-yne catalyzed by Pd single crystals and foils. The Pd( I 11) plane was much more selective (87%) toward cis-2-hexene formation than was the Pd( 110) (only 37%). It should be stressed here that the reaction was assisted by the P-PdH phase formation. The palladium hydride is catalytically active, and its reversible transformation into palladium during the reaction is responsible for morphological changes of an active surface [reconstruction to Pd(l1 l)]. More importantly, these transformations produce very fine cracks in Pd crystallites. As a result, the originally smooth Pd( I 1 1) surface turns rough, generating defect sites more active (and less stereoselective) in the half-hydrogenation of 2-hexyne. It looks that in using these rather meager data on catalytic hydrogenation on well-defined surfaces of unsupported Pd for probing more complex Pd surfaces (as we have in supported catalysts), one has to be very cautious and not forget the possible effect of Pd hydride formation. The /3-PdH may be more easily formed in massive Pd specimens (36; but see also 37, 38), whereas for highly dispersed Pd catalysts we can deal with the absence of the hydride phase (see Section IV,A).

B. REACTIONS OF ALKANES WITH H, To the best of our knowledge, there are no published data on the reactions of alkanes with H, over palladium single crystals. Deuterium exchange reactions with various alkanes over palladium films were extensively investigated by Kemball et u f . (39-42). Because those results were already discussed in several excellent review articles (43-46), we do not intend to dwell upon analyzing all the data at greater length. We only specify here the basic characteristics of Pd films exhibited in D,/alkane reactions:

1 . Palladium is a very active metal (cf. other Group VIII metals) in the D, exchange of all hydrocarbons when the so-called ap process [step (c)

52

ZBIGNIEW KARPINSKI

and its reverse, Scheme 11 is possible. Such a situation exists in the case of extensive exchange of n-alkanes and so-called one-set exchange of cyclanes. 2. The aa-diadsorbed species (adcarbenes) either do not exist or are in much lesser proportion than other species on a Pd surface during exchange reactions, even at relatively high temperatures (>200"C) (47).It should be noted that these reactions on nickel or rhodium occur via aa-diadsorbed species (39, 45). 3. It is rather generally agreed (46) that a very extensive two-set exchange of cyclanes on Pd occurs via a roll-over mechanism proposed by Burwell (48, 49). 4. The ay-diadsorbed species are also not plentiful on Pd surfaces in the presence of excess hydrogen (deuterium) (50). Recent data on Pd( 1 11) epitaxially oriented films confirm the characteristics mentioned above (51, 52). A stepwise (simple) mode of D, exchange with methane (at 300°C) and neopentane (at 230-275°C) predominates over Pd(ll1). On the other hand, two-set exchange of cyclopentane with D, is very extensive, suggesting that no special coordinatively unsaturated Pd sites are needed for the roll-over process to occur (52, 53). Hydrogenolysis, isomerization, and cyclization reactions proceed at higher temperatures than do those usually employed in D,/exchange studies. Therefore, the aa-, ay-, andlor aay-adsorbed species, previously ruled out as important intermediates in D, exchange on Pd, may now play some role in alkane conversions. The consensus is that these multiply bonded species are important intermediates in the hydrogenolysis of alkanes (54, 55). Pd is the poorest among platinum metals in forming aadiadsorbed complexes (50, 56). Therefore, not surprisingly Ir, Rh, and even Pt are more active than Pd in many hydrogenolyses (54-58). However, the presence of low-coordinated atoms (such as sites B and C in Fig. 1) may bring about the formation of adcarbenes or adcarbynes, and, in consequence, a C - C bond breaking. Most often, selective terminal demethylation occurs [n-pentane (59),n-hexane (60-62), or n-heptane (57)l. The mechanism of bond-shift isomerization of alkanes is still a subject of controversy. Several suggested mechanisms involved an intermediacy of multiply bonded species [Anderson-Avery (63), Muller-Gault ( 6 4 , and Garin-Gault (65) mechanisms] (Fig. 2). More recent data (51, 52, 66-68) seem to support the 1,2 bond-shift isomerization involving monoadsorbed alkyl species (69). Figure 2 shows these mechanistic ideas. The use of deuterium as an isotopic tracer in a study of mechanisms of alkane conversions is generally unrealistic, because at a comparable temperature a D,/alkane exchange is considerably faster than an alkane

53

CATALYSIS BY PALLADIUM

Me

'1 \

HzC

I

\CH

Me

HC

/ \

Me

HzC

-----*

H zc

\*-\

Me 'C'

Me

HzC d

CH z

Me

HzC

!I' /*\

I

Me

'*' FIG.2. Mechanisms of bond-shift isomerization on metal catalysts. (a) Anderson-Avery mechanism (63). (b) Muller-Gault mechanism for isomerization of neopentane on Pt (64). (c) McKervey-Rooney-Samman mechanism (69). (d) Muller-Gault mechanism of isomerization of isobutane on Pd (64).(e) Clarke-Rooney mechanism (46).(0 Garin-Gault mechanism (65).

54

ZBIGNIEW KARPINSKI

conversion. Thus the distribution of deuterium in the products of alkane conversion reaches an equilibrium level and, as a result, has no value for mechanistic considerations. However, in a few cases it was possible to employ successfully deuterium as a tracer for obtaining mechanistically meaningful information (66-68). Karpinski and Guczi (68) succeeded in reducing the rate of the deuterium exchange sufficiently to study the mechanism of isomerization of neopentane over Pt films. This was achieved by drastic sintering of a Pt film for 3 h at 527”C, in D, (4 torr). Appreciable amounts of monodeuteriated isopentane formed in the reaction products strongly support a mechanism involving an adsorbed alkyl species as the intermediate in the isomerization of neopentane, for example, the McKervey-Rooney-Samman mechanism (Fig. 2c) (46, 69). On the other hand, drastically sintered Pd(ll1) films still gave overly rapid D,/neopentane exchange to isomerization (51, 52). However, when the exchange reaction is carried out at temperatures only slightly lower than those at which the isomerization begins, the mode of exchange may give some indirect indications as to the population of adsorbed species at high temperatures. The (1 11)-oriented films of Pd gave good selectivity for isomerization of neopentane at -290°C (7040%) (51, 52). On the other hand, the contribution of multiple exchange (via aa and/ or a y species) was negligible at 230-275°C. Also, methaneil), exchange on Pd( 1 I 1) goes via adsorbed alkyl species in a similar temperature range ( 5 2 ) . The experiments mentioned above show that the chemisorbed carbene species are not plentiful on the surface of the Pd( 11 1) film at relatively high temperatures. Therefore, the reaction of isomerization of neopentane over Pd(l1 I ) would go via an adsorbed alkyl species (69). More recently, Finlayson et al. (66)performed deuterium tracer studies using 2,2,4,4-tetramethylpentane as a model reactant, and proved that the bond-shift isomerization to 2,2,5-trimethylhexane may go via a monoadsorbed intermediate on Pd (as well as on Pt, Ir, and Rh). The authors show that in the mechanisms of “nondestructive” reactions such as cyclization and bond-shift isomerization of alkanes, only one metal atom is required as the catalytic center. In the early 1970s it was known that palladium films isomerize n-butane and isobutane (63, 70) but not neopentane (63).Similarly, only pure cracking of neopentane was observed on Pd/SiO, catalysts (71). The observed inability of Pd to catalyze neopentane isomerization was confirmed in the case of more complicated molecules having a quaternary carbon atom (64). On the other hand, platinum seemed to be the unique metal in isomerizing neopentane. Comparing the catalytic behavior of both metals, Gault came to the conclusion that palladium, due to its well-established propensity toward formation of n-olefinic and n-allylic adspecies, can

CATALYSIS BY PALLADIUM

55

isomerize n-butane or isobutane, but not neopentane (72), whereas Pt may isomerize neopentane either via cwy or metallocyclobutane precursors. On the other hand, if the McKervey-Rooney-Samman mechanism is valid (involving the organometallic complex, very similar to those in the olefin-metal complexes, Fig. 2c), platinum should be a better catalyst than palladium because Pt forms much stronger bonds with olefins than does Pd (46). More recent experiments with sintered Pd films (73) and Pd powders (74, 75) in a static system showed that the activity of this metal in the isomerization of neopentane was small. Although hydrogenolysis always predominated, the initial product distribution (i.e., selectivities) showed nonnegligible (-5%) amounts of isopentane. However, in the case of Pd films, this isomerization did not last long: at a somewhat higher conversion level (-2-3%), their isomerizing activity died off very quickly (73). The possibility of killing the sites active in isomerization by carbonaceous residues is ruled out on the basis of the results of a flow reactor study of n-pentane conversion over Pd/Si02 ( 5 9 ) : hydrogenolysis suffers much more than does isomerization with increasing time on-stream. Returning to the static reactor study with sintered Pd films ( 7 3 , it should be stressed that after reaching a conversion level of -2-3%, no appreciable decline in overall activity (now only pure cracking) has been observed, which suggests an absence of substantial self-poisoning. Therefore, we believe that the continuous decrease of the amount of isopentane in the reaction products in a static reactor experiment is due to secondary hydrogenolysis of isopentane (primary product). This may be explained by a much higher adsorption coefficient of isopentane versus neopentane on a Pd surface. Hence, after getting a nonnegligible concentration of isopentane in the gas phase, its readsorption leads to its cracking (76). In 1980 it was shown that the epitaxially oriented Pd(l1 l)/mica film is very selective in the isomerization of neopentane [selectivity of 70-80% (51, 5 2 ) ] . This fact, confirmed later [(75, 77), when a somewhat lower isomerization selectivity, 56%, was found], was in distinct disagreement with earlier data concerning the isomerizing behavior of palladium (63, 71). It was thought that a very smooth Pd(ll1) surface does not contain highly unsaturated surface sites needed for hydrogenolysis (such as tenters C and/or B in Fig. 1). Further work furnished evidence that the state of surface topography of Pd may influence the course of neopentane conversion (75). As was mentioned, films of Pd/mica or Pd/Pyrex, which had been carefully annealed in H, , catalyzed the isomerization reaction. The same films after reconstruction induced by special treatments lost their good isomerization selectivity, becoming very active in hydrogenolysis. Among these pretreatments is an oxidation (at high temperature) and

56

ZBIGNIEW KARPINSKI

reduction (at mild temperature) cycle, which, according to other literature data (78-80), should produce highly defective Pd surfaces. Also, Pd transformation into the P-PdH phase and its decomposition back to Pd (avoiding excessive heat treatment) should disintegrate larger crystallites, introducing a considerable density of defects (5, 21, 81). As is seen in Fig. 3 both kinds of pretreatments lower the initially high isomerization selectivity down to the level of a few per cent. On the other hand, when these defective specimens have been annealed in H, at -500°C overnight, a high isomerization selectivity level is recovered. The same holds for the catalytic behavior of initially unsintered (after deposition) Pd films that later were subjected to annealing in H, at 500°C for overnight (Fig. 3B). One may argue that Pd powders, which exhibited a modest selectivity for neopentane isomerization (74, should become more selective after annealing at higher temperatures. However, although it has been shown that some increase in isomerization selectivity does appear on annealing [up to -30%, mainly due to a drastic decrease of hydrogenolysis; see Fig. 4, pretreatment (c)], the level of isomerization selectivity characteristic of Pd(ll1) films (70-80%) cannot be achieved (75). Very severe H, pretreatment at 550-700°C of Pd powders produced rather high isomerization selectivity (50-70%), but this selectivity increase could not be exclusively attributed to surface smoothing but rather to the formation of Pd silicide due to the reaction of Pd powder with silica [Fig. 4, pretreatment (d) and (e)]. The silica had originated from the reactor walls and quartz wool plugs protecting the catalyst bed or was intentionally added to the Pd powder (75). If Pd silicide is responsible for high selectivity toward isomerization, then, returning to the results on Pd(ll1) film, one might argue that palladium surface of epitaxially oriented-on-mica films, condensed at 400-500°C and annealed in H2 at this temperature (51, 52, 73, contains impurities originating from the mica base (K, Mg, Al, Si?) and this is a reason for higher isomerization selectivity. Such an explanation should not be ruled out without checking catalytic behavior of real single crystals of Pd, cleaned in UHV. Unfortunately, such experiments are still missing. It should be emphasized that the procedure of preparation of (1 11)-oriented Pd/mica films was adopted from the work of Christmann and Ertl (82), who obtained clean Pd( 111) and Pd-Ag( I 11) films in UHV by annealing Pd/mica deposits at temperatures up to 400°C. LEED indicated that the degree of ( 1 11) surface orientation was very high whereas Auger electron spectra of the films did not show any impurities. There was an important difference in the two preparations. The films of Christmann and Ertl (82) were annealed in UHV, whereas Karpinski, in order to avoid contamination inherently associated with heating in HV conditions, annealed in ultrapure H, (51,52, 75). The difference in chemical environment may have

A

85.2

s 'O0I 80

69.1

67.5

1 .o

ol

iM

Pretreatment of Pd/mico f i l m

(el

Pretreatment of Pd/Pyrex film

Fic. 3. Isomerization selectivity in the reaction of neopentane with H, on palladium films (at -29OoC, unless otherwise stated). (From Ref. 75.) (A) Pd(l1 l)/mica film, after pretreatments in O2 and/or H2 ("C, hours): (a) fresh Pd( 1 1 I), (b) O,, 290, 0.25; H,, 290, 0.6, (c) O,, 500, 0.5; Hz, 500, 13.5, (d) 02,500, 0.5; H2, 300, 0.17, and (e)0 2 , 460, 0.33; Hz, 500, 15. (B) PdiPyrex film: (a) unsintered film (reaction at 268"C), (b) 0,, 284, 0.25; H2, 500, 17.5 (at 285"C), (c) as in pretreatment b, but reaction at 303"C, (d) O , , 300, 0.5; H2, 500, 19 (at 300"C), (e) O z r480, 1 ; H,, 490, 18.5 (at 303"C), and (f) after decomposition of P-PdH (reaction at 298°C).

58

ZBIGNIEW KARPINSKI

s

-

h

Patladium

Pd silicide

80

C

0

L

aJ

t

(01

(bl

(cl

(d)

(e)

Pretreatment of Pd powder FIG. 4. lsomerization selectivity in the reaction of neopentane with H, on Pd powders after various pretreatments (75); pretreatments in O2 and H2 (“C, hours). (a) Pd powder: 0,. 300, 0.25; H 2 , 300, 1 ; H,, 400, 1 (reaction at 286°C). (b) A Pd + S O z physical mixture: 0 2 , 300, 0.25; H,, 500, 2 (reaction at 300°C). (c) Pd powder: 02,300, 0.25; H,, 550, 10 (reaction . 0.25; H,. 700, 12 (reaction at 355°C). (e) A Pd + S O , at 330°C). (d) Pd powder: 0 2 300, physical mixture: O,, 300, 0.25; H,. 550, 15 (reaction at 330°C).

important consequences: in UHV, one may expect diffusion of impurities (e.g., sulfur) into the bulk, whereas heating in H, may cause segregation of impurities onto the surface. Also, it is not absolutely certain whether thick (-100 nm) epitaxial films of Pd on mica and real bulk single crystals of Pd, after cleaning by Ar ion sputtering, are rough on an atomic scale, having only plateaulike regions or microfacets (83). On the other hand, a high-temperature pretreatment in H2 is known to sinter drastically both unsupported (84) as well as alumina-supported Pd (78, 79). Summing up, it is conceivable that both the state of surface topography as well as the presence of other elements on the Pd surface are decisive for its activity in alkane reactions. However, two things seem obvious and unquestionable. First, all these catalytic variations result from the fact that hydrogenolysis, rather than isomerization, is sensitive to changes in

CATALYSIS B Y PALLADIUM

59

surface morphology of Pd. This is not new, because it has been known for many years that alkane hydrogenolysis is “more” structure sensitive than isomerization (85). Second, the apparent controversy on the catalytic behavior of palladium in neopentane isomerization ( 5 1 , 5 2 , 6 3 , 6 5 ,71) may be understood in this way: Pd is cupable of isomerization, and rupture of hydrocarbon adspecies to surface carbenes is very pronounced on defective Pd, but not on smooth Pd(ll1) surfaces (66, 86). Pa51 and TCtCnyi (87) reported that Pd black had a rather pronounced activity in the skeletal rearrangement of 2,2-dimethylbutane. They suggested two types of bond-shift mechanisms: one is responsible for benzene formation under low H, pressure [involving a McKervey-Rooney-Samman type of intermediate (Fig. 2c)l and the other gives skeletal isomers at higher H, pressure [presumably by the Garin-Gault mechanism (Fig. 2f)l. Attaching more significance to hydrogen pressure effects, the authors compared the catalytic behavior of Pd black with that of Pt black and concluded that both metals behave similarly in skeletal reactions of Ch alkanes. Under conditions of relative deficiency of H, (usually, in the kinetic study of conversions of alkanes over metals, the H, : alkane ratio is about 10 or higher) and at higher temperatures, when some dehydrogenation is expected, the chain-lengthening homologation is observed (86). On a Pd film, n-pentane, and to a lesser extent isopentane, can be transformed into benzene. The mechanism for selective multiple homologation was suggested via addition of surface methylene to the unsubstituted vinylic carbon of a-olefin, forming intermediate metallocyclobutanes. The latter may further hydrogenate to produce the next higher homologue or isomerize to the corresponding a-olefin, which in turn may repeat the same reaction (Scheme 2). On the other hand, Sarkany (88) observed that on a Pd catalyst, chain lengthening of small alkanes was commensurable to their isomerization. Therefore, participation of adcarbenes in the isomerization (as well as in homologation) is feasible under hydrogen-deficient conditions. However, RCH=CH*

RCH-CH,

I *=CH2

1

I

-

-

j*-T

RCH2CHyCH2 I

1

RCHlCHzCHi

+ CHI=*

RCH2CH-CH2

etc.

I

*-cH~

1

LR C H ~ C H ~ C H ~ C H ,

SCHEME 2.

60

ZBIGNIEW KARPINSKI

because neopentane neither isomerizes nor “homologates” under these conditions on Pd (88), it is hardly possible to employ this kind of “carbene” mechanism for isomerization of neopentane. Similarly, the homologation reaction route is not a preferred one on platinum, which is the best isomerizing metal.

C . HYDROGENATION OF CO For thermodynamic reason, synthesis of methanol from syngas must be performed at higher pressures. Therefore, although palladium has been found to be an efficient catalyst for CH,OH production (89), in a lowpressure study only methanation can be observed. Also, in the absence of a support, palladium was never found to be a good methanol synthesis catalyst (90-93). Both Pd black (93) and a high-purity Pd powder (94) were one or two orders less active than Pd/SiO, catalysts (95). The presence of an adequate carrier or promoter is essential for obtaining a good methanol yield. Although Pd( 111) single crystal decomposes methanol (96) and the principle of microscopic reversibility states that this would also be capable of synthesizing CH,OH, a more rapid formation of CH4 via oxygen-containing intermediates may drive the overall reaction course. In this respect it should be said that even open faces such as the (210) and the stepped (001) surface do not dissociatively adsorb CO at 25-125°C (97). This suggests that unsupported Pd is a rather poor methanation catalyst. Under 1 atm total pressure in a CO + H, mixture, the Pd black catalyst (210-nm crystallites) produces methane but, here again, the activity level is about two times lower than that of Pd/SiO, catalysts (4.6-nm Pd particle size), and about two orders of magnitude less active than Pd/ Al,O, catalysts (4.8-nm Pd particle size) (98). It therefore seems that the effect of dispersion here is not pronounced with respect to the support effect. Silica, as an inert support, does not influence the activity of Pd to the same extent as does more the acidic alumina. Finally, it should be mentioned that the lack of kinetic data for singlecrystal surfaces of Pd encouraged Hicks and Bell to calculate specific activities of Pd(l11) and Pd(100) planes from the results obtained on Pdl SiO, and Pd/La,O, catalysts (99). The distribution of these planes was inferred from the IR spectra of adsorbed CO, based on the relative intensities of the B, and B, (bridging) bands (100). In the case of the Pd/La,O, catalyst, the methanol turnover frequency depends on the crystallographic orientation of the metal surface: for Pd( 100) planes, turnover frequency is s - ’ . However, as Hicks 18 x lo-, s - ’ ; for Pd(ll1) planes, 6.5 x

CATALYSIS BY PALLADIUM

61

and Bell (99) admit, these activity levels reflect the situation in which metal-support effects exist. In the case of silica-supported Pd, the respective turnover frequencies for both planes are a factor of 7.5 lower than for La,O,-supported Pd.

111.

Catalysis by Electron-Deficient Palladium

In this section we shall try to review available data on catalytic properties of electron-deficient palladium on supported catalysts. The term “electron deficient” we shall consider to mean very small clusters of Pd on various supports, and/or palladium ions stabilized by virtue of their presence in an appropriate chemical environment. As was mentioned in Section I, we shall limit ourselves to three classes of heterogeneously catalyzed reactions: hydrogenation of unsaturates, reactions of alkanes with H,, and hydrogenation of CO. Although the literature data covers homogeneous catalysis over Pd(I1) or Pd(1) ( l o ] ) , and Pd(0) in a liquid phase, our intention is to show the unique properties of the electron-deficient Pd species as compared with the catalytic behavior of “fully reduced” heterogeneous Pd catalysts. There are at least three reasons for this. First, an exceptional catalytic performance of electrondeficient palladium may be of technological importance. Second, such results should provide relevant information for the interpretation of metal particle size effects or structure sensitivity in the three classes of catalytic reactions. Finally, in the course of developing our idea about using a catalytic reaction as a kind of probe for characterizing surfaces of supported metals (in this case, palladium), we believe that very pronounced differences in the catalytic behavior, if any, might be suggestive of the existence (and perhaps also the relative abundance) of electron-deficient Pd species in complex, supported Pd catalysts (Section IV,B). Of course, such a chemical probing would be possible only if we can prove, by means of other (mainly physical) techniques, that the existence of electron-deficient palladium in supported palladium is possible. Therefore, the organization of this section is as follows: First, we discuss the results of XPS studies of electronic properties of small Pd particles deposited on various supports. Then we examine other evidence for the existence of positively charged Pd species using other techniques, such as electron spin resonance (ESR) and infrared (IR) spectroscopy of adsorbed CO. Finally, catalytic consequences of the appearance of positively charged species in the Pdlsupport catalysts will be demonstrated.

62

ZB I GNI EW KARPINSKI

A. XPS STUDYOF SMALLPALLADIUM PARTICLES Baetzold used extended Huckel and “complete neglect of differential overlap” (CNDO) procedures for computing electronic properties of Pd clusters (102, 103). It appeared that Pd aggregates up to 10 atoms have electronic properties that are different than those of bulk palladium. d-Holes are present in small-size clusters such as Pd, (atomic configuration 4d0)because the diffuse s atomic orbitals overlap strongly and form a lowenergy symmetric orbital. In consequence, electrons occupy this molecular orbital, leaving a vacant d orbital. For a catalytic chemist the most important aspect of these theoretical studies is that the electron affinity calculated for a 10-atom Pd cluster is 8.1 eV. This value, compared to the experimental work function of bulk Pd (4.5eV), means that small Pd clusters would be better than bulk metal as electron acceptors. The electronic structure of small Pd clusters has been the subject of various experimental studies involving modern techniques of surface analysis such as X-ray and ultraviolet photoelectron spectroscopy and Auger electron spectroscopy. Several authors reported that the core electron binding energies (being very sensitive to the valence band structure) for small Pd clusters supported on conducting surfaces (carbon) and insulators (SO,, A1,0,, and zeolites) generally diminish with the increase of the cluster size, achieving a value characteristic of bulk Pd at -4-5 nm [Pd 3d5,,: 335.0 (104) t 335.2 eV (ZOS)]. Figure 5 shows a plot of several literature data. Two different interpretations of the binding energy shift were suggested. First, the shift may be a result of a size dependence of the initial-state electronic structure. Specifically, changes in the number of valence d electrons with size are thought to be responsible for the observed shifts. Second, the shift may be due to variation in final-state relaxation processes. Each of these interpretations were examined by Mason (106), who showed that the final-state effects are of only minor importance on supports such as carbon and silica. On the other hand, Kohiki and Ikeda (107) are of the opinion that the initial-state effect is responsible for the core electron binding energy shift for small Pd clusters on the conductive amorphous carbon substrate, whereas the photoemission final-state relaxation processes give rise to the binding energy shifts for the small clusters supported on insulators such as SiO, and AI,O,. Wertheim and co-workers attribute the latter effect to the electrostatic effect of the positive hole, unneutralized during the photoemission time (108),and believe that Pd clusters supported on amorphous carbon exhibit XPS shifts primarily due to the charge left on the cluster in the final state by the photoemission process (109). However, in conclusion of his other work (ffO),Kohiki states that

63

CATALYSIS BY PALLADIUM

337

+ A

4

> a

-x

U m T3 Q

x

," 336 W

C llJ

m

.-C

M+ 0

D C ._

m

-

bulk 335

0

2

4

Pd 6

8

Pd particle size, n m FIG.5. Dependence of binding energy of Pd 3d,,? on Pd particle size. 0, Pd/A120, (from Ref. 116); A , Pd/Si02 (from Ref. 116); A, Pd/C (from Ref. 112a); +, Pd/C (from Ref. 112b); and 0, Pd/Y (from Ref. 113; estimated particle size - I nm).

positive core electron binding energy shifts in very small Pd clusters supported on Al,O, and SiO, arise predominantly from the initial-state effect. Similar conclusions have been drawn by several other authors (111-1f6). Most of them argue that if the screening effect responsible for the relaxation processes were incomplete, the change of substrate (C, A1203, SiO,, or zeolite) should produce large differences in core level shifts. Ryndin et ul. present a universal curve correlating the Pd 3dS,, binding energy shift with the Pd particle size (116). Similarly, Hub et ul. (115) report a satisfactory agreement between chemical shift on Pd/Al,O, with the respective value reported by Takasu et al. (112), about 1 eV, for the Pd/C system. Figure 5 shows that the comparison of several experimental works seems less satisfactory because of somewhat different values

64

ZBIGNIEW KARPINSKI

of core level shifts for various Pd/support systems with the highest metal dispersions. It is obvious that in the particle size range <1 .5 nm, this value is known with rather limited accuracy. However, it should be stressed that for the 1.5 to 4.0-nm size range, it is in reasonable agreement as to the binding energy shift versus Pd particle size (Fig. 5). Summing up, although the basic assignment of the positive binding energy shift of Pd core levels to final- or initial-state effects is still debated (117, 118), it is clear that XPS, apart from electron microscopy, is a sensitive method to monitor average-sized metal particles. Thus it is particularly useful for investigating phenomena such as metal particle growth or redispersion brought about by thermal or chemical treatments ( I 19). Accordingly, Legare et al. ( I 19) observed positive Pd 3 4 , binding energy shifts when palladium deposited on y-alumina was heated in ultrahigh vacuum at 500°C or was exposed to oxygen at room temperature and then heated again in UHV at 500°C for 3 h. These shifts were not attributed to the formation of PdO, but to the redispersion of palladium material over the support. If this is so, one should ask what would be the mechanism of this phenomenon, i.e., what are the support sites that stabilize small Pd clusters and what is the chemical nature of these small Pd entities. We shall return to this problem in Section III,B. From this discussion one may conclude that, in spite of the present difficulties in interpreting XPS and UPS spectra, small Pd clusters, unsupported or supported (no matter what support is used), would contain fewer d electrons relative to the bulk metal. Next we shall furnish electron spin resonance and infrared evidence for the existence of electron-deficient Pd species in alumina- and zeolite-supported Pd catalysts.

B. ESR AND IR EVIDENCE FOR THE EXISTENCE OF ELECTRONDEFICIENT Pd SPECIES I N SUPPORTED CATALYSTS A vast majority of the literature data for unsupported and supported Pd catalysts indicate total reduction of palladium to the metal at fairly low temperatures. Only in a few cases the presence of higher temperature, temperature-programmed reduction (TPR), peaks was observed (120). However, several recent XPS, E S R , and IR studies report the presence of ionic species in highly dispersed Pd/support catalysts that have been subjected to H2 treatment at relatively high temperatures, say, at 2250°C. An important question is whether these ionic species play a significant role in heterogeneous catalysis. In a majority of basic and applied works in the field of heterogeneous catalysis, a support is a necessary catalyst component to obtain and keep extremely high metal dispersion. Differ-

CATALYSIS BY PALLADIUM

65

ences in the relative resistance to metal sintering caused by the use of various supports must be at least in part ascribed to differences in the interaction between metal particle and the carrier surface. In extreme cases, such an interaction may lead to the formation of ionic metal species. Huizinga and Prins detected (by ESR) the presence of positively charged Pt in Pt/A1,0, and indicated that the charged species were located close to the metal-support interface (121). This would suggest a positive correlation between concentration of metal ions and the dimension of the metal-support interface. When one achieves very small metal clusters (the percentage exposed, almost loo%), the relative proportion of metal ions should be much higher than in the case of a poorly dispersed metal on a support. Now the question arises as to whether one deals exclusively with metal ions after getting almost atomic dispersion. The information we previously presented indicates that relative electron deficiency may be an inherent property of very small Pd particles. Therefore, the catalytic behavior of metal ions would be more clearly marked for highly dispersed metal/support catalysts. Here we shall try to examine ESR and IR evidence for the existence of electron-deficient Pd species in supported Pd catalysts subjected to more or less careful reduction in hydrogen. The electron spin resonance as a method enabling identification of Pd+ and Pd3+ species was used in a number of cases to detect these species in supported Pd catalysts. Especially in the case of palladium-loaded zeolite X and Y the presence of ionic Pd species in mildly reduced samples has been confirmed (122-126). However, after applying higher reduction temperatures (>25OoC),paramagnetic signals disappeared. It must be noted that more drastic reduction of Pd/zeolite usually leads to serious metal sintering (Pd particles 2 nm in size). Unfortunately, in the case of alumina-supported Pd, the ESR investigations are complicated by the fact that alumina may give quite a strong, broad ESR signal, which coincides with the signal characteristic of Pd+ (127, 128). The intensity of this broad signal is due to various paramagnetic impurities (mainly iron) and depends on a thermal treatment (with evacuation) of alumina (127). Nevertheless, the presence of Pd species was postulated in a number of cases involving Pd/Al,O, catalysts (129-134) (Fig. 6). From the majority of works concerning alumina- and zeolite-supported palladium, an anisotropic signal with 8 1 = 2.10-2.15 and gll= 2.3-3.1 should be due to Pd+ (122-126, 129, 131, 133,134). Pd3+ produces an isotropic signal with giso= 2.23 (122-126). A more recent ESR study by Parvulescu et al. (132) identifies Pd+ species in 0.1, 0.3, and 0.5 wt% Pd/AI,O, after reduction at 400°C for 4 h. The ratio of relative intensities of Pd+ species was not proportional to the Pd loading (i.e., 1 : 3 : 5) but was approximately 1 : 1.7 : 2, indicating that +

I &

0

m

I

n

m

0

-

0

P

CATALYSIS B Y PALLADIUM

W

H

+ C,H,Li

+ PdCI,

U 4 L i

--*

-

1

+ C,H,,,

(itO-),PdC12

li.

1;

C-OLi

(M-),Pd

67

~

~

+ xLiCl

+ ( 2 - x)HCI

(3)

SCHEME 3.

only some part of Pd exists as paramagnetic Pd+. According to the previously mentioned work of Huizingaand Prins (121), Pd+ should be located in close proximity to the metal-support interface. Recently, Margitfalvi et al. (135) reported results on preparation and characterization of aluminasupported Pd catalysts prepared via an anchoring technique according to Scheme 3. The authors claim that catalysts prepared in this way, after hydrogen treatment at higher temperatures (not specified), contain a considerable part of Pd in its higher oxidation state (ESR and XPS characterization). Zakumbaeva et al. (130) showed that the oxidation state of palladium depends very much on the reduction conditions. In the case of a 4-wt% Pd/AI,O, catalyst, complete reduction of Pd2+ to Pd' was achieved at -300°C. The process of the Pd2+disappearance is not associated with the appearance of Pd+ ions observed above 300°C. The authors believe that the Pd+ ions result from the interaction between metal and support, as above 300°C alumina loses its hydroxyl groups. An interesting finding is that after reduction at 500°C and cooling in H,, 18% of palladium was as Pd+ ions (by ESR). A still higher concentration of Pd+ (27%) was in torr) at this the catalyst reduced at 500°C and evacuated (down to temperature. Fewer Pd+ ions were formed when the catalyst was reduced at 500°C and was cooled down in H, to room temperature and then subjected to contact with air (22%). The XPS results of Legare et al. (119) should be recalled here (cf. Section 111,A); they showed an increase in binding energy of the Pd 3 4 , core level after heating Pdly-Al,O, layers in UHV at 500°C. Although the authors attributed that fact merely to redispersion of palladium, we think that such an interpretation is not in conflict with the results and conclusions of Zakumbaeva et al. (130). Legare el al. (119) believe that the positive shift in binding energy is too small to be attributed to the formation of PdO. Also due to the mentioned problems in the interpretation of binding energy shifts of Pd supported on insulators (initial-state versus final-state effects), the authors do not wish to discuss the relative electron deficiency of small Pd clusters. However, it is possible that these small Pd clusters, which are redispersed over alumina as a result of the pretreatment in vacuum at 500"C, carry a positive charge. Pd', confirmed by ESR, is a feasible species.

-

68

ZBIGNIEW KARPINSKI

To the best of our knowledge no paramagnetic palladium species have been reported for reduced Pd/SiO, catalysts. We shall show a little later that in the case when silica is employed as a carrier, the presence of other elements such as Ca and Mg helps in stabilizing palladium ions after reduction in H,. It appears that alumina, unlike silica, interacts strongly with Pd. Bychkova et al. investigated the acceptor properties ofthe surface of the Pd/Al,O, catalyst by an ESR spin-probe method using 2,2,6,6tetramethyl-1-piperidinyloxyradicals, which form complexes with coordinatively unsaturated (cus) AI3+ ions. It appears that introducing Pd to A1,0, causes both a significant decrease in the concentration of As:l sites (136) and a shift in distribution of acceptor sites toward weaker centers (137). The conclusion that palladium particles in zeolites may carry a partial positive charge follows from the IR study of CO adsorption. This adsorbate can be considered to be a probe of the electronic state of palladium. Namely, the shift toward higher frequencies of the CO linear band (for Pdo-CO it appears at 52100 cm-') reflects a decrease in the back donation of electrons from Pd to CO. Along with such an interpretation, Figueras et al. (138) detected the presence of electron-deficient Pd species in Pd/ HY but not in Pd/Si02. More recently, Lokhov and Davydov (139) confirmed the presence of positively charged Pd species apart from PdO in reduced (at 300°C) Pd/Y samples and ascribed a 2120- to 2140-cm-' band to Pd+-CO complexes (Fig. 7). Similarly, Romannikov e t al. (140) report that adsorption of CO on Pd/Y samples reduced at 300°C produces IR bands at 32100 cm-' ascribed to Pd+-CO and Pd"+-CO complexes. The presence of similar bands was confirmed recently by Sheu et al. (141). In addition, it was shown that CO release from a Pd,(CO), cluster is likely due to interaction of the Pd with zeolite protons, because the IR band of the zeolite 0-H group decreases when CO is released and increases when CO is added to the cluster (141, 142). We have already shown that carefully reduced low-content metal-loaded Pd/Al,O, catalysts exhibited ESR signals ascribed to Pd+ ions. Several IR studies of the adsorbed CO on such samples confirm the presence of electron-deficient Pd (114, 133, 143-145). It should be stressed here that several other IR studies of adsorbed CO on Pd/AI,O, catalysts do not report bands at 22100 cm-', suggesting the absence of unreduced palladium (146-149). However, it seems relevant to remark here that a majority of these works concern catalysts with higher percentage Pd loadings [2 wt% (146, 147) or even 9 wt% (148, 149)], thus are not necessarily characterized by high values of metal dispersion. On the other hand, even in the IR studies of low-content metal-loaded Pd/Al,O, catalysts (114, 133, 143-145), the intensities of Pd"+-CO or Pd+-CO bands were always

CATALYSIS BY PALLADIUM

1900

2000 2100 Wavenumber,

69

2200 cm-l

FIG.7. IR spectra of absorbed CO on reduced (at 300°C) Pd/Y catalyst: curve 1, background; curve 2, after CO adsorption (15 torr); curve 3 , after additional evacuation. (From Ref. 139.)

found to be low compared to the PdO-CO band: not surprisingly, therefore, electron-deficient Pd has not been seen in the IR study of less dispersed, high-content metal-loaded Pd/Al,O, samples. The presence of a very small amount of electropositive Pd species (compared to totally reduced PdO) should not be neglected. There is no doubt that the IR of the adsorbed CO probes the surface sites, which might be active centers for catalytic reactions. As long as the relative turnover frequencies for individual kinds of active sites are not known, their presence should not be neglected, even if they are not plentiful. On the other hand, the ESR experiments monitor the concentration of Pd+ ions both on the surface and in the bulk of metal particles. Therefore, ESR results are more relevant for catalysis in the case of very small metal particles (dispersion of -100%). Heating palladium in UHV at 500°C leads to metal redispersion over yalumina [XPS (ff9);see also Section III,A]. Purging the Pd/Al,O, catalyst with ultrapure helium or argon at 500-600°C produces Pd+ species seen by ESR (134). IR spectra of the adsorbed CO on the catalyst pretreated in Ar at 600°C for 17 h show the presence of Pd+-CO and Pd"+-CO bands

70

ZBIGNIEW KARPINSKI rn

N N

N -t

-.

0

m

m

8 a:

d

0

C

2L : Z 0

In

zs

0

6

8 W hl

9

0 I

2310

2190

2070

1950

Wavenumber, cm-’

2310

2190

2070

1950

Wavenumber, cm-’

FIG.8. FTIR spectra of CO adsorbed on 0.97 wt% Pd/AI2O, after reduction at 600°C for 17 h followed by Ar purging at 600°C: (a) I-h purging, (b) 17-h purging. (From Ref. 143.)

(Fig. 8 ) (143). The authors discuss the mechanism of the formation of electropositive Pd species via an oxidation with surface hydroxyl groups according to the redox reaction [Eq. (4)]: n(OH, )

+ M”+

(O?-),,M”’+ (n/2)H1(3)

(4)

This reaction certainly occurs for less noble metals such as Mo ( / S O ) , Fe ( / 5 1 ) , or Ni (152) supported on alumina. In the case of noble or near-noble metals, such an oxidation seems less likely, but its feasibility could be increased by the involvement of very small metal aggregates. It should be mentioned that considerable gain in entropy due to the formation of gaseous H, makes this reaction more feasible at very high temperatures. Tzou et al. (153) found that Pt atoms located in sodalite cages in Y zeolite are oxidized by the surface hydroxyls. An important question is how these ionic Pd species are stabilized on the surface of y-alumina but not on SiO,. It is known that y-alumina, unlike silica, has unoccupied (by Al’+) octahedral sites. It has been suggested

CATALYSIS BY PALLADIUM

71

FIG.9. Model of superactive Pd"+ site stabilized on y-alumina surface (143).

(143) that Pd"+ species are stabilized in these vacant sites (Fig. 9). The recent EXAFS study by Lesage-Rosenberg et al. (154) shows that the fixation of the active cation (from the Pd acetonate precursor) on alumina involves octahedral aluminum vacant sites of the carrier, forming locally aphase close to an aluminate. On the other hand, silica has only tetrahedral sites, which are fully occupied, so the stabilization of Pd"+ species is more difficult than in the case of y-Al,O,. Another model seems to explain the presence of electropositive Pd species in zeolite-supported Pd catalysts. A crucial difference between Pd/SiO, and Pd/NaY is that reduction of Pd2+ions (introduced by cation exchange) yields H + ions only in the case of Pd/NaY, whereas water is formed when PdO/SiO, is reduced. It appears plausible that positively charged particles are formed when Pd particles interact with some of these protons in NaY (155). As two protons are formed per Pd atom and as the reduced atoms will migrate and form clusters, it is probable that the cluster will be located in a supercage, which held originally at least one Pd2+ion. After reduction, two H f are present in this cage, compensating the two negative charges of the zeolite matrix. These protons can either be fixed on 0,- ions of the cage wall or interact with the Pd, particle inside the cage. In addition, other protons could migrate into the cage in exchange from Na + migrating out. The resulting particle, consisting of n Pd atoms and z hydrogen atoms, will have a positive charge + z . This charge, of course, will not be localized on the H atoms but will be smeared out over the Pd,-HZ particle. This model fits with Gallezot's finding that the degree of electron deficiency increases with the acidity of the zeolite (156). Recently it has been suggested that this positively charged (Pd,-HZ)Z+species is responsible for very high activity for neopentane conversion over Pdi NaY (157) (Section III,C,2.). Neither ESR nor IR spectra of adsorbed CO indicate that electron-

72

ZBIGNIEW KARPINSKI

deficient Pd species are present in silica-supported Pd catalysts after careful reduction. However, the appearance of ionic palladium may be brought about either by an addition of other elements such as Ca or by the influence of reactive environment (oxidative in nature). Ponec and co-workers (158-160) emphasized the essential role of Pd ions in methanol synthesis. They furnished several forms of evidence (ESR and IR of CO and extraction with acetylacetonate) of the presence of palladium ions in Mg-doped Pd/SiO, catalysts. Similarly, the presence of calcium in Pd/CaX catalysts (treated in H2)accounts for the presence of Pd+ ions, which are not seen in the reduced Pd/NaX samples [Kevan et al. (124-12611. C. CATALYTIC BEHAVIOR OF ELECTRON-DEFICIENT PALLADIUM 1.

Hydrogenation of Unsaturated Hydrocarbons

Usually the catalytic hydrogenation of unsaturated hydrocarbons over Pd catalysts is performed at low temperatures, most often somewhere between room temperature and 100°C. The partial pressure of hydrogen is rather high, even if the hydrogenation is realized under atmospheric pressure. Therefore, as is mentioned in Section II,A, regarding unsupported Pd, the role of the phase transformation leading to formation of the p-palladium hydride phase should be taken into account, as the catalytic behavior of P-PdH may differ (and it actually does) from that of pure Pd. This phenomenon is expected to complicate the relation between catalytic activity/selectivity and Pd particle size, because the structure-sensitivity relation [understood in the usual geometric or electronic terms (161)] is complicated by the fact that the chemical composition of the catalyst changes with the Pd dispersion (Pd or p-PdH) (162-164). Our present analysis will take into account only examples of catalytic activity that has been directly correlated with the appearance of electron-deficient Pd species. Hydrogenation of benzene is generally regarded as a structure-insensitive reaction over supported Pd catalysts (161). In the presence of Pd catalysts, the reaction is carried out usually at about 100°C (and above), hence the problem of p-PdH formation may be disregarded here (if Phydrogen 5 1 atm). Figueras et d.(138)found that the activity of Pd/zeolite catalysts is considerably higher than in the cases of Pd supported on silica, magnesia, or alumina. This result was confirmed later by others (140). This support effect appears only on solids that exhibit strong acceptor sites. The observed sequence of activities NaX < NaY < CaY < MgY < CeY HY - LaY corresponds to the known sequence of acidic or oxidizing properties of these zeolites. Accordingly, after Pd deposition on

-

CATALYSIS BY PALLADIUM

73

silica-alumina, the number of electron acceptor sites decreases, suggesting that Pd should acquire a (partial?) positive charge. The shift of the IR band of linearly bonded CO toward higher frequencies confirmed the presence of electron-deficient palladium species in Pd/Y (138). The metal dispersion measured by H,-0, titration was poor (12-16%), but electron microscopy showed the presence of very small metal particles ( < I nm) along with big Pd crystallites (50 nm). We believe that the pronounced activity of zeolite-supported palladium has to be attributed to these very small Pd particles. Alumina-supported Pd catalysts exhibited a similar low level of activity as silica- and magnesia-supported samples. This fact does not confirm our earlier considerations (cf. Section II1,B) as to the possible stabilization of electron-deficient palladium species on alumina. However, the authors (138)did not specify the level of metal dispersion in their Pd/ A1,0, catalysts, thus any further conclusions can hardly be ventured. In a more recent paper from the same laboratory (165), the authors report that alumina-supported Pd catalysts exhibit constant (insensitive to metal dispersion) activity, which is about three times higher than the activity level of silica-supported Pd. Although the authors argue that this activity difference may be due to the presence of an iron impurity in commercial Davison 70 silicagel, Vannice and Chou (166) do not regard poisoning by impurities from the support as a good explanation, because TOF values should not change for a structure-insensitive reaction as the metal surface is covered. More pronounced differences between catalytic behavior of very small versus large Pd particles are observed in hydrogenation of alkenes and alkynes. Here a considerable caution should be exercised in order not to compare the “pure” Pd samples with those in which P-PdH exists (we simply treat them as different catalytic materials, for example, palladium and silver). This is very difficult because the hydrogenation reaction proceeds at ambient temperatures. Nevertheless, it has been shown that very small Pd particles, electropositive in nature (as indicated by the accompanying XPS measurements), exhibited much higher activity than did large Pd crystallites in the hydrogenation of but-I-ene (115). However, in the case of hydrogenation of alkynes [but-I-yne (115) or vinylacetylene (116)], small Pd particles are less active than the larger ones because of too strong alkyne adsorption by small electron-deficient Pd clusters. These catalytic results were interpreted in terms of the adsorption strength of the reactant molecules. For a given metal (thus also for Pd), the adsorption strength of the reactant varies in the following order: aromatic < olefin < diolefin < alkyne. Too strong or too weak adsorption means a decrease in activity, suggesting a kind of a “volcano” curve for correlating activity with the adsorption strength. Figure 10 shows such a relation, indicating

74

ZBI G N I E w KARPI NSKI

>

FIG. 10. Schematic graph of activity in hydrogenation reaction as a function of the adsorption coefficient on large Pd particles (thick arrows) and on small Pd particles (thin arrows). (From Ref. 115.)

the reactants that should be more readily hydrogenated on small electrondeficient Pd clusters. Clearly, only olefins and, to alesser extent, aromatics belong to this class of reactants. Whereas for alkenes the situation seems to be proved, in the case of hydrogenation of aromatics the existing results are not convincing. Another indication that electronic properties of Pd may be important in hydrogenation reactions originates from the work of Carturan el af. (167), who investigated palladium supported on vitreous materials in hydrogenation of phenylacetylene. A relatively better catalytic activity of the catalyst with smaller alkaline content (Na,O) suggests that an electron transfer from Pd to the support is smaller in the case of less “acidic” (containing more alkaline) supports. Similar metal particle sizes (2.8-3.4 nm) exhibited by all the catalysts rule out an explanation that takes into account a surface sensitivity of this reaction. 2. Reactions of Alkanes with Hydrogen Although by 1972 Dalla Betta and Boudart (168) and later Foger and Anderson (169) showed that electron-deficient platinum (in zeolites Y : L a y , Cay, and Nay) is much more active than Pt/SiO, (169) and Pt/Al,O,

75

CATALYSIS B Y PALLADIUM

(168) in the reaction of neopentane with hydrogen, no similar evidence has been furnished for palladium catalysts until recently. Both the palladium catalysts supported on y-alumina (143) and o n Y zeolite (157, 170) appeared more active than silica-supported ones, and this difference could not be interpreted in terms of different metal dispersion or additional activities for y-alumina or Y zeolite. For Pd/A1203, very high cracking activity was achieved after prolonged purging of the catalyst in helium at 600°C (143). In Section III,B we mentioned that the IR spectra of the adsorbed CO on such catalysts revealed the presence of electron-deficient Pd"+ species. Accordingly, it has been suggested (143) that Pd"+ ions are sites of high cracking activity. The mechanism would be as follows [Eqs. ( 5 ) and (@I. CH3 Pd"?

+ C(CH:), + Hz+

CH,

+ Pd'"-"+-H + CH:-Ct

I

I

(5)

CH?

where n = 1 or 2. This step is favored by the stability of the tertiary carbenium ion and the strong Pd'" I ) + - H bond. The next step is ~

CH3 Pd'"-"+-H

I

+ CH3--C'

--j

I

Pd"+

+ (CH&CH

(6)

CH:

This step is facilitated by C-H bond formation; it regenerates the Pdn+ site. The reaction mechanism is supported by recent results of alkane activation by Pd+ ions in the gas phase. According to Tolbert er al. (I71),Pd' ions exhibit uniquely high Lewis acidity in the activation of neopentane. The reaction proceeds via an insertion of Pd+ between the methyl and tert-butyl fragments of the neopentane molecule, just as shown above. Catalytic superactivity of electron-deficient Pd for neopentane conversion was recently verified for Pd/NaHY (157, 170). The reaction rate was positively correlated with the proton content of the catalyst. Samples that contained all the protons generated during H2 reduction of the catalysts were two orders of magnitude more active than silica-supported Pd. Samples prepared by reduction of Pd(NH3):+NaY displayed on intermediate activity. It was suggested that Pd-proton adducts are highly active sites in neopentane conversion. With methylcyclopentane as a catalytic probe, all Pd/NaY samples deactivated rapidly and coke was deposited. Two types of coke were found (by temperature-programmed oxidation), one of

76

ZBIGNIEW KARPIfiSKI

which correlated with the proton concentration in the catalyst (170). The presence of (Pdn-HJZ+ adducts is substantiated by the results of an IR study of CO adsorption on Pd/NaY (141, 142) (Section 111,B). 3 . Hydrogenation of CO

Two reactions of CO with H, are to be considered: formation of methane (called methanation) and synthesis of methanol. Under sufficiently high pressure, both of them proceed over supported Pd catalysts, with the activity and selectivity dependent on choice of the support and the presence of various promoters (92, 93, 95, 158-160, 172, 173). In this section we shall refer only to those papers in which an essential role of electrondeficient Pd species was demonstrated. There is a consensus that in the synthesis of methanol from syngas, the presence of Li+ (95, 159, 173), Mg2+,or La3+ (95, 158-160, 172, 173) ions promotes the activity of Pd supported on SiO,. Different interpretations of this activity enhancement have been suggested and until now the basic reason for it was uncertain. The effect of metal dispersion and promoter ions was carefully investigated by Kelly et al. (95). The authors tend to believe that the morphology changes due to supports and promoters play a very important role in achieving good activity and selectivity toward methanol formation. However, one very attractive interpretation (which probably needs more evidence) involves the crucial role of Pd"+ species (158-160). Several very important items of evidence have been furnished (160). First, Driessen et al. (158) found very good correlation between activity and the concentration of Pd extractable by acetylacetonate (as Pd"+ ions) from Pd-MgO/ SiO, and Pd-La,O,/SiO, catalysts (Fig. 11). Second, the presence of palladium ions in the active catalysts were confirmed by ESR and IR spectra of adsorbed NO and CO (159). Two further proofs are indirect. Namely, using the same series of Pd-MgO/SiO, catalysts, Hindermann et al. (174) found a good correlation between concentration of detected (by chemical trapping) surface formyl intermediates with the methanol activity (dependent on the percentage MgO). That finding in conjunction with the evidence of Driessen et al. (158) produces a good correlation between the concentration of formyl intermediates with the relative amounts of Pd" ions (160). Finally, quantum mechanical calculations revealed that the formyl formation is difficult on a neutral or negatively charged Pd atom, whereas it is easy on a positively charged palladium (175). Vannice and co-workers (176) found that methanation over Pd catalysts is essentially a surface-insensitive reaction; however, the activity depends upon the support used. Alumina-supported and silica/alumina-supported catalysts were 10 times more active than the silica-supported ones, resembling very much the corresponding relation reported for benzene +

-

77

CATALYSIS BY PALLADIUM

0.4 -

CH,OH act.

tV O1 0.3-

0.2-

0.1 -

0

I

0

1

1

1 O/O

Pd n+

3

FIG. 11. Activity in CHIOH formation at 215°C as a function of the relative amounts of Pd"+,extractable from Mg- and La-promoted silica-supported Pd catalysts. 0, Mg promoter; 0, La promoter (percentage of CO converted into all products); 0 , La promoter (percentage of CO converted into methanol). (From Ref. 158.)

hydrogenation. This result may be related to the mechanism of CO hydrogenation over Pd catalysts suggested by Driessen et al. (158).If most of the CH, formed on the Pd catalyst originates from 0-containing intermediates bound to Pd"+ centers, then it seems understandable from the foregoing discussion that the use of alumina and silica/alumina as supports for palladium provides more active methanation catalysts than does Pd/SiO,. IV. Catalysis over Supported Palladium

Three issues are to be considered in this section. First, we shall discuss briefly the rather general phenomenon of an apparent structure insensitivity or a rather mild structure sensitivity of palladium compared with other

78

ZBIGNIEW KARPINSKI

catalytic metals (such as Pt, Ir, and Rh). We shall show several examples of structure-insensitive reactions catalyzed by palladium catalysts, whereas the same reactions exhibit structure sensitivity when catalyzed by other metals. We shall try to suggest an explanation for such a behavior. In several cases an apparent structure sensitivity is caused by secondary phenomena, such as changes in the catalyst surface generated by a reaction environment. As was mentioned previously, many of the alkene or alkyne hydrogenation reactions were performed under conditions of higher H, partial pressure (in the vicinity of I atm) and low temperature. Thus, the transformation Pd + P-PdH phase should take place under such conditions. From the majority of published work it follows that this transformation is possible unless the Pd particles are very small. Therefore, for highly dispersed Pd catalysts, at low temperatures, the deviation from structureinsensitive behavior can be due to a changed catalyst composition. When the temperature is sufficiently high to prevent the formation of the pPdH phase, unique properties of small Pd clusters versus Pd catalysts, characterized by moderate and low levels of metal dispersion, may result from variations in electronic properties of Pd particles. Second, we shall present our approach as to how to probe the palladium surfaces in more complex Pd/support catalysts, especially when the socalled metal-support interactions are expected. We shall develop our idea of how to use such chemical probes as a catalytic reaction (alkane catalytic conversion) or chemisorption in order to see important changes in the catalytic behavior. When possible, an adequate reference to available data from more sophisticated physical techniques is made. The third aspect considered will be concerned with an attempt to present our view on the problem of the evolution of supported Pd catalysts in the course of catalytic reactions. Here again, we shall limit our considerations to reactions of unsaturated hydrocarbons, alkanes, and CO with Hz.

A.

STRUCTURE SENSITIVITY OF PALLADIUM-CATALYZED REACTIONS

Since the late 1960s there has been some interest in the concept of a structure-sensitive reaction in heterogeneous catalysis (177, 178). In the case of supported metal catalysts, structure sensitivity is visualized as a dependence of metal particle size and catalytic behavior in a given reaction (activity and selectivity). Almost all of the possible kinds of relationships were reported in the past. Recently, Che and Bennett reviewed this problem ( / 6 / ) .Our intention here is not to repeat most of their analysis, rather we shall try to present our view on the general characteristics of palladium versus other platinum metals.

CATALYSIS BY PALLADIUM

79

In analyzing the problem of the structure sensitivity of a reaction catalyzed by Pd catalysts, it is convenient to classify the reactions in terms of the temperature range in which the reactions proceed and/or in which Pd catalysts were pretreated in H,. The effects of reaction (pretreatment) temperature and the presence of hydrogen in a reaction system should be treated as follows: 1, Possibility of a transformation of palladium into the P-PdH phase in hydrogen at lower temperutures. Such a transformation may take place in the reaction system during low-temperature hydrogenation (179). Small and large Pd particles may be transformed into P-PdH in different yields, producing a catalytic material entirely different from the initial one. Therefore, a structure sensitivity of the formation of P-PdH may be involved in an apparently observed “structure sensitivity” of reactions carried out at lower temperatures in the presence of H2. Palczewska (180) analyzed the problem of an apparent structure sensitivity in some hydrogenation reactions catalyzed by Pd catalysts. The attention was focused on the problem of P-PdH formation and its possible impact on the catalytic activity of Pd catalysts. The point is that the phase transformation proceeds easily in the case of medium and large Pd crystallites, whereas it does not for highly dispersed Pd particles [say, below 2 nm (36)]. If this is so, to analyze catalytic data for differently dispersed Pd catalysts, one compares samples in which the variation in chemical composition is large (from pure Pd to the P-PdH phase). Recent results suggest that PdH may be formed also in very small Pd particles [-1 nm (37, 38)]. While waiting for confirmation of these results, which are in distinct disagreement with earlier data (36), one can speculate that a variable (with Pd dispersion) proportion of the 0-PdH formed in the catalyst may be the reason for the observed structure sensitivity. In several papers a rather mild structure sensitivity, within the range of Pd dispersion, has been reported when the existence of P-PdH phase is possible (181-185). In some cases, the larger palladium hydride crystallites were less active than the smaller ones. If hydrogenation is effected via the desorbing “hydride” hydrogen, its flux would be smaller at lower temperatures, when P-PdH is more stable (180). This stability may also be a function of Pd particle size. Therefore, it is impossible to propose a universal curve of structure sensitivity for hydrogenation reactions, when the influence of P-PdH phase is marked. The stability of the p-PdH phase under a given H, partial pressure is a function of metal particle size; moreover, different unsaturates (alkenes, dienes, and alkynes) may “extract” hydrogen more easily from more dispersed specimens. A rather modest structure sensitivity observed in the hydrogenation of

80

ZBIGNIEW K A R P I ~ ~ S K I

an acetylene-ethylene mixture at 25°C over Pd/AI,O, catalysts was partly attributed to the structure sensitivity in the P-PdH formation (186).However, in addition, the strength of acetylene chemisorption should increase with decreasing Pd dispersion. The increasing coverage by acetylenic species [most probably, ethylidene species (31)Jshould inhibit adsorption of ethylene, thus ensuring good selectivity for half-hydrogenation. 2. Palladium is a metal characterized by relatively low lattice energy. This metal is the least dense and has the lowest melting point among the platinum group elements. Therefore, its surface may be subjected to greater modifications than is possible for the surface of Ir, Rh, or Pt (lattice energy for Pd, Pt, Rh, and Ir is 90, 122, 127, and 152 kcal/mol, respectively). These transformations may be realized in the following three ways: 0 At temperatures when the P-PdH phase may be formed and decomposed, several such cycles may produce considerable structural changes in the Pd particles: an increase of ( I 11) texture is observed (5, 21, 187) and the density of defects is increased (5, 21, 81). 0 At higher temperatures, in the presence of H,, a considerable sintering of Pd is observed, leading to the formation of bigger, more stable Pd crystals (84, 188). 0 At sufficiently high temperatures, due to not too strong cohesion, the surface Pd atoms may acquire convenient positions to form a bond with reacting hydrocarbon molecule (189). This concept, called “extractive chemisorption,” was introduced by Burwell et al. (190, 191) as a possible cause of absence of steric hindrance in adsorption and reaction of some complex organic molecules. It was proposed that in chemisorption one or two metal atoms were displaced above the initial planar level, leading to increased bonding to the surface for low-dispersion catalysts. An extension of this concept to the problem of structure sensitivity allows one to explain several cases of the relatively mild (or absent) structure sensitivity in many reactions catalyzed by Pd catalysts. The hydrogenolysis of cyclopentane is structure sensitive on Rh/AI,O, (at 225°C) but is apparently insensitive on Pd/AI,O, and Pd/SiO, (at 290°C) (192). Similarly, hydrogenolysis of methylcyclopentane is structure sensitive on Rh/AI,O, and Rh/SiO,, but is apparently structure insensitive over Pd/AI,O, (192). Dehydrogenation of methylcyclohexane in the presence of H, at 390°C over SO,-, AI,O,-, and P0,Al-supported Pd catalysts (193) showed moderate changes in the Pd particle size (threefold); the identity of the support causes a similar kind of change. Karpinski et ul. (189) found that the reaction of neopentane with H, over a series of Pd/SiO, catalysts is structure insensitive and exhibits very mild changes in activity and selectivity. The same reaction performed by

CATALYSIS BY PALLADIUM

81

others (85) on supported Pt catalysts shows structure sensitivity: both turnover frequency and selectivity change with metal dispersion. The hydrogenolysis of methylcyclopropane is quite structure sensitive on Rh/SiO, (194),whereas over Pd/SiO, the variations in turnover frequencies with Pd particle size are not very pronounced (182). Hydrogenolysis of ethane appeared structure sensitive over Rh/SiO, catalysts (195), with a maximum for an intermediate level of dispersion. For Pd catalysts similar data do not seem to exist; nevertheless, a very recent study by Gao and Schmidt (80) has revealed that the activity of the Pd/SiO, catalyst exhibits only small changes after oxidation and lowtemperature reduction cycling. Although such an activation produces highly reactive, low-coordination sites, the Pd surface becomes annealed at temperatures well below the temperature employed in the kinetic study (-200°C). On the other hand, the more refractory Ru, Rh, and Ir can retain their high activity during hydrogenolysis at 300°C. Another example of a structure-insensitive reaction catalyzed by Pd/ SiO, catalysts is the neohexane/D, exchange (191). The reaction performed at 105°C in D, (684 torr) should not be assisted by the P-PdD formation. For the prevailing range of metal dispersion, turnover frequency was fairly constant, -3 x 10- s-I. Only for very highly dispersed catalysts (H/Pd, = 0.791 and 0.967) was the activity about three times higher. In addition, the effect of metal dispersion on product distribution (especially, the relative yield of ethyl-d, in neohexane) did not confirm expectations that larger Pd particles should contain a larger proportion of more densely packed planes, such as ( I 1 1) and (IOO), and thus create more steric hindrance from the bulky tert-butyl group. The authors advanced a hypothesis that the bounding faces of the Pd crystallites might contain many defects where the reaction would proceed without special steric hindrance. Another suggested possibility (191)is that an extractive chemisorption would reduce adsorbate-surface hindrance. Benzene hydrogenation carried out at 140°C on a series of Pd/SiO, and Pd/AI,O, catalysts is structure insensitive (165) [in agreement with others (196, 197)],whereas very small Rh particles exhibited much lower activity than moderately and poorly dispersed Rh/AI,O, catalysts (192). 3. For very small P d particles (<2 nm in size), especially when supported on silica-alumina or alumina, the electron-deficient P d species present, stabilized by the support, may explain drusticully dgferent catalytic properties of highly dispersed Pd catalysts. Specific activity may be higher or lower than that of medium-dispersion (and low-dispersion) catalysts, depending on the adsorption strength of reacting molecule with the catalyst. If the adsorption strength of Pdo-adsorbate is weak, then the presence of Pd"+ in the catalyst may be advantageous. However, too

82

Z B I G N I E W KARPINSKI

strong a chemical bond between Pd"+ and adsorbate may lower the reaction rate, e.g., due to hindered product desorption or self-poisoning (Section lll,C,l; Fig. 9).

B.

CHEMICAL PROBING OF Pd SURFACES

Here we would like to present our approach to probe palladium surfaces in supported Pd catalysts. A special interest was to investigate Pd/support catalysts in the state of metal-support interactions (10). The idea was to use such a reaction, which is sensitive to small changes in the morphology of Pd surfaces. Hydrogenolysis of alkanes, proposed by others, seemed a good choice (198, 199), because the rate of this class of reactions exhibits large variation with the surface morphology. As was mentioned in Section Il,A, palladium generally exhibits low activity in alkane hydrogenolyses. Therefore, if variations in an absolute rate of this reaction are not very pronounced, it would appear more suitable to use selectivity as a convenient diagnostic parameter. Such a possibility provides the use of the reaction of neopentane conversion with H7_.I n Section U,B, we presented results on the catalytic behavior of Pd films and powders. Big differences in the catalytic behavior of various Pd surfaces (literature data) suggest that the reaction of neopentane may serve as a good probe reaction. The Pd(1 I ])-oriented films showed very high selectivity toward isomerization, whereas Pd powders and rough Pd films showed more pronounced hydrogenolysis. lt should be emphasized that neopentane, unlike other alkanes, is unable to form olefins or carbenium ions without changing its carbon skeleton. Therefore, the catalyst deactivation is negligible with this probe molecule [e.g.,for Pd/NaY (157)]. This fact permits the attribution of the observed variations in the catalytic behavior to changes in the catalyst morphology. Our intention was to employ the reaction of neopentane for probing the Pd surface, especially when the so-called strong metal-support interactions were expected. Such interactions may be manifested both in topographical (sintering and pillbox structure) as well as in chemical changes (decoration by support species and formation of an intermetallic compound) of the surface of Pd particles, and it would be difficult to decide u priori to what extent each of these factors may contribute to the observed changes in the catalytic behavior. Our goal was to try to separate these two factors. The effect of the surface topography of unsupported Pd specimens was described in Section I1,B. It should be recalled here that Pd ( 1 I I ) films exhibited high isomerization selectivity (-70-80%), whereas powders and

83

CATALYSIS BY PALLADIUM

.

5

0

10

0 Pd-Au 0 Pd-Cu

20

30 atom

VO

40

50

60

70

gold or copper

F I G . 12. Catalytic activity of Pd-Au(1 I I ) alloys (52) and Pd-Cu(I I I ) alloys (77) for the isornerization of neopentane at 290°C.

films that were unsintered or reconstructed by various chemical treatments showed isomerization selectivity between 0 and 30%. Figures 3 and 4 show the selectivity variations for unsupported Pd specimens. In order to study the effect of the surface composition of Pd catalysts, we diluted the surface of a Pd film with a Group IB metal (Au, Ag, or Cu) while maintaining the ( 1 1 1 ) texture. The changes in catalytic behavior are seen in Fig. 12. For catalytically active Pd-IB ( I I I ) alloy films, isomerization selectivity stays high (-80%), but also some activity increment for -10 atom% of the Group 1B metal is seen (52, 77) (Fig. 12). Because the exact surface composition is not known (some Group IB metal segregation to the surface is expected), we can not draw any conclusion about the active site size requirements. However, because isomerization suffers with alloying generally much less than does hydrogenolysis, the first reaction, most probably, requires smaller active ensembles (200-203). Therefore, in the case of Pd-supported catalysts in the SMSI state, decoration of the Pd surface by inactive species originating from the support might cause an increase in isomerization.

84

ZBIGNIEW KARPIfiSKI

Of course, before one should try to explain changes in the catalytic behavior of Pd/support catalysts in the SMSI state, it is necessary to determine these changes experimentally. After high-temperature reduction (between 500 and 600"C), several supported-Pd catalysts experienced changes (204).They are shown in Fig. 13. Pd/SiO,, Pd/TiO,, and Pd/CeO, exhibited marked changes in selectivity. For silica- and titania-supported Pd catalysts, the increase in isomerization selectivity was due to a drastic decline of hydrogenolysis. Pd/TiO, catalysts in the SMSI state exhibited a considerable decrease in hydrogenolysis of ethane (I 98, f99). Because the selectivity changes appeared after high-temperature reduction, an important question was to separate a possible catalytic effect of hydrogen left in the catalyst. This rather general phenomenon was exhaustively discussed by Paal and Menon (205). Here, with respect to this problem, it is relevant to mention that in the reaction of neopentane with H, on a silica-supported Pd catalyst, the hydrogenolysis is more strongly affected by the partial pressure of H2 than isomerization [more negative exponent on H? pressure in experimental power law (206)l. In the case of neopentane conversion, the effect of hydrogen left in the Pd/ SiO, catalyst was indeed large and, unfortunately, catalytic consequences of its presence were similar to those expected from morphological changes in the Pd surface in the SMSI state (isomerization selectivity increase, Fig. 13A: pretreatments b and e). It seemed advisable, therefore, to eliminate the hydrogen effect from the catalyst after high-temperature reduction (HTR), without destroying the SMSI state. Very careful flushing with Ar (or He) at 600°C of the Pd/SiO, catalyst pretreated in H, at 600°C removed strongly held hydrogen, but the very high selectivity for isomerization remained very high (204, 207) (Fig. 13). Our route in probing Pd surfaces is shown in Fig. 14. Certainly, this approach is oversimplified and does not take into account several important factors. In particular, control of the surfaces of unsupported samples, which have a low surface area, was not adequate. Reasonably good reproducibilities of kinetic results was the only means of control over obtaining samples characterized by a similar surface morphology. However, there is a need for verification of our results by study on real single-crystal Pd surfaces, with agood control of the surface and the gas phase. In particular, it seems that very careful surface analysis would exhibit the presence of mica-base constituents (Mg, Si, and Al) in the Pd surface layer. Also, a better control of surface topography is desirable. Defects introduced by various techniques (see Section I1,B) should be identified and their density should be determined. Development of transmission electron microscopy would bring more information in this respect in the near future. At present, on the basis of the experiments shown in Fig. 14, it seems

B

A 100

[

92.2

>;

Pd/titania

loo

s

Pd/cerio

80 70.6

c

._ > ._

60.0

c

60

w

-aJ

C

C

W

W

2 c

50.0

1

m

40

.-0

43.2

40

c

0

0

N

N .L

aJ

5 -

20

VI

n

(a)

lbl

Ic)

Id)

P r e t r e a t m e n t of Pd/silica

fe)

If)

fa)

(bl

lc1

P r e t r e a t m e n t of Pd c a t a l y s t

FIG. 13. Isomerization selectivity in the reaction of neopentane with HZ over supported Pd catalysts. (A) Silica-supported Pd, after pretreatments (reaction temperatures: 250-300°C). The level of 30%, marked as an upper selectivity level, is characteristic of sintered Pd powder; pretreatments in H?, He, or 0, (“C, hours). (a) H?, 300, 1; (b) H:, 450, 3; (c) H,, 450, 3; He, 500, I ; (d) H 2 , 450, 17; He, 500, 1; (e) H , , 600, 17; (f) H2, 600, 17; He, 600, 1 . (B) Titania-supported Pd (at 220-300°C): (a) H z , 250, I ; (b) H,, 500, 3; He, 500, 1 ; (c) 02,400, 0.5; H 2 , 250, 1. Ceria-supported Pd (at 230-300°C): (d) H,, 250, 1; (e) H,, 500, 1; He, 500, 1 ; (f) O,, 300, 0.5; H?, 250, 1. (Taken from Ref. 204.)

(a) how to use information collected for model surfaces in probing Pdlsupport catalysts 7

changes in isomerization selectivity, Pd/

Pd surfaces do not (63, 71)

1 m similarly i to Pd-Au

effect of surface campmition 7

YES

1111) . . 177) . ,

I

I

-

7 H, leh after HTR increases Sis, but after careful purge in Ar. 51s still v. hioh (2U7)

othereff8cts

7

e.g. hydrogeneffect? YES, but warablg

even better fhan Fd nst

,’

& dispersing alloy

Pd-IB over SiO,

i 1. Pd-Agpowder mare sekctive (58%) than Pd powder (5 35%) (207), lessman W-IE (111)

Pd ( l t l ) / m i c aisomerizes neoC,H, (Si = 70%) (51).other

Pd surfacesdo not (63,71)

EFFECT OF SUPPORTING ?

‘diluting Pd surface with I3 on SiO,

dilutiw Pd surfax with IB POSITIVEEFFECT !

surface composition = coml

I

effect of surface tapogra;

7

unorientedldefected Pd filmsshow low

further effect ? powdenzing ?

isomerization, Si = 0-150’’ (75)

NO !

pd powdersshow

S- isornerization selectiviv Si = 10 i30% (75)

dispersing Pd over SiO, (non-SMSI)

>

-

’I]

ANYEFFECT?

I I

YES ! Pd/SiO, behaves similarly in D, range

it i s w e t o sinter Pd powder in order to Wt isomeri2afionselectivity characteristic of

Pd powders reduced at 500” C show Si = 355

Pd (111) !

(75)

effect of Pd dispersjon ?

t”

7.1-79.1~ psi

NEGLIGIBLE !

FIG. 14. Chemical probing of Pd surfaces in various Pd catalysts (also in the SMSL state)

CATALYSIS BY PALLADIUM

87

that both surface topography and surface composition of Pd-based catalysts play an important role in determining their catalytic behavior in the reaction of neopentane with H,. Therefore, a considerable increase of isomerization selectivity caused by high-temperature reduction of the Pd/ SiO, catalyst may follow from both topographical and chemical changes in the metal surface. Isomerization selectivity of the Pd/SiO, catalyst after HTR reaches the level characteristic of Pd(1 I I ) films (-80%). On the other hand, drastic annealing of Pd powders in H, at 500°C increases isomerization selectivity, but only up to -30%. Also, an apparent structure insensitivity in this reaction exhibited by a series of Pd/SiO, catalysts (reduced at 300"C), characterized by metal dispersion between 7.1 and 85.6% (189), suggests a minor effect of Pd topography. As mentioned previously, real single-crystal study with better control of surface cleanliness and structure should verify our view as to the activity of Pd(l11) surfaces. On the other hand, the effect of inert additives to the Pd surface certainly brings about the increase in isomerization selectivity (films and powders). The missing links in the scheme shown in Fig. 14 are results of kinetic studies using Pd-IB/SiO, catalysts reduced to below the temperature at which the metal-support interactions begin. If, after introducing a Group 1B metal (Cu or Au), isomerization will reach a selectivity level characteristic of Pd/SiO, catalysts transformed into the SMSI state, then such a result should indicate that the Pd surface of the SMSI Pd/SiO, contains some foreign species responsible for such catalytic consequences. These species (necessarily originated from the silica support) must be rather uniformly distributed (not as patches), breaking the contiguity of Pd surface atoms. Such experiments are underway in the author's laboratory. As was mentioned previously, such a chemical probing constitutes an alternative to more sophisticated physical methods, and is particularly useful when we deal with more complex, highly dispersed metal-support catalysts. However, it is not advisable if one is not making use of other available techniques to verify conclusions that follow from kinetic studies. The comparison of chemisorption of H, (dissociative) with that of CO (two modes: linear and bridging) on Pd permits the determination of whether the contiguity of Pd atoms in the surface layer is interrupted after HTR. Because the X-ray diffraction spectra showed formation of the Pd,Si phase in the physical mixture of Pd powder and silica gel (Fig. 15), the presence of silicon atoms in the metal surface seemed an immediate indication of the catalytic changes. Figure 15 shows that more CO with respect to H, is chemisorbed on the surface of the Pd/SiO, catalyst that experienced HTR. Typically, for medium- and low-dispersion Pd catalysts after lowtemperature reduction (LTR), H/Pd, is not lower than CO/Pd, (Fig. 16); hence it may be concluded that after HTR the contiguity of surface Pd

88

ZBIGNIEW KARPtfiSKI

-200-Pd

-111tcc

=!

0

>;

Pdg Si (orthorhombicl

c .v)

C al c C

I , I

230

01

.

2

201

I

50

.

I

48

.

L

46

,

I

I

I

.

I

I

I

44 42 40 30 Diffraction a n a l e (2Bl/deg

1211

I

Y

I

I

36

I

I

34

FIG. 15. X-Ray diffraction spectrum of a physical mixture of Pd powder with silica gel after high-temperature reduction at 600°C. (Taken from Ref. 207.)

atoms is disturbed (hydrogen chemisorption is suppressed; on the contrary, CO may be adsorbed in a linear mode). Careful infrared study of CO chemisorbed by Pd/SiO, catalysts in an SMSI versus a non-SMSI state verified that after HTR, silicon species are distributed in the Pd surface layer. For the catalysts reduced at 300°C (LTR), the B/L intensity ratio (B = bridging CO; L = linearly bound CO) is a monotonic function of Pd particle size (Fig. 17). On the other hand, the B/L ratios for Pd/SiO, catalysts that experienced HTR show considerable departure from this universal curve (Fig. 17) (208). Of course, a relatively higher proportion of linearly bound CO for Pd/SiO, catalysts in the SMSI state is believed to follow from the existence of silicon, rather uniformly interdispersed in the metal surface, resembling the case of CO adsorption o n silica-supported Pd-Ag alloys (209). After HTR of the Pd/SiO, catalyst, oxygen is consumed (at room temperature) in a much greater amount than is needed to cover the palladium surface: the additional 0, is required for silicon oxidation of silicon to SO,. If the oxidation process is restricted only to surface Si species, then the oxygen consumption suggests the presence of Pd,Si (207).

89

CATALYSIS BY PALLADIUM

I

0.70

0.68

0.52

0 L

0

0.41

+ j -

0.40

0.34

Q

I 0.20

0.00 la)

Ib)

(C)

(d1

Pretreatment of Pd/siLica FIG.16. Chemisorptions of H, and CO over differently pretreated Pd/SiO, ; pretreatment in O,, H,, and Ar ("C, hours): (a) after 0,. 300.0.5;H,, 300, I ; Ar,450, 1 ; (b) after O,, 300, 0.5;H,,450,3; Ar, 500. I; (c) after O?, 300,0.5;H ? , 450, 17;Ar. 500, 1; (d) after 0,, 300, 0.5;H,, 600, 17;Ar, 600, I; (e) sample d 0,. 450. I; H,,300, I; Ar, 450, 1. (From Ref. 204.)

+

A complete restoration of the low level of isomerization selectivity was not possible by oxidation and low-temperature reduction (at 300°C). Hence, it was thought that some oxidized silicon species stay on the metal surface after regeneration (207). Suggested mechanism of interactions between Pd and SiO, during reduction at high temperatures is presented in Fig. 18. In the case of the Pd/TiO, catalyst, similar but somewhat lower changes in the isomerization selectivity after HTR (at 5OOOC) were observed (204). If Ti plays a role similar to that of Si (i.e., an inert diluent), then we can deduce that either the metal surface is covered to a lesser extent by Ti than by Si species or, even more feasible, some of the Ti agglommerates on the Pd surface [as in the case of Rh/TiO, (212)],whereas Si is more evenly distributed on the Pd surface, interrupting the contiguity of Pd atoms more efficiently than in the case of Pd/TiO,. The Pd/CeO, catalyst is a special case. HTR reduction (at 500°C) brings

90

ZBIGNIEW KARPINSKI

1

\

0 0.00

0.20

0.40

Dispersion o f

0.60

0.80

1.00

Pd P a r t i c l e s

FIG.17. The effect of Pd dispersion (in Pd/SiO?catalysts) on the B/L ratio. (a) LTR, 0.76 wt% Pd/SiOz; (b) HTR, 0.76 wt% Pd/SiO,; (c) regenerated (by oxidation and LTK),0.76 wt% Pd/SiO, ; (d) LTR. I .S8wt% PdiSiO?; (e) HTR, 1.58 wt% Pd/SiO, ; and ( f ) regenerated. I .S8wt% Pd/SiO?.0, Pd (208);+, Pd silicide (208);A,Pd/SiO, (210);U, Pd/SiO?(211);A , Pd/SiOz (148).(Reproduced from Ref. 208.)

about a decrease in isomerization selectivity. On the other hand, these catalysts, reduced at 250"C, exhibit quite a high selectivity, -50%. The decrease in isomerization selectivity accompanied by decrease in overall activity (204) suggest that the Pd surface is partially covered by cerium (ceria?) species after reduction at 250°C. More drastic reduction probably causes substantial blocking of the free (active) Pd surface by cerium, which seems very inactive in this reaction. Changes in isomerization selectivity in the reaction of neopentane with H, may be a convenient diagnostic parameter to determine if Pd interacts with a support. Sometimes such changes are very small, for example, the case of differently treated Pd/AI,O, catalysts (204). However, in this case,

91

CATALYSIS B Y PALLADIUM

0 Pd

m sio, oxygen vacancy ( o v l Pd atom trapped in ov

I Pd,Si,

L T R sample

n

d sio,

SiO?

"regenerated" sample

I

O2

sio,

HTR sample FIG.18. Model of interactions between palladium and silica in the course of high-temperature reduction and regeneration. (From Ref. 207.)

overall activity experiences large changes, especially after purging the catalyst with Ar at 600°C (followed by reduction at the same temperature). A substantial growth in activity was interpreted in terms of formation of Pd"+ species, which are superactive in hydrocracking (143) (Section IIl,C,2). C.

EVOLUTION OF Pd CATALYSTS I N THE COURSE OF

A

REACTION

The problem of evolution of Pd catalysts in the course of a reaction, important from a technological standpoint, was a subject of numerous papers (213-216). Here, we would like to emphasize only some aspects of the problem.

92

ZBIGNIEW K A R P I I ~ S K I

The Pd-support catalyst during a reaction may undergo various kinds of changes. At higher temperatures, in the presence of hydrogen, metal sintering seems inevitable, if, of course, the surface has not been considerably decreased as a result of prereduction in H,. On the other hand, if a reaction is carried out in H, at a lower temperature (at -60°C or below), when a /I-PdH formation and its decomposition (at the surface, the situation is a dynamic one) take place, one should expect considerable reconstruction of the Pd surface. Reorientation into the (1 11) structure is possible; however, many cracks in Pd crystallites develop, increasing the density of surface defects (5,21,81). In the process of hydrogen incorporation into Pd, the reacting hydrocarbon molecules (and solvent, if used) may play a role of a promoter or inhibitor (179); hence the presence of organic solvents and/or special additives (such as quinoline or lead in the Lindlar Pd-Pb catalyst) may be responsible for morphological changes in the Pd surface layer (5, 21, 217). During a catalytic reaction, the hydrocarbon molecule, after adsorption, undergoes transformation into the product, which further desorbs from the catalyst. However, as a result of dehydrogenation and polymerization, it may also form carbonaceous residues. This coke may be “softer” or “harder,” depending on the reaction conditions. During acetylene hydrogenation, buildup of higher molecular weight hydrocarbons (oligomers) on the surface of Pd/Al,O, catalysts reduces the selectivity toward halfhydrogenation (218). On the other hand, Al-Ammar and Webb (219) suggest that dissociatively adsorbed acetylene, which undoubtedly is a coke precursor, forms the primary layer on which associatively adsorbed acetylene is hydrogenated, with hydrogen originating from the primary layer. In addition, an incorporation of elemental carbon into a palladium lattice during hydrogenation of acetylene has been reported (220, 221). The amount of carbon incorporated reached the level of 0.13-0.15 (CIPd ratio). Another aspect of evolution of Pd catalysts during a reaction is the change in the electronic properties of palladium. Namely, both in the reaction of ethylene dimerization (in the absence of H, in the gas phase) and in the reaction of CO with H,, drastic changes in the activity and selectivity during initial stage of the reaction were correlated with the fact of formation of electron-deficient Pd species. In the reaction of CO + H, over well-reduced Pd and Pd-MgO/SiO, catalysts, during the initial period of time on-stream (the first to 2 h), the activity toward methanol formation is developed (158). During a similar period, the methane yield is stabilized at a low level. Because the authors correlate the activity for methanol formation with the existence of Pd“+ species (see Section III,C,3), the conclusion is that these active centers

CATALYSIS BY PALLADIUM

93

are formed during the induction period via reaction of Pdowith the reaction mixture. On the other hand, the participation of Pd2+ and/or Pd+ species as active sites in the reaction of dimerization of ethylene was suggested some years ago (222). Recently, two independent groups established that Pd + species in PdJX and Pd/Y zeolites are sites of highest activity (124-126, 223). Starting with a Pd2+/Yor Pd2'/X catalyst, they monitored its activity in the course of reaction. Parallel study with XPS, ESR, and solid-state NMR confirmed that Pd+ is an active species in the reaction. Its concentration grows with time on-stream. V. Conclusions-Suggestions for Future Work Although the specificity of heterogeneous Pd catalysts has been recognized for many years, the explanation of its action is still far from being satisfactory. There is no doubt that at lower temperatures, in the presence of hydrogen, the formation of p-palladium hydride has a great effect on the reaction course and/or on the reconstruction of the Pd surface. Another important aspect is the presence of electron-deficient palladium in highly dispersed Pd catalysts, although, at present, all the collected evidence (Section 111) should be verified by more sophisticated physical techniques. In particular, an interpretation of the positive shift of electron core-levels (XPS) of metals highly dispersed over insulating carriers should be given an unequivocal explanation (initial-state versus final-state effect). Chemical probing of Pd surfaces in supported catalysts is a useful alternative to physical studies. Several reactions are very sensitive to the kind of Pd surface, making them genuine diagnostic tools for surface characterization. Unfortunately, unlike in the case of platinum, comprehensive information for well-defined Pd surfaces is not plentiful. This fact makes all comparisons with supported catalysts, which are complex by nature, more difficult. Hence, studies involving various model forms of Pd catalysts, such as flat, stepped, and kinked surfaces of palladium, are urgently needed. Evidence for such a need follows from a somewhat strange fact that the estimation of specific activities of various crystallographic planes of Pd used to be indirectly obtained from catalytic studies of supported palladium (99). A relative population of (1 1 1) and (100)planes in supported Pd catalysts was inferred from the IR study of adsorbed CO, by comparing the results obtained with respective data reported for single crystals (100). Although the results of such an estimation may be correct, another work (191) warns us that the bounding faces of Pd crystallites

94

ZBIGNIEW KARPINSKI

might be defective and contain more highly coordinatively unsaturated sites than plane atoms. Now the question arises as to whether these highly unsaturated metal sites control the overall activity of the whole metal surface, even if they are not plentiful. The turnover frequencies of various sites may differ considerably from each other. More attention should be attached to the problem of p-PdH involvement in the catalyzed reaction. Careful studies should reveal whether the pPdH phase is formed or decomposed under specified reaction conditions. Surface fugacity of hydrogen in the reaction conditions is difficult to predict, and the presence of a reactant hydrocarbon and/or solvent may inhibitlor accelerate the process of incorporation of hydrogen into the bulk of Pd. Another problem is to verify the recent suggestion as to the formation of Pd hydride also in the case of very small Pd particles. In conclusion, irrespective of the relevance of the experiment of Fleischmann and Pons (224), palladium should remain an important objective of fundamental and technological studies, certainly in the field of heterogeneous catalysis. REFERENCES 1 . Maitlis, P. M., in “The Organic Chemistry of Palladium Vol. 2: Catalytic Reactions.” Academic Press, New York, 1971. 2 . Rylander, P. N., “Hydrogenation Methods.” Academic Press, London, 1985. 3. “The Catalytic Reaction Guide.’’ Johnson Matthey, Royston, England, 1988. 4 . Freifelder, M., in “Practical Catalytic Hydrogenation,” p. 96. Wiley (Interscience), New York, 1971. 5 . Maier, W. F., in “Catalysis of Organic Reactions” (P. N. Rylander, H. Greenfield, and R. L. Augustine, eds.), p. 233. Dekker, New York, 1988. 6 . Ertl, G., Prigge, D., Schloegl, R . , and Weiss, M., J. Catal. 79, 359 (1983). 7. Ertl, G., J. Vac. Sci. Techno/. A l , 1247 (1983). 8. Kirch, G., Schwab, E., Wicke, E., and Zuchner, H., Proc. Int. Congr. Curd., 8th, West Berlin 4, 209 (1984). 9. Beard, C. B., and Ross, P. N., J. Phys. Chem. 90, 6811 (1986). 10. Tauster, S. J . , Fung, S . C., and Garten, R. L., J . A m . Chem. Soc. 100, 170 (1978). 11. Davis, S. M., Zaera, F., and Somojai, G. A., J . Am. Chem. Soc. 104, 7453 (1982). 12. Davis, S. M., Gillespie, W. D., and Somorjai, G . A . , J . Card. 83, 131 (1983). 13. Davis, S . M., Zaera, F., and Somojai, G. A , , J . Catal. 85, 206 (1984). 14. Davis, S. M., Zaera, F., Gordon, B. E., and Somorjai, G . A,, J. Card. 92,240 (1985). 15. Garin, F., Aeiyach, S . , Legare, P., and Maire, G., J. Card. 77, 323 (1982). 16. Dauscher, A . , Garin, F., and Maire, G., J. Catal. 105, 233 (1987). 17. Gentle, T. M . , and Muetterties, E. L., J . Phys. Chem. 87, 2469 (1983); Rucker, T. G., Logan, M. A., Gentle, T. M., Muetterties, E. L., and Somojai, G. A,, J . Phys. Chem. 90, 2730 (1986); Logan, M. A., Rucker, T. G., Gentle, T. M., Muetterties, E. L., and Somorjai, G.A., J . Phys. Chem. 90, 2709 (1986). 18. Tysoe, W. T., Nyberg, G . L., and Lambert, R. M., J.C.S.Chem. Commun. p. 623 (1983); Surf. Sci. 135, 128 (1983). 19. Sasselmann, W., Woratschek, B . , Ertl, G., and Kiippers, J., Surf. Sci. 130,245 (1983).

CATALYSIS BY PALLADIUM

95

20. Massardier, J., Bertolini, J. C., and Renouprez, A., Proc. Int. Congr. Catal., 9th, Calgary 3, 1222 (1988). 21. Ulan, J. G., Maier, W. F., and Smith, D. A,, J . Org. Chem. 52, 3132 (1987). 22. Ledoux, M. J., Nouu. J. Chim. 2, 9 (1978). 23. Ledoux, M. J., Gault, F. G., Bouchy, A., and Roussy, G., J . C . S . Faraduy 174,2652 ( 1978).

24. Horiuti, J., and Polanyi, M., Truns. Faraday Soc. 30, 1164 (1934). 25. Burwell, R. L., Jr., Arc. Chem. Res. 2, 289 (1969). 26. Rooney, J . J., J. C a d . 2, 53 (1963). 27. Mohsin, S. B., Trenary, M., and Robota, H. J., J. Phys. Chem. 92, 5229 (1988). 28. Zaera, F., and Somojai, G. A., J. A m . Chem. Soc. 106, 2288 (1984). 29. Hattori, T., and Burwell, R. L., Jr., J. Phys. Chem. 83, 241 (1979); Beebe, T. P., Jr., and Yates, J . T., Jr., J. Am. Chem. Soc. 108, 663 (1986). 30. Burwell, R. L., Jr., Langmuir 2, 3 (1986). 3Z. Gates, J . A., and Kesmodel, L. L., J . Phys. Chem. 76, 4281 (1982); Surf. Sci. 124, 68 (1983); Kesmodel, L. L., Waddill, G. D., and Gates, J. A.,Surf. Sci. 138,464 (1984). 32. Tysoe, W. T., Nyberg, G. L., and Lambert, R. M., J. Phys. Chem. 90, 3188 (1986). 33. Margitfalvi, J., Guczi, L., and Weiss, A. H., J . Catal. 72, 185 (1981). 34. Moses, J. M., Weiss, A. H., Matusek, K., and Guczi, L., J . Catal. 86, 417 (1984). 35. Hub, S., and Touroude, R., J. Catal. 114,411 (1988). 36. Boudart, M., and Hwang, H. S . , J. Catal. 39,44 (1975). 37. Moraweck, B., Clugnet, G., and Renouprez, A., J. Chim. Phys. Phys.-Chim. B i d . 83, 265 (1986).

38. Jobic, H., and Renouprez, A., J. Less-Common Met. 129, 311 (1987). 39. Kemball, C., Proc. R . Soc. London, S e r . A 217, 376 (1953). 40. Anderson, J. R., and Kemball, C., Proc. R . Soc. London, Ser. A 223, 361 (1954). 41. Kernball, C., Trans. Faraday Soc. 50, 1344 (1954). 42. Gault, F. G., and Kemball, C., Trans. Faruday Soc. 57, 1781 (1961). 43. Kemball, C . , Adu. Cutal. 11, 223 (1959). 44. Rooney, J . J., Chem. Br. 2, 242 (1966). 45. Kemball, C., Card. Rev. 5 , 33 (1971). 46. Clarke, J. K. A., and Rooney, J . J., Adu. Catal. 25, 125 (1976). 47. Guczi, L., and Karpinski, Z., J . Card. 56, 438 (1979). 48. Schrage, K . , and Burwell, R. L., Jr., J . A m . Chem. Soc. 88, 4549 (1966). 49. Burwell, R. L., Jr., and Schrage, K., Discuss. Faraday Sue. 41, 1438 (1966). 50. Brown, R., Kemball, C., and Sadler, I. H., Proc. I n t . Congr. Card., 9th, Calgary 3, 1013 (1988).

51. Karpinski, Z., Nouu. J . Chim.4, 561 (1980). 52. Karpifiski, Z., J . Catal. 77, 118 (1982). 53. Clarke, J. K. A., McMahon, E., and 0 Cinneide, A. D., Proc. In?. Congr. Catal.. 5th, Miami Beach, Flu., 1972 1,685 (1973); Plunkett, T. J . , and Clarke, J. K. A., Z . Phys. Chem. 88, 329 (1974). 54. Sinfelt, J. H., Adu. Catal. 23, 91 (1973). 55. Anderson, J. R., A d v . Catal. 23, 1 (1973). 56. Van Broekhoven, E. H., and Ponec, V . , J . Mol. Catal. 25, 109 (1984). 57. Carter, J. L., Cusumano, J. A,, and Sinfelt, J. H., J . Cutal. 20, 223 (1971). 58. Kikuchi, E.,Tsurumi, M., and Morita, Y.,J . Catal. 22, 226 (1971). 59. Karpinski, Z., and Koicielski, T., J . Catal. 63, 313 (1980). 60. Matsumoto, H., Saito, Y . , and Yoneda, Y.,J. Catal. 22, 182 (1971). 61. Hajek, M., Corolleur, S., Corolleur, C., Maire, G., 0 Cinneide, A., and Gault, F. G., J . Chim. Phys. 71, 1329 (1974).

96

ZBIGNIEW K A R P I I ~ S K I

62. PaA, Z., and Tettnyi, P., React. Kiner. Catal. Lett. 12, 131 (1979). 63. Anderson, J. R., and Avery, N. R., J . Catal. 5 , 446 (1966). 64. Muller, J. M., and Gault, F. G., Proc. I n t . Congr. Catal., 4th, M o s c o w , 1968 Pap. No. 15 (1971). 65. Garin, F., and Gault, F. G., J . A m . Chem. Soc. 97, 4466 (1975). 66. Finlayson, 0. E., Clarke, J. K. A., and Rooney, J . J., J . C . S . Faraday I80, 191 (1984). 67. Clarke, J. K. A . , Hegarty, B. F., and Rooney, J. J., J . C . S . Faraday I84,2511 (1988). 68. Karpinski, Z . , and Guczi, L., J.C.S.Chem. Commun. p. 563 (1977). 69. McKervey, M. A., Rooney, J. J., and Samman, N. G., J . Catal. 30, 330 (1973). 70. Anderson, J. R., and Avery, N. R. J . Catal. 2, 542 (1963). 71. Boudart, M., and Ptak, L . D., J . Catal. 16, 90 (1970). 72. Gault, F. G., Adu. Catal. 30, l(1981). 73. Karpinski, Z . , and KoScielski, T., J . Catal. 56,430 (1979); KoScielski, T., and Karpinski, Z., 2. Phys. Chem. 111, 125 (1978). 74. Sarkany, A,, Guczi, L., and Tetknyi, P., Acta Chim. Acad. Sci. Hung. 96, 27 (1978). 75. Karpinski, Z . , Juszczyk, W., and Pielaszek, J . , J . C . S . Furaday Z83, 1293 (1987). 76. Leclercq, G., Leclercq, L., and Maurel, R., J . Catal. 44, 68 (1976). 77. Karpinski, Z., Juszczyk, W., and Stachurski, J., J.C.S.Furaday I81, 1447 (1985). 78. Chen, .I. J . , and Ruckenstein, E . , J . Phys. Chem. 85, 1606 (1981). 79. Ruckenstein, E., and Chen, J. J., J . Colloid Interface Sci. 86, 1 (1982). 80. Gao, S . , and Schmidt, L. D., J . Catal. 115, 356 (1989). 81. Pielaszek, J., Catal. Lett. 1, 117 (1988). 82. Christmann, K . , and Ertl, G., Surf. Sci. 33, 254 (1972); Thin Solid Films 28, 1 (1975). 83. Vook, R . W., De Cooman, B. C., and Vankar, V. D., Scanning Electron Microsc. 4, 1635 (1983), and discussion following this paper. 84. Paal, Z., in “Hydrogen Effects in Catalysis” (Z. Paal and P. G. Menon, eds.), p. 293. Dekker, New York, 1988. 85. Boudart, M., Aldag, A. W., Ptak, L. D., and Benson, J . E., J . Catal. 11, 35 (1968). 86. O’Donohue, C., Clarke, J. K. A., and Rooney, J. J., J.C.S. Faraday 176, 345 (1980). 87. PaB1, Z., and Tttenyi, P., Appl. Cutal. 1, 9 (1981). 88. Sarkany, A., J.C.S. Faraduy Z82, 103 (1986). 89. Poutsma, M. L., Elek, L. F., Ibarbia, P. A., Risch, A. P., and Rabo, J. A., J . Card. 52, 157 (1978). 90. Fischer, F., Tropsch, H., and Dilthey, P., Brennsr.-Chem. 6, 265 (1925). 91. Kratel, R., Ph.D. Thesis, Tech. Univ. of Berlin-Charlottenburg, 1937. Cited in Ref. 92. 92. Fajula, F., Anthony, R. G., and Lunsford, J. H . , J . Catal. 73, 237 (1982). 93. Ryndin, Y. A., Hicks, R. F . , Bell, A. T., and Yermakov, Y.I., J . Card. 70,287 (1981). 94. Sudhakar, C., and Vannice, M. A., J . C a r d . 95, 227 (1985). 95. Kelly, K. P., Tatsumi, T., Uematsu, T., Driscoll, D. J., and Lunsford, J . H., J . Card. 101, 396 (1986). 96. Kok, G. A., Noordermeer, A,, and Nieuwenhuys, B. E., Surf. Sci. 135, 65 (1983). 97. Bader, S. D., Blakely, J . M., Brodsky, M. R., Friddle, R. J . , and Panosh, R . L., Surf. Sci. 74, 405 (1978). 98. Vannice, M. A . , J . Catal. 40, 129 (1975). 99. Hicks, R. F., and Bell, A. T., J . Catal. 90, 205 (1984). 100. Bradshaw, A. M., and Hoffmann, F. M., Surf. Sci. 72, 513 (1978); Hoffmann, F. M., Surf. Sci. Rep. 3,107 (1983); Ortega, A., Hoffmann, F. M., and Bradshaw, A. M., Surf. Sci. 119, 79 (1982). 101. Temkin, 0. N., and Bruk, L. G., Usp. Khim. 52, 206 (1983).

CATALYSIS BY PALLADIUM

97

102. Baetzold, R. C., J . Chem. Phys. 55, 4363 (1971). 103. Baetzold, R. C., J . Catal. 29, 129 (1973). 104. Wagner, C. D., Riggs, W. M., Davis, L. E., Moulder, J. F., and Muidenberg,

G. E., “Handbook of X-Ray Photoelectron Spectroscopy.” Perkin-Elmer, Eden Prairie, Minnesota, 1979. 105. Nefedov, V. I., “X-Ray Photoelectron Spectroscopy of Chemical Compounds,” p. 26. Khimia, Moscow, 1984. 106. Mason, M. G . , Phys. Rev. E 27, 747 (1983). 107. Kohiki, S . , and Ikeda, S., Phys. Rev. E 34, 3786 (1986). 108. DiCenzo, S. B., and Wertheim, G. K., Comments Solid State Phys. 11, 203 (1985). 109. Wertheim, G . K., DiCenzo, S. B., and Buchanan, D. N. E., Phys. Rev. E 33, 5384 (1986). 110. Kohiki, S . , Appl. Surf. Sci. 25, 81 (1986). 111. Takasu, Y . , Unwin, R., Tesche, B., Bradshaw, A. M., and Grunze, M., Surf. Sci. 77, 219 (1978). 112. Takasu, Y., Akimaru, Y., Kasahara, K., Matsuda, Y., Miura, H., and Toyoshima, I., (a) Proc. Int. Conf. Vac. Metall., 7th, Tokyo Pap. 23-2, p. 719 (1982); (b) J . Am. Chem. Soc. 104, 5249 (1982). 113. VCdrine, J., Dufaux, M., Naccache, C., and Imelik, B., J.C.S. Faraday 174,440(1978). 114. Tarasov, A. L., Shvets, A. A,, Zaitsev, A. V., and Kazanskii, V. B., Kinet. Katal. 27, 1181 (1986). 115. Hub, S., Hilaire, L., and Touroude, R., Appl. Catal. 36, 307 (1988). 116. Ryndin, Y.A., Nosova, L. V., Boronin, A. L., and Chuvilin, A. L., Appl. Catal. 42, 131 (1988). 117. Wertheim, G . K., and DiCenzo, S. B., Phys. Rev. E 37, 844 (1988). 118. Fritsch, A., and Legare, P., Surf. Sci. 162, 742 (1985). 119. Legart, P., Finck, F., Roche, R., and Maire, G., Z . Phys. D:At., Md. Clusters 12, 19 (1989). 120. See, e.g., Vasudevan, S., Cosyns, J., Lesage, E., Freund, E., and Dexpert, H., in “Scientific Bases for the Preparation of Heterogeneous Catalysts” (G. Poncelet P. Grange, P. A. Jacots, eds.), p. 463. Elsevier, Amsterdam, 1983; Lieske, H., Lietz, G., Hanke, W., and Volter, J., Z . Anorg. Allg. Chem. 527, 135 (1985). 121. Huizinga, T., and Prins, R., J. Phys. Chem. 87, 173 (1983). 122. Naccache, C., Dutel, J. F., and Che, M., J . C a r d 29, 179 (1973). 123. Ben Taarit, Y., VCdrine, J. C., Dutel, J. F., and Che, M., J. Magn. Reson. 31, 251 (1973). 124. Michalik, J., Lee, H., and Kevan, L., J . Phys. Chem. 89, 4282 (1985). 125. Michalik, J., Narayama, M., and Kevan, L., J. Phys. Chem. 89, 4553 (1985). 126. Gosh, A. K., and Kevan, L., J. Phys. Chem. 93, 3747 (1989). 127. Berger, P. A., and Roth, J. F., J. Catal. 4, 717 (1965). 128. Geus, J. W., Proc. Int. Congr. Catal., 3rd, Amsterdam, 1964 1, 597 (1965). 129. Turkevich, J . , Nozaki, F., and Stamirez, D., Proc. Int. Congr. Catal., 3rd, Amsterdam, 1964 1, 586 (1965). 130. Zakumbaeva, G . D., Naidin, V. A., Dagirov, T. S., Litvyakova, E. N., and Kurashvili, L. M., Kinet. Katal. 23, 126 (1982). 131. Shubin, V. E., Shvets, A. A., Savel’eva, G. A., Popova, M. N., and Kazanskii, V. B., Kinet. Katal. 23, 1153 (1982). 132. Parvulescu, V., Sandulescu, I., Grecu, N., Contescu, C., and Nicolescu, I. V., Rev. Roum. Chim. 33, 3 (1988). 133. Rakai, A., Ph.D. Thesis Univ. P. et M. Curie, Paris, 1989.

98

ZBIGNIEW KARPINSKI

134. Karpinski, Z., Juszczyk, W., Herbich, J., and Michalik, J., Catul. Lett. 5, 197 (1990). 135. Margitfalvi, J . , Gobolos, S . , Kern-Talas, E., Hegedus, M., and Simon, P., Acta Chim. Hung. 124, 181 (1987). 136. Bychkova, T. V., Lunina, E. V., Selivanovskii, A . K., Nikiel, E. A , , Smagina, T. Y . ,

and Strakhov, B. V., Zh. Fiz. Khim. 55, 2599 (1981). 137. Bychkova, T . V., Lunina, E. V., Selivanovskii. A. K., and Strakhov, B. V., Zh. Fiz.

Khim. 55, 2697 (1981). 138. Figueras, F., Gomez. R., and Primet, M., Adu. Chem. Ser. No. 121, 480 (1973). 139. Lokhov, Y . A . , and Davydov, A . A . , Kinet. Katal. 21, 1515 (1980). 140. Romannikov, V. N., Ione, K . G . , and Pedersen, L. A , , J . Catal. 66, 121 (1980). 141. Sheu, L.-L., Knozinger, H., and Sachtler, W. M. H., Catal. Lett. 2, 129 (1989). 142. Sheu, L.-L., Knozinger, H., and Sachtler, W. M. H., J . A m . Chem. Soc. 111, 8125

( I 989). 143. Juszczyk, W., Karpinski, Z., Ratajczykowa, I., Stanasiuk, Z., Zielinski, J . , Sheu, L.-L., and Sachtler, W. M. H . , J . Catal. 120, 68 (1989). 144. Martin, M. A . , Pajares, J. A , , and Gonzalez-Tejuca, L., Z . Phys. Chem. 140,107 (1984). IR bands of Pd+-CO visible in Fig. 2 (sample la, p. 112) of this paper, although not discussed in the text. 145. Tarasov, A . L . , Shvets, V. A . , and Kazanskii, V. B., Kinet. Katal. 30, 396 (1989). 146. Vannice, M. A., Wang, S.-Y., and Moon, S. H., 1. Card. 71, 152 (1981). 147. Vannice, M. A . , and Wang, S.-Y., J . Phys. Chern. 85, 2543 (1981). 148. Palazov, A . , Chang, C. C., and Kokes, R. J., J . Card. 36, 338 (1975). 149. Palazov, A . , Kadinov, G., Bonev, C., and Shopov, D., J . Catul. 74, 44 (1982). 150. Burwell, R. L., Jr., and Brenner, A,, J . Mol. Card. 1, 77 (1975/76). 151. Brenner, A . , and Hucul, D. A., Inorg. Chem. 18, 2836 (1979). 152. Bjorklund, R. B., and Burwell, R. L., Jr., J . Colloid Interface Sci. 70, 383 (1979). 153. Tzou, M. S . , Teo, B. K., and Sachtler, W. M. H., J . Catal. 113, 220 (1988). 154. Lesage-Rosenberg, E., Vlaic, G., Dexpert, H., Lagarde, P., and Freund, E., Appl. Catal. 22, 21 I (1986). 155. Homeyer, S. T., and Sachtler, W. M. H., in “Zeolites: Facts, Figures, Future” (P. A . Jacobs and R. A . van Santen, eds.), Part B, p. 975. Elsevier, Amsterdam, 1989. 156. Gallezot, P., Catal. Rev.-Sci. Eng. 20, 121 (1979). 157. Homeyer, S. T., Karpinski, Z., and Sachtler, W. M. H., Recl. Trau. Chim. Pays-Bas 109, 81 (1990). 158. Driessen, J. M., Poels, E. K., Hinderrnann, J. P., and Ponec, V . , J . C u r d 82, 26 (1 983). 159. Poels, E. K . , Ph.D. Thesis, Leiden, 1984. 160. Van der Lee, G., and Ponec, V., Cazul. Reu.-Sci. Eng. 29, 183 (1987). 161. Che, M., and Bennett, C. O., A d a Catal. 36, 55 (1989). 162. Nandi, R. K., Pitchai, R., Wong, S . S . , Cohen, J. B., Burwell, R. L., Jr., and Butt, J. B., J . Catal. 70, 298 (1981). 163. Carturan, G., Facchin, G., Cocco, G., Enzo, S., and Navazio, G., J . Catal. 76, 405 (1982). 164. Borodzinski, A., DuS, R., Frak, R., Janko, A . , and Palczewska, W., Proc. Int. Congr. Catal., 6th, London, 1976 2, 150 (1977). 165. Fuentes, S. , and Figueras, F., J . C . S . Furuday 174, 174 (1978). 166. Vannice, M. A., and Chou, P., Proc. I n t . Congr. Catal., 8th, West Berlin 5,99 (1984). 167. Carturan, G., Facchin, G., Gottardi, V . , Guglierni, M., and Navazio, G., J . Non-Cryst. Solids 48, 219 (1982). 168. Dalla Betta, R. A . , and Boudart, M . , Proc. Int. Congr. Catal., 5th, Miami Beach, Flu., 1972 2, 1329 (1973).

CATALYSIS BY PALLADIUM

99

169. Foger, K., and Anderson, J. R., J. Card. 54, 318 (1978). 170. Homeyer, S . T., Karpinski, Z., and Sachtler, W. M. H . , J . C a r d . 123, 60 (1990). 171. Tolbert, M. A., Mandich, M. L., Halle, L. F., and Beauchamp, J. L., J . A m . Chem. Soc. 108, 5675 (1986). 172. Kikuzono, Y., Kagami, S. , Naito, S . , Onishi, T., and Tamaru, K . , Faraday Discuss. Chem. Soc. 72, 135 (1981). 173. Deligianni, H., Mieville, R. L., and Peri, J . B., J . Catal. 95, 465 (1985). 174. Hindermann, J. P., Kiennemann, A,, Chakor-Atami, A., and Kieffer, R., Proc. Int. Congr. Cntal., 8th, West Berlin 2, 163 (1984). 175. Anikin, N . A., Bagatoriants, A. A , , Zhidomirov, B. M . , and Kazanskii, V. B., Russ. J . Phys. Chem. 57, 393 (1983). 176. Wang, S.-Y., Moon, S. H . , and Vannice, M. A,, J. Card. 71, 167 (1981). 177. Boudart, M., Adu. Catal. 20, 153 (1969). 178. Bond, G. C., Surf. Sci. 156, 966 (1985). 179. Palczewska, W., Adu. Card. 24, 245 (1975). 180. Palczewska, W., in “Hydrogen Effects in Catalysis” (Z. Paal and P. G. Menon, eds.), p. 373. Dekker, New York, 1988. 181. Rorris, E., Butt, J. B., Burwell, R. L., Jr., and Cohen, J. B., Proc. Int. Congr. Catal., 8th, West Berlin 4, 321 (1984). 182. Pitchai, R., Wong, S. S . , Takahashi, N., Butt, J. B., Burwell, R. L., Jr., and Cohen, J. B . , J . Catul. 94, 478 (1985). 183. Nandi, R. K., Molinaro, F., Tang, C . , Cohen, J . B., Butt, J. B., and Burwell, R. L . , Jr., J. Carol. 78, 289 (1982). 184. Gubitosa, G., Berton, A., Camia, M., and Pernicone, N . , in “Preparation of Catalysts” (G. Poncelet, P. Grange, and P. A. Jacobs, eds.), Vol. 3, p. 431. Elsevier, Amsterdam, 1983. 185. Boitiaux, J. P., Cosyns, J., and Vasudevan, S . , in “Preparation of Catalysts” (G. Poncelet, P. Grange, and P. A. Jacobs, eds.), Vol. 3 , p. 123. Elsevier, Amsterdam, 1983; Appl. Catal. 6, 41 (1983); Appl. Catal. 15, 317 (1985). 186. Sarkany, A,, Weiss, A. H., and Guczi, L., J. Card. 98, 550 (1986). 187. Janko, A., Palczewska, W., and Szymerska, I . , J . Catal. 61, 264 (1980). 188. Ruckenstein, E., and Sushumna, I., in “Hydrogen Effects in Catalysis” (Z. Paal and P. G. Menon, eds.), p. 259. Dekker, New York, 1988. 189. Karpinski, Z., Butt, J . B., and Sachtler, W. M. H., J. Catal. 119, 521 (1989). 190. Kung, H . H., Pellet, R. J . , and Burwell, R. L., Jr., J. Am. Chem. SOC.98,5603 (1976). 191. Eskinazi, V., and Burwell, R. L., Jr., J . Catul. 79, 118 (1983). 192. Del Angel, G. A., Coq, B., Ferrat, G., Figueras, F., and Fuentes, S . , Surf. Sci. 156, 943 (1985). 193. Corma, A., Martin, M. A., Pajares, J. A , , Perez-Pariente, J., Avalos, M., and Jose Yacaman, M., J. Mol. Catal. 48, 199 (1988). 194. Karpinski, Z., Chuang, T.-K., Katsuzawa, H., Butt, J. B., Burwell, R. L., Jr., and Cohen, J. B . , J . C a r d . 99, 184 (1986). 195. Yates, D. J. C., and Sinfelt, J. H., J. Card. 8, 318 (1967). 196. Moss, R. L., Pope, D., Davis, B. J., and Edwards, D. H., J . C a d . 58, 219 (1979). 197. Aben, P. C., Platteew, J . C., and Stouthamer, B., R e d . Trau. Chim. Pays-Bas 89,449 ( 1970). 198. Bracey, J . D., and Burch, R., J. C a r d . 86, 384 (1984). 199. Anderson, J. B. F . , Bracey, J. D., Burch, R., and Flambard, A. R., Proc. Int. Congr. C a r d , 8rh. West Berlin 5 , 1 1 1 (1984). 200. Clarke, J . K . A., Chem. Rev. 75, 291 (1975).

100 201. 202. 203. 204. 205. 206. 207. 208. 209. 210. 211.

212. 213. 214. 215. 216. 217. 218. 219. 220. 221. 222.

223. 224.

ZBIGNIEW KARPINSKI

Sachtler, W. M. H., and van Santen, R. A., Adv. Catal. 26, 69 (1977). Ponec, V., Carat. Rev.-Sci. Eng. 11, 1 (1975); Adv. C a r d 32, 149 (1983). Karpidski, Z., and Clarke, J. K. A., J.C.S. Faraday 171, 893 (1975); 71,2310 (1975). Juszczyk, W., Karpiriski, Z., Pielaszek, J., Ratajczykowa, I., and Stanasiuk, Z., Proc. Inr. Congr. Carat., 9rh, Calgary 3, 1238 (1988). PaB1, Z., and Menon, P. G., Card. Rev. Sci. Eng. 25,229 (1983);Paal, Z., in “Hydrogen Effects in Catalysis” (Z. Pa61 and P. G. Menon, eds.), p. 449. Dekker, New York, 1988. Juszczyk, W., and Karpinski, Z., Reacr. Kinet. C a r d Lett. 37, 367 (1988). Juszczyk, W., and Karpinski, Z., J. Carat. 117, 519 (1989). Sheu, L.-L., Karpinski, Z., and Sachtler, W. M. H., J . Phys. Chem. 93,4890 (1989). Soma-Noto, Y . , and Sachtler, W. M. H., J . Catal. 32, 315 (1974). Hicks, R. F., Yen, Q.4.. and Bell, A. T., J . Catal. 89, 498 (1984). Ichikawa, S., Poppa, H., and Boudart, M., 1.Carat. 91, I (1985); Ichikawa, S . , Poppa, H., and Boudart, M., in “Catalytic Materials: Relationship between Structure and Reactivity” (T. E. Whyte Jr., R. A. Dalla Betta, E. G. Derouane, and R. T. K . Baker, eds.), ACS Symp. Ser. No. 248, p. 439. Am. Chem. SOC.,Washington, D.C., 1984. Schepers, F. J., van Senden, J. G., van Broekhoven, E. H., and Ponec, V., J. Carat. 94,400 (1985). Bartholomew, C. H., Card. Rev.-Sci. Eng. 24, 67 (1982). Butt, J . B., in “Catalysis-Science and Technology” (J. R. Anderson and M. Boudart, eds.), Vol. 6, p. 1 . Akademie-Verlag, Berlin, 1985. Butt, J. B., in “Catalyst Deactivation” (E. E. Petersen and A. T. Bell, eds.), p. 3. Dekker, New York, 1987. “Deactivation and Poisoning of Catalysts” (J. Oudar and H. Wise, eds.). Dekker, New York, 1985. Schlogl, R., Noack, K., Zbinden, H., and Reller, A., Helv. Chim. Acta 70,627 (1987). Heck, R. M., and Smith, T. G., Ind. Eng. Chem., Process Des. Dev.9, 537 (1970). Al-Ammar, A. S., and Webb, G . , J.C.S. Faraday 174, 657 (1978). Stachurski, J . , and Frackiewicz, A., J. Less-Common Met. 108, 249 (1985). Ziemecki, S. B., Jones, G. A., Swartzfager, D. G., Harlow, R. L., and Faber, J., Jr., J . Am. Chem. Soc. 107, 4547 (1985). Lapidus, A. L., Mal’tsev, V. V., Shpiro, E. S., Antoshin, G. V., Garanin, V. I., and Minachev, K. M., Izv. Akad. Nauk SSSR, Ser. Khim. p. 2454 (1977); Lapidus, A. L., and Mal’tsev, V. V., Acta Phys. Chim. 24, 195 (1978). Bai, X . , Wang, G., and Zheng, L., Proc. Ini. Congr. Carat., 9rh, Calgary 1,340 (1988). Fleischmann, M., and Pons, S., J . Electroanat. Chem. 261, 301 (1989).

ADVANCES IN CATALYSIS, VOLUME 37

The Bond-Order Conservation Approach to Chemisorption and Heterogeneous Catalysis: Applications and Implications EVGENY SHUSTOROVICH Corporate Research Laboratories Eastman Kodak Company Rochester, New York 14650

I. Introduction The worldwide value of products made via catalytic technology is in excess of one trillion dollars per year (1). In particular, one-sixth of the value of all goods manufactured in the United States involves catalytic processes (2), and heterogeneous catalysis leads the way. This tremendous economic impact is the major thrust of intensive efforts in both industry and academia to understand heterogeneous catalytic processes at the molecular level. Still, this understanding remains frustratingly discrete and insufficient. As a result, mechanisms of most heterogeneous catalytic reactions are not known at the microscopic level and, therefore, remain highly speculative (1-5). Understanding mechanisms of chemical reactions is impossible without knowledge of reaction energy profiles. Ultimately, these profiles should be calculated by quantum mechanical techniques. So far, the progress has been impressive for gas-phase reactions (6,7) but rather modest for surface reactions. Fundamentally, the main problem is the absence of a generally accepted microscopic theory of chemisorption resulting in a variety of competing models designed to treat separate aspects of chemisorption and surface reactivity (8-17). In practical terms, the main trouble is the low accuracy of calculated heats Q of adsorbate chemisorption (see Section 101 Copyright 0 1990 by Academic Press, Inc. All rights of reproduction in any form reserved.

102

EVGENY SHUSTOROVICH

m

p,

Gasphase

x

Ec

QP

w

Chernisorbed

I

ctata

Reactant@) ---c

Transition state * Product(s)

Reaction coordinate

FIG.1 . Schematic energy profiles for the gas-phase versus surface reaction [R, reactant(s); TS, transition state; P, product(s)]. The relevant activation barriers AE; and AET usually are dramatically different because of different chemisorption energies of the reactant QR and transition state QTs.

V). As illustrated in Fig. 1, without knowledge of Q, even the most accurate gas-phase reaction energy profiles are useless for projecting activation barriers AE* for adsorbate transformations. While waiting for the advent of efficient quantum mechanical techniques, common sense requires one to look for working alternatives. If not microscopic modeling, let it be phenomenological. Because we are interested in reaction energetics, the phenomenological model should ultimately be based on thermodynamics. To have a chance to succeed, such a model should have a rigorous mathematical framework and make use of well-defined parameters, preferably only observable ones. And in order to catch the experimentalist’s eye, theoretical constructs should be simple enough, ideally at a “back-of-the-envelope’’ level. Below we discuss such a phenomenological construct, namely, the bond-order conservation-Morse potential (BOC-MP) model (16-1 9 ) .This analytic model proved to be efficient for treating energetics of atomic and diatomic adsorbates on transition metal surfaces, particularly, the heat of chemisorption Q and activation barriers AE* for dissociation and recombination. Recently, the BOC-MP model has been extended to treat energetics of polyatomic adsorbates as well, which made it possible to analyze mechanisms of catalytic heterogeneous reactions of practical importance. The goal of this article is to survey major applications and implications of BOC-MP modeling to chemisorption and heterogeneous catalysis. We shall begin by highlighting the BOC-MP analytic formalism and interrela-

BOND-ORDER CONSERVATION

103

tions. Next, we shall illustrate the efficiency of the BOC-MP calculations of Q and AE* by comparing our projections with experiment. Then, based on these calculations, we shall discuss the mechanisms of complex catalytic heterogeneous reactions involving several competing pathways. The examples include hydrogenation of CO on the platinum-group metals (19a,c), as well as transformations of C, hydrocarbons (196,e) and decomposition of HCOOH on transition metal surfaces (194. We shall compare the BOC-MP projections with experiment and other theoretical results. Finally, we shall summarize strengths and weaknesses of the BOC-MP modeling and comment on its implications for understanding metal surface reactivity.

11.

BOC-MP Formalism: A Brief Reminder

The BOC-MP formalism has been extensively published (f6-19) and summarized in two review articles (Z6-13, thus here a brief reminder will suffice. Only the latest developments leading to a more accurate treatment of the activation dissociation barriers (189 and the heats of chemisorption of radicals (194 will be reviewed in detail.

A. BASICMODELASSUMPTIONS We consider chemisorption of an adsorbate X, atomic A or molecular AB, on transition metal M surfaces. High coordination of the M atoms [up to 12 nearest neighbors in the close-packed lattices, e.g., face-centered cubic (fcc) and hexagonal close packing (hcp)] makes the M-M and M-A forces quasispherical when the total energy E depends mainly on the bond distance r . Of the supportive arguments, one can cite the following: (1) Both transition and simple s' (1A and 1B group) metals have the same densely packed [hcp, fcc, or body-centered cubic (bcc)] crystal lattices (20).(2) The d-orbital anisotropy is averaged out, and for many purposes, the d band may be effectively represented by a degenerate s-type band (21). (3) For various cases of transition and simple metallic binding, including chemisorption, there exists an apparently universal relation between E and r (21~2,22). To deal with quasispherical interactions, one should define a model potential relating E to r or some convenient function of r , for example, the two-center M-A bond order x, x = exp[ - ( r - r,)/a]

(1)

104

EVGENY SHUSTOROVICH

which is an exponential function of the M-A distance r , and r, and a are constants. In order to reproduce the total energy E minima, the model potential must include both attractive and repulsive forces. For two-center interactions, the simplest general potential of this kind is the Morse potential (2.9, including only linear and quadratic terms in x , namely, E(x) = - Q(x>= - Q o ( 2 ~- x2)

(2)

where Q, is the M-A equilibrium bond energy. The total energy E(x) [Eq. ( 2 ) ] has only one minimum at the equilibrium distance r,, when the bond order x = 1, by definition [Eq. ( l ) ] . If we want to describe many-center Mn-A interactions (n is the coordination number) in the Morse-potential fashion, the simplest scheme is pairwise additivity of all the two-center M-A contributions to Q and x , namely.

and n

x(n) =

2 xi

(4)

i= 1

Here the simplest way to proceed is to keep the Morse parameters (Q,, r,, and a ) the same for both the isolated M-A bond and each additive M-A contribution within Ma-A so that these contributions would differ by their bond orders xi (i = I , 2 , . . . , n ) only. Experimentally, Q(n> increases less than linearly with n

Q, < Q(n) < nQo9

n

>1

(5)

which requires imposing some constraints on the allowed values of xi < 1 (if xi = 1 , then Q(n> = nQa). The simplest assumption is that the total bond order x [Eq. (4)] does not change with n ; that is, x is conserved and normalized to unity n

x = ~ x i = x , = l i= 1

for any n L 1 [cf. Eq. ( l ) ] . Finally, we have to choose the values of n in Mn-A. The simplest assumption is to limit n to nearest-neighbor metal atoms. For instance, for A/fcc(100), the maximum n = 4 can be reached in the hollow site, but n = 2 and 1 in the bridge and on-top sites, respectively. In summary, our model assumptions are as follows:

BOND-ORDER CONSERVATION

105

1. Each two-center M-A interaction is described by the Morse potential [Eqs. (1) and ( 2 ) ] . 2 . For a given M,-A, n two-center M-A interactions are additive. 3. Along a migration path up to dissociation, the total M,-X bond order is conserved and normalized to unity. (The analytic form of BOC depends on X, namely, Eq. (6) for adatoms X = A or Eq. (8) for admolecules X = AB.) 4. For a given M,-A, n is limited to nearest neighbors. Assumptions 1-4 are the rules of the game. They are the simplest logical possibilities. The rest is straightforward algebra. The most definitive results have been obtained for chemisorption on flat symmetric surfaces with a regular unit mesh M,, say, an equilateral triangle M, for fcc(ll1) or hcp(001), a square M, for fcc(lOO), etc. Consider the major findings concerning surface energetics. B. DIATOMIC ADSORBATES In the following BOC-MP interrelations, the basic energetic parameter is the Morse constant Q, [cf. Eq. ( 2 ) ] ,which corresponds to the maximum M-A two-center bond energy QOA. The value of QOA, although not directly observable, can be easily scaled from the experimental heat of atomic chemisorption QA (atomic binding energy) identified with the M,-A bond energy Q(n), namely, QA =

Q(n>= Q o A ( ~ - I/n)

(7)

which is the exact result of bond energy variation under the BOC condition of Eq. (6). Let us clarify that, because Q(n)monotonically increases with n [cf. Eq. (7)], the value of Q(n) reaches the absolute maximum in the hollow n-fold site, which makes this highest coordination site always preferred. 1. Heat of Chemisorption

For molecular AB chemisorption, BOC for M,-AB reads as n

(XAi i= 1

+ XB;) + X A B

=

I

which implies that the heat of molecular M,-AB chemisorption QAB relates to heats of chemisorption of the coordinated atoms QA, QB, and the A-B gas-phase dissociation energy DAB. In order to obtain the value of QAB, we

106

EVGENY SHUSTOROVICH

should maximize the total M,-AB bond energy under the BOC condition of Eq. (8a). Unlike Eq. (6) for the atomic case (M,-A), however, Eq. (8a) introduces too many variables to make the general analytic solution for QABpossible. For realistic calculations, the number of the variable bond orders xi should be reduced. This can be done either by neglecting some metal-adsorbate interactions (here, the specific M,-AB geometry provides the guidance) or by using the effective group terms, say the effective I xA,.In the last case, Eq. (8a) can be rewritten atomic bond order xA = as

xy=

XA

+ x g + XAB = 1

G3b)

The practical difference of using Eq. (8a) versus Eq. (8b) is that these BOC conditions introduce different Morse constants to describe the effective interaction of an atom A (or B) in a molecule AB with a metal surface, namely, the two-center A-M energy QoAfor Eq. (8a) and the polycenter A-M, energy Q A for Eq. (8b). Because the resulting approximate formulas for Q A B will be different, we should also make a decision about which molecules AB can be best described by each approximation. We begin with the BOC condition of Eq. (8a). If AB is monocoordinated (q')to a surface with the A end down, one can neglect the M-B interactions. Then for the on-top coordination M-A-B (qlpl, where the subscript on p represents the number of the metal atoms involved), the variational procedure gives + DAB)

QAB 5 Q?~A/(QoA

(9)

and for the n-fold M,-AB coordination (q'p,),

It should be stressed that Eqs. (9) and (10a) contain some neglected negative terms, so that QAB,ncan change nonmonotonically with n . In general, admolecules with larger values of D A B prefer lower coordination sites, as will be discussed in detail in Section II1,A. If AB is coordinated parallel to a surface, via both A and B (dicoordination $), the bridge mode A-B

/ M

\

M

appears to be the general prototype (q2,u2),with the bonding energy

BOND-ORDER CONSERVATION

107

where a =

QiA(QO~ + ~QoB)/(QoA + QoB)~

and

For a homonuclear.A,, Eqs. (12a) and (12b) reduce to a = b = a, =

(3/4)QOA

and Eq. (1 1) reduces to

Because, within the BOC-MP approximations, the bridge $p2 and hollow > 2) modes are indistinguishable, we shall henceforth use the notation r12p2only. In the extreme case Of DAB = 0, Eqs. ( 1 1 ) and (14) become, respectively,

q2p, ( n

QAB

=u

+b

(15)

and QA,

=

2a0 = (3/2)Q0,4

(16)

Obviously, if A and B are free atoms, this q2p2geometry, corresponding to the on-top atomic coordination, will be energetically unfavorable, because adatoms always prefer higher coordination sites [cf. Eq. (7)]. However, if atoms A and B are linked via some atom X, a chelate structure A-X-B

/ M

\ M

can legitimately be formed. Equations (15) and (16) can be applied to chelate chemisorption if the parameters a and b take into account formation of the A -X and B - X bonds with the respective energies DAX and DBX. Thus, the problem is reduced to the known case of mono(qIpl) coordinations M - A - X and M - B - X, when the modified Morse constants are [cf. Eq. (9)]

and

108

EVGENY SHUSTOROVICH

leading to the modified parameters [cf. Eqs. (12a,b)]

uf2(af+ 26') (a' b')* br2(b'+ 2a') bx = (a' b')2 a, =

+

+

and the chelate heat of chemisorption QAB(X)

= ax + 6,

(19)

or QA,(x)

= 2ax

(204

Now let us turn to the BOC condition of Eq. (8b). Although Eq. (8b) does not explicitly depend on n , it assumes the best possible coordination within the M,-AB unit mesh, which reflects in the use of the experimental heats of atomic chemisorption, QA and Q B , as the Morse constants in energy calculations. For the monocoordination ~ ' pin, M,-AB, the variational procedure now leads to an expression QAB

= Q ~ / ( Q A + DAB)

( 1Ob)

which is an analog of Eqs. (9) and (10a) employing the parameter QA instead of QoA.Other analogs of the above formulas for QAB can be obtained in a similar way. In particular, Eq. (20a) for the symmetric chelated chemisorption (q2p,)transforms into Q ~ , c x )= 2ax

= (312)a' =

(3/2)Q',/(Q, + DAJ

(20b)

Because QA is almost twice as large as QOA [cf. Eq. (7)], we shall refer to the relevant expressions for QAB, say Eqs. (10a) and (20a) versus Eqs. (lob) and (20b), as the weak and strong chemisorption, respectively. Of course, along with the weak and strong M,-AB bonding, one can imagine the intermediate one, which may be described by interpolating between the two extremes. In particular, for the monocoordination ($p,) M,-AB, we shall simply average Eqs. (10a) and (lob) as

109

BOND-ORDER CONSERVATION

We shall discuss the assignment of AB molecules into each category in Section 1II.A. 2. Activation Barriers for Dissociation and Recombination The use of Eq. (8b) is also intrinsic for the BOC-MP treatment of AB dissociation. Consider the activation barrier AEXB,, for dissociation B, when AB approaches a surface from the gas phase. The AB, + A, traditional one-dimensional Lennard-Jones (LJ) potential diagram refers AEXB,, to the intersection point of the molecular AB and atomic A + B curves, as shown in Fig. 2a.Thus, the transition state (TS) where the configuration switch occurs corresponds to x x i = 0, reducing Eq. (8b) to

+

1

x A f XB =

(8C)

The relevant variational procedure gives the BOC activation barrier AEXk; corresponding to the LJ picture (18b)

AEXik,; = D A B- (QA +

+ Q A Q B ~ Q A + QB)

QB)

For homonuclear dissociation A2,g- A, AEzy =D

-

(214

+ A,, Eq. (21a) simply becomes

(3/2)Q~

(22a)

In reality, however, there is no such thing as the LJ intersection, because the AB and A + B energy profiles are multidimensional hypersurfaces (24).The minimally adequate representation is two-dimensional, as shown in Fig. 2b. It is immediately clear that in the transition state, should have a nonzero value and, therefore, A E i k i [Eq. (21a)l should typically be an overestimation. How can the value of AEXk; be corrected? For xi; = c > 0, from Eq. (8b), we have XA

+xs= 1 - c

(84

and the A-B dissociation activation barrier becomes (184

AEZB,, = (1 - C)’DAB- (QA +

QB)

+ (1 + C)’

QA

+

(21b) QB

or, for homonuclear A-A dissociation,

Because for all diatomic A-B molecules, DAB B QAQB/(QA + QB), by comparing Eq. (21a) with Eq. (21b), we find AEXB,, < AEWk;, in agreement

+M

QA,

A i;-iXB

R

b

FIG. 2. Chemisorption and dissociation of a diatomic AB. (a) The traditional Lennard-Jones one-dimensional potential diagram E versus R , where R is an ill-defined reaction coordinate, say the AB-surface distance. (b) The conventional two-dimensional potential diagram E versus R(x, y ) . The reaction coordinates are the A - B distance (x) and the AB-surface distance (y). The energy minima correspond to the (nondissociated) molecular chemisorbed state DAB + QABand atomic (dissociated) chemisorbed state QA + QB, the maximum to the transition state (TS) with some finite A -B bond length. (c) The multidimensional BOC potential diagram, similar to (b), but the reaction coordinate is the A - B bond order XAB. The M, - AB bond order is conserved to unity (xA + xAB+ xg = I ] up to the transition state where 1 > xi; = c > 0 and 6 = 1/2(AE2hjg + QAB). See text for notations and explanations.

BOND-ORDER CONSERVATION

111

DAB

FIG. 2. (continued)

with our expectations. Still, we cannot calculate AEZB,, until we know the values of c , which, in principle, may vary from one system AB/M to another. So, we should find some other way around. To accurately calculate AE);B,g, we must know details of the multidimensional AB, and A, + B, energy profiles, which we do not know anyway. Obviously, the dissociation of AB must occur when the total AB energy is intermediate between that of the AB chemisorbed state and the LJ dissociation point. The simplest interpolation puts the dissociation point just in the middle of this energy interval, QAB + AE);&i (cf. Fig. 2a). Then, we have (18i)

Let us stress that Eq. (21c) explictly shows how the molecular chemisorption energy QAB can affect the AB dissociation, the effect totally missing in the one-dimensional LJ picture described by Eq. (21a). Also, if we know AEZB,,, from experiment or from Eq. (21~1,by using Eq. (21b) we can determine c and, thus, get an idea of how much the gas-phase A - B bond should be stretched to dissociate on a metal surface. The multidimensional (non-LJ) BOC picture is summarized in Fig. 2c.

112

EVGENY SHUSTOROVICH

One can add that the dissociation barrier AEXB,sfrom a chemisorbed state will be larger than AEXB,, (one- or multidimensional) just by the amount of the molecular heat of chemisorption Q A B ,

AEXB.~ = AEXB,~ +

QAB

(234

Combining with Eq. (21c), we get

AEXB,, = 1/2(AE

+ QAB)

For the reverse reaction of recombination of chemisorbed A and B, the LJ activation barrier is particularly simple [cf. Eq. (21a)l:

Clearly, if Q A is much larger than Q B , AE2:kJ will be close to Qe, the heat of chemisorption of the weaker bound partner. If Q A = Q B , we simply have AEiLy = (1/2)&. Obviously, the activation recombination barrier AEz$ cannot be smaller than AH = AHAB - A H A + B , the difference between the enthalpies of the reactant AB, (-AHA* = D A B + Q A B ) and the products A, and B, (-AffA+B = Q A + QB). Thus, the BOC barrier [Eq. (24)] is only the necessary (minimal energy) condition for recombination, which may be sufficient if

but not sufficient if QAQB/(QA + QB)< AH. In such cases, the recombination barrier may be assumed to be the enthalpy difference, AE;:kJ = AH. In general, for the non-LJ dissociation described by Eqs. (21b) and (21c), the recombination barriers AE2-B can be calculated only from the relevant thermodynamic relations. Remember that for the recombination of chemisorbed A, and B, to chemisorbed AB, or gas-phase AB,, the activation barriers AE;-Bss and AEZ-B,, may be the same or different, depending on the sign of the gas-phase dissociation barrier AEzB,,, namely if AEXB,, > 0

(26a)

AEg-B,, < AEz-B,g if AEXe,, < 0 More specifically, Eq. (26a) can be rewritten as

(274

AEX-B?, = AEi-B,, and

113

BOND-ORDER CONSERVATION

c.

POLYATOMIC

ADSORBATES

In the zero-coverage extreme, the BOC-MP analytic results are exact for atomic A adsorbates and are well defined for diatomic AB adsorbates. Formally, Eqs. (9)-(14) can be applied to polyatomic admolecules provided they are treated as quasidiatomics-in other words, if we define the effective parameters QA, QB, and DAB for the relevant molecular fragments A and B. Actually, we have already done such partitioning while extending Eqs. (1 1)-(14) to Eqs. (17)-(20). Now let us discuss the partitioning problem in a more systematic way. 1. Partitioning of Two-Center Bond Energies

Consider first those applications of Eqs. (9)-(14) to polyatomic admolecules AB when bond energy partitioning does not affect the results. The simplest case is monocoordination q'p, M,-AB via an atom A , where B is a set of atoms directly bound to A, for example, as in HCO (Ia), NH, (II), or CH, (IIIa-c):

Ia

I1

IIIa

Illb

IIlc

Here, the values of QAB can be calculated directly from Eqs. (lOa)-(lOc), where QOA and QA refer, as earlier, to an atom A, but DAB means the total AB bond energy, because this is the energy of all bonds formed by A. Similarly, Eq. (14) for homonuclear q 2 p 2dicoordination A-A

/ M

\ M

can be used for chemisorption of symmetric molecules, such as H C s C H or H,C=CH, coordinated via atoms C with the C-C vector parallel to a surface. In this cases, QOAin Eq. (14) stands for the atomic carbon value of Qoc = Qc/(2 - l / n ) [cf. Eq. (7)] and DA, for the total energy of all bonds formed by each C atom (which can be easily found from a simple

I14

EVGENY SHUSTOROVICH

thermodynamic cycle; see Section III,A, 1). Another example is coordination of a linear symmetric triatomic molecule CO, as M

/o-c-O

M ‘

Now in Eq. (14) Q O A refers to coordinated atoms 0, but D A , means the total CO, bond energy (the energy of all bonds formed by both atoms 0 or, equivalently, by the atom C ) . Let us return to the q ’ p nmonocoordination M,-AB via an atom A, but now B is a molecular fragment such as OH (IV) or CH, (V or VI):

QI

I

C

I C

1v

V

I

I

VI

Ib

Then the values of Q A B in Eqs. (10a)-( t Oc) will depend on the effective twocenter A-B (C-0, C-C) bond energies, which makes Q A B sensitive to the energy partitioning. This sensitivity will become even higher for q2p, dicoordination described by Eq. ( I I). For example, if we apply Eq. ( I 1 ) to formyl HCO (Ib), D A B will have the meaning of the C=O bond energy and the parameters a and b will relate to an atom 0 and a “quasiatom” CH, respectively. Because the parameter b includes the C H bond energy [cf. Eqs. (17) and (18)], both DABand b will now depend on the way of partitioning of the total HCO bond energy into the H-C and C=O contributions. The main problem here is not computational, however, but conceptual. Remember that partitioning of the total bond energy in a polyatomic molecule into fragments and, ultimately, into two-center contributions has no physical significance. This partitioning makes sense only if it serves some practical purpose. In our case, the purpose is to fit best Eqs. (10)-(27). So, we are free to develop any working partitioning scheme, but for the sake of generality, the same scheme should be used for all related molecules. Earlier, we suggested such a scheme ( ] a h ) , which involves the averaging of chemical structures representing limits (upper and lower) of two-center bond energies. Qualitatively, the results were usually reasonable, but quantitatively, the averaging procedure was loaded with uncertainties, reducing its heuristic value. Most recently, we have found a way to eliminate the bond energy averaging altogether in calculating Q A B and AEZBfor polyatomic molecules, and henceforth we shall stay with this straightforward approach. Before we proceed, some discussion is necessary.

115

BOND-ORDER CONSERVATION

2. Purtitioning, Dissociation, and Disproporfionafion Consider the simplest case of a triatomic molecule A-B-C. Its total bond energy DAB, is, by definition, the heat of dissociation into the constituent atoms DAIK

+B +C

ABC-A

We want to partition DAB,into the sum of A-B such that

DABC

= DAB

+ DBc,

and B-C

DAB > 0,

bond energies,

DBC> 0

(28)

The simplest way to define DAB and DBC in Eq. (28) is to take DaB from dissociation I

DAB

ABC-A

+ BC

and Dkc from dissociation I

DBC

BC-B

+C

The other equivalent way is to define DBC as I1

DBC

ABC --+ AB

+C

and DAB as I1

DAB

AB-A+B

It is obvious, however, that

DaB f D i B ,

DLC f DEc

because the A-B bond in a diatomic molecule AB and that in the AB diatomic fragment of a triatomic molecule ABC are never the same. This is true even for symmetric molecules ABA such as H20, CO,, or NO,. For example, during dissociation, I27

+ 0257-c + 0 + 0

CO2-C0

the first C-0 bond cleavage takes D&, = 127 kcal/mol, but the second one is DEo = 257 kcal/mol (25), although both C-0 bonds in a linear C 0 2 are equivalent, O=C=O, with the average C-0 bond energy

D,, = 1/2(Di0

+ D&)

=

1/2(127 + 257) = 192 kcal/mol

For nonsymmetric triatomic A-B-C and all polyatomic molecules, the

116

EVGENY SHUSTOROVICH

bond energy averaging becomes totally arbitrary. Little wonder, even the elaboratored averaging techniques proved to be only partially efficient (18), and we do not want to employ them anymore. For the sake of uniformity, it would be a great advantage to use the same value of DAB in calculations of both AEXB and QAB.Moreover, we want to use the same formalism, Eqs. (10)-(27), to describe both diatomic and polyatomic molecules. To fulfill these conditions, the simplest and most uniform way to proceed is to define the A-B bond energy DAB as the difference between the total gas-phase energies of AB and the dissociated fragments A and B. The only limitation of this definition is that it cannot be applied to exothermic dissociation of polyatomic molecules, where DABbecomes negative. Indeed, in Eq. (28), if ABC -+ A + BC is exothermic, it means DAB, < DB, and, therefore, DAB < 0, which makes no physical sense. For example, hydroxycarbyne COH, which is an isomer of formyl HCO, has the total gas-phase bond energy DCoH= 233 kcal/mol (26a). Thus, dissociation COH + CO + H is exothermic, leading to the negative and therefore meaningless value of DOH = 233 - 257 = -24 kcalimol. Another example is dissociation of a formate radical HCOO to H and CO,, which appears to be practically thermoneutral. [This follows from the combined results of the ab initio spin-density configuration interaction (SDCI) calculations (266) and the experimental (photoionization mass spectrometry) studies (26c) of the gas-phase potential surface of the system . COOH was measured to lie below H + C O Z e H C O O ~ C O O HNamely, H + CO, by 10 kcal/mol (26~1,but HCOO was calculated to lie above COOH by 10 kcal/mol(26b).l Thus, DHcoo = DcoZ = 384 kcal/mol(25), and the bond energy partitioning DHco0= DHC + Dcozgives the nonphysical value of DCH= 0. With the exception of gas-phase exothermic dissociations, our choice of DAB in polyatomic molecules is uniquely defined. Because the overwhelming majority of dissociation processes are endothermic (and this is exactly the reason why catalysts are often required to make these processes happen), they are within the scope of our modeling. The latter makes it possible to treat a variety of disproportionation reactions A

+

BCG[A,..B...C]*AB

+C

(2%

that proceed without major geometric reorganization of the fragments via the transition state [A..*B..C], where A may be either an atomic or molecular adsorbate and BC a diatomic or polyatomic admolecule. Arranging in Eq. (29) the disproportionating species such that D = DA

+ DBc

-

DAB - Dc > 0

(2W

117

BOND-ORDER CONSERVATION

we can treat A + BC as a quasimolecule (with the total gas-phase bond energy D A + B C = D A + DBcand the total heat of chemisorption Q A + B C = Q A + Q B C ) that dissociates into twofragments, AB and C (with the relevant energies D A B , Q A B , Dc, and Q,). Then, the activation barrier AE? for the forward reaction in Eq. (29) may be calculated as the dissociation barrier AETAB,,, which, from the gas phase,will be [cf. Eq. (21c)l AE$%, = A E T A B p , = 1/2(D +

QAB

+ Qc

- QA

- QBC - QAB

- Qc) (29b)

or, from the chemisorbed state [cf. Eq. (23b)], AE$., = AETAB,,,, = 1 / 2 ( D

+

QAB

+ Qc

+

QA

+

QBC - QAB -

Qc)

(29c) For the reversed reaction in Eq. (29), the activation barrier AET may be calculated as the relevant recombination barrier AE2B-C from Eqs. (26) and (27). 111.

Basic BOC-MP Applications

Energy profiles of surface reactions reflect changes in the total energy of chemisorbed species under their interactions and transformations. Because the gas-phase bond energies D for molecules of interest are usually known or can be reasonably calculated by the available quantum chemical techniques (6), one needs only to add the heat of chemisorption Q to obtain the total bond energy in the chemisorbed state, D + Q (cf. Fig. I). Thus, the critical point in calculations of surface energy profiles is the accuracy of calculations of Q. Within the BOC-MP model, this critical importance of Q is further stressed by the fact that various activation barriers AE* (for migration, dissociation, and recombination) explicitly relate to Q (16, 17). So, we shall begin by testing the accuracy of the BOC-MP estimates of Q for a variety of diatomic and polyatomic molecules. Then, we shall see how accurate the BOC-MP projections of AE* are for various surface reactions of dissociation and recombination. A.

HEATSOF CHEMISORPTION

Conceptually, the most important model conclusion is that for a diatomic AB, the heat of molecular chemisorption Q A B relates to both the heats of chemisorption of coordinated atoms Q A and Q B and the A-B bond energy

118

EVGENY SHUSTOROVICH

TABLE I Experimental Heats of Atomic Chemisorption QA on Some Metal Surfaces" Atom

Metal surface

H

Ref.

0

W(110) Fe( 110) Ru(001) Rh(l11) Ir(lI1) Ni(l1l) Pd(l1 I ) Pt( 1 I I ) Cu(ll1) Ag(ll1) Au(ll0)

68 64 67 61 58 63 62 61 56 552 558

30 30 31 32 33 30 30 34 30 35u 35b

-125 ( I 18)' 100 102 93 1 I5 87 85 103y

80 575

C

Ref.

N

Ref.

36 37 38 39 30 40 30 30

155

30 30

-

-

(200)h (200)h

I27 135 I30 I16

30 30 30 42

171d (160)' (150)' ( 120)' (<120)h -

41

30 356

140 -

-

-

-

In kilocalories/mole. The assumed values (see text). For polycrystalline Fe. ,I Ref. 43. For a polycrystalline surface.

DAB.Within the BOC-MP framework, the values of Q A ( Q B ) and DAB are not calculated, but are taken from experiment, which forms the thermodynamic basis of the BOC-MP model. So before considering the model projections on Q A B , let us summarize the necessary background concerning QA (QB). Table I lists the experimental values of Q A for major adatoms such as H, 0, N, and C on some close-packed metal surfaces (30-43). Typically, the heat of atomic chemisorption QAdecreases while going from the left to right along a transition series and from the top to bottom of a column. This decrease A Q A is the least pronounced for monovalent H when, within the series Pt-Ni-W, A& does not exceed 7 kcal/mol [QH = 61, 63, and 68 kcal/mol for Pt( I1 l ) , Ni( 1 1 l), and W(l lo), respectively]. For divalent 0 and trivalent N , however, the changes in Q A from Pt to Ni to W become very large, up to A Q A = 40 kcal/mol; i.e., Q, = 85-125 kcal/mol and QN = 115-155 kcal/mol. For tetravalent C , the experimental measurements have been reported only for Ni(ll1) and Ni(100), giving Qc = 171 kcal/ mol(43). So, for other metal surfaces we are to use extrapolated estimates of Q,. For C, we assume a somewhat larger spread in QA compared with 0 and N; that is, AQ, = 50 kcal/mol, from Q, = 150 kcal/mol for Pt( 1 11)

BOND-ORDER CONSERVATION

119

to 200 kcal/mol for W( 1 lo), with the experimental value of Q, = 171 kcal/ mol for Ni(lI1) lying in between. Before we proceed to calculations of Q A B , we should establish criteria of assigning AB molecules to weak, strong, and intermediate M,-AB chemisorption [described, for example, by Eqs. (lOa), (lob), and ( ~ O C ) , respectively]. Here we shall rely on the basic quantum chemical rules. If the A-B bond significantly exhausts the valence capabilities of the A and B atoms, the M,-AB bonding can be expected to be relatively weak. The best candidates appear to be AB molecules with a closed electronic shell, such as H,, N,, CO, NH,, or H,O, or with unpaired electrons occupying the substantially delocalized molecular orbitals, such as NO or 0,. An important indicator of the weak M,-AB bonding is relative insensitivity of Q A B to the coordination site (on-top versus twofold bridge versus n-fold hollow). On the other hand, AB molecular radicals, in which unpaired electrons mainly retain their atomic character, such as CH, CH2 NH, OH, or OCH,, are deemed to chemisorb much more strongly, resembling the patterns of atomic chemisorption, including the distinct preference for the n-fold hollow coordination. The intermediate M,-AB bond strength may be expected for monovalent radicals AB, where A is a tri- or tetravalent atom, say N or C in NH,, CH,, or HCO. 1. Weak Bonding Now we are ready to calculate Q A B for a variety of molecules AB chemisorbed in different coordination modes. We begin with weakly bound diatomic and polyatomic molecules for which the values of Q A B do not depend on bond energy partitioning, and, therefore, the validity of Eqs. (1Oa) and (1 I ) can be tested directly. For the q'p, monocoordination M,-AB via A, Eq. (10a) formally projects that Q A B , , will increase as n increases. However, because Eqs. (9) and (10a) were simplified by neglecting the negative terms, also increasing in absolute value with n (18d), we expect the effective compensation resulting in the weak dependence of Q A B on n , the dependence being the weaker the larger is the value of D A B . A good example is CO with the huge dissociation energy D,, = 257 kcal/mol. Under chemisorption of CO on various metal surfaces, the energy differences among the on-top, bridge, and hollow sites are so small that at higher coverages and temperatures, some or all of these sites can coexist (44-46). For example, for CO/Pt( 1 1 l), Qco decreases in the order on-top > bridge > hollow, but within A Q < 1 kcal/mol (47). The similar difference of A 5 1 kcal/mol manifesting in the coverage and temperature

120

EVGENY SHUSTOROVICH

dependence of the preferred sites was found for CO/Rh(l00) (48), CO/ Ni(100) (49), CO/Ni(llO) (50), CO/Ni(lll) (51a,b),and CO/Cu(lll) (52). For CO/Ni(l 1I), the Qco order hollow > bridge > on-top (51a,b)is reversed compared with that for CO/Pt(11l), which demonstrates the lack of correlation between Qco and n. Because Eq. (10a) is inadequate for judging how QAB,n depends on n , the critical question is which value of n should be chosen in Eq. (10a) to best reproduce the experimental values of QAB.Because eq. (1Oa) assumes the bond-order conservation for the coordinated atom A, it infers that, other conditions being equal, the stronger (multiply) bound AB molecules will prefer the lower coordination but the weaker (singly) bound molecules will prefer the higher coordination. Thus, for the triply bound CEO with D,, = 257 kcallmol, the value of Q,, should be calculated for n = 1, reducing Eq. (10a) to Eq. (9). For CO/Ni(lll), where Q, = 171 kcal/ mol, Eq. (9) gives Qco = 29 kcal/mol, in excellent agreement with the experimental value of 27 kcal/mol (30). Unfortunately, the lack of the experimental values of Q, for all metals but Ni prevents further comparisons of Qco for other metal surfaces. Within the assumed range of Q, = 150-200 kcal/mol, however, Eq. (9) projects the values of Qco to be within 25-38 kcal/mol, which nicely reproduces the observed range of Q,, = 25-40 kcallmol for transition metal surfaces (30, 46). Consider now chemisorption of molecules coordinated via N, such as NO or NH,. With DNH3= 279 kcalimol (25), NH, is the classically triply bonded molecule N=H,. On the other hand, with D,, = 151 kcal/mol (25),NO is basically doubly bonded as N=O. Thus, in Eq. (10a) it seems reasonable to use n = 1 for NH, but n = 2 for NO. Because the experimental values of the heat of atomic nitrogen chemisorption (unlike carbon!) are known for many surfaces, we can calculate QNOand QNH,for a variety of surfaces. The results are compiled in Tables I1 and 111, where they are compared with experiment (30, 53-62). The agreement is excellent, the error typically not exceeding 10-15%. For example, for NO on Pt(II1) and Pd(l1 I ) , we find QNo= 26 and 32 kcal/mol, whereas the experimental values are 27 and 31 kcal/mol, respectively (30). Similarly, for NH, on Pt(ll1) and Ni(lll), we calculate QNH, = 13 and 18 kcallmol, to be compared with the experimental values of 12 (56) and 20 (57) kcal/mol, respectively. Now let us say a few words about chemisorption of H 2 0 and methanol CH,OH coordinated via 0. For these saturated molecules, one can assume the on-top (q'p,) coordination. With D,,, = 220 kcal/mol and for the range of Q, = 80-100 kcal/mol for close-packed surfaces of such metals as Ag, Pt, Pd, Rh, and Ru, Eq. (10a) then gives QH,, = 9-13 kcal/mol. Measurements of QH2, are obscured by the concomitant formation of

121

BOND-ORDER CONSERVATION

TABLE I1 Heats of Molecular Chemisorption Qas: Diatomic Molecules" QAB

Experimental values of System CO/Ni( I 1 1) COIFe(ll0) CO/M NOIPt(ll1) NO/Pd(lll) O,/Pt(lll) O,/Ag(l 10) N,/Pt(Ill) N2/Ir(l10) N,/Ni(100) N2/Ni(l10) N,/Fe(lll)

Coord. type q'pl q'p, qlpI q'p2 q1p2 q2p2 q2p2 qzp[ q2pf q2pi

q2A q2p;

QAh

QBb

171 200 150-200

(115) (125) (85-125) (85) (87) 85 80 116 127

116

130 85 80 116 127 135 135 140

135

135

140

Calc.

DABL M-AB 257 257 257 151 151 119 119

226 226 226 226 226

29d 38 25-38d 26d 32d 11' 10' 11' 13' 14' 14' 15'

(M-BA)

Exp.

Ref.

(15)d

27 36 26-40 27 31 9 10 9 11

30 53 30, 46 30 30 54 55 66a 66a 666 66c 66d

(17) (8-17)d (15)d

-

-

-

11 11

8

' See text for explanations and notations. All energies in kilocaloriedmole. From Table I and Ref. 30. From Refs. 25 and 29. Eq. (10a). Eq. (14). Assuming the same energy as in the experimentally observed coordination q'p,,. Coexists with the coordination qlpn(66e).

hydrogen bonds between H,O molecules. With these corrections, the experimental estimates for smooth surfaces of the late transition metals and silver are QH,o = 10-12 kcal/mol (59), in full agreement with the BOC-MP projections. In CH,OH, the total bond energy formed by 0 is D,,H,(cH,, D C H 3 0 H - D C H 3 = 487 - 293 = 194 kcal/mol. For the same range of Q,, = 85-100 kcal/mol, Eq. (IOa) now gives QcH,OH = 11-14 kcalimol. For comparison, the experimental values of QcH,oH are 1 1 kcal/mol for Pt( 1 1 I) (60) and 12 kcal/mol for Rh( I 1 1 ) ( 6 1 ) . Again, the agreement with experiment is excellent. Coming back to diatomic molecules AB, one can wonder why CO is always coordinated via C and NO via N but not via 0. From Eq. (IOa), it is obvious that in the upright M,-AB geometry AB will be coordinated to M, through the atom whose heat of chemisorption is larger, namely, M,-A-B if QA > Q,. Because QA increases in the order 0 < N < C (cf. Table I), the preferred coordination is that via C in CO and via N in NO.

122

EVGENY SHUSTOROVICH

TABLE 111 Heuts of Molec,ulur Chemisorption QAs :Polvcitotnic Molecules"

System NH,/Ni(lll) NH,/Pt( 11 1) H,O/Pt( 11 1) H,O/M CHIOH/Ni( I 1 I ) CHIOHIRh(l1 I ) CHjOH/Pt(l I I ) C,H,IM C,HZ/M CO,/M R,CO/Pt( I 1 1) R,CO/Ru(OO 1) R~COIRU(OO I) HCOOHIAu(l10) "

Coord. t Y Pe

Experimental values of QBh

QAh

135 1 I6 85 80- 100 1 I5

DAB

Calc.

Exp.

279' 279' 220' 220'

I 8d 13d I Id 9-13* 18d 14d I Id 12- 15'' 14-18" 4-6" 1I" 15" 16- 18'' 12L

20 12 12 10-12 14' 12 11 11-13

194'

I00 85 150-170

150- 170 85-1 15 x5

150-170 150-170

194' 194' 3.50': 311#

85-115

384'

67' 75

178' 178" 178' (222, 207)'

-

100 100

75

QAB

-

5-6 12 16 16 13

Ref.

57 56 58 59 62 61 60 68- 70

-

72-74 60, 76 75 75 78

See text for explanations and notations. All energies in kilocalories/mole.

' From Table I. ' From Ref. 25.

Eq. (10a). Compare Table V111. For Ni( 100). Compare Table XI. Eq. (11). ' For a quasiatom B = (CH,),C with Qos metals (18f). From Eq. (30). Eq. (19). "

'

=

40 kcalimol, which is the average value for the Pt-group

J

As seen from Table 11, the coordination via oxygen, M-0-C or M-0-N, appears to be less favorable by 10-20 kcal/mol. Another question is when and how AB, say CO, can be coordinated parallel (tilted) to a surface. By comparing Eqs. (10a) and (1 I ) , one can get an idea of which coordination, q 1or q2,is more favorable. Given the approximate character of the equations, our primary goal is to discern periodic trends. For CO on late transition metals, the estimates show (18g) that the monocoordination ql (via C) is always preferred, in full agreement with experiment (44-46). Still, the ql versus q2energy differences A Q are not great; i.e., AQ < 5 kcal/mol (1813). Most important, the differences A Q decreases as one traverses from right to left along the transition metal series so that in the middle of the period, the ql and q2energies seem to

BOND-ORDER CONSERVATION

123

converge and may even be reversed. This model conclusion is consistent with the recent findings that even on smooth surfaces with no steric constraints, the tilted and normal coordinations of CO can coexist, the tilted one being slightly more favorable, on Cr(ll0) (63),Fe(100) (64),and Mo(100) (65). Quantitative comparisons of the 7' versus q2 coordination should be made with caution, however, because Eqs. (10a) and (1 1) have been obtained in somewhat different approximations. The most important difference is that Eq. (10a) for the upright M,-A-B coordination explicitly neglects the (repulsive) contribution from an atom B, but Eq. (1 I ) for the side-on /A-B\M M-

coordination explicitly treats both the A and B contributions. The neglect of the B contribution in Eq. (10a) is well justified if QA9 Q B , but becomes a poor approximation if QA = Q B (184. Thus, Eq. (10a) is inappropriate for homonuclear molecules A,, which prevents the meaningful 7'versus q2 comparisons for molecules such as N2 or 0,. Put simply, we cannot calculate QA2if A, chemisorbs normal to a surface, for example, as N, on many metal surfaces (66). At the same time, if A, is known to be chemisorbed parallel to a surface, as 0, on Pt( 1 11) (54) or Ag( 110) (55), we have a good equation [Eq. (14)] in place. As seen from Table 11, here we obtain Q,, = 10-11 kcal/mol, in excellent agreement with the experimental values of 9 kcal/mol for Pt( I 1 1) (54) or 10 kcal/mol for Ag(ll0) (55). Similar to CO, the normal (YIP,) and parallel (q2p,)coordinations of N, can be assumed to be close in energy. For example, for N,/Fe( 11 I ) , both the q' and q2 modes were found to coexist (66d,e). So, we will use Eq. (14) for calculating QN, regardless of the actual coordination of N,/M, as illustrated in Table I1 for Pt, Ir, Ni, and Fe surfaces. With the exception of N,/Fe, the agreement with experiment is good again. As noted in Section II,B, Eq. (14) can equally be applied to symmetric polyatomic molecules of the A, type, such as H,C-CH, (x = 2, 1) or O=C=O. For CH,CH,, in Eq. (14), D,, stands for the total energy of all bonds formed by each atom C. From DcH,cH, = 538 kcal/mol (25) and DcH, = 183 kcal/mol, we find D,, = 538 - 2(183) = 172 kcal/mol, which makes DA, = 183 + 172 = 355 kcal/mol. Similarly, for CHCH, where DC-CH = 392 kcal/mol (25) and DCH = 81 kcal/mol, we obtain Dcc = 392 - 2(81) = 230 kcal/mol and DA, = 81 + 230 = 31 1 kcal/mol. Table 111lists the values of QcH,cH, for a variety of metal surfaces. For the range of Qc = 150-170 kcal/mol covering the late transition metals (see Section III,A), we obtain for ethylene chemisorption QCHZCH2= 12-15 kcal/mol,

124

EVGENY SHUSTOROVICH

in agreement with the experimental range of 11-13 kcai/mol for Pt (67a), Pd (68),Ru (69), and Ni (70). Please note that the large change of AQc = 20 kcal/mol results in the small change of A Q C H 2 C H 2 = 3 kcal/mol (almost an order of magnitude smaller!), which makes the inaccuracy of the extrapolated values of Qc practically irrelevant. At the same time, one should not overestimate the accuracy of the experimental values of QC,H, obtained in various desorption studies and depending on ethylene coverage and the choice of the preexponentials in the Arrhenius-type kinetic analysis [cf. Eq. (32)]. For instance, for Pt(ll1) the range of Qc,H4 = 9-18 kcal/mol has been reported (67). Our calculations for acetylene give slightly higher values of QcHcH = 14-18 kcal/mol. Because of the lack of direct experimental data on Q C H C H , we shall postpone a discussion until Section IV,B. Finally, let us calculate Qco, for CO, linearly chemisorbed on transition metal surfaces (71).In Eq. (14), DA, now means the total CO, bondenergy, which is Dco2 = 384 kcal/mol. For the range of Qo = 85-115 kcallmol, corresponding to the late transition metals, we find Qco, = 4-6 kcal/mol, in excellent agreement with the measured values of 5-6 kcalimol for Rh (72a), Pt (72b,c), and Cu (73).Again, it is remarkable that the molecular heat of chemisorption Qco, is 15-20 times smaller than the relevant atomic heat Qo, and the large spread of AQo = 30 kcalimol translates into the meager change of AQcoz = 2 kcal/mol. The general conclusion from Eqs. (9)-( 14) is that the molecular heat of chemisorption Q A B rapidly decreases as the gas-phase dissociation (total bond) energy DABincreases. The values of Q A s are smaller than Q A ( Q B ) , typically by a factor of 5-10 but sometimes even 15-20. For this reason, the periodic changes in Q A B for molecules such as CO, NH,, NO, H20, CZH4, and C,H2 are expected to be small and potentially irregular, unlike the large and systematic variations in Q A observed for the relevant multiply bonded adatoms A. 2. Strong Bonding Table IV lists the values of Q A B for some strongly bound radicals. Although the direct experimental data are practically not available (radicals usually decompose while being desorbed), the accuracy of these calculations can be verified via the calculated activation barriers AE*, which we shall discuss later. Here we only mention that for OH and NH on Pt (l l l ), with Q, = 85 and QN = 116 kcal/mol (see Table I), Eq. (1Oc) gives QoH = 39 and Q N H = 68 kcalimol, respectively, in excellent agreement with the experimental ranges of 36-45 kcallmol for OH on Pt(l11) (153a,b) and 63-69 kcal/mol for NH on Pt wire (153c,d),found in laser-induced fluorescence studies by Lin and co-workers (153).

125

BOND-ORDER CONSERVATION

TABLE IV Heats of Molecular Chemisorption QAAB:Strongly Bound Radicals"

System

OHIM OCH,/M CH/M CH,/M CHy'M HCOIM NH/M NH2/M HCOO/M

Coord. type

D*Bh 102 90' 81 183 293 274 81 169 166'

QA'

80- 125 80- 125 150-200 150-200 150-200 150-200 115-155 115-155 80- 125

QAB

35-69d 38-73d 97- i 42d 68- 104d 38-62J 40-65 f 68- 1 02d 36-57f 39-81'

See text for explanations. All energies in kilocalories/mole. From Ref. 25. From Table I. Eq. (lob). The C - 0 bond energy [see Table VlII for OCH3 and Eq. (30) for HCOO]. f Eq. (10c). Eq. (20b). "

For chemisorption of CH, radicals, our model suggests to use Eq. (lob) for CH and CH,, but Eq. (1Oc) for CH,, projecting for all the species that preferred hollow site. In particular, for Ni( 1 1 1) we obtained QCH, = 1 16, 83, and 48 kcal/mol for x = 1, 2, and 3 , respectively (see later, Table VIII). For comparison, in the latest and most complete ab initio complete active space self-consistent field-configuration interaction (CASSCF-CI) cluster-type calculations of CH, on Ni(lll), Siegbahn et al. (74a,b) have found the hollow site to be universally preferred with QCH, = 122,85, and 49 kcal/mol for CH, CH,, and CH,, respectively.

3. Monocoordination Versus Dicoordination Let us turn now to mono- versus dicoordination of polyatomic species chemisorbed on a metal surface, say Ni(ll1). In the closed-shell formaldehyde CH20, the C-0 bond energy is Dco = 361 - 183 = 178 k calh o l (see later, Table VIII), and we shall accept this value also for acetone (CH,),CO. For the q l p ncoordination of H,CO via 0,we can assume either on-top ( n = 1) or bridge ( n = 2) sites, giving the values of QH,co = 19 and 22 kc a l h ol , respectively. For the q2p2coordination via both 0 and

126

EVGENY SHUSTOROVICH

C, by using eq. (11) we find QHlc0 = 21 kcal/mol. We conclude that the q' and q2 coordinations of H,CO on Ni(1 I I ) will have approximately the same chemisorption energy. Although direct experimental data are lacking, the above conclusion may be compared with the experimental findings for the less active surface Ru(OOI), where both 7' and q2 H2C0 and (CH,)2C0 are formed with QR,cO = 10-16 kcal/mol estimated for the monolayer state (75a,b). For Ru(OOI), where Qo = 100 kcal/mol, Eq. (10a) gives QH2co= 15-17 kcal/mol, in agreement with the above experimental estimates. Similarly, both the 7'and q2coordinations of acetone (CH,),CO = 12 and 16 kcal/mol, respectively (7.5~). coexist on Pd(lI1) with Q~CH3,1C0 Also, q' (CH3),C0 coordinated via 0 was found on Pt(ll1) with Q(CH1)ZCO = 12 kcal/mol (60, 76). For Pt(lI1) and Pd(l1 I ) , where Qo = 85-87 kcal/mol, from Eq. (1Oa) we find QR2co= 12-14 kcal/mol, again in agreement with experiment. A more difficult test is calculations of QAB for chelated adsorbates, for example, formic acid HCOOH and formate HCOO, coordinated via 0 atoms, because we have to assign all the two-center C-0, C-H, and 0-H bond energies. Because DHCOOH,D,,,,, and DHcO are 481,384, and 274 kcal/mol, respectively, we obtain DOH = 481 - 384 = 97 kcal/mol and, for the 7' monocoordinated HCOO, Dco = 384 - 274 = 110 kcal/ mol. By the same token, the total bond energy formed by the hydroxyl oxygen in HCOOH will be Do,H,,,Ho, = 481 - 274 = 207 kcal/mol. For q2dicoordinated HCOO, both C-0 bonds are equivalent, but in order to determine the values of D,, we should first assign the value of DcH. Because the decomposition HCOO -+H + CO, is practically thermoneutral, it leads to the nonphysical value of DCH = 0 (see Section II,C,2). So, we should make a reasonable guess. This C-H bond is formed by the tricoordinated C in the sp2 hybridization state, where the closest analogs of formate HCOO appear to be formyl HCO and formaldehyde H,CO. In all three molecules, the C-H bond has the same length (-1.1 1 A), and the HCO bond angles are also very similar (125-127") (6,26b). The values of DCHin HCO and H,CO, determined for the dissociations HCO -+ H + CO and HICO -+ H + HCO, are 17 and 87 kcal/mol, respectively (6). We assume the effective value of DCH in HCOO to be intermediate between those in HCO and H,CO, and we simply interpolate it as the arithmetic mean D,H = 1/2(17 87) = 52 kcal/mol. For the C-0 bonds in HCOO, this makes Dco = 1/2(384 - 52) = 166 kcal/mol. Finally, assuming DcH to be the same in both HCOO and HCOOH and by using Do,H,,cHo) = 207 kcalimol (see above), we find for the carbonyl C-0 bond in HCOOH, Dco = 481 - 52 - 207 = 222 kcal/mol. The bond energy partitioning in HCOOH and in the nonsymmetric (q')and symmetric ( q 2 )coordinations of HCOO is shown in Eq. (30).

+

BOND-ORDER CONSERVATION

127

Because HCOOH is a closed-shell molecule, we use Eq. (19) with D,, = 222 kcal/mol and D,, = 207 kcal/mol. For Ag(l1 l), where Qo = 80 kcal/ mol, Eq. (19) gives QHcooH = 13 kcal/mol, to be compared with the crude experimental estimate of 10 kcal/mol for Ag(ll0) (77). For another noble metal surface, Au(ll1) with Qo = 75 kcal/mol, we obtain QHcooH = 12 kcal/mol, in excellent agreement with the experimental value of 13 kcal/ mol for Au( 110) (78). Similarly, we can calculate QHcooH for any transition metal surface but the relevant experimental values are not available for comparison because HCOOH readily decomposes on these surfaces (see Section IV,C). For the formate radical HCOO, we use Eq. (lob) with DAB= 110 kcal/ mol for the q 1 coordination and Eq. (20b) with D,, = 166 kcal/mol for the q2coordination. We have calculated the relevant values of QHcoo for Ag(lll), Ni(lll), Fe(llO), and W(110). Because no direct experimental data are available, we shall discuss our projections later in the general context of HCOOH decomposition on metal surfaces. Here we shall only mention that for HCOO the q2 coordination was found to be always preferred over the q ’ coordination, in full agreement with experiment (see Section IV,C). A related question is the q ’versus q2mode of chemisorbed C,H, hydrocarbons (x = 1,2) coordinated via C atoms. Because of some uncertainties in the “quasiatomic” parameterization of CH and CH, groups in the q2 geometry of the H,CC radicals, the relevant values of QHXCC may be less accurate. Nevertheless, the general conclusion, similar to what we found for R2C0, is that the 7’ and modes have close energies (I&). Experimentally, both modes were found, usually with the slight preference of the q2 mode. For HCC, the corroborating evidence has been obtained by electron energy loss spectroscopy (EELS) on Pd(ll1) (79), Ir( 111) (80),and Ru(001) (81, 82). For H,CC, one can cite EELS data for Ru(OOl)p(2 x 2)O (83) and Pt(ll1) (84) and the I3C NMR data on supported Pt particles (85). BARRIERS OF SURFACE REACTIONS B. ACTIVATION

In this section we discuss the accuracy of the BOC-MP projections concerning the activation barriers of dissociation and recombination surface reactions. We shall begin with diatomic adsorbates AB e A f B.

128

EVGENY SHUSTOROVICH

TABLE V Observed Range of kfor A2 Dissociution"

Experimental values of

H,

Fe(l1l)

N2

0 2

Ni(l1l) Ni( 110) Cu(100) W(110) Fe(l10) Fe( 100) Fe(1ll) Pt(ll1)

62 63 62 56 155 138 140 139 85

104 I04 104 I04 226 226 226 226 I19

-0 2 -0

5

- 10 8 2.5 - 0.8 -1

1.7 1.6 I .7 1.7 I .4 1.6 1.6 I .6 I .4

" See text for notations and Table I 1 in Ref. 16 for sources of the experimental values of QA, D,,,and AEX,. All energies in kilocaloriesimole. From AEX, = D,, - kQA [to be compared with the theoretical (LJ) value k = 1.5 in Eq. (22a)l.

'

1. Diatomic Adsorbates The easiest test concerns Eq. (22a) describing the LJ-type dissociation. The equation establishes the linear correlation between the dissociation barrier AE;, for a homonuclear admolecule A, and the atomic heat of chemisorption QA with the slope of k = 3/2. As seen from Table V, for H,, 0,, and N, dissociated on surfaces of metals as varied as Fe, Ni, Cu, W, and Pt, the experimental values of k lie within the range k = I .4-1.7, that is, within 10-15% of the theoretical LJ value of k = 1.5. It should be stressed that, unlike similar linear relations between the activation barriers and the heats of reactions (Bronsted, Polanyi, Frumkin-Temkin-Semyenov, etc.), Eq. (22a) is not a postulate but a corollary of the general principle (BOC-MP) applied to the one-dimensional dissociation AB, -+ A, B,. The error in k of 10-15% makes one feel that Eq. (22a) is basically valid. Quantitatively, however, the error in AEW,, becoming of the order of (0.15-0.25)QA, may reach 10-20 kcal/mol or even more. An important point is that for most systems, with the exception of N,/W(110) and O,/ Pt(l1l), we find k > 1.5, so that the LJ activation dissociation barriers are typically much larger than the experimental barriers, as we anticipated in Section II,A,2. Table VI illustrates this point and shows that the use of Eq. (21c) instead of Eq. (22a) leads to significant improvement, the devia-

+

129

BOND-ORDER CONSERVATION

TABLE VI Dissociation Activation Barriers AE& :Non-L J Corrections'

AB

Surface

DAB

QA

QB

QAB

H,

Fe(ll1) Ni(ll1) Cu(100) Fe(ll0) Fe( 100) Fe(1ll) Ni(ll1) Ni(100) W(110) Fe(lI1) Mo( 100)

104 104 104 226 226 226 257 257 257 257 257

62 63 56 138 140 139 171 171 200 (200 (200

62 63 56 138 140 139 115 130d 125 12.5)' 125)'

7 7 5 8 8 8 27 30 21C 32" 16"'

N,

CO

Ch

0.08 0.06 0.09 0.04 0.05 0.06 0.10-0.12' 0.09-0.1 1' 0.04-0.07* 0.04-0.07* 0.04-0.07*

Eq. (21a)

Eq. (21c)

Exp.

11

2 1.5 7.5 6.5 4.5

0 2 5 8 2.5 -0

10 20 21 18 19 40 30 9 9 9

5

6.5 -0 -6 - 12 -4

-

-6;

-;lf

15' - 12' - 2" -

The values of QN2and Qco were taken from experiment, but QHZ values (not known experimentally) were calculated from Eq. (14). See Tables 1, 11, and V11 and text. All energies in kilocaloriesimole. Eq. (21b) or (22b). ' For AE;':o,g= 5-0. Ref. 152. ' For AE?,,, = 0-( - 5). From the measured value of AE&\ = 23-24 kcalimol (86a,e) via Eq. (23b). Ref. 866. "ForAEZog = - % - I S ) . ' From the measured value of AEE,,,, = 6 kcalimol (866) via Eq. (23b). Assumed to be the same as for W(110).See text. Ref. 86c. From the measured value of AE&, = 20 kcallmol (86c)via Eq. (23b). Ref. 86d. From the measured value of AE?,,, = 14 kcal/mol ( 8 6 4 via Eq. (23b).

'

J

'

tions from the experimental values of AEiz,gnow being not more than 2-3 kcal/mol. In this respect, it might be desirable to revisit experimentally the interrelated values of AEZ,,gand QA for O,/Pt( 1 11) and N,/W( 1 lo), where surprisingly, AE;E2,B > AEZ;iJ.As a precedent, one can mention that for N,/Pt ( I 11) the value of QN has recently been corrected from 127 (30) to 116 kcal/mol (42), which means an increase in AEG$ [cf. Eq. (22a)l by -16 kcal/mol. Equation (21c) is also superior to Eq. (21a) in describing AE&g , for CO dissociation on active transition metal surfaces. Here chemisorbed CO thermally dissociates while being desorbed, which hints AEZB,gI0 [cf. Eq. (33)], and only Eq. (21c) produces consistent results. In particular, for CO/Ni(100), W(110), Fe(lll), and Mo(100), we find the range of

I30

EVGENY SHUSTOROVICH

A E & g = 0 - -12 kcal/mol compared to the experimental range of - 2 - - 15 kcal/mol (86a-e). Let us clarify that these experimental estimates have been obtained via the common Arrhenius-type analysis [see Eq. (32)] and, therefore, are well suited for comparison with our theoretical values of AEZB. At the same time, we make no comparison with the activation energies of C-0 dissociation on Ni obtained in molecular-beam studies ( 8 3 ,because these energies have no direct relevance to the thermal dissociation barriers (see below). As Tables VI and 1X illustrate, Eq. (21 c) correctly reproduces the periodic trends in dissociation of CO, when the metal activity rapidly diminishes along the series W, Fe > Ni, Ru 9 Pd > Pt (88). For recombination A, + B, -+ AB,3,, the counterpart of Eqs. (21c) and (23c) is Eq. (26b). For CO on Ni(100), by using Eq. (26b) with the experimental values of Q,, Qo, and Q,, from Table VI, we calculate the recombination barrier AEZ-o,\ = 44 kcal/mol, in excellent agreement with the experimental estimate of 43 kcal/mol by Astaldi et al. (86e). For CO on Fe( 1 1I), where the values of Q, and Q, are extrapolated (see Table VI), we find the recombination barrier AE&,,, = 56 kcal/mol compared to the experimental estimate of 48 kcal/mol by Whitman et al. (86c). Similarly, let us calculate AE&H,sfor 0, + H, -+ OH, on Cu(ll1). By using Q, = 103 kcal/mol, QH = 56 kcal/mol (see Table I), and QoH = 52 kcal/mol [from Eq. (lob)], we obtain AE&, = 21 kcal/mol, very close to the experimental value of 22 kcal/mol by Mesters et al. (86f). Finally, by using the experimental (or calculated) values of AEZ,,, and Eqs. (21b) and (22b), we find that the bond orders .xi; in the transition states fall in the narrow range of = 0.05-0.10. Because for diatomics AB the common range of the Morse parameters [Eq. (I)] is a = 0.3-0.5 A (89), we project the typical expansion of the gas-phase A - B bond length in the transition state by Ar = 0.6-1 .O A. One can mention that the CASSCF-CI ah initio calculations of chemisorptive dissociation 0, on large Ni,3-45clusters by Panas et al. (90) gave the close range of Ar -0.54-0.72 A.

2. Polyatomic Adsorbates Now we turn to polyatomic molecules, which we shall treat as quasidiatomic. Remember that in partitioning a polyatomic molecule AB into two quasiatomic fragments A + B, the values of QA and QB usually are not known from experiment but are calculated with inevitable uncertainty. Furthermore, the master BOC equation [Eq. (S)] can be explicitly written for a polyatomic AB only in a rather crude fashion when we have to use some group terms [such as xA or xg in Eq. (8b)], which further blurs the

BOND-ORDER CONSERVATION

131

detailed picture of the AB, versus A, + B, energy profiles. Thus, if the analytic non-LJ corrections have a clear physical sense and Eq. (21c) appears to be preferred for diatomic molecules, it is not that clear for polyatomic molecules whether Eq. (21a) or (21c) may prove to be more accurate. For the sake of uniformity, the use of Eq. (21c) would be of great advantage, and indeed, we shall see that Eq. (21c) gives remarkably reasonable results for a broad variety of polyatomic molecules. But we have also found a few cases in which Eq. (21a) works better, for example, for dissociation of triatomic linear molecules A-B-C such as 0-C-0 and N-N-0. The preference of Eq. (21a)for C0, dissociation may be well anticipated. It has been shown [see, for example, the high-resolution electron energy loss spectroscopy (HREELS) studies of CO, on Re(001) (71a), ultraviolet/ X-ray photoelectron spectroscopy (UPS/XPS) studies of CO, on Fe( 111) and Fe(ll0) (37), and computer simulations for CO, on Pt(l11) (716)] that the molecule is practically undistorted (symmetric and linear) in the ground chemisorbed state but strongly distorted (nonsymmetric and bent) as an intermediate preceding the dissociation CO,,, + CO, + 0,. So, there is a good reason to believe that in the transition state the coordinated C-0 bond is strongly expanded [by 0.12 (716)] and becomes very weak. But the weaker the C-0 bond (xco = 0), the more accurate is Eq. (21a). The dissociation-recombination reaction COZ3, CO, + 0, has been thoroughly studied on a variety of metal surfaces (91-95). Because Q, is known only for Ni, we shall not calculate the values of Qco but take them from experiment. In Eq. (21a), DAB= DCO,- Dco = 384 - 257 = 127 kcal/mol. Also, because the thermodynamic criterion [Eq. (25a)l is valid, we can use Eq. (24) directly to obtain AEi-B.The calculated values of the dissociation AE&,, and recombination AE&,, barriers are given in Table VII and compared with experimental data at low coverages. The agreement is excellent, the error being less than 10% (1-3 kcal/mol). In particular, as Eq. (24) predicts, the recombination barrier AE&co is very sensitive (and rather close, indeed) to Qco but insensitive to Q , (because Q, % Qco). For example, for Ag(l10) and Pt(l1 I), the values AE&., = 5 and 25 kcalimol follow Qco = 6.5 and 32 kcal/mol but not Q, = 80 and 85 kcal/mol, respectively. For the same reason, at high oxygen coverages on Pt surfaces, the barrier AE&co drops dramatically, by a factor of 2-3 (91, 96), because Qco drops this way (18c, 18e). Another example is the recombination-decomposition reaction

A

*

NO,

+ N, S N 2 0 ,

-+

NZ,g + 0,

(31)

carefully studied on Pt( 111) (97), Rh( 111) (98), and Rh( 100) (99). From Eq. (23a), the dissociation barrier AE,*,2)0,s appears to be strongly negative,

132

EVGENY SHUSTOROVICH

TABLE VII Dissociation und Recornhinution Barriers AE* j b r Some Sttrfuce Reucrion,ns Involving Triuromic Adsorhum"

Experimental values of Reaction

CO,,,

-+

CO,

+ 0,

+ O,+COz,,

CO,

+N,4

NO,

NlO,

4

N?,,

N,O,"

+ 0,

Surface

DABh

Rh( I I I ) Re(OOI) Rh(lll) Pd( 1 I I ) Pt( I I 1) Ag(I 10) Rh(lll) Rh(100) Pt(l1 I ) Rh(l I I ) Pt( I I I )

127 127 -

-

40 40

AE*

QA"

Ref.

QBd

Ref.

32 29 32 34 32 6.5 26 25 27 9 9

92 Y4 92 30

102 127 102

39 94 3Y

87

30 30 Y7 98 30 87 87

85 80 (128)" (131)''

30 30 30

116

102 85

-

-

42 39 30

Calc. 17" - 5' 24' 24 I 23 6.0' 22 21f 22, - 63" - 46" f

Exp.

Ref.

17

Y-3

so

94 72 91 91

27 25 25 5.3 21 21' 20'

95 98 99 97

(' See text for notations and explanations. All energies in kilocalories/mole. From Ref. 25. A = CO or NO. 'f3 = O o r N . ' Eq. (21a). Eq. (24). Followed by nonactivated decomposition N,O, Nz,g + 0,. 'I Extrapolated values (cf. Table 1). ' The reported experimental values were AE* = 28-29 kcal/mol for the assumed slandard preexponential u , = 1OI3 s - ' , but the barrier will be AE* = 20-21 kcal/mol, if one assumes the preexponential v , = 2 x lo9 s - ' found for the same reaction on Rh(l 1 1 ) (Y8). I'

---f

that is, nonexistent. This model projection is in full agreement with extremely facile decomposition of N20, assumed to be the reaction transition state (97-99). If so, the recombination barrier AEi-N, becomes the activation barrier of the whole reaction [Eq. (31)]. Accordingly, from Eq. (24) we arrive at A E i - N O = 21-22 kcal/mol, in excellent agreement with the experimental values of 20-21 kcal/mol. Again, as Eq. (24) predicts, the are close to the respective values of QNo = 25-27 kcal/ values of AEiwNO mol but have no distinct relation to the values of Q N = 116-131 kcal/mol. It is worth repeating that the high accuracy of Eq. (24) for atomic-molecular recombination A-BC stems from the strong inequality QA P QBC leading to AEi-,, < QBc. Because QBc G QB, Qc (cf. Section IILA), the recombination barrier AEi-Bc will be confined to a small range of a few kilocalories/mole and show a very weak (periodic) dependence on the heats of chernisorption of the atomic constituents QA, QB, and Q,.

133

BOND-ORDER CONSERVATION

TABLE VllI Heats of Chernisorption (Q) und Total Bond Energies in the G a s Phase (DJ and Chemisorhed (D + Q) States on the Platinum-Group Metals"

C CH CHI CH3 CH4 H 0

OH OH2 OCH, CHiOH

co

HCO HZCO

81 183 293 398

-

I02 220 383 487 257 214 361

171 116 83 48 6' 63 115 61 17 65 18 27d 50 19

171 197 266 34 I 404 63 I I5 163 237 448 505 284 324 380

160 I06 75 42 6'' 62 87 40 10 43 I1 34d 44 12

160 187 258 335 404 62 87 I42 230 426 49n 29 I 318 373

150 97 68 38 6' 61 85 39 10

41 11

32d 40 11

150 I78 25 1 33 1 404 61 85 141 230 424 498 289 3 I4 372

' See text for the relevant formulas and explanations. All energies in kilocaloriesimole. Ref. 25. For the values of DAB# D used in calculations of QAB,see Table 111 (CH,OH and H2CO) and Table IV (OCH,). ' Taken as the experimental value of Q,,, = 6 kcal/mol on Rh [Brass, S . G . , and Ehrlich, G . , Surf. S c i . 187, 21 (1987)l. Experimental values (30).

Now let us proceed to polyatomic molecules and consider first the C-H bond cleavage in alkanes. For the dissociation CH4 +. CH, + H, DCH= 398 - 293 = 105 kcal/mol. For CH4on Ni( 11 1) and Ni( IOO), where QCH,,Q C H , , and Q H are 6, 48, and 63 kcal/mol, respectively (see Table VIII), Eq. (21c)gives AEM C.H,g = 7 kcalimol. This value may be compared with the activation energigs of 12 and 6 kcal/mol for methane decomposition on Ni( 111) and Ni( IOO), respectively, measured in thermokinetic studies by Beebe et al. (100). It should be stressed that a comparison made between measured thermal sticking coefficients (100) and those calculated from molecular-beam studies of CH4on Ni(l11) (IOla)and Ni(100) (1Olb) showed correlation only with the Ni(l11) beam results (100). But even here the activation energies that appeared in the molecular-beam experiments should be transformed to fit a Maxwell-Boltzmann distribution to become relevant to thermal dissociation, and this transformation effectively scales down the activation energy of CH, dissociation on Ni(ll1) from >I8 kcal/mol (IOla) to 12 kcal/mol(100). In general, because molecular-beam

I34

EVGENY SHUSTOROVICH

experiments probe mostly translational activation of chemisorption, they may be not informative about the dissociation, which requires vibrational excitation, especially the “nonactivated” thermal dissociation with ALE:,, 5 0 (see the above discussion of CO on Ni, Fe, Mo, and W). In Section IV we shall discuss numerous examples of C-H bond cleavage in both hydrocarbons and their oxygenated derivatives. Special attention will be given to stability and transformations of radicals such as methoxide CH,O, commonly observed on the late transition metals (102). Similarly, we shall also discuss the C-C bond scission in various C2H, hydrocarbons. Here one example will suffice. Consider the C-C bond scission in ethylidyne H,CC formed under dehydrogenation of ethylene on many transition metal surfaces (103, 104). By using D,, = 117 kcal/ mol, we find from Eq. (23c) for dissociation CH,C, -+ CH,,, + C,, ALE& = 11, 8 , and 5 kcal/mol for Pt(lll), Ni(lll), and Fe(llO), respectively (see Table XIII). These calculated values can be compared with the experimental temperature-programmed static secondary ion mass spectrometry (TPSSIMS) activation energies of CH,C decomposition, found ) 12 kcal/mol for Ru(OO1) (105h), to be 17 kcal/mol for Pt(l11) ( 1 0 5 ~and the latter surface being intermediate in activity between Pt( I I 1) and Ni(lI1). Thus, the model predicts that stability of CH,C should be sensitive to metal composition, namely, the stability decreases as the metal activity increases. Indeed, ethylidyne has been readily observed under decomposition of ethylene on the close-packed surfaces (103) and metal particles (85b, 1 0 4 ~of) Pt, Pd, Rh, Ru, and Ir, only at high adsorbate coverages on Ni (104b,c) and Co (1044, but not on more active metals. In the next section we shall see how our model projections can shed light on mechanisms of heterogeneous catalytic reactions.

IV. Mechanisms of Catalytic Heterogeneous Reactions The BOC-MP method provides reasonably accurate estimates of the heats of chemisorption Q and the dissociation and recombination barriers AE* for various molecules and molecular fragments. Combined with the knowledge of the molecular total bond (gas-phase dissociation) energies, this allows one to construct potential energy profiles of surface reactions. The question now is how these energy profiles relate to reaction rates. Remember that the reaction rate constant r of an elementary process is described in the Arrhenius form as (5, 106) r = A exp( - AE*/RT)

(32)

where A is a preexponential factor and AE* is an activation energy. Be-

BOND-ORDER CONSERVATION

135

cause of their exponential effect, the values of AE* are usually critical. At room temperature, an increase of AE* = 1 kcal/mol leads to a fivefold decrease in the reaction rate. The BOC-MP model provides the barriers AE*. Clearly, in order to make a meaningful comparison between two reactions based solely on the values of AE*, the values of A should be close. This condition is most likely to be met for a given reaction on a given surface, say fcc( 11 l), but of two different metals, say Pt( 1 1 1) versus Ni(ll1). Related to this is a comparison of rates, rdeqversus rd,,,, of two first-order reactions, desorption AB, -+ AB, and dissociation AB, -+ A, + B,. Because of the fundamental inequality of the preexponentials Ades S Adlc, [Usually Ade5/Adlss 103-104 (62, 102d, 107)] and because AE&, = AEXB,, = AEZR,,+ Q A B [Eq. (23a)I and AE2e, = Q A B , the typical condition of prevailing dissociation is

> rdes

<0

(33) Let us clarify that Eq. (33) may be valid only at low enough temperatures, making the exponential factor decisive, but at high temperatures the preexponential factor becomes more important and the desorption prevails (62). The second-best case to consider is two reactions of the same order differing by one reactant only, for example, C, + H, S CH, versus CH,,,. Finally, if a vacant surface site S is treated as a CH, + H, reactant, one can compare dissociation reactions with recombination reactions, e.g., S + CO, S C, + 0, versus H, + CO, e HCO, + S. Here, however, the conclusions based on AE" will be the least definitive, because the differences in the nature of surface sites (on-top, bridge, or hollow) and in the effective number of the sites involved in the reaction may strongly affect the value of A . Another principal limitation is that the discussed BOC-MP interrelations [Eqs. (7)-(27)] have been obtained for low adsorbate coverages (formally, in the zero-coverage limit). Steady-state conditions of real heterogeneous reactions usually correspond to substantial coverages of coadsorbed species. Coverage and coadsorption effects change the heats of chemisorption and the activation barriers in a complex (nonlinear) way. These effects have also been treated within the BOC-MP framework (18c,e). Qualitatively, most of the results are unambiguous and consistent with experiment. In particular, tk,e model predicts (18c,e)that, beginning with a certain coverage, the heat of chemisorption decreases as the coverage increases, and this effect is especially pronounced for molecules coadsorbed with multiply bonded atoms, say for CO + S or CO + 0. Furthermore, there is growing evidence that not only AE*, but also the preexponential A , may vary strongly with coverage, often in a puzzling way (107). Under such circumstances, we will proceed with zero-coverage rdivs

if

AEXB,g

136

EVGENY SHUSTOROVICH

estimates in the hope that they qualitatively reflect the relative energetics of elementary steps comprising the reaction mechanism at the steady-state conditions. One more comment seems necessary. The Arrhenius expression [Eq. (32)] is commonly used to describe the rates of nonelementary reactions including several steps. In this case, the measured value of AE* is the apparent (global) activation energy, which is the resultant of sums and differences (with some coefficients) of activation energies of elementary steps whose rates contribute to the global rate (108). In our model approach, we calculate AE* for elementary steps only. Thus, there is no direct and simple way to compare our calculated barriers with the apparent barriers of nonelementary processes. This is particularly true for energy estimates made from the thermal-stability thresholds of chemisorbed species. OF CO ON A. HYDROGENATION

THE

PLATINUM-GROUP METALS

Hydrogenation of CO on transition metal surfaces produces a variety of hydrocarbons and their oxygenated derivatives (88, 109). Owing to its fundamental and practical importance, this catalytic heterogeneous process has been studied intensively. Here we shall limit our discussion to hydrogenation of CO over the platinum-group metals leading to C, species such as methane (CH,) and methanol (CH,OH). This process shows distinct periodic regularities; namely, CH, has been produced on Ni, Pd, and Pt (110), but CH,OH has been produced only on Pd and Pt (111). So, we wonder what might be possible pathways for CH, and CH,OH synthesis and how they depend on metal composition. More specifically, the critical questions are as follows: (1) Why cannot the C-0 bond be retained on nickel catalysts? (2) How does C-0 bond cleavage occur, directly from CO or from partially hydrogenated species H,CO (hydrogen-assisted C-0 cleavage)? Table VIII lists the total bond energies of reactants and products as well as conceivable intermediates in the gas-phase (D)and chemisorbed ( D + Q) states on Pt(1 I l), Pd(l1 I ) , Ni(l11). Tables IX and X summarize the activation barriers of the conceivable elementary steps leading to CH, and CH,OH, respectively. As seen from Table VIII, in the gas phase the methanation reaction CO + 3H, + CH, + H20, with AH = -49 kcalimol, is much more exothermic than is the methanol formation CO + 2H, -+CH,OH, with AH = -22 kcalimol. Thus, the only chance to selectively produce methanol is to have a catalyst on which hydrogenation of H,CO species (x =

137

BOND-ORDER CONSERVATION

TABLE IX Activation Barriers for Forward (AEF) and Reversed (AET) Elementary Reaciions for Methanation CO + H2 + CH4 + H 2 0 on N i ( l l 1 ) and Pd(ll1)” AEf Reaction

AET

Ni

Pd

Ni

50 40 15 9 9 9 35 46 24 79 51 22 0

35 5 23 24 8

Pd

~~

33 42 17 12 14 14 23 35 18 46 29 28 10

~

6 5 24 24 10 16 0 2 8 0

14

0 23 28 II 16

0

13 21

15 26

‘ See Table VIII for the values of Q and D + Q used in calculations of AE*. The dissociation barriers from Eq. (21c) or (23b), disproportionation barriers from Eq. (29c), and recombination barriers from Eq. (26b) or (27b). All energies in kilocalories/mole. TABLE X Activation Barriers for Forward ( A E j ) and Reversed (AET) Elementary Reactions for Methanol Formation CO + Hz -+ CH30H on N i ( l l 1 ) and Pd(1ll)u AEF

AET

Reaction

Ni

Pd

Ni

CO, + H, ~2 HCO, HCO, + H, $ H2C0, HZCO, S CH?,, + 0, HZCO, + H, F? CH,O, CHjO, S CHj, + 0, CHjO, + H, F? CHjOH, CHjO, H, S CHjOH, CHIOH, F? CH, , + OH, CHjOH, + CHj,, + OH,

23 33 24 5 13 24 19 -4 14

35 16 34 10 16 7 7 10 22

0 26 23

+

See footnote a in Table IX.

10

21 -5 13 17 13

Pd 0 9 6 1 12 6 17 0 0

138

EVGENY SH USTOROVICH

0-3) is kinetically preferred over the H,C-0 bond cleavage. Even then, because the thermodynamic and kinetic factors appear to compete, the same metal catalyst may produce both CH, + H 2 0and CH,OH, depending on the reaction conditions and the catalyst structure. On the other hand, if the barriers for the H,C-0 bond dissociation are smaller than those for the H,CO hydrogenation, the thermodynamic and kinetic factors are in accord, and a catalyst may selectively produce methane only. We begin by considering the pathways for the formation of CH4and CH,OH on Ni. As noted above, for Ni(l11) we obtained the gas-phase dissociation barrier AEZ,, = 6 kcal/mol. Thus, for direct dissociation CO,+ C, + 0, we find AE& = 33 kcal/mol Land 30 kcal/mol for Ni(100)], and the reaction is slightly exothermic by AH = -2 kcal/mol (see Table VIII). In principle, the carbidic carbon C, can also be formed via the hydrogenassisted C-0 bond cleavage CO,

+ H , + [ H . . . C O ] , + C , + OH,

(34)

which is moderately endothermic by A H = 13 kcal/mol (cf. Table VlII). From Eq. (29c), the activation barrier for this disproportionation reaction will be 29 kcal/mol, which is smaller by 4 kcal/mol than the barrier for direct CO, dissociation. The results are consistent with the fact that carbidic carbon C, is formed o n Ni while heated either in clean CO or in H, + CO, but with H, the CO dissociation proceeds much faster (2, 110). It is also of relevance that in the FT-IRAS study of the CO methanation on Ru(OOI), which is very similar to the methanation on Ni surfaces (112),Hoffmann and Robbins (112) suggested the formation of the H - - - CO complex to reconcile all experimental findings. Once C, is formed, the progressive hydrogenation C, + H, -+ CH, H, -+ CH2,, + H, + CH,,, + H, + CH, takes place where the activation recombination barrier AE:,C-H decreases with x ; namely, AE$,C-H = 42, 17, 12,and 14kcal/molforx = 0, 1,2,and3,respectively,forbothNi(lII) and Ni(100). In principle, the CH, intermediate can also be obtained by the disproportionation H, + CO, -+ CH, + 0, with the comparable barriers of 46 and 42 kcal/mol for Ni( 1 1 1) and Ni(100), respectively. Thus, we project that methanation on Ni is structure insensitive and the ratelimiting step (if any) is the formation of CH, via either the hydrogenation of C, or the HC-0 bond cleavage, in agreement with experiment. In particular, Campbell and Goodman (86a) have found that potassium promotion of CO hydrogenation on Ni( 100) decreases the activation barrier of CO dissociation (from 23 to 10 kcal/mol) but does nor alter the apparent activation energy of methanation, which eliminates CO dissociation as the rate-limiting step. Furthermore, as has been shown by Yates et af. ( I f O c ) ,

+

BOND-ORDER CONSERVATION

139

for the hydrogenation of CH,<, compared to the hydrogenation of CO, on Ni( 11 l ) , the overall CH, production rate per Ni site increases by about six (!) orders of magnitude, which places the rate-limiting step of methanation much earlier than the final hydrogenation step. Finally, the methanation kinetics is virtually identical for single-crystal surfaces Ni( 111) and Ni( 100) and for Ni/A120,, which is a highly dispersed supported Ni catalyst (2, 110).

We can readily understand why Ni is not effective in forming CH,OH. First of all, the dissociation HCO, + C, + OH, with AE* = 18 kcal/mol is strongly preferred over the competing hydrogenation HCO, + H, -+ H,CO, with AE* = 33 kcal/mol. Furthermore, if the C-0 bond could somehow survive up to the formation of methoxide CH,O, [as was found by Glugla et al. on Ni/AI,O, (1106)],it will be easily cleaved as CH,O, -+ CH,,, + 0, with AE" = 13 kcal/mol. The alternative hydrogenation to methanol CH,O, H, -+ CH,OH, would have required the distinctly larger activation barrier AE" = 19 kcal/mol, not to mention that CH,OH, cannot be desorbed without decomposition, because AE$,,CO)H = - 5 kcal/mol [cf. Eq. (33)]. This value is in excellent agreement with the experimental laser-induced desorption (LID) estimate of - 5 kcal/mol found for CH30H/Ni(100)by Hall (62).Thus, on Ni the kinetic preference of (2-0 (and 0-H) dissociation further enhances the thermodynamic preference of the CH, + H 2 0 over CH,OH formation. It is another story for CO hydrogenation on Pd (and Pt). First of all, on Pd(ll1) the formation of carbidic carbon is highly endothermic and requires much higher activation barriers, namely, 50 kcal/mol for CO, + C, + 0, or 51 kcal/mol for CO, + H, -+ C, OH,. The subsequent hydrogenation of C to CH, would further require the barriers of 40, 15, 9, and 9 kcallmol for x = 1, 2, 3, and 4, respectively. Thus, the major route appears to be the progressive hydrogenation CO, + H, -+ HCO, -+ H$O, --j H,CO,, which occurs with barriers of 35, 16, and 10 kcalimol, respectively. Unlike Ni, now the hydrogenation HCO, + H, + H2C0, is preferred over the dissociation HCO, -+ C, + OH, (the barriers are 16 and 24 kcal/mol, respectively). Also, in contrast to Ni, hydrogenation of methoxide CH,O, + H, + CH,OH, requires a lower barrier (7 kcalhol) than that (16 kcal/mol) for the C-0 bond cleavage to CH,,, 9 + 0,, and now methanol can be effectively desorbed [AE&,CO)H,g S 01. Thus, on Pd the thermodynamic and kinetic preferences for the formation of CH,OH and CH, + H 2 0 appear to be in conflict, which makes it possible for Pd to be a selective catalyst for either CHIOH or CH,. This ambivalence seems very likely because differences in the activation barriers for the C-0 dissociation and relevant hydrogenation are not large, being, for example, only 8-9 kc a l hol for HCO, and H3C0, (see Tables IX and X).

+

+

140

EVGENY SHUSTOROVICH

Consistently, Pd metal has been found to be a good methanol synthesis or methanation catalyst, the selectivities varying from 100 to 0% depending on the reaction conditions, the nature of the support, promoters, etc. (111). The important point, however, is that unsupported Pd, either as Pd powder (1116)or the single-crystal Pd(ll0) surface (111f), may be an active methanol synthesis catalyst. Finally, we project that for the CO methanation the relative significance of hydrogen-assisted H,C-0 bond cleavage (the value of x) increases from Ni to Pd, in agreement with experiment (113), particularly with H,/D, isotopic studies by Mori el al. (1134. As a related subject, consider briefly the insight that the BOC-MP modeling can bring into decomposition of methanol CH,OH on metal surfaces. In the gas phase, of two possible dissociation routes, CH,OH + CH, + OH versus CH30 + H, the relevant dissociation bond energies are Dco = 93 kcal/mol and DOH = 104 kcal/mol (cf. Table VIII), which makes the formation of CH, and OH preferred. In the chemisorbed state, however, we find (see Table VIII and X) that the formation of CH,O, and H, is more exothermic and requires a smaller barrier than that for CH,., and OH,, although the differences in the barriers are not large, comprising 1-5 kcallmol. Consistently, various studies have shown that decomposition of CH30H, occurs via the methoxy CH,O, intermediate on a large number of clean surfaces of Fe, Mo, W, Ru, Ni, and Pd as well as oxygenmodified surfaces of Pt, Cu, and Ag (62, 102). At the same time, in the recent XPS and secondary ion mass spectrometry (SIMS) study of thermal decomposition of CH,OH on Pd( 1 1 1) by Levis et al. (102c),both H,CO-H and H,C-OH bond cleavage have been reported depending on CH,OH exposure. Our further model projection that the formation of CH,O, is followed by stepwise hydrogen abstraction, CH,O,+ CH,O, + ( 3 - x)H,, x = 0-2, has been corroborated by EELS dataof Bhattacharya et af. (102e) for Pd( 110) and pulsed-field desorption mass spectrometry (PFDMS) data of Kruse et al., (102f)for Ru and Rh stepped single-crystal surfaces. The BOC-MP model also sheds light on the role of preadsorbed oxygen in facilitating the 0-H bond cleavage in CH,OH and other alcohols (and organic acids). This role is not just to shift the dissociation equilibrium by removing H,, but, most importantly, the formation of H,O, substantially improves the thermodynamic conditions of the 0-H dissociation, which becomes critical on low-activity metals such as Ag or Au. As follows from the data in Tables I , IV, and VIII, on Ag(ll1) the direct dissociation CH,OH, CH30, + H, is highly endothermic by A H = 24 kcal/mol. At the same time, the oxygen-assisted dissociation 2CH,OH, + 0, -+ 2CH,O, + H,O, is practically thermoneutral with AH = 3 kcal/mol. Accordingly, the activation barrier for the direct dissociation CH,OH, --z CH,O, + H, is calculated to be 23 kcalimol [cf. Eq. (23b)], whereas the

-

BOND-ORDER CONSERVATION

barrier for the oxygen-assisted dissociation CH,OH, OH, is 18 kcal/mol [cf. Eq. (29c)l.

B. TRANSFORMATIONS OF C,H,

ON

141

+ 0, +-CH,O, +

TRANSITION-METAL SURFACES

Surface reactions of hydrocarbons are another class of catalytic heterogeneous processes of great fundamental and technological importance (114, and here C,H, species are among the best studied. Some of them, such as ethane (CH,CH,), ethylene (CH2CH,), or acetylene (CHCH), are stable in the gas phase and are therefore quite familiar, but others, such as ethylidyne (CH3C), vinylidene (CH,C), or acetylide (CHC), have been identified only in their chemisorbed states. Similar to hydrogenation of CO, we are mostly interested in periodic (relative) changes in catalytic behavior of transition metals. Specifically, we choose the sequence from Pt to Ni and further to Fe or W, the latter two being simulated as the same model metal Fe/W [with the model parameters averaged over the (close) parameters of real Fe and W]. The BOC-MP calculations have been made for the smoothest (most densely packed) surfaces, namely, fcc Pt(11 1), Ni(l1 I ) , and bcc Fe/W(110). Our discussion will be organized along the following lines: ( I ) the thermochemistry of C,H, species in the chemisorbed versus gas-phase states; (2) the influence of metal composition on the general patterns of C,H, and C2H, decomposition; (3) the effects of metal composition and the structure of C2H, species on the activation energy for C-H and C-C bond cleavage, Tables XI and XI1 list total bond energies in the gas phase ( D ) and chemisorbed (D + Q ) states for all C,H, species (x = 0-6). The calculated activation barriers AE* for C-C and C-H bond cleavage and recombination for chemisorbed C,H, species are summarized in Table XIII. All the discussion below will refer to chernisorbed species if not stated otherwise. Of general model conclusions, perhaps the most important is the following: many reorganizations of C2H, species, being highly endothermic in the gas phase, typically become exothermic on transition metal surfaces. Some examples are given in Table XII. In the gas phase the ground state of C2H3 is vinyl, H,C=CH, which is lower than ethylidyne, H,C-C (the excited state), by 45 kcal/mol. Under chemisorption, this isomerization becomes exothermic by 8, 15, and 25 kcal/mol for Pt, Ni, and Fe/W, respectively. Similarly, isomerization of acetylene to vinylidene HC=CH + H,C=C is highly endothermic in the gas phase ( A H = 44 kcal/mol) but becomes highly exothermic by 13-41 kcal/mol on the metal surfaces studied. This makes it comprehensible why H,CC and H2CC are

142

EVGENY SHUSTOROVICH TABLE XI Total Bond Energies in the Gas-Phase (D)and Cheinisorbed ( D on Some Transition Metals"

+

&:H,

+ (8

QC~H,'

C2Hx H,C - CHj H3C - CH2 H3C - CH H2C = CH2 H3C - C H2C =rCH H2C = C HC = CH HC=C CH3 + CH3 CH3 + CHI CHI + CH CHI + C CH2 + CH2 CH2 + CH CHI + C CH + CH CH + C C + C CH4 + CH4

DCZH,I)

674 576 466? 538 3761 421g 348" 392 259' 586 476 374 293 366 264 183 162 81 0 796

Q) States + Qc~H, x)QHd

-

FelW

Ni

Pt

FelW

Ni

Pt

6 64 107 20 141 71

5 49 85

5 39 70 12 97 44 71 14 69 76 106 135 188 136 I65 218 I94 247 300 12

812 838 837 822 847 822 854 813 827 842 840 842 885 838 840 883 842 885 928 810

805 814 803 805 806 79 1 813 788 784 808 796 790 827 784 778 815 772 809 846 808

801 798 780 794 778 770 785 772 755 784 765 753 786 746 734 761 722 755 788 808

110

25 I06 124 166 204 262 208 246 304 284 342 400 14

15

I I5 55 87 18 84 96 131 I64 219 166 199 254 232 287 342 12

' All energies are in kilocalories/mole. The parameters used: Qc = 150, 171, and 200 and Q H = 61, 63, and 66 for Pt, Ni, and FelW, respectively. See text for notations and explanations. From Ref. 25 with corrections and additions specified below in footnotes e-i. ' Eqs. (10)-(20). Normalized for the stoichiometry C2HB.For C2H, (or CH, + CH,_,) the remaining (8 - x) atoms H are assumed to be atomically adsorbed. ' For the average ('.'A) state of CH3CH, for which D c H ~ C H -~ DCH3c~= 72 ( 2 8 ~ ) . For the ground ('A) state of CH3C, for which DCH~CH = 45 (28a). From DcH~cH~ - D C H Z C H = 117 (27). For the ground ( ' A ) state CH*C, for which D ~ H-~ DCH~C H = 44 (28). From DCHCH - DCHC = 133 (27).

'

often observed in the chemisorbed states, unlike the gas phase. Moreover, we can project the relative stability of these and other C2H, species, depending on metal composition. Experimental data to quantitatively verify the model projections are mainly limited to the least active surface Pt( 1 1 I), and they are consistent

143

BOND-ORDER CONSERVATION

TABLE XI1 Enthalpies of C2H, lsomerization Reactions: Gas-Phase versus Chemisorbed States AH"

Reaction

C2H4 C2H3 C2H2

Gas phase

H,C=CH, + H$--CH HzC=CH + H3C--C HC=--CH + H2C=C

12 45 44

Fe/W(lIO)

Ni(lll)

I5 - 2s -41

- 1s

-

2 - 25

Pt(lll) 14 -8 - 13

" In kilocaloriesimole; From Table XI.

with our numerical estimates. We have already said (see Section II1,A) that the BOC-MP value of Q,,,, = 12 kcal/mol for Pt(ll1) lies in the middle of the reported experimental range of 9-18 kcal/mol(67). Furthermore, our calculated activation barriers for C-H bond cleavage in C,H, (C,3-H) and C2H4 (C,z-H) chemisorbed on Pt(l11) are 13 and 25 kcal/ mol, in close agreement with the experimental estimates of 12-15 and 27-30 kcal/mol, respectively, reported by Campbell (86g). Let us begin with the energy profile for ethylene C2H4. Because we cannot directly estimate the isomerization activation barriers, we shall refrain from mechanistic speculations about how C2H4 transforms into ethylidyne CH,C, either via ethylidene CH,CH or vinyl CH,CH, namely isornerizdtiun

dehydrogenation

CHzCH2

-]CHIC dehydrogendliun

CHzCH

isornerlzdtlon

-n

The formation of CH,C from C2H, is strongly exothermic on Fe/W, almost thermoneutral on Ni, and moderately endothermic on Pt (AH = -25, - 1, and 16 kc a l hol , respectively). Obviously, the lower limit of the activation barrier to form ethylidyne from ethylene is just the enthalpy difference for the surface reaction C,H, -+ CH,C + H . On Pt(lll), we thus project AE* 3 AH = 16 kcal/mol for the zero coverage extreme, to be compared with the activation barrier of 14-18 kcal/mol (67b-d) measured by various experimental techniques for high (usually, saturation) coverages. Once CH,C is formed, however, its fate appears to be sensitive to metal composition. On Pt( 1 1 I), we predict CH3C to be rather stable, because the calculated C-C bond-scission barrier is AE& = 11 kcal/mol. But on Ni( 11 1) and Fe/W(1lo), we predict this barrier to become smaller, 8 and 5 kcal/mol, respectively. Thus, the model conclusion is that the stability of CH,C decreases in the order Pt > Ni > Fe/W, in agreement with experiment (103-105), as we discussed before. Furthermore, on

144

EVGENY SHUSTOROVICH

TABLE XI11 Actiuution Barriers fbr Forwurd and Reversed Reuctions of Chemisorhed C2Wr

CHjCH,,, CH,CHi CH3CH2

* *

* &

CH2CH2,, CHZCH2 CH&H

+ + + + * + S

* CH2CH

S F?

+ CH,C

+ $

CHCH, CHCH

+ +

* &

CHZC CHC

* & *

CHiCHz + H CH1CH2 + H CHI + CH2 CHZCH2 + H CHiCH + H CH2 + CHI CH2CH + H CH2 + CH, CHzCH + H CH, + CH CH,CH + H CH,C + H CH2 + CH CHCH + H CH2C + H CH, + C CHzC + H CH + CH CHC + H CH + CH CHC + H CH2 + C CHC + H CH + C

98 98 I00 38 110

172 117 I72 117 92 45 90 157 29 73 83 28 230 133 230 I33 165 89 178

5

-3 3

10

18

24

16 21 -2 -3 18 17 19 25 17 21 14 5 5 2 -4 - 12 21 13 20 34 13

11

24 17 7 32 22 23 21 19 31 9 7 8 15

19 1 37 19 27 32 22

8 13 33 4 25 36 13 48 25 27 18 20 38 5 9 11

13 36 11

50 25 32 31 29

32 29 20 0 20 36 20 34 17 24 10 27 39 5 37 43 9 54 39 50 27 49 4 71

19 19 6 2 13

II 8 II 8 10 9 22 18 6 29 29 22 21 18 21

18 29 3 47

10 10

0 0 7 0 1 0 1 0 8 18 2 7 24 19 20 0 8 0

8 14 1 29

" See text for notations and explanations. The barriers for the gas-phase ethane, ethylene, and acetylene are also added. All energies in kilocalories/mole. The difference between the gas-phase total bond energies of the reactant and products (see Table XI). ' Eq. (21c) or (23b). The values of Q A , QB, and QAB are from Table XI. Eq. (26b) or (27b).

Pt(lll), the C-C bond cleavage CH3C + CH, + C (AE& = 11 kcali mol) appears more favorable than dehydrogenation, CH,C + CH,C + H (AE& = 13 kcal/mol), so that the molecule will retain most of its hydrogen up to the point of C-C bond scission, in full agreement with the I3C NMR experiment (85b,c). It is worth repeating that even small differences in AE* may have profound kinetic consequences. At room temperature, other conditions being equal, an increase of AE* = 2 kcal/mol leads to a 30-fold decrease in the reaction rate [cf. Eq. ( 3 2 ) ] .

BOND-ORDER CONSERVATION

145

For a given surface, acetylene chemisorbs more strongly than ethylene, but not much. Our estimates are Qc,H, = 14,18, and 25 kcal/mol compared with Qc,H, = 12, 15, and 18 kcal/molfor Pt(lll), Ni(lll), and Fe/W(l10), respectively. Experimental data on QczHz[usually from temperature-programmed desorption (TPD) spectra] are not available because C,H, begins to decompose before it desorbs. Consistently, we found the gas-phase C-H bond cleavage to be nonactivated on Ni and especially on Fe/W, namely, AEzH,g = 1 and -12 kcal/mol, respectively. On Pt, where AE$H,g = I 1 kcal/mol, the first surface reaction may be the distinctly exothermic isomerization CHCH + CH,C, with AH = - 13 kcal/mol (unfortunately, as for other isomerization processes, we do not know how to calculate the activation isomerization barrier AE*). Indeed, the formation of the vinylidene CH,C intermediate on Pt(ll1) was first suggested by Ibach and Lehwald (84) from EELS spectra and confirmed on Pt particles by the I3C NMR analysis by Wang et al. (85a,c>.However, this isomerization is not a favorable route for C-C bond scission, because CH,C --j CH, + C would require AE& = 32 kcal/mol. Dehydrogenation CHCH + CHC + H may occur first (AE& = 25 kcal/mol), followed by the C-C bond scission CHC ---z CH + C (AE& = 29 kcal/mol). Thus, contrary to decomposition of CH3C, one can expect a loss of hydrogen before the C-C bond rupture, in agreement with the I3C NMR data by Wang et al. (85c) for C,H, on supported Pt. Another alternative appears to be disproportionation 2C2H2 + CH3C

+ CHC

for which, from Eq. (29c) and Table XI, we obtain AEt,S = 26 kcal/mol. Consistently, such disproportionation has been found on Pt( 1 11) by Avery (l15a) with the estimated activation barrier of -23 kcal/mol(115a, 11%). On Fe/W(l lo), the situation looks rather different because for CHCH not only C-H but also C-C bond cleavage seems to be nonactivated from the gas phase (AE&-,g = - 4 kcal/mol). Thus, on Fe surfaces, one can expect acetylene to decompose rapidly into CH, fragments. This model projection is consistent with the fact that under heating of chemisorbed CHCH, only CH, intermediates have been observed on various Fe surfaces by Erley et al. (116a) and Seip et al. (116b). Between Fe/W and Pt, the decomposition products may be a variety of C,H, and CH, species, depending on metal composition and reaction conditions. For example, on Ni surfaces, rapid decomposition of CHCH to (partly) HCC and (mainly) CH, species was reported by Lehwald and Ibach (117) and Stroscio et al. (118). At the same time, on Ru(OOl), the activity of which is intermediate between that of Pt( 111) and Ni(l1l), the whole set of H,CC species, x = 1, 2, or 3, resulting from isomerization (CHCH + CH,C),

146

EVGENY SHUSTOROVICH

dehydrogenation (CHCH + CHC + H), and rehydrogenation (CHCH + H + [CH,CH] -+CH,C) was identified by EELS by Parmeter et al. (81) and Jacob et al. (82). In general, by comparing possible C-C bond-scission routes on Pt versus Ni versus Fe/W (cf. Table XIII), one can easily see that the hydrogen content (x) in the hydrocarbon species C,H, undergoing this scission decreases in the order Fe/W > Ni > Pt, in agreement with vast experimental observations cited by Sinfelt (36). Another model projection is the smallness of most activation barriers for hydrogenation of C,H, species. In general, for all hydrogenation reactions, the calculated barriers decrease in the order Fe/W > Ni > Pt, which explains an increase in the hydrogenation ability of transition metals in the direction Fe/W < Ni < Pt, making Pt overall the best hydrogenation catalyst (119). The above analyses (see also Section IV,A) shed light on the mechanism of the Fischer-Tropsch synthesis of higher hydrocarbons (mostly, alkanes and a-olefins) from CO and H, on transition metal catalysts, typically Fe based (88, 109). For the direct dissociation of CO from the gas phase CO, + C, + 0, on Fe(1 lo), we calculate the very negative activation barrier AEZ,,, = - 11 kcal/mol [see Tables I and I1 and Eq. (21c)l. For CO on Fe(1 I I ) , we similarly find AE$o,g = - 12 kcal/mol, practically indistinguishable from the experimental estimate of - 12 kcal/mol (see Table VI). Thus, the BOC-MP calculations project very easy dissociation of CO on Fe surfaces, leading to high coverage of carbidic carbon. Consistently, CO has been found to spontaneously dissociate on Fe( 111) (86c) and Fe(100) (88d). In the latter case, the Fe(100)-c(2 x 2)C,O structure was formed corresponding to @c,o = 1/2. For such high coverages, the initial values of Qc, Q,, and Qco significantly decrease (18c, 18e) becoming 178, 82, and 20 kcal/mol, respectively (cf. Tables 1 and 2 in Ref. 18e). From Eq. (21c), we now find AE&, = 16 kcal/mol, which is close to the CO dissociation barriers on Pd and Pt surfaces (see Tables VIII and 1X). Consistently, on Fe( 100)-c(2 x 2)C,O only molecular (nondissociative) chemisorptions of CO have been observed with Qco = 20-24 kcal/mol (88d). Thus, we project that hydrogenation of carbidic carbon to form CH, species, polymerization of the CH, species leading to C-C chain growth, and chain termination leading to the products (followed by their desorption) will occur on carbided iron surfaces where the reaction energetics resembles that on metallic Pd or Pt surfaces. Table XI11 clearly shows that the activation barriers for all processes of recombination and desorption are much smaller on Pt than on Fe. Moreover, from Table XI11 it follows that on a pure Fe surface such as Fe(llO), the desorption energies for

BOND-ORDER CONSERVATION

147

both alkanes(C2H,) and olefins (C2H4) are larger than the C-H bond dissociation barriers, which makes it impossible to desorb the hydrocarbon products without serious decomposition. Summing up, the BOC-MP analysis reveals that metallic Fe is necessary to produce the abundance of carbidic carbon from CO, but the Fischer-Tropsch synthesis of hydrocarbons occurs on carbided Fe surfaces. This general model conclusion is in full agreement with diverse experimental data (88, 109). In particular, by using AES and XPS techniques, Krebs et al. (88e) have found that the maximum turnover number (TON) for methane is nor representative of a clean Fe surface but rather of a surface covered by a large amount of carbidic carbon. Some specific features of the Fischer-Tropsch synthesis can also be projected. We see from Table XI11 that of all possible CH,-CH, recombinations, the CH,-CH, one has the smallest barrier. Thus, we predict that C-C chain growth should occur predominantly via CH, insertion into the metal-alkyl bond, in agreement with numerous experimental studies (@a, 88’ 109b). Furthermore, because for all of the metal surfaces the activation barrier for the dehydrogenation C,H5+ C2H4+ H is smaller than that for the hydrogenation C,H, + H =+ C,H,, we project that the primary products of the Fischer-Tropsch synthesis should be a-olefins rather than alkanes, again in agreement with experimental observation (109b). As a related subject, one can add that the earlier version of BOC-MP modeling has been successfully used by Modak and Khanra to elucidate the mechanisms of methanation and Fischer-Tropsch reactions (120a)as well as of alkane conversion (1206) on Ni-Cu alloys; by Sen and Vannice (120c) to study hydrogenation of acetone over platinum catalysts; by Goddard et al. (1204 to make kinetic simulations of ethane hydrogenolysis over Pd, Pt, Ir, and Co; and by Schoofs et al. (120e) to calculate the activation barrier of dissociation of methane on transition metal surfaces.

C . DECOMPOSITION OF HCOOH

ON

TRANSITION-METAL SURFACES

Decomposition of formic acid on transition and posttransition metal surfaces has drawn a great deal of interest in recent years (121-137). The process shows distinct periodic regularities and, therefore, is well suited for the BOC-MP analysis. We want to understand the mechanism of HCOOH decomposition, in particular (1) why formate HCOO is the prevailing intermediate, (2) what is the preferred coordination mode, y ] or y2, for HCOO, and (3) how HCOO decomposes further into CO, and CO. As a periodic series, we shall take that from Ag( 111) to Ni( 111) to Fe/

148

EVGENY SHUSTOROVICH

TABLE XIV Total Bond Energies in the Gas-Phase (0)and Chemisorbed (D + Q ) States on Some Metal Surfaces" Ni

Ag

Species

Coord. type

Dh

Q 152' 80 ' <120' 6d 35 <27 3 34 39 14

H 0 C

-

co

-

11' 11'

257 102 274 384 384/ 384/ 48If

OH HCO CO2 HCOO HCOO HCOOH

-

q2 11' q2 q2

D

+Q

4 2 80 (120 263 137 1301 387 418 423 495

Q

63' 115' 171' 21' 61 50 6 59 71 26

D

Fel W

+ 63 1 I5

171 284 163 324 390 443 455 507

Q

Q

66' 125' 200' 36' 69 65 8 66 81 31

D

+

Q

66 I25 200 293 171 339 392 450 465 512

See text for notations and explanations of the calculated values of Q. The experimental values of Q were used for CO. All energies in kilocalorieslmole. Ref. 25. From Table 1. Ref. 30. ' From Table 11. f See Eq. (30) for the bond energy partitioning used in calculations of Q .

W( 110). The relevant parameters and calculated values of D,Q, and AE* are listed in Tables XIV and XV. The first question is which decomposition route is preferred, via formate or formyl, HCOO, HCOOH,

HCO,

+ H,

+ OH,

Clearly, for all metal surfaces, the formation of formate is much more favorable, both thermodynamically and kinetically, the difference between the two activation barriers increasing from 9 to 13 to 37 kcal/mol along the series Fe/W < Ni < Ag. Qualitatively, this model prediction is in full agreement with the fact that a formate species appears to be a ubiquitous metastable intermediate in the decomposition of HCOOH on Mo (122), Ru (123), Rh ( 1 2 4 , Ni (125), Pd (126), Pt (127), and Cu (128), but aformyl species has never been observed (102a, 129). From Table XIV one can see that the dissociation HCOOH, HCOO, + H, is exothermic on both Fe/W and Ni but is endothermic on Ag, namely AH = - 19, - 11, and 20 kcal/mol, respectively. Accordingly,

-

I49

BOND-ORDER CONSERVATION

~

~

AEp

AI$I

Reaction

Ag

e HCOO, + H, e HCO, + OH, HCOOH, s HCOO, + H,

HCOOH,

97 105

97

HCO, + OH, HCOO, z HCO, + 0, s CO, + O H , S CO?, + H,' CO?, s CO, + 0,' HCO, s CO, + H,

10s 110

@

25 0 127 17

6 43 20 57 42 23 0 44 0

Ni 15 -2 11 24 25 13 5 6 0

-

Fe/W

Ag

NI

-22 -13 9

22 12

0 0 0 0 0 0

x

I6

37 6 22 4 9 5 2

I 2

0 14

23

18

IS

FeIW SO

29 28 16

21 I1 1' 27 22

See text for explanations. All energies in kilocalories/mole. The difference between the gas-phase total bond energies of the reactant and products (see Table X I V ) . I ' Eq. (21c) or (23b). The values of (IA, Q , , and QAH are from Table XIV. Eq. (26b) or (27b). '~Because this reaction in t h e gas phase is practically thermoneutral (see Table XIV), the formal value of D,,, is equal to zero, so that the values of AE7,appear to be underestimated. See text. Compare Table VII for a more accurate treatment via Eqs. (21a) and (24). See text. "

the gas-phase 0-H dissociation barriers AE$H,g are negative on Fe/W and Ni but are positive on Ag (see Table XV). This model conclusion agrees well with experimental observation that on transition metal surfaces, HCOOH readily dissociates at or betow room temperature (102a, 122-128), but on noble posttransition metals such as Ag (77) and Au (78), dissociation of HCOOH requires preadsorption of oxygen (similar to dissociation of CH,OH, as discussed in Section (IV,A). Once HCOO, is formed, it can be, in principle, q2 dicoordinated via both 0 atoms or q' monocoordinated via one 0 atom, that is

-

r)?

7)'

As seen from Table XIV, the q2 mode is always preferred, its relative preference increasing in the series Ag < Ni < Fe/W from 5 to 15 kcall mol. Indeed, on clean metal surfaces, the bridge q2 coordination has

150

EVGENY SHUSTOROVICH

always been observed (122-131). [A rare exception may be Cu(lO0) (132), for which a monodentate formate species has been suggested, coexisting with the bidentate species at certain coverages and temperatures.] At the same time, on precovered metal surfaces, when the q2coordination may be sterically hindered, a monodentate formate species was observed by Avery (133) on O/Pt(lll), by Venkert et al. (134) on O/Ag(llO), and by Madix et al. (125) on C/Ni(llO) [for a general discussion see Ref. (102a)l. Now we wonder how HCOO, decomposes. The first step is to compare the C-0 versus C-H bond cleavage, namely,

t

HCO,

HCOO,

CO,

+ 0,

+ OH,

CO?,

+

(3.0

H,

As seen from Table XV, of the first two reactions relating to the C-0 bond cleavage, the dissociation HCOO, + CO, OH, is preferred over HCOO, .-+ HCO, + 0,, the activation barriers AET being smaller by 10-19 kcal/mol. Even if HCO, is formed, it is expected to decompose spontaneously to CO, and H, [hE&,,,., = 01 on all the metal surfaces. Unfortunately, we cannot reliably calculate the values of AET for the C-H bond cleavage HCOO, + CO,,, + H,, because in the gas phase this dissociation reaction is practically thermoneutral (see Table XIV), which makes the formal value of DCH equal to zero. So, the relevant values of AET in Table XV seem to be underestimated and suited for qualitative comparisons only. Nevertheless, it is fair to conclude that the decomposition HCOO, -+ CO,,, + H, appears to be preferred, but the competitiveness of the other decomposition route HCOO, + CO, + OH, rapidly increases in the order Ag 4 Ni < Fe/W. Consistently, the observed experimental regularity is that CO, is the sole product of HCOOH decomposition over Au(ll0) (78), Ag(llO), Cu(lOO), and Pt(ll1) (102a), whereas CO, along with CO is formed over Ru(OOl), Ni(ll0) (102a), Ni(ll1) (136), Fe(100), and W(100) - (5 x l)C (137), and only CO over Mo(100) ( 1 0 2 ~ ) . At the quantitative level, consider the recombination CO, + OH, + HCOO,, which has been studied by temperature-resolved electron energy loss spectroscopy (TREELS) on Rh( 100) ( 1 0 2 4 . For the experimental value of Q, = 102 kcal/mol (see Table I), we find QOH = 51 kcal/mol and QHcoo = 58 kcal/mol. Then, by using the experimental value of Qco = 32 kcal/mol (see Table VII), from Eq. (26b) we calculate the recombination activation barrier to be AE&.-oH = 10 kcal/mol, in excellent agreement with the experimental estimate of 8 kcallmol (1024. One can add that for CO, dissociation to CO, and O,, we find the activation barrier to rapidly decrease in the periodic manner from 44 kcal/mol on Ag to 1 kcal/mol on Fe/W. This agrees well with the experimental observation that the

+

BOND-ORDER CONSERVATION

151

probability of CO, dissociation on transition metal surfaces has a strong periodic dependence (71, 93, 94) with the gas-phase dissociation barrier of 17 kcal/mol for Rh(l1 I) (93), but with no barrier for Re(001) (94). V. Comparisons with Other Theoretical Techniques Historically, most of the quantum mechanical modeling of chemisorption has been made on metal clusters (8-11, 138). Here, as everywhere in quantum chemistry, for calculations to be sophisticated and, it is hoped, accurate, they should be confined to small enough systems, that is, to small adsorbates and small clusters. But the smaller the cluster, the more indirect are results for simulation of chemisorption on semiinfinite surfaces. There is ongoing debate over the efficiency and accuracy of various quantum mechanical techniques for calculations of surface energetics (8-17, 138-141), and it is not our intent to become embroiled in this heated discussion. As we stressed in Section I, the phenomenological thermodynamic BOC-MP approach is complementary to microscopic quantum mechanical approaches to chemisorption and heterogeneous catalysis. For this reason, our primary concern is the fit of the BOC-MP calculations of Q and AE" with experiment, as we have discussed in Section IV. Still, for a better perspective, it might be informative to give a few examples of how other techniques deal with surface energetics compared with the BOC-MP method. The smallest adsorbates are atoms. For hydrogen, many quantum chemical techniques, both cluster- and band-structure types, produce the values of QHwith the accuracy of a few kilocaloriesimole (141). For other atoms the results may not be that accurate. For example, in the recent ub initio calculations of C/Ni(100) by Chiarello et af. (142), the calculated value of Q, = 293 kcal/mol exceeds the experimental value of 171 kcaVmol(43) by more than 120 kcal/mol. This is especially frustrating because accurate theoretical calculations of Q, for various metal surfaces appear to be the only alternative to alleviate the lack of experimental data on Q,. Anyway, in the BOC-MP approach, the values of QA are simply taken from experiment, so that there is no point for comparison. Of diatomic species, the best-studied molecule is CO, where ab initio cluster-type calculations of Qco are widely used (143). Still, one of the most accurate self-consistent field-configuration interaction (SCF-CI) allelectron calculations by Bagus et af. of the cluster Ni,-CO to mimic COINi(100) gave Qco = 13 kcal/mol (1434, to be compared with the experimental value of 30 kcal/mol (and the BOC-MP estimate of 27 kcall mol). A much simpler and rather popular technique is the atom superposi-

152

EVGENY SHUSTOROVICH

tion electron delocalization molecular orbital (ASED-MO) method (144), where the extended Huckel (EH) binding energies are complemented by some (empirically scaled) repulsive interactions. The ASED-MO calculations by Tomanek and Bennemann of the Ni4-CO cluster to mimic CO/ Ni(ll1) gave Qco = 12 kcal/mol ( 1 4 9 , which is smaller by a factor of 2 than the experimental value of 27 kcal/mol (the BOC-MP estimate is 29 kcal/mol). At the same time, the ASED-MO calculations by Anderson and Awad of the Pd,,-CO cluster to mimic CO/Pd(lll) produced Qco = 65 kcal/mol (144a), which is twice as large as the experimental value of 34 kcal/mol (30). Here, the anisotropy of Qco within the range hollow > bridge > on-top was found to be A Q = 15 kcal/mol(144a), which exceeds the experimental range for Pt-Ni (47-51) by an order of magnitude. The ASED-MO calculations by Mehandru and Anderson of the Cr,,-CO cluster to mimic COICr(ll0) projected Qco = 134 kcal/mol for the parallel geometry and Qco = 90 kcal/mol for the perpendicular geometry (1446). Experimentally, the two geometries coexist at 120 K and differ in energy by not more than I kcal/mol (63). Also, although the experimental value of Qco for Cr(ll0) is not known (owing to easy dissociation of CO), it can be thought to be within the common range of Qco = 25-40 kcal/mol on transition metal surfaces (cf. Table 11), including, in particular, Fe( 110) (5.9, Mo(ll0) (90), and W(110) (30). Similarly, the ASED-MO calculations by Mehandru and Anderson (1444 of Fen ( n = 21-27) clusters to mimic CO on Fe(llO), Fe(100), and Fe(l1 I ) gave Qco = 70,86, and 90 kcal/mol, respectively. The experimental value, known only for Fe( 1 lo), is Qco = 36 kcal/mol(53), which makes the error AQco = 70 - 36 = 34 kcal/mol. In their cluster modeling, Mehandru and Anderson found the q2coordination to be preferred for CO on Fe(100) and Fe( I 1 I), with the q2 versus q 1energy difference AQ,, = 12 and 20 kcal/mol, respectively. Experimentally, the q2and q' orientation states for CO on Fe(100) (64)and Fe( 1 11) (146) were found to be sequentially filled and coexisting, very much similar to the behavior of CO on Cr(llO), where AQco = 1 kcal/mol. For comparison, for CO/Cr(llO) and CO/Fe(llO), the BOC-MP values are Qco = 36-38 kcal/mol (see Table 11) and AQ,, (q2versus 7') = 1-3 kcal/mol (28d). For chemisorption of 0, on Ag, two most recent ah initio calculations by McKee (247)and Upton et al. (148) found 0, to be endothermically chemisorbed, the conclusion being independent of the cluster size, for both Ag, (147) and Agz4(148). Only when special corrections had been made [for the electron affinity of 0, and the ionization energy of Ag, (147) or by assuming bonding in an excited state of Ag24 (148)], the observed exothermicity of 0, chemisorption on Ag surfaces has been reproduced.

BOND-ORDER CONSERVATION

153

For comparison, the BOC-MP method gives straightforwardly Q,, = 10 kcal/mol versus the experimental value of Q,, = 10 kcal/mol (55). Let us move to polyatomic adsorbates such as ethylene and acetylene. Because here calculations for small metal clusters may be particularly inappropriate, we shall consider the results for large clusters and semiinfinite metal slabs. For ethylene on Pt, ( n = 16-19) clusters simulating Pt(l1 l), the ASED-MO calculations by Kang and Anderson ( 1 4 4 ~pro) duced QC2H4= 47 kcal/mol, which is four times as large as the experimental value of 11 kcal/mol (67) and the BOC-MP estimate of 12 kcal/mol. For acetylene, the same calculations gave Qc,H, = 49 kcal/mol, which appears to be strongly overestimated as well (see discussion in Section IV,B). At the same time, the ASED-MO difference A Q = Qc,H2 - Qc,H4 = 2 kcali mol is the same as found in the BOC-MP calculations. For chemisorption of acetylene on Pt(l1 l), within the EH method, the cluster Pt, ( n = 2-11) calculations by Gavezzotti and Simoneta (149) projected very weak binding, with QC,H, = 1-4 kcal/mol, but the metalslab calculations by Silvestre and Hoffmann (150) suggested extremely strong binding, with Qc2H2= 109 kcal/mol, creating the calculated EH range Qc,H, = 1-100 kcal/mol of two orders of magnitude. Strong overestimation of the values of Q for molecular species is rather common, not only for simple techniques such as the EH or ASED-MO methods, but also for sophisticated techniques such as the generalized valence bond-configuration interaction (GVB-CI) method. Consider, for example, the GVB-CI studies by Upton (151) of the decomposition of CH30H on Ni(100), using a 20-atom model for bulk Ni. For an atomic 0, the heat of chemisorption in the fourfold hollow site was found to be Q , = 105 kcal/mol, in reasonable agreement with experimental estimates of 122-129 kcal/mol by Egelhoff (152).But for a hydroxyl OH in the same hollow site, Upton found QOH = 133 kcal/mol(151), larger than the atomic binding energy Q, = 105 kcalhol, which is incredible from the viewpoint of valence theory and inconsistent with the fundamental experimental fact that, for any AB molecule coordinated via A, QA9 QAB(cf. Tables 11 and 111). Although the value of QoH was not reported for Ni, it was measured by Hsu and Lin (153a,b)for Pt( 1 11) by using the laser-induced fluorescence technique, namely, QoH = 36-45 kcal/mol (153a,h).Thus, the GVB-CI value of QoH = 133 kcal/mol for Ni(lO0) appears to be overestimated at least by a factor of 3, that is, by 80-90 kcalimol. Not surprisingly, the GVB-CI calculations projected the decomposition CH,OH, +. CH,., + OH, to be thermodynamically much more favorable than CH,OH, + CH,O, + H, (151), which is inconsistent with experiment when the methoxide formation is overwhelmingly observed (102). For comparison, for

154

EVGENY SHUSTOROVICH

Pt(1 I I ) , the BOC-MP value of QoH = 39 kcallmol (see Table XI) is just in the middle of the experimental range of 36-45 kcal/mol (which, in turn, is much smaller than Qo = 85 kcal/mol). Consistently, the BOC-MP model projected the decomposition CH,OH, -+ CH,O, + H, to be preferred, both kinetically and thermodynamically. The above examples illustrate the point that although the available quantum chemical techniques may be fairly accurate in calculating the atomic QA and some molecular QAB binding energies, more often they are not. The uncertainties and errors in QAB may reach lens of kilocalories. What makes things worse, these errors do not appear to be cancelled out for the relative energies such as the anisotropy AQAB. Experimentally, AQ,, due to the metal adsorption site (on-top versus bridge versus hollow) or the adsorbate coordination mode (7'versus 7') does not exceed 1 kcal/ mol for various transition metal surfaces (47-51, 63, 6 4 ) . However, these tiny effects within AQco < 1 kcal/mol were interpreted and projected by the calculations (144) wherein the values of Qco erred by 30-50 kcal/mol and the projected anisotropy was AQco = 15-45 kcallmol. The activation barriers AE* for dissociation and recombination belong to the same realm of relative energies as A&,. For this reason, we shall not discuss here purely numerical calculations of AE*. Remarkably, many authors tried to conceptualize their computational results in terms of simple analytic models, which have no direct relation to the computations. For example, the effective medium theory (EMT) is a band-structure model with a complex and elaborated formalism including many parameters (154). Nevertheless, while reviewing the numerical EMT applications to surface reactions, Norskov and Stoltze (155) discussed the calculated trends in the activation energies for AB dissociation in terms of a oneparameter model (unfortunately, no details were provided) projecting AEiB to vary as Nd(10 - N d ) ,where Nd is the d band occupancy [cf. Eqs. (21a)-(21c) of the BOC-MP theory]. As far as phenomenological modeling is concerned, an excellent review of earlier thermodynamic approaches to chemisorption and surface reactivity was given by Benziger (156), who also developed some general thermodynamic criteria for dissociative versus nondissociative adsorption of diatomic and polyatomic molecules on transition metal surfaces ( I 37, 156). In particular, for quantitative estimates of QA, A = C, N, or 0, Benziger (156) used the heats of formation of bulk metal carbides, nitrides, and oxides. The BOC-MP approach is different, however, not only analytically but also in making direct use of experimental values of QA. Finally, one should mention that some version of bond-order conservation, known as the bond-energylbond-order (BEBO) method, has been applied to chemisorption bonding and surface reactivity by Weinberg and

BON D-ORDER CONSERVATION

155

Merrill (157). The basic assumptions and mathematical formalism of the BOC-MP and BEBO methods are quite different, however. Most important, in the BEBO method, following Lewis and Pauling, the bond order x is defined as the number of shared electron pairs, so that x may be smaller than, equal to, or larger than unity, reflecting fractional, single, o r multiple A-B bonding, respectively. Furthermore, the BEBO method makes use of the power function E(x) = - Q, x where p is some empirical constant. J’,

VI. The BOC-MP Model: Comments and Summary The major strength of the BOC-MP model is that it is an analytic model, based on a few well-defined assumptions and producing straightforward and falsifiable (in the Popperian sense) results. The crucial model assumption is the BOC at unity in its pairwise additive form [Eq. (6)]. For various linear three-center A B C interactions, such a form (xAB+ xBc = 1) was already assumed (158) and shown to be very accurate, both computationally (159) and experimentally (160). Most recently, Van Santen has shown (161) that the BOC at unity is equivalent to the normalization condition of the covalent resonance structures within the valence bond theory. In a sense, we simply postulate the similar BOC for many-center M,-A (spherical) interactions. Furthermore, Morse potentials were chosen not just because of their simplicity but because, within the BOC framework, they are well suited to describing the energetics of chemisorption. The reason here is that the zero-energy gap between the occupied and vacant parts of the metal band eliminates the repulsive terms in the chemisorptive M,- ,-A-M (n z 1) interactions, making them always attractive (18g). But such an attractive M,-A interaction can always be described by an effective Morse potential when the total A-M, bond order is conserved at unity (18g). Finally, the assumed confinement of n in the chemisorbed M,-A site to one unit mesh M, reflects the known efficiency of the nearest-neighbor approximation in many problems of metallic binding (21a). In the zero-coverage extreme, the BOC-MP interrelations are exact for atomic adsorbates and well defined for molecular adsorbates, the same analytic formalism being used to treat both diatomic and polyatomic molecules. Moreover, these interrelations are expressed in terms of observables only (the heats of chemisorption and various constants), which makes comparison with experiment direct and unambiguous. With rare exceptions, the agreement with experiment is remarkably good, qualitatively and quantitatively. Of course, with such general and uniform modeling, some of the calcu0 . .

3..

156

EVGENY SHUSTOROVICH

lated numbers may be not satisfactory. For example, both Eqs. (10a) and (14) strongly overestimate QN, for N,-Fe (cf. Table 11) and several activation barriers in Tables IX, X, XII, and XV (e.g., for the decomposition of CH, OCH,, or HCOO) apparently beg for improvement. One should stress, however, that in surface science the perfect theoretical fit to experimental numbers may be not only impossible but also undesirable (162).Much too often, the experimental data are subject to change because of better control of many unsuspected or neglected factors such as surface defects, impurities, coverage effects, postulated Arrhenius activation barriers and preexponentials, etc. Also, various complementary techniques usually do not probe and measure the same characteristics. For these reasons, what we wanted to achieve was the qualitative consistency of model projections with diverse and representative experiment, of course with the acceptable numerical accuracy. The BOC-MP model is a simple, truly “back-of-the-envelope’’ model that can be directly used by practitioners in the field. It efficiently describes and interrelates a wide variety of chemisorption phenomena. Most important, the model maps out metal surface reactions providing insight into both regularities and details. We have considered a number of examples of different complexity. In principle, any metal surface reaction can be treated this way. The only requirement is to retain the rigor and simplicity of the model projections. As John von Neumann put it, if a style, classical in the beginning, turns to resemble baroque, this is a sign of danger.

VII. Conclusion The theorists are incurable optimists in their belief that there exists a comprehensible order of things. The only problem is to find a framework within which the intricacies of real phenomena are coherently interrelated. For chemisorption phenomena on transition metal surfaces, including surface reactivity, the BOC-MP model appears to provide such a framework. ACKNOWLEDGMENTS

This work was begun during my apprenticeship with the late Earl L. Muetterties. 1 am indebted to many of my colleagues and friends for sharing their knowledge and vision with me. Here my special thanks are to John T . Yates, Jr., Gerhard Ertl, and Michel Boudart, who have been the most sympathetic listeners and incisive critics of my work, and to Alexis T. Bell for the sheer scientific pleasure of our collaboration. REFERENCES 1 . Roth, J. F., in “Catalysis 1987, Studies in Surface Science and Catalysis” (J. W. Ward,

ed.), Vol. 38, p. 925. Elsevier, Amsterdam, 1988. 2. Goodman, D. W., and Houston, J . E., Science 236, 403 (1987).

BOND-ORDER CONSERVATION

157

3. (a) Briggs, D., Surf. Sci. 189/190,801 (1987);(b) Sinfelt, J. H., J . Phys. Chem. 90,471 I ( 1986). 4. (a) Mills, G. A., in Ref. I , p. 905; (b) McHenry, K. W., in Ref. 1, p. 917.

5. Boudart, M., and Djega-Mariadassou, G., “Kinetics of Heterogeneous Catalytic Reactions.” Princeton Univ. Press, Princeton, New Jersey, 1984. 6. Hehre, W. J., Radom, L., Schleyer, P. R., and Pople, J. A , , “Ab Initio Molecular Orbital Theory.” Wiley, New York, 1986. 7. (a) Truhlar, D. G., and Steckler, R., Chem. Rev. 87, 217 (1987); (b) Dunning, T. H., Jr., Harding, L. B., Wagner, A. F., Schatz, G. C., and Bowman, J. M., Science 240, 453 (1988). 8. “The Nature of the Surface Chemical Bond” (T. N. Rhodin and G . Ertl, eds.). NorthHolland Publ., Amsterdam, 1979. 9. “Theory of Chemisorption” (J. R. Smith, ed.). Springer-Verlag, Berlin, 1980. 10. “Quantum Chemistry: The Challenge of Transition Metals and Coordination Chemistry” (A. Veillard, ed.). NATO AS1 Series. Reidel, Dordrecht, Netherlands, 1986. 11. (a) Messmer, R. P., Surf. Sci. 158, 40 (1985); (b) Smith, J. R., Arlinghaus, F. I., and Gay, J. G., J . Vuc. Sci. Technol. 18, 411 (1981). 12. (a) Mattson, A , , Panas, I., Siegbahn, P., Wahlgren, U., and H. Akeby, Phys. Rev. B 36,7389 (1987); (b) Panas, I., Schule, J., Siegbahn, P., and Wahlgren, U., Chem. Phys. Lett. 149,265 (1988); ( c )Panas, I., Siegbahn, P., and Wahlgren, U., Theor. Chim. Acta 74, 167 (1988); (d) Panas, I., Schule, J., Brandenmark, U., Siegbahn, P., and Wahlgren, U., J . Phys. Chem. 92, 3079 (1988): (e) Blomberg, M. R. A , , Lebrilla, C. B., and Siegbahn, P. E. M., Chem. Phys. Lett. 150, 522 (1988). 13. Freeman, A. J.. Fu, C. L., and Wimmer, E., J . Vac. Sci. Technol. A4, 1265 (1986). 14. Feibelman, P. J., Phys. Rev. B 35, 2626 (1987). 15. Shustorovich, E., Baetzold, R. C., and Muetterties, E. L . , J . Phys. Chem. 87, 1100 (1 983). 16. Shustorovich, E., Suyf. Sci. Rep. 6, 1 (1986). 17. Shustorovich, E., Acc.Chem. Res. 21, 183 (1988). 18. (a) Shustorovich, E., J . A m . Chem. Soc. 106,6479 (1984): (b) Shustorovich, E., Surf. Sci. 150, LI 15 (1985);(c)Shustorovich, E., Surf. Sci.163, L645(1985);(d) Shustorovich, E., Surf. Sci. 163, L730 (1985); (e) Shustorovich, E., Surf. Sci. 175, 561 (1986); (f) Shustorovich, E., Surf. Sci. 176, L863 (1986);(g) Shustorovich, E., Surf. Sci. 181, L205 (1987); (h) Shustorovich, E., Surf. Sci. 205, 336 (1988); (i) Shustorovich, E., J . Molec. Catul. 54, 301 (1989). 19. (a) Shustorovich, E., and Bell, A. T., J . Curd. 113, 341 (1988); (b) Shustorovich, E., and Bell, A . T., Surf. Sci. 205, 492 (1988); (c) Shustorovich, E., and Bell, A. T . , J . Cutal. 121, 1 (1990): (d) Shustorovich, E., and Bell, A. T., Surf. Sci. 222, 371 (1989); (e) Shustorovich, E., and Bell, A. T., Su$ Sci. in press. 20. Somorjai, G. A . , “Chemistry in Two Dimensions: Surfaces.” Cornell Univ. Press, Ithaca, New York, 1981. 21. (a) Abel, G. C., Phys. Rev. B 31,6184 (1985); (b) Skriver, H. L., Phys. Rev. B31, 1909 (1985); (c) Haydock, R., Philos. Mug. 35, 845 (1977); 38, 155 (1978). 22. (a) Rose, J. H., Smith, J. R., and Ferrante, J . , Phys. Reu. B 28, 1835 (1983); (b) Rose, J. H., Smith, J. R., Guinea, F., and Ferrante, J., Phys. Rev. B 29, 2963 (1984); (c) Ferrante, J . , and Smith, J. R., Phys. Reu. B 31, 3427 (1985). 23. Morse, P. M., Phys. Rev. 34, 57 (1929). 24. See, e.g., (a) Harris, J., Holloway, S., Rahman, T. S . , and Yang, K., J . Chem. Phys. 89,4427 (1988); (b) Luntz, A. C., Williams, M. D., and Bethune, D. C., J . Chem. Phys. 89,4381 (1988); (c) Asscher, M., Becker, 0.M . , Haase, G., and Kosloff, R., Surf. Sci. 206, L880 (1988).

158

EVGENY SHUSTOROVICH

25. “CRC Handbook of Chemistry and Physics,” pp. F171-190. CRC Press, Boca Raton, Florida, 1984- 1985. 26. (a) Lathan, W. A., Curtis, L. A., Hehre, W. J . , Lisle, J . B., and Pople, J. A,, Progr. Phys. Org. Chem. 11, 175 (1974); (b) Schatz, G . C., Fitzcharles, M. S . , and Harding, L. B., Furuday Discuss. Chem. Soc. 84, 359 (1987); (c) Ruscis, B., Schwarz, M., and Berkowitz. J., J . Chem. Phys. 91, 6780 (1989). 27. Shiromaru, H., Achiba, Y . , Kimura, K . , and Lee, Y. T., J . Phys. Chem. 91, 17 (1987). 28. (a) McKee, M. L., and Shevlin, P. B., J . Am. Chem. Soc. 107, 5191 (1985): (b) Carrington, T., Jr., Hubbard, L. M., Schaeffer, H. T., 111, and Miller, W . H., J . Chem. Phys. 80, 4347 (1984). 29. Cotton, F. A., and Wilkinson, G . , “Advanced Inorganic Chemistry,” 3rd Ed., pp. 106-107. Wiley (Interscience), New York, 1972. 30. Ertl, G., in Ref. 8, Chap. 5, Tables 5.1-5.5. 31. Feulner, P., and Menzel, D., Surf. Sci. 154, 465 (1985). 32. (a) Yates, J. T., Jr., Thiel, P. A.. and Weinberg, W . H., Surf. Sci. 84, 427 (1979); (b) Ehsasi, M., and Christmann, K . , Surf. Sci. 194, 172 (1988). 33. Engstrom, J . R., Tsai, W., and Weinberg, W. H., J . Chem. Phys. 87, 3104 (1987). 34. Gdowski, G. E., Fair, J. A., and Madix, R. J., Surf. Sci. 127, 541 (1983). 35. (a) Zhou, X.-L., White, J. M., and Koel, B. E., Surf. Sci. 218, 201 (1989); (b) Sault, A. G., Madix, R. J., and Campbell, C. T., Surf. Sci. 169, 347 (1986). 36. Engel, T., Niehus, H., and Bauer, E . , Surf. Sci. 52, 237 (1975). 37. Pirner, M., Bauer, R., Borgmann, D., and Wedler, G., Surf. Sci. 189/190, 147 (1987). 38. Surnev, L . , Rangelov, G., and Bliznakov, G., Surf. Sci. 159, 299 (1985). 39. Fisher, G. B . , and Schmeig, S. J., J . Vuc. Sci. Techno/. A l , 1064 (1983). 40. Brennan, D., Hayward, D. O., and Trapnell, B. M. W., Proc. R . Soc. London, Ser. A 265, 81 (1960). 41. Giamello, E., Fubini, B., Lauro, P., and Bossi, A . , J . Cutul. 87, 443(1984). 42. Vajo, J. J., Tsai, W., and Weinberg, W. H., J . P h y s . Chem. 89, 3243 (1985). 43. Isett, L. C . , and Blakely, J . M., Surf. Sci. 43, 493 (1974). 44. Hoffmann, F. M., Surf: Sci. R e p . 3, 107 (1983). 45. Biberian, J . P., and Van Hove, M. A., Surf. Sci. 118, 443 (1982); 138, 361 (1984). 46. Ishi, S., Ohno, Y . , and Viswanathan, B., Surf. Sci. 161, 349 (1985). 47. (a) Hayden, B. E., and Bradshaw, A . M., Surf. Sci. 125,787 (1983): (b) Mieher, W. D., Whitman, L. J., and Ho, W., J . Chem. Phys. 91, 3228 (1989). 48. Gurney, B . A . , Richter, L. J., Villarubia, J. S . , and Ho, W., J . Chejn. Phys. 87, 6710 (1987). 4 9 . Andersson, S., Solid State Commun. 21, 75 (1977). 50. Bauhofer, J . , Hock, M., and Kuppers, J., Surf. Sci. 191, 395 (1987). 51. (a) Tang, S. L., Lee, M. B., Yang, Q. Y . , Beckerle, J. D., and Ceyer, S . T., J . Chem. Phvs. 84, 1876 (1986); (b) Surnev, L., Xu, Z., and Yates, J. T., Jr., Surf. Sci. 201, 1 , 14 (1988). 52. Raval, R., Parker, S. F., Pemble, M. E., Hollins, P., Pritchard, J . , and Chesters, M. A., Surf. Sci. 203, 353 (1988). 53. Wedler, G., and Ruhmann, H., Surf: Sci. 121, 464 (1982). 54. (a) Gland, J . L., Sexton, B. A . , and Fisher, G., Surf Sci. 95,587 (1980); (b) Campbell, C. T., Ertl, G., Kuipers, H., and Segner, J., Surf. Sci. 107, 220 (1981). 55. (a) Barteau, M. A., and Madix, R. J., Surf. Sci. 97, 101 (1980); (b) Campbell, C.T., Surf. Sci. 157, 43 (1985). 56. Tsai, W., Vajo, J. J., and Weinberg, W. H . , J . Phys. Chem. 89, 4926 (1985). 57. Klauber, C., Alvey, M. D., and Yates, J. T., Jr., Surf. Sci. 154, 139 (1985).

BOND-ORDER CONSERVATION

58. 59. 60. 61. 62. 63. 64.

65. 66.

67.

68. 69.

70. 71. 72. 73. 74.

75.

76. 77. 78. 79. 80. 81.

82. 83.

159

Fisher, G. B., and Gland, J. L., Surf. Sci. 94,446 (1980). Thiel, P. A., and Madey, T. E., Surf. Sci. Rep. 7, 21 1 , 262 (1987). Sexton, B. A., and Hughes, A. E., Surf.Sci. 140, 227 (1984). Solimosi, F., Berko, A., and Tarnoczi, T. I., Surf. Sci. 141, 533 (1984). Hall, R. B., J . Phys. Chem. 91, 1007 (1987). Shinn, N . D., and Madey, T. E., J. Chem. Phys. 83, 5928 (1985). (a) Benndorf, C., Kriiger, B., and Thieme, F., Surf. Sci. 163, L675 (1985); (b) Moon, D. W., Bernasek, S . L., Lu, J. P., Gland, J. L., and Dwyer, D. J., Surf. Sci. 184, 90 (1987). Fulrner, .I. P., Zaera, F., and Tysoe, W. T., J . Chem. Phys. 87, 7265 (1987). (a) Nieuwenhuys, B. E., Surf. Sci. 105, 505 (1981); (b) Griinze, M. J., Fuhler, J., Neumann, M., Brundle, C. R., Auerbach, D. J., and Behm, J., Surf. Sci. 139, 109 (1984); (c) Brubaker, M. E., and Trenary, M., J. Chem. Phys. 85, 6100 (1986); (d) Griinze, M . , Golze, M., Hirschwald, W., Freund, H.-J., Pulm, H., Seip, U . , Tsai, M. C., Ertl, G., and Kiippers, J., Phys. Reu. Lett. 53,850 (1984); (e) Freund, H.-J., Bartos, B., Messrner, R. B., Griinze, M., Kuhlenbeck, H., and Neumann, M., Surf. Sci. 185, 187 (1987). (a) Salmeron, M., and Somorjai, G. A., J . Phvs. Chem. 86, 341 (1982); (b) Windham, R. G., Bartram, M. E., and Koel, B. E., J. Phys. Chem. 92 (1988) 2862; (c) Mohsin, S . B . , Trenary, M., and Robota, H., Chem. Phys. Lett. 154 (1989) 511; (d) PettietteHall, C. L., Land, D. P., McIver, R. T., Jr., and Hemrninger, J. C., J . Phys. Chem. 94 ( 1990) 1948. Tysoe, W. T., Nyberg, G. L., and Lambert, R. M., J. Phys. Chem. 88, 1960 (1984). Hills, M. M., Parmenter, J. E., Mullins, C. B., and Weinberg, W. H., J . A m . Chem. Soc. 108, 3554 (1986). Zuhr, R. A,, and Hudson, J. B., Surf. Sci. 66, 405 (1977). (a) Asscher, M., Kao, C.-T., and Somorjai, G. A,, J. Phys. Chem. 92,2711 (1988); (b) Kwong, D. W. J., DeLeon, N., and Haller, G. L . , Chem. Phys. L e f t . 144, 533 (1988). (a) Weinberg, W. H., Surf. Sci. 128, L224 (1983); (b) Matsushima, T., Surf. Sci. 127, 403 (1983); (c) Norton, P. R., Surf. Sci. 44, 624 (1974). Hadden, R. A,, Vandervell, H. D., Waugh, K. C., and Webb, G., Proc. Inr. Conf. C a d . , 9th, Ottawa 4, 1835 (1988). (a) Schule, J., Siegbahn, P., and Wahlgren, U., J . Chem. Phys. 89, 6982 (1988); (b) Siegbahn, P., and Wahlgren, U., in “Reaction Energetics on Metal Surfaces: Theory and Applications,” (E. Shustorovich, ed.), Chap. I , VCH, New York, 1990. (a) Anton, A. B., Avergy, N. B., Toby, B. H., and Weinberg, W. H., J . A m . Chem. Soc. 108, 684 (1986); (b) Anton, A. B., Parmeter, J. E., and Weinberg, W. H., J . Am. Chem. Soc. 108, 1823 (1986); (c) Davis, J. L., and Barteau, M. A., Surf.Sci. 208,383 (1989). Avery, N. R., Weinberg, W. H., Anton, A. B., and Toby, B. H., Phys. Rev. L e t f . 51, 682 (1983). Barteau, M. A,, Bowker, M., and Madix, R. J., Surf. Sci. 94, 303 (1980). (a) Outka, D. A., and Madix, R. J., Surf. Sci. 179, 361 (1987); (b) Outka, D. A., and Madix, R. J., J . Am. Chem. Soc. 109, 1708 (1987). Kesmodel, L. L., Waddill, G. D., and Gates, J. A., Surf. Sci. 138, 464 (1984). Marinova, T. S . , and Kostov, K. L., Surf. Sci. 181, 573 (1987). Parmeter, J. E., Hills, M. M., and Weinberg, W. H., J . Am. Chem. SOC. 108, 3563 (1986); 109, 72 (1987). Jacob, P., Cassuto, A., and Menzel, D., Surf. Sci. 187, 407 (1987). Hills, M. M., Parmeter, J. E., and Weinberg, W. H., J . Am. Chem. SOC.109,597 (1987).

I60

EVGENY SHUSI‘OROVICH

8 4 . lbach, H., and Lehwald, S., J . Vtrc. S c i . Techno/. 15. 407 (1978). 8 5 . (a) Wang, P.-K., Slichter, C . P.. and Sinfelt, J. H., I’lrvs. Rcv. L e t / . 53, 82 (1984): ( b ) Wang. P . - K . , Slichter, C. P., and Sinfelt, J . H., J . Pl7y.s. Clicrti. 89, 3606 (1985); (c) Wang, P.-K., Anserniet, J.-P., Kudaz, S. L., Wang, Z.. Shore. S . , Slichter, C. P., and Sinfelt, J. H., Scienc~c~ 234, 35 (1986). 86. (a) Campbell, C. T., and Goodman, D. W . . Sirrj: Sci. 123,413 (1982); (b) Urnbach, E., and Menzel. D., S N I : ~Sci. : 135, 199 (1983): ( c ) Whitman, L. J.. Richter, 1,. J., Gurnery, B. A., Villarruhia, J . S., and Ho, W., J . CIicvn. Pliv.~.90, 2050 (1989); (d) Semancik. S., and Estrup, P. J., S q f i Sc,i. 104, 261 (1981); (e) Astaldi, C . , Santori, A,, Della Valle, F . , and Rosei, K., S r r t f Sci. 220, 322 (1989); (f) Mesters, C. M. A. M . . Vink. 7’. J., Gijzenian. 0. I..J., and Gem. J . W., S r r & Sc,i, 135, 428 (19x3); (g) Campbell, C. T., Book c$Ah.s/rwrs, 1YYAC.S N i t / / .M t ~ r / i n pBo.v/m. , April 22-27, 1990, Coll. Div., paper 120. 87. (a) Lee, M. B., Beckerle, J . I>,, Tang, S. L., and Ceyer, S. T., J . C h ~ iPliys. . 87, 723 (1987);(b) Steinbruck. H. P., D’Evelyn, M. P., and Madix, K. J . . Siirf: Sci. 172, L56l (1986). 88. (a) Biloen. P., and Sachtler. W. M . H., A h . Cutrrl. 30, 165 (1981); (b) Rofer-DePoorter. C. K . , C h i n . Rev. 81,447 (1981); ( c ) Broden, G . , Rhodin, T. N . , Bruker, C . , Benbow, K.,and Hurych, Z., S I I I : ~Sci. : 59, 593 (1976); (d) Vink, T. J.. GQzeman, 0. L. J . , and Geus, J. W.. Siirf: S c i . 150, (1985) 14; ( e ) Krebs, H . J . . Bonzel. H. P.. and Gafner, G . , S w f : Sci. 88, (1979) 269; ( f ) Brady, R.C.. and Pettit, K.,J . A m . Thc~m. Sot,. 102 (1980) 6181; 103 (1981) 1287. 89. See, e . g . . (a) Dunning, T . H . , J r . , Harding. I,.B., Bair, K. A , , Eades, K. A,. and Shepard, K. I,.. J . P1iy.s. Clictn. 90, 344 (1986): (b) Leroy, G . , and Sana. M . . 1.M o l . .S/r/rcf.(TIiroc,licrn.)136, 283 (1986):(c) Burgi. H . - B . , and Dunitz, J . D., J . A m . Ckem. S o c . 109, 2924 ( 1987). YO. Panas, I.. Sieghahn, P.. and Wahlgren, U . , J . C h o n . Pliys. 90, 6791 (1989). 91. Ertl, G . , in “Catalysis: Science and Technology” (J. R. Anderson and M . Boudart, eds.), Vol. 4. Chap. 3. Springer-Verlag. Berlin. 1983. 92. Thiel, P. A., Williams. E. D.. Yates. J . T., Jr., and Weinberg. W. H.. Surf: S c i . 84, 44 (1979). 93. Goodman, D. W . . Peebles, D. E.. and White, J. M., S w f : Sci. 140. L239 (1984). 94. Peled. H.. and Asscher, M . , S w f . .Qi. 183, 201 (1987). 95. Bowker, M . , Barteau. M. A , , and Madix, R. J., S w f : Sci. 92, 528 (1980). 96. Engstrom, J. R., and Weinberg, W . H., Plivs. R P U .L e t t . 55, 2017 (1985). 97. Cornrie. C. M . , Weinberg, W . H.. and Lambert, R. M . . S w f . Sci. 57, 619 (1976). 98. Root, T. W.. Schmidt, L. D.. and Fisher, G . B., Sirrf: S c i . 134, 30 (1983). YY. Ho, P., and White, J. M., Surf: Sci. 137, 103 (19x4). 100. Beebe, T. P . . Goodman, D. W.. Kay, B. D., and Yates. J. T.. Jr.. J . C l i ~ w tPhvs. . 87, 2305 (1987). 101. (a) Ceyer. S. T.. Beckerle, J. D., Lee. M . B., Tang, S. L . . Yang, Q.Y., and Himes, M . A . , J . Vat. Sci. T d i n o l . AS, 501 (1987);(b) Hamza. A. V., and Madix. K. J . . Sio:f. Sci. 179, 25 (1987). 102. (a) Canning. N . 11. S., and Madix. R . J . . J . Phvs. Chcvn. 88, 2437 (1984); (b) Hrbek. J., DePaola. R.,and Hoffmann, F. M.. Sirr:f: Sci. 166,361 (1986);(c)Levis, K . J., Zhicheng, J., and Winograd. N.. J . Am. C l i w ~Sot,. . 110, 4431 (1988); 111, 4605 (1989): ( d ) Ho. W.. J . Phvs. Clirm. 91, 766 (1987); (e) Bhattacharya. A. K.. Chester, M . A,, Pernhle. M . E., and Sheppard, N . , Surf: Sci. 206, L845 (1988); (f) Kruse, N., Chuah, G.-K.. Abend. G . , Cocke. D. L., and Block, J. H., Surf.Sci. 189/190,832 (1987);(g) Guo, X . , Hanley. L.. and Yates, J . T.. Jr., J . A m . Chrm. Soc. 111, 3155 (1989); ( h ) Lu, J.-P..

BOND-ORDER CONSERVATION

161

Albert, M. A,. Bernasek, S. L., and Dwyer, D. S., Surf: S c i . 218, 1 (1989); 221, 197 (1989). 103. See discussion in (a) Stuve, E. M.. and Madix, R. J., J. Phys. Chem. 89, 105 (1985); (b) Henderson, M. A., Mitchell, G. E.. and White. J. M., Sit$ Sci.203, 378 (1988). 104. (a) Beebe. T. B., Jr., and Yates, J. T.. Jr., J . Phys. Chem. 91, 254 (1987); (b) Zhu, X.-Y., and White, J. M., Su$ Sci. 214, 240 (1989); (c) Lapinski, M. P., and Ekerdt, J. G., J. Phys. Chem. 92, 1708 (1988);(d) Anderson, K. G.. and Ekerdt, J. G., J . Cutul. 116, 556 (1989). 105. (a) Ogle, K. M.. Creighton, J. R., Akhter, S., and White, J. M., Surf. Sci. 169, 246 (1986); (b) Greenlief, C. M.. Radloff, P. L., Zhou, X.-L., and White, J. M., Surf Sci. 191, 93 (1987). 106. Satterfield, C. N., “Heterogeneous Catalysis in Practice.” McGraw-Hill, New York, 1979. 107. Seebauer. E. G., Kong, A . C. F., and Schmidt, L. D., Siqf. Sci. 193, 417 (1988). 108. See, e.g., (a) Vannice, M. A., and Twu. C. C., J. Cum/. 18,213 (1983); (b) Stoltze, P., and Norskov, J. K., J. C u t d . 110, I (1988). 109. (a) Ponec, V., Crrtul. R e v . 18, IS1 (1978); (b) Bell, A. T., Cutul. Rev. 23, 23 (1981). 110. (a) Goodman, D. W., A N . Chem. Res. 17, 194 (1984); (b) Goodman, D. W., Kelly, R. D., Madey. T. E., and White, J. M., J . Cutd. 64,479 (1980);(c) Yates, J. T., Gates, S . M., and Russell, J. M.. Jr., Surf. Sci. 164, L839 (1985); (d) Glugla, P. G., Bailey, K. M., and Falkoner, J. L., J. Cutul. 115, 24 (1989). 111. (a) Poutsma, M. L., Elek, L. F.. Ibaria, P. A , . Risch. A . P., and Rabo, J. A , , J. Curd. 52, 168 11978); (b) Ryndin, Y. A . , Hicks. R. F.. Bell, A . T., and Yermakov, Y. 1.. J. Cutul. 70,287 (1981); (c)Kikuzono. Y., Kagami, S.. Naito, S., Onishi, T., and Tamaru. K., Discuss. Fwuduy Soc. 72,735 (1981 ); (d) Fajula, F., Anthony, R. G., and Lansford, J . H . . J . Cntul. 73,237 (1982);(e) Vannice, M. A., J . Ctrtol. 50,228 (1977);(f) Berlowitz, P. J., and Goodman, D. W., J. Curd. 108, 364 (1987). 112. (a) Hoffmann, F. M., and Robbins, J . L., J. Electron Spec.trosc. ReIut. Phcnom. 45, 421 (1987); (b) Hoffmann. F. M., and Robbins, J. L., Proc. Inr. Congr. Curd. Yth. Ottuwu 3, 1144; 5, 373 (1988). 113. (a) Ho, S. V., and Harriott. P., J. C u t d . 64,272 (1980); (b) Wang, S.-Y., Moon, S. H., and Vannice, M. A., J. Cutul. 71, I67 (1981);(c) Rieck. J . S.. and Bell, A. T., J. C u r d . 96,88 (198.5);99,262 (1986); (d) Mori, T.. Miyamoto. A., Niizuma, H., Takahashi, N., Hattori. T., and Murakami, Y., J . Phys. Chem. 90, 109 (1986). and references therein. 114. See, e.g., Davis, S. M.. and Somorjai, G. A , , in “The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis” (D.A . King and D. P. Woodruff, eds.). Vol. 4, p. 217. Elsevier, New York, 1982. 115. (a) Avery, N. K., Lungrniiir 4,445 (1988); (b) Carter, E. A., and Koel, B. E., S w f . Sci. 226, 339 (1990). 116. (a) Erley, W., Baro, A. M., and Ibach, H., Surf: Sci. 120,273 (1982);(b) Seip, U., Tsai, M.-C., Kiippers, J., and Ertl. G., Surf. Sci. 147, 6.5 (1984). 117. Lehwald. S . , and Ibach, H., Srtrj: Sci. 120, 273 (1982). 118. Stroscio, J . A., Bare, S. R.. and Ho, W.. Surf. S c i . 148, 499 (1984). 119. See. e.g., Peterson, R. J., “Hydrogenation Catalysts.” Noyes Data Corp.. Park Ridge, New Jersey. 1977. 120. (a) Khanra, B. C., and Modak, S., Chcm. Phys. L e u . 143, 390 (1988); (b) Modak, S . , and Khanra, B. C., Surf.Sci. 197,361 (1988); (c) Sen. B., and Vannice, M. A , , J. Curd. 113,52 (1988);(d) Goodard. S. A., Amiridis, M. 0 . . Rekoske, J. E., Cardona-Martinez. N.. and Dumesic, J. A . , J . Cutrrl. 117, I55 (1989); (e) Schoofs, G. R., Arumainayagam, C. R., McMaster, M. C., and Madix, R. J., Surf. Sci. 215, I (1989).

I62

EVGEN Y SH USTOROVICH

121. Madix. R. J.. Adu. Ctrrul. 29, I (19x0). 122. Miles, S. L . . Bernasek, S . L., and Gland, J. L., S w f : Sci. 127, 271 (19x3). 123. Avery, N. K..Toby. €3. H.. Anton, A. B., and Weinberg, W. H., Srrrf: .Sc,i. 122, L574 (1982). 124. Solimosi, F . , Kiss, J . , and Kovacs. I . , J . Phvs. Clic~n.92, 796 (1988). 125. Madix, K. J . . Gland, J. B . , Mitchell, G . E., and Sexton. B. A , , Srrrf:St,i. 125,481 (198.3). 126. Egawa, C.. Doi, I . , Naito, S . . and l’amaru, K., Srr(ff: Sc,i. 176, 491 (1986). 127. Hoffmann, P., Bare, S . R., Richardson. N . V.. and King, D. A., Surf: Sc,i. 133, L459 (198.3).

128. (a) Hayden. B. E., Prince, K., Woodruff, 11. P . , and Bradshaw, A. M., Sin 589 (19x3); (b) Iglesia, E., and Boudart, M.. J . Ctirtrl. 81, 214 (1983). 129. See discussion in Crowell, J. E., Chen. J . G., and Yates, J. T., Jr., J . Clicvn. Plrys. 85, 3111 (1986). 130. (a) Puschmann, A , , Haase. J., Crapper, M. D., Riley, C . E . , and Woodruff, D. P., Phys. Rev. Lert. 54, 2250 (1985); (b) Crapper, M. D., Riley. C. E., Woodruff, D. P.. Puschmann, A , , and Haase, J . , Srrr:f:Sci. 171, I (1986). 131. Jones, T. S . . Richardson, N. V.. and Joshi, A. W . . Sin:/: S(.i. 207, L948 (19x8). 132. (21) Sexton. B. A., Srrr:f: Sci. 88, 319 (1979);( b ) Dubois, L. H.. Ellis, T. H.. Zegarski, B . R., and Kevan, S. D.. Srrrf: Sci. 172, 385 (1986). 133. Avery, N. K., Appl. Srrr:f: Sc.i. 11/12, 774 ( 1982). 134. Venkert, A., Ihriel, M. P.. and Talianker, M., J . Lo.ss-Cornrnon Mct. 103, 361 (1984). 13.7. Jorgensen, S . W., and Madix, K. J . , Srrrf: S(.i. 183, 27 (1987). 136. (a) Benziger. J . B., rind Schoofs, G. K., J . Phys. Cl7crn. 88, 4439 ( 1984); (b) Erley. W., and Sander, D., J . Vcrc. Sci. T d i n o l . A7, 2238 (1989). 137. Benziger. J. B . , and Madix, R. J., J . Ctrrtrl. 74, 67 (1982). fM. Shustorovich, E., in Ref. 10, pp. 445-464. 139. Hoffmann, K . , “Solids and Surfaces: A Chemist’s View of Bonding in Extended Structures.” VCH. New York, 1988. 140. Hoffmann. K . . R P U .Mod. Plij1.s. 60, 601 (1988). 141. Christmann, K . , Srnf Sci. R e p . 9, I , 130-147 (1988). 142. Chiarello, G . , Andzelm. J . , Fournier, R . , Kusso, N., and Salahub, D. R.. Strr:f: St,i. 202, L621 (1988). 143. (a) Allison, J . N., and Goodard, W. A., S w f . Sc,i. 110, L615 (1981); (b) Bagus, P. S . , and Herniann, K., Pltys. Reu. 5 33,2987 (1986): ( c ) Bauschlicher, C. W., Jr.. J . C h n . Pkys. 85,354( 19x6); (d) Bagus, P. S . , Bauschlicher, C . W.. Jr., Nelin, C. J., Laskowski, B. C., and Seel, M., J . Cl7cvn. Plrys. 81, 3594 (1984). 144. (a) Anderson. A. B., and Awad, M. K., J . Am. C h ~ 7 7 So(.. . 107, 7854 (1985); (b) Mehandru. S. P . , and Anderson, A. B . , Sin:/: Sci. 169, L281 (1986):(c) Kang, [I. B., and Anderson. A. B . , Srrr:f: S c i . 155, 639 (1985); (d) Mehandru, S . P., and Anderson, A. B., S u r f : Sci. 201, 345 (1988). 145. Tomanek, D., and Brennemann, K. H., Srccf: Sci. 127, LI I I (1983). 146. Seip, U.. Tsai, M.-C., Christmann, K., Kuppers, J., and Ertl. G., Surf: Sci. 139, 29 (1984). 147. McKee, M. L., J . C l t m . Pliys. 87, 3143 (1987). 148. Upton, T. H., Stevens, P., and Madix, R. J., J . ClicJm.P h y s . 88, 3988 (1988). 149. Gavezzotti, A., and Simonetta, M., Srruf. Sci. 99,453 (1980). 150. Silvestre, J., and Hoffmann, K., Lungrnrrir 1, 621 (1985). 151. Upton, T . H . , J . V N C .Sci. Techno/. 20, 527 (1982). 152. Egelhoff, W. F., Jr., J . Vtrc. S c i . T(~clrno1.A 5 , 700 (1987). 153. (a) Hsu, D. S . Y . , and Lin, M. C., J . Chem. P h v s . 88, 432 (1988); (b) Hsu. D. S. Y..

BOND-ORDER CONSERVATION

I63

Hoftbauer, M. A.. and Lin, M. C.. L ~ n g m r i r 2 , 3 0 2(1986);(c) Selwyn, G. S., and Lin. M. C.. Chrm. Plzys. 67, 213 (1982); (d) Selwyn, G. S . . Fujimoto, G. T., and Lin, M. C.. J . Phys. C/w,ti. 86, 760 (1982). 154. (a) Norskov. J. K., and Lang, N . D.. Pliys. Rau. B 21,2136 (19x0): (b) Stott, M. J . . and Zaremba, E.. Plrys. Rev.B 22, 1564 (19801. 155. Norskov, J. K., and Stoltze, P.. Swf: Sc;. 189/190, 91 (1987). 156. (a) Benziger. J. B . . App1. Surf: Sci. 6, 105 (1980); ( b ) Benziger, J . B . , Ph.D. ‘Thesis, Stanford Univ.. Slanford, California, 1979. 157. Weinberg, W. H.. and Merrill, K. P., J . Vrrc.. Sci. T d r n o l . 10, 89 (1973); Srrrf: Sci. 33, 493 (1972); 3Y, 206 (1973); 41, 312 (1974); J . Ctrttrl. 28, 459 (1973): 40, 268 (1975). 158. (a) Johnston. H. S . , and Parr, C., J . A m . C l r ~ mSoc. . 85, 2544 (1963); (b) Marcus. K. A,, J . Phys. Cliew~.72, X91 (1968). 159. See, e.g., (a) Wolfe, S.. Mitchell, I). J . , and Schlegel. H . B., J . A m . Clicwr. SOC. 103, 7692, 7694 (1981); (b) Dunning, T. H . , Jr., Harding, L. B., Bair, K. A,, Eades. K. A,. and Shepard, R. L., J . Phys. Chrin. 90, 344 ( 1986). 160. Dunitz. J . D., “X-Ray Analysis and the Structure of Organic Molecules,” pp. 341-360. Cornell Univ. Press. Ithaca, New York. 1979. 161. Van Santen, R. A,, R d . Truu. Clrini. Prry-Btrs. 109, 59 (1990). 162. See, esp., Cerny, S., in “The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis” (D. A. King and D. P. Woodruff, eds.), Vol. 2, I . pp. 42,43,51. 52. Elsevier, New York. 1983.

This Page Intentionally Left Blank

ADVANCIIS IN C A I A I . Y S I S . VOLUME 37

Solid Superacids KAZUSHI ARATA

1.

Introduction

Solid acids have been extensively studied and used as catalysts or catalyst carriers in chemical industry, particularly in the petroleum field, for many years. Many kinds of solid acids have been found; their acidic properties on catalyst surfaces, their catalytic action, and the structure of acid sites have been elucidated for the last 30 years, and those results have been reviewed by several workers, especially by Tanabe (1-3). Among a large number of solid acids, catalysts with fairly high surface acidity, stronger than Ho = -8.2, are the binary oxides of SO,-AI,O,, Ti0,-ZrO,, SO,-TiO,, and Si0,-ZrO,; in particular, SO,-AI,O, bears strong acih sites on its surface and has been used in various organic reactions. The highest acid strength of Si0,-AI,O, has been determined so far to be Ho = - 12 ( 4 ) , a value in the range of superacidity. Following Gillespie’s definition, any acid may be termed a s u p m x . i d when its acidity is stronger than that of 100% H,SO,, i.e., Ho S - 12 (5, 6). Such a superacidity has been reached by a number of systems, which are generally made up by mixing a fluorine-containing Bronsted acid (HF, HSO,F, CF,SO,H, etc.) and a fluorinated Lewis acid (BF,, SbF,, TaF,, etc.). The systems have been developed since the 1960s, when Olah’s studies of obtaining stable solutions of electron-deficient ions, particularly carbocations, focused interest on very high-acidity nonaqueous systems (7). Superacids were extensively studied only in the last 20 years, and acidities up to times that of sulfuric acid have been obtained. The interest in such superacids arises from development of new areas of study as well as from interest in basic research and in synthetic applications, for example, the hydrocarbon chemistry field. The research articles have been recently reviewed by Olah (8, 9 ) . The preparation and use of strong solid acids and superacids are active 16.5

Copyright b l 1990 by Academic Press. Inc. All rights of reproduction in any form reserved.

166

KAZUSHI ARATA

areas of research for isomerization, cracking, hydrocracking, dehydration, alkylation, acylation, converting methanol to gasoline, etc. Because of the reported advantages of solid catalysts, recent research has focused on the preparation and characterization of stronger solid acids. Replacement of homogeneous liquid acids by heterogeneous solid acids as catalysts in the chemical industry is expected to bring about the ease of separation from the reaction mixture, which allows continuous operation, as well as regeneration and reutilization of the catalyst. Furthermore, use of heterogeneous solid catalysts can lead to additional advantages, e.g., no corrosion of the reactor and no environmental problem in the disposal of the used catalyst. Novel organic syntheses that are not possible in the usual acidic media can be accomplished in superacids, including syntheses of economically important hydrocarbons. The remarkable ability of superacids to bring about hydrocarbon transformations, even to activate methane, can open up new fields in chemistry. In consideration of the exceptionally high activity of liquid superacids and their application in hydrocarbon chemistry, it is not surprising that research was extended to prepare solid superacids. As for chemical applications of liquid superdcids, efforts were originally made to attach the superacids to solid materials; the results are found primarily in the extensive patent literature. In the field of catalysis chemistry, solid acid systems stronger than acidic oxides, such as silica-alumina and zeolites, have been developed recently and are categorized as solid superacids. The search for solid superacids became active in the early 1970s, and now is the age of superdcid, particularly in the field of solid acid catalysts. Several methods are available for estimating acidity in solution for the homogeneous superacids: spectroscopy ( U V , NMR), electrochemical methods, chemical kinetics, and heats of protonation of weak bases ( 9 ) . Due to the heterogeneity of solid superacids, accurate acidity measurements are difficult to carry out and to interpret. The most simple and useful way to estimate the acidity of a solid catalyst is to test its catalytic activity in well-known acid-catalyzed reactions; we usually compare the activity with that of SO,-AI,O,. The acid strength of a solid is defined as the ability of the surface to convert an adsorbed neutral base into its conjugate acid (2). If the reaction proceeds by means of proton transfer from the surface to the adsorbate, the acid strength is expressed by the Hammett acidity function Ho, Ho = pK, + log[B]/[BH+], where [B] and [BH'] are, respectively, the concentrations of the neutral base (basic indicator) and its conjugate acid, and pK, is pKBH+.If the reaction takes place by

I67

SOLID SUPERACIDS

TABLE I Busic Indicutors U.srd,for Meu.surrment of Supercicid Strength Indicator" p-Nitrotoluene m-Nitrotoluene p-Nitrofluorobenzene p-Nitrochlorobenzene m-Nitrochlorobenzene 2.4-Dinitrotoluene 2,4-Dinitrofluorobenzene I ,3,5-Trinitrotoluene

P K, -

I I .35

-

11.99

12.44 12.70 13.16 13.75 - 14.52 - 16.04 ~

~

~

(' Color of base form, colorless; color of acid form, yellow.

means of electron pair transfer from the adsorbate to the surface, Ho is expressed by Ho = pK, + log[B]/[AB], where [AB] is the concentration of the neutral base that reacted with the Lewis acid or electron pair acceptor, A. From the above theory, the color of suitable indicators adsorbed on a surface can give a measure of acid strength; if the color is that of the acid form of the indicator, then the value of the Ho function of the surface is equal to or lower than the pK, of the conjugate acid of the indicator (2). Lower values of Ho correspond to greater acid strength. Thus, for indicators undergoing color changes in this way, the lower the pK,, the greater is the acid strength of the solid. Table I shows the Hammett indicators used for the measurement of superacid strength, which we have usually used up until now. The indicators with higher pK, values are 2,4,6-trichlorobenzene (pK, = - 16.12), (2,4-dinitrofluorobenzene)H+ ( - 17.35), (2,4,6-trinitrotoluene)Hf ( - 18.36), and (p-methoxybenzaldehyde)H+ ( - 19.50) ( l o ) ,but we have not used these yet. This review summarizes the recent works on syntheses of solid superacids and their catalytic action, including Lewis acids and liquid superacids in the solid state, as discussed in Sections 11-IV. Sections V1 and VII describe new types of solid superacids we have studied in this decade: sulfate-supported metal oxides and tungsten or molybdenum oxide supported on zirconia. Perfluorinated sulfonic acid, based on the acid form of DuPont's Nafion brand ion membrane resin, is also gaining interest as a solid superacid catalyst; Nafion-H-catalyzed reactions are reviewed in Section V.

I68

K A Z U S H l ARATA

II. Liquid Superacids Supported on Solids Solid acid catalysts generally do not show intrinsic acidity comparable with liquid superacids, but high temperatures can be used to achieve catalytic activity, different from the unstable superacids. To obtain solid superacid catalysts, efforts were primarily made to bind physically or chemically the superacids to solid supports. There have been considerable difficulties in achieving this goal. For example, BF,-based systems such as HF-BF, are difficult to adsorb onto solid supports because of the highly volatile BF,, but SbF,, TaF,, and NbF, have much lower vapor pressures and are more adaptable to being attached to solids ( 8 ) .Extremely reactive HF-SbF, and HS0,F-SbF, can be also attached preferentially to fluorinated alumina and graphite. The primary studies to obtain the solid superacid catalyst of liquid superacids attached to solid supports are found in the extensive patent literature (I 1-23). BF, attached to ion-exchanged resin: BF, or SbFSsupported on graphite; AIF, or HSbF, supported on A1?03; SbF, or TaF, supported on SiO,, Al20,, zeolites, and SiO2-AI20,; HF-SbF, or HS0,F-SbF, supported on metals, alloys, SiO,, A120,, active carbon, graphite, Si02-AIZ0,, and polymer: and CF,SO,H-SbF, supported on AIZO,are examples of the systems. On such carriers, at temperatures as low as 70"C, HF-SbF, and HS0,F-SbF, readily isomerize straight-chain alkanes such as n-hexane or n-heptane, and similar systems are also effective as alkylation catalysts for alkanes with alkenes, especially for isobutane with butenes for obtaining gasoline materials with high octane value. In the case of the latter, sulfuric acid and HF alkylation processes have been used for the production of such alkylates: these processes employ liquid-liquid catalytic systems, which are known to be expensive and troublesome because of problems such as maintaining an acid/hydrocarbon emulsion, product separation, and waste disposal. Gates and co-workers have prepared a superacid catalyst by the reaction of anhydrous AICI, vapors with macroporous, sulfonated polystyrene divinylbenzene (24-27). HCI was evolved, and a bound complex was formed that incorporated S, Al, and CI in the ratio 2 : I : 2 . The resin was found to be capable of isomerizing and cracking n-butane and n-hexane at temperatures of 85-100"C. but the polymer was unstable, rapidly losing CI during operation. The author, afterward. prepared catalysts from several Lewis acids (TiCI,, SbF,, and BF,) and the resin support; the catalysts were active in the presence of small amounts of HCI cocatalyst for the reaction of n-butane, but rapid deactivation resulted from loss of hydrogen halide (28). The activities of the catalysts were compared to the acid strengths of unsupported conjugate Lewis

169

SOLID SUPERACIDS

TABLE I1 Reuction of Butune over ShF5-Trerrted Metul Oxides ut 20°C Products (%Y Metal oxide TiOz Si02 SnOz MgO AlzO, SiOz-TiO2 Si02-AI,0, Si0?-AI20, TiOz-ZrO, SiO?-ZrOz HY CaY CeY LaY 4A 5A 13X

Time (h)

280 280 24 24 280 280 280 20 20 20 20 20 20 20 23 23 23

C,

Cl

i-C,

C,

i-C5

2.2-DMB

C,

IS.1 0.3

6.4 6.9 1.1

59.1 54.8 29.3 24.2 21.1 58.1 47.7 41.0 30.3 14.2 0.7 0.6 0.4 0.7 1.3 1.7 27.8

1.1 1.0 0.2 0.1 0.1 0.6 1.1 0.6 0.2

13.4 6.8 1.3 0.8 1.3 4.8 8.1 4.8 1.3 0.7 -

2.0 2.9 0.1 0. I 0.1 I .0 3.7 I .4 0.3 0.2 -

I .o 2. I 0.1 -

-

0.1 0. I 0.1 0.2 4.5 -

1.3

0.1 21.9 4.4 1.8 0.7 0.4 0. I 0.1 0.1 0.2

0.1

0.1

-

0.1

-

0.2 1.0

-

0.5 2.6 0.7 0.1 0. I

-

-

-

" c,,methane; C,, propane; i-C4, isobutane; C,, pentane; i-C5. iropentane; 2,2-DMB. 2.2dimethyl butane; C6. hexane.

acid analogs indicated by the Hammett acidity function, and Ho values of the catalysts prepared from BF,, AICI,, and SbF, were estimated to be - 11.4, - 13.5, and - 15, respectively. Tanabe and co-workers provided a systematic investigation of Lewis acid-treated metal oxides (29-33). The SbF,-treated oxides such as TiO, and Si02, and mixed oxides such as Si02-A120,, Si02-Ti0,, TiO,-ZrO,, were found to be effective in the isomerization and cracking reactions of butane and other alkanes. The catalysts were prepared by exposing the powdered metal oxides to SbF, vapor followed by evacuation of excess SbF,; prior to exposure to the vapor of SbF,, all the metal oxides were outgassed at 500°C. The adsorption-desorption cycle was repeated a number of times in order to substitute the surface hydroxyl group with fluorine and to let SbF, adsorb on the substituted surface. The product distributions obtained over the catalysts are summarized in Table I1 for the reaction of butane (29, 3f, 32). The activities varied with the kinds of metal oxides that were treated with SbF,. SbF,/ SO,-TiO, showed the highest activity, and SbF,/TiO, was highly selective for the skeletal isomerization of butane, the selectivity being 72%. On

I70

KAZUSHI ARATA

treatment with SbF,, TiO?, SO,, SnO,, MgO, A120,, Si0,-AI2O3, SiO,-TiOz, and Ti0,-ZrO,, 13 x molecular sieves became highly active catalysts, whereas ion-exchanged Y-type and A-type molecular sieves showed rather low activities. The acid strengths were determined by observing the color change of the Hammett indicators to be - 14-52< Ho 5 - 13.75 for SbF,/SiO2-AI20, and - 13.75 < Ho 5 - 13.16 for SbFJ Si02-Ti0, and SbF,/AI,O,. Conversions of several alkanes proceeded at room temperature or below, over the metal oxides treated with SbF,; the conversion rates of methylcyclopenalkanes were in the following order: cyclohexane tane > hexane > pentane -- 2-methylbutane > butane = 2-methylpropane > propane + 2,2-dimethylpropane -- ethane --- methane (31, 32). The 1R spectra of pyridine adsorbed on SbF,/SiO2-AI20, showed that both Briinsted and Lewis acid sites were present on the surface when Si0,-A1,03 was treated with SbF, at low temperatures (below 100°C),but only Lewis acid sites were present when treated at 300°C (30, 32). The reaction of SbFSwith Si0,-AI,O, is considered as follows: 2-

Br6insted site

I

6+ H \

H 6- I OSbFc 0

Lewis site

i

I

I

6+

1

I

I

8SbFc

-0-Si-O-Si-O- Al-O-

Lewis site I

OSbF4 F

+

i

I

6+

I

I

i

-O-S~-O-Si-O-Al-O-

6-

SbFc

A comparison of the reactivity of SbF,-treated metal oxides with that of HS0,F-, SbC15-, and HS0,F-SbF, (magic acid)-treated catalysts showed that the former was by far the best catalyst for reaction of alkanes (31,32). Tracer studies of conversion of alkanes catalyzed by the superacids were performed; it was suggested that the reactions proceeded by carbenium ion mechanisms in which the reactions were initiated by abstraction of H - from the reactants (33). Namba and co-workers prepared SbF,/HY-zeolite catalysts and examined them in the reaction of pentane. Although the catalysts were quite active in the initial stage, they were completely deactivated afterward by the formation of heavy hydrocarbons (34).Krzywicki and co-workers also

SOL1 D S UPE RACI DS

171

prepared superacid sites on the surface of Al20, by sublimation of P,O, and AICI, onto it (Ho 5 - 13.7) (35). 111.

Superacid-Intercalated Graphites

The intercalation of metals and salts in the lattice of graphite is well known (36, 37). Salts most easily intercalated are chlorides of transition metals such as FeCI, or CoCI2. Graphite possesses a layered structure that is highly anisotropic, and it consists of sheets of sp2 carbon atoms in hexagonal arrays with a C-C bond distance of 1.42 A consistent with a one-third double-bond and two-thirds single-bond character ( 9 ) . The distance between the layers is 3.35 A and is in accord with the fact that the graphite sheets are held together only by weak Van der Wads forces. Lalancette and Lafontaine found that SbF, can be intercalated easily in the lattice of graphite simply by heating a mixture of SbF, and graphite at 110°C for a few days (38).All operations were performed under dry nitrogen, in a dry box; the graphite was thoroughly dried by heating in uuc'uo at 150°C for at least 24 h. X-Ray powder diffraction showed the typical pattern of an expanded lattice, the strong band for pure graphite at 3.35 A being almost completely eliminated by the intercalation and a new band appearing at I I . 10 A. The difficulties encountered in handling liquid superacids and the need for product separation from catalyst in batch processes stimulated research in the synthesis of graphite-intercalated superacids as solid catalysts. In addition to SbF, (39-46), other superacids such as NbF, (471, HF-SbF, (45),AICI, (48, 4 9 ) , AIBr, (491, and Br,-AlBr, (45)have also been synthesized as graphite intercalates and used for various superacidcatalyzed reactions. The isomerization of a series of cyclic and bicyclic saturated hydrocarbons over SbF,-intercalated graphite was achieved at or below room temperature without the ring opening and cracking reactions, and the thermodynamic equilibrium was reached for the isomers in all cases (39). Interconversion between cyclohexane and methylcyclopentane also yielded the thermodynamic equilibrium mixture. It was shown that the SbF,-intercalated graphite efficiently promotes disproportionation of various alkylbenzenes by simple mixing at room temperature (4f). The isomerization of methylpentanes was carried out over the catalyst at room temperature, - 30, and - 17°C in a continuous flow system; a careful study of the kinetically controlled product distribution was performed to obtain information for the reaction path (42, 43). The skeletal rearrangements of I3C-labeled2-methyl,3-methylpentaneand

I72

KAZUSHI ARATA

2,3-dimethylbutane were studied using the same catalyst under similar conditions (44).The isomerization process, which involves only intramolecular rearrangements of the hexyl cations, was fully described by considering 1,2-alkyl shifts of methyl and ethyl groups and rearrangements via (protonated) cyclopropane rings. Yoneda and co-workers found graphite-SbF,, -SbF,-HF, and -AIBr,-Br, catalysts to be effective for the isomerization and cleavage reactions of pentane and hexane (45).The HF-SbFS catalyst gave skeletal isomerization products exclusively from pentane and hexane; the Br,-AIBr, catalyst showed a highly catalytic activity to yield C4-C, alkanes, formed via the p-cleavage reaction of oligomeric alkyl cations derived from starting alkanes. Olah and his group investigated the direct ethylation of methane with ethylene using "C-labeled methane over solid superacid catalysts such as TaF,/AIF,, TaF,, and SbF,-graphite, and found that methane is ethylated by ethylene to give propane over the catalysts (46).A very low concentration, < I .O mol%, of ethylene compared to methane was necessary in order to minimize the self-condensation cracking of ethylene. Lalancette rt ul. studied the catalytic activity of AIC1,-intercalated graphite for the alkylation of aromatics with ethyl bromide, ethylene, propylene, and isobutylene and compared it with pure AICI,; the intercalate was a milder catalyst than AICI, and gave less polysubstituted products (48) . The major drawback in the extended use of this catalyst system is its relatively short lifetime and ease of deactivation. An example for such a deactivation is shown in Table Ill, where reactions were carried out in a flow system, in the gas phase, and in the temperature range of 160-180°C at atmospheric pressure (49). Although the initial conversions were high, the catalyst was totally deactivated after a period of 6-8 h. A possible reason is leaching of the metal halide from the graphite layers by the feed. Three possible reasons for such a deactivation have been given by Heinermann and Gaaf in the case of the SbF, catalyst (40):

I . Leaching out of SbF,. 2. Reduction of Sb(V) to Sb(II1). 3. Poisoning of the acidic sites by carbonaceous products.

IV. Aluminum Halide-Metal Salt Mixtures

Ono and co-workers have shown that the mixtures of aluminum halides with metal salts such as AlCI3-Ti2(SO4),,AlBr,-Ti2(S04),, AICI,-CuSO,, and AIC1,-CuCI, are active for the isomerization of paraffins at room

173

SOLID SUPERACIDS

Conversion (76) Onstream time (h)

Ethylation

Tranaethylation

1

60.4 62.9 43 8

45.0 70.4 66.9 70.3 56.9 16.2 13.8 15.2 12.7 10.4

2 3 4 5 6 7 8 9 10

33.1

26.5 6.8 2.6

2.2 I .7 I .s

temperature (50-54). The catalysts were prepared by kneading a mixture of aluminum halide and a dehydrated metal salt in a porcelain mortar in a dry nitrogen atmosphere. The isomerization of pentane was carried out with a series of mixtures containing aluminum chloride or bromide with sulfates of metals such as Ti, Fe, Ni, Cu, AI, and others; the most effective catalyst was an equimolar mixture of AIBr, and Ti2(SO4),with a conversion of 86% and a selectivity to isopentane of 99% at room temperature (50, 51). In the vapor phase conversion, however, the main product was isobutane. The AICI,-CuSO, mixture was more thoroughly investigated (52). The catalytic activity of the mixtures for the isomerization of pentane was found to be proportional to the amount of CuSO, and also to the specific surface area of the CuSO, used. The acidity was estimated to be - 14.52 < Ho < - 13.75. It was concluded that the active species were located on the surface of the CuSO,. The mixtures of AIC1,-CuCI,, AICI,-TiCI,, and GaCI,-CuSO, were also found to be highly active for the pentane isomerization (53,549. The X-ray diffraction revealed the formation of new compounds by the reaction of the was two components of the mixtures; in the case of AICI,-CuCI2, CU(AICI,)~ isolated and was found to be highly active for the isomerization (54). V. Nafion-H (Perfluorinated Resin Sulfonic Acid)

Nafion-H is a registered trademark of E. 1. DuPont de Nemours & Co. and has been shown to be a useful and versatile acid catalyst for organic

I74

KAZUSHl ARATA

reactions. It is effective in a wide range of liquid and gas-phase reactions, including dehydration of alcohols, rearrangements, electrophilic substitutions on aromatic nuclei, and polymerizations. Nafion resins are copolymers of tetrafluoroethylene and monomers such acid and were first syntheas perfluoro-3,6-dioxa-4-methyl-7-octensulfonic sized by DuPont chemists.

where m = 5-13.5, n = -1000, and z = I , 2, 3, . . . . A series of compositions may be produced in which m can be as low as 5 and as high as 13.5. The lower value of m corresponds to an equivalent weight (EW) of 950 and the higher value to 1800. In terms of ion exchange capacity the range is from 1.05 meq/g to 0.55 meq/g. The system separates into and hydrophilic (-S03H) regions, and the hydrophobic (-CF,CF2-) superacidity of the sulfonic acid group is attributed to the electron-withdrawing ability of the perfluorocarbon chain. A convenient solid of perfluorinated-sulfonic acid can be made readily from DuPont’s commercially available Nafion brand ion membrane resins. Powder granules of the 1200-EW polymer, Nafion 501, have been used most frequently in catalytic applications; the price in the K + form of the perfluorosulfonic salt, SOIX, was $650/kg in 1981. Because only the potassium salt derivative is commercially available, the salt is converted to the free sulfonic acid by treatment with mineral acid. A standard procedure for the conversion is described below. This procedure also serves to regenerate the resin in various catalytic cycles. A slurry of 100 g of perfluorinated ion-exchange polymer in 130 ml concentrated HCI (36%) and 400 ml distilled water is stirred at 50-60°C for 2 h. The aqueous acid is decanted away from the resin and fresh acid/ water solution added. The mixture is heated and stirred for another 2 h. After the third exchange, the resin is isolated by filtration and washed with distilled water until the washings are neutral. The resin is dried in a vacuum oven under N ? at - I 10°C for 6 h. Conversions 295% are obtained with this procedure. The Nafion-H given above exhibits acidic character comparable to 100% H,SO,. The physical and chemical properties of the resins as reported by the manufactures are summarized in Table IV (55). It has been suggested that Nafion-H has an acidity between - I 1 and - 13 ( 8 , 91, and we also determined it to be - 12 by the visual color change method of the

-

I75

SOLID SUPERACIDS

TABLE IV Typicul Properties of Perfliiorinuted Ion Exclicingc Polymcvs"

Property Physical form Particle size Equivalent weight Thermal stability Acid strength (Ho) Surface area (mVg) Bulk density (g/ml) True density (giml)

s01x

51 IX

Granules 12 mesh ( I .7 mm) to 140 mesh (0.105 mm) 1200 (0.83 meq/g) 180-200°C - 1 1 to -I5 0.02 I .0 I .9

Granules 12 to 140 mesh I100 (0.9 meq/g) 180-200°C - 1 1 to -15 0.02 I .0 1.9

" The properties shown in the table are for the free sulfonic acid material.

Hammett indicators; the resin dried at IOWC changed slightly the color of p-nitrotoluene (pK, = - 11.35) in benzene solvent (56). The maximum continuous operating temperature of Nafion-H is about 175°C in anhydrous system. However, the maximum operating temperature could be higher in aqueous and organic systems with proton-donating solvents. Nafion-H is also stable in corrosive environments. These unique properties have led to an extensive development of reactions in synthetic organic chemistry. Table V summarizes several reactions that have been demonstrated on a laboratory scale; 1 know of no industrialized chemical process using Nafion as a superacid catalyst. Although many of the reactions were carried out with stirring a mixture of reactants and Nafion-H, several alkylation, disproportionation, rearrangement, and esterification reactions were performed by means of the flow-reaction method in the liquid or gas phase. For instance, in the esterification of carboxylic acids with alcohols, when a mixture of the acid and alcohol was allowed to flow over a NafionH catalyst at 95-125°C with a contact time -5 s, high yields, usually 290%, of the corresponding ester were obtained (82).It had been found that no reactivation of the catalyst was needed because the catalytic activity of the Nafion remained unchanged for prolonged periods of operation. The topic of catalysis with Nafion has recently been reviewed in detail (86).Apart from using Nafion-H primarily as a solid superacid catalyst, a number of reports have described the use of functionalized Nafion derivatives by metal cation exchange to achieve various types of organic reaction. These include a bifunctional catalyst (acid and cation site), a heterogeneous pertluorosulfonate salt (only cation sites), and a trifunctional

I76

KAZUSHI A R A T A

TABLE V Krrictions Crittilvzcd hv N(!fioti-H

Reaction class Alkylation ArH + KCH=CH, + ArCHKCH, ArH + KX + ArK + HX ArH + CH,OH + ArCH, + HzO ArOH + CH30H + CH,ArOH + H?O ArH + CICOOK --z ArK + HCI + CO? Acylation ArH + KCOCl+ ArCOK + HCI C,H$ + (KCO),O -+ C,H,SCOR + KCOOH Nitration ArH + HNO,--t Ar-NO? + H?O ArH + C,HYONO'+ ArNO: + C4HqOH Disproportional ion PhMe? + PhBrMe,-, PhSrMe? + PhMq ?PhK,--t PhRi + PhR 2C,H,,Me,~.,, c6H,,4'Me6-,,-' + C6H,,.IMe,-,,+I ArH + C,,HyNO?-+ ArNO? + C14Hl,, Oligomerization Styrene + dirners. [rimers Rearrangement m-PhMe? + o-. p-PhMe, Ally1 alcohols + aldehydes Pinacols pinacolones (CH2),,C(OH)CCH (CH?),,_,CHCCOMe Condensat ion PhCHO + 2PhYNH: + PhCH(PhYNH'): + HzO Z(CH3):CO --+ (CH,),CCHCOCH> + HZO R'CH,COK' + (CH'O),, -+ CH2(OCH?)2CK'COR' XCH,)?CO --+ I.3.S-PhMe7 + 3H,O El herification KOH + CH,(OMe),-. KOCH!OMe + MeOH K?C(OH)CH,CHKCR,OH cyclic ether + HIO Esterification R'COOH + K'OH -+ R'COOK? + HIO Hydration K'R?COCK3R4 t H 2 0 RIR'C(OH)C(OH)R'RJ Diacetate KCHO + (MeCO),O + RCH(OCOMe)? RIR?CO + HC(OMe), -+ KIR'C(OMe)l + CHO(0Me)

-

--j

---f

---f

---f

Reference 49.57 58 59 60

61 62 63 64 65, 66

67 68 69

70 71 72 73 74 7.7

76 77 78 79

80 8l 8-3

83

84 85

I77

SOLID SUPERACIDS

catalyst (acid and two different cation sites) (87). Catalysis with metal cation-exchanged resins has recently been reviewed (88). Superacid -MY'-+ (acid site)

Bifunctional -M!++ (acid and cation site)

I I

M;'

Sulfonate (cation site)

VI.

Trifunctional (acid and two cation sites)

I

M2' 1

Bis(sulfonate) (two cation sites)

Sulfate-Supported Metal Oxides

It is known that sulfur treatment of catalysts changes the catalytic behavior significantly in some cases (89-91). Several examples of sulfurpromoted catalysts using H2S are a selective synthesis of acetic acid by Rh-Ir-Mn-Li/SiO, (92),the hydrogenation of SRC-11 heavy distillate (93) and of monoaromatic hydrocarbons (94, 95) over Co-Mo/Al,O,, and the dealkylation of cumene over metal Y zeolites (96). The promoting action of SO2 was also observed with cation-exchanged zeolites for the doublebond isomerization of butenes and for the dehydration of 2-propanol (97-99). The addition of SO:- enhanced the activities of Moo,, VO,/TiO,, and ZrO, for the reduction of NO, with NH3 or H?; the positive effect of SO:- on the activity was due to acidity change on the surface of the catalysts (100, 101). We have synthesized solid superacids, which can be used at temperatures of over 5OO0C, by the SO:- addition to several metal oxides. Before discussion of synthesis of the superacid, 1 would like to mention briefly the way to achieve success in this study. Benzylation of toluene with benzyl chloride, which is a typical example of Friedel-Crafts alkylation, is known to be catalyzed by Lewis-type superacids such as AIC13 and BF,. This type of catalyst has been mostly used for the Friedel-Crafts reaction, which is one of the most studied of organic reactions. This reaction was performed over several metal oxides and sulfates, and iron sulfates showed an unexpected effectiveness for the reaction (102-104). The catalytic activities of FeSO, and Fe,(SO,), for the reaction were examined in detail; the activities were remarkably dependent on calcination temperature, the maximum activity being observed with calcination at 700°C (105-107). Catalytic actions analogous to the above case were also observed with other Friedel-Crafts reactions, the benzoylation of toluene with benzoyl chloride (108),the isopropylation of toluene with isopropyl halides (109), and the polycondensation of benzyl chloride (110).

I78

KAZUSHI ARATA

The material obtained by calcination at 700°C showed also the highest activity among the catalysts treated at various temperatures in other cases, including alkylations of aromatics with alkyl chlorides catalyzed by calcined iron sulfate treated with hydrogen chloride (111, 112), and acetylations of aromatics with acetyl halides or acetic anhydride catalyzed by calcined iron sulfate activated by exposure to a mixture of benzyl chloride and the aromatics (113, 114). Among the results of our investigation of Friedel-Crafts reactions in the heterogeneous system that we have dealt with in our laboratory in the past 20 years, iron oxides obtained by calcining iron hydroxides, which were precipitated by hydrolyzing FeCI, and Fe(NO1),with ammonia, showed high catalytic activities for the reactions, though commercial iron oxides were inactive; the studies were the alkylation and acetylation of toluene with alkyl and acetyl chlorides ( 1 1 3 ,the synthesis of thermally stable oils by the benzylation of biphenyl with benzyl chloride (116). and the polycondensation of benzyl chloride (117). Tanabe rt nl. studied in detail the catalytic action and properties of metal sulfates; most of the sulfates showed the maximum acidity and activity by calcination at temperatures below 500"C, with respect to the surface acidity and the acid-catalyzed reaction (118,119). Other acidcatalyzed reactions were studied with the FeSO, catalyst together with measurement of the surface acidity of the catalyst: the substance calcined at 700°C showed the maximum acidity at Ho S 1.5 and proved to be the most active for the polymerization of isobutyl vinyl ether, the isomerization of d-limonene oxide, and the dehydration of 2-propanoI(120-122). It is of interest that the catalyst calcined at a slightly higher temperature, 75OoC,was completely inactive and zero in acidity in spite of the remarkable activity and acidity when heat treated at 700°C. Surfaces of FeSO, and Fe2(S0,), calcined at various temperatures were examined by X-ray photoelectron spectroscopy and X-ray diffraction (XPS and XRD) (123, 124). Figure 1 shows one of the XPS spectra, 0 Is. The single peak at 531.6 eV for FeSO, (500"C), FeSO, (6500C), and Fe,(SO,), (500°C) is attributed to the sulfate oxygen; the main peak appeared at 529.9 eV for FeSO, (700"C), and Fe,(SO,), (700°C) is assigned to be the oxide oxygen. The samples, both sulfates calcined at 700"C, also show a shoulder peak around 532 eV, and this peak is assigned to be the sulfate oxygen. On the basis of the results obtained by XPS and XRD, it was concluded that iron sulfates heat treated at 700°C have the following surface properties: ( I ) the sulfates completely decompose to form a-Fe,O, at 675-70OoC, and afterward the crystallization proceeds rapidly; and (2) the slight amount of sulfur (0.15 wt%) remaining after the decomposition of the sulfate salts exists mainly as SO:- on the surface.

179

SOLID SUPERACIDS

I

I

I

I

534

532

530

523

534

532

530

523

Binding energy, eV

FIG. I . XPS 0 Is spectra of iron sulfates calcined at 500-700°C. ( A ) FeSO, (500°C); ( B ) FeSO, (650°C); (C) FeSO, (700°C); (D) Fe,(SO,), (500°C); (E) Fe,(SO,), (700°C).

A. PREPARATION OF SUPERACIDS I.

SupUte-Supported Iron

Oxidc

On the basis of the above results for the surface properties of iron sulfates treated at 700"C, sulfate-treated iron oxides were differently prepared, and the catalytic effect of sulfate ion on solid acid catalysts was examined. It was found that remarkable increases in the surface acidity and in the catalytic activity of Fe,O, result from the sulfate ion treatment followed by heat treatment (125, 126). The catalyst was prepared as follows. Iron hydroxides (2 g) were exposed to aqueous sulfuric acid or aqueous ammonium sulfate (30 ml) on a filter paper. After drying, the materials were powered and calcined in air. The catalytic activities of various iron oxides with or without the sulfate treatment for the reaction of 2-propanol are summarized in Table V1. The Fe,O, catalysts treated with 0.25-0.5 M H,SO, showed unexpectedly high activities for the reaction, though the reaction did not occur at all over the catalysts prepared by calcining Fe(OH),. The treatment with ammonium

180

KAZUSHI ARATA

TABLE VI Drhvdrution .f'2-Proprrnol

so3 Catalyst

Treatment

Calcination temp. after treatment ("C)

Fe?O,-l

Without treatment Without treatment 0 . 5 M H,SO, 0 . 5 M (NH&SO, 0 . 5 M H2S0, 0.5 M H$O, 0.25 M HZSO, 0.5 M H$O,

300 500 500 500 550 600 500 500

Fe20,-II

Conversion 0 0 54 58 17 0 85 79

(%)

content (Wt<%o)" 2.33 2.06 0.89 0 3.43 4.80

'' Estimated from weight decrease at 500-800"C in TGA.

sulfate also enhanced the activity to the same extent as that with sulfuric acid. Temperature gradient analysis (TGA) data of the catalysts treated with sulfate ion showed a weight decrease at 550-750°C; this decrease was caused by decomposition of the sulfate to form SO,, the catalytic activity being related with the amount of sulfate adsorbed on t h e surface of Fe,O,. The Fe,O,-I1 catalyst [prepared by hydrolyzing Fe(NO,), with ammonia] subjected to the similar treatment showed higher activity and higher SO, content than the Fe,O,-I catalyst. The quantity of S was estimated by chemical analysis to be 0.92 and 0.23 wt% for the Fe,O,-I1 catalysts treated with 0.25 M H,SO, and calcined at 500 and 600"C, respectively; specific surface area of the former catalyst was 77 m'/g, whereas that of Fe,O, without the sulfate treatment was 37 m2/g, the large increase of the area being observed (56). The activity and the SO, content were found to depend greatly on the calcination temperature of the hydroxide before the treatment. By experiments using differential thermal analysis (DTA) and XRD, it was revealed that sulfate ion was not adsorbed on the crystallized oxide, but it was adsorbed on the amorphous or especially the hydroxide form, and hence the amount of adsobed anion was closely related to the catalytic activity. The XPS spectra of Fe,O,-I1 catalyst were completely consistent with those of iron sulfates calcined at 700°C as shown in Fig. I , with the shoulder peak at 532. I eV being observed; the sample was evacuated and measured at room temperature (56). Benesi prepared sulfuric acid mounted o n silica gel whose surface acidity was quite strong, Ho < -8.2 (127, 128). In this case, the material was prepared by mounting the acid on silica gel by an impregnation procedure

SOL1 D SUPERACL D S

181

60

50

b?

40-

f W 0)

d

30

-

0 0

u

5 20 .-

-

L v) 0)

c >

10-

0 Cal c i no t i on temperature, " C F I G . 2. Reaction of 2-propanol to propylene over Fe,O,-l treated with 0.5 M H2S0, ( 0 ) or (NH&SO, (0) and crrlcined at various temperatures. Pulse reaction conditions: He carrier 30 cmimin, pulse size 0.4 pI (liquid), catalyst 3 0 mg, temperature 170°C.

followed by drying at 120°C, compared with calcination at high temperature for the present catalysts; the high acidity of Benesi's sample is concluded to be of course due to the mounted acid. Sulfuric acid and ammonium sulfate are known to decompose at temperatures below 300°C. The catalytic activities of the Fe,O,-I catalyst for the reaction of 2propanol are shown as a function of calcination temperature ofthe catalyst in Fig. 2 (56).The maximum activity was observed with calcination at 300 and 450-500°C for the sample treated with sulfuric acid. It is considered that the former activity was due to the catalytic action by the mounted acid, but the latter one was based on acid sites created by strong interaction of the ion with the support. In the case of the sample treated with ammonium sulfate, the maximum activity was observed with calcination at 500"C, and the calcination at temperatures below 300°C did not give activity. The results seem to be reasonable judging from the absence of catalytic effect of ammonium sulfate. The catalytic action of the Fe,O,-1 treated with sulfate ion of different concentrations was examined; the maximum activity was observed at 2% of the SO, content or 0.5 M concentration of both HJO, and (NH&SO,. The material treated with sulfate ion of 0.5 M concentration showed IR

I82

K A Z U S H I ARATA

adsorption bands at 980-990, 1030-1060, 1120, and 1220 c m . ~ 'which , are assigned to the bidentate sulfate coordinated to metal elements, whereas the samples treated with 1.5 and 3 M sulfate ion showed IR spectra quite similar to those of iron sulfate, as shown in Fig. 3 (126, 129).The treatments with SO,, H,S, and SO3also promoted the catalytic activity, the 1R spectra of those substances being similar to spectrum D or E in Fig. 3 (56, 126). The present catalysts showed quite high activities for the dehydration of ethanol, much higher than that of Si02-A1,0,, which is well known as a catalyst with one of the highest surface acidities (126). Thus, the catalytic action for the reaction of butane, which is catalyzed by superacids, was examined, and it was found that the present catalysts are active for the skeletal isomerization of butane to isobutane at room temperature or even at 0°C ( 4 ) . Si0,-AI20, was totally inactive for the reaction. The acid strength of the SiO2-A1,O, used was in the range of - 12.70 < H o 5 - 1 I .35 (4). Consequently, the present catalysts are concluded to be solid superacids. The material prepared by hydrolyzing FeCI, with ammonia followed by treating with H2S04and calcining had a tendency to be converted to iron sulfate. The catalysts obtained by using 0.25 and 0.5 M H,SO, showed IR spectra similar to spectrum C in Fig. 3 (56);the latter sample gave an XRD pattern of a mixture of a-Fe?O, and Fe,(SO,), forms, whereas the sample prepared from Fe(NO,), and 0.5 M H2S0, gave only the oxide form, without the sulfate form (56).

2. SO,lTiO,, ZrO,, HfO,,SnO,, SiO,, und AlzOj On the basis of the results of the sulfate-supported iron oxide, this preparation method of catalyst was applied to other metal oxides. The activity enhancement of metal oxides by sulfate addition was not observed for MgO, CaO, CuO, NiO, ZnO, CdO, AlzO,, La,03, MnO,, Tho,, Bi,O,, and CrO,, but it was observed for TiO?, ZrO,, SnO,, and SiO, (129). Solid superacids with an acid strength of up to Ho 5 - 16.04 were obtained by adsorbing sulfate ion onto hydroxides or oxides of Ti, Zr, Sn, and Si followed by calcination above 500°C (130-136). The superacid of TiO, was prepared as follows. H,TiO,-1 was obtained by dissolving titanium(1V) isopropoxide in aqueous nitric acid, hydrolyzing with aqueous ammonia, washing the precipitate, and drying at I00"C. H,TiO,-IL was prepared by hydrolyzing aqueous TiCI, with aqueous ammonia followed by washing and drying. The catalyst was treated with sulfate ion by pouring 0.5 M H,SO, (30 ml) onto the dried titanium material (2 g) on a filter paper, drying, and calcining in air. The catalysts t h u s

SOLID SUPERACfDS

1400

1200

1300

183

800

!Jave number, cm-'

FIG.3. I R spectra of Fe,(SO,), (A) and Fe,O,-I catalysts treated with 3 M H2S0, ( B ) , I .5 M H2SO4 (C), 0.5 M H2S04 (D), and 0.5 M (NH&S04 ( E ) . Calcination temperature, 500°C.

prepared from H,TiO,-I and -11 were referred to as Ti0,-I and -11, respectively. Both catalysts were active for the skeletal isomerizations of butane and isobutane at room temperature (130). The reaction of butane was carried out under pulse reaction conditions; the catalytic activities were shown as a function of calcination temperature of the catalyst (Fig. 4) (129). Butane was converted to isobutane and propane, and the catalytic activity of Ti0,-I was higher than that ofTiO,-II. The maximum activity was observed with calcination at 525°C. The quantity of S was estimated to be 2.1 I and 0.01% for the Ti0,-1 catalyst calcined at 525 and 650"C, respectively (56). The activity enhancement of TiO, by an addition of ammonium sulfate was also reported by Tanabe et uf. (137). The superacid of ZrO, was prepared in the same manner as that of TiO, (131, 132). Zr(OH), was obtained from ZrOCI, and ZrO(NO,),, and the catalysts prepared from those are referred to as Zr0,-1 and Zr02-11,respec-

I84

KAZUSHI ARA TA

450

500

550

600

650

Calcination temnerature, "C FIG. 4. Reaction of butane at 150°C over S0,/Ti02-I (0) and SO4/TiO2-II( 0 ) .Pulse reaction conditions: He carrier 10 cmimin, pulse size 0.05 rnl (gas). catalyst 0.3 g.

tively. The maximum activity for the reaction of butane was observed with calcination at 625-650°C for Zr0,-I and 575°C for Zr0,-11 (132). The quantity of S was estimated to be 2.8, 2.2, and 0.2 wt% for the ZrOz-l catalyst treated with 0.5 M HISO, and calcined at 500, 6.50, and 800"C, respectively (129). The catalyst, Zr0,-I heat treated at 650"C, converted propane to methane (5. I %)and ethane ( I .3%) at 280°C (132).The reaction of butane was carried out in a recirculation reactor at 25°C over ZrOz-I calcined at 650°C; pentane and isopentane were observed as products in addition to propane and isobutane (132). The relation between the precalcination temperature of Zr(OH), before the treatment and the catalytic activity for the reaction of butane was examined, and the results are shown in Fig. 5. It is seen that the catalysts prepared by heating Zr(OH), at 100 and 400"C, then treating each with 0.5 M H2S04and finally calcining at 500"C, gave almost the same conversions, but the activities decreased greatly by calcination of Zr(OH), over 450°C. DTA ofZr(OH), showed that this material crystallizes at 410°C. Thus, the treatment with sulfate ion on the crystallized oxide is not effective, as was observed in the case of Fe201(129). Hafnium is the third element in the same group with Ti and Zr, the transition metals of the Periodic Table; it is predicted that HfO, is enhanced in acidity by sulfate addition up to superacidity. The catalyst, which was obtained by exposing Hf(OH),, prepared by the hydrolysis of HfCI,, to 1 M H,SO, and then calcining, was active for the skeletal isomerization of

185

SOLID SUPERAClDS

H

40 -

al C

eax

30-

D

0 c

s 2 0 C

0

.-

t

-

J

3

n

r .-:: al

20

10-

> C u 0

100

203

300

400

500

600

Heatlng temperature of Zr(OH), , " C

FIG. 5 . Relation between calcination temperatures of Zr(OH)4 before the sulfate treatand propane ( 0 ) . ment and their catalytic activities for reaction of butane to isobutane (0) ( A ) Crystallization temperature of ZrOz determined by DTA. Pulse reaction conditions: He carrier 3 cm/min, pulse size 0.04 ml (gas), catalyst 0.3 g. temperature 250°C.

butane (138). The maximum activity was observed with calcination at 700"C, and this activity was close to that of the ZrOz-l catalyst calcined at 650°C. Tanabe et af. applied the preparation method of superacid to tin oxide, but the highest acid strength was not superacidic, Ho 5 -8.2 (139). We synthesized the SnO, catalyst with an acid strength of Ho 5 - 16.04 by a preparation method different from the previously mentioned cases; tin hydroxide was obtained from the solution with pH 10 followed by exposure to 3 M H,SO, and calcining at 550°C. The catalyst was active for the reaction of butane at room temperature (135). The SiO, catalyst, which was obtained by exposing silica gel to SO2CI2 followed by drying in a vacuum and calcining in air at 400"C, showed high activity for the dehydration of ethanol, much higher than that of Si02-A1203,the temperature difference between both catalysts to get the same conversions being over 40°C (136). The silca gel was prepared by dissolving Si(OC2H,), in water with a few drops of HN03, stirring until the gel formation, and drying at 100°C. The maximum activity was observed with calcination at 400°C. Treatment with H,SO, and gelation using NH, instead of HNO, were not effective to generate superacidity. It is assumed from DTA and IR experiments that the surface structure is the same as that to be described for other superacids. In the case of the sulfate-treated superacids of Fe, Ti, Zr, Hf, Si, and

186

KAZUSHI A R A T A

Sn, superacid sites were not created by the treatment of sulfate ion on the crystallized oxides but rather on the amorphous forms, followed by calcination to the crystallization. The superacid of Al,O, was prepared from the crystallized oxide, y-Al20,; highly active catalysts were obtained by treatment on the crystallized oxide rather than on the amorphous one (140). The most active catalyst was obtained by exposing y-Al,O, to 2.5 M H,SO, followed by calcining in air at 550-650°C. The catalyst was active for the benzoylation of toluene with benzoyl chloride or benzoic anhydride in a heterogeneous system. Many binary metal oxides are known to have a surface acidity higher than that of each component oxide (2).One of the mixed oxides with high surface acidity is Ti0,-ZrO,, whose catalytic activity has been examined in various reactions as well as in the dehydration of ethanol (141). Thus, the present method of catalyst preparation was applied to TiOz-ZrO, to get higher acidity, but an activity enhancement more than that of each component oxide was not observed by the addition of sulfate ion (56).The sulfate treatment was also attempted on Zr02-Sn0,, but results were not obtained (142). n. Acidity. Acid strength of the superacids was examined by the visual color change method using the Hammett indicators, where indicator dissolved in solvent is added to the sample as powder form placed in nonpolar solvent (2), but the TiOz and ZrO, superacids (white color) were immediately colored in organic solvents (benzene, toluene, hexane, carbon tetrachloride, etc). However, sulfuryl chloride was found to be a solvent suitable for the acid strength determination of the present superacid catalysts; guaranteed-grade sulfuryl chloride was dried over silica gel before use (131). Afterward, cyclohexane was also found to be suitable (135). The results with the indicators m-nitrochlorobenzene (pK, - 13.16), 2,4-dinitrotoluene ( - 13.75), 2,4-dinitrofluorobenzene( - 14.52), and I ,3,5trinitrobenzene ( - 16.04) are summarized in Table V l l ( 5 6 , 129, 132, 135, 140).The acid strength is estimated to be Ho S - 16.04for Zr02-l(650"C), Zr0,-11 (575"C), and SnO, (550°C) and -16.04 < Ho S -14.52 for Ti0,-I and TiO,-II (525°C) and A120, (650°C). I t is of interest that the Zr0,-I catalyst, even when heat treated at quite a high temperature (SOO"C), is still a superacid with an acid strength higher than Ho = - 13.75, Acid strength of the HfO, catalysts could not be measured by the visual color change method of Hammett indicators because the materials change color (to yellow) after calcination. The maximum activity was observed with calcination at 700"C, and its catalytic activity for the reaction of butane was close to that of the Zr0,-1 (650°C) catalyst. Thus, the catalyst treated at 700°C is considered to hold the acid strength close to Ho = - 16 on the surface (138). The superacid of FezO, is also colored (brown); it is

I87

SOLID SUPERACIDS

pK, value of Hamrnett indicatorh

Catalyst"

ZrO,-I

ZrO,-ll TiO,-l

Ti02-Il

SnO? AI,O,

-

(500°C) (650°C) ( 800°C ) (575°C) (650°C (500°C) ( 525°C) (600°C) (500°C) (525°C) (550°C) (650°C)

13.16

~

13.75

+

~

14.52

+

+

+

I

+

t

+

+

t

+ +

+ -

16.04

?

+ + +

+

-

?

2

+ +

" Calcination temperature given in parentheses. Solvent is sulfuryl chloride except for SnO, and AI,O,, for which the solvent is cyclohexane. ZrO,-l, Zr02-ll, TiO,-I, and TiO,-ll prepared from ZrOCI?, ZrO(NO,)?. 'li(O-i-pr)4. and TiCI,, respectively. Acidic color of the indicator was observed distinctly ( + ), slightly (+), or not at all ( - ) on the surface.

*

considered from the catalytic activity that the catalyst treated at 500°C bears the surface acidity close to Ho = - 13 (4). Among the superacids we have synthesized so far, the ZrO, catalyst showed the highest acid strength as well as the highest activity. ZrO, is known to hold basic sites together, with acidic ones on the surface (143, 144). The generation of highly superacidic sites might be due to the acid-base bifunctional nature of ZrO,, especially its basic nature. 6. Surfucc. Area. Specific surface areas of the catalysts are shown in Table VIII, together with those of pure metal oxides (129, 135, 140). It is noteworthy that the areas of the catalysts are much larger compared with those of the oxides without the sulfate treatment, except for the A1,0, catalyst; in particular, the large increase in surface area was observed on the highly active and acidic catalysts. The XRD spectra of the SnOzcatalysts showed the degree of crystallization of the sulfated SnO, to be much lower than that of SnOz without the sulfate treatment (145). The analogous phenomena were also observed with the ZrO, and TiO, catalysts (129). Amorphous hydrated zirconium oxide crystallizes at 410"C, and the ZrO, heat treated at 500°C consists of

188

KAZ U SH 1 ARATA

TABLE V l l l Surfircr Arerrs of'the Cutulvsts Surface area (rn'/g) Catalyst ZrO?-l

Zr02-lI TiOz-l

Calcination temp. ("C) 500 650 800 575 650 500 525 600

With SO:-

Without SO:-('

I x7 I24 41

I00 50 28 64 44 x7 63 55 75 71 29 28 21 250 253 149

I36 84 I12 I44 I00 I I7

TiO?-ll

500

525

90

SnO?

500

I47 I66 135 I61 15 I I10

A120,- I A1103-2 AI,O,-3

550 600 650 650

650

Metal oxides without the sulfate treatment.

a mixture of tetragonal and monoclinic forms: the former converts to the latter form by calcination over 5OO0C, and only the monoclinic form is developed after calcination over 700°C (f29,146).The XRD spectra of the ZrO,-I catalysts are shown in Fig. 6. The degree of crystallization of SO4/ ZrO, calcined at 500°C is almost the same as that of ZrO, treated at 350°C. The XRD pattern of SO,/ZrO, heated at 650"C, whose superacidity was highest, was completely a tetragonal form; the pattern of the sulfated material calcined at 800°C was almost coincident with that of ZrOZtreated at 650°C. Figure 7 shows the spectra of the Ti0,-I catalysts. The crystallographic phase transformation of TiO, is known to be an anatase form to a rutile one by calcination. It is found that the temperature of crystallization or phase transformation of the anatase to rutile form for SO,/TiO, is -200°C higher than that for pure TiO,: the pattern of SO,/TiO, treated at 525°C most active among the TiOz catalysts, was a pure anatase system. Measurement by scanning electron microscope of the ZrO, catalysts was performed; the micrographs are shown in Fig. 8 (56). It is seen that the samples with the sulfate treatment were cracked into fine particles in comparison with those of the substances without the sulfate treatment.

I89

SOLID SUPERACIDS

Tetraqonal

(B)

1 1 1 1 1 1 30

40

2 e/deg

50

30

40

50

2 e/des

FIG.6. XKD profiles of SO,/ZrO,-l and ZrO,-l. Calcination temperature: ( A ) 500°C. ( B ) 650"C, (C) 800°C.

This must be a reason, as well as the retardation of crystallization, why specific surface areas of the catalysts are much larger than those of the oxides, which have not undergone the sufate treatment. Specific surface areas of the A1203 catalysts were much smaller than those of the oxides without the sulfate treatment, as shown in Table VIII. These catalysts were prepared from the crystallized oxides, and highly active catalysts were obtained by the treatment on the crystallized oxide rather than on the amorphous one. It is considered that the large difference in the surface area before and after treatment is brought about by the different method of catalyst preparation (/do). c . Note. There are several helpful details concerning the preparation of the superacids under discussion (147, 148). ( I ) The catalyst is usually

I90

KAZUSHI ARAT A

TIO,-I -

L. Anatase

u 20

30

40

2 e/deg

u 20

30

40

2 e/deg

Fiti. 7. X K D profiles of SO,/TiO,-I and TiOl-I. Calcination temperature: (A) 300°C. (B) S00"C. (C) S2S°C, (I)) 600°C.

calcined in a Pyrex tube and sealed in an ampule until use, to avoid humidity. The appearance of the catalyst with the sulfate treatment differs greatly from that without the treatment. With treatment, the catalyst is a finely powdered solid that coats the wall of the glass ampule, whereas without treatment, this is not the case. This is a way to confirm whether superacidity has been generated. ( 2 ) The catalyst acquires the hue of the deactivated color indicator when heated in a vacuum at temperatures above 250°C; in particular, the Fe,O, superacid should be evacuated at temperatures below 100°C. (3) The catalyst obtained by the treatment with sulfuric acid is usually higher in activity than that obtained with ammonium sulfate treatment. (4) The superacids of Fe,O, and SnO, show oxidizing

SOLID SUPERACIDS

FIG.8 . AEM micrographs of ZrO?-I (A) and SO,/ZrOz-I ( B ) calcined at 650°C.

191

I92

KAZUSHI A R A T A

action at temperatures above 100°C. The oxidizing effect of those catalysts is described later. ( 5 ) The catalysts obtained from isopropoxide of Ti and nitrates of Fe and Zr as starting materials are high in activity and easy to prepare.

3.

Ti(SO,), und Zr(SO,),

It is assumed from the generation of superacidity on Fe,O,, TiOz, and ZrO, and the catalytic action of iron sulfates heated at high temperature that titanium and zirconium sulfates generate high acidity when heated at elevated temperatures; the results agree with the expectation (149). The catalytic activities of Ti(SO,), and Zr(SO,), for the reaction of cumene are remarkably dependent on calcination temperature; the maximum activity was observed with calcination at 625°C for Ti(SO,), and 725°C for Zr(SO,),. The latter material showed an acid strength of - 13.16 < Ho I - 12.70 and activity for the reaction of pentane. XRD analysis showed Ti(SO,), calcined at 625°C and Zr(SO,)Zat 725°C to be a mixture of their crystallized oxides and sulfate forms.

Zr 3d3/2, 512

0 Is __

536

532

528

188

Binding energy,

184

180

eV

FIG. 9. XPS spectra of ZrOz catalysts. (A) Zr(SO,)?; ( B ) S0,/Zr02-I (6SO"C): (C) Z r 0 2 - l (650°C).

193

SOLID SUPERACIDS 0 1s -

A

A A

T i 2p1/2, 3/2

(A)

(B)

(C)

u534

530

467

B i n d i n g energy,

463

459

eV

FIG. 10. XPS spectra of TiO, catalysts. ( A ) Ti(S0,):; ( B ) SO,/TiOl-I (525°C): (C) TiOz-l (525°C).

B. STRUCTURE OF ACIDCENTERS Experiments using XPS were carried out in order to elucidate the surface property; the spectra of the ZrO, catalyst are shown in Fig. 9 (150, 151).The spectra of Zr 3 4 , , Zr 3d5,,, and 0 I s for SOJZrO, treated at 650°C were consistent with those for ZrO,, i.e., 184.1, 182.0, and 530.3 eV binding energy (BE), respectively. The former sample also showed a shoulder peak at 532 eV, which was similar to that of Zr(SO,),. The single peak of S 2p7,?was observed at 169.3 eV, whose value agreed with that of Zr(SO,),. Thus, it is concluded that the surface is composed of ZrO, and SO,. The XPS spectra of the TiO, catalyst are shown in Fig. 10 (152). The spectra of Ti 2 p , / ,and Ti 2p3,,for S0,/Ti02treated at 525°C were consistent with those for TiO, and Ti(SO,),, 464.6 and 458.8 eV BE, respectively, whose values also agreed with those in the literature (153). The main peak at 530.1 eV for SO,/TiO, is assigned to be the oxide oxygen; a shoulder peak around 532 eV is attributed to the sulfate oxygen, BE of S 2p being 168.6 eV. Thus, the surface appears to be composed of TiO, and Ti(SO,),. In the case of the A120, superacid, the XPS spectra also showed the possibility of bearing A1,03 and AI,(SO,), on the surface (140).

194

KAZUSHl ARATA

1500

1300

1100

900

Wave number, cm-I

(B), SO,/TiO,-I (S25"C) FIG.1 I . IR spectra of S04/ZrOz-l(650°C) (A), SO,/ZrO~-l(X0O0C) (C), and S0,/Sn02 (550°C) (D).

The IR spectra of the superacids of Zr02, TiO,, and SnO, are shown in Fig. 11. The samples showed the spectra to be different from those of metal sulfates (shown in Fig. 3); all the materials showed absorption bands at 980-990, 1040, 1130-1 150, and 1210-1230 cm-', which are assigned to the bidentate sulfate coordinated to metal elements (154). The IR spectra of pyridine adsorbed on SO,/Zr0,-1(6SO0C) are shown in Fig. 12, where both the pyridinium ion (at 1540 c m - ' ) and coordinately bonded pyridine (at 1440 cm-') are observed, the decrease in the former band being seen together with the increase in the latter band after evacuation at high temperature (spectrum C) (150, 151). Spectrum D shows the changes that occur upon the addition of water to the sample; the increase in the 1540-cm-' band indicates that a considerable amount of Bronsted acid has been formed, and the decrease in the 1440-cm-' band shows a

I95

SOLlD SUPERACIDS

I

1600

I

I

I

1500

I 1400

Wave number, cm-I

FIG. 12. IR absorption spectra for pyridine on S0,/Zr02-l (6SO"C). (A) Background; (€3) equilibrated with pyridine, evacuated at 150°C; (C) after evacuation at 350°C; (D) after exposing to H?O (3 mm Hg)and evacuating at room temperature.

concomitant decrease in Lewis acidity. The results show the easy conversion of Lewis sites to Bronsted sites by water molecules. Analogous results were also observed with the TiO, superacid (56). The relation between the preheating temperature of SO,/ZrO,-1 in a reactor under the pulse reaction conditions before reaction and activity for the reaction of butane was examined; the results are shown in Fig. 13. The maximum activity was observed with heating at 35OoC, and the activities decreased by heating at 400-600°C. It is considered that Bronsted sites, created by adsorption of water on Lewis sites, were decreased by heating at the high temperatures in the He flow, thus the reaction of butane catalyzed by Bronsted acid was restrained. In fact, the activities on heating at 500 and 550°C were raised to those around 300-400°C by regeneration of the Bronsted sites (A' and B' in Fig. 13), where the catalyst was treated by moistening with water at 120°C after the reaction, followed by heating again in the He flow at 300°C for I h and performing the reaction at 130°C. It is indicative that Lewis and Bronsted sites on the present catalysts are easily changeable by adsorption or desorption of water molecules, the reaction of butane being also catalyzed by the Bronsted site (150, f5f).

I96

KAZUSHl ARATA

From the above results, the surface structure appears to be SO, combined with Zr elements in the bridging bidentated state, as Okazaki et d . proposed in the case of titanium oxide with sulfate ion (155, 156). The double-bond nature of the complex is much stronger compared with that of a simple metal sulfate; thus, the Lewis acid strength of Zr4+becomes remarkably stronger by the inductive effect of S = 0 in the complex, as illustrated by arrows in the previous scheme. If water molecules are present, the Lewis acid sites are converted to Bronsted acid sites (129, 151, 157). Using 1R spectroscopy, Tanabe ef al. proposed a structure for the acid sites on sulfur-promoted iron oxide to be chelating bidentate complexes (158-160).

Saur and others studied the structure of sulfated alumina and titania and postulated that in the absence of water three oxygens of the sulfate are bonded to A1 or Ti, whereas in the presence of H,O this is converted to a bridged bidentate sulfate, thus accounting for the increased Bronsted acidity (161). M-0 \ M-O-S=O / M-0

+ H20

M-OH M-0

\ /OH

s

M-0

/ No

M-0

M-0

\ Sa. 0- H +

/

0

Recent Laser Raman spectroscopic studies of sulfated zirconia indicate a monodentate structure (162).

c.

REACTIONS

CATALYZED

BY SUPERAClDS

About 10 years have passed since this study began to be seriously undertaken, but the usage of solid superacids as catalysts is still limited. Table IX summarizes the acid-catalyzed reactions on sulfated metal oxides, i.e., cracking, isomerization, alkylation, acylation, esterification,

I97

SOL1 D S UPERACI DS

be a, 8 0

43 4

n

v) 0

.-

40 2

0 8

.-

m L a, >

8

0 0

Heating temperature o f c a t a l y s t ,

"C

FIG. 13. Catalytic activities of SO,/ZrO?-I (650°C) preheated at various temperatures in the He flow before reaction (butane at 130°C). ( O ) ,Heat-treated at various temperatures. (0). Exposed to water by injection at 120°C and treated at various temperatures. (A' and B'), After the reaction of butane ( A and 5 ) . water ( I PI) was injected at 120°C followed by heating the catalyqt at 300°C for I h and performing the reaction under the same conditions.

polymerization, oligomerization, and oxidation, and illustrates the range of reactions that occur. In the reaction of pentane in the vapor phase, the reaction is known to occur in the consecutive steps of pentane -+ isopentane 3 isobutane. Thus, in order to suppress the reaction of isopentane -+ isobutane for obtaining better isopentane selectively, the reaction should be carried out with a short contact time and at a low temperature using a highly acidic catalyst (163). Cyclopentane was converted to propane, butane, isobutane, pentane, and isopentane, and neopentane was converted to methane, ethane, and propane (129). In the alkylation of isobutane with butenes, the ZrO, superacid was highly active for the reaction in the heterogeneous system, the yield of alkylates being almost analogous to the case of H,SO, in the homogeneous system (165). 2,2,3-Trimethylpentane, whose octane value is high, was produced more than was found in the case with H,SO, and HF. Homogeneous reactions using H2S0, and AICI, in industrial processes give rise to many problems that must be solved, for instance, corrosion of the reaction vessels or reactors and difficulties in treating the acids after reaction because of environmental problems. In particular, the disposal of liquid acid catalysts such as H,SO, requires expensive treatment to make

I98

KAZUSHI A R A T A

TABLE 1X Reuctions Cutulyzed by Solid Superucids Type of reaction Cracking

Catalyst SO4/MeO,

Reaction Propane + ethane + methane Pentane i-butane Pentane + propane + i-butane Cyclopentane + i-butane Neopentane methane + propane Butane + i-butane -+

---f

Isomerization

Al kylation

Acylation

Pentane -+ i-pentane Methane + ethylene -+ C1-C7 hydrocarbons ;-Butane + butenes + C,-C,, alkylates o-Xylene + styrene -+ phen ylxyl ylethane Toluene + PhCOCl -+ o - , m-,pmethylbenzophenone + HCI Toluene + (PhCO),O+ methylbenzophenones + PhCOOH Toluene PhCOOH -+ methylbenzophenones + H,O Toluene + CHtCOOH+ methylacetophenones + H,O Chlorobenzene + o-Chlorobenzoyl chloride C, C,-OH + CH3COOH + esters n-Octyl-2-ethylhexyl alcohol + terephthalic, phthalic acid + Ethanol + acrylic acid + Methanol + salicylic acid +

+

Esterification

Polymerization

-

Ethyl, methyl vinyl ether poly(ethy1, methyl vinyl ether) Oligomerization I-Octene, I-decene, P-pinene -+ dimer, trimer Oxidation Butane, i-butane + CO + C 0 2 Ethylene + H,O + acetaldehyde acetone Cyclohexanol --+ cyclohexanone -+

+

ZrOz SnO, Zr(S04h ZrO, ZrOz Fe,03, TiOz ZrOz, HfO,, SnO? ZrO, ZrO,

References 132 135 149 129 129 4 , 130 131, 138 I63 I64

ZrO?

165

ZrOz

166

A1203,TiOz, Zr02, HfOz AlzO,, ZrO, ZrOz ZrOz

140, 167, 171, 172 140, 168, 169, 170 168, 169, I 70 168, 169, I 70

ZrO, 171, 172

Fe,03

I 73 168, 169, I 74 168, 169, 174 168, 169, I 74 I 75

TiO,, ZrO,, SnO,

145, 176

ZrO, Ti02, ZrO, TiOz. ZrOz TiOz. ZrO,

Sn02

177 145

SnO,

145

Fe203

SOLID SUPERACIDS

199

disposal environmentally safe. Hence, the application of solid catalysts instead of H,SO, and AICI, in various chemical processes is highly desirable. It appears that the present superacids can be used, considering the acid strength. Two reactions were chosen for the investigation, i.e., acylation and esterification. In Friedel-Crafts acylations the reaction is generally performed with an acid chloride as the acylating reagent and with AICI, as the catalyst; the catalyst melts and becomes inactive owing to its coordination to the carbonyl group of the product as the reaction proceeds. The acid chloride is synthesized from the corresponding acid; thus using the carboxylic acid directly as the acylating agent could be advantageous. Acid sites stronger than Ho - 15 were needed for the formation of the acyl cation (RCO') from a carboxylic acid; the reactivity with acylating reagents was (PhCO),O > PhC0,H > PhCOzEt > PhC0,Me (168, 169, 170). It was found that the present catalyst could be used in acetylation with carboxylic acids in the vapor phase. In esterification, repeating the reaction of ethanol with acetic acid with the used catalyst was examined; no reactivation of the ZrO, catalyst was needed because the catalytic activity remained unchanged for the repeated operation (173).However, the TiO, catalyst was deactivated by the esterification of phthalic acid with n-octyl alcohol because of the elimination of sulfate ion on the surface, the activity of ZrO, remaining unchanged for the repeated operation without the sulfate elimination (168, 169). The discrepancy in the preparation temperature of TiO, and ZrOz catalysts is 150°C,and it is considered that the catalyst prepared at a high temperature is more stable and effective as a solid catalyst. The Fe,O, superacid was found to be quite effective for oxidation of hydrocarbons to CO and CO, when the reaction was performed at temperatures above 100°C. The catalyst gave a 29% conversion for the reaction of butane at 300°C to form CO and CO, in the ratio 4 : 6 under the conditions in which none of the reactions occurred at 300°C over Fe,O,, without the sulfate treatment (177). The decrease in oxygen of the catalyst surface was observed together with the complete recovery of activity by supply of 0,. The catalyst was entirely poisoned by the addition ofpyridine, the oxidation being related to the surface acidity. The activity enhancement of oxidation by the sulfate addition was also observed with the SnO, superacid (135, 145). Iron and tin oxides are known to be oxidation catalysts; thus those superacids would be the oxidation catalysts with superacidity.

D. APPLICATION OF THE PRESENT CATALYST SYSTEM Selenium and tellurium belong to the same group as S in the Periodic Table, and their oxygen compounds, analogous to sulfuric acid, are called

200

KAZUSHI ARATA

Products Catalyst SeO,/ZrO,

Cr,O, CrzO1/ZrO,

Pt/AI?O, RhlA120, Pt-Re/Al,O,

Calcination temp. ("C) 600 650 700 7.50 700 600 700 800 900 700 700 700

Benzene

By-products

27 38 34

9 12 9

16

7 0.3 I0 II 8 4

0.2 48 52 41 22 3 36 9

35 46

25

oxoacids. The preparation method of the catalyst was applied to selenate and tellurate ions; Zr(OH), was exposed to a 0.05 M aqueous solution of selenic (H,SeO,) or telluric (H,TeO,) acid followed by calcination in air. The materials caused propan-2-01 to convert to acetone with 100% selectivity by oxidative dehydrogenation (178, 1 79). The catalysts also converted CO to CO,. It is of interest that treatment with sulfate ion followed by calcination creates superacid sites on the ZrO, surface, whereas treatment with selenate or tellurate ion produces dehydrogenation catalysts without any acidic action. The SeO,/ZrO, and Te0,/Zr02 catalysts were reduced with H2 and then were used for the reaction of hexane; the results are shown in Table X (180). The SeO,/ZrO, catalyst was quite effective for the dehydrocyclization of hexane to benzene; high activities were observed on calcination at 600-700"C, the selectivity being up to 84%. Reforming catalysts, Pt/AI2O3, Rh/Al,O,, and Pt-Re/Al,O,, were much lower in selectivity; these results probably reflect the fact that the reduction was carried out at high temperature (550°C) and in the absence of hydrogen. XPS experiments showed the surface structure to be Se combined with Zr elements as Se'-. in the bridging bidentate state, Zr-Se-Zr (180, 181). A dehydrogenation catalyst for alkanes was also obtained by exposing Zr(OH), to 0.05 M (NH,),CrO, followed by calcination in air at 600-800°C and reduction at 550°C (quantity of Cr, 0.5 wt% after reduction); this catalyst converted hexane to benzene with a selectivity of up to 84%, as shown in Table X (182). The yield of benzene at 550°C was steady up to

SOLID SUPERACIDS

20 I

6 h with 76-77% selectivity in a flow system at atmospheric pressure. Specific surface area of the catalyst calcined at 700°C was 57 m2/g;that of ZrO, without the chromate treatment was only 15 m2/g.This large increase in area is similar to that of the sulfate superacid. XPS of the catalyst showed the active site for dehydrogenation to be Cr4’, created by reduction of CrO,/ZrO,: 578.3 eV for the binding energy of Cr 2 p j n , intermediate between the values of CrO, (580.2 eV) and Cr,O, (576.8 eV) (182). The Fe,O, catalysts treated with tellurate ion are effective for the selective conversion of ethanol to acetone (183). The synthesis of acetone from H 2 0+ Me,CO CO, 4H,) is of interest from the ethanol (2EtOH point of view of using “biomass” as a chemical resource. The loading of Pt on the SO,/ZrO, catalyst showed high performance in the skeletal isomerization of alkanes; the loaded catalyst gave a steady yield of isopentane from pentane up to 100 h, with an equilibrium mixture of isomers (184, 185). The catalyst also produced gasoline substances, with a higher octane value, from light naphthas (186). It was proposed that the Pt/SO,/ZrO, catalyst was prevented from deactivation by the continuous formation of Bronsted acid sites, caused by interaction between hydrogen molecules and the strong Lewis acid sites (187).

+

+

+

VII. Superacids by Metal Oxides Although the sulfate superacids are stable enough because of preparatory heat treatment at elevated temperatures, elimination of the sulfate is sometimes observed during reaction as a result of catalyst deactivation, especially in a solid-liquid system. It is hoped to synthesize superacids with the system of metal oxides. We have succeeded in preparing another type of superacid, not containing any sulfate ion but consisting of metal oxides, which can be used at temperatures over 800°C (188-192). The catalyst was prepared as follows. Zr(OH), was impregnated with aqueous ammonium metatungstate [(NH,),(H,W,,O,,) - nH,Ol followed by evaporating the water, drying, and calcining in air at 600-1000°C. The concentration was 15 wt% W based on the hydroxide and 13 wt% W after calcination at 650-950°C. The analogous superacid was also formed by the kneading method with tungstic acid (H,WO,), which is insoluble in water. The catalysts were quite effective for the benzoylation of toluene with benzoic anhydride and for the reaction of pentane; the maximum activity was observed with calcination at surprisingly high temperatures of 800-850°C for both reactions. The Si02-AI,0, catalyst was totally inactive under the same conditions. The catalyst treated at 800°C was active for

202

KAZUSHI ARATA

I

20

I 30

1 40

I 53

I 60

2 e/dee

FIG.14. XRD profiles of WO,/ZrOz catalysts. (A) WO1 (700"C), ( B ) WO,/ZrO? (800°C). (C) WO,/ZrOz ( I0OO"C): ( D) WO,/ZrO, prepared by calcining Zr(OH), at 700°C. impregnating with the tungstate, and calcining at 700°C.

isomerization of butane at 50°C and for pentane at 30°C and XPS showed this catalyst to be WO, supported on ZrO, [W0,/Zr02(80WC)] (192). The acid strength of this catalyst was estimated to be Ho 5 - 14.52 by a color change method using Hammett indicators. The catalysts prepared by heating Zr(OH), at 100-300"C, then impregnating each with the tungstate and finally calcining at 800°C, gave almost the same activities, whereas the activities decreased greatly with calcination of Zr(OH), at temperatures over 400°C, the crystallization temperature of ZrO,, as was observed with the sulfated superacids. XRD measurement of the catalysts was performed; the spectra are shown in Fig. 14. The present catalysts show the crystallographic phase transformation of ZrO, to be completely a tetragonal form when calcining up to 900°C and a monoclinic form after calcination at 1000°C. The catalyst prepared by

SOLID SUPERACIDS

203

precalcining Zr(OH), at 700°C to ZrO,, followed by impregnating and calcining at 700"C, showed the pattern to be a monoclinic system in addition to the crystallized WO, (Fig. 14D), while the materials prepared similarly from Zr(OH),, dried at IOWC, gave 100% tetragonal forms without the crystallization of WO,, from 600 to 800-850°C of calcination (Fig. 14B); the pattern of the former was almost coincident with that of the latter material calcined at 1000°C (Fig. 14C), whose catalytic activity was quite low. Therefore, it is concluded that superacid sites are not created by impregnation on the crystallized oxide, but on the amorphous form, whose calcination then converts to the crystalline form; i.e., tungsten oxide combines with zirconium oxide to create superacid sites at the time when a tetragonal system is formed (188, / 9 0 ) . This preparation method of catalyst was applied to molybdenum oxide; Zr(OH), was impregnated with molybdic acid (H,MoO,) dissolved in ammonia water followed by evaporating the water, drying, and calcining in air (191). The concentration was 5 wt% Mo metal based on the hydroxide. The catalyst was effective for the benzoylation of toluene with benzoic anhydride, which did not occur over SiO,-AIZO,; high activities were observed on calcination at 750-800°C. Because the catalyst was colored (yellowish green), the acid strength was not estimated by the visual color change method using the Hammett indicators. It is considered that the catalyst bears a surface acidity higher than Ho = - 12.70 (-- - 13),judging from the reaction results. XPS spectra of the sample treated at 800°C [Mo03/Zr0,(800"C)] were consistent with those of Moo, and ZrO, ( / 9 2 ) . Superacid sites were not created by impregnation of the molybdate on the crystallized oxide, but rather on the hydroxide, as was observed in the case of the WO, catalyst. The XRD pattern of the inactive material prepared from the crystallized oxide was also completely different from that prepared from the hydroxide, as shown in Fig. 15, i.e., monoclinic for the former and tetragonal for the latter ( / 9 / ) . Specific surface areas of the WO, and MOO, catalysts are shown in Table XI. It is noteworthy that the areas of both catalysts are large compared with those of the oxides without tungsten, and compared to molybdenum oxides, as was observed with the sulfate superacids, especially in the case of calcination at 800°C. W, Mo, and Cr belong to the chromium group in the transition metals of the Periodic Table. It is predicted that ZrO, is enhanced in acidity by the addition of CrO, up to superacidity, but CrO, has not been found to be the third additive, the relative activity of WO,/ZrO,, MoO,/ZrO,, Si02-A120,, and CrO,/ZrO, for the oligomerization of I-decene, one of the acid-catalyzed reactions, being 30 : 20 : 11 : 2 (56). The SO,/ZrO,-I catalyst with the highest acidity, prepared by calcination

204

KAZUSHI ARATA

I

I

20

30

I 40 2

I 50

I 60

I 70

e/des

FIG.IS. XRD profiles of MoO,/ZrOz catalysts. ( A ) MoO,/ZrO, prepared by impregnation of the molybdate on the crystallized ZrOz (calcined at 700"C), followed by calcination at 800°C; (€3) MoO,/ZrO, (800°C).

at 65OoC, was shown to contain 2.2 wt% S . This value is estimated to be the amount of sulfur that, as the sulfate, strongly interacts with ZrO, in monolayer, and an excess of the sulfate is decomposed to form sulfur oxides. In fact, fuming gas is observed while calcining the catalyst. The 2.2 wt% S is equivalent to 13 wt% Wand 6.6 wt% Mo, these values being just equal to the supported quantities of the present catalysts. Sublimate based on an excess of molybdenum oxide was also observed with calcination of the MOO, catalyst.

Surface area (m?/g) Calcination temp. ("C) 600 700 800 900

WO,/ZrOz

MoO,/ZrO?

ZrO?

44

68 60

34

3')

35 30

58

15 6

7

2

SOLID SUPERACIDS

205

VIII. Aluminum Halides Supported on Alumina Okazaki and co-workers have prepared a superacid of AlF,-supported alumina by treatment of A1,0, with CF,CI (193-198). They found that AI,O,-Cr,O, treated with chlorofluoromethanes, especially CF,CI, promoted the conversion of CH,OH to olefins and that the A1 component was essential to the catalytic activity for the conversion, the F component being predominantly bound to Al, not to Cr ( I93). A1,0, was then submitted to the surface treatment with CF,CI at 400-420°C in a circulation reactor; the material was active for isomerization of paraffins such as butane, pentane, and hexane at temperatures as low as 0"C, and the surface acidity was estimated to be - 14.52 < Ho 5 - 13.75 (194). The XPS analysis showed the catalyst to be AlF, formed on the A1203,but a rather unstable AIF, on the less-crystallized A120, structure. The catalyst was highly active for reactions of benzene with compounds containing the CF, group, i.e., CF,CH=CH,, CF,COCF,, CF,CFOCF,, and CF,CHO (195). The same treatment was carried out with H-mordenite and gave appreciable enhancement of activities for isomerization of xylenes ( I96), alkylation of chlorobenzene with CH,OH (197), and isomerization of o-chlorotoluene (198). Ayame and co-workers recently prepared a superacid of chlorinated alumina (199-203). A120, was chlorinated by heat treatment with CI, gas at temperatures above 800°C in a circulation reactor; the material was active for isomerization of paraffins such as butane, pentane, and cyclohexane (201).The chlorinated alumina showed a surface acidity due to the Lewis type of Ho 5 - 14.52 (202, 203). Drago and Getty have prepared solid acid catalysts of AIC1,-metal oxides (204). The catalyst was prepared by reacting a metal oxide (SiO,, AI,O,, B,O,, Ti02, or MgO) with aluminum chloride (Al2CI6)in refluxing carbon tetrachloride. The catalysts converted n-hexadecane to propane, butane, and isobutane at IO0"C. IX. Conclusion and Prospects

1 have attempted to present the recent works on syntheses of solid superacids and their catalytic action, together with some information on their surface structure. Solid acid systems stronger than the acidic oxides, such as silica-alumina and zeolites, which were used extensively as catalysts in the past, have been developed recently and are categorized as solid superacids. However, those superacids are much inferior in acid strength compared to liquid superacids. The Hammett acidity function of

206

KAZUSHI ARATA

“magic acids” such as HS03F-SbF, and HF-SbF, reaches about - 2 5 , which acidity is difficult to estimate at present, being more than 1O’times stronger than the solid superacids already mentioned. Solid catalysts can be used at elevated temperatures, though their acidities are much weaker than those of liquid ones. From this point of view, solid superacids based on Lewis acids and liquid superacids discussed in Sections l L 1 V are not sufficiently stable; Nafion-H is also unsatisfactory, its maximum operating temperature being below 200°C. A new type of the sulfate-supported metal oxides is more stable because of preparatory heat treatment at high temperatures, but elimination of the sulfate is sometimes observed during reaction, thus it is hoped to synthesize superacids with the system of metal oxides. Another type of superacid, tungsten or molybdenum oxide supported on zirconia, has been prepared by a new preparation method, and its stability is satisfactory so far. It is hoped that t h e preparation method will be extensively applied to other metal oxides for new solid superacids. There are many reactions in which solid superacids might be expected to perform as effective catalysts. The convenience of using solids in place of corrosive liquids will undoubtedly provide an incentive to further studies in the use of solid superacids. 1 hope this article has indicated some of the stimulating aspects for the investigation of solid superacid catalysts. REFERENCES 1 . Tanabe, K., “Solid Acids and Bases, ‘Their Catalytic Properties.” Tokyo Kodansha,

Academic Press, New York, 1970. 2 . Tanabe. K., in “Catalysis-Science and Technology” ( J . R. Anderson and M. Boudart, eds.). Vol. 2. Springer-Verlag. Berlin, 1981. 3 . Tanabe, K., Proc. Int. Congr. Cotiil., 9th. Culgur?,5 , (1988). 4 . Hino, M., and Arata, K., Clwt~r.Lett. p. 1259 (1979). 5 . Gillespie, K. J.. ACT. Che,n. Res. 1, 202 (1968). 6. Gillespie, R. J . , and Peel. T. E., A h . Phvs. Org. Chc.tn. 9, 1 (1972). 7. For an early review see Olah, G . A.. Angcw. r/7l’tn., Int. Ed. ERR/.12, 173 (1973). 8. Olah, G. A., Prakash, G. K. S., and Sommer. J . , Science 206, 13 (1979). 9. Olah, G . A., Prakash, G. K. S., and Sommer, J.. “Superacids.” Wiley, New York, 198.5.

10. Tanabe. K.. and Noyori, R.,in superac acid-Superbase," p. 5. Kodanasha Sci. Tokyo, 1980. If. Kemp, J . D., U S . Pat. 3,852,371 (1974). 12. Kemp, J . D., U . S . Pat. 3,979,476 (1976). 13. Bloch, H. S., Br. Pat. 1,389,237 (1972). 14. Kodewald, P. G . , and Whitehurst, D. D.. U.S. Pats. 3,925,194, 3,925,251 (1975). 15. Oelderik, J . M . , U.S. Pat. 3.394,202 (1968). 16. Oelderik, J . M., Mackor, E. L., Plattecuw, J . C., and van der Wiel, A,, U.S. Pat. 3,201,494 (1965). 17. Oelderik, J . M.. Can. Pat. 794.400 (1968).

SOL1D S U PERACIDS

207

18. Bloch, H. S . . U.S. Pat. 3,678,120(1972). 19. Olah, G. A . . U.S. Pats. 3,708,553, 3,766,286 (1973). 20. Gardner, L. E., U.S. Pats. 3,864,425, 3,878,261 (1975).

21. 22. 23. 24.

Crathorne, E. A,, U.S. Pat. 3,975.299 (1976). McCaulay, D. A., US. Pat. 4,035,286 (1977). Rodewald. P. G . , U.S. Pats. 3,962,133. 3,984,352 (1976). Magnotta, V . L., Gates, B. C., and Schuit, G. C. A.. J.C.S. Chern. Comnrrn. p. 342 (1976). 25. Magnotta, V. L., and Gates, B. C., J . Polym. Sc,i., P ~ / V t r iCliem. . 15, 1341 (1977). 26. Magnotta. V . L., and Gates, 9. C., J . C u t d . 46, 266 (1977). 27. Fuentes, G. A., and Gates, B. C., J . C a t d . 76, 440 (1982). 28. Dooley. K. M., and Gates, B. C., J . C(rtcr/.96, 347 (1985). 29. Tanabe, K., and Hattori, H., CIiern. Lett. p. 625 (1976). 30. Takahashi, 0.. Yamauchi. T . , Sakuhara, T., Hattori, H . , and Tanabe, K., B d I . Chrrn. Six,. J p n . 53, 1807 (1980). 31. Takahashi. O., Ph.D. Thesis, Hokkaido Univ., 1980. 32. Hattori. H., Takahashi, 0.. Takagi, M., and Tanabe, K., J . C{rtd. 68, 132 (1981). 33. Takahashi, O., and Hattori, H., J . C a r d . 68, 144 (1981). 34. Namba, S.. lhara, K., Sakaguchi. Y ., and Yashima, T., J . Jpn. Per. Inst. 23, 142 (1980). 35. Krzywicki, A . . Marczewski, M., and Malinowski, S., Reuct. Kinrr. Crrtd. Lctt. 8, 25 (1978). 36. Croft, R. C.. Airst. J . Clicrn. 9, 201 (1956). 37. Sudo, M., Ichikawa, M., Soma, M., Onishi, T., and Tamaru, K.. J . Phvs. Cliein. 73, 1174 (1969). 3X. Lalancette. J . M., and Lafontaine, J., J.C.S. C l i c ~ iCornm/m. . p. 815 (1973). 39. Laali, K . , Muller. M., and Sommer, J., J . C.S. Cliern. C(J/n/rI/4n.p. 1088 (1980). 40. Heinermann. J. J. L., and Gaaf, J . , J . Mol. Ctrtrrl. 11, 215 (1981). 41. Laali, K., and Sommer, J., Nouu. J . Chirn. 5, 469 (1981). 42. Le Normand, F., Fajula, F., and Sommer. J.. Nortu. J . Chirn. 6 , 291 (19x2). 43. Le Normand. F., Fajula, F., Gault. F., and Sommer, J., Nocru. J . Chirn. 6,411 (1982). 44. Le Normand, F., Fajula, F.. Gault, F., and Sornmer, J . , N O N UJ. . Cliirn. 6,417 (1982). 45. Yoneda. N., Fukuhara, T., Abe, T., Suzuki, A,, and Kudo, K . . Chmn. Lrit. p. 1485 (1981). 46. Olah, G. A., Felberg, J . D., and Lammertsma. K.. J . A m . Cliern. So(.. 105,6529 (1983). 47. Olah, G. A., and Kaspi, J . , J . Org. Chern. 42, 3046 (1977). 48. Lalancette, J . M., Fournier-Breault, M. J . , and Thiffault. R., Cun. J . Chrm. 52, 589 (1974). 49. Olah, G. A.. Kaspi, J., and Bukala, J., J . Org. C/zern.42, 4187 (1977). 50. Ono, Y., Tanabe, T., and Kitajima, N., Chem. Lett. p. 625 (1978). 51. Ono. Y., Sakuma, S . , Tanabe, T., and Kitajima, N., Chern. Lett. p. 1061 (1978). 52. Ono, Y.. Yamaguchi, K., and Kitajima, N., J . Catcrl. 64, 13 (1980). 53. Ono, Y., Tanabe, T., and Kitajima. N., J . Curd. 56, 47 (1979). 54. Ono, Y., Kitajima, N., Nakajima, T., and Keii, T., Proc. I n / . Congr. Crrttrl., 7th, 7 o k y o B23 (1980). 55. Product information by E. I . DuPont de Nemours & Co., Plast. Prod. Resins Dept., D-13114 Wilmington, DE 19898 (1981). 56. Hino, M., and Arata, K., unpublished observations. 57. Hasegawa, H., and Higashimura, T., Polyrn. J . 12, 407 (1980). 58. Olah, G. A . , and Meidar, D., Nouu. J . Chirn. 3, 269 (1979). 59. Kaspi, J., Montgomery, D. D., and Olah, G. A., J . O r g . C h m . 43, 3147 (1978).

208

KAZUSHI ARAI'A

60. Kaspi, J . . and Olah, G. A.. J . O r g . Chem. 43, 3142 (1978). 6 ) . Olah, G. A.. Meidar, D., Malhotra, R . , Olah, J . A., and Narang. S . C.. J . Catul. 61, 96 (1980). 62. Olah. G. A.,Malhotra, R . , Narang, S . C., and Olah, J. A , , Synrhc~sisp. 672 (1978). 63. Konishi. H., Suetsugu, K., Okano, T., and Kiji, J.. Bull. C'Iiem. Soc. J p n . 55, 957 (1982). 64. Kameo, T.. Hirashima, T., and Manabe, O., Nippon K u g u k u Koislri p. 414 (1983). 65. Olah, G. A., and Narang. S . C., Syrirhcvis p. 690 (1978). 66. Olah, G. A., Malhotra, R., and Narang, S . C . , J . Org. Chrm. 43, 4628 (1978). 67. Olah, G. A.. Meidar. D., and Olah. J . A,, Nonu. J . Chim. 3, 275 (1979). 68. Olah, G . A., and Kaspi, J., Nortu. J . Clrim. 2, 585 (1978). 69. Olah, G. A , , and Kaspi, J . , Norm. J . Cliirn. 2, 581 (1978). 70. Olah, G. A., Narang, S. C.. Malhotra. R., and Olah. J. A , , J . A m . Chcwr. S o c . 101, 1805 (1979). 71. Hasegawa, H., and Higashimura. T . , I-'o!vm. J . 11. 737 (1979). 72. Beltrame, P., Beltrame, P. L., Carniti. P., and Nespoli, G., I d . Ertg. C/wm.. Prod. Res. Dru. 19, 205 (1980). 73. Olah, G . A., Meidar, D., and Liang G., J . Org. Clwm. 43, 3890 (1978). 74. Olah, G . A,, and Meidar, D., S,ynthe.sis p. 358 (1978). 75. Olah, G. A., and Fung, A. P., S,vnr/iesi.s p. 473 (1981). 76. Tanaka, M.. and Sekiguchi. T.. Nippot? K t i g h t Kttishi p. 1499 (1983). 77. Piltrnan, C. U., Jr., and Liang. Y. F.. J . Org. Clrem. 45, 5048 (1980). 78. Delrnas. M.. Denis. A., Gorrichon, J . P., and Gaset, A,. Svnth. Commrin. 10, 517 ( I 980). 7Y. Olah. G. A., and Ip, W. M., Neit>J.Chcwz. 12, 299 (1988). 80. Olah, G . A.. Husain. A.. Gupta, B. G . B., and Narang. S . C.. Sy/?t/itl.sis p. 471 (19x1j. 81. Olah, G. A., Fung, A . P.. and Malhotra, K., Syrirhi>si.sp. 474 (1981). 82. Olah. G. A., Keumi, T., and Meidar, D., Svni/7c~sisp. 929 (1978). 83. Olah. G. A., Fung, A. P., and Meidar, D., Svnfliesis p. 280 (1981). 84. Olah, G . A., and Mehrotra. A. K., .Syntlic,si.s p. 962 (1982). 8s. Olah. G . A., Narang, S. C., Meidar. D., and Salem. G. F., Synthesis p. 282 (1981). 86. Olah, G. A.. lyer. P. S.. and Prakash, G . K. S., Synthesi.~p. 513 (19x6). 87. Waller, F. J . . and Warren Van Scoyoc, K., Chrm/~c.li p. 438 (1987). 88. Waller. F. J., Cuttrl. R ~ u . - S c i . Eng. 28, I (1986). 89. Madon, K. J., and Shaw, H., C(rtal. Rtw.-S~i. E i i p . 15, 69 (1977). 90. Oudar, J.. C u f d . Reu.--Sci. E n g . 22, 171 (1980). 91. Bartholomew, C. H., Agrawal, P. K., and Katzer, J . R . , Adu. Curd. 31, 135 (1982). 92. Nakajo, T.. Arakawa, H.. Sano. K., and Matsuhira, S.. C k m . L e t / . p. 593 (1987). 93. Yamada, M., Obara, T., Shindo. T.. Yan. J., and Amano. A,. A r c , / 67, 298 (1988). 94. Yarnada, M.. Saito, A., Wakatsuki. T.. and Obara, T.. C / w m . Lett. p. 571 (19x7). 9.5. Yamada, M., Obara. T., Yan. J . , and Hatakeyama, S., J . J p n . Pet. Inst. 31, I18 (1988). 96. Sugioka. M., Hosotsubo, T., and Aoniura. K., J.C.S. Chew. Commrtn. p. 54 (1976). 97. Ishinaga. Y . , Otsuka. K.. and Morikawa. A.. Bid/. C'lrem. SOC.Jprr. 52, 933 (1979). 98. Otsuka, K., Wada. Y., Tanaka, K., and Morikawa, A.. E d / . Chem. Soc. Jpn. 52, 3443 (1979). 99. Otsuka, K., and Morikawa. A,. J . Curd. 56, 88 (1979). 100. Okazaki. S . , Kumasaka. M., Yoshida. J., Kosaka, K., and Tanabe. K., /nil. Errg. Chrm., Prod. R e s . D c u . 20, 101 (1981). 1 0 1 . Chinone, T., and Okazaki, S . , Nippori Ktrptrkrr Kaishi p. 1327 (1978). 102. Takeshita. T., Arata, K.. Sam, T.. and Tanabe, K., Kopyo Kuguktc Zu.sshi 69, 916 (1966).

SOLID SUPERACIDS

209

Arata, K., Takeshita, T., and Tanabe, K., Shokuhoi 8, 226 (1966). Arata. K., Azurni, N., and Sawarnura, H., Bnll. Chem. Soc. Jpn. 48, 2944 (1975). Arata, K., and Toyoshirna, I., Chem. Left. p. 929 (1974). Arata, K., and Toyoshorna, I., Shokuhui 17,98P (1975). Arata, K., Sato, K., and Toyoshirna, I., J. Curd. 42, 221 (1976). Arata, K., Yabe, K., and Toyoshirna, I., J. Cu/ul. 44, 385 (1976). Arata, K., and Toyoshima, I., J . C u r d . 47, 109 (1977). Arata, K., Fukui, A.. and Toyoshirna. I . , J.C.S.Chem. Commun. p. 121 (1978). Hino, M., and Arata, K., Chern. Lett. p . 277 (1977). Arata, K., Hino, M.. and Yabe, K., B i d . Chem. Soc.. J p n . 53, 6 (1980). Hino, M.. and Arata, K., Chem. L e a . p. 325 (1978). Arata. K., and Hino, M.. Bull. Chrm. Sot,. Jpn. 53, 446 (1980). Arata, K., and Hino, M.. Chrm. Lett. p. 1479 (1980). Hino, M., and Arata, K., Bid/. Chcvn. Soc. Jpn. 54, 31 I (1981). 117. Hino, M., and Arata, K., Chem. Lett. p. 1141 (1979). 118. Tanabe, K., and Takeshita, T.. Adu. Cutul. 17, 315 (1967). 119. Takeshita, T., Ohnishi, R., and Tanabe, K., Cutul. Rev.-Sci. Eng. 8, 29 (1973). 120. Hino, M.. and Arata. K., J. Polvm. Sci., Pol.vm. Lett. Ed. 16, 529 (1978). 121. Hino, M., and Arata, K., J. Polym. Sci., Polyn. Chrm. Ed. 18, 235 (1980). 122. Arata, K., and Hino, M., Bull. Chcm. Soc. Jpn. 53, 535 (1980). 123. Arata, K., Yabe, K., Hino, M., and Toyoshirna, I., Shokuhui 19, 246 (1977). 124. Yabe, K., Arata, K., and Toyoshirna, I . , J. Ccrtul. 57, 231 (1979). 125. Hino, M., and Arata, K., Shokuhai21, 217 (1979). 126. Hino, M., and Arata, K., Chem. Lett. p. 477 (1979). 127. Benesi, H. A., J. A m . Chrm. Sor. 78, 5490 (1956). 128. Benesi, H. A.. J. Phvs. Chem. 61, 970 (1957). 129. Hino, M., Ph.D. Thesis, Hokkaido Univ.. 1982. 130. Hino, M., and Arata, K., J.C.S. Chem. Cominun. p. 1148 (1979). 131. Hino, M.. Kobayashi, S., and Arata, K., J. A m . Chew. Soc. 101, 6439 (1979). 132. Hino, M.. and Arata, K., J.C.S. Chem. Commun. p. 851 (1980). 133. Arata, K., Hino, M., Hisamitsu. T., and Mukai, Y., Jpn. Pat. 59-6181 (1984). 134. Arata. K., Hino, M., Hisamitsu, T . , and Mukai, Y., Jpn. Pat. 59-40056 (1984). 135. Matsuhashi, H., Hino, M.. and Arata, K., Chcjm. Lett. p. 1027 (1988). 136. Matsuhashi, H., Hino, M., and Arata, K.. Hokkuido Sittnmrr Mert. Chem. Soc,. J p n . Abstr. E l 5 (1989). 137. Tanabe, K., Itoh, M., Morishige, K., and Hattori, H., in “Preparation of Catalysts” (B. Delrnon, P. A. Jacobs. and G. Poncelet. eds.), p. 65. Elsevier, Amsterdam, 1976. 138. Arata, K., and Hino, M., Rcarr. Kine!. Critul. Lett. 25, 143 (1984). 139. Wang, G., Hattori, H., and Tanabe, K., Chrm. Let/. p. 277 (1983). 140. Arata, K., and Hino, M., Appl. Cutul. 59, 197 (1990). 141. Arata, K., and Sawarnura, H., Bull. Chc~m.So(,.Jpn. 48, 3377 (1975). 142. Wang, G . , Hattori, H.. and Tanabe, K.. Chem. Lett. p. 959 (1983). 143. Yamaguchi, T., Nakano, Y., and Tanabe, K., Bull. Chrm. Soc. Jpn. 51, 2482 (1978). 144. Nakano. Y., lizuka, T., Hattori, H.. and Tanabe. K., J. Catul. 57, I (1979). 145. Matsuhashi. H.. Hino. M., and Arata, K . , Pro<,.I n / . S y m p . Acid-Busr C ~ t u l .S. ~ i p p o r o , J p n . p. 357 (1988); A p p / . Cutul. 59, 205 (1990). 146. Livage. J.. Doi, K., and Mazieres, C., 1.Am. Cerum. Soc. 51, 349 (1968). 147. Arata, K . , in “Shokubai Kohra” (Catal. SOC.Jpn.. ed.), Vol. I I , p. 31. Kodansha Sci., Tokyo, 1986. 148. Arata, K., and Hino, M.. Shokrchui 24, 241 (1982). 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116.

210

KAZUSHI ARA TA

149. Arata, K., Hino, M., and Yamagata, N.. Birll. Chcim. .%Jl'. Jpn. 63, 244 (1990). 150. Hino. M., Arata, K., and Yabe, K., Shokrrhai 22, 232 (1980). 151. Arata, K., and Hino, M., Hyornen 19, 75 (1981).

152. Yabe, K.. Hino. M . , and Arata, K., unpublished observations. 153. Hendrickson, D. N.. Hollander, J. M., and Jolly, W . . Inorg. Cl7rm. 8, 2642 (1969). 154. Nakamoto. K., Fujita, J . , Tanaka, S . , and Kobayashi, M . , J . A m . Chem. SO(.. 79,4904

(1957). 155. Kurosaki, A., and Okazaki, S., Nippon Kugukic Kuishi p. 1816 (1976). 156. Okazaki. S.. Chinone, T., and Kurosaki, A , , Nippon KtigriXrr Kiiishi p. 1282 (1977). 157. Tanabe, K.. Proc,. Symp. 1nd.-llniu. Coop. Cliem. P r o g . . Texus A & M Uniu. p. 71 (1984).

158. Jin. T., Machida, M., Yamaguchi. T., and Tanabe, K.. Inorp. Chprn. 23,4396 (1984). 159. Yamaguchi, T.. Jin. T . . and Tanabe, K.. J . Phvs. Chem. 90, 3148 (1986). 160. Jin, T., Yamaguchi. T.. and Tanabe, K . , J . P h y s . Chc~m.90, 4794 (1986). 161. Saur. O., Bensitel. M . , Mohammed Saad, A. B., Lavalley. J . C . , Tripp, C. P.. and Morrow, B . A.. J . Cutiil. 99, 104 (1986). 162. Hatakeyama, K . . and Hisamitsu, T . , unpublished observations. 163. Hino. M . . and Arata, K., Recicr. Kine/. Curd. Lori. 19, 101 (1982). 164. Scurrell, M. S . , Appl. Curd. 34, 109 (1987). 165. Hatakeyama, K., Suzuka. T . , and Ozaki, H., Nail. Mecv. Per. Soc. Jpn.. 28ih, AXirci Abstr. 38 (1986). 166. Rajadhyaksha, R. A,, and Chaudhari, D. D., Ind. Ej7p. C h o n . , Prod. Rrs. DCW.26, 1743 (1987). 167. Arata. K . , and Hino, M., Nurl. MCW.Ccrral. SOC.Jpn., Humamar.srt Abstr. C37 (1984). 168. Arata, K., and Hino. M., .%+(JkithUi 25, 124 (1983). 169. Arata, K., and Hino, M., Proc. Sou.-Jpn. Setnitr. Curd.. 7tk, lrkrrrsk p. 7 (1983). 170. Hino, M., and Arata, K., J.C.S. C h e w . Coinmun. p. I12 (1985). 171. Yamaguchi. T., and Tanabe, K., P r o i ~ .Sou.-Jpn. Semin. Cuiitl., 7rh, lrkutsk p. 20 (1983). 172. Tanabe, K., Yamaguchi, T., Akiyama, K., Mitoh, A,, Iwabuchi. K . , and Isogai, K . , Pro(,. 1nf. Congr. Cutul.. Rrh, Wc'st Berlin. p. V-601 (1984). 173. Hino, M.. and Arata, K., Chem. Letr. p. 1671 (1981). 174. Hino, M., and Arata, K., Appl. C u m / . 18, 401 (1985). 175. Hino, M., and Arata, K., Chern. Lett. p. 963 (1980). 176. Arata, K., and Hino. M.. Norl. Mrer. Critirl. Soc. Jpn., 58111, Nugoyu Abstr.. p. 24 (1986). 177. Hino, M., Kobayashi. S., and Arata, K., Recicr. Kiner. C u r d . Leu. 18, 491 (1981). 178. Hino, M . , and Arata, K., J . C . S . Chern. Comrnun. p. 1037 (1984). 179. Hino, M., and Arata, K., Jpn. Pat. (Kou Kai Shou) 59-92023 (1984). 1x0. Hino. M., and Arata, K.. Chern. Lett. p. 1483 (1985). 181. Arata. K . . and Hino, M., Shokuhai 28, 413 (1986). 182. Hino, M., and Arata, K., J . C . S . Chrrn. Comrnirn. p. 1355 (1987). 183. Hino, M., and Arata, K., J . C . S . C h m . Cornmin. p. 1168 (1988). 184. Baba, S., Shibala, Y., Takaoka, H., Kimura. T., and Kohsaka, K., Jpn. Pats. (Toku Kai Shou) 61-68137, 61-68138 (1986). 18.5. Haba. S., Shimizu, T., Takaoka, H., Imai, T . . and Yokoyama, N . , Null. Mrrr. Pet. S o c . Jpn., 16rh, Nugoyii Abstr., p. 68 (1986). 186. Imai, T.. Nojima, S., Shimizu, T., and Baba. S., Nuil. Mrrr. Coral. Sot,. Jpn., 62nd. Sendui Abstr. 38307 (1988). 187. Ebitani, K., Konishi, J., Horie, A., Hattori, H.. and Tanabe, K., Pro(.. I n / . Symp. Acid-Base CutuI., Holikuido Uniu. p. 491 (1988).

SOLID SUPERACIDS

21 1

188. Arata. K., and Hino, M., Proc. I n ( . C o n g r . C u t d . , Y t h , Crilgciry p. 1727 (1988). 189. Arata, K., and Hino. M., Jpn. Pat. (Toku Can Shou) 63-114790 (1988). 190. Hino. M.. and Arata, K.. J.C.S. Chrm. Cornrwim. p. 1259 (1988). 191. Hino, M., and Arata, K.. ChcJm.Lrtt. p. 971 (1989). 192. Arata, K.. and Hino. M., Sliokuhrii 31, 477 (1989). 1Y3. Kurosaki. A., and Okazaki, S., Bull. Chem. So(.. Jptz. 56, 1279 (1983). 194. Kurosaki, A., and Okazaki, S., Chern. L P U . p. 1741 (1983). IY5. Takusari, H., and Okazaki, S.. Nippon Kugiikrr Kuishi p. 2016 (1985). 1Y6. Okazaki, S., Sudoh, T., Toba, R., and Kurosaki, A,. Siiokuhai 25, 103 (1983). 197. Horie. 0.. and Okazaki, S., C k m . Lrtt. p. 1089 (1986). 1%. Okazaki. S., and Jouhouji, H., Bull. ~ / 7 1 w l SOL.. , J p n . 59, 1931 (1986). 199. Ayame, A.. and Izumizawa, T.. Hyoiwen 27, 640 (1989). 200. Ayame, A , , Ohta, K., lzumizawa, T., Sawada, G.. Zhang, G., Sato, H., and Kakizaki, H.. Shokirhui 30, 72 (1988). 201. Ayarne, A , , and lzumizawa. T., J.C.S. C/icvn. Cotnrnrtn. p. 645 (1989). 202. Ayarne, A., Sawada, G., Sato, H., Zhang, G . , Ohta, T.. and Izumizawa. T., Appl. Ciitul. 48,25 (1989). 20-3. Ayame, A., Izumizawa, T., and Kakizaki, H., Pro(.. Acid-Base Cirtul. I n / . S-vmp., I b k y o p. 371 (1989). 204. Drago, R. S., and Getty, E. E., J . A m . Chrm. S o l . . 110, 331 I (1988).

This Page Intentionally Left Blank

ADVANCES IN CATALYSIS, VOLUME 37

Oscillatory Catalytic Reactions at Single-Crysta I Surfaces G. ERTL

I.

Introduction

If a chemical reaction is operated in a flow reactor under fixed external conditions (temperature, partial pressures, flow rate etc.), usually also a steady-state (i.e., time-independent) rate of reaction will result. Quite frequently, however, a different response may result: The rate varies more or less periodically with time. Oscillatory kinetics have been reported for quite different types of reactions, such as with the famous Belousov-Zhabotinsky reaction in homogeneous solutions ( I ) or with a series of electrochemical reactions (2). In heterogeneous catalysis, phenomena of this type were observed for the first time about 20 years ago by Wicke and coworkers ( 3 , 4 ) with the oxidation of carbon monoxide at supported platinum catalysts, and have since then been investigated quite extensively with various reactions and catalysts (5-7). Parallel to these experimental studies, a number of mathematical models were also developed; these were intended to describe the kinetics of the underlying elementary processes and their solutions revealed indeed quite often oscillatory behavior. In view of the fact that these models usually consist of a set of coupled nonlinear differential equations, this result is, however, by no means surprising, as will become evident later, and in particular it cannot be considered as a proof for the assumed underlying reaction mechanism. Mechanistic studies with “real” catalysts near atmospheric pressure conditions are complicated by several factors: the surface structure and composition will be inhomogeneous and hence also the reactivity may be spatially different. In addition, the heat released by the reaction may change the (local) temperature, and as a consequence, kinetic oscillations are frequently associated with strong nonisothermal effects. These prob213

Copynghl 61 IW by Academic Prrrr. Inc. All rights (11 reproduction in any kirm reserved.

214

G . ERTL

lems can be overcome by using well-defined single-crystal surfaces at partial pressures that are low enough to render any temperature variations caused by the exothermicity of the reaction negligible, and, in addition, allow in sitrr investigations of the microscopic properties of the surfaces. The first report taking this “surface science approach” was published in 1982 for the oscillatory CO oxidation on Pt( 100)( 8 ) ,which system was then investigated further in great detail, both experimentally and theoretically. Subsequently, similar phenomena were observed with a series of other systems and their number will certainly increase further. There is certainly not a single microscopic mechanism underlying these phenomena, not even for the same reaction, if conducted under different conditions. For example, under low-pressure conditions, kinetic oscillations may be associated with adsorbate-induced restructuring of the surface ( 9 ) , whereas at higher pressures “chemical” transformations, such as suggested by the “oxide” (10-13) or “carbon” models (14, 15), may come into play. Quite generally, temporal oscillations of the reaction rate are coupled to periodic switching of the surface between states of high and low activity, whereby these transformations proceed via autocatalytic processes. Theoretical description has to be performed within the framework of nonlinear dynamics (16). In brief, knowledge about the microscopic elementary steps for a particular system permits formulation of a set of rate equations for the variation of the surface concentrations of the species involved in the overall reaction, x i , (the state variables) with time, in terms of the external parameters p k (temperature, partial pressures) and also of all state variables x , (17): ~ dx;/dt

=

F(Xi, p &

(1)

This represents a set of coupled nonlinear ordinary differential equations whose solutions (trajectories) will for t + 53 approach so-called limit sets, which may be either fixed points (the usual steady states) or limit cycles (periodically oscillating state variables and hence also reaction rates), or even more complex (mixed-mode oscillations, quasiperiodicity, or chaotic behavior). This scenario of temporal self-organization will be illustrated in the following discussion by a series of experimental examples. Equation ( I ) contains only the variation with time of the state variables, which are hence assumed to behave uniformly in space. For a macroscopic system (such as a single-crystal surface with say 30 mm2area or even a supported high-area catalyst), however, some type of spatial coupling between different parts will be necessary in order to yield temporal self-organization of an integral quantity such as the reaction rate. This communication is achieved by transport processes such as heat conductance (under noniso-

OSCILLATORY CATALYTIC REACTIONS

215

thermal conditions), coupling between surface diffusion and reaction, or via the gas phase through modulation of the partial pressures associated with a varying reaction rate. Account can be taken in terms of a generalized In this respect Eq. (1) is incomplete diffusion process with coefficient Di. and has to be extended into partial differential equations that take into account the tornporul as well as spatial uuriation of the state variables (18).

dxjldt

=

F ; ( X ,pa) ~ , + Di(d’x,ldr’)

(2)

where r is a space coordinate. If the dominating process is much faster than the typical time scale of the oscillations, Eq. (2) reduces to Eq. ( I ) and the whole surface is in fact oscillating in phase. Otherwise, spatial self-organization will be associated with wave propagation phenomena. Again these general conclusions will be illustrated by detailed observations. In the following work, we will restrict our discussion essentially to processes occurring at well-defined single-crystal surfaces under ultrahighmbar. As vacuum (UHV) conditions and with partial pressures of mentioned above, additional effects will come into play at higher pressures and with polycrystalline or even supported catalysts, which will not be treated explicitly here. B y far the most extensive investigations, both experimentally and theoretically, have been performed with the CO/Oz reaction on Pt( 100) and Pt( I 10) surfaces. The description of these systems will form the main part of this article. II. Catalytic Oxidation of CO On Pt(100) and Pt(ll0) Surfaces

A.

GENERAL

The mechanism of the reaction CO + t0, + CO, on surfaces of the platinum-group metals has been explored in detail and proceeds via the following steps (19):

co + * O? + 2*

COad + o,,

k?

COad

k 2 2o,,

(b)

3

(c)

COX+ 2*

The star denotes schematically a free adsorption site, which, however, has a different meaning for the two adsorbates. Adsorbed CO tends to

216

G . ERTL

form densely packed layers that, beyond a certain critical coverage, inhibit dissociative oxygen chemisorption. The adsorbed 0 atoms, on the other hand, form rather open structures, which still permit CO adsorption even if the surface is saturated with Oad.While the initial (i.e., at zero coverage) sticking coefficient for CO is near unity for all platinum surfaces, the initial sticking coefficient for dissociative oxygen adsorption is strongly affected by the structure of the surface-a factor that is of vital importance for the occurrence of kinetic oscillations. The rate parameters (including adsorption energies, etc.) of the above reaction scheme are all dependent on coverage, so that no simple analytical expression for the overall rate can be formulated. Nevertheless, the steady-state kinetics could be successfully modeled mathematically, even for high-pressure conditions, on the basis of detailed knowledge about these parameters derived from UHV singlecrystal studies (20, 2 0 . The qualitative features pertinent to the present topic can even satisfactorily be modeled by an ansatz based on modified Langmuir kinetics for the above reaction scheme:

whereby the coverages O,, and 0, are determined by the following two differential equations

The rate constants k , and k , for CO and O2 adsorption, respectively, are determined by the respective initial sticking coefficients (i.e., at zero coverage). The coverage dependence of CO adsorption follows a precursor kinetics that can be modeled by the exponent r [being between 3 and 4 for Pt(ll0) (22)]. The kinetics of oxygen adsorption is modeled by the requirement for two neighboring empty adsorption sites, which can be occupied by either 0 or CO. We will return to this model later at various points. Figure 1 shows the variation of the stationary rate of CO, formation with CO pressure at fixed temperature and 0, pressure for both a wellannealed, flat Pt( I 10) surface (solid line) and the same surface exhibiting periodic arrays of monoatomic steps forming facets (broken line) (23).At low CO pressures, the surface is largely covered by Oad,and the rate increases linearly with pco. Nearly every CO molecule striking the surface

217

OSCILLATORY CATALYTIC REACTIONS I

I

I

I

Pt (1101 Po2= 1.5 ~ l 0Torr -~ Ts =L80K

2.0 -

J

L

W

.

/faceted 1.5

V 0

-

/HA---

0

I

I

I

I

1

2

3

L

I

5

-d

I

6

7

8

PCOI 10-5Torr FIG. I . Steady-state rate of CO? formation as a function of CO pressure on a flat (solid line) and a stepped (dashed line) Pt(ll0) surface. (From Ref. 3Y.)

is adsorbed and will rapidly be consumed by reaction with neighboring adsorbed 0 atoms. Hence the rate is limited by CO adsorption in this range. Because the CO sticking coefficient is not affected by t h e presence of steps or other surface defects, there is also n o difference between the two types of surfaces, and one would denote the reaction as “structure insensitive.” With increasingp,,,, however, the concentration of adsorbed CO molecules also increases, which now begins to inhibit adsorption of oxygen. Hence the rate passes through a maximum and decreases about Y l/pco with further increase of the CO pressure due to progressing inhibition of O2 adsorption. In this range obviously the rate is limited by oxygen adsorption. Because the 0, sticking coefficient is markedly higher at surface defects than on flat low-index terraces, for the same set of reaction conditions the rate will always be higher with the facetted surface in this range: now the reaction is structure sensitive. More generally, the surface can (for certain sets of parameters) be in two states, with either predominant CO or 0 coverage. Upon variation of one of the partial pressures, for example, a more or less abrupt transition between these states occurs, whereby the system may also show pronounced hysteresis effects-it exhibits bistability. Figure 2 reproduces corresponding experimental data for Pt( I lo), where the stationary work function change A@ (oxygen coverage) is plotted as a function of pco at fixed po, and for various temperatures (24). Kinetic oscillations are ob-

-

218

G . ERTL

FIG.2. CO oxidation on Pt( 110): variation of the work function A@ with CO pressure and temperature at constant O2 pressure. (From Ref. 24.)

served within narrow ranges of the transition region. With increasing temperature the A@-jumps between the two states become smaller and finally disappear in a so-called cusp. The qualitative features of these findings (with exception of the oscillations) including the cusp bifurcation are completely reproduced by numerical solution of the set of Eqs. (4) and (3,as shown in Fig. 3 (16). This underlines the basic validity of this simple mathematical model. The qualitative features of the r ( p c o ) curve of Fig. I are the same for all types of platinum metal surfaces and will serve as a guide for exploring the mechanisms of kinetic oscillations occurring with this reaction. The shape of this curve can be modeled quite easily on the basis of the underlying reaction mechanism (8, 25). These equations are nonlinear in nature and can, for example, describe hysteresis effects observed in the transition region between high and low reactivity, but they will not produce sustained temporal oscillations. The latter require additional processes for which numerous suggestions and speculations can be found in the literature (8).

OSCILLATORY CATALYTIC REACTIONS

219

FIG.3 . Steady-state oxygen coverage during catalytic CO oxidation as a function ofpco and T a s calculated from Eqs. (3) and (4) with parameters adopted from experimental data for Pt(l10). (From Ref. 16.)

At a first glance, also the experimental situation looks complex. No oscillations, under any conditions, were found with Pt( 1 1 1 ) (13, 25-27). With Pt(lO0) (9,13,26-33), ( 1 10) (23,2#,3#),and (210) (35-37) oscillations occur beyond the maximum of the steady-state reaction rate, i.e., under conditions for which oxygen adsorption is rate limiting and the surface is CO covered to an appreciable extent. To complicate things further, with Pt(2 10) a fairly long “induction period” is required for establishing the oscillations, during which also the overall reactivity changes (37, 38). Similar slow changes in reactivity during the course of the reaction are, under certain conditions, also observed with Pt( 110) (39),which, in turn, also affect the oscillatory properties (23). The solution to this apparent puzzle lies in the different microscopic surface processes governing the macroscopic kinetics. The densely packed Pt( I 1 I ) surface remains unaltered under the conditions of the reaction. Its kinetics can hence be described within the framework of the aboveformulated sequence of reaction steps without any additional variations of the kinetic parameters. As mentioned before, the solution of these

220

G . ERTL

equations will show no oscillatory behavior, in agreement with the experimental findings. The clean Pt(100) and ( I 10) surfaces, on the other hand, are reconstructed, i.e., the configuration of atoms in the topmost layer differs from that in the corresponding bulk plane. This reconstruction is lifted by adsorbed CO as soon as its coverage exceeds a critical value. These structural transformations are associated with changes of the oxygen sticking coefficient and hence of the reactivity. (Remember that the oscillations occur under conditions for which oxygen adsorption is rate limiting!) Periodic switching between these states of low and high reactivity gives rise to the observed modulation of the reaction rate. With Pt( I lo), the surface may, under certain conditions (also outside the oscillatory range), undergo a further continuous structural transformation leading to facets, consisting of a step and terrace structure whose consequence on the steady-state kinetics becomes evident from Fig. I and which, in turn, also affects the oscillatory kinetics on the flat ( I 10) terraces. The clean Pt(210) surface is nonreconstructed and there is also no indication for a CO-induced alteration of the atomic positions within the unit cell. In this case, however, under reaction conditions, continuous faceting, again with creation of other crystal planes, preferably ( I lo), takes place. The latter may then in turn be subject to periodic structural transformations, which then accounts for the observed integral oscillatory behavior. The following discussions will present a detailed account of the experimental evidence for the briefly sketched effects as well as of their consequences with respect to temporal self-organization.

B.

MECHANISM OF KINETIC OSCILLATIONS Pt(100) A N D Pt(l10) SURFACES

ON

Figure 4 shows a typical example of sustained kinetic oscillations occurring for particular conditions (pco, pol, and 7') during the catalytic CO oxidation on a Pt( 110) surface (40). The measurements were performed with an UHV system acting as flow reactor, where the COz partial pressure is directly proportional to the rate. The simultaneously recorded CO pressure oscillates with the same period and with amplitudes of about 1%. whereby pco shows a minimum whenever the reaction rate is maximum. The work function A@ varies parallel to the rate R . This quantity is essentially determined by the oxygen coverage. Because under oscillatory conditions the rate is determined by oxygen adsorption (see above), it becomes plausible why A@ and R vary in phase. The key for understanding the occurrence of kinetic oscillations with

OSCILLATORY CATALYTIC REACTIONS

22 I

Ts = 480K

Po2 = 1.6xlO“Torr PCO= 5.1~10‘~Torr

P

0

0

I

I

200 -C

400 Timeis1

I

600

FIG.4. Sustained oscillations during the CO/@ reaction Pt( 110). Kate o f C 0 2formation function A?) as a function of time. (From Ref. 40.)

( p ~ ~ CO ) ~ )pressure, , and work

Pt(l10) and Pt( 100) lies in the structural and adsorptive properties of these surfaces. If these are clean, the atomic arrangement in the topmost layer is not that of the corresponding bulk plane ( I x 1 structure; see Fig. 51, but both surfaces are reconstructed. The Pt(l10) surface is of the I x 2 “missing row” type (4/-43), whereas Pt( 100) exhibits a quasihexagonal (“hex”) configuration of surface atoms ( 4 4 , 4 5 )(Fig. 5 , right side). In both cases the reconstruction is lifted by adsorption of CO if its coverage exceeds a certain critical coverage. The driving force for this process has to be sought in the energy gain by the more favorable I x I configuration offered to the adsorbate, which overcompensates the energy necessary for altering the geometry of the clean surface. The energetic situation for Pt(100) is depicted schematically by the diagram of Fig. 6 (46).The heat

222

G . ERTl

-

1 x 1 Surface

ieconstructed Surface

hex

ill01

1i 7

Pic;. 5 . Ball models of the three most densely packed surfaces of Pi. (Left) Nonreconstructed bulk terminalion. (Right) Reconstructed surface (schematic).

of adsorption for CO on the 1 x I Pt( 100) surface is 37 kcal/mol, but only 27 kcal/mol on the hex phase. Therefore, nuclei of the 1 x I phase, on which CO forms a c2 x 2 structure (with local coverage O S ) , appear on the hex surface as soon as the total coverage exceeds a critical value of about 0.08. This process proceeds even at 150 K and exhibits no noticeable activation energy (46). Due to the higher adsorption energy for CO o n the 1 x 1 phase, additional molecules adsorbing on hex patches will diffuse to the boundaries of these c2 x 2-1 x 1 islands, where they will be trapped. As a consequence, the adjacent sites will also be converted into 1 x 1 and the fraction of the total area covered by this phase will grow with increasing total coverage. In other words, the surface will consist of two phases: 1 x 1, with local CO coverage 0 = 0.5, forming a c2 x 2 overlayer structure, and hex, with a much lower CO coverage. If, on the other hand, the local CO coverage on the I x I surface drops below about 0.3 (e.g., by thermal desorption or reaction), this phase becomes metastable and it transforms

OSCILLATORY CATALYTIC REACTIONS

223

T FIG. 6 . Schematic energy diagram for the CO-induced transformation of the surface structure o f Pt(100). (From Ref. 46.)

back into the hex structure, which process is thermally activated with an energy of about 25 kcal/mol (47, 48). The atomic density of the hex phase is about 20% higher than that of the 1 x I phase. As could be demonstrated by scanning tunneling microscopy (STM) ( 4 9 ) , during the CO-induced hex -+ I x I transformation, these additional atoms are “squeezed” out from the surface layer, on top of which they are aggregating to new small I x I patches, a result which could also be successfully modeled by computer simulations (50). Similarly, the I x 2 structure of the Pt( 110) surface transforms into the nonreconstructed I x I phase upon CO adsorption ( 2 2 , 5 / - 5 6 ) . Now the reconstructed phase has a 50% lower atomic density, and surface diffusion will be restricted to very short distances if the temperature is not high enough. How this problem is overcome is demonstrated by the STM images reproduced in Fig. 7 (57). Adsorption of CO on 1 x 2 Pt( 110) at room temperature causes homogeneous nucleation (initiated by local adsorbate density fluctuations) of very small 1 x 1 patches, simply by jumps of surface atoms over one to two lattice spacings. The first nuclei are formed after the critical coverage of about 0.2 is reached, and further increase of the CO exposure causes the formation of additional nuclei rather than growth of the already existing ones. At higher temperatures (such as during kinetic oscillations), on the other hand, the enhanced surface mobility of the Pt atoms leads to the formation of more extended 1 X 1 regions.

224

-

G . ERTL

10A

(a )

(b)

FIG.7. Scanning tunneling microscopy image from a Pt(l10) surface showing nucleation of the CO-induced 1 x 2 + I x I transformation (a) and corresponding ball model (b). (From Ref. 57.)

The two phases (reconstructed and nonreconstructed) of the respective Pt surface exhibit, in addition, marked differences with respect to oxygen adsorption. For Pt(100), the sticking coefficient s,, for the 1 x I phase is about 0.1, but is almost negligibly small on the hex phase (48, 58-62). This difference is less pronounced for Pt(ll0). Reports on s,, for the 1 x 2 phase vary between 0.15 and 0.5 in the literature (63-66), and it is suggested to be larger by about a factor of 2 for the nonreconstructed plane (66). On the basis of this information, the following mechanism for the occurrence of kinetic oscillations [exemplified for Pt( loo)] may be suggested: remember that the external parameters (po,, pco, and T) are adjusted in a way that the surface is largely CO covered and oxygen adsorption is limiting the rate of CO, formation. The surface will be largely transformed into the c2 x 2-1 x I phase with O,,, = 0.5. Such an adlayer, however, does not completely inhibit O2adsorption and subsequent reaction, mainly due to the inevitable presence of surface defects (67). At these sites dissociative oxygen adsorption proceeds with a high sticking coefficient, and the adsorbed 0 atoms react with adjacent CO molecules to yield CO,, which is immediately released into the gas phase. In this way two empty

OSCILLATORY CATALYTIC REACTIONS

225

adsorption sites on the 1 x I phase are created, and in an autocatalytic process the uptake of oxygen by the surface and hence also the rate increase. Because CO is rapidly reacted off by O,,, the steady-state coverage of CO,, on the I x I patches will, however, now drop below its critical value (Oa, will not stabilize the I x I phase!) and the surface transforms into the hex phase. Here the oxygen sticking coefficient is negligibly small, the rate decays, and the CO coverage "recovers" beyond the critical value for the hex -+ I x I transformation, which process then completes one cycle of the oscillations. This mechanism implies periodic transformation of the surface structure between the largely CO-covered c(2 x 2)-1 x I and the essentially adsorbate-free hex phase, with the metastable I x I surface exhibiting appreciable 0,, concentration as an intermediate. This conclusion could be verified experimentally by probing the structural properties in situ by means of low-energy electron diffraction (LEED) (28). Figure 8 shows the variation of the work function (reaction rate) as a function of time together with the intensities of selected LEED beams reflecting the presence of the respective surface phase for a Pt( 100) surface under properly chosen stationary conditions. Both sets of data vary periodically with identical time constants, thus demonstrating the coupling between rate oscillations and structural transformations. More specifically, the intensity of the c2 x 2 phase, for example, stays at a constant high level for several minutes, whereas those for the hex phase and of the T i beam (0coverage) are very low, signaling that the surface area probed by the LEED beam ( - I mm?) is essentially CO covered and exists as c(2 x 21-1 x 1 phase. Subsequently the rapid decay of the c2 x 2 intensity is accompanied by a steep rise of the T i intensity, reflecting the autocatalytic removal of CO,, and the intermediate buildup of a high O,, concentration on the metastable 1 X 1 phase. The latter continuously transforms into the hex phase, whose _intensity grows to a maximum, paralleled by a decrease of the 1 1 intensity, until CO adsorption on the hex phase causes its rapid transformation back into the c2 x 2-1 x I phase. The shape of the work function trace does not correlate directly with the fairly abrupt LEED intensity changes. This effect is due to the different areas probed by both techniques and is related to the problem of spatial self-organization, which will be discussed later. Theoretical description of this system could successfully be achieved in terms of a set of four coupled differential equations, describing the variation with time of the CO coverages on the I x 1 and hex phases, respectively, that of the 0 coverage on I x I phase (oxygen adsorption on hex is neglected), and finally of the fraction of the surface being present in the I x I configuration, i.e., describing the CO-induced hex e 1 x I structural

226

Ci. E R I L

Torr Torr

t

300-

.-

200-

s

I -

u C

2

100-

%3

0-

-I > o

c U

ul

z w

I-

E

__ 0

0

5

10

time

-

15

[minl

FIG.8. Sustained kinetic oscillations during the COiO? reaction on Pt(100). (From Ref. 28.) (a) Variation of the work function (integral reaction rate) with time. (b) lntensities of selected low-energy electron diffraction beams.

transformations (29). All the kinetic parameters and critical coverages involved were deduced from independent experiments, so that no further adjustment of parameters was necessary. The qualitative agreement of the oscillatory solutions with the experimental data was rather satisfactory. Recently the mathematical model was further refined, however, without any substantial alterations of the conclusions to be drawn (68). Basically the same mechanism prevails with the oscillatory CO oxidation on Pt(l lo), however, with two important differences: now the sticking coefficient for 0, differs for both surface modifications only by about a

227

OSCILLATORY CATALYTIC REACTIONS

factor of 2. This has consequences on the width of the partial pressure range over which oscillations may occur, which, in turn, is crucial for the mechanism dominating spatial self-organization. Second, under certain conditions the Pt( 110) surface may undergo faceting, which effect-if it occurs parallel to kinetic oscillations-also influences the oscillations. The latter aspects will be treated separately later. More recently, a kinetic model for describing the occurrence of temporal oscillations on Pt( 110) was developed, which is more simplified than that originally derived for Pt( loo), but nevertheless reproduces the qualitative features well and enables, in addition, analysis of further effects, such as forced oscillations, etc. (69). It starts from Eqs. (4) and ( 5 ) outlined previously, which were the basis for successful description of the overall kinetics. The system was then extended in order to include the CO-induced structural transformation of the surface as well as the different oxygen sticking coefficients on the two modifications. If we denote the fraction of the surface being present in the nonreconstructed ( I x I) structure by a , I - LI will be the fraction being reconstructed ( I x 2). The lifting of the reconstruction starts at a (total) coverage O,, = 0.2 and is completed at Oco = 0.5; it occurs with a rate constant k,, for which an estimate is also experimentally accessible. Hence a third equation can be formulated -k,a

c

for Oco 5 0 . 2 for 0.2 < Oco < 0.5

dt k,(l

-

a)

for Oco 2 0 . 5

(6)

-

The cases of O,, < 0.2 and O,, > 0.5 describe the 1 x I I x 2 and 1 x 2 -+ I x I transitions up to completion, respectively, in terms of simple first-order kinetics. The range between OCo = 0.2 and 0.5 was fitted by a third-order polynom, with the coefficients qi adjusted to the experimental data of Gritsch rt al. (57) in such a way that the resul.ting function is differentiable everywhere. Equation ( 5 ) was modified in order to take into account the different oxygen sticking coefficients on the 1 x I and I x 2 patches. If their ratio is denoted by a [which was taken to be 1.5 from the experimental estimates (66)1, the temporal variation of the oxygen coverage now reads

whereby k, contains the 0, sticking coefficient on the clean 1 x 2 surface. The kinetics of CO adsorption and desorption is, within this simple model,

228

G . ERTL

.6

-

.5

L

$

.2

0

"

.l

IO

15

20

25

30

35

40

45

50

55

60

t Is1 FIG.9. 'Theoretical data for the variation ofthe oxygen ((=)(,)andCO (@Ico) coverages, as well as of the fraction of the surface present in the nonreconstructed 1 x I structure ( a ) as a function of time, as evaluated from the solution of Eqs. (4). (Sa), and ( 6 ) . modeling the oscillations for the CO/O, reaction on Pt( I 10). Parameters adopted from experimental data for po? = 5 x lo-' torr, pco = 2.3 X torr, and T = 540 K. (From Kef. 6Y.)

assumed not to be affected by the structure of the surface, and hence Eq. (4) remains unaltered. Analysis and numerical solution of this new set of equations [Eqs. (4), (5a), and (6)] revealed indeed oscillatory behavior within narrow ranges of external parameters in agreement with experimental evidence. A typical set of resulting time series for the variations of the 0 and CO coverage and of the ( I x I ) fraction u is reproduced in Fig. 9. The success of this simple model is considered as additional confirmation of the conclusion, that the CO-induced structural transformation of the surface causing an alteration of the oxygen sticking coefficient is the basic physical reason for the occurrence of kinetic oscillations.

C.

CONDITIONS FOR T H E OCCURRENCE OF

AUTONOMOUS

OSCILLATIONS Once the external parameters ( p c o , po,. and T ) have been established to initiate autonomous kinetic oscillations, these can usually be sustained for periods of time as long as desired, provided that the surfaces are prevented from becoming contaminated (in particular by carbon, originating from spurious traces of hydrocarbons in the feed gas mixture) and that the partial pressures are not drifting off (31). With Pt(l10) a complication may arise insofar that during the course of the reaction the surface struc-

229

OSCILLATORY CATALYTIC REACTIONS

500 250 -

'+

h

T 100 L 50 -

@

I

0

a"

1x2-I

XI

in co vis. L E E D in C0/02

- osc. 0

't%

10 -

+

v,.

b '

A-0

1

25

52.5

1

+'

1.6

1.8

2 .o

2.2

1000K T FIG. 10. Pt(l10). Plot of log pro versus liT; + , the completion of the I x 2 + I x 1 transformation of the surface structure with a pure CO atmosphere as monitored by LEED; A and V,the corresponding data in the presence of an additional O? pressure of 4 x lo-' torr. The latter coincide with the oscillatory range (horizontal lines) for the same set of control parameters, whereas the former agree with the high-temperature limit for the occurrence of oscillations at the respective CO pressure (0). (From Ref. 40.)

ture changes continuously by faceting. This effect will be discussed separately in Section II,F and does not play a role in the present context. Kinetic oscillations with the present systems were typically established in the 10-s-10 torr total pressure range and for temperatures between about 400 and 550 K. With Pt(llO), these oscillations were generally confined to rather narrow ranges of parameters. If, for example, PO, and Twere kept fixed, the oscillatory range of the CO pressure could be varied within only afew percent. For agivenpo,, on the other hand, the oscillatory CO increases with increasing T up to ahigh-temperature limit, as can be seen from the log pco versus l / T plot of Fig. lO(40) (dashed line). These data points coincide with those marking the completion of the I x 2 -+ I x I transformation of the surface structure as monitored by LEED under identical conditions (24).The isostere marking the high-temperature limits for the oscillations (obtained by varying po,) coincides, on the other hand, with the conditions under which the CO-induced I x 2 -+ 1 x I transformation in u purr CO utmospherc. is completed. The slope of this line agrees with the heat of CO adsorption on t h e Pt( 110) 1 x I surface (43. 53,63, 70). In other words, if at a given CO pressure the temperature is increased beyond this limit, the CO coverage will never exceed the critical value necessary for the 1 x 2 -+ 1 x 1 transformation and hence also no kinetic oscillations occur. Although for Pt( 100) the mechanism underlying the kinetic oscillations

230

G . ERTL

100

i-

T=L80K

pcol10-6Torc1

FIG. I I . Conditions pco and po, at fixed T = 480 K for the occurrence of oscillations in the CO/Oz reaction on Pt( 100) and-Pt( 110) surfaces. ‘The dashed line marks the lower limit for the CO pressure to cause the hex + I x 1 structural transformations on Pt( IOO), which also coincides with the lower boundary of the oscillatory range for this plane. (From Ref. 40.)

is basically very similar, there are pronounced differences with respect to the conditions for their occurrence. This becomes, for example, evident from Fig. I I , comparing those conditions for Pt( 100) and ( I 10) at fixed T = 480 K in a po, versus pco diagram (40).For a given po,, for example, with Pt( I lo), the variation i n pco must not exceed 2%-that is why the existence range is in this case simply represented by a line, whereas for Pt( loo), pco may be varied by up to a factor of 3. Although the existence range for oscillations is rather broad in the latter case, in a log pro versus l / T plot the conditions for the hex -+ I x I transition when cooling the sample in a pure CO atmosphere were found to coincide with the oscillatory conditions (9), which was the first experimental hint for the close coupling between these phenomena. As can be seen further from Fig. I I , there is no overlap of the oscillatory conditions between the two planes and there exists a low-pressure limit at a given temperature. All these observed features underline the fundamental importance of the adsorbate-induced surface structural transformations and can be qualitatively rationalized in terms of the outlined mechanism. For example, in order to obtain oscillations at given T and po?,the CO pressure (causing the respective CO coverage) must be high enough in order to lift the surface reconstruction, but, on the other hand, also sufficiently low to permit subsequently the reactive removal of the CO adlayer by oxygen.

23 1

OSCILLATORY CATALYTIC REACTIONS

Pt (1001

I

I

I

2 1

0

I

0.5

I

I

1

1.5

I

2

I

I

2.5

3

Ig FIG.12. Computer simulations of the oscillatory range. Variation of the high and low CO pressure boundaries (at fixed poz and 71 with the ratio a ofthe sticking coefficients for oxygen on the nonreconstructed and reconstructed surfaces. respectively. (From Ref. 4 0 . )

Both limits will increase with increasing po,, however, with different slopes, so that an intersection occurs that marks the lower limit of po, and pco, below which oscillations are no longer possible. The difference between these slopes is primarily due to the difference in oxygen sticking coefficient (rate of removal of adsorbed CO) for the reconstructed and nonreconstructed surface phases. The width of the oscillatory range is of direct relevance for the dominating mechanism of spatial self-organization, as will be outlined in Section, 11, G. Its difference between Pt(l10) and (100) is again a consequence of the varying adsorptive properties for oxygen with these two surface (40, 69). Figure 12 shows the result of computer simulations, based on the oscillatory mechanism outlined above and with suitably chosen kinetic parameters, for the upper and lower CO pressure limits for the occurrence of oscillations at fixed po, and T , as a function of the ratio a of the O2 sticking coefficients on the nonreconstructed and reconstructed phases, respectively. For Pt(100), a > 10’ and hence a wide range for pco is predicted for which oscillations occur, but it will be much narrower for Pt( 110) with a = 2, and it finally disappears completely if the difference in oxygen sticking coefficient vanishes (i.e., if a + I ) .

232

G . ERTL

D.

N O N L l N E A R DYNAMICS:

TEMPORAL SELF-ORGANIZATION

With Pt( loo), the temporal oscillations are usually irregular, which is connected with the problem of spatial self-organization. With Pt( 1 lo), on the other hand, for wide ranges of conditions the whole surface area oscillates essentially in phase so that spatial inhomogeneities are not of significant importance and the behavior of the system may indeed be satisfactorily modeled by ordinary differential equations, as presented in the preceding section, i.e., without inclusion of spatial variables. The oscillations are frequently quite regular, but may also exhibit particular features that are characteristic for systems obeying nonlinear dynamics, i.e., features that may be generally described by a system of coupled nonlinear ordinary differential equations of the type in Eq. ( I ) as outlined in Section I .

-_ dt

-

F(xi, p k )

The solutions of these equations (the trajectories) will for long times (i.e., after transient effects associated with switching on the external parameters have decayed) approach so-called limit sets, which may be classified into fixed points (stationary states), limit cycles (periodic oscillations), mixedmode oscillations, quasiperiodic oscillations, and chaotic behavior. Transitions between these states may occur upon variation of the external parameters pnand are called bifurcations. Experimental evidence for these effects with the system CO + O,/Pt( I 10) will be briefly presented without going further into details of the underlying general theory (see 16, 17). A typical sequence of limit sets is reproduced in Fig. 13, in which the response of the system was monitored through recording the variation of the work function A@(reaction rate) as a function of time for constant parameters T = 540 K and po, = 1.5 x torr, and pcO was varied torr (71). For pco = 3.9 x 10-j torr (curve between 3.9 and 3.4 x a) A @ stays constant, as is characteristic for a Lwctiotw-ystate, representing the “normal” state of an open system under fixed conditions. Upon slight reduction of pcO, regular oscillations with small amplitude appear (b). Their period remains essentially constant, but the amplitude grows by further decreasing the CO pressure (c). This is characteristic for a socalled supercritical Hopf bifurcation, marking here the transition between a fixed point and a limit cycle of harmonic oscillations. At even lower pco, period doubling occurs (d), i.e., the amplitudes are alternating between large and small, followed by a second period doubling (e), and finally a transition to irregular behavior with nonsystematic variation of the amplitudes (f) takes place. Below pco = 3.4 x 10 torr the amplitudes decrease

-’

233

OSCILLATORY C A I A L Y I'IC REACTIONS

F ~ t i .13. Kinetic oscillations during the C O / 0 2reaction on Pt( 110) at 7 = 540 K , po, = 7.5 x torr, and for varying pc.o.(From Ref. 7/.) ( a ) pco = 3.90 x l o - ' torr; conslmt behavior (fixed point). (b) pco = 7.X4 x torr: onset of harmonic oscillations with small amplitudes (Hopf bifurcation). ( c ) p c o= 3.60 x 10 ' toi-r: harmonic oscillation with increased tom; amplitude. ( d ) p c o = 3.61 x lo-' torr; first period doubling. (e) pro = 3.52 X second period doubling. (f) pcli = 1.42 x l o - ' torr: apcriodic (chaotic) behavior.

and another stationary state is reached. This i s an example for the transition of a system from regular oscillations via a sequence of period doublings t o the state of determinisric chrros (Feigenbaum scenario) (72, 73). (Note that for- expcrimental reasons only the first two period doublings were clearly discernible, which should continue further with periodicities 8, 16, . . . up to the chaotic state.) Plotting the maxima and minima of the A 4 amplitudes at various pco yields the bifurcation diagram of Fig. 14. lfone denotes the CO pressure for which the ith pcriod doubling occurs by p i . theory (72. 7-31 predicts a value of 3.73 for the ratio ( y 2 p , ) / ( p i p?), which is compatible with the experimental value of 4 -+ I . Another way of presenting time series of the type shown in Fig. 13 consists in deriving from them the attractors by applying the time delay method (74, i.e., by plotting A@(r) versus A @ ( r + 71, whereby T is an arbitrary (but fixed) time, or-in three dimensions-by plotting A@([) versus A@(r i- 7) versus A @ ( r 27). Attractors of the latter type constructed from some of the time series of Fig. 13 are reproduced in Fig. IS (71). Diagram a represents the simple periodic oscillations with the width ~

+

234

G . ERTL

60 -

20 -

LO

0-

-20

-

-10 -

-

-60

3.9

I

I

I

I

I

3.8

3.7

3.6

3.5

3.1

c -

pco IW5 Torr 1

FIG. 14. Bifurcation diagram for the oscillatory C O / 0 2 reaction on Pt( 110) as derived from the experimental data of Fig. 13. (From Ref. 7 / . )

of the attractor being due to the noise of the underlying experimental data; b arises from twofold periodic, c arises from fourfold periodic, and d arises from chaotic behavior (“strange” attractor). Chaos has to be clearly distinguished from random (noise) behavior, which is often not trivial on the basis of a limited set of experimental data, but which can be performed by careful mathematical analysis that provides quantitative data characteristic for the chaotic state, such as Lyapunov exponents and correlation integrals (75, 76), as was done for the present system (71). Although the regular oscillations and their transition to chaos are observed at fairly high temperatures, at which the surface structure remains unaffected from slow changes due to faceting, such effects can interfere at lower temperatures and are probably also responsible for the occurrence of rnixc.d-tnode oscillutions observed under certain conditions ( 2 4 ) . A typical time series belonging to this category is shown in Fig. 16. These types of oscillations are periodic mixtures of two modes of oscillations, namely, a small harmonic one with short periods and a large one exhibiting rather sudden changes (relaxation oscillation) (77). The amplitudes of the rapid oscillation increase, in this case, continuously (this corresponds in phase space to a spiraling of the trajectories) and then change abruptly to the slowly varying high level of the relaxation model. This is sometimes called the “hour glass” effect, because of the characteristic shape of the associated trajectories in phase space. Experimental data such as those of

OSCILLATORY CATALYTIC REACTIONS

235

FIG.15. Three-dimensional attractors constructed by the time-delay method from several of the experimental time series reproduced in Fig. 13. (From Ref. 71.)

Fig. 16 are only rarely strictly periodic, because usually rather small fluctuations in the external parameters are sufficient to trigger abrupt changes. However, in principle, mixed-mode oscillations belong to the category of multiple-periodic limit cycles. If the behavior is governed by two incommensurate frequencies, i.e., the ratio of two periodicities is an irrational number. This situation is denoted by quasiperiodicity and has been realized experimentally with periodically forced oscillations, as will be described next.

236

G . ERTL

160

120 80

40 I

1

2

3

4

5

6

7

0

Q I 0 1 1 1 2 1 3 1 4 1 5

t [loo SQC] FIG. 16. A typical example for “mixed-mode” oscillations with the CO/Oz reaclion o n Pt(l 10); p02 = 2.0 x torr. pco = 6.6 x lo-’ torr, T = 531 K . (From Ref. 2 4 . )

E. FORCED OSCILLATIONS So far, only situations were considered for which the external control parameters pk were kept constant, i.e., time independent, giving rise to the autonomous temporal behavior of the system. If, on the other hand, one (or several) of these parameters is modulated periodically, the system will respond with forced behavior. The resulting phenomena have been studied extensively for homogeneous reactions (78-86), as well as for several heterogeneously catalyzed systems (87-90). However, the latter suffered from the fact the autonomous oscillations were usually not very regular and not characterized by a well-defined frequency v,. It is, however, just the ratio of v, over the frequency of the external modulation v p that (together with the forcing amplitude A ) determines the nature of the dynamic response. The same shortcoming was encountered in a study with the CO oxidation on Pt(100 (33),wherein fairly large forcing amplitudes (of about 10%) were needed in order to cause appreciable effects, which were essentially restricted to responses of the systems with the same frequency as applied externally (harmonic entrainment). A very rich scenario was, on the other hand, observed with Pt(l10) if perturbed under conditions for which regular autonomous oscillations with well-defined frequencies existed (9/-93). Typically, these autonomous oscillations were established at fixed external parameters and then one of the partial pressures was periodically modulated by use of a feedbackregulated gas inlet system with frequencies up to 0.5 s - ’ and relative amplitudes around 1% (31, 33). Following the pioneering mathematical treatment of forced nonlinear oscillations by Kai and Tomita (80), the results can be rationalized in terms of a dynamic phase diagram chardcterizing the response of the system as a function of the amplitude A and of the period of the pressure modulation T,, with respect to that of the

237

OSCILLATORY CATALYTIC REACTIONS

‘/‘AWJ

a: subh.

112 213

I

1

b: superh.

./.

312 513

2

FIG.17. Experimental dynamic phase diagram for forced oscillations in the CWO?reaction on Pt(l10). Periodic modulation of the 0: pressure with amplitude A (as a percentage of the constant base pressure) and period length, T,,, with respect to that of the autonomous oscillations, T<,.(From Ref. 9/ .) Fixed parameters, for the subharmonic (superharmonic) range: P O , = 3.0 (4.15) x torr, T = 525 (530) K. torr, pc0 = 1.6 (2.1) x

autonomous oscillation Tc,. Figure 17 shows the results of a series of experiments in which the 0, pressure was modulated, while pco and T were kept fixed. Data for T,,IT, < I were taken at somewhat different conditions as those for T,,/T, > 1 , because the range of attainable external frequencies was experimentally limited. The unshaded areas denote conditions for which the system responds with a fixed phase difference to the perturbation (entrainment); within the shaded regions this phase difference varies continuously (quasiperiodicity). In particular one may distinguish between three situations, which are examined in the following discussions.

1.

Hurmonic Entrainment

The range around T,,IT, = I (which widens with increasing amplitude A) is characterized by the same frequency of modulation and response. A typical time series of this category is reproduced in Fig. 18. After switching on a periodic modulation of the 0, pressure with I .2% amplitude, that of the system increases considerably and after a short transient period it is phase locked and oscillates with the same period as that of the modulation, T, = Tex,whereby T, is different from the period To of the unperturbed oscillation. Similar to a linear oscillator, the phase shift varies between 0

238

0.ERTL

4.05 CI

0

a

3.95

50.

I

40

30 20

I

100

t csec:

200

FIG. 18. Forced oscillations with Pt(I 10): harmonic entrainment. Modulation of the 0: pressure by 1.2% with a frequency of 0.20 s - ' , which is close to that of the autonomous torr, pco = 2 x lo-' torr, 7 = oscillations (0.16 s - I). (From Ref. Y3.) pO? = 4 X S30 K.

*

and 7~ with decreasing T,, (i.e., increasing modulation frequency) and becomes 7 d 2 for T,, = To, whereas the amplitude of the response becomes maximum-as for classical resonance behavior.

2 . Siihhurmonic and Superhurmonic Entrainment Outside the region of harmonic entrainment the phase diagram contains further entrainment bands with phase locking, which are characterized by kT,, = &T,,whereby k and & are small integers. For k/& > I the entrainment is called superharmonic; for k / & < I it is subharmonic. Figure 19 reproduces a series of typical time series belonging into this category: 1/2 subharmonic, as well as 2/1 and even 7/2 superharmonic. Note that with the latter the amplitudes of the response are of unequal height, but the same pattern of k amplitudes repeats in regular intervals of & periods of the modulation. For harmonic entrainment there existed a single fixed phase shift between perturbation and response, but now the time series

(D

c

2 - i n FIG. 19. Forced oscillations with Pt(l10): entrainment. (From Ref. 93.) {a) L/2 subhartorr. T = 525 K, To = I 1 s, T,, = torr, pco = 1.7 x monic behavior: prl. = 3 x 5.5 Y , A :1.2%. (bj 2/ I superharmonic behavior: p0, = 4.15 X 1V5t o n , pro = I .7 X ion. T = 530 K, To = 4.0 s, T,, = 9. I s, A = 1.2%.-(c) 7/2 superharmonic behavior: poz = 4.15 x IO-'torr, pco = 2.0 x IO-' torr, T = 530 K , To = 5.4 s, T,, = 21.3 s, A = 1.2%.

240

G . ERTL

are characterized by a small number ( k ) of phase shifts during ti periods of the modulations. 3.

Quasiperiodic Behavior

The shaded regions in Fig. 17 outside the entrainment bands mark conditions for which the phase relation between modulation and response varies continuously. This is due to the fact that k l e is now no longer the ratio of two small integers. The time series of the response consists no more of a single self-repeating pattern, but the temporal behavior has now to be described by two periodicities, which are incomrnenwrate, i.e., their ratio is an irrational number. An example is reproduced in Fig. 20. An enlarged section (Fig. 20b) illustrates that here upproximately 25 periods of the response coincide with 12 of the perturbation, i.e., a situation that is slightly outside a simple 2/1 entrainment. These 25 periods mark the overall time (called the beat period) during which approximately the sequence of phase shifts repeats. Twoother features that are observed experimentally and are in agreement with theoretical prediction shall only be mentioned briefly. First, the main frequency of the response is not equal to the frequency of the autonomous, i.e., unperturbed oscillation, but is still influenced by the frequency of the modulation, even if no entrainment occurs-an effect that is called “frequency pulling” (94).Second, the response of the switchingon of a periodic perturbation is usually rather rapid (cf. Fig. 18). However, near an entrainment edge (i.e., close to conditions for which the behavior of the system becomes quasiperiodic), the transient time becomes longer and longer, an effect that is denoted as “critical slowingdown” (85,95).It has its counterpart with static phase diagrams, wherein phase transitions are associated with an increase of the correlation length (96). The qualitative features of the experimental phase diagram of Fig. 17 are in agreement with the general conclusions from theoretical treatments (78-83, 97). With respect to the present particular system, Kasai et al. (98) solved the system of equations, originally established for description of the kinetic oscillations on Pt( 100) (29),with periodic pO,modulation and could qualitatively reproduce some of the entrainment bands. Figure 21 shows the result of a recent theoretical treatment based on Eqs. (4) ( 5 4 , and (6) derived for Pt(1 lo), which inter alia were able to reproduce the occurrence of a Hopf bifurcation of the autonomous system (69). The calculated diagram contains, apart from the entrainment bands and regions of quasiperiodicity at small forcing amplitudes A , further complex fine structure for larger A , for which complete bifurcation analysis revealed, e.g., Feigenbaum as well as intermittency transitions to chaos. This under-

130

ia0

430

300 t (sac)

sOO

--

I -

I

25

b

'5 30 -

-E

4

a

20-

5-I

I

50

I

100

I

J

150

t Isecl

F I G .20. Forced oscillations with Pt(l10): quasiperiodic behavior. (From Ref. Y3.) (a) Full time series; (b) enlarged section. pOq = 4 x l o - ' torr, pco = 2 x I O - ' torr, T = 530 K , To = 5.2 S, TeX= 1 1 . 1 s , A = 0.6%.

242

G. ERIL ....,... ns

____ pd

-snp

0.2

0.5

1.0

1.5

2.0

Fic;. 21. Theoretical dynamic phase diagram for forced oscillations on Pt( 110). evaluated from Eqs. (4). (5a). and ( 6 ) . The calculations have been performed up to larger forcing amplitudes than were experimentally accessible (compare with Fig. 17). For small A , the bands characterizing entrainment and quasiperiodicity are discernible, whereas for larger A more complex bifurcations result. Types of bifurcations: ns, Neimark-Sacker; pd, period doubling; snp, saddle-node of periodic orbits. (From Ref. 69.)

lines the rich variety of dynamic phenomena that may result from solution of a quite simple system of coupled, nonlinear differential equations.

F. FACETING Figure 1 showed data for the rate of CO, formation on both a flat Pt( I 10) surface and a plane consisting of a sequence of steps and terraces. At high p c O , i.e., at conditions for which the rate is limited by oxygen adsorption, the stepped surface exhibited considerably higher activity. I n fact, the formation of these facets on the initially flat surface occurred during the course of the reaction and was associated with a continuous increase of the rate at fixed external conditions (39). The progress of the structural transformation could be followed by recording the continuous splitting of low-energy electron diffraction (LEED) beams (Fig. 2 2 ) , which data indicate the formation of periodic arrays of steps and terraces approaching inclined (430) and (210) plane orientations, respectively (39, 99). New adsorption states for oxygen with enhanced sticking coefficients are associ-

OSCILLATORY CATALYTIC REACTIONS

*...*

1.1

!

I

-+”

243

Pt(llO1 Ts = 5LO K P~:1.0 xlO-‘To:r Pco:9.7x10”Torr E =67eV

I

I

0.1

-

s~iZlo)--

O,

- covered I surlocc

A

l

h

t

f

0.1

1k.l

RECIPROCAL SPACE

[OOl]-zone

__c

FIG.22. LEED intensity profiles for Pt( 110) in a COiO? mixture showing continuous splitting due to formation of facets. (From Ref. 39.)

244

G . ERTL

ated with these altered surface structures, which accounts for the increased reactivity (100, 101). Variations of the topography of a catalyst surface during the reaction are observed quite frequently (102-104) and were reported as early as 1922 by Langmuir (105) for the catalytic oxidation of CO at a Pt wire. Recent investigations of Pt single-crystal spheres by means of reflection electron microscopy revealed that this effect is not restricted to the Pt( 110) surface, but occurs also with Pt( 100) and (to a much lesser extent) even with Pt( 1 I I ) (106). The faceting process under discussion in the present context does not occur by interaction of either one of the reactants alone, but requires conditions of ongoing reaction. In turn the flat surfaces are restored by thermal annealing, if the flow of one of the reactants is switched off. For Pt(llO), at sufficiently high temperature and low pressures the rate of annealing exceeds that for the creation offacets, and hence uniform kinetic oscillations with long-time stability, as described in the preceding sections, may be observed. Obviously these structural transformations are not driven by equilibrium thermodynamics (i.e., the tendency for lowering the surface free energy) but by kinetics. The increase in rate associated with the progressing faceting indicates an increase in net entropy production, in contrast to nearequilibrium processes for which entropy production always approaches a minimum. Part of the energy released by the ongoing reaction is used to create spatial differentiation of the reacting medium, for which the term “dissipative structures” was coined by Prigogine and his school (107). More specifically, if we return to the fundamental Eq. (2) describing both the temporal and spatial evolution of a nonlinear system of coupled variables

there may exist solutions for thex, that are constant in time (i.e., stationary states) but vary with the spatial coordinates r . The formation of such stationary spatial patterns in a reaction-diffusion system from an initially completely uniform situation is known as Turing instability (108, 109), and the formation of periodic facets during the course of the CO oxidation on Pt( I 10) can be considered as experimental verification of this effect. From the microscopic point of view, again the CO-induced I x 2 -+ 1 x 1 structural transformation of the Pt( 110) surface (as also underlying the mechanism of temporal oscillations) is of crucial importance for the development of facets, as becomes evident from the fact that this effect is restricted to conditions of high stationary CO coverages. Simply speaking, CO adsorption lifts the I x 2 reconstruction and simultaneously creates

245

OSCILLATORY CATALYTIC REACTIONS

[I 101

Number of Cycles

[iioi

-

-

“

4

3

-

6000-

GOO0 -

2ol .-

$ 3 3000 - 1 1500 -

0

0 I

,

300

-

defect sites with enhanced oxygen sticking coefficient, at which the adsorbed CO is reacted off again, so that the surface tends to transform back into the 1 x 2 phase. In this way the surface is continuously “ploughed” during the reaction. In fact, the resulting surface morphology is not randomly roughened but is characterized by a tendency for the formation of a periodic hill-and-valley structure, as was concluded from analysis of the LEED data (99). This effect could qualitatively nicely be reproduced by computer simulations based on the Monte Carlo technique (110). In the I x 2 transformations underlying model it was assumed that by the I x I monoatomic steps were either created or annihilated in order to compensate the differing atomic densities of the two phases. A series of typical results is reproduced in Fig. 23. With conditions for which the CO coverage fluctuates around the values for the 1 x 2 S 1 x 1 transitions, a periodically faceted structure develops from the initially flat surface with a period length of about 40 lattice sites. After switching off the gas flow, the flat surface is restored by thermal ordering-just as observed experimentally. This demonstrates that indeed spatial patterns with rather long period lengths may develop under reaction conditions, even if (as underlying the computer model) only interactions between nearest neighbors are invoked. The formation of facets may take place under conditions for which kinetic oscillations also occur, and of course the latter will be affected,

246

G. E R 7 L

h

PI (1101

5,.1.25xlO-'Torr

covered surloce belore CO inlroduction

---Clod

, PCO*2.7xlU'brr

Ts = 3 7 0 K

--c I

t =lhr period -GO s

0

a

period-11s

.-*

100s

Time

FIG. 24. Gradual increase of the oscillating period due to simultaneous progress of faceting of a Pt( 110) surface. (From Ref. 23.)

since faceting alters, e.g., the kinetics of oxygen adsorption (23). As can be seen in Fig. 24, for example, the period of the oscillations may vary substantially with progressingfaceting. Aglance at Fig. I , however, makes plausible that, at fixed external parameters, a nonoscillating system may even slowly move into conditions for oscillations. This was the case with the example reproduced in Fig. 25 ( 2 4 , where adjustment of the new pressure conditions (arrow) did not cause an instantaneous response of the rate to a new stationary value. Instead, it started to increase slowly due to the onset of faceting, and after about 10 min kinetic oscilhtions with increasing amplitude (and also increasing period, similar to Fig. 24) developed.

G . SPATIOTEMPORAL SELF-ORGANIZATION The kinetic oscillations occurring on Pt( 1 10) at high temperatures (at which the additional effect of faceting is unimportant) are usually rather regular, but with Pt( IOO), regular oscillations are the exception. This difference is essentially a consequence of the different mechanisms that govern spatial coupling of temporal oscillations between different regions of the

247

OSCILLATORY CATALYTIC REACTIONS

300 250 h

> E 200 v

8

a 150 100

so 2

4

6

8

10

12

14

16

18

20

22

t (100 sec) torr, T = 470 K. At the point FIG.25. CO/Oz reaction on Pt(ll0); pco = 2.3 x marked by an arrow, poq was changed from 1.5 to 2.0 x 10.‘ torr. The stationary reaction rate starts to increase slowly due to faceting until oscillatory conditions are reached. (From Ref. 24.)

surfaces. Without the operation of an efficient coupling mechanism a macroscopic system would not be able to exhibit temporal self-organization of an integral property such as the reaction rate (or, as with the present example, the work function, which is recorded as an integral quantity over a surface area of about 30 mm’), but otherwise the different parts would oscillate with uncorrelated phase and the net result would simply be an averaged constant stationary value. The general solutions of the fundamental systems of nonlinear equations [Eq. ( 2 ) ]will be of the type wherein the state variables are dependent both on time and space, which will manifest in the form of wave propagation. Coupling between several parts of the system will be transmitted through the generalized diffusion coefficient D.If the associated transport process proceeds on a time scale comparable to or slower than the period of the temporal oscillation, macroscopic wave propagation phenomena are to be expected, as, for example, realized with the Belousov-Zhabotinsky

248

G . ERTL

reaction, for which concentration differences can easily be made visible through changes in color (111-113). If, in this case, the solution is stirred, the whole system changes simultaneously and periodically its color-now the transport process (convection)-occurs much faster than the temporal concentration changes caused by the chemical reaction. The possibility for the formation of spatiotemporal patterns with heterogeneously catalyzed reaction systems was pointed out early (114-1 16). Here the following transport mechanisms may be responsible for spatial coupling: a . Hecit Conductcrncv. Because chemical reactions are associated with heat effects, local differences in reaction rate will also cause local temperature differences, which may, by heat conductance, propagate wavelike across the macroscopic system and synchronize, e.g., separate catalyst particles on an oxide support. This is usually the mechanism dominating kinetic oscillations near atmospheric pressure, for which temperature excursions exceeding 100 K were observed (86,117-125). Recent computer simulations revealed that coupling between different oscillators (modeling, e.g., individual catalyst pellets) by heat conductance may even give rise to chaotic behavior of the integral reaction rate (126). With the low-pressure experiments at massive single-crystal samples (exhibiting high heat conductivity), however, the heat released by the reaction was at least one order of magnitude smaller than that continuously fed into the crystal in order to keep its temperature constant to within 0.05 K. Therefore strictly isothermal conditions within these limits were fullfilled with the experiments under discussion here and effects due to heat conductance can be neglected. h. Coupling ihrough thc G N SPhuse. Any variation of the reaction rate will cause changes in the partial pressures of the species involved in the reaction, whose magnitude depends on the flow rate, etc., and which with the present systems could reach values of the order of 1% during kinetic oscillations, as can be seen, for example, from Fig. 4. Because any variation of the partial pressure under low-pressure conditions (< torr) propagates with the mean molecular velocity of about 1000 m/s, it will reach any other part of the system almost instantaneously, i.e., within a time (< 10 s), which is orders of magnitude shorter than the periods of oscillation. This is the counterpart of the stirring effect with the BZ reaction, and similarly the whole system is expected to oscillate essentially in phase. This is actually the mechanism dominating the regular high-temperature oscillations on Pt( 1 10) (40),which becomes plausible based on the following arguments: ( I ) any increase of the reaction rate will cause a lowering of the CO pressure, which in turn favors a higher reaction rate; (2) the

-'

OSCILLATORY CATALYTIC REACTIONS

249

oscillations always occur under conditions for which oxygen adsorption is rate limiting, which becomes enhanced by a reduction of the CO coverage (pressure), as demonstrated by Fig. I . That means there exists a positive feedback, which may also trigger less reactive regions. As outlined in Section II,C, with Pt( 110) the oscillations are restricted to a very narrow pressure range, which is a direct consequence of the fairly small difference in oxygen sticking coefficient for the two surface modifications, and the experiments on forced oscillations described in Section II,E demonstrated that the system responded sensitively to periodic partial pressure modulations with amplitudes < 1%. The variation of the partial pressures under stationary conditions of autonomous oscillations may become of comparable magnitude, and hence it is quite obvious that feedback effects through the gas phase will be of importance. To a first approximation, spatial differences of the surface concentrations will be of minor influence. This justifies the neglect of any spatial effects in the corresponding theoretical formulation, as was done with the system of ordinary differential equations [Eqs. (4), (5a), and (6)] to model the temporal oscillations with Pt(l10). A perfectly homogeneous surface is expected to oscillate, under these conditions, uniformly in phase, whereby statistical fluctuations will trigger the oscillations at random locations. On the other hand, because the oscillatory properties of Pt( I 10) are affected by the defect structure (cf. section ILF), spatial differences might still occur, and the following experimental evidence was found: measurements in which the oscillations were monitored at different parts of the surface by means of two Kelvin (A@) probes revealed that the oscillations always exhibited the same frequency, but were frequently of different shape and amplitude and had a nonvanishing phase difference. This indicates that the two regions differed slightly in their adsorptive properties and were mutually affecting each other through the gas phase (127).Current investigations by means of photoemission electron microscopy (PEEM; see later) suggest, on the other hand, that even with Pt( 110) frequently not pure gas-phase coupling prevails, although conditions may be found for which indeed macroscopic areas switch their adsorption state almost instantaneously in phase (128). c. Coupling by Surjiuce DiffmsionlReaction. This process dominates with Pt(100), for which plane coupling through the gas phase plays no significant role. It is associated with the propagation of concentration waves across the surface and may give rise to interesting phenomena of spatiotemporal pattern formation. Due to the large difference in oxygen sticking coefficient between the hex and I x I phases of Pt(100), the oscillatory conditions extend over a rather broad partial pressure range. External forcing of the oscillations requires amplitudes >5% of the pressure modulation, and hence the communication between different parts of

250

G . ERTI.

the surface through the much smaller pressure fluctuations associated with the reaction under stationary conditions are without any relevance. Local differences in coverage initiate surface diffusion, and this process will be coupled to the ongoing reaction and gives rise to the occurrence of “chemical waves” (125’). Effects of this type were reported as early as 1906 by Luther (130). Experimental verification of spatial pattern formation and wave propagation associated with the oscillatory CO oxidation on Pt(100) was first achieved by rastering the electron beam used for LEED across the surface while simultaneously the LEED pattern was continuously monitored by a computer-controlled video camera (27, 30). I n this way the structural properties of a 30 mm2 surface area could be monitored within 10 s with a lateral resolution of about 0.5 mm. A typical set of data associated with large-amplitude kinetic oscillations is reproduced in Fig. 26. It shows the LEED intensity distributions over the 4 x 7-mm’ surface area from the c2 x 2 (CO coverage) and the hex (adsorbate free) phases with progressing time, from which the large-scale wave propagation process becomes clearly evident. These data explain also the different shapes of the LEED intensity and work function oscillations reproduced in Fig. 8. The LEED intensity varies fairly abruptly as soon as a wavefront passes the small area (-0.2 mm’) probed by this technique, while A@ records the integral properties of a much larger area of about 30 mm’. Recent experimental developments enabled considerable improvement in both temporal and lateral resolution. The various surface phases differ considerably with respect to their work function, which effect gives, in turn, rise to differing electron yields emitted from the surface under the influence of U V irradiation. If this radiation is focused onto a small spot ( - I p m diameter) of the surface that is continuously rdstered, an image representing lateral work function variations can be constructed. This is the principle of scanning photoemission microscopy (SPM) (131).Alternatively, a larger area of the surface may be illuminated and the emitted electrons are imaged on a screen through suitable electron optics, photoemission electron microscopy (PEEM) ( / 3 2 - / 3 5 ) ,a simplified version of which was recently applied to monitor spatiotemporal pattern evolution with high spatial as well as temporal (-20 ms) resolution (136). Figure 27 shows a spatial pattern recorded by SPM from a Pt(lO0) surface during kinetic oscillations; the grey scale represents the varying electron yields (from dark = high work function = oxygen-covered region to bright = low work function = adsorbate-free region). The wave nature of the pattern reminiscent of the concentric rings frequently observed with BZ reaction becomes clearly evident. A sequence of images recorded with the PEEM technique at various time intervals is reproduced in Fig. 28

OSCILLATORY CATALYTIC REACTIONS

0

x

'

25 I

I

v

A

'

20

W

z

c (

+ 40

60

t 80

FIG.26. LEED intensity distributions over a 4 x 7-mm' Pt( 100) sample from spots characterizing the CO c2 X 2 structure (left) and the hex structure (right) during kinetic oscillations, illustrating the propagation of waves of structural transformations across the surface. (From Ref. 30.)

(136).The irregular shapes of the propagating wave fronts are mainly due to spontaneous nucleation induced by random fluctuations. These data explain why the integral temporal behavior of this system as monitored through the overall reaction rate is mostly irregular and not of the type of harmonic oscillations, as is more characteristic for Pt( 110). More regular temporal behavior of Pt( 100) was, on the other hand, always found when the waves emanated repeatedly from an edge zone of the sample (as, for example, with Fig. 26), where the higher defect concentration with its enhanced oxygen sticking coefficient could act as trigger zone.

252

M

c

.-C

x

8 I)

FIG.28. Sequence of snapshots of the spatial pattern evolution on a Pt(100) surface during oscillatory CO oxidation as recorded by photoemission electron microscopy. (From Ref. 136.)

254

G . ERTL

0.0 0 1

h

0

T

10

20

30

LO

compartment number FIG.29. Theoretical (one-dimensional) spatial pattern evolution on Pt( 100). (From Ref. 29.)

Theoretical description of the spatiotemporal self-organization for Pt( 100) was achieved both in terms of solution of a suitable set of partial differential equations (29) as well as with computer simulations using the cellular automata technique (32). The former approach was based on a one-dimensional model in which only diffusion of adsorbed CO was taken into consideration, because O,, is known to be far less mobile. The surface was represented by 40 compartments, from which one at the edge was associated with an increased oxygen sticking coefficient, and adsorbed CO was to diffuse between neighboring compartments under the influence of a concentration gradient. A typical set of resulting spatial patterns at various times is reproduced in Fig. 29; it shows, e.g., the formation and propagation of CO and hex waves in analogy to the experimental data of Fig. 26. Qualitative similar features were found in the (two-dimensional) computer simulations (32). A series of resulting “snapshots” for the distribu-

255

OSCILLATORY CATALYTIC REACTIONS

a

b

C

d

0-1x1

co-1x1

e

d

h

0-1x1

co-1x1

FIG. 30. Computer simulation by the cellular automaton technique of spatial pattern evolution during CO oxidation on Pt( 100). (From Ref. 3 2 . )

tion of adsorbed CO and 0 (both on the I x 1 phase) is shown in Fig. 30, from which again the effect of wave propagation becomes evident. These waves were preferentially triggered at defect sites (i.e., in regions with an increased oxygen sticking coefficient), but they were also found to form spontaneously by fluctuations on a perfectly uniform surface. The highresolution images recorded with the PEEM technique as reproduced in Fig. 28 revealed quite frequently that wave nuclei did not necessarily emerge repeatedly within the same regions of the surface. This points to the importance of spontaneous symmetry breaking even under experimental conditions, which are, of course, never completely defect free. Qualitatively, the formation of chemical waves with the present system may be made plausible in the following way. We consider two adjacent regions on the surface, one with a lower CO concentration, onto which oxygen adsorption and reactive removal (the autocatalytic step!) is possible (I), and the other one with a higher CO concentration, which inhibits oxygen adsorption (11). Driven by the existing concentration gradient, adsorbed CO will diffuse from 11 to 1, the effect being that the reacting zone now extends into region 11 and the wave propagates. According to this model it should also be possible to initiate propagating reaction fronts even if the system does not oscillate autonomously, by a

256

G . ERTL

pt (100)Single Crystal

I/

Laser Spot

Pt (100)/c0/02 T =L80K 7

-'d

N

po2 =0.1xXiLmbar pco=3.7x10~5mbar

Laser

i

I

0

= 0.53 MW/cm2

LO

0

2

L

I

l

l

,

,

,

6

0

10

12

1L

16

t Iminl FIG.3 I . Initiation of a chemical wave on Pt( 100) by rapidly dewrbing CO locally by means of a laser pulse. The integral rate of CO? formation increases immediately after the pulse. while the local A@ signal is delayed due to the finite velocity of wave propagation. (From Ref. 137.)

l o c u l perturbation enabling triggering through a momentary increase of the reactivity. This could be realized experimentally by thermally desorbing CO from a small spot of the surface by means of an infrared laser pulse while the system was operated under stationary conditions ( pco, po,, and T ) close to those for autonomous oscillations (137). The propagation of the resulting chemical waves was monitored by a small work function probe mounted at a distance of several millimeters from the point of irradiation, and the total reaction rate was recorded by means of a quadrupole mass spectrometer. Figure 3 1 shows a typical result together with a sketch of the experimental arrangement. The laser beam desorbs CO from a spot of about 0.5 mm? during a fraction o f a second. As can be seen, the rate of COz production starts to increase immediately after the laser flash, passes through a main

OSCILLATORY CATALYTIC REACTIONS

257

maximum and through two additional weak maxima, and then returns to its initial value. The work function probe, on the other hand, shows a change of its signal only about 1.5 min after the laser shot-this is obviously the time required by the reaction front with its increased oxygen coverage (enhanced A@) to reach the probe. The A@ signal again exhibits three maxima with successively lower amplitude before it returns to its initial value. This result indicates that in this case the laser shot initiated the emanation of several consecutive wavefronts with decaying amplitude. Other examples include the creation of a single wavefront or of only a weak transient increase of the rate without a wavefront reaching the work function probe (damped waves). The different possible situations are illustrated schematically in Fig. 32. The velocity of propagation of the wavefront was found to be around 2 mm/min at 480 K and to increase with temperature. These values are of the same order (albeit associated with some scatter) as those derived from the imaging of spatiotemporal pattern evolution during autonomous oscillations (27, 138). With a polycrystalline Pt film and the same reaction but in the atmospheric pressure range, Dath and Dauchot (122) derived (through measuring local changes of the electrical resistivity) wave velocities of the order of 60 mm/min. Even if with the present system the external conditions were adjusted in a way to enable the initiation of wave propagation, additional requirements had to be fullfilled. The laser power had to exceed a threshold in order to cause sufficient CO desorption, and a second laser pulse only excited another wave if the preceding excursion had decayed completely, i.e., the system exhibits a refractory period. The observed phenomena clearly belong to the category of trigger waves (129,139),which are governed by coupling between reaction and diffusion. Quite remarkably, the essential criteria of these features were recognized in the first report on this subject by Luther (130), namely, the need for an autocatalytic process, a threshold of the intensity of the perturbation, and the existence of a refractory period. He even presented the basic equation for the velocity of wave pro a ation, which was derived in detail only much later (129, 130): up = D K , with D being a diffusion coefficient and K an “effective” (pseudo-first-order) rate constant. If one inserts for D the coefficient for surface diffusion of adsorbed CO, and for K the effective value resulting for oxygen adsorption (the rate-limiting step), an estimate for up of about I mm/min results (140), which indeed agrees with the order of magnitude of the experimental findings. Recent PEEM experiments revealed that the wavelike propagation of coverage variations across the Pt( 100) surface during the CO oxidation are not restricted to the occurrence of autonomous oscillations or to local

J=

258

G . ERTL

Laser pulse

Kelvin probe

0 CO

covered surface 0 covered surface

I'

I'

I'

1'

I'

I'

single pulse wave

multiple pulse wave

damped wave

FIG.32. Schematic sketch of the various types of waves that may be excited by laser desorption on Pt(100). (From Ref. 137.)

perturbations near oscillatory conditions, but can quite generally be produced by changing one of the external parameters such as the temperature, which will cause a variation of the stationary state of the surface (141). As outlined above, for the regular oscillations on Pt( 100)at high temperatures, spatial coupling will predominantly be mediated through the gas phase, which process is occurring so rapidly that no phase difference between different parts of the surface are to be expected. However, with this system wave phenomena were also observed by means of PEEM under conditions for which the temporal oscillations were less regular

OSCILLATORY CATALYTIC REACTIONS

259

FIG.33. Spatial patterns during the oscillatory CO oxidation on Pt(l10). The two PEEM images were recorded at an interval of I min (From Ref. 136.)

(128, 136). Under certain conditions the switching of a large (-I mm’) surface area from one laterally uniform state to another one was found to proceed via a very rapidly (-1 mmisec) propagating extended wavefront. Most probably this phenomenon belongs to the category of kinematic waves (129, 142), which do not involve mass transport between different parts of the reacting medium, but are essentially an optical illusion caused by somewhat differing temporal oscillatory behavior of adjacent regions. Another type of phenomenon was characterized by the simultaneous appearance of nuclei at various points of the surface (indicating gas-phase coupling) whose further growth proceeded then much more slowly by trigger wave propagation. Sometimes patterns of particular beauty evolved in this way, and an example is reproduced in Fig. 33 (136).The two images were taken at an interval of I min and exhibit a nice spiral-type pattern evolution, even if one takes some electron optical distortions of the outer regions into account. The mode of coupling between different parts of the surface is also of relevance for the occurrence of temporal oscillations with polycrystalline samples, even if heat conduction as the usually dominating mechanism can be ruled out. Large-amplitude oscillations have, for example, been observed with a polycrystalline Pt ribbon under isothermal conditions (143). This suggests coupling between different crystallites of the same surface orientation through the gas phase, because differently oriented

260

G . ERTL

crystal planes will hardly oscillate under the same conditions (as can be seen from Fig. 1 I ) , so that synchronization via surface diffusion seems to be unlikely. No wave propagation effects were indeed found for the oscillatory CO oxidation in the I-torr pressure range on supported Pt catalysts under isothermal conditions ( / 4 3 u ) ,which supports this conclusion. Because surface defects have a pronounced influence on the oxygen sticking coefficient, these may of course also alter the conditions for the occurrence of oscillations from those determined for flat single-crystal surfaces. This became evident in connection with the effect of faceting of the Pt( 110) surface as described in Section 1I.F. I n order to model more closely the situation of a polycrystalline surface, recently a series of experiments with a cylindrical Pt single crystal were performed (37). The sample had its axis parallel to the [OOl] orientation and exhibited (loo), ( 1 lo), and (2 10) orientations and their continuous transitions. [The oscillatory behavior of Pt(210) will be discussed later.] It turned out that the pressure range for oscillations in the (100) region was considerably broader than with a flat Pt(100) crystal, which has to be attributed to the existence of defects with varying density. In addition, reaction fronts were found to propagate into adjacent regions, which alone would not have been oscillating. Widely separated Pt(l10) regions, on the other hand, coupled through the gas phase. In fact, conditions could be found for which almost the whole surface area was oscillating, although only one orientation was in its genuine existence region for autonomous oscillations. Because of the fairly large total surface area of the sample (-5 cm2) the concomitant variation in CO partial pressure became as large as up to 25% which, of course, suffices by far for coupling through the gas phase.

111. Other Systems

A. CO OXIDATION ON Pt(210) The mechanism of the kinetic oscillations occurring with the CO + Oz reaction on clean Pt(100) and Pt(l10) surfaces was based on the reversible transformation of the surface structure by the presence of adsorbed CO and by an associated variation of the oxygen sticking coefficient that increased upon CO-induced lifting of the reconstruction of the clean surface. The most densely packed Pt( 1 1 I ) surface is not reconstructed and its structure is also not affected by CO adsorption. Accordingly, kinetic oscillations with a clean Pt(1 I I ) surface (i.e., for partial pressure torr) could never be observed (13, 26, 27, 38). Again no reconstruction

OSCILLATORY CATALYTIC REACTIONS

1000 time

26 1

2 alo

(secl

F I G . 34. Development of kinetic oscillations with the CO/O, reaction o n Pt(210). After establishing the indicated conditions at point C, the system slowly evolved oscillations with continuously changing periods and amplitude. (From Ref. 38.)

that might be affected by adsorption of CO is observed with the rather open Pt(210) surface, and, in addition, the oxygen sticking coefficient on the clean surface is already so high (near unity) that no substantial further increase by a hypothetical CO-induced variation of the surface structure is feasible (38, 144).Quite surprisingly, this plane may exhibit nevertheless kinetic oscillations under low-pressure conditions and without t h e possible interference by self-poisoning, as first observed by Ehsasi et d.(35, 36, 38) and later confirmed with a cylindrical Pt single crystal exhibiting (210) regions as well as with another flat Pt(210) sample (37, 144). Interestingly, it was found that after establishing the proper external conditions the oscillations did usually not start immediately, but only after some delay time, during which presumably the surface underwent a continuous change of its properties. A typical time series is reproduced in Fig. 34 (38).At point C the indicated partial pressures were adjusted and the system started to oscillate with small amplitudes and short periods, and subsequently both features grow gradually. This behavior resembles closely the effects observed with Pt( 110) when kinetic oscillations were paralleled by facet formation (see Fig. 25) (23). With Pt(l10) the growth of higher index facets was found to lead to an incrc~aseof the steady-state rate of CO, formation due to the enhancement of the oxygen sticking coefficient with respect to that of the flat Pt(ll0) plane (37, 39). With Pt(210)just the opposite effect was observed. The stationary reaction rate decreased with time, but could be restored to its initial value by annealing the sample (without the reaction mixture)-just as with Pt(l lo), for which

262

G . ERTL

thermal annealing restored the flat plane. Pt(210) exhibits the highest sticking coefficient for oxygen among the planes of the [OOI] zone [which includes ( 1 lo)] (37, 38, 144). Hence the formation of other planes under reaction conditions will be associated with a decrease of the 0’ sticking coefficient and thus accounts for the observed reduction in activity. LEED investigations revealed that such afaceting process indeed also takes place with Pt(210). The observed splitting of certain diffraction spots indicates the formation of facets with predominant ( 1 10) and (310) orientations. In situ LEED observations showed, in addition, that during kinetic oscillations the intensities of the half-order diffraction spots from the ( I 10) facets also vary periodically with time. This suggests that even with Pt(210) basically a mechanism similar to that with Pt(l10) and (100) is responsible for the occurrence of low-pressure oscillations. It is in fact not the Pt(2 10) surface that exhibits oscillatory properties, but facet planes of the ( I 10) type, which are formed during the course of the reaction (144). Interestingly, as with Pt(l10) (39), the formation of facets on Pt(210) seems again not to be driven by thermodynamics, because it was not observed upon heating the surface in a pure oxygen atmosphere (144),but requires the continuous flow of free energy associated with the ongoing reaction.

B. CO OXIDATION ON Pd( 110) In contrast to Pt( 1 lo), the Pd( 110) surface is not reconstructed. Although t h e surface may undergo reconstruction under the influence of intermediate CO coverages ( / 4 5 )[i.e.,just opposite to Pt(l lo), where CO removes an existing reconstruction], there is little possibility for further increase of the already very high (-0.9) oxygen sticking coefficient (146).The starting situation hence resembles that for Pt(210), and again kinetic oscillations are observed (35,36, 147, 148) that cannot simply be attributed to adsorbate-driven transformations of the (two-dimensional) surface unit cell. The phenomenology is, however, quite different to that observed for Pt(2 10). There is no slow variation of the steady-state activity coupled to facet formation and, even more important, the oscillations require minimum O2 partial pressures, around torr. In addition, the necessary ratio po21 pco was always found to be high, on the order 10’ to 10’. Ehsasi r t cil. (147) followed the occurrence of these oscillations over a wide range of oxygen pressures, between and I torr, and found no indication for a change of the phenomenology at higher pressures. The temperatures could, on the other hand, be fairly low-oscillations are even observed near room temperature. At a first glance, the need for a fairly high oxygen

263

OSCILLATORY CATALYTIC REACTIONS

1

I

-600

-soo'

t "

I

100

'

"

'

I

200

'

"

t i m e Ls

"

300

"

"

400

I

FIG. 35. The CO/Oz reaction on Pd(l 10). Slight variation of the conditions caused a transition from harmonic to period-doubling behavior. (From Ref. 147.)

pressure appears surprising, because the oxygen sticking coefficient on the clean surface is quite large and closer inspection reveals that indeed the oscillations occur under conditions for high oxygen coverages. Figure 35 shows a typical example for a time series in which slight variation of one of the external parameters caused a transition from regular harmonic behavior to period doubling (147). Besides even more complex time series, irregular oscillations suggesting chaotic behavior were also observed, but have not yet, however, been fully analyzed as has Pt(l10). Apart from the lower limit for the oxygen pressure, a higher limit for the temperature (which increases with increasing p o l ) was also found to exist. Figure 36 shows the variation of both the reaction rate and the work function for fixed po2 and T when pco was either continuously raised or lowered, with the oscillatory range being shaded. The oscillations occur near the rate maximum, where a steep increase of A@ reflects the transition from an essential 0-covered to a CO-covered surface. [Note that with this system CO adsorption causes a larger work function increase than does oxygen adsorption ( / 4 9 ) . ]The shape of the rate v e r s u s p o curve is similar to that found generally (cf. Fig. I for comparison); however, the hysteresis behavior is quite different from the situation for platinum. The direction of the hysteresis is just reversed: lowering the pco with Pt causes a lower reaction rate due to the inhibiting effect of the CO adlayer, whereas with the present system it becomes even higher. This unexpected behavior is

264

G . ERTL,

Po? = 6 8 ~ 1 Torr 0~~ T =396K

' OI

0.9

Pco /

Torr

Fic. 36. Hysteresis effects with the work function A 4 and the rate of CO?formation on a Pd(l10) surface upon raising and lowering the CO pressure, while T and pO2were kept fixed. (From Ref. 148.)

found only for conditions of low T and high po,, for which oscillations may also occur, and both phenomena are indeed dosely linked to each other. The key for understanding these effects lies in the participation of a subsurface oxygen phase. At high O2 exposures oxygen atoms penetrate below the Pd(l10) surface and give rise to a complex LEED pattern and a reversal of the work function change. This species shows up in thermal desorption spectroscopy (TDS) as a distinct state (PI) and should not be confused with a true bulk oxide phase (148, 150). It has a lower reactivity toward CO impinging from the gas phase than the chemisorbed 0 species on the surface (p2 state in TDS). This becomes evident from Fig. 37, showing the evolution of COz as a function of time following exposure of a Pd( I 10) surface in low (solid line) or high oxygen concentration (dashed line) to a constant CO pressure at 540 K. With the chemisorbed phase (0= 0.50, characterized by a c2 x 4 LEED pattern) the rate increases instantaneously to its maximum and then decays due to depletion of the

OSCILLATORY CATALYTIC REACTIONS

0

10

20 Time 1s

265

30

FIG.37. CO titration curves of a Pd( 110) surface with small (solid line) and large (broken line) preexposure to oxygen. (From Ref. 148.)

adlayer, whereas with the subsurface oxygen phase (0= I .44, complex LEED pattern) the reactivity is fairly sluggish and attains its maximum only after some time. This effect is due to the lower adsorption energy for CO on the surface exhibiting the subsurface oxygen phase, and hence at 540 K the steady-state concentration of adsorbed CO will be reduced. This also explains why in Fig. 36 the rate to the left of the maximum is lower ifp,, is increased instead of decreased. In this range the rate is determined by CO adsorption, and if one starts from low pcO the surface will be in the subsurface oxygen state. As an additional effect, the presence of the subsurface oxygen species reduces substantially the sticking coefficient for further oxygen uptake by the surface (151 1. At high pco, on the other hand, the subsurface region will be depleted of oxygen by the presence of large concentrations of adsorbed CO and hence lowering of the CO pressure will yield a higher reactivity (remember that now oxygen adsorption is rate determining!) than ifp,, is changed in the opposite direction.

266

G . ERTL

The mechanism of kinetic oscillations involves again a periodic switching of the surface between states of low and high reactivity. If one starts with a CO-covered surface, the stored subsurface oxygen will segregate to the surface and reacts off. The decrease of the subsurface concentration will be associated with an increase of the oxygen sticking coefficient from the gas phase, which in turn will reduce the CO coverage. As a consequence, the surface will switch from an essentially CO-covered to a mainly 0-covered state. With the O;,, concentration being high, the subsurface reservoir becomes filled again-the oxygen sticking coefficient drops and at least the rate of CO adsorption will dominate again and the initial situation is restored. This mechanism resembles closely the “oxide” mechanism proposed earlier by Sales rt al. (11, 152) in order to explain kinetic oscillations at various platinum metals under high pressure conditions, with the important difference that the subsurface oxygen species under discussion here is distinctly different from a genuine bulk oxide phase. From in situ 1R spectroscopic investigations during the oscillatory CO oxidation on supported Pd particles at atmospheric pressures, Schuth and Wicke (153) concluded that the mechanism is based on a CO-induced transformation of the state of surface oxygen, which is in full accordance with the presented conclusions. Suhl (154) and Volokitin et c d . (155) performed a general mathematical analysis of kinetic models based on the existence of two types of oxygen that were shown to exhibit the possibility for the occurrence of oscillations. Quite recently Bassett and Imbihl (156) developed a full kinetic model based on the sketched mechanism, i.e., including the formation o f a subsurface oxygen species. The three basic steps [Eqs. (a)-(c) in Section Il,A] are supplemented by the process

*

Owh

(4

which leads to three differential equations for the variation of the concentrations of CO,,, Oad,and Osub.co, formation is assumed only to take place between CO,, and o l d , and the sticking propability for the latter species is set to decrease exponentially with the concentration of Osub.By using realistic values for the individual rate constants (which to a large exent were known from independent experiments), solution of these equations indeed provided close agreement with the experimental findings, including the shape of the hysteresis loop and the existence range for the occurrence of oscillations. The observation that O2 pressures >lo-’ torr (at 400 K) are required is simply a consequence of the kinetic parameters governing the formation (and removal) of the subsurface oxygen species. Spatial coupling with this system occurs again through the gas phase, as

267

OSCILLATORY CATALYTIC REACTIONS

was demonstrated by recent experiments with two separated samples (157).

C. THEREACTION NO

+ CO + 2N2 +

CO,

ON

Pt(100)

The first indications for the occurrence of kinetic oscillations with the NO + CO reaction on Pt(100) were reported in 1980 by Singh-Boparai and King (158).Phenomenaof this kind had been observed with polycrystalline torr range) even earlier by Adlhoch and platinum (at pressures in the Lintz (159), who speculated on the possible participation of a periodic variation of the surface structure (160). More systematic investigations with Pt( 100)were performed more recently by Schwartz and Schmidt (161, 162), who suggested that the oscillations are again driven by the hex + 1 X 1 structural transformation of the surface. These authors went even one step further and speculated that this principle might be common to all oscillatory reactions occurring o n platinum catalysts, whereby with polycrystalline samples the fraction of the total area exhibiting (100)orientation should be responsible for the overall effect. This is certainly not the case, and even for Pt(100) single-crystal surfaces the mechanism for the NO + CO reaction differs basically from that for the CO + 0, reaction, as was most recently explored in our laboratory (140, 163). There exist several major differences in the phenomenology of these two reactions. With the CO + 0, reaction, autonomous temporal oscillations are usually irregular, but can be sustained for deliberately long periods of time. With the NO + CO reaction, on the other hand, the shapes of the oscillations are usually regular, but these are damped, i.e., decay after several periods. In addition, with the latter reaction oscillations may be obtained in considerably lower pressure ranges (-lop7 torr). A typical trace of the rate of CO, formation under strictly isothermal conditions as a function of time is shown in Fig. 38, together with the intensities of selected LEED beams of the reconstructed (hex) and nonreconstructed surface phases (163).The oscillations were initiated by a rapid lowering of the temperature to a new stationary value and decay after a few periods. Intermediate short increase of the temperature by a few degrees and return to the initial value triggers the oscillations again. A similar observation was also made by Schwartz and Schmidt (162). The decay of the amplitude is definitely not due to poisoning or other irreversible changes of the surface, but has to be attributed to insufficient coupling between various parts of the surface area, which may be restored by small temperature variations. (To be more precise, the oscillations are only

268

G . ERTL

0

pNo = 4 ~ 1 0 ~ ~ r n b a rpco

~ x I O rnbar -~

4

b

2

T

394 K

0 tirne(min.1

FIG. 3X. NOKO reaction at Pt(100) showing damped oscillations after lowering the temperature to the indicated value. (From Ref. / 6 / . )

triggered by a decrease of temperature.) As can be seen in Fig. 38, the reaction rate (integrating over 30 mm2 surface area) reaches its constant value earlier than the intensity of the ( 1 8 ) LEED beam, which probes only about 0.1 mm’. In fact, this system responds very sensitively to periodic modulation of the temperature by forced oscillations. Modulation with an amplitude of 2 K suffices to establish substained undamped oscillations. This becomes evident from Fig. 39, which shows how the oscillations are damped if the temperature is held strictly constant (a),but become sustained under the influence of a weak periodic temperature modulation (b). Upon variation of the frequency, pronounced resonance behavior is observed. A remarkable consequence of this effect is that even random fluctuations of the temperature by about t1 K suffice to establish sustained oscillations (Fig. 40) with irregular amplitudes. Obviously the system selects its autonomous frequency from the Fourier spectrum of the random fluctuation. This leads to the conclusion that in this case spatial selforganization and hence large-amplitude rate oscillations are sensitively affected by coupling via temperature differences. A rapid lowering of the temperature will trigger the onset of oscillations simultaneously at various parts of the surface, which, however, will soon drift out of phase due to slight frequency differences if spatial coupling is insufficient. This was the case in our experiments, where the temperature

269

OSCILLATORY CATALYTIC REACTIONS

Pt (100)+ NO+CO v,=32 rnHz , pNO=Gx10-7rnbar, ~ ~ ~ = 3 x l O - ~ r n m b a r

y t-

,!&

: 1 [ [ 2g L

-

w t Irninl

FIG.39. NO/CO reaction on Pt(100). (From Ref. / 6 / . )(a) Damped oscillation of the rate if the temperature is held strictly constant. (b) Periodic modulation of the temperature by 2 K causes forced oscillations with appreciable amplitude.

Pt (100 I

+

NO * CO

Tmean=392K, pNO=4~10-7mbar, pCO:3~10'7rnbor

-I

I

'd

1

1 1

5

~

5 rnin

t [rnin]

FIG.40. Random temperature fluctuations by * I K suffice to establish sustained oscillations for the NOiCO reaction on Pt(100). (From Kef. 161.)

270

G. ERTL

was kept constant to within 0. I K and variations of the partial pressures never exceeded 0.5%. Schwartz and Schmidt (161), on the other hand, reported temperature fluctuations up to about 2 K and 3-10% variations of the partial pressures associated with t h e rate oscillations. It is thus plausible with their measurements that the oscillations sometimes continued for longer times, albeit they were always damped again. Undamped oscillations had been reported by Adlhoch et af. (160) for a Pt ribbon operated in the 10 ' torr range and by Schiith and Wicke (124) for a supported Pt catalyst working near atmospheric presure. An estimate for the former case yields temperature variations of the order of 10 K due to the exothermicity of the reaction; in the latter case even periodic changes by 5-25 K were measured-quite obviously heat conductance is efficient enough to synchronize the oscillatory behavior of these systems. With the low-pressure reaction at Pt( loo), for fixed NO and CO partial pressures (whose ratio ranges typically between 4 : 3 and 3 : l ) , two temperature regions exist, within which oscillations may occur and are separated by a roughly 30 K wide range characterized by stationary behavior. Although the mechanism responsible for the oscillations in the higher temperature range (around 440 K for pressures in the lo-' to IO-'torr range) is still unclear, it could be elucidated for the low-temperature branch (around 400 K) by combined application of LEED and rate measurements. In this case the oscillations are definitely not linked to a periodic hex I x I structural transformation of the surface. The coverages are always high enough to keep the surface in the nonreconstructed state. As becomes evident from Fig. 38, the LEED pattern did not exhibit any evidence for the intermediate appearance of the hex phase during the oscillations. The reaction proceeds through the following steps:

Formulation of the temporal variations of the coverages of CO, NO, and 0 in terms of three coupled differential equations (the recombination of 2N,, and desorption of N2 is much faster than the other processes and can hence be left without explicit consideration) leads indeed to oscillatory solutions without the need for additional inclusion of a surface-phase transition step. The physical reason lies in the fact that dissociation of adsorbed NO (step g) needs another free adsorption site and is inhibited if the total coverage exceeds a critical value [The adsorptive properties of

OSCILLATORY CATALYTIC REACTIONS

27 I

CO and NO on Pt( 100) are very similar, and both components are forming a mixed phase (164). However, NO dissociates on Pt(100) I x I but CO does not.] This effect is also responsible for the extremely narrow temperature-programmed reaction (TPR) peaks observed upon continuously increasing the temperature o f a NO + CO-covered surface (“surface explosion”) (165, 166). Qualitatively, the occurrence of oscillations can be rationalized in terms of the following steps: (1) At first the total coverage is so high, that NO dissociation is inhibited. (2) Once it starts by formation of free sites due to desorption or reaction between Od, + CO,,, the latter step will be autocatalytic and hence NO dissociation speeds up and the coverages of both NO and CO drop. ( 3 ) Because p N O> pco, there will be more O,d produced than reacted off by CO, thus the oxygen coverage increases and finally inhibits NO dissociation. (4) The NO coverage will increase again as well as that of CO (after consumption of the adsorbed oxygen), and one reaches the initial situation again. More detailed analysis reveals that the situation is somewhat more complicated; the occurrence of oscillations requires in fact a strong enough dependence of t h e rate constants for NO and CO desorption on coverage, as experimentally realized (140). but the basic mechanism is of the outlined nature. It does not necessarily involve the structural transformation of the Pt( 100)surface, as suggested previously (124, 161, 162). IV. Conclusions

Among oscillating chemical systems, catalytic reactions at well-defined single-crystal surfaces under low-pressure conditions are conceptually particularly attractive for the following reasons: The reacting medium is strictly two-dimensional and uniform on the mezoscopic length scale, which is relevant for spatial pattern formation. (Microscopic nonuniforrnities such as atomic steps may, nevertheless, be of major importance for the magnitude of kinetic parameters as well as for nucleation phenomena, as demonstrated.) In addition, the low partial pressures cause only minor heat production by the reaction, so that usually isothermal conditions can be kept to within a very good approximation. By combining kinetic measurements with surface-sensitive techniques, it became possible to explore the elementary steps underlying the complex temporal and spatial effects in Fair detail, which information, in turn, then could be used for theoretical modeling. The wide variety of phenomena, which were observed even for this class of (in principle) rather simple reaction systems, is a consequence of the rich scenario of effects predicted by nonlinear dynamics as well as of

272

G . ERTL

the various mechanisms governing the temporal as well as spatial selforganization of the different reactions. For the systems described here, the following mechanisms were found to determine the occurrence of isothermal kinetic oscillations: 1 . Adsorbate driven structural transformation (CO oxidation on platinum). 2. Formation and depletion of a subsurface oxygen species [CO oxidation on Pd( 1 lo)]. 3. Autocatalytic surface reaction [CO NO reaction on Pt( loo)].

+

In addition, the formation of new crystal planes (faceting) under the influence of the ongoing reaction may create conditions suitable for kinetic oscillations [cf. the CO oxidation on Pt(210)I or alter the properties of an oscillatory system [CO oxidation o n Pt(1lo)]. Further mechanisms may come into play at higher pressures, where the surfaces may undergo more profound chemical transformations, such as exemplified by the proposed “oxide” (10-13) or “carbon” (14, 15) models. It should be pointed out, however, that even for supported catalysts at high pressures, the same mechanism as derived from low-pressure single-crystal studies may be prevailing, as, for example, suggested for the CO oxidation on platinum (167). All these mechanisms have in common that the catalyst’s surface may periodically switch between states of high and low reactivity whereby an autocatalytic process has to be involved. The observation of macroscopic rate oscillations requires synchronization of various parts of the surface, which may be achieved through one of the following transport processes: J I . Heat conductance. This process will usually dominate with “real” catalysts near atmospheric pressure conditions, but was shown also to govern the CO + NO reaction on Pt(100). In this case sustained oscillations were observed under the influence of small ( ? I K) temperature fluctuations, but under strictly isothermal conditions these were damped. 2. Coupling through the gas phase. Any variations of the partial pressures associated with changing reaction rate will be transmitted practically instantaneously to other parts of the system. The oscillatory CO oxidation on Pt( 1 10) may be very sensitively affected by such pressure fluctuations and hence belongs in this category. 3. Coupling between surface diffusion and reaction. If this mode is dominating [such as with the CO oxidation on Pt( loo)], “chemical” waves propagating across the surface will give rise to spatiotemporal pattern formation.

OSCILLATORY CATALYTIC REACTIONS

273

Additional systems will certainly be explored in the future, and further refinement of experimental tools as well as of theoretical analysis will provide even more detailed insight into a fascinating area. REFERENCES 1. See, e.g., Zhabotinsky, A. M.. Brr. Blin.senges. Plzys. Chem. 84, 303 (1980); Gurel. O., and Gurel, D., “Oscillations in Chemical Reactions.” Springer-Verlag. Heidelberg, 1983. 2. Franck, U . F., Ber. Bitnsenges. Phys. Chem. 84, 334 (1980). 3. Hugo, P., Brr. Bunsenges. P h y s . Chrm. 74, 121 (1970). 4. Beusch, H., Fieguth, P., and Wicke, E., Chem.-lng.-’IPc/i.44, 445 (1972). 5. Sheintuch, M., and Schmitz, R. A., Cotal. Rev.-Sci. Eng. 15, 107 (1977). 6 . Slinko, M. G., and Slinko, M. M., Catid. Rev.-Sci. Eng. 17, I19 (1978). 7 . Razon, F., and Schmitz, R. A., C ~ r u lReu.-S(,i. . Eng. 28, 89 (1986). 8 . Ertl, G . , Norton, P. R., and Riistig. J., Phys. Rev. Letr. 49, 177 (1982). 9. Imbihl. R., in “Optimal Structures in Heterogeneous Reaction Systems” (P. J. Plath, ed.), p. 26. Springer-Verlag. Heidelberg, 1989. 10. Wicke, E.. Kummann, Keil, W., and Schiefler, J., Ber. Bunsenges. P h y s . Chem. 84, 315 (1980). 11. Sales, B. C.. Turner, J . E.. and Maple, M . B., Sirrf S r i . 114, 381 (1982). 12. Lindstrom, T. H.. and Tsotsis, T . T., Sur:f: Sci. 150, 487 (1985). 13. Yeates, R. C., Turner, J. E., Gellman, A . J.. and Somorjai, G. A , , S14rf: Sci. 149, 175 (1985). 14. Burrows, V. A . , Sundaresan, S . , Chabal, Y. J., and Christman, S. B., Surf. Sci. 180, 110 (1987). 15. Collins, N. A., Sundaresan, S . , and Chabal, Y . J . , Surf. Sci. 180, 136 (1987). 16. Eiswirth, R . M., Krischer, K.. and Ertl, G., in “Dynamics and Kinetics of Interface Reactions” ( M . Grunze and H. J., Kreuzer, eds.). A p p l . P h y s . A , , in press. 17. Thompson. J. M. T.. and Stewart, H. B . , “Nonlinear Dynamics and Chaos.” Wiley, New York, 1987. 18. Kuramoto, Y ., “Chemical Oscillations, Waves and Turbulence.” Springer-Verlag, Heidelberg, 1984. 19. Engel, T., and Ertl, G . , A h . C u r d . 28, I (1979). 20. Oh, S. H., Fischer, G. B., Carpenter, J . E., and Goodman, D. W., J . Catal. 100, 360 (1986). 21. Sault, A. G., and Goodman, D. W., Adv. C h i w . Pliys. 76, 153 (1989). 22. Comrie, C. M., and Lambert, R. M., J . C . S . Fartiday f 7 2 , 1659 (1986). 23. Ladas. S., Imbihl, R., and Ertl, G . , Surf: Sci. 198, 42 (1988). 24. Eiswirth, M., and Ertl, G., Surf. SLY. 177, 90 (1986). 25. Ehsasi, M., Matloch, M., Frank, O., Block, J. H., Christmann. K.. Rys. F. S., and Hirschwald, W., J . Chern. Pliys. 91, 4949 (1989). 26. Norton, P. R., Bindner, P. E., Griffiths. K., Jackman, T. E., Davies, J. A . , and Riistig, J . , J . Chem. Plivs. 80, 3859 (1984). 27. lmbihl, R . , Cox, M. P . , and Ertl., G., J . Chern. P h y s . 84, 3519 (1986). 28. Cox, M. P., Ertl, G . . Imbihl, R., and Riistig, J.. Surf. Sci. 134, L517 (1983). 29. Imbihl, R., Cox, M. P., Ertl, G., Miiller, H . , and Brenig, W.. J . Chem. Phys. 83, 1578 (1985). 30. Cox, M. P., Ertl, G . , and lmbihl, R . , Phys. Reu. Lett. 54, 1725 (1985).

274

G . ERTL

31. Eiswirth. M., Schwankner. R. J., and Ertl, G., Z . P h y s . Chem. 144, 59 (1985). 32. Moller. P . , Wetzl, K., Eiswirth, M., and Ertl, G.. J . C h m . Phys. 85, 5328 (1986). 33. Schwanker, R. J . , Eiswirth. M., Miiller, P., Wetzl. K., and Ertl, G., J . Chetn. P h y s . 87, 742 (1987). 34. Vishnevskii, A . L., and Savchenko, V. I . . Riwct. K i n e t . Cutul. Lcit. 38, I67 (1989). 35. Ehsasi, M., Block, J. H.. Christmann, K.. and Hirschwald, W., J . Voc. S c i . Twlinol. AS, 821 (1987). 36. Ehsasi, M., Rezaie3ere.j. S . , Block. J . H., Christmann, K., and Hirschwald, W.. Pro(.. f n t . Congr. Cutd., 9 r k Cir1gury 3, 1299 (1988). 37. Sander, M., Bassett, M. R.. Imbihl, R., and Ertl, G., Vacr4utn in press. 38. Ehsasi. M., Ph.D. Thesis, F U , Berlin. 1989. 39. Ladas, S . , Imbihl, R., and Ertl, G., Surf. Sci. 197, 153 (1988). 40. Eiswirth, M.. Moller, P., Wetzl, K., Imbihl, R., and Ertl, G., J . C l i m . Pl~ys.YO, SIO (1989). 41. Kellog, G. L . , Phys. R P U .L e t t . 55, 2168 (1985). 42. Niehus, H., S u r f Sci. 145, 407 (1984). 43. Jackman, T . E.. Davies, J. A., Jackson, D. P., Unertl, W. N.. and Norton, P. R . , SurJ: Sci. 120, 389 (1982). 44. Van Hove. M. A., Koestner, R. J., Stair, P. C., Biberian, J . P., Kesmodel. L. L., and BartoS. 1.. and Somorjai. G. A., S i r t j : ScY. 103, 189 (1981): 218. 45. Heilmann, P., Heinz, K., and Miiller, K., Sitt$ Sci. 83, 487 (1979). 46. Behm. K. J . . Thiel, P. A., Norton, P. R., and Ertl, G., J . C h m . Phvs. 78, 7438; 7448 (1983). 47. Heinz, K., Lang, E., Strauss. K . , and Muller, K., App1. Surf: S(,i. 11/12,61 I (1982). 48. Norton, P. R., Davies, J. A,, Creber, 0. K . , Sitter, C. W., and Jackman. T . E . , Surf: Sci. 108, 205 (1981). 49. Behm, R. J . , Hosler, W., Ritter, E.. and Binnig, G . , Pliy.s.’Reu.L a / / .56,228 (1986). 50. Reynolds, A. E., Kaletta, D., Ertl, G . , and Behm, R. J., Swf:f: Si,i. 218, 452 (1989). 51. Bonzel, H. P., and Ferrer S.,S w f Sci. 118, L263 (1982). 52. Imbihl. R., Ladas, S . . and Ertl. G.. Surf: Si,i. 206, L903 (1988). 53. Hoffmann, P.. Bare. S . R. and King, D. A., S w f : Sci. 117,245 (1982): 144,347 (1984). 54. Hayden, B. E.. Robinson, A. W., and Tucker, P. M., S~tyf.Sci. 192, 163 (1987). 55. Fenter. P., and Guslaffson. T . , Phvs. Rru. B 38, 10197 (1988). 56. Freyer. N., Kiskinova. M.. Pirug. G., and Bonzel, H . P., Appl. P h y s . A A39, 209 (1986). 57. Gritsch. T., Coulman. D.. Behm, R. J., and Ertl. G . . Plzys. Riw. Lett. 63, 1086 (1989). 58. Kneringer, G . , and Netzer, F . , Surf. Sci. 49, 125 (1975). 59. Pirug. G., Broden. G., and Bonzel, H . P., Proc. In!. Vuc,. Congr. 7th. Vicnnir p. 907 (1977); Cliem. P h y s . Lett. 73, 306 (1980). 60. Barteau, M. A.. KO, E. I., and Madix, R. J . , S ~ r fSci. : 102,99(1981); 104, 161 (1981). 61. Griffiths, K., Jackrnan, T . E., Davies, J. A,. and Norton, P. R.. Surf. Sci. 138, 113 (1984). 62. Norton, P. R., Griffiths, K . , and Bindner. P. E., Surf: Sci. 138, 125 (1984). 63. Wilf, M., and Dawson, P. T., Surf. Sci. 65, 399 (1977). 64. Ducros, R., and Merrill, R. P., Surf. Sci. 55, 227 (1976). 65. Sundaram, V. S . . and Dawson, P. H., S w f : Sci. 146, L593 (1984). 66. Freyer, N . , Kiskinova, M., Pirug, G.. and Bonzel. H . P.. S w f : Sci. 166,206 (1986). 67. Behm, R. J., Thiel, P. A., Norton, P. R., and Bindner. P. E.. Surf: Si,i. 147, 143 (1984). 68. Andrade, R. F. S., Dewel, G., and Borckmans, P., J . Chern. Phys. 91, 2675 (1989). 69. Krischer, K., Ph.D thesis, F U Berlin, in preparation.

OSCILLATORY CATALYTIC REACTIONS

275

Fair, J., and Madix, R. J . , J . Chern. Phys. 73, 3480 (1980). Eiswirth, M., Krischer, K . . and Ertl, G., Siri:f: Sci. 202, 565 (1988). Schuster, H. G . , “Deterministic Chaos.” Physik-Verlag, Weinheim, 1984. Wiggins, S., “Global Bifurcations and Chaos.” Springer-Verlag, Heidelberg, 1988. Takens, F., in “Dynamical Systems and Turbulence” (D. A. Rond and L. S. Young, eds.), p. 336, Springer-Verlag, Heidelberg, 1981. 75. Eckmann, J. P., Rev. Mod. Phys. 53, 643 (1981). 76. Grassberger, P., and Procaccia, I., P h y s . Rev. Lett. SO, 346 (1983). 77. Barkley, D., J . Chrm. Phys. 89, 5547 (1988). 78. Schneider, F., Annu. Rev. Phys. Chem. 36, 347 (1985). 79. Rehmus. P., and Ross, J . , in “Oscillations and Travelling Waves in Chemical Systems” ( R . Field and M. Burger, eds.), p. 288. Wiley. New York, 1985. 80. Kai, T.. and Tomita, K., Proy. 7heor. Phys. 61, 54 (1979). 81. Vance, W., and Ross, J . , J . Chern. Phys. 88, 5536 (1988). 82. Tsarouhas, G., and Ross, J . , J . Chem. Phys. 117, 6538 (1987). 83. Kevrekides, I. G., A r k , R.,and Schmidt, L. D., Chern. Eny. Sci. 41, 1549 (1986). 84. Dolnik, M., Schreiber, I . , and Marek, M., Physicrr 21D, 78 (1986). 85. Pugh, S . A., Schell, M., and Ross, J . , 1. Chem. P h y s . 85, 868 (1986). 86. Hudson, J. L., Lamba, P., and Mankin, J. C., J . Phvs. Chem. 9U, 3430 (1986). 87. Marek, M., in “Temporal Order” ( L . Rensingand N. I. Jaeger, eds.), p. 105. SpringerVerlag, Heidelberg, 1985. 88. Onken, H. U . , and Wicke, E., Bar. Bunsenges. Phvs. Chern. 90, 976 (1986). 89. Capsaskis, S. C., and Kenney. C. N., J . Phvs. Chem. 90, 4631 (1986). 90. Jaeger, N . I . , Plath, P. J., and Svensson, P., in “Spalial InhomogeneitiesandTransient Behavior in Chemical Kinetics,” (P. Gray, G. Nicolis, F. Baras, P. Borckmans, and S . K . Scott, eds.). p. 593. Manchester University Press, Manchester, England, 1990. 91. Eiswirth, M., and Ertl., G., Phvs. Rev.Let/. 60, 1526 (1988). 92. Eiswirth, M., and Ertl., G . , Appl. Phys. [Port] A A47, 91 (1988). 9.1. Eiswirth, M.. Moller, P., and Ertl, ti., Sitrf: Sci. 208, 13 (1989). 94. Freund, A . , Buchholz, F., and Schneider. F. W., Ber. Bitnsenges. Phys. Chem. 89, 657 (1985). 95. Rehmus, P., and Ross. J., J . Chem. Phys. 711, 3747 (1983). 96. Pfeuty, P., and Toulouse, G., “Introduction to the Renormalisation Groups and Critical Phenomena.” Wiley, New York. 1978. 97. Vance, W.. and Ross, J . , submitted for publication. 98. Kasai, H . , Yamamoto, T., and Okiji, A,, S w f : Sci. 220, L709 (1989). 99. Falta, J . . Imbihl, R., and Henzler, M.. Phys. Reu. Lett. 64, 1409 (1990). 100. lmbihl. R., Sander, M., and Ertl, G . , Sryf: Sci. 204, L701 (1988). 101. Vishnevskii, A. L.. and Savchenko, V. I . , Re[ict. Kinel. Curd. Lett. 38, 159 (1989). 102. Flytzani-Stephanopoulos, M., and Schmidt, L. D., Prog. Sucf Sci. 9, 83 (1979). 103. Kalwey, A. K., Gray, P., Griffiths, J. F., and Hasko, S. M., N u t w e (London) 313, 668 (1985). 104. Somojai, G . A,, and Van Hove, M. A . . Prog. S u ~ fk i . 30, 201 (1989). 105. Langmuir, I., Truns. Furuduy Soc. 17, 607 (1922). 106. Uchida, Y., Lehmpfuhl, G . , and Imbihl, R.. Surf. Sci.. in press. 107. Nicolis, G., and Prigogine, I., “Self-Organisation in Nonequilibrium Systems.” Wiley, New York, 1977. 108. Turing, A. M., Philos. Mug. 327, 37 (1952). 109. Haken, H., “Advanced Synergetics.” Springer-Verlag, Heidelberg, 1983. 110. lmbihl, R., Reynolds, A. E., and Kaletta, D., to be published. 70. 71. 72. 73. 74.

276

G . ERTL

1 1 1 . Zaikin, A. N., and Zhabotinsky, A. M., Nuturc (London) 225, 535 (1970).

Winfree. A. T., Science 175, 634 (1972). Vidal, C., and Hanusse, P., I n / . Rru. Phys. Clrem. 5, 1 (1986). Pismen, L. M., and Kharkats, Y . I . , Dokl. Akod. Nuuk S S S R 178, 901 (1968). Luss, D., Bailey, J. E., and Sharma, S . , Cliem. Eng. Sci. 27, 1555 (1972). Tsotsis, T. T., and Schmitz, R. A., C1wrn. E n g . Sci. 27, 1555 (1972). Kaul, D. J . , and Wolf, E. E., J . C a r d . 91, 216 (1985); 93, 321 (1985). Sant, R., and Wolf, E. E., J. Ccirril. 110, 249 ( 1988). Padberg, G., and Wicke, E., ChPrn. Eng. Sci. 22, 1035 (1967). Jaeger. N. I.. Moller, K.. and Plath, P. J., J.C.S. Furuday 82, 3315 (1986). Brown, J . R.. O’Netto, G. A., and Schmitz. K. A., i n “Temporal Order” ( L . Rensing and N. I . Jaeger. eds.), p. 86. Springer-Verlag, Heidelberg, 1985. 122. Bath, J. P., and Dauchot, J. P., J . Cutul. 115, 9 7 (1989). 123. Onken, H. U.. and Wicke, E., Z . Phys. Chern. 165, 23 (1989). 124. Schuth, F., and Wicke, E., Ber. Biinsenges. P h y s . Chem. 93, 191 (1989). 125. Tsai, P. K., Maple, M. B., and Herz, R. K., J . Cam/. 113, 453 (1988). 126. Schuth, F.. Song, X . , Schmidt, L. D., and Wicke, E., J. Chrrn. Phys. 92,745 (1990). 127. Imbihl, R . , Ladas, S. , and Ertl, G., Surf Sf.i. 215, L307 (1989). 128. Rotermund, H. H., Jakubith, S. , von Oertzen, A,, and Ertl. G., to be published. 129. (a) “Oscillations and Travelling Waves in Chemical Systems” (R. J. Field and H. Burger, eds.). Wiley, New York, 1984; (b) Ross, J . , Muller, S . C.. and Vidal, C.. Scic,nce 240, 460 (1988). 130. Luther, R., Z . Elckrroc&vn. 12, 596 (1906); see also Arnold, K., Showalter, K., and Tyson, J., J . Chern. Edrrc. 64, 740 (1987). 131. Rotermund, H . H., Ertl, G., and Sesselmann, W., Sirrf. S c i . 217, 383 (1989). 132. Bethge. H., and Klauna, M., Ulrrornicroscopy 11, 207 (1983). 133. Mundschau, M . , Bauer, E., and Swiech, W., Surf: Sci. 203, 412 (1988). 134. Kordesch, M. E., Engel, W., Lapeyre, G . J., Zeitler, E., and Bradshaw. A. M., A p p l . Phys. A49, 399 (1989). 135. Griffith, 0. H.. and Rempfer, G . F.. Adv. Opt. Electron Mic,rosc. 10, 269 (1987). 136. Rotermund, H. H.. Engel, W., Kordesch, M. E., and Ertl. G.. Notrtre (London) 343, 355 (1990). 137. Fink, T., Imbihl, R., and Ertl. G., J . Chem. Phvs. 91, 5002 (1989). 138. Rotermund, H. H., Jakubith. S.. von Oertzen, A,, and Ertl, G., J . Chem. Phys. 91, 4942 (1989). 139. Reusser, E. J., and Field, R. .I., J. A m . Clwm. SOC. 101, 1063 (1979). 140. Fink, T., Ph.D. Thesis, FU, Berlin, 1990. 141. Mundschau, M.. Kordesch, M. E., Kausenberger, B., Engel, W.. Bradshaw, A. M., and Zeitler, E., Surf. Sci. 227, 246 (1990). 142. Ortoleva, O., and Ross, J., J. Chrrn. Phvs. 60, 5090 (1974). 143. Lindstrom, T. H., and Tsotsis, T. T., Surf: Sci. 167, L194 (1986). 143u. Razon, L. F., Chang. S.-M., and Schmitz, R. A . , Chern. Eng. S c i . 41, 1561 (1986). 144. Sander, M., Ph.D. Thesis, TU, Berlin, in preparation. 145. Raval, R., Haq, S. , Harrison, M. A., Glyholder, G.. and King, D. A,. submitted for publication. 146. Goschnik, J.. Wolf, M., Grunze, M., Unertl, W. N., Block, J. H., and LobodaCackovic, J., Surf S c i . 178, 831 (1986). 147. Ehsasi, M.. Seidel, C., Ruppender, H., Drachsel, W., Block, J. H., and Christmann, K., Surf Sci. 210, L198 (1989). 148. Ladas, S., Imbihl, R.. and Ertl, G., Surf. Sci. 219, 88 (1989). 112. 113. 114. 115. 116. 117. 118. 119. 120. 121.

OSCILLATORY CATALYTIC REACTIONS 149. 150. 151. 152. 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167.

277

He, J. W., and Norton, P. R.,J . Chem. Phys. 89, I170 (1988). He, J. W., and Norton, P. R., Surf. Sci. 204, 26 (1988). Turner, J. E., Sales, B. C., and Maple. M. B., Surf. Sci. 103, 54 (1981). He, J. W., Mernmert, U., and Norton, P. R., J . Chem. Plzys. 90, 5088 (1989). Schiith, F., and Wicke, E., Ber. Bunsenges. Phys. Chem. 93, 491 (1989). Suhl, H., S14rf. Sci. 107, 88 (1981). Volokitin, E. P., Treskov, S. A,, and Yablonskii, G. S., Surf. Sci. 169, L321 (1986). Bassett, M. R., and Imbihl, R.,J. Chem. Phys., in press. Ehsasi, M., Frank, O., Block, J. H., and Christmann, K., Chem. Phys. Left. 165, 115 (1 990). Singh-Boparai, S . P., and King, D. A., Proc. I n t . Congr. Surf. Sci., 4rh, Cunnes p. 403 ( 1980). Adlhoch, W., and Lintz, H . G . , 2. Phys. Chem. 103, 207 (1976). Adlhoch, W., Lintz, H. G., and Weisher, T., Surf Sci. 103, 576 (1981). Schwartz, S. B., and Schmidt, L. D., Surf. Sci. 183, L269 (1987). Schwartz, S . B., and Schmidt, L. 13.. Surf. Sri. 206, 169 (1988). Dath, J.-P., Fink, T., Irnbihl, R., and Ertl, G., to be published. Banholzer, W. F., and Masel, R. I . , Surf. Sci. 137, 339 (1984). Lesley, M. W., and Schmidt, L. D., Su$ Qi. 155, 215 (1985). Fink, T., Dath, J.-P., Bassett, M. R., Irnbihl. R., and Ertl, G., Vucuum in press. Wicke, E., and Onken, H. U., in “From Chemical to Biological Organisation” (M. Markus. S. C., Miiller, and G. Nicolis, eds.). p.68. Springer-Verlag. Heidelberg, 1988.

This Page Intentionally Left Blank

ADVANCtS IN CATALYSIS. VOLUME 37

Role of Sulfur in Catalytic Hydrogenatio n Reactions J . BARBIER, E. LAMY-PITARA, AND P. MARECOT Uniuersiic: (ic Poitiers Luhorutoirr dc, Crrfulysc, en Chiinia Chgriniyue CNRS U . R . A . 350 86022 Poitirrs. Frtrnce AND

J . P. BOITIAUX, J. COSYNS, AND F. VERNA 1n.sfitrtt Frcrnsuis du P<’trolr 92506 Ruril Mtrlrnaison, A-cincr

1.

Introduction

One of the most severe causes of metallic catalyst poisoning is the adsorption of species containing sulfur; sulfur-containing compounds are present in natural sources of hydrocarbons and therefore will necessarily be found in industrial feedstocks. The poisoning of metallic catalysts by sulfur has been extensively studied, essentially on nickel, palladium, and platinum, for numerous reactions and consequently under very different experimental conditions, particularly for temperatures ranging from 300 to 1300 K. Experiments carried out at high temperatures (1-4) have shown that poisoning can be balanced by the following reaction:

Such desorptions, depending on the metal-sulfur binding energy and the thermodynamic properties of the surface-adlayer interface, play an important role in controlling the sulfur coverage ( 5 ) . The distribution of adsorbed sulfur on the metallic surface and the resulting effect on activity and selectivity of the metal depends on this coverage (6). At low coverage, for which interactions between adjacent adatoms are negligible, sulfur 279 Copyi-ighi Q 1990 by Academic Press. Inc. All rights of reproduction in any form reserved.

280

J . B A R B I E R et al.

atoms adsorbed on a heterogeneous surface are located o n sites that have the highest binding energies. At higher coverages, sulfur atoms could be gathered into islands (7). Catalytic reactions performed at low temperatures (below 400 K) show that adsorption of sulfur-containing compounds is then irreversible, the equilibrium [Eq. ( l ) ] being wholly displaced toward the left. As an example of low-temperature catalytic reactions, hydrogenation of unsaturated hydrocarbons is the most important industrial application. Chemical industrial needs are mainly for unsaturated hydrocarbons, which have reactivities that enable polymer or petrochemical product development. All the processes developed for the production of olefins, diolefins, and aromatics give a mixture of unsaturated hydrocarbons, which are not valuable as such; further hydrogenations are necessary to obtain usable products for refining and chemical industry. Sulfur is generally considered to be a poison of hydrogenation catalysts. But in the case of hydrodehydrogenation reactions, this compound can also be used as a modifier of selectivity or even, in some cases, as an activator. To increase catalytic activity for hydrogenation reactions in the presence of sulfur compounds, i.e., to enhance sulfur resistance of metallic catalysts, extensive research has been undertaken on sulfur poisoning. Before discussing the effect of sulfur on catalytic activity and the selectivity of metal during hydrogenation reactions, it is necessary to have a clear idea of the thermodynamics, the adsorption, and the arrangement on the surface. Therefore, we shall first deal with the adsorption of sulfur on metallic catalysts, then with the effect such poisoning has on the adsorption of reactants. Finally, the last two paragraphs will be devoted to the modifications induced by sulfur adsorption on activity and selectivity of metallic catalysts. Overall, this article develops four different areas of interest: adsorption of sulfur on metallic catalysts, effects of preadsorbed sulfur on the reactants’ adsorption, influence of sulfur on activity, and influence of sulfur on selectivity. II. Sulfur Adsorption on Metals

A. THERMODYNAMICS OF SULFUR 1.

ADSORPTION

AjJnity of Sirlfiur f o r Met&

Adsorption of sulfur on monocrystalline, polycrystalline, and supported metallic catalysts has been extensively studied. For all adsorption reactions, thermodynamics states that AH,,, = Etl - Ed (where AH,,, is the enthalpy of adsorption, E;, is the activation energy of adsorption, and E,,

S U L F U R IN CATALYTIC HYDROGENATION REACTIONS

28 1

is the activation energy of desorption). In the case of metallic catalysts, the activation energy of sulfur adsorption is negligible. As a result, binding energy can be evaluated from desorption or adsorption studies. Such isotherms allow evaluation, using the Van't Hoff equation, of the chemisorption energy for a given sulfur coverage. Such techniques were used by Benard and Oudar and co-workers (8) on polycrystalline Cu catalysts ( 9 ) , on monocrystals of Ag (/O),and on other metals ( I I ) . On the low-index Ag monocrystals, the bonding energy ofchemisorbed sulfur follows the sequence ( 1 10) > (100) > ( I I I ) , showing that the sulfur affinity changes with respect to the crystallographic orientation under consideration and therefore on the basis of the coordination number. On nickel catalysts, which have been extensively studied (12-/4), McCarty and Wise (14) showed that in a temperature range from 373 to 873 K , sulfur coverage of nearly half a monolayer can be reached with H,S partial pressures as low as 1-10 ppb. Such results, indicating that the equilibrium H2S H,(g) + S(a) is totally displaced toward the right side, show the very high affinity of sulfur for transition metals and thus the difficulties in avoiding sulfur poisoning in metallic catalysis. Moreover, for coverage close to I , a sudden decrease of the adsorption enthalpy (Fig. I ) can be explained by adsorption of species such as HS or undissociated H,S. A study of the nickel-sulfur interactions shows that the adsorbed state is energetically more stable than the bulky Ni,S2 sulfide (14). The same result was found for Ir catalysts (15). This shows that the contact of a metal with H2S will lead to a widely covered surface without any sulfur dissolution in the metal. The chemisorption energies of sulfur were also defined on Pt (/6), Ir (15), Ru (f7),and Fe and Co (18). For example, in the case of Pt, which is known as more resistant than Ni to sulfur poisoning, sulfur is weakly chemisorbed ( f 6 ) . Nevertheless, a comparison of the enthalpy of sulfur chemisorption on different metals is made difficult by the decrease of the binding energy with increasing sulfur coverage. On platinum, McCarty et ul. (16) showed that - A H " decreases from 122 to 26 kJ mol-' when sulfur coverage is increased from 0. I to 0.6. Consequently, acomparison of sulfur chemisorption energy on different metal surfaces has to be made at the same coverage. It was suggested by Alstrup et uf. (19) that the heat of adsorption varies linearly with coverage according to the Temkin law, but Wise et af. (20) proposed a modified Temkin equation taking into account a linear variation of the enthalpy and entropy of sulfur chemisorption. PH>F ~

P H7

=

8 exp

AH"(1 RT(I

+ SO) + Be)

-

AS"(I + A @ ) R

282

J . BARBIER

-200

c

E" .

-150

Ul.

I __~

f

-I Y

v

I-

4 I -loo

L't

~

Z

0

F

a

[I -50-

0

a

0 010203040506070809 1 1 1 12131415161718

SULFUR COVERAGE FIG. I . Adsorption heat of sulfur o n nickel versus coverage.

It is claimed that the data are better described by such a modified equation. It appears that, for all metals studied, the binding strengths of sulfur decrease with increasing sulfur coverage. Such an evolution can be easily explained, but with regard to the effect of coverage on the entropy variation and therefore on the mobility of adsorbed sulfur, results are more complex and show that increasing coverage can act in an opposite way on the mobility of adsorbed sulfur for Ir, Pt, and Ru (20).

2.

Interaction of Metul-Suljiu

The upper plateau of the isotherm of sulfur chemisorption allows definition of a surface saturation state of sulfur for each metal. Such a state can be reached in a large range of temperatures and partial pressures of hydrogen sulfide. Using %, Oudar ( 2 1 ) listed the values obtained on different metals for a maximum concentration of sulfur before the appearance of solid sulfide. On the (100) faces of nickel and platinum, this saturation state corresponds to one sulfur atom for two accessible metallic atoms. On the (1 11) faces, it is slightly lower than one sulfur atom for two metal atoms. On the ( 1 10) faces, it is, respectively, equal to 0.71 on nickel

S U L F U R IN CATALYTIC HYDROGENATION REACTIONS

12

283

i

n 0

180°C

:

I

I

I

I

1

1

5

10

15

20

25

30

500°C

35

TIME (hours)

FIG.2. Desorption of H2S under hydrogen at 773 K . 0 , Pt/AI,O,; H, A1203(1% CI); dashed line, initial amount of sulfur on Pt/AI2OI;dotted line, initial amount of sulfur on AlzOl (1% CI).

and to 0.81 on platinum. In conclusion, there is some evidence that sulfur preferentially adsorbs on sites of lowest coordination, such as corner and edge sites, on stepped single-crystal surfaces, and, by implication, on equivalent sites of small metal crystallite surfaces. The surface saturation by sulfur has to be compared to the “irreversible” adsorbed sulfur introduced by Menon and Prasad (22)and Apesteguia et al. (23). The study of H,S adsorption on supported catalysts was carried out by Menon and Prasad (22)and Apesteguia et al., Parera et al., and Barbier et a/. and Marecot (23-25). For alumina supports, it was shown (23-25) that chlorine inhibits the adsorption of H,S on the support. Yet this adsorption on pure alumina is wholly reversible at 500”C, as is shown in Fig. 2. On Pt/AI,O, at 500°C. only a fraction of the adsorbed sulfur is quickly desorbed in a hydrogen atmosphere. This result enabled the preceding authors (22-25) to develop the notion of “reversible” and “irreversible” adsorbed sulfur. The irreversible form, which does not exist on pure alumina, would interact with the metal. The quantity of irreversible sulfur, determined after 30 h of desorption under hydrogen flow at 500”C, does not depend on the sulfiding conditions (Table I). After 30 h of treatment at 500°C in hydrogen flow, the sulfurized Pt/

J . BARBIER el (11.

TABLE 1 Tofu1 Sulfur cind “Irrcwersihle Sulfur o n Cl~lorinutedAI103 und PrIAI,O,” ”

g S/g catalyst x 10’

Al?Oi, 15% CI

H?S in hydrogen (%)

Total sulfur

8.0 3.5 1.5 0.5 0.02

0.120 0.091 0.085 0.059 0.029 ~~

‘‘ “Irreversible”

Pt/A120,

“Irreversible” sulfur

0.151

0.121 0.085 0.055

-

0.003 ~

“Irreversible” sulfur

0.172

0.004 0.006 0.004

~

Total sulfur

0.026 0.029 0.028 0.025 0.028

~~~~~

sulfur determined after a 30-h desorption under H I at 773 K .

A1,0, catalyst has a sulfur coverage of almost 0.4 atom of sulfur per accessible metal atom. This notion of irreversible sulfur was extended to other metals and, particularly, to bimetallic catalysts ( 2 6 ) . Whatever the nature of the metal, the irreversible sulfur coverage varies in the range of 0.4-1 sulfur atom by accessible metallic atom. Thus the metal-sulfur bond has such a strength at these coverages that the lifetime of adsorbed sulfur on the catalyst reaches a high value, and, even in a flow without any poison, no regeneration occurs after a reasonable lapse of time. According to Schultze and Koppitz (271, when a species S’ comes in close contact with a metal M, a charge transfer of electrons is possible (depending on the chemical nature of the adsorbent and of the adsorbate): M + SL M - S L + + r - . This charge transfer can be only partial, resulting in a partial adsorption valency, which depends on the difference of electronegativities between the metal substrate and the adsorbate ( 2 7 ) . Kornyshev and Schmickler claimed (28)that the most important parameters that determine the charge of the adsorbate are the atom ionization energies, the work function of the metal, and the electronic affinities. This last parameter is taken into account by Barbier rt ul. (25) to explain the adsorption behavior of sulfur on noble metals. Following this explanation, the smaller the difference of electronic affinities between sulfur and metal, the more covalent the metal-sulfur bond. On the contrary, the bond could be polarized for large differences in electronic affinities. For the platinum-sulfur system, the electronic affinities are, respectively, equal to 2.12 and 2.08 eV ( 2 9 ) .Thus the Pt-S bond is essentially a covalent one, whereas for Ir the electronic affinities are more different (2.08 eV for sulfur and I .6 e V for Ir). In such conditions, --j

S U L F U R I N CATALYTIC HYDROGENATION REACTIONS

285

sulfur on iridium is adsorbed at a negative oxidation state. Therefore chemisorption of sulfur on metallic catalysts will depend on the electronic properties of the metal. Adsorption of an electron acceptor-like sulfur will be enhanced on metals of low electronic affinities (30). It was shown by Marecot rt al. (3f)on a series of supported metallic catalysts with comparable dispersions that the “irreversible sulfur” coverage increases from Au to Re when electronic affinities decrease in the same sequence. On the other hand, for supported catalysts, the electronic state of metallic particles can be modified not only by the acidic properties of the support but also by the dispersion of the metal. Cini (32),by a theoretical calculation on isolated metallic clusters, shows that a decreasing particle size from 20 to 10 A of Ag clusters induces a decreasing electronic affinity of almost 1 eV. Vedrine et af. (33) and Foger and Anderson (34), using XPS data, pointed out that small platinum crystallites deposited on very acidic supports are electron deficient. As a matter of fact, the two effects (dispersion and support) could act in an opposite way. So the electronic density of the metal for supported catalysts, and, of course, the affinity of the catalysts for sulfur, will be the result of these two phenomena. From a practical point of view, when platinum is deposited on an acidic support (such as a reforming alumina), the support effect plays a leading role and small platinum particles are more resistant to sulfur adsorption than are bulky catalysts (25).It is worth noting that such a result is at variance with the conclusion previously described on the effect of coordination on singlecrystal surfaces. In this case we concluded that, without support effect, sulfur is preferentially adsorbed on platinum at sites of lowest coordination, such as ledge or edge sites, which are, of course, more numerous o n small metallic particles. On tile other hand, when iridium is deposited on a low-acidic support, the size effect prevails over the support effect, because the amount of sulfur adsorbed on small metallic crystallites is more important than it is on the large ones (25). Finally, when platinum is deposited on the same low-acidic support, by a compensation between size and support effect, the sulfur coverage will be almost the same, no matter what platinum dispersion there may be (25, 35).

B. CHEMICAL STATEOF ADSORBED SULFUR Generally speaking, the toxic effect of a molecule on a catalyst is related to its ability to create a strong chemisorption bond with the catalyst surface. This implies that it must form an actual chemical bond between the adsorbate and the adsorbent. Then a sufficient condition for a compound to

286

J . BARRIER t‘t 121.

present some toxicity is that it possesses a doublet of free electrons in its valence layer (36,37).Thus Maxted (36)provided the toxic effect of sulfurcontaining compounds with two doublets of free electrons on sulfur or with a free doublet on sulfur. The toxic effect of this group of compounds disappears when, through chemical treatment, usually oxidation, the free electrons, which allow bonding with the metal, are themselves involved in the formation of a stable compound (shielded structures). The examples of detoxication reported by Maxted and Marsden are numerous (38-40). As an example, the metallic activity of thiophenepoisoned catalysts can be regenerated by oxidation of thiophene to sulfone through hydrogen peroxide treatment. Conversely, a reducing treatment can turn a nontoxic compound such as sulfate ion into a toxic compound, e.g., hydrogen sulfide (41). Therefore, at high temperatures, under hydrogen flow, sulfur-containing compounds will be reduced and (or) hydrogenolyzed to yield hydrogen sulfide, and thus, at such temperatures, the only sulfiding agent will be hydrogen sulfide; the initial state of the poison will thus no longer have any effect on the deactivation of the catalyst. On the other hand, at low temperatures, adsorption and toxicity properties of sulfur compounds will be largely modified by the nature of the molecule under consideration (36). I.

Hydrogen Suljide

The chemisorption of sulfur from mixtures of H,S and H, has been widely studied; we have discussed some of the results. Nevertheless, introduction of “irreversible” and “reversible” adsorbed sulfur, which is in line with adsorption stoichiometries varying from more than I to 0.4 sulfur atom by accessible platinum atom, shows that different adsorbed species are involved in sulfur chemisorption. In fact, electrooxidation of adsorbed sulfur o n platinum catalysts occurs at two different electrochemical potentials (42); in the same way, two different species of adsorbed sulfur were identified on gold by electrochemical techniques and XPS measurements ( 4 3 , 4 4 ) .By use of 35S(45)it was pointed out that, according to the experimental conditions, reducible PtS, or nonreducible PtS monolayers can be created. Adsorption of sulfur introduced as sulfide ions was studied by electrochemical techniques. Such sulfide species are oxidized during their adsorption and the mean valency of the adsorbed sulfur species, X,,varies with the coverage 8,and also with the coefficient of roughness, r ( 4 6 , 4 7 ) .At the same time, the mean stoichiometry of adsorption varies from 2 to 3 sulfur atoms by accessible platinum atom at low coverage, to 1 sulfur atom at high coverage. At high rates of coverage, the number of electrons

SULFUR I N CATALYTIC HYDROGENATION REACTIONS

287

required to induce electrooxidation of adsorbed sulfur to sulfate increases sharply (Fig. 3 ) (48).I n the same way the relative reflectivity of the surface decreases sharply, in the same range of sulfur coverages (Fig. 3 ) . Such results are ascribed to the formation of multilayers of polysulfides at coverages as high as 0.7. Finally, ultraviolet-visible reflectance spectroscopy of sulfur adsorbed on platinum catalysts at different coverages shows that adsorbed species are made of neutral sulfur and polysulfides [twodimensional (at low coverage) and three-dimensional (at high coverage)]. Koestner et al. (49) noted the presence of HS. and H2S at low temperatures on (1 1 I ) monocrystalline platinum. Such species were isolated on various other metals (50-53). Ultraviolet photoelectron spectroscopy (UPS) was used to define the change in the electron density of states of the valence band associated with sulfur adsorption. The results indicate that the 3d and 4s electrons of nickel are involved in the chemisorption band with the 3p electrons of sulfur (37). On the other hand, the change in the work function resulting from the chemisorption of sulfur can be used to calculate the charge transfer between the metal atoms and the adsorbed sulfur. On nickel (loo), during adsorption of a half monolayer, charge transfer is less than 0. I electron, showing that bonding of sulfur to nickel is essentially covalent. Such a transfer varies with the nature of the metal, with the adlayer coverage, and with the single-crystal metal surfaces under consideration. Nevertheless, measurements on Ni ( 5 4 , Cu (21,55), Pt (56), and Ru (57) confirm the covalent nature of the S-metal bond. It is obvious that such charge transfers, evaluated by spectroscopic techniques, or valencies of adsorbed sulfur calculated from electrochemical techniques, are mean values corresponding to various adsorbed species simultaneously covering the catalyst.

2.

Other SuUur-Contuining Compounds

Previously we mentioned studies of adsorption of sulfur introduced as H,S and the thermodynamics of the sulfur-metal bond. But a great variety of sulfur compounds can be present in the feeds of industrial hydrogenations. The variation of the poisoning effect as a function of the sulfur compound type has been recognized for a long time, but is not yet fully understood. It is thus of the greatest interest to study the adsorption of different types of sulfur compounds. With regard to high-molecular-weight sulfur compounds, Maxted (36) demonstrated that at low temperatures the specific toxicity of sulfur increases with the size of the molecule. The molecule is adsorbed by means of sulfur atom anchorage, around which the free rotation of the carbon chain inhibits the adsorption of reactants on all of the adjacent surface.

J . BARBIER el

0

0.1

0.2

0.3

0.5

0.4

d.

0.6

0.8

0.7

0.9

1

1.1

Ns/Npt 60

50

40

4

2 30

z

20

10

0

0

- ;

0,l

II

0,2

0,3

I

1

0,4

0.5

I

0,6

1

0,7

I

I

0.8

0,9

1

Ns/Npt FIG.3. ( A ) Variation of the relative reflectivity of a platinum surface as afunction of the coverage degree by adsorbed sulfur species. ( B ) Number ofelectrons exchanged per platinum site occupied by sulfur (N,,,) as a function of the degree of coverage.

SULFUR LN CATALYTIC HYDROGENATION REACTIONS

289

It was pointed out that adsorption of propanethiol or hexanethiol occurs according to the following reaction (58): RSH -+ Mad,+ H + + e - ; desorption can occur by an electrooxidation, RS,,, + 3 H 2 0-+ RSO,H + 5H+ + 5 e - . A comparative study of the amount of thiol adsorbed and the effect of this adsorption on hydrogen chemisorption showed that one molecule of thiol is adsorbed on one accessible platinum atom (58). In the early 1960s, Bourne rt al. (59) studied the gaseous-phase adsorption of thiophene, thiocycloalkanes, alkylsulfides, thiols, and disulfides o n dispersed nickel catalysts over an inert mineral silicate (sepiolite). They observed that, under hydrogen, every sulfur compound undergoes some dissociation on the metal, giving mainly the corresponding alkane and sulfur, which combines with the metal. But they found significant differences in the behavior of the various species. They classified the sulfur compounds in two groups: the first group includes thiophene, thiocycloalkanes, and di-n-butylsulfide. For this group they found that the extent of sulfiding is limited to the surface atoms, and the maximum value is one sulfur atom for every four nickel atoms. By contrast thiols and disulfides progressively sulfide the nickel, which results in the formation of a bulk sulfide, probably Ni,S,. More recently, De Plaen (60) studied the adsorption of thiophene in liquid phase on various metals of Group V I l l (Ni, Rh, and Pt) under hydrogen pressure. It was observed that butane and sulfur are the only products of the dissociative adsorption, which can be written as follows: XM

+ yC,H,S +

3yH?+ M,S,.

+ JC,H~"

Here also, the metal sulfiding is limited to the metal surface atoms and the maximum stoichiometry of the sulfur uptake was found to be one sulfur atom for every two metal atoms. The adsorption of thiophene on supported palladium has been studied (61, 6 2 ) . The studies were performed under conditions used for industrial hydrogenations: liquid phase, low temperature, and hydrogen pressure. The carrier is a special inert alumina with large pores (greater than 10 nm) and a surface area of less than 100 m'/g. After thiophene addition, a rapid formation of butane was observed, followed by a slower one, which finally reached a plateau (Fig. 4). Hydrogen pressure has a positive influence on the maximum value of butane obtained. Neither butadiene nor butenes were detected. Moreover, no butanethiol or H,S was found in the hydrocarbon mixture: the only sulfur compound in the mixture after the butane evolution had stopped was thiophene. Several catalysts with various degrees of palladium loading and metal dispersions were submitted to thiophene adsorption. The dissociatively

290

J . BARBIER

et ul.

2

w c -

1

0

20

40

60

I 80

1

TIME (HOURS) FIG.4. Dissociative adsorption of thiophene on a Pd/AI?OIcatalyst in heptane. ( I ) P I bar, T = 20°C; (2) P = 20 bars, T = 3S"C.

=

adsorbed sulfur (measured at the plateau of butane formation) and the total sulfur adsorbed are plotted in Fig. 4 as a function of the superficial palladium. It can be observed that the amounts of adsorbed sulfur (total and dissociative) are directly proportional to the quantity of superficial palladium atoms, corresponding to a ratio of SlPd equal to 0.5. Therefore, it appears clearly from all the observations that sulfur adsorption from thiophene onto palladium is restricted to the surface area of metal particles, with part of this surface being sulfided through the decomposition of the sulfur compound. Figure 5 presents the formation of butane obtained o n dispersed palladium catalysts from the adsorption of several sulfur compounds: thiophene, thiophane, dibutylsulfide, butanethiol, dibutyldisulfide. Two groups can be observed: (1) thiols and disulfides that undergo a rather low-level dissociation in butane and (2) thiophene, thiophane, and dibutylsulfide, which are dissociated to a much larger extent. The dissociation of dibutylsulfide produces a quantity of butane twice that produced from thiophene and thiophane, indicating that the two bonds (sulfur - butyl radical) are broken altogether. From observation of the first parts of the butane formation curves (initial time), it appears that the dissociation rate of thiophene is the highest one.

SULFUR IN CATALYTIC HYDROGENATION REACTIONS

0

m

DIBUTYLSULFIDE

h

0

1 10

n

0 U

a w

z

5 3 m

29 1

/

k

:p.

THIOPHENE

THIOPHANE

BUTANETHIOL

El

I

m

-,

0

500

-

8

.

I

DIBUTYLDISULFIDE

I 1000

1500

I 2000

TIME (minutes) FIG.5 . Formation of butane as a function of time, during the dissociative chemisorption of different sulfur compounds on Pd/AI2O3( T = 20"C, P = 1 bar, solvent = heptane).

Comparison with dissociation rates of thiophane and butanethiol, which are significantly lower, gives interesting insights into the dissociation mechanism of thiophene. It is clear that thiophene dissociation does not occur by intermediate hydrogenation to thiophane followed by dissociation to the corresponding thiol. The most probable mechanism involves immediate breaking of the two C-S bonds; the unsaturated hydrocarbons produced remain strongly chemisorbed and are only desorbed through hydrogenation to butane. This mechanism can be presented as follows (asterisks denote adsorbed forms):

m

Figure 5 shows that the dissociative chemisorption of butanethiol is much lower than that of thiophene. But these results were obtained at a

292

J . BARBIER C f U / .

I

- - _ _ - -_ - - -

G-

O v

-

3

'/2

15

X

0 E

v

0

- 1

0

I

I

I

I

20

40

60

80

I

100

120

140

TIME (hours) Piti. 6 . Formation of butane as a function of time, during the dissociative chemisorption of burylmercaptan on Pd/A1201.( I ) P = 1 bar. 7 = 20°C; (2) P = 10 bars, 7 = 35°C: ( 3 ) P = 25 bars. T = 50°C.

low temperature (20°C) and at atmospheric pressure. Figure 6 presents the dissociation of butanethiol to butane at various temperatures and pressures. We observe a large increase oft he dissociation rate: moreover, the maximum value of sulfur coverage (from dissociation) tends to reach 0.5, as with thiophene. Therefore, the value of 0.5 for S/Pd seems to be the limit of surface sulfiding for palladium in the range of studied conditions. Verna (62) studied the dissociative adsorption of thiophene o n platinum, palladium, and rhodium dispersed on alumina. Figure 7 and Table I1 present the dissociative chemisorption of thiophene to butane o n the three metals. The sulfur coverage of platinum is very low compared to the other metals. The sulfur coverage on palladium is about 2.5 times higher than on platinum; such a value is similar to the one found by Mathieu and Primet (63). As we saw previously, the adsorption of sulfur on metallic surfaces has been extensively studied in the gas phase with mixtures of H,/H,S. On the contrary, adsorption of sulfur compounds in hydrocarbon mixtures and especially unsaturated hydrocarbons has seldom been studied. Never-

SULFUR IN CATALYTIC HYDROGENATlON REACTIONS

293

a a

RHODIUM a

A

PALLADIUM

-

m

rn

rn

rn

PLATINUM

Y

0

500

1.500

TIMgTmin)

2 I00

FIG.7. Formation of butane as a function of time, during the dissociative chemisorption of thiophene on different metals.

theless, these conditions are the only realistic means to study the effect of sulfur on hydrogenation and dehydrogenation reactions. Boitiaux et al. and Verna (61,62)studied the dissociative chemisorption of thiophene in the presence of isoprene and of the corresponding methylbutenes o n palladium. Figure 8 records the results of a typical experiment carried out with a mixture of thiophene and 3 mol% isoprene in heptane. The evolution of isoprene, isopentenes, and isopentane is shown as a

Sirlfirr Couercige

Catalyst Pt Pd Rh

TABLE 11 Deducted from Butune Fnrmcrtion

Superficial metal (atg/g catalyst)

Sulfur from butane (atg S/g catalyst)

S/metal

2.09 x 3.55 x 10-5 1.49 x

1.9 x 10-6 9.4 x 1 0 - 6 8.0 x

0. I 0.25 0.5

294

J . BARBIER

et ul.

PARAFFIN

BUTANE A S/Pds (MOL.1

___--la-‘-114

P= 10 Kg/cmz

5.10-5-

- 1/11

T= 35°C

+

TIME IN MINUTES

-1-

TIME IN HOURS

Fie. 8 . Dissociative adsorption of thiophene on Pd/AI?O, as a function of the nature of hydrocarbon on Pd/AI?03( P = 10 kg/crn’. 7 = 35°C).

function of time. The formation of butane from thiophene is also given. I t appears that initially butane is not formed, as long as isoprene remains available. When isoprene is almost completely hydrogenated, a rapid formation of butane is observed up to a value of 1 atom of sulfur for every 8 superficial palladium atoms. This value remains constant as long as olefins (isopentenes) are present in the mixture. When the olefins have nearly disappeared, butane formation increases again, to reach a ratio of S/Pd equal to 0.25. Thus clearly, the sulfided state of the palladium surface depends deeply on the nature of hydrocarbon in competition of adsorption. In conclusion, on a working catalyst (in hydro-dehydro reactions), the state of a sulfided surface depends on ( I ) the nature of the sulfur compounds, ( 2 ) the nature of the hydrocarbons, and (3) the nature of the metallic phase.

111.

Sulfur Effect on Adsorption of the Reactants

In heterogeneous catalysis, chemisorption of reagents is a preliminary step of the surface reaction. Thus every discussion about catalytic modification (activity, selectivity, and lifetime of the catalyst), induced by poison deposition, has to be carried out in parallel with a study~ofthe poison

S U L F U R IN CATALYTIC HYDROGENATION REACTIONS

295

effect on the reactant adsorption. By its irreversible adsorption, sulfur is very suitable for coadsorption studies. CO and HZadsorptions on sulfurized metallic catalysts have therefore been extensively studied. Bonze1 and Ku were the first to show that on Pt( I 101, adsorbed sulfur considerably weakens the CO-metal bond (7, 64). By thermodesorption of adsorbed CO on platinum that is partially deactivated by sulfur, they showed that the higher the sulfur coverage, the lower the CO desorption temperature. Such results can be explained by assuming that on a heterogeneous metallic surface area, sulfur is preferentially adsorbed o n sites that have the strongest affinity for CO. According to Glowski and Madix (65), sulfur chemisorption selectively blocks the adsorption of CO at the edges because of its preference for edge sites. Otherwise, the decrease of the CO binding energy as a function of sulfur coverage can result from a variation of the electronic density of the partially poisoned metal or from lateral interactions between adsorbed CO and sulfur species. On Pt( 1 1 I), preadsorbed sulfur decreases the initial adsorption binding energy of CO. Thus at zero coverage, the extrapolated value of adsorption energy of CO is equal to 19 kcal/mol on Pt( I I I ) with a sulfur ( 2 x 2 ) overlayer; on clean Pt( I I I ) the same energy is almost 32 kcal/mol (66). At such low coverages long-range electronic effects have to be considered. On the other hand, as the CO concentration increases, its binding energy decreases rapidly for the sulfurized Pt( I I I ) surface, whereas for the clean catalyst, in the same coverage range, the energy is almost constant. Such a large decrease was explained by a repulsive interaction between CO molecules and adsorbed sulfur (66). In the same way, preadsorbed sulfur is able to decrease the binding energy of chemisorbed hydrogen on various metals. On supported metallic catalysts the strength of the hydrogen-metal bond can be evaluated by thermodesorption of molecular hydrogen (67).The shift of the desorption peaks can therefore be used to define the effect of sulfur preadsorption on hydrogen chemisorption. Table Ill shows that sulfurization of the metal results in ( I ) a decrease of the hydrogen adsorption capacity and ( 2 ) a decrease of the hydrogen binding energy (67). In the liquid phase, the adsorption of hydrogen can be studied by an electrochemical technique, cyclic voltammetry. At various sulfur coverages, introduced as H,S, the amount of adsorbed hydrogen can be calculated by integration of its electrooxidation peaks following the reaction Had\ e H + + P (68). Finally, by using the Nernst equation, which correlates the electrochemical potential with hydrogen pressures, isotherms of hydrogen chemisorption at different sulfur coverages can be brought out. Such isotherms consist of two parts in relationship with strongly (low hydrogen coverage) and weakly (high hydrogen coverage)

296

J . BARBIER

et al.

TABLE 111 Effect of Sidfiir on ihr Hydrogen Adsorption Cupucity und t h e Hydropen Binding Energy of Plutinicm- und Iridiurn-Supported Cutulysts Atoms Hlg catalyst x 10-l' Catalyst 0 . 6 PtlSiO? 0.6 Pt/AI,O, 0 . 6 lr/SiOz 0 . 6 Ir/AlzO,

AHads(kJ mol-')

Fresh catalyst

Sulfurized catalyst

Fresh catalyst

Sulfurized catalyst

14.24 16.93 12.60 24.88

I .72 3.33 I .72 3.00

44.7 38.5 61.9 47.6

32.2 32.6 43.9 27.6

adsorbed hydrogen. Each of these parts complies with the Frumkin law and allows evaluation of the free energy of hydrogen chemisorption. At low hydrogen coverage (hydrogen strongly adsorbed), the affinity of hydrogen for partially sulfurized platinum is almost constant with an increasing sulfur coverage. On the other hand, sulfur is able to decrease the free energy of weakly adsorbed hydrogen from - 14 kJ/mol for OS = 0 to -0.35 kJ/mol for 8, = 0.76 (Table IV) (69). By the same technique, the chemisorption of olefinic compounds can be studied. The catalyst coverage by such compounds can be evaluated by the displacement of adsorbed hydrogen and allows drawing the isotherms of chemisorption. Figure 9 gives the evolution of the adsorption free energies of maleic acid and dimethyl maleic acid as a function of sulfur coverage. Sulfur induces an increasing adsorption energy of dimethyl maleic acid when the affinity of partially poisoned platinum surfaces for maleic acid is almost constant. This difference can be explained by the

TABLE 1V Free Energy of' Adsorption qf Hydropen A d s o r b e d on Puriiully Sulfurized Pluiinitm

Sulfur coverage

Strongly adsorbed hydrogen (-AC"/kJ mol-')

Weakly adsorbed hydrogen (-AC"/kJ mol-I)

0 0.09 0. I6 0.35 0.76

37 31 30 40 38

14 12 I 7 0.35

S U L F U R 1N CATALYTIC HYDROGENATION REACTIONS

:o

20

297

I 0

0.1

0.2

0.3

0.4

0.5

SULFUR COVERAGE FIG.9. Variation of the free energy of adsorption of maleic as a function of sulfur coverage.

(m) and dimaleic ( 0 )acids

electronic densities calculated on each olefinic carbon atom, which is equal to 0.05~.and 0.02r, respectively, for dimethyl maleic and maleic acids (70). Thus sulfur acts as an electron acceptor compound and decreases, through its adsorption, the electronic density of the unpoisoned metallic surface area and promotes the adsorption of olefinic compounds of high electronic densities. Such electronic transfer induced by sulfur adsorption was also pointed out by using cinnamic acid as a probe molecule (48). The UV-visible reflexion spectra of adsorbed cinnamic acid on nonpoisoned and partly poisoned platinum catalysts shows that adsorption on pure platinum induces a shift of the peaks toward the higher wavelengths and an appearance of fine structure. Sulfurization of platinum induces a further enhancement of higher wavelength peaks. Binding energy of cinnamic acid is thus increased by sulfur adsorption on Pt catalysts. This increasing adsorption energy of olefinic compounds by sulfurization of metals can induce a partial desorption of sulfur or else a surface reaction giving thiols, as was pointed out by Oudar e t ml. (71) in the case of butadiene chemisorption at I00"C on Pt(l10) and Pt( I I I ) faces. As another example, the possibility for sulfur to modify the interactions

298

J . BARBIER

n = 0.1

et a /.

.

9

0

9

9

>

0

3

-2

-4 I

I

-3

-4 LOG Pc

FIG.10. Determination of the partial order with respect to cyclopentane; O.., Pt/AI20,; A , sulfurized Pt/AI20,.

of a metallic catalyst with other adsorbates was shown in the kinetic study of cyclopentane hydrogenolysis (72). Figures I0 and I I show the influence of the partial pressure of hydrocarbons and hydrogen on the rate of hydrogenolysis of cyclopentane carried out on Pt/AI20, and sulfurized Pt/AI,O,. In the case of the sulfurized catalyst the partial order with respect to

0

>

-1

u

I? -2

-3

.,..

FIG. 1 1 .

olysis;

-3

-2

-1

0

LOG PH2

Effect of the hydrogen partial pressure o n the rate of cyclopentane hydrogenPt/AI,O,; A, sulfurized Pt/AI?O,.

S U L F U R IN CATALYTIC HYDROGENATION REACTIONS

299

TABLE V EffPct.7 of the Sulfurizution oj' Plcitinum" Catalyst

K

Pt/AI,O, Pt(S)/AI,O,

I100 130

(AJA,,F 5.5 t 0.3 1.5 t 0.3

Effects are o n the rate constant K (molecules hour-' accessible Pt atoms-' atmospheres-') and on the ratio of the adsorption equilibrium constants of cyclopentane and hydrogen (cyclopentane hydrogenolysis).

cyclopentane is positive and close to 0.6; on the other hand, o n Pt/AI20, the order is much lower and close to 0.1. Besides, the partial order with respect to hydrogen cannot be determined because the reaction rate goes through a maximum when the hydrogen pressure varies. However, attention should be paid to a shift of the maximum toward weak hydrogen pressures on the Pt(S)/AI2O3catalyst, with regard to a maximum observed for the Pt/AI,O, catalyst. The mathematical exploitation of this kinetic study (Table V) shows that sulfur modifies both the rate constant of the reaction and the ratio of the adsorption equilibrium constants of hydrogen and cyclopentane, showing that hydrocarbon adsorption is more strongly inhibited than that of hydrogen by sulfur deposition. This conclusion enables explanation of the initial higher order of cyclopentane on the sulfurized platinum, but also the appearance of the inhibiting effect of hydrogen for a lower pressure on the Pt(S)/AI20, catalyst (72). When the sulfur-containing molecule is a thiol (which in aqueous liquid phase is adsorbed dissociatively on platinum as we saw previously), the free energy of adsorption, -AGO, of olefinic compounds, such as maleic acid, increases with a decreasing acidity of the adsorbed thiol under consideration. Adsorption of olefinic compounds, which occurs with electronic transfer from the double bond to the metal, is therefore enhanced by adsorption of thiols according to the following sequence H2S < CH,(CH,),SH < CH&CH,),SH, showing a higher charge transfer from the metal to the sulfide-adsorbed radical in relationship to the basicity of the conjugated bases CH,(CHJ2S- and CH,(CH2)$ - (58). The effect of sulfur on adsorption and decomposition of formaldelyde was studied by Abbas and Madix (73). On Pt(l I I ) adsorbed sulfur reduces the amount of adsorbed formaldelyde. This was not explained by a charge

300

J . BARBIER 6‘t 121.

transfer, but by assuming a direct interaction between sulfur and t h e oxygen of the methoxide intermediate resulting from the H2C0 surface decomposition. In conclusion, coadsorption of sulfur with different molecules induces a decrease of the adsorption capacity (by a geometrical effect) and a change in the binding energy (by a ligand effect). This change can be a decrease (H2, CO, or saturated hydrocarbon) or an increase (olefinic compounds with high electronic densities on the double bond). Such results are consistent with the following effects:

1. A preferential adsorption of sulfur on some special sites that can exhibit a higher bond strength for adsorption of given reactants. 2. An electronic effect due to the variation of the electronic states density of the metal atoms in the vicinity of sulfur atoms. 3. An effect due to lateral interactions between adsorbed sulfur and a coadsorbed molecule. Such interactions can enhance or reduce the adsorption free energy of the reactant under consideration.

IV. Effect of Sulfur Adsorption on the Catalytic Activity On metallic catalysts, sulfur is strongly adsorbed, and even if only minute amounts are found in the feedstock, accumulation can occur on a significant part of the metallic surface area. In the adsorbed state, the poison molecule will deactivate the surface on which it is adsorbed: then the toxicity will depend on the number of geometrically blocked metal atoms. On the other hand, the chemisorption bond between the poison and the metal can modify the properties of the neighboring metallic atoms responsible for the adsorption of reactants. If the interaction between the poison and the metal is weak, the structure of the metal will remain unchanged, but it can induce a perturbation all around the adsorption site, which will be able to modify the catalytic properties of this surface. Yet if the interaction between the metal and the adsorbate is strong, it can go as far as to modify the metal-metal bond. The mobility of the surface atoms can be increased and a new superficial structure can appear. In order to obtain deeper insight into such complex effects, the definition of the deactivation extent induced by sulfur adsorption can provide a lot of information about its action. The simplest technique to characterize a poison is to define its “initial toxicity” as the number of accessible metal atoms deactivated through adsorption of t h e first molecule of poison (74).

30 1

SULFUR IN CATALYTIC HYDROGENATION REACTIONS

TABLE V1 Determinution of the Initiul Toxic,ity c?f f o r Different Reuctions Reaction

Sii&ir

Initial toxicity of sulfur 2.0 0.2 I .7

Hydrogenation of benzene Monoexchange ofbenzene Epirnerizdtion of dirnethyl cyclohexane (453 K ) Hydrogenolysis of cyclopentane Monoexchange of cyclopentane Multiple exchange of cyclopentane Epirnei-ization of dimethyl cyclohexane (383 K )

9.0 0.5 9.0 8.9

Table VI gives the values obtained for various hydrocarbon reactions and shows that toxicity of sulfur can vary greatly according to the reaction taken into consideration. On the other hand presulfiding the reforming catalysts causes different deactivation effects on various reactions. Thus hydrogenation and dehydrogenation reactions are less modified than the hydrogenolysis reaction is. Yet hydrogenolysis of linear hydrocarbons is by far the most sensitive reaction to sulfur poisoning. Finally, a comparison between three different metallic catalysts, Pt/AI,O,, Pt-Re/Al,O,, and Pt-Ir/Al,O,, shows that, at comparable sulfur coverage, the Pt-Re catalyst is the most sensitive to this poison. On the contrary, addition of small quantities of iridium to platinum seems to increase its thioresistance (Table V11). The toxicity of

TABLE V11 Rtrrio of the Activities o f Frrsli ( r ) und Sirlfrtrized ( r ' ) Catulysts f o r Dgjierent Reuctions rir'

Pt/AI,O, Reaction

(OS = 0.39)

Benzene hydrogenation (373 K) Cyclohexane dehydrogenation (573 K ) Cyclopentane hydrogenolysis (573 K ) Ethane hydrogenolysis (633 K)

3.0 1.7 25 280

Pt-Re/AI?O, (0, = 0.44) 10 2.3 50

400

Pt-lr/AI?O3 (8, = 0.45) 3.0 1.5 13 75

302

J . BARBIER C‘t U l .

sulfur will thus depend on the reaction under consideration but also on the nature of the metallic catalyst (75). Further, Maxted (36) studied the influence of several sulfur compounds on the activity of platinum black for the hydrogenation of crotonic acid in the liquid phase. He noticed that between 15 and 50°C the toxicity remains constant for a sulfur compound, pointing out the irreversibility of sulfur adsorption. Conversely, the toxicity of various compounds increases with the molecule size. For molecules containing two sulfur atoms, losing all freedom of rotation through this double adsorption, t h e toxicity is less than for molecules of the same length containing only one atom of sulfur. Finally, the toxicity of sulfur can be changed by varying the experimental conditions, for example, the reagent concentration. The hydrogenation of maleic acid, carried out in the liquid phase, can be held up as an example. At a high reagent concentration, the catalyst is highly covered with unsaturated compound and the maleic acid order tends to zero (76). Under such conditions, the initial toxicity of the poison will be limited to t h e inhibition of the hydrogen adsorption (77). In the case of sulfur-poisoned platinum catalysts, for which each sulfur atom at low coverage inhibits the hydrogen adsorption on two platinum atoms, the toxicity is two. At low maleic acid concentration, the kinetic order of the hydrogenation reaction with respect to the reagent is one and in such conditions the rate of the reaction is r = KO-)HC3c,,efinr. Sulfur adsorption acts on both the adsorption of hydrogen and the adsorption of the olefinic compound. So in such conditions the deactivation law is given by the following relation: rIr(, = k ( I - O S ) ’ ,in good agreement with a hydrogenation site made by an ensemble of 5 ? I accessible platinum atoms (78). Pinol et d.(79)examined how the presulfiding with H,S of the Pt( I 10) crystal face changes the hydrogenation of butadiene. The activity decrease is proportional to the sulfur coverage and the activation energy of the hydrogenation does not change. They infer, consequently, that the sulfur acts only as a geometric hindrance. Such a similar conclusion has been drawn by Zrncevic et (11. (80) for the hydrogenation of benzene on supported nickel: the activity decrease is linearly related to the sulfur content in the thiophene form. Pradier ot ul. (8f)do not confirm the results obtained in the hydrogenation of butadiene on Pt( I 10)when they consider the (100)face. The activity decrease is no longer proportional to the sulfur coverage. They consider that the sulfur action changes with the crystallographic face: the sulfur would adsorb on the diene sites over the (100) face, but on the hydrogen activation sites, over the ( 1 10)face. Moreover, the complexity of the sulfur action is not only seen by changing the substrate or the catalyst. It is

SULFUR IN CATALYTIC HYDROGENATION REACTIONS

303

also experienced by changing the test procedure (82):the deactivation of styrene hydrogenation into ethylbenzene is only noticed for amounts greater than 100 ppm. But a preaddition of quantities close to 50 ppm makes the catalyst unaffected by a second addition of sulfur. The nature of the unsaturated hydrocarbon has a very important role in the sulfur action: Berenblyum et a / . (83) have reactivated a palladium catalyst, poisoned with thiols, through the interaction with phenylacethylene; the presence of acetylenics together with low levels of sulfur even activate the nickel sites activity for acetylene hydrogenation (84, 85). Boitiaux rt a / . (61) have examined the influence of palladium sulfuration on the hydrogenation and isomerization of I-butene, 1,3-butadiene, and 1-butyne. The tested catalysts have been sulfided with thiophene to obtain an atomic ratio (sulfur per surface palladium) varying between 0 and 0.5. The thiophene in heptane solution is put in contact with the reduced palladium catalyst at S O T , under 2 MPd hydrogen pressure. The butane evolution is followed during the sulfiding step (see above) and a control of total sulfur adsorption is performed by the analysis of the heptane after the sulfiding step and through X-ray fluorescence after the reaction step. The measure of activity on these pretreated catalysts gives a direct access to the real toxicity of sulfur for the specific reaction. Figure 12 emphasizes the turnover numbers of the I-butene hydrogenation and isomerization versus the sulfur level. The sulfiding of the metal deactivates the catalyst for both reactions. Nevertheless, they are quantitatively not similarly affected: the hydrogenation shows a toxicity of 5 and the izomerization of 2. The rates are not proportional to the free surface portion, which would be indicated by a toxicity of 1. The addition of one sulfur atom deactivates more than one palladium atom. For 1,3-butadiene hydrogenation, the toxicity of sulfur is 3 (Fig. 13). which is lower than the toxicity for olefin hydrogenation. The hydrogenation of I-butyne has also been studied for various ratios of sulfur over palladium. As was already published (86), the 1-butyne hydrogenation rate increases with time. The same effect has been observed on sulfided palladium. The turnover number is consequently presented for I-butyne hydrogenation versus the sulfur content for various I-butyne conversions (see Fig. 14). During the first minutes of reaction (0-25% conversion), the toxicity of sulfur appears close to 1; the rates are proportional to the free surface. However, at higher conversion, the rate becomes independent from the sulfur ratio. The toxicity is zero. Such a phenomenon could be due to a destabilization action of the acetylenic on the metal-sulfur bond. Such an action has been evidenced: some sulfur goes back into the solution (61). The extraction of sulfur produced by the I-butyne has been quantified

304

J . BARBIER

0

0.2

et al.

0.4

0.6

0.a

SPd, FIG.12. Effect of degree of sulfuration of palladium on the rates of hydrogenation and isomerization of I-butene ( T = 22°C. P = 10 bars); 0 , hydrogenation; W, isomerization

from an initial sulfur content S/Pd = 0.5. The sulfur content of the catalyst after the hydrogenation has been determined through direct analysis of the catalyst and by calculation from the sulfur content of the solution (Table VIII). In conclusion, the observed toxicities of sulfur are very much dependent on the type of hydrocarbon and on the reaction. Except for l-butyne, the toxicities are greater than 1; the sulfur deactivates not only the metal site where it is adsorbed but also the neighboring sites. The isomerization of 1-butene through double-bond shift is less affected than is its hydrogenation. Figure 12 indicates that the hydrogenation becomes zero for sulfur ratios higher than 0.5, but the isomerization still exists at these levels of sulfur. One can consequently consider that the Ibutene still adsorbs on a palladium surface that is almost completely sulfided. Moreover, the hydrogenation and the bond shift reactions go through the same primary reactive intermediates (87). We cannot thus explain the differences in toxicities by different actions of sulfur on the I butene adsorption. The only possible action concerns the surface species involved in the interconversion or in the hydrogenation. The hydrogenation is fast if the hydrocarbon species are sufficiently fed

SULFUR IN CATALYTIC HYDROGENATION REACTIONS

0

0.1

0.2

0.3

0.5

0.4

305

0.6

Wds FIG. 13. Effect of degree of sulfuration of palladium on the hydrogenation rate of 1,3butadiene ( T = 30"C, P = 10 bars).

with activated hydrogen; the isomerization does not consume hydrogen. That could mean that the sulfur affects essentially the hydrogen activation. Such an interpretation is in accordance with the results of Oudar et al. ( 7 f u , b ) ,who demonstrated that the H,/D, reaction is more affected by sulfur than the butadiene hydrogenation is. However, the rate of H,-D, equilibration reaction (at 125°C) on the Pt( 111) face, modified by sulfur adsorption, is increased at low sulfur coverages showing a maximum at eS= 0.11 (88, 89). In the same way, an activation effect of sulfur, adsorbed at low coverages on polycrystalline

TABLE Vlll Initial und Final Sulfur Content o j Palladium Catalyst Used in the I-Butyne Hydrogenation

Initial sulfur

Final sulfur (Catalyst analysis)

Final sulfur (Solution analysis)

0.50

0.35

0.37

306

J . BARBIER

0

0.2

et ul.

0.4

0.6

0.8

1

WPds FIG. 14. Effect of the degree of palladium on the hydrogenation rate of I-butyne, for different conversions ( T = 20°C, P = 10 bars).

platinum, is found for the electrocatalytic reaction of hydrogen evolution (6%. The different toxicities found for I-butene, 1,3-butadiene, and I-butyne hydrogenation can be explained by assuming that the energetic adsorption of unsaturated hydrocarbons destabilizes the metal-sulfur bond producing a real desulfurization with I-butyne. The destabilization exists also with the butadiene, as has been shown on platinum (71). As we saw previously on Pt catalysts at low temperatures in an aqueous liquid phase, thiols are adsorbed without decomposition of the sulfide radical. Thus on platinum we can study the effect of t h e chain length on toxicities of sulfur-containing molecules. For the hydrogenation reaction, like hydrogenation of maleic acid, at a high concentration (zero order with respect to this reagent), one molecule of thiol (hexanethiol or propanethiol), by adsorption of the sulfur atom on one platinum atom, inhibits the adsorption of one atom of hydrogen. Thus in this kinetic field the initial toxicity of thiols, whatever the length of the chain, is equal to 1. On the other hand, at low maleic acid concentrations, the toxicities of different sulfides are very different [see Table 1X (5S)l.

SULFUR IN CATALYTlC HYDROGENATION REACTIONS

307

TABLE IX Initiul Toxiciiies of Dqfereni Sulfur Compounds f o r Maleic Acid Hydrogenfition

Sulfur compound

Initial toxicity

According to Maxted (36), at low temperatures the molecule would be adsorbed through sulfur atom anchoring. Around this point the hydrocarbon chain could, thanks to its free rotation, inhibit adsorption over the whole adjacent surface. A comparison of the toxicities of hexanethiol and propanethiol shows that the ratio of these toxicities is exactly equal to the ratio of the surface area covered by the two molecules. 1.3 tpropanethiol - - = 0.26 t hexanethiol 5

82 A’ 314 A’

On the other hand, a comparison with the toxicity of H,S, which is higher than the toxicity of propanethiol and hexanethiol, shows that such an explanation is not sufficient. Thus, for sulfur compounds, it is obvious that toxicities cannot be only defined by a geometrical blocking of a metallic surface area. It is obvious that modifications induced by sulfur adsorptions, such as charge transfer or interaction between the adsorbed sulfur and the reagent, which alter the bond strength of the different reagents, will result in a change of the specific activity of the unpoisoned metallic surface and accordingly the apparent initial toxicity will also be changed. We saw previously that for maleic acid adsorption the higher the acidity of the thiol the lower the increase of binding energy of maleic acid. Such modifications are able to change the specific rate of the reaction. So comparison of the toxicities of different compounds cannot be limited to the comparison of the surface area blocked by the adsorbed poison, as was proposed by Maxted. In conclusion, the toxicities of sulfur compounds are the result of several effects, suggesting that sulfur induces both geometric and electronic limitations or enhancements of the surface reactivity. Finally, some unsaturated compounds are able by their high energy of chemisorption to displace adsorbed sulfur and to avoid sulfur poisoning.

308

J . BARBIER

V.

et al.

Effect of Sulfur Adsorption on Catalytic Selectivity

Modification of selectivity induced by sulfur adsorption has been widely studied ( 4 / , YO, 75) and can be evaluated by comparison of the initial toxicities (see Table VI) for different parallel reactions. As an example, sulfur on platinum does not inhibit similarly benzene hydrogenation and the exchange with deuterium reactions. The H/D exchange is less deactivated than the hydrogenation. For instance, one sulfur atom (introduced as H,S) poisons five accessible platinum atoms for benzene hydrogenation, whereas it poisons only two platinum atoms for exchange. The mechanism of selectivity modification induced by sulfur adsorption is far from being explained. Each of the previously introduced models (geometry, ligand effect, and restructuring) is sufficient alone to explain modifications of catalytic selectivity. It was mentioned previously that sulfur is preferentially adsorbed on some special metallic sites. Sulfur would thus show an important toxicity for any reaction occurring on the sites where it is adsorbed and a negligible toxicity for all “structure-sensitive” reactions occurring on the sites where the poison does not adsorb. On the other hand, if sulfur adsorption, according to numerous studies, such as those by Oudar or G. A. Somorjai (21, 9 / 4 3 ) , is able to induce surface reconstruction, the change in selectivity could be explained by assuming that at least one of the considered reactions is a “structure-sensitive” reaction that preferably occurs on some monocrystalline orientations. Finally, in addition to being the case for the two previously introduced models, the change in selectivity can be a consequence of a charge transfer created by sulfur adsorption. Such transfers are able to change the specific properties of the unpoisoned surface and the thermodynamics of the different reactants’ adsorption, which is equivalent to modifications of the relative rates of competitive surface reactions. Competitive reactions and essentially competitive hydrogenations were often used to discuss the extent of the electronic transfers induced by poison adsorption. For instance, two model molecules with different electronic densities are chosen, e.g., benzene and toluene. In this case the electronic donor properties of the methyl group increase the electronic density on the insaturated bonds. During competitive hydrogenation of benzene and toluene, sulfur adsorption poisons the two reactions but is less toxic for toluene than for benzene hydrogenation (94-96). Sulfur, by its adsorption as an electron acceptor, is able to decrease the electronic density of the unpoisoned metallic surface area and could favor the adsorption of the reactant with the highest donor properties enhancing the hydro-

SULFUR IN CATALYTIC HYDROGENATION REACTIONS

309

TABLE X Evolution of the Ratio q f Adsorption Eyirilihriirm Constants und of Selectivities .for Competitive HydrogenationU Sulfur coverage 0 0.05 0.11 0.20

K , $ K ~ ~ 2.14 0.20 0.02 0.01

SMIDMM

2.60 0.92 0.73 0.50

Maleic (M) and dimethyl maleic (DMM) acids on partially sulfurized platinum catalysts. (I

genation of toluene in competition with benzene. In the same way, the competitive hydrogenation of dimethyl maleic and maleic acid can be used in the liquid phase (46).We saw previously that each olefinic carbon atom bears almost 0.05 electron for the dimethyl maleic acid and 0.02 electron for maleic acid. During hydrogenation of these two compounds on different platinum catalysts, partly poisoned by increasing sulfur coverages, the selectivity of the catalyst, introduced as the ratio of maleic acid hydrogenation over dimethyl maleic acid hydrogenation, decreases. Such variation can be explained by a change of the equilibrium constants of adsorption of these two olefinic compounds (Table X). This result is in good agreement with the electronic transfer, from the metal to the adsorbed sulfur atom, previously described for competitive hydrogenation of benzene and toluene. During olefinic hydrocarbon hydrogenation, a parallel reaction can occur: the double-bond shift. As we saw previously, this isomerization is activated by sulfur as compared to hydrogenation (97-99). For I-butene transformation on presulfided palladium, Fig. 15 gives the 2-butenelbutane ratio versus the sulfiding extent of palladium. On the other hand, the trans/ cis ratio of butenes increases from 1.4 on pure palladium to a plateau at 2.9, which is close to the thermodynamic equilibrium. A large variation of the isomerization/hydrogenation ratio, going from 1 . 1 to 44 for a sulfur coverage of 0.5, has also to be mentioned. The influence of sulfur introduced in thiophenic form to the hydrocarbon feedstock (2-methyl-l-butene), on presulfided catalysts, has been studied (61). The 2-methyl-I-butene is isomerized to 2-methyl-2-butene and is hydrogenated to isopentane. The influence of thiophene on the isomerizationlhydrogenation ratio is indicated in Table XI for two presulfiding values: 0.125 and 0.33.

310

J . BARBIER

et al.

50

3

40

32

I

/ - - - b

30

(v

2

5!

+

8

8 \

(\1

v2 *O

v

1

10

n 0

T

I

I

I

I

0.1

0.2

0.3

0.4

0.5

0 0.6

WPds FIG. IS. Effect of degree of sulfuration of palladium on the selectivity of hydrogenation and isomerization o f I-butene ( T = 2 T C , P = 10 bars; m,A,thiophene; A,H,S; 0, DMDS).

The added sulfur seems to be of little importance in comparison with the sulfur deposited during the presulfiding. The catalyst that has been sulfided to 0.125 does not decompose further thiophene in presence of methylbutene. This result is very clearly shown in Fig. 8, where the maximum sulfiding ratio in presence of methylbutenes is close to 0.125. On the other hand, diolefin hydrogenation in the presence of sulfur has been largely studied from a practical view, because all selective hydrogenations of pyrolytic gasoline are designed for hydrogenate diolefins in the

TABLE XI IsomerizutionlHydro~enalionof 2-merhyl-I-burene

Quantity of sulfur in the reaction mixture Presulfuration extent

0 ppm

10 ppm

30 ppm

0.125 0.33

3.5 27

6.5 31

2s

11

SULFUR IN CATALYTIC HYDROGENATION REACTIONS

31 I

TABLE XI1 Isoprene Hydrogetwtion on Pulludiurn Black: Injurnce of Dietliylsulfide Compound/event

s = 0

S

=

8000 ppm

ic5 3MeBel 2MeBel 2MeBe2 iPr

12.5 17.2 23.4 35.0 11.9

I .6 36.9 23.8 26.6 11.1

Olefin yield 1.2 addition 1.4 addition

85.8% 53.7% 46.3%

98.2% 69.5% 30.5%

presence of various sulfur compounds. The presence of sulfur increases the selectivity of diolefin hydrogenation, because less paraffin is formed through overhydrogenation (59, 100-1U3). The influence of sulfur on this selectivity is very clearly demonstrated in Tables XI1 and XI11 [see Kitayama and Hayakawa (104) and Le Page (105)l.In both tables the olefin yield is seen to be greatly increased by the presence of sulfur. However, this is not the only consecutive selectivity to be changed by sulfur: in the competitive hydrogenation of aromatics, olefins, and diolefins, the aromatic and olefin hydrogenations are completely stopped by sulfur. In the isoprene hydrogenation results of Table XII, the ratio of the 1,4 addition/ 1,2 addition decreases on palladium with sulfur addition. Such a result is not at all in accordance with other studies dealing with nickel

TABLE XI11 Injiunce o j Sulfur Cotnpounds on Nickel Cutulyst Selectivity

Compound

Feedstock

Benzene Cyclohexane C yclohexene Hexene Isoprene Hexane Methylbutenes lsopentane

50 20 IS 5 10 0 0 0

Product without sulfur

Product with 100 ppm thiophenic sulfur

20

50

65

20

0 0 0

15 4

5 0 10

0.3 1

9.5 2

312

J . BARBIER

et

Ul.

TABLE XIV Buiunediene Hydrogenation over Nickel, Presulfurizc~dwith H2S Sulfur level (S/Ni) 0 0.20 0.22 0.2s 0.35 0.38 0.5 I 0.62

I-Butene

/ran.\-2-butene

cis-2-butene

6.5 41 39 33 30 2.5 21

28 51 52 57 61 68 62 64

7

27

x 9 10 9

7 II 9

(90). Okamoto et ul. (106) have found on sulfided nickel an improved I ,4 addition with butadiene (Table XIV). The improvement of I ,4 addition has been explained (106-108) by an electronic transfer from the nickel to the sulfur, which changes the reaction mechanism: on sulfided nickel the 7~ allylic species are favored to the detriment of u-rr olefinic bonds. Boitiaux et al. and Verna (61, 62) confirmed this variation of selectivity in the butadiene hydrogenation for the parallel reactions ( I ,2 and 1,4 addition) on presulfided palladium. Figure 16 shows the 2-butene/l-butene and the trans-butenelcis-butene ratios versus the sulfurization extent of the surface palladium. The sulfur decreases the trans/cis ratio and favors the I ,4 addition. However, the sulfurization of the surface palladium has a detrimental effect on the consecutive hydrogenation (Fig. 17): the olefin yield is lower on sulfided palladium. Such a result is not in accordance with the literature cited above and this difference can be explained by differences in experimental conditions: the improved consecutive selectivities were obtained with the sulfur present in the feedstock, i.e., in adsorption competition, when the detrimental effect on the consecutive hydrogenation was pointed out on partly sulfided palladium without sulfur compounds in the feedstock. The selectivity for olefin production in the hydrogenation of isoprene is increased by the presence of sulfur in the substrate (Fig. 18). So, on palladium catalysts, the role of Pd-S (presulfiding) is different from the role of sulfur-containing molecules in competition for adsorption with the unsaturated hydrocarbon.

VI. Concluding Remarks

The large binding energies of adsorbed sulfur on all metallic catalysts can explain the high toxicities of this compound for various reactions.

313

SULFUR IN CATALYTIC HYDROGENATION REACTIONS 10

2

9

a

cv

mocv \

-cv Y

1.5

92

7

+ %

52

'cv

v

6

D.5 5

A

0

I 0.1

I 0.2

I 0.4

I

0.3

I 0.5

0.6

3

WPds FIG. 16. Variation of selectivities for the formation of different olefines as a function of degree of sulfuration ( T = 3WC, P = 10 bars).

0

25

50

75

BUTADIENE CONVERSION (%)

100

.,

FIG.17. Evolution of the amount of olefines, formed during the butadiene hydrogenation, for different degrees of sulfuration: 0, without sulfur: A,SIPd, = 1/8: 0 , S/Pd, = 1/4; SIPd, = 1/2 ( T = 30"C, P = 10 bars).

314

J . BARBIER

et al.

H2 consumption

(%I

75 Isoprene = 100 cc.

50

25

0

50 H2 consumptian(%I

100

FIG.18. (A) Hydrogenation of isoprene on Raney nickel without thiophene; P, isopentane; 2-1,2-methyl-l-butene;2-2,2-methyl-2-butene; 3-1,3-methyl-l-butene. (B) Hydrogenation of isoprene on Raney nickel in the presence of thiophene.

S U L F U R IN CATALYTIC HYDROGENATION REACTIONS

315

Nevertheless, two factors strongly influence the heat of sulfur chemisorption on metal surfaces: relative coverage and crystallographic structure. Thus sulfur chemisorbs at high coordination sites and, as a result, a selective poisoning of structure-sensitive reactions, preferentially catalyzed by these sites, may occur. Such a simple geometrical model can be used to explain change in selectivities induced by sulfur adsorption. On the other hand, by a “ligand effect,” the reactivity of sites located at varying distances from the sulfur-occupied site may be affected. As a proof of charge transfer, adsorbed sulfur is able to decrease the binding energy of adsorbed hydrogen when the free energy of adsorption of olefinic compounds can be increased on partially sulfurized metallic catalysts. In hydrogenation conditions the effect of sulfur adsorption is the result of interactions between the metal, the hydrocarbon, and the sulfur-containing compound. As a consequence, for a given metal, the sulfur coverage, and its effect on the activity and selectivity of the unpoisoned metallic surface area, will be defined by the nature of the hydrocarbon. Such interaction is not only a simple competitive chemisorption but a mutual effect. The sulfur-containing compound and the unsaturated hydrocarbon are able to modify their own interactions with the metallic catalyst. As an example, during catalytic hydrogenation, the deactivation induced by sulfur addition can be changed by the nature of the unsaturated hydrocarbon under consideration-the higher the unsaturation degree, the lower the toxicity. Moreover, adsorption of deeply unsaturated compounds is able to induce desorption of adsorbed sulfur species and so the higher the unsaturation degree, the lower the sulfur coverage. In the case of butene, hydrogenation toxicity of sulfur compounds is higher for hydrogenation than for isomerization. Such low toxicity for isomerization points out that, even in the presence of sulfur-containing compounds, butene is adsorbed o n the catalyst. So, the higher toxicity for hydrogenation than for isomerization may be explained by an inhibiting effect of sulfur on the hydrogen activation, corroborating the decreasing binding energy of hydrogen adsorbed on sulfided catalysts. REFERENCES 1 . Hettinger. W . P., Keith, C . D.. Gring. J . C . . and Teter, J . N . , Ind. Eng.Chern. 47,719 (1955). 2. Hanner, E.. Kaufmann, H., and Leibnitz, E., Freihrrg. For.\clz. 67. 329 (1964). 3. Minachev. Kh. M., Kondratev, D. A., and Slyunyaev, P. J., K i n r f . Kntul. 2 ( S ) , 690 (1961). 4. Minachev, Kh. M., and Isagulyants, G. V . , Proc. I n / . Congr. Card.. 3rd, Amsrerdum. IY64, p. 308 (1965). 5 . Simpson, H . D., Adu. Cliem. S r r . 143, 39 (1975). 6 . Hegedus, L., and McCabe, R., Curul. Rru. Sci. Eng. 23 ( 3 ) . 377 (1981).

3 16

J . BARBIER

et al.

Bonzel, H. P., and Ku, R. J., Chern. Phys. 58, 4617 (1973). Benard, J . , Cutul. Reu. 3, 93 (1969). Oudar, J . , C . R . Acad. Sci. 249, 91 (1959). Benard, J . , Oudar, J., and Cabane-Brouty, F., Surf. Sci. 3, 359 (1965). Benard, J., Oudar, J., Barbouth, N., Margot, E., and Berthier, Y., Surf. Sci. 88, L. 35 (1979). 12. Rostrup-Nielsen, J. R., J . C a r d . 21, 171 (1971). 13. Oliphant, J . C., Fowler, R. W., Paunel, R. B., and Bartholomew, C . H., J . Cutul. 51, 229 (1979). 14. McCarty, J . G., and Wise, H., 1. Chern. Phy,v. 72, 6332 (1980). 15. McCarty, J. G., and Wise, H., J. Curd. 94, 543 (1985). 16. McCarty. J. G., Sancier, K. M., and Wise, H., J . Curul. 82, 92 (1983). 17. McCarty, J. G., and Wise, H., J. Cht,rn. Phys. 74, 5877 (1982). 18. McCarty, J . G., and Wise, H., J . Clzern. Phys. 76, 1162 (1982). 19. Alstrup. I.. Rostrup-Nielsen, J. R., and Raen, S., Appl. Curd. 1, 303 (1981). 20. Wise, H., McCarty, J. G., and Oudar, J., in “Deactivation and Poisoning ofCatalysts” (J.Oudar and H. Wise, eds.), p. I . Dekker, New York, 1985. 21. Oudar, J., Cutul. Reu. Sci. Eng. 22, 171 (1980). 22. Menon. P. G., Prasad, J., Proc.I n / . Congr. Curd.. 6rh,London, 1976. p. 1061 (1977). 23. Apesteguia, C. R., Barbier, J . , Plaza de 10s Reyes. J. F., Garetto, T. F., and Parera, J . M., Appl. Cutul. 1, 159 (1981). 24. Parera, J. M., Apesteguia, C. R., Plaza de 10s Reyes, J. F., and Garetto, T. F., R e o c t . Kine/. Cutul. Lett. 15, 167 (1980). 25. Barbier, J., Marecot, P., Tifouti, L., Guenin, M., and Frety, R., Appl. Cutul. 19, 375 (1985). 26. Apesteguia, C., and Barbier, J., Reucr. K i n e / . Critcrl. Lett. 19, 351 (1982). 27. Schultze, J. W., and Koppitz, F. D., Electrochim. Acru 21, 327 (1976). 28. Kornyshev, A. A., and Schmickler, W.. in “Electrodynamics and Quantum Phenomena,” Telavi (URSS), abstr., p. 195 (1984). 29. “Usuel de Chirnie Generale el Minerale” (F. Busnot and M. Besnard, eds.), 1. Dunot, Bordas, Paris. 30. Gallezot, P., Datka, J . , Massardier, J., Primet, M., and Imelik, B., Proc. I n t . Congr. Catal., 6/11. London, 1976 2 , 696 (1977). 31. Marecot, P., Mahoungou. J . R., and Barbier, J . , unpublished results. 32. C h i , M., J . Cutul. 37, 187 (1975). 33. Vedrine, J. C., Dufaux, M., Naccache, C., and Imelik, B., J.C.S. Furuduy Truns. I 74, 440 (1978). 34. Foger, K. E., and Anderson, J . R., J. Coral. 54, 318 (1978). 35. Apesteguia, C. R., Garetto, T. F., Brema, C. E., and Parera, J . M., Appl. Curd. 10, 291 (1984). 36. Maxted, E. B., Adu. Cutul. 3, 129 (1951). 37. Bartholomew, C. H., Agrawal, P. K.. and Katzer, J. R.. Adv. Catal. 31, 135 (1982). 38. Maxted, E. B., J . SOC. Chern. Ind. (Lond(~ri) 67, 93 (1948). 39. Maxted, E. B., J . Soc. Chern. 624 (1947). 40. Maxted, E . B., and Marsden, A.. J. Chern. Soc. 23 (1946). 41. Maurel, R., Barbier, J., and Leclercq, G., J. C u r d . 37, 324 (1975). 42. Contractor, A. Q.. and Lal, H . , J. Electrounul. Cliern. 96, 175 (1979). 43. Lamy-Pitara, E., and Barbier, J., Elec~troc~hirn. Acru 31, 717 (1986). 44. Buckley, A. N., Hamilton, 1. C., and Woods, R., J. Electrounul. Chem. 216,213 (1987). 45. Hordnyi, C., and Ritzmayer, E. M., J. Electrounal. Chem. 206, 297 (1986). 7. 8. 9. 10. 11.

SULFUR IN CATALYTIC HYDROGENATION REACTIONS

317

46. Lamy-Pitara, E., Bencharif, L., Lghouzouani, L., Tainon, Y., and Barbier, J . , in “Catalyst Deactivation 1987” (B. Delmon and G. F. Froment, eds.), p. 331. Elsevier, Amsterdam, 1987. 47. Lamy-Pitara, E., Tainon, Y., and Barbier, J., J. Chimie Phys. 85 (9), 861 (1988). 48. Lamy-Pitara, E., Tainon, Y., Beden, B., and Barbier, J., J . Electroanal. Chern. 279, 291 (1990). 49. Koestner, R. J . , Salmeron, M., Kollin, E. B., and Gland, J. L., Surf. Sci. 172, 668 (1986). 50. Saleh, J. M., Kemball, C., and Roberts, M. W., Trans. Faraday Soc. 57, 1771 (1961). 51. Den Besten, I. E., and Selwood, P. W., J . Cuial. 1, 93 (1962). 52. Rudajevova, A., Pour, V., and Regner, A., CON.Czech. Chern. Comm. 38,2566 (1973). 53. Bordoli, R. S., Vickerman, J. C., and Wolstenholme, J., Surf. Sci. 85, 244 (1979). 54. Peralta, L., Berthier, Y.,and Oudar, J., Proc. Colloy. Phys. Chim. Surf. S o l . , 14th, Antibes, 1978, p. 83 (1978). 55. (a) Delescluse, P., Thesis, Paris, 1979; (b) Delescluse, P., and Masson, A., Surf. Sci. 100, 423 (1980). 56. Pradier, C. M., and Berthier, Y . , Bull. Soc. Chim. (France) 3, 317 (1985). 57. Kelemen, S. R., and Fisher, T. E., Surf. Sci. 87, 53 (1979). 58. Lamy-Pitara, E., Tainon, Y., and Barbier, J., unpublished results. 59. Bourne, K. H., Holmes, P. D., and Pitkethly, R. C., f n t . Congr. Catal. Amsterdam, 1965, p. 1400 (1965). 60. De Plaen, E., Thesis, Louvain, 1979. 61. Boitiaux, J . P., Cosyns, J., and Verna, F., in “Catalyst Deactivation 34th” (B. Delmon and G. H. Froment, eds.), p. 105. Elsevier, Amsterdam, 1987. 62. Verna, F., Thesis, Paris, 1988 (Techn. ed.). 63. Mathieu, M. V., and Primet, M., C . R . Acad. Sci. (Paris) 300(2), 12, 561 (1985). 64. Bonzel, H . P., and Ku, R., J. Chem. Phys. 59, 1641 (1973). 65. Glowski, G. E., and Madix, R. J . , Surf. Sci. 115, 524 (1982). 66. Kelemen, S . R., Fisher, T. E., and Shwarz, J . A., Surf. Sci. 81, 440 (1979). 67. Mahoungou, J. R., Thesis, Poitiers Mars, 1990. 68. Woods, R., in “Electroanalytic Chemistry” (A. J. Bard, ed.), Vol. 9, pp. 1-162. Dekker, New York, 1976. 69. Lamy-Pitara, E., Lghouzouani, L., Tainon, Y., and Barbier, J., J. Electroanul. Chern. 260, 157 (1989). 70. Lamy-Pitara, E., Lghouzouani, L., Sieiro, C., and Barbier, J., J . Chim. Phys.. unpublished results. 71. (a) Oudar, J., Pinol, S., Pradier, C. M., and Berthier, Y., J. C u r d . 107,445 (1987); (b) Oudar, J., Pradier, C. M., Vassilakis, D., and Berthier, Y., Catal. Lett. 1, 339 (1988). 72. Apesteguia, C., and Barbier, J., Bull. Soc. Chim. (Frunce), 5-6, 1-165 (1982). 73. Abbas, N. M., and Madix, R. J., Appl. Surf. Sci. 7, 241 (1981). 74. Maurel, R., and Barbier. J., J . Chim. Phys. 73 (11-12), 995 (1976). 75. Apesteguia, C. R., and Barbier, J., J. Cutal. 78, 352 (1982). 76. Barbier, J., and Gonzalez, M. E., J. Chim. Phys. 76(3), 219 (1979). 77. Lamy-Pitara, E., Bencharif, L., and Barbier, J., Appl. Catul. 18, 117 (1985). 78. Lamy, E., Barbier, J., and Lamy, C., J. Chim. Phys. 77(10), 967 (1980). 79. Pinol, S . , Berthier, Y., and Oudar, J., C . R . Acud. Sci. (Paris). 300(2), 12, 561 (1985). 80. Zrncevic, S . , and Gomzi, Z., Chern. Eng. Sci. 38(8), 1351 (1983). 81. Pradier, C. M., Margot, E., Berthier, Y., and Oudar, J., Appl. Cuf. 31, 243 (1987). 82. L’Argentiere, P. C., and Figoli, N. S., J. Chrrn. Technol. Biotechnol. 38, 209 (1987). 83. Berenblyum, A. S., Mund, S. L., and Karelskii, V. V., Kinet. Caral. 76(3), 626 (1985).

318

J . BARBIER

et al.

Takenchi, A , , Tanaka, K., Toyoshima, I . , and Miyakara, K., J . Curd. 40, 94 (1975). Takenchi, A,, Tanaka, K., and Miyakara, K., J. Card. 40, 101 (1975). Boitiaux, J . P., Cosyns, J., and Vasudevan, S., Appl. Catal. 6, 41 (1983). Guisnet, M., Perot, G., and Maurel, R., J . Chem. Phys. 6 , 1059 (1972). Pradier, C. M., Margot, E., Berthier, Y., and Oudar, J., C.R. A c a d . Sci. (Paris),294, 2, 1321 (1982). 89. Oudar, J . , in “Metal-Support and Metal-Additive Effects in Catalysis” (B. lmelik rt ul., eds.), p. 235. Elsevier, Amsterdam, 1982. 90. Boitiaux, J . P., Cosyns, J . , and Martino, G., in “Metal-support and Metal-Additive Effects in Catalysis” (B. Imelik ef a/. eds.), p. 355. Elsevier, Amsterdam, 1982. 91. Somojai, 0.A., J . C a r d 27, 453 (1972). 92. Barbier, J . , in “Deactivation and Poisoning of Catalysts” (J. Oudar and H. Wise, eds.), p. 108. Dekker, New York, 1985. 93. Rhead, G. E., S u ~ Sci. . 15, 353 (1969). 94. Cosyns, J., Franck, J . P., and Gil, J. M., C.R. A c a d . Sci. (Paris) p. 287C (1978). 95. Tran Mahn Tri, Massardier, J., Gallezot, P., and Imelik, B., in “Metal-support and Metal-Additive Effects in Catalysis” (B. lmelik el a/., eds.), p. 149. Elsevier, Amsterdam, 1985. 96. Barbier, J., Marecot, P., and Tifouti. L.. React. Kinet. Cutal. Lett. 32, 269 (1986). 97. Burwell, R. L., Proc. I n f . Cong. Curd.. Paris, 1960, p. 987 (1960). 98. U.S. Patent 3,215,751. 99. Assefi, M.. Thesis, Paris, 1979. 100. Daimai-Imetik. G . , Rousseau, J., and Bertolini, J. C . , Lr Vide 27, Supp. 158, 36(1972). 101. Ger. Patents 1062693 and 1443512. 102. Fr. Patent 1,518,809. 103. Duyverman, C. J., Vlugter, J . C., and Van de Weerdt, W. J . , Inr. Cong. C u r d . , Amsrrrdam. p. 1416 (1965). 104. Kitayama, Y. and Hayakawa, M., Chem. Lett. 181 (1973). 105. Le Page, J . F., Appl. Hrrerng. C u r d . , p. 339, Techn. ed. (1987). 106. Okarnoto, Y . , Fukinio, K., Imanaka, T., and Teranishi, S., J. C u r d . 74, 173 (1982). 107. George, M . , Moyes, R. B., Ramanarao, D., and Teranishi, J . Card. 74, 173 (1982). 108. Burden, A. G . , Grant, J . , Martos, J., Moyes, R. B., and Wells, P. B., Discuss. Furaduy Soc. 72, 95 (1981). 84. 85. 86. 87. 88.

Index conversion, SbF5-treated metal oxides, I70

A

Acetylene, hydrogenation, 5 1, 92 Acid centers, structure, sulfate-supported metal oxides, 192-196 Acidity estimating, 166 sulfate-supported metal oxides, 186-187 surface, monolayer dispersion, 34-35 Acid strength, 166-167, 186-187 Activation barrier chemisorbed, C2Hx, 141, 144 C-H bond cleavage, 143 C2H, hydrogenation, 146 CO hydrogenation, 136-137 determination, 154 diatomic adsorbates, dissociation and recombinatioi~,109-1 13 HCOOH decomposition, 149-150 surface reactions, BOC-MP, 127-134 bond energies, 133 C 0 2 dissociation, 131 diatomic adsorbates, 128-130 heat of chemisorption, 133 non-LJ corrections, 128-129 polyatomic adsorbates, 130-134 triatomic adsorbates, 131-132 Adsorption ethylene, 49 monolayer dispersion, 33-34 AICI3, 168, 172-173 AICI3-CuC12, 173 AIC13-CuS04mixture, 173 Alcohols heat of chemisorption, 120-121 0-H bond cleavage, 140-141 AlF3 superacid, 205 Alkanes C-H bond cleavage, 133-134

dehydrogenation 2oo-201 hydrogenolysis, 82 reactions with hydrogen electron-deficient Pd, 74-76 reactions with H2 over unsupported Pd, 51-60 bond-shift isomerization, 52-53 deuterium as isotopic tracer, 52, 54 2,2-dimethylbutane skeletal rearrangement, 59 multiply bonded species, 52 neopentane isomerization, 54-58 Alkylation, isobutane, 197 Alkynes, hydrogenation, 73 A1203superacid, 6-7, 186, 189-190 Alumina, aluminum halides supported on, 205 y-Alumina, 65-66, 70-71 Aluminum halides, 172-173, 205 Arrhenius expression, 134, 136 ASED-MO, 151-153 Attractors, three-dimensional, 233-235

B Benzene, 72-73, 173 Benzylation, toluene, 177-178 BF3, liquid superacids attached to solid supports, 168 Bifurcation diagram, oscillatory CO/02, 233-234 Binding energy chemisorbed sulfur, 281 hydrogen, sulfur effect, 295-296 ZnO/Si02, 21-22 Binding energy shift, Pd, 62-64 319

INDEX

B/L intensity ratio, Pd dispersion effect, 88,90 Bond energy C2H, species, 141-142 gas phase and chemisorbed states, 133 HCOOH decomposition, 148-149 two-center, partitioning, BOC-MP, I13 Bond-energyhond-order method, 154-155 Bonding energy, BOC-MP, 106-107 Bond-order conservation, 101-103 Bond-order conservation-Morse potential (BOC-MP) model, 102-103, 155-156 applications, see also Heat of chemisorption surface reaction activation barriers, 127-134 assumptions, 103-105, 155 catalytic heterogeneous reactions, see Catalytic heterogeneous reactions comparisons with other techniques, 151155 ASE D-MO, 151-153 BEBO, 154-155 diatomic species, 151-152 effective medium theory, 154 EH method, 153 GVB-CI, 153 polyatomic species, 153-154 quantum chemistry, 15 1- I54 SCF-CI, 151 thermodynamic approaches, 154 diatomic adsorbates, 105-1 13 bonding energy, 106-107 bridge mode, 106-107 chelate structure, 107-108 dissociation and recombination activation barriers, 109-1 13 heat of chemisorption, see Heat of chemisorption Lennard-Jones potential diagram, 109110fl multidimensional potential diagram, 111 many-center M,-A interactions, 104, I55 methanol decomposition on metal surfaces, 140 Morse potential, 104 polyatomic adsorbates, 113-1 17 strength, 155 two-center M-A bond order, 103-104 zero-coverage extreme, 155

Bond-shift isomerization, alkanes, 52-53 Br2-A1Br3catalyst, 172 Bronsted sites, SOJZr02, 194 Butadiene, hydrogenation, sulfur effect, 302-303 1,3-Butadiene, hydrogenation, 48, 303,305 Butane formation, form adsorbed sulfur compounds, 289-293 isomerization, 183-184 reaction over SbF5-treated metal oxides, 169-170 reaction to isobutane and propane, 184185 Butanediene, hydrogenation, over presulfurized nickel, 31 1-313 Butanethiol, dissociation, 292 1-Butene, hydrogenation and isomerization, 303-304, 309-3 10 C

Carbidic carbon, 138, 146-147 Catalytic heterogeneous reactions, 134-15 I Arrhenius expression, 134, 136 C2H, transformation on transition-metal surfaces, 141-147 CO hydrogenation on platinum-group metals, 136-141 coverage and coadsorption, 135 HCOOH decomposition on transitionmetal surfaces, 147-151 heat of chemisorption, 135 reaction rate constant, 134-135 C-C bond cleavage, 143-145 C-H bond cleavage alkanes, 133-134 activation barriers, 143 gas-phase, 145 HCOO, decomposition, 150 C H K , formation, 143-144 Chelated adsorbates, heat of chemisorption, 126-127 Chemisorption, see also Heat of chemisorption acetylene, 49, 51 CO, 119-120 extractive, 80 Hz and CO, 87-89 CzH, bond energies, 141-142

32 1

INDEX

chemisorbed, activation barriers, 141, 144 hydrogenation, activation bamers, 146 isomerization enthalpies, 141, 143 transformation on transition-metal surfaces, 141-147 Cinnamic acid, adsorbed, sulfur effect, 297 Close-packed monolayer model, Moo3, 7 CO, also see Oscillatory catalytic reactions adsorbed Pd/A1203,FTIR spectra, 69-70 adsorption, IR, 68-69 chemisorption, 87-89, 119-120 coverage, oscillations on Pt surfaces, 228 heat of adsorption, Pt(100) surface, 221222 hydrogenation, 60-61 electron-deficient Pd, 76-77 on platinum-group metals, 136-141 oxidation oscillatory, 226-227, 250-25 1, 259 on Pd(l10), 262-266 on Pt(210), 260-262 pressure range, autonomous oscillations on Pt, 229-231 sticking coefficient, 216-217 titration curves, Pd(l10) surface, 264265

co2 dissociation, 131 formation faceted Pt, 242-243, 245 on Pd(1 lo), 263-264 steady-state rate, 216-217 sustained oscillations, 220-22 1 under isothermal conditions, 267-268 heat of chemisorption, 124 C-0 bond cleavage, 138, 150 CO-metal bond, adsorbed sulfur, 295 Crotonic acid, hydrogenation, 302 C~ClZ/Ali03,DTA, 31 CuC12/y-A1203, 8-9, 38 CuC12/~-AlzO3,22-23, 24-28, 33-34 Cu3d, XPS spectra, 21-22 Cyclopentane, hydrogen01 ysis, 298-299 D Dehydration, 2-propano1, 180 Dehydrogenation, alkane catalyst, 200-201 Desorption, H2S, 283

Detoxication, sulfur, 286 Deuterium, as tracer, 52, 54 Diatomic adsorbates, see also Bond-order conservation-Morse potential (BOCMP) model activation bamers of surface reactions, 128-130 heat of chemisorption, 120-121, 151-152 Dicoordination, versus monocoordination, BOC-MP, 125-127 Diethylsulfide, isoprene hydrogenation effect, 310-311 Differential thermal analysis, monolayer dispersion, 3 1 Dimaleic acid, free energy of adsorption, 296-297 2,2-Dimethylbutane, skeletal rearrangement, 59 Diolefin, hydrogenation, presence of sulfur, 309-311 Dispersion capacity, Moo3 on y-AI2O3,6-7 Disproportionation, polyatomic adsorbates, BOC-MP, 115-117 Dissociation activation barriers, 109-113, 128-129 barriers for triatomic adsorbates, 131132 butanethiol, 292 C02, 131 k range, 128 polyatomic adsorbates, BOC-MP, 115117 thiophene, 291 Dissociative chemisorption, thiophene, 292-294 Double-bond shift, 309

E Effective medium theory, 154 EH method, 153 Electron-deficient palladium catalysis by, 61-77 catalytic superactivity for neopentane conversion, 75-76 CO hydrogenation, 76-77 ESR and IR evidence for existence of, 64-72 hydrogenation of unsaturated hydrocarbons, 72-74 metal sintering, 65

322

INDEX

properties, 61 reactions of alkanes with hydrogen, 74-76 XPS study, 62-64 supported Pd, 8 1-82 Electron spectroscopy, monolayer dispersion, 19-22 Electron spin resonance evidence for existence of electrondeficient PD species, 64-72 Pd+ and Pd3+species identification, 6567 Enthalpies, C2H, isomerization, 143 Esterification, sulfate-supported metal oxides, 199 q2 coordination, formate species, 149-IS0 r)' and v2energies, 122-123 q'pncoordination, 125-126 Ethane, hydrogenolysis, 81 Ethylation, 172- 173 Ethylene, adsorption as function of CuCl content, 33-34 EXAFS, detectability, 46 Extractive chemisorption, 80

F Faceting, 242-246, 272 FeC13/gy-Alz03, Mossbauer spectra, 29-30 Fe203superacid, 199-201 Fe03/g y-A1203,Mossbauer spectra, 30 Fe203-lcatalyst, 181- 183 FeS04, 177-179 Fe2(S0d3 calcination, 178-179 catalytic activities, 177- 178 IR spectra, 182-183 Fischer-Tropsch synthesis, 147 Forced oscillations, oscillatory catalytic reactions, 236-242 Formaldehyde, adsorption and decomposition, sulfur effect, 299-300 Formic acid, decomposition, see Catalytic heterogeneous reactions Free energy of adsorption, maleic and dimaleic acids, 296-297 Friedel-Crafts acylations, sulfate-supported metal oxides, 199 Friedel-Crafts reactions, sulfate-supported metal oxides, 177-178

FTIR spectra, CO adsorbed Pd/AI2O3,6970

G Gas phase coupling, 248-249, 272 Gas-phase dissociation barrier, 112-1 13 Gas-Phase reaction energy profiles, 101I02 Graphites, superacid-intercalcated, 171173 GVB-CI, 153 H H2, see also Alkanes, reactions with H2 over unsupported Pd chemisorption, 87-89 reactions with unsaturated hydrocarbons over unsupported Pd, 48-51 relative deficiency, chain-lengthening homologation, 59-60 Hammett acidity function, 166-167 Harmonic entrainment, Pt, 237-238 HCOOH, see Formic acid H2-D2equilibration reaction, sulfur effect, 305-306 Heat conductance, 248, 272 Heat of adsorption: 221-222, 281 Heat of chemisorption BOC-MP, 117-127 chelated adsorbates, 126-127 CO, 119-120 C02, 124 diatomic adsorbates, 105-109, 120-121 r)' and q2 energies, 122-123 H 2 0 and alcohols, 120-121 metal surface values, 118-119 monocoordination versus dicoordination, 125-127 NO, 120-121 polyatomic molecules, 122-124 strong bonding, 124-125 weak bonding, 119-124 catalytic heterogeneous reactions, 135 gas phase and chemisorbed states, 133 theoretical calculations, 151 Heterogeneous catalysts, see Monolayer dispersion active components, 1-3

323

INDEX

dispersion on internal surface of zeolite, 17-21 HgC12 on active carbon, 10, 12 highly active monolayer-dispersed catalyst preparation, 34-35 impregnation method, 16-17 monolayer model, 1-2 supported metal particle preparation, 37-39 zeolite internal surface modification, 3940 Hexane, reaction at 550" C, 200 Hexanethiol, adsorption, 289 Hex-2-yne, hydrogenation, 5 I HfOz superacid, acid strength, 186-187 HF-SbF5 catalyst, 172 HgC12, on active carbon, X-ray diffraction, 10, 12 High-energy electron diffraction, monolayer dispersion, 30-31 H20, heat of chemisorption, 120-121 Homologation, chain-lengthening, relative deficiency of Hz, 59-60 H2S, 283-286-288, 307 H2S04, disposal, 197, 199 Hydrocarbons oxidation, 199 synthesis, solid superacids, 166 unsaturated, hydrogenation, 72-74 Hydrodesulfurization catalyst, preparation, 34-35 Hydrogen adsorption capacity, 295-296 binding energy, sulfur effect, 295-296 partial pressure and cyclopentane hydrogenolysis, 298-299 reactions with alkanes electron-deficient Pd, 74-76 Hydrogenation acetylene, 51, 92 alkynes, 73 benzene, 72-73 butadiene, 303, 305, 31 1-313 1,3-butadiene, 48, 303, 305 I-butene, 303-306, 309-310 carbidic carbon, 146-147 C2H,, activation barriers, 146 CO, 60-61 electron-deficient Pd, 76-77 on platinum-group metals, 136-141

crotonic acid, 302 diolefin, presence of sulfur, 309-31 1 ethylene, 49 hex-Zyne, 51 isoprene, 310-311, 313-314 maleic acid, 302, 306-307 2-methyl-l-butene, 309-3 10 olefins, 49 ratio of adsorption equilibrium constants, 309 unsaturated hydrocarbons, 72-74 Hydrogenolysis, 81-82, 298-299

1

Impregnation method, heterogeneous catalysts, 16-17 Intensity ratio MoOJy-AI2O3, ISS, 22-23 Raman, 24 XPS, versus Moo3content, 19-21 Ion-scattering spectroscopy, monolayer dispersion, 23 IR spectra CO adsorption, 68-69 evidence for existence of electrondeficient PD species, 64-72 Fe2(S0& and Fe20,-I catalysts, 182-183 low-content metal-loaded Pd/AIzO3, 6869 pyridine, 193, 196 Zr02, Ti02 and Sn02 superacids, 193, 195 Iron oxide, sulfate-supported, preparation, 179-182 Isobutane, 183-184, 197 Isomerization butane, 183-184 I-butene, 303-304, 309-310 C2H, enthalpies, 141, 143 2-methyl-1-bu tene, 309-3 10 methylpentanes, 171-172 neopentane, 54-58, 83 pentane, 173 selectivity, supported Pd, 84-85, 87, 9091 Isoprene, hydrogenation, 3 10-3 I 1, 3 13314

324

INDEX

K Kinematic waves, 258-259

L Langmuir kinetics, modified, 216 Laser desorption, Pt(100), wave types, 257 LEED intensity, 242-243, 250-251 Lennard-Jones potential diagram, 109-1 10 Lewis acids, solid superacids from, 168169 Lewis acid sites, SOdZr02, 194 Liquid superacids, supported on solids, 168-171

M Magic acids, 206 Maleic acid free energy of adsorption, 296-297 hydrogenation, 302, 306-307 McKervey-Rooney-Samman mechanism, 53-55 Metallic catalyst poisoning, 279-280, 300301 Metal oxides, see also Sulfate-supported metal oxides Lewis acid-treated, 169-170 superacids by, 201-204 Metals affinity of sulfur, 280-282 interaction with sulfur, 282-285 Methanation, 76-77, 138-139 Methane, ethylation, 172 Methanol activation bamers, 136-137 decomposition on metal surfaces, 140 formation, activation barriers, 136-137 synthesis, electron-deficient Pd, 76-77 Methylation, toluene, para selectivity, 3940 2-Methyl- I-butene, isomerization/hydrogenation, 309-310 Methylc yclopropane, hydrogenolysis, 8081 Methylpentanes, isomerization, 171-172 Mixed-mode oscillations, 234-236 Monocoordination, versus dicoordination, BOC-MP, 125-127

Monolayer-dispersed catalysts, highly active, preparation, 34-36 Monolayer-dispersed oxides and salts, 3637 Monolayer dispersion, 2-4 adsorption, 33-34 capacities, 13-14 close-packed model, 7 C U C I ~ / ~ - A I8-9 ~O~, differential thermal analysis, 31 dispersion capacity, 6-7 EXAFS, 26-29 HgClz on active carbon, 10, 12 high-energy electron diffraction, 30-3 1 ion-scattering spectroscopy, 23 MoOJy-AIZ03, 4-5 Mossbauer spectra, 29-30 Raman spectroscopy, 24-26 saltly-Al2O3,8, 10 spontaneous, 12- 16 coverage, 13-15 direct migration across particles, 1516 heat treatment role, 15-16 surface bond, 12-13 systems displaying, 8-9 static secondary ion mass spectrosCOPY, 22-23 Tammann temperature, 16 XRD patterns, 13, 15-16 support, 4 surface acidity, 31, 33 temperature for dispersion, 10-1 1 transmission electron microscopy, 30-32 UV diffuse reflectance spectroscopy, 26 X-ray photoelectron and auger electron spectroscopy, 19-22 XRD quantitative phase analysis, 5-6 ZnAc2, Moo3,and HgCI2 on silica gel, 10-1 1 Monolayer model, heterogeneous catalysts, 1-2 Moo3, 6-7 MoOdy-Al203,4-5, 22-26 Mo03/Si02,relation between acidity and Moo3 content, 31-33 Mo03fli02,threshold value, 24 M o O ~ i 0 3XPS , intensity ratio, 19, 21 MOO3/Zr02, 203-204 Morse potential, 104

325

INDEX

Mossbauer spectra, monolayer dispersion, 29-30 N

NaCVNaY zeolite, 18-19 Nafion-H, 174-177 Neohexane/D2 exchange, 81 Neopentane conversion, catalytic superactivity of electron-deficient Pd, 75-76 isomerization, 54-58, 83, 90-91 Nickel methanation on, 138-139 presulfurized, butanediene hydrogenation, 311-313 selectivity, sulfur effect, 310-31 I sulfur adsorption, 281-282 Ni/La20,/y-A1203, XRD, 36 NiOly-AI2O3, 16-17, 27-28 NiO/g y-AI2O3,30-32, 37-39 NO, heat of chemisorption, 120-121 NO + CO 1/2N2 + COz, on Pt(100), 267-271 Nonlinear dynamics, oscillatory catalytic reactions, 232-236

-

work function variation with CO pressure, 2 17-21 8 faceting, 242-246 forced oscillations, 236-242 rate equations, 214-215 spatiotemporal self-organization, see Platinum, 246 temporal self-organization, 232-236 Oxidation CO, see Oscillatory catalytic reactions on Pd(1 IO), 262-266 on Pt(210), 260-262 hydrocarbons, 199 state, Pd, 67 Oxides monolayer-dispersed, 36-37 spontaneous dispersion onto zeolites, 18, 20 Oxidelzeolite, dispersion capacities, 19, 21 Oxoacids, 199-200 Oxychlorination catalyst, preparation, 37 Oxygen adsorption, Pt(100), 224 coverage, 218-219, 227-228 Oxygen sticking coefficient, 260-262, 266

P 0 0 2

pressure, kinetic oscillations on Pd( 110). 262 sticking coefficient, 216-2 17 subsurface phase, Pd(1 lo), 264-266 0-H bond, cleavage, 140-141 Olefinic compounds, adsorption, sulfur effect, 296-300 Olefins, hydrogenation, 49 Oscillatory catalytic reactions, 213-215, 271-272, see also Platinum catalytic CO oxidation on Pt(1 I I ) and Pt(l10) surfaces C 0 2 formation, 216-217 kinetic oscillation mechanism, 220228 modified Langmuir kinetics, 216 steady-state oxygen coverage, 218219 steps, 215 sticking coefficients, 216-217

Palladium, 45-47, see also Supported palladium; Unsupported palladium active sites, 49-50 binding energy shift, 62-64 catalytic properties, 45 characteristics exhibited in D2/alkane reactions, 51-52 CO hydrogenation, 139-140 electron-deficient, see Electron-deficient palladium electronic properties, 74 formation of electropositive species, 70 future work, 93-94 model of superactive site stabilized on yalumina, 70-71 oxidation state, 67 sintered films, 55 small clusters, electronic structure, 62 sulfuration, 303 sulfur content, I-butene hydrogenation, 304-305 versatility, 47

326

INDEX

Pd( 1 lo), CO oxidation, 262-266 CO titration curves, 264-266 kinetic model, 266 kinetic oscillations, 262-263 subsurface oxygen phase, 264-265 work function and reaction rate, 263264 Para selectivity, toluene methylation, 3940 Partitioning, polyatomic adsorbates, BOCMP, 115-117 PdIAI203, 65-66, 68-69, 75 Pd/Ce02, 89-90 p-PdH phase, Pd transformation, 79-80 Pd(l 1])/mica film, epitaxially oriented, 5556 Pd/NaY, 7 1 Pd/Si02, 66, 71, 88, 90 Pd/TiOz, 89 Pentane, 170-171, 173, 197 Perfluorinated resin sulfonic acid, see Nafion-H Platinum autonomous oscillation conditions, 22823 I ball models of densely packed surfaces, 221-222 energy diagram, CO-induced transformation, 221-223 faceting, 242-247 forced oscillations, 236-242 dynamic phase diagram, 236-237 harmonic entrainment, 237-238 quasiperiodic behavior, 240-242 subharmonic and superharmonic entrainment, 238-240 kinetic oscillations, 220-228 ball models of densely packed surfaces, 221-222 energy diagram of CO-induced transformation, 221-223 faceting, 245-247 kinetic model, 227 limit sets, 232-233 nucleation of CO-induced transformation, 223-224 oxygen adsorption, 224 oxygen coverage, 227-228 periodic transformation of surface structure, 225

phases of surface, 222-223 Pt(210), 261 sustained oscillations, 220-221 work function variation, 225-226 nucleation of CO-induced transformation, 223-224 partially sulfurized, adsorbed hydrogen, 296 reflectivity variation with sulfur coverage, 287-288 spatiotemporal self-organization, 246260 chemical wave formation, 255-258 color changes, 247-248 computer simulations, 254-255 coupling by surface diffusionireaction, 249-250 coupling through gas phase, 248-249 heat conductance, 248 kinematic waves, 258-259 oscillatory CO oxidation, 250-251, 259 spatial patterns, 259 SPM, 250-253 surface defects, 260 theoretical spatial pattern evolution, 254 trigger waves, 258 types of waves excited by laser desorption, 257 sulfurization effects, 299 temporal self-organization, 232-236, 244 Pt(100), NO + CO + 1/2N2 + COZ. 26727 I Pt(llO), 295, 302 Pt(l1 I), preadsorbed sulfur, 295 Pt(210), CO oxidation, 260-262 Platinum-group metals CO hydrogenation, 136-141 heat of chemisorption and bond energy, 133 Polyatomic adsorbates activation barriers of surface reactions, 130-1 34 BOC-MP, 113-1 17 heat of chemisorption, theoretical calculations, 153-154 weakly bound, heat of chemisorption, 122-124 Propanethiol, adsorption, 289 2-Propanol, 180-182

INDEX

Pdy-Al2O3, redispersion of Pt, 38-39 Pyridine, iR absorption spectra, 193, 196

Q Quantum chemistry, heat of chemisorption determination, 15I - 154 Quasiperiodic behavior, Pt, 240-242

R Raman spectroscopy, monolayer dispersion, 24-26 Raney nickel, isoprene hydrogenation, 313-314 Recombination, activation barriers BOC-MP, 109-113 triatomic adsorbates, 131-132 Redispersion, through oxidation-reduction cycle, 38-39

S Salt/y-A1203,X-ray diffraction, 8, 10 Salts, 18, 20, 35-36 Salt/zeolite, dispersion capacities, 19, 21 SbFS, 168-172 SbF5/HY-zeolitecatalysts, 170- 171 Scanning photoemission microscopy, kinetic oscillations, Pt(100), 250-253 Self-consistent field-configurationinteraction, 151 SEM, Zr02 superacid, 189, 191 Silica, model of interactions with Pd, 89, 91 Single-crystal surfaces, see Oscillatory catalytic reactions Sintering, 38, 65 S O 2 superacid, preparation, 185-186 SiO2-AI20,,reaction with SbF5, 170 Skeletal rearrangement, 2,2-dimethylbutane, 59 Sn02 superacid, 185, 188, 193, 195 Solid superacids, 165-167, see also Sulfate-supported metal oxides acidity estimation, 166 acid strength, 166-167 advantages, 166 aluminum halide-metal salt mixtures, 172-173

327

catalyst of liquid superacids, 168-171 Hammett indicators, 167 intercalcated graphites, 171- 173 by metal oxides, 201-204 Nafion-H, 174-177 novel organic syntheses, 166 SO$Ti02 preparation, 182- I84 SOJZr02, 193-194, 197, 203-204 Spatiotemporal self-organization, see Platinum Static secondary ion mass spectroscopy, monolayer dispersion, 22-23 Sticking coefficient, CO, 216-217 Subharmonic entrainment, Pt, 238-240 Sulfate-supported iron oxide, preparation, 179- 182 Sulfate-supported metal oxides, 177-201 acid center structure, 192-196 acidity, 186- 187 aluminum halides supported on alumina, 205 applications, 199-20 1 autonomous oscillation conditions, 22823 1 catalytic action and properties, 178 esterification, 199 Friedel-Crafts acylations, 199 Friedel-Crafts reactions, 177-178 industrial processes, 197, 199 preparation catalyst appearance, 190, 192 iron oxide, 179-182 S0dTi02,Zr02, Hf02, Sn02, and AI2O3,182-192 Ti(S04)2and Z T ( S O ~ )192 ~, reactions catalyzed by, 196-199 surface area, 187-190 toluene benzylation, 177-178 Sulfur adsorbed, chemical state, 285-294 butane Formation, 289-293 butanethiol, 29 1-292 gaseous-phase adsorption, 289 high-molecular-weight compounds, 287, 289 hydrogen sulfide, 286-288 thiophene, 289-293 adsorption thermodynamics, 280-285 affinity for metals, 280-282 catalytic activity effect, 300-307

INDEX

activity ratio of fresh and sulfurized catalysts, 301-302 1,3-butadiene hydrogenation, 303, 305 1-butene, hydrogenation and isomerization, 303-306 chain length effect, 306 crotonic acid hydrogenation, 302 deactivation extent, 300-301 experimental conditions effects, 302 H2-D2equilibration reaction, 305-306 maleic acid hydrogenation, 302 Pd, 303 catalytic selectivity effect, 307-313 butanediene hydrogenation, 3 11-3 13 1-butene hydrogenation/isomerization, 309-3 10 diolefin hydrogenation, 309-31 1 double-bond shift, 309 electronic transfer induction, 308 isoprene hydrogenation, 313-314 2-methyl- I-butene hydrogenation/ isomerization, 309-3 10 ratio of adsorption equilibrium contents, 309 surface reconstruction, 308 chemisorbed, binding energy, 281 coverage, deducted from butane formation, 292-293 detoxication, 286 initial toxicity, 301, 306-307 interaction with metals, 282-285 metallic catalyst poisoning, 279-280 reactant adsorption, 294-300 cyclopentane hydrogenolysis, 298-299 formaldehyde adsorption and decomposition, 299-300 hydrogen adsorption capacity, 295296 hydrogen binding energy, 295-296 olefinic compounds, 296-300 on Pt(1 lo), 295 on Pt(lll), 295 partially sulfurized Pt, 296 sulfurization effects on Pt, 299 reversible and irreversible, 283-284 surface saturation state, 282-283 toxicity, 286, 308 Superacids, 165, see also Liquid superacids; Solid superacids Superharmonic entrainment, Pt, 238-240

Supported catalysts, electronic state of metallic particles, 285 Supported metals, 37-39, 46 Supported palladium, 77-78 chemical probing of surfaces, 82-91 B/L intensity ratio, 88, 90 H2 and CO chemisorption, 87-89 isomerization selectivity, 84-85, 87, 90-91

model of interactions between Pd and silica, 89, 91 neopentane isomerization, 83 route, 84, 86-87 X-ray diffraction spectrum, 87-88 evolution in course of reaction, 91-93 lattice energy, 80-81 presence of electron-deficient Pd, 81-82 structure sensitivity of reactions. 78-82 transformation into P-PdH phase, 79-80 Surface acidity, monolayer dispersion, 31, 33 Surface diffusion/reaction coupling, 249250, 272 Surface saturation state, sulfur, 282-283

T Tammann temperature, 16 Temporal self-organization, see Platinum Thermal annealing, platinum, 244 Thiophene adsorption, 289-290 dissociation, 291 dissociative chemisorption, 292-294 TiOz superacid, 182-184, 188, 190, 192195 TiOz-ZrOs superacid, preparation, 186 Ti(S04)2superacid, preparation, 192 Toluene, 39-40, 177-178 Transethylation, benzene, 173 Transition-metal surfaces C2H, transformations, 141- I47 HCOOH decomposition, 147- 15 1 Transmission electron microscopy, monolayer dispersion, 30-3 1 Triatomic adsorbates, dissociation and recombination barriers, 131-132 Trigger waves, 258

329

INDEX

U Unsupported palladium, catalysis over, 47-61, see also Alkanes CO hydrogenation, 60-61 Horiuti-Polanyi mechanism, 49 unsaturated hydrocarbons with HI, 4851

UV diffuse reflectance spectroscopy, monolayer dispersion, 26

W Work function, 217-218, 225-226, 263-264 W03/ZrOz,201-204

X X-ray diffraction C~CIzly-AlzO3,8-9 MoOJy-Al203, 4-5 Mo03/Zr02, 203-204 NaCI/NaY zeolite, 18 Ni/y-A1203, 37 patterns, 2-3, 10-11 Pd powder with silica gel, 87-88 quantitative phase analysis, monolayer dispersion, 5-6 salt/y-Alz03,8, 10 Ti02 superacid, 188, 190

WO3/ZrOI, 202-203 Zr02 superacid, 188-189 X-ray photoelectron spectroscopy calcined iron sulfates, 178-179 detectability, 46 monolayer dispersion, 19-22 small Pd particles, 62-64 TiOz superacid, 192, 194 ZrOz superacid, 192-193 p-Xylene, formation, 39

Z Zeolite, 17-21, 39-40 ZnO/SiOz, binding energy, 20-21 Zr02 superacid, see also S04/Zr02W03/ ZrOz acid strength, 187 hexane reaction, 200 IR spectra, 193, 195 isobutane alkylation, 197 preparation, 184-185 Pt loading, 201 SEM, 189, 191 XPS spectra, 192-193 XRD spectra, 188-189 Zr(OH)4, precalcination temperature, 184185 Zr(S04)2superacid, preparation, 192

This Page Intentionally Left Blank