The Chemical Physics of Solid Surfaces, Vol. 10: Surface alloys and alloy surfaces

THE CHEMICAL PHYSICS OF SOLID SURFACES T H E C H E M I C A L P H Y S I C S OF S O L I D S U R F A C E S Volume 1...

Author: D. P. Woodruff

44 downloads 588 Views 29MB 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

THE

CHEMICAL

PHYSICS

OF SOLID

SURFACES

T H E C H E M I C A L P H Y S I C S OF S O L I D S U R F A C E S

Volume 1 CLEAN SOLID SURFACES Volume 2 A D S O R P T I O N AT S O L I D S U R F A C E S Volume 3 CHEMISORPTION SYSTEMS Volume 4 F U N D A M E N T A L S T U D I E S OF H E T E R O G E N E O U S CATALYSIS Volume 5 S U R F A C E P R O P E R T I E S OF E L E C T R O N I C M A T E R I A L S Volume 6 COADSORPTION, PROMOTERS AND POISONS Volume 7 PHASE TRANSITIONS AND ADSORBATE R E S T R U C T U R I N G AT M E T A L S U R F A C E S Volume 8 G R O W T H A N D P R O P E R T I E S OF U L T R A T H I N E P I T A X I A L LAYERS Volume 9 OXIDE S U R F A C E S Volume 10 SURFACE ALLOYS AND ALLOY SURFACES

TH E CH EMICAL PHYSICS OF SOL! D SU RFACES

EDITED D.P.

BY

W O O D R U F F

B.Sc. (Bristol), Ph.D., D.Sc. (Warwick)

Professor of Physics, University of Warwick

VOLUME

I0

SU RFACE ALLOYS AN D ALLOY SU RFACES

2002

ELSEVIER AMSTERDAMSAN DIEGO

BOSTON - SAN

- LONDONFRANCISCO

NEW - SINGAPORE

YORK-

OXFORD - SYDNEY-

- PARIS TOKYO

ELSEVIER SCIENCE B.V. Sara Burgerhartstraat 25 P.O. Box 211,1000 AE Amsterdam, The Netherlands 92002 Elsevier Science B.V. All rights reserved. This work is protected under copyright by Elsevier Science, and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use. Permissions may be sought directly from Elsevier Science via their homepage (http://www.elsevier.com) by selecting 'Customer support' and then 'Permissions'. Alternatively you can send an e-mail to: [email protected], or fax to: (+44) 1865 853333. In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London Wl P 0LP, UK; phone: (+44) 207 631 5555; fax: (+44) 207 631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of Elsevier Science is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission ofthe Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier Science Global Rights Department, at the mail, fax and e-mail addresses noted above. Notice No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made.

First edition 2002 Library of Congress Cataloging in Publication Data A catalog record from the Library of Congress has been applied for.

British Libary Cataloguing in Publication Data A cataloque record from the British Library has been applied for.

ISBN 0-444-51152-0 (Vol. 10) ISBN 0-444-41971-3 (Series) O The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed in The Netherlands.

Contributors to Volume I0

D.A. ADAMS

Institute of Physics and Astronomy, University of Aarhus, DK-8000 Aarhus C, Denmark

J.N. ANDERSEN

Department of Synchrotron Radiation Research, Institute of Physics, Lund University, S-223 62 Lund, Sweden

C.J. BADDELEY

School of Chemistry, University of St Andrews, St Andrews, Fife KY 16 9ST, UK

U. BARDI

Dipartimento di Chimica, Universith di Firenze, Via G. Caponi 9, 50014 Firenze, Italy

C.J. BARNES

School of Chemical Sciences, Dublin City University, Dublin 9, Republic of Ireland

J.C. BERTOLINI

Insitut de Recherches sur la Catalyse- CNRS, 2, avenue Albert Einstein, F-69626 Villeurbanne Cedex, France

G. BOZZOLO

Ohio Aerospace Institute, 22800 Cedar Point Rd., Cleveland, OH 44142, USA and NASA Glenn Research Center, Cleveland, OH 44135, USA

J.E. GARCES

Ohio Aerospace Institute, 22800 Cedar Point Rd., Cleveland, OH 44142, USA and Centro Atomica Bariloche, 8400 Bariloche, Argentina

J. HRBEK

Department of Chemistry, Brookhaven National Laboratory, Upton, New York 11973, USA

Y. JUGNET

Insitut de Recherches sur la Catalyse- CNRS, 2, avenue Albert Einstein, F-69626 Villeurbanne Cedex, France

G.L. KELLOGG

Sandia National Laboratories, Albuquerque, NM 87185-1415, USA

M. POLAK

Department of Chemistry, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel

H. NIEHUS

Humboldt-Universit~t zu Berlin, Institut fur Physik, Oberfl~chenphysik und Atomsto6prozesse, InvalidenstralSe 110, D- 10115 Berlin, Germany

J.K. NORSKOV

Centre for Atomic-scale Materials Physics and Department of Physics, Technical University of Denmark, DK-2800 Lyngby, Denmark

J.A. RODRIGUEZ

Department of Chemistry, Brookhaven National Laboratory, Upton, New York 11973, USA

vii A.V. RUBAN


L. RUBINOVICH

Department of Chemistry, Ben-Gurion University of the Negev, Beer-Sheva 84105, Israel

M. SCHMID

Institut fur Allgemeine Physik, Technische Universit~it Wien, A- 1040 Wien, Austria

H.L. SKRIVER


S. SPELLER

Research Institute for Materials, University of Nijmegen, Toernooiveld 1, 6525 ED Nijmegen, The Netherlands

P. VARGA

Institut fur Allgemeine Physik, Technische Universit~it Wien, A- 1040 Wien, Austria

E. VLIEG

NSRIM Department of Solid State Chemistry, University of Nijmegen, Toernooiveld 1, 6525 ED Nijmegen, The Netherlands

D.P. WOODRUFF

Physics Department, University of Warwick, Coventry CV4 7AL, UK

...

Vlll

Preface During the late 1960s and 1970s the commercial availability of ultra-high vacuum (UHV) systems allowed the development of a plethora of new techniques which were devised to probe materials in a surface-specific fashion, and this in turn led to the creation of modem surface science; the study of the structural, electronic and chemical properties of extremely well-characterised surfaces on an atomic scale. When David King and I first conceived this series of volumes in the later 1970s our objective was to recognise the growing maturity of this new scientific discipline which was already starting to apply these techniques in a combined fashion to understand surface processes. In the 20 years since the first volume was published, this perception has certainly proved to be well-founded, and while new techniques have continued to appear, they have rapidly been assimilated into the general armoury of methods (the increasing pervasiveness of scanning probe microscopies is evident again in the current volume), and it is the combination of methods which has proved most effective. The topic of the present volume, Sufluce Alloys and Alloy Sufluces, provides new insights into a mixture of old and new problems. The surfaces of bulk alloys have long been known to be of practical interest for their chemical properties, be it novel activity or selectivity to certain reactions in a way which differs from the constituent elements in isolation or novel passiveness to corrosion. It has also long been known that the surface composition of such alloys commonly differs from that of the underlying bulk, and some of the basic thermodynamics of this segregation is far from new. Nevertheless, our understanding of these chemical and physical phenomena is far from complete, and the application of surface science methods to investigate these phenomena is a manifestation of a general trend to the study of surfaces of increasing complexity. A phenomenon which has been fully recognised far more recently is that of surface alloy formation - the intermixing of substrate atoms and adatoms in the outermost atomic layer, or few atomic layers of a solid, to form a stable ultra-thin alloy phase which may be in equilibrium with an essentially elementally pure substrate and may even involve the intermixing of elements which are immiscible in the bulk. There are now many examples of these surface alloys, and quite complex alloying and de-alloying behaviour may be observed as a function of surface stoichiometry. It is this combination of surface alloys and alloy surfaces which is addressed in the chapters of this volume. The first three chapters, by Ruban, Skriver and N~rskov,by Bozzolo and Garces, and by Polak and Rubinovich, are concerned with different theoretical descriptions of some of these phenomena from which one gains physical

1X

insight and predictive powers into the mixing, segregation and ordering phenomena. There follows a series of chapters based on experimental studies of surface composition, ordering and structure based on a variety of different materials and techniques. Schmid and Varga show, in particular, the remarkable power of scanning tunnelling microscopy, when chemical discrimination of the elemental components of an alloy surface is possible, to gain an atomic-scale understanding of some of the effects of segregation and ordering. Kellogg describes phenomena particularly in the Cu/Pb system based on information from many techniques but including the application of low energy electron microscopy to follow the processes of alloying and de-alloying. Speller and Bardi and Adams and Andersen describe the results of extensive structural studies of Pt-Sn alloys and Al-alkali surface alloy phases respectively. The latter systems, involving mixtures of superficially 'simple' metals, show a remarkably rich range of ordering and intermixing phenomena. Woodruff and Vlieg describe some detailed quantitative structural aspects of some metallic surface alloys including the systematics of surface alloy layer atomic rumpling and associated effective atomic radii, while Barnes surveys the structural aspects of surface alloys on Cu(100). Finally, in this group of chapters, Niehus discusses some results on the surface order of bulk alloys, especially Cu3Au and related systems, and ordered overlayers on these surfaces. The final group of chapters by Bertolini and Jugnet, Rogriguez, Hrbek and Baddeley address issues more directly related to the chemical properties of these surfaces, the first three of these chapters being concerned directly with the relationship of the nature of the surface alloys (and alloy surfaces) and their reactivity, while Baddeley turns the problem round in addressing the issue of adsorbate-induced modification of surface segregation; not only does the alloy surface modify the reactivity, but also the reaction modifies the surface alloy. March 2002

D.P.Woodruff

Contents Preface

viii

Chapter 1 (A.V. Ruban, H.L. Skriver and J. Norskov)

Local equilibrium properties of metallic surface alloys 1. Introduction 2. Surface energy 2.1 Monoatomic solids 2.2 Alloys 3. Stable surface alloy configurations 4. Generic classes of surface alloying 4.1 Mixing energy 4.2 Segregation energy 5. General trends for the surface mixing energies in transition metal alloys 6. General trends for the surface segregation energies in transition metal alloys 7. Island formation: multilayer versus monolayer growth 8. Bulk-type ordered surface alloys 9. Alternative ordered structures on the surface Acknowledgement References

1 2 2 5 7 8 9 10 11 13 15 19 23 27 28

Chapter 2 (G. Bozzolo and J.E. Garces)

Atomistic modelling of surface alloys 1. Introduction 2. The BFS method 2.1 Calculation of the BFS strain energy 2.2 Calculation of the BFS chemical energy 2.3 The BFS reference state in surface alloys vs. epitaxial growth 3. BFS modelling of surface alloys 3.1 Calculational procedure 3.2 Au/Ni(110) 3.3 Pd/Ni(110) 3.4 Pd/Cu(100) 3.5 Pd/Cu(110) 3.6 Cu/Pd(110) 3.7 Pt/Cu(100) 3.8 Au/Cu(100) and Au/Cu(110) 3.9 Cu/Ni(110) 3.10 (Cu, Au)/Ni(110) 4. Conclusions Acknowledgements References

30 36 39 44 45 47 48 51 60 62 68 68 68 72 78 79 82 83 83

Chapter 3 (M. Polak and L. Rubinovich) Alloy surface segregation and ordering phenomena: recent progress 1. Overview 2. Segregation in multi-element alloys 3. Surface segregation in ordered alloys 3.1 Prediction of order/segregation interplay by means of a simple model 3.1.1 Equiatomic binary alloys 3.1.2 Non-equiatomic binary alloys 3.2 Case studies 3.2.1 Compositional variations in Cu3Au(100) and CuaPd(100) 3.2.2 Surface order in PtsSn(111) and Co3Pt(111) 3.2.3 Segregation characteristics of aluminide surfaces 4. Segregation in a bi-phase binary alloy 5. Summary References

86 90 96 97 97 99 101 101 104 105 109 113 115

Chapter 4 (M. Schmid and P. Varga) Segregation and surface chemical ordering- an experimental view on the atomic scale 1. Introduction 2. Chemical discrimination on bimetallic surfaces with atomic resolution by STM 2.1 True topographic effect 2.2 Difference in local electronic density of states 2.3 Tip-surface interaction 3. Segregation on alloys- surface and subsurface composition 3.1 Segregation 3.2 Preferential sputtering and segregation in the altered layer 4. Chemical ordering of alloy surfaces 4.1 Bulk chemical order 4.2 Fundamentals of surface chemical order 4.3 Chemical order of close-packed alloy surfaces 4.4 fcc(100) surfaces 4.5 Site-specific segregation 5. Implications for adsorption on alloys 5.1 Chemical affinity 5.2 The ensemble effect 5.3 The ligand effect 6. Conclusions Acknowledgement References

118 120 121 123 125 127 127 128 130 130 131 134 140 141 144 144 145 147 148 149 149

xii

Chapter 5 (G.L. Kellogg) Surface alloying and de-alloying of Pb on single-crystal Cu surfaces 1. Introduction 2. Experimental and theoretical techniques 2.1 Experimental 2.2 Theoretical 3. Atomic structure, surface alloying and de-alloying 3.1 Pb on Cu(111) 3.2 Pb on Cu(100) 3.3 Pb on Cu(110) 3.4 Pb on stepped surfaces of Cu 3.5 Summary of Pb surface alloy and overlayer structures on single-crystal surfaces of Cu 4. Concluding remarks Acknowledgements References

152 154 154 157 158 158 165 172 175 178 178 180 180

Chapter 6 (S. Speller and U. Bardi) Surface alloys and alloy surfaces: the platinum-tin system 1. Introduction 2. Methods 3. The platinum-tin system 3.1 Low index surfaces of the Pt3Sn alloy 3.1.1 Pt3Sn(111) 3.1.2 Pt3 Sn(001) 3.1.3 Pt3Sn(110) 3.2 Surface alloys obtained depositing tin on platinum surfaces 3.2.1 Sn-Pt(111) 3.2.2 Sn-Pt(100) 4. Discussion 4.1 Surface atomic structure of the bulk Pt3Sn alloys 4.2 Defects and disorder on Pt3Sn alloy surfaces 4.3 Multilayer and single layer surface alloys 5. Conclusion Appendix: Notes on nomenclature References

184 185 190 191 191 197 202 207 207 209 210 212 215 217 219 220 221

Chapter 7 (D.L. Adams and J.N. Andersen) Alkali-aluminum surface alloys 1. Introduction 1.1 Background 1.2 Present work 2. Experimental methods 2.1 LEED measurements 2.2 LEED analysis 2.3 The surface structures of clean AI(111), (100) and (110)

225 225 226 228 228 228 229

xiii 2.4 Core-level measurements 3. Adsorption on AI(111) 3.1 AI(111)-(2x2)-Rb and Cs phases formed at 100 K 3.2 AI(111)-(~/3x~/3)R30~ Rb and Cs phases formed at 100 K 3.3 AI(111)-(4x4)-Na phase formed at 100 K 3.4 AI(111)-(~/3x~/3)R30~ Na, K and Rb phases formed at 300 K 3.5 AI(111)-(2~/3x2~/3)R30~ phase formed at 300 K 3.6 AI(111)-(2x2) -Na phase formed at 300 K 3.7 Ternary surface alloys formed by coadsorption on Na and K, Rb or Cs on AI(111) at 300K 4. Adsorption on AI(100) 4.1 AI(100)-(2x2) -Na phase formed at 100 K 4.2 Al(100)-(~/5x~5)R26.6~ phase formed at 240 K 4.3 AI(100)-c(2x2)-Li and Na phases formed at 300 K 4.4 Al(100)-c(2x2) -2Li phase formed at 400 K 5. Adsorption on AI(110) 5.1 AI(110)-c(2x2)-Li and Na phases formed at 300 K 5.2 AI(110)-(4xl)-3Na phase formed at 300 K 6. Phase transitions 6.1 AI(111)-(~/3xx/3)R30~ and Rb 6.2 AI(100)-c(2x2)-Na 6.3 AI(100)-(q5xq5)R26.6~ 7. The role of DFT calculations 7.1 AI( 111)-(q3xq3)R30~ and K 7.2 Al(111)-(2x2)-Na 7.3 Al(100)-(~/5x~/5)R26.6~ 7.4 AI(100)-c(2x2)-Na 7.5 Al(100)-c(2x2)-Li 7.6 AI(100)-c(2x2)-2Li 8. Summary and conclusions Acknowledgements References

229 233 234 235 235 237 240 241 243 245 246 246 247 248 253 253 254 257 258 261 262 264 264 266 267 268 268 269 270 273 273

Chapter 8 (D.P. Woodruff and E. Vlieg) The structure of surface alloy phases on metallic substrates 1. Introduction 2. Case studies 2.1 Cu(111)/Sb and Ag(111)/Sb: interracial stacking faults 2.2 Ni(111)/Pb: a case of strongly suppressed surface alloy rumpling 2.3 Mn and non-magnetic metals on Cu(100), Ni(100) and Pd(100): effect of local magnetic moments 2.4 Surface alloys formed by Sn on Cu, Ni, Pt and Rh surfaces: effect of changing substrate lattice parameter and surface orientation on rumpling amplitude 3. Interatomic distances in surface alloys 4. More complex systems 5. Conclusions References

277 278 278 286 288 291 293 298 301 302

xiv

Chapter 9 (C.J. Barnes) Surface alloy formation on Cu{100} 1. Introduction 2. Cu{ 100 }-c(2x2)-X (X=Au,Pd,Mn) surface alloys 2.1 Geometric and electronic structure 2.2 Growth mechanism of Cu{ 100}-c(2x2) surface alloys 3. Surface alloy formation upon Co, Fe and Ni adsorption 4. Surface alloy formation upon alkali and alkaline earth metal adsorption 4.1 The Cu{ 100}/Li surface alloy: the coverage dependent (2xl)---)(3x3)---~(4x4) transition 4.2 The Cu { 100 }-c(2x2)-Mg surface alloy 5. De-alloying transitions: adsorption of group IIIA, IVA and VA metals 5.1 The Cu { 100 }/Pb system 5.2 De-alloying in the Cu{ 100}/Bi system 5.3 Surface alloy formation in the Cu { 100 }/In and Sn systems 5.4 De-alloying transitions for transition metal adsorbates 6. Underlayer 2D alloys and overlayer to underlayer transitions 6.1 The Cu{ 100 }-c(2x2)-Pd overlayer to underlayer transition 6.2 Cu{ 100 }/Pt: the Cu{ 100 }-c(2x2)-Pt underlayer alloy 6.3 Cu{ 100 }/Ir: the unusual case of p(2xl) underlayer formation 7. Formation of ordered multilayer alloys 7.1 The Cu { 100 }-p(2x2)- 1 ML Pd surface alloy 7.2 The Cu{ 100 }-c(2x2)-Pt multilayer alloy 7.3 The Cu{ 100}-(4x2)-pgg-Mn structure 8. Conclusions Acknowledgements Reference

305 308 308 315 322 326 326 331 333 334 339 341 343 345 345 347 349 351 351 355 356 358 359 359

Chapter 10 (H. Niehus) Surface and sub-surface alloy formation connected with ordered superstructures

1. Introduction 2. Surfaces of ordered bulk alloys 2.1 Preparation dependent surface composition: NiAI 2.2 Surface properties of alloys with identical surface composition 2.2.1 Cu3Au(110) 2.2.2 Cu3Au(100) 3. Surface alloys of bulk immiscible constituents 3.1 Sub-surface alloy formation: iridium on Cu(100) 3.2 Intermixing versus phase separation: copper on Ir(100)-(5xl) 4. Alloy surfaces as substrates for ordered superstructures 4.1 Vanadium on Cu3Au(100) 4.2 Vanadium oxide on Cu3Au(100)-O 5. Summary Acknowledgement References

364 366 366 372 373 375 378 378 389 393 394 396 399 400 400

XV

Chapter 11 (J.C. Bertolini and Y. Jugnet) Surface structure and catalytic activity of palladium overlayers with 1,3butadiene hydrogenation 1. Introduction 2. Experimental approach 3. The 1,3-butadiene hydrogenation reaction 4. Surface and reactivity of Pd based alloy surfaces 4.1 General points 4.2 Surface composition and reactivity of Pd5Ni95 and Pd5Pt95 polycrystals 4.3 Influence of the surface orientation on reactivity 4.3.1 A solid solution in the whole range of composition: PdsNi92(111 ) and (110) 4.3.2 A system with a tendency to ordering: Pd50Cu50(111) and (110) 5. Surface and reactivity of Pd deposits 5.1 Pd in compression on Ni and Cu 5.1.1 Case thermodynamically favouring A on B: Pd on Ni Pd on Ni(111) Pd on Ni(110) 5.1.2 Case of A on B unfavourable: Pd on Cu(110) 5.2 Pd in tension on Au(110) 6. Summary and conclusion Acknowledgements References

404 407 409 413 413 414 418 418 421 423 423 423 423 424 428 430 433 434 435

Chapter 12 (J.A. Rodriguez) Electronic and chemical properties of palladium in bimetallic systems: how much do we know about heteronuclear metal-metal bonding? 1. Introduction 2. Photoemission studies 3. Thermal desorption studies 4. CO chemisorption studies 5. Models for bimetallic bonding 6. Theoretical studies 6.1 Charge redistribution in bimetallic bonding 6.2 Core-level and valence-band shifts 6.3 CO chemisorption 7. Conclusion Acknowledgement References

43 8 439 445 448 454 455 455 458 460 462 462 462

XV1

Chapter 13 (J.A. Rodriguez and J. Hrbek) Interaction of sulphur with bimetallic surfaces: effects of structural, electronic and chemical properties 1. Introduction 2. Repulsive interactions between gold and sulphur on transition metal surfaces 3. Interaction of sulphur with Ag/Ru(0001) and Cu/Ru(0001) 4. Admetal promoted sulphidation of Pt(111) and Mo(110) 5. Bimetallic bonding and the prevention of sulphur poisoning 6. Conclusion Acknowledgement References

466 467 475 482 488 492 492 492

Chapter 14 (C.J. Baddeley) Adsorbate induced segregation at bimetallic surfaces 1. Introduction 1.1 Bimetallic surface chemistry - traditional ideas 1.1.1 Ensemble effects 1.1.2 Electronic effects 2. Adsorbate induced segregation 2.1 Thermodynamic considerations 3. Techniques for characterising adsorbate induced segregation 3.1 Photoelectron spectroscopies 3.1.1 X-ray photoelectron spectrscopy (XPS) and Auger electron spectroscopy (AES) 3.1.2 Photoelectron microscopy (PEEM, SPEM) 3.2 Ion scattering spectroscopies 3.2.1 Low energy ion scattering (LEIS) 3.2.2 Medium energy ion scattering (MEIS) 3.3 X-ray absorption spectroscopies 3.3.1 Extended X-ray absorption fine structure (EXAFS) 3.4 Vibrational spectroscopies 3.4.1 Infra-red spectroscopy 3.5 Other techniques 3.5.1 Scanning tunnelling microscopy (STM) 3.5.2 Low energy electron diffraction (LEED) 3.5.3 Nuclear magnetic resonance (NMR) 4. Conclusions References

508 508 509 510 515 515 516 516 517 517 521 522 522 523

Index

527

495 495 497 499 500 500 505 505 505

9 2002 Elsevier Science B.V. All rights reserved.

Surface Alloys, and Alloy Surfaces D.P. Woodruff, (Editor)

Chapter 1

Local equilibrium properties of metallic surface alloys A. V. Ruban, H. L. Skriver, and J. K. NOrskov Center for Atomic-scale Materials Physics and Department of Physics, Technical University of Denmark, DK-2800 Lyngby, Denmark INTRODUCTION A great variety of structures are formed after deposition of one (or several) metals on the surface of another [1]. The deposited metals may form alloys with each other or they may form islands with some microstructure [7,8] with the substrate in the first or deeper layers [ 1-6]. Alloy formation at the surface may be observed even in those cases where there is phase separation in the bulk [9-11 ]. If the size mismatch between the deposited and substrate atoms is large, misfit dislocation structures may be formed [ 12-14]. A detailed theoretical prediction of such structures from very general considerations based on first-principles total energy calculations, is very demanding, since it includes the kinetics of the deposition, growth, and diffusion in the system under the relevant external conditions. Nevertheless, there are some surface alloys, the structures of which, although being metastable, mostly reflect the thermodynamics of the ground state of the system. This is so, since at ordinary temperatures the entropy driven diffusion of the deposited material into the bulk is very slow, and, hence, at time scales which are long in terms of surface kinetics, but short in terms of bulk diffusion, a local equilibrium may be established in the surface region [ 15,16] In such cases a local-equilibrium structure may be obtained theoretically by minimization of the free energy of the system under the constraint of a fixed alloy composition in the surface region [8,17-24]. Although this approach is very similar to the one used for bulk systems, it should be modified due to the specific features introduced by the surface. First of all, since the structure of the underlying bulk system is fixed, it acts as the source of an external field for the surface alloy, creating, for instance, epitaxial strain. Secondly, since the surface is an open system, it allows the formation of a great variety of different structures, which may not have any connection at all to the crystal structure of the substrate. Finally, the surface is a spatially inhomogeneous system, and thus different alloy components have their own

specific preference for different parts of the surface region, which will cause the segregation of alloy components to the various layers. The surface segregation phenomena play a major role in establishing the composition of the surface alloy in each layer, and therefore any thermodynamic-like consideration of the surface alloy formation should start by investigating the segregation behavior of the alloy components. This behavior is in fact naturally incorporated into the theory if, instead of the free energy of the surface region, the surface free energy is considered [15]. In general the surface energy is a complex function of the structure, composition, and alloy configuration in each layer of the surface region, and thus the optimization to find the equilibrium state should be made in the phase space of all these parameters. Moreover, for different amounts of the deposition element there may exist different equilibrium phases or mixtures of them. However, to categorize in a simple way the behavior of surface alloys the surface energy of a monolayer of a pseudomorphic random alloy of the deposited and substrate elements can be used. The main advantage of the surface energy curve of a random surface alloy is the fact that its general features can be described by only two physically well-determined parameters" the solution and segregation energies of the deposited element in the surface layer of the substrate. Four different combinations of theses two parameters lead to four generic cases of surface alloy behavior. Another advantage of this kind of theoretical consideration is the fact that such a surface energy curve, or the corresponding solution and segregation energies, may easily be obtained in first-principles or ab initio calculations based on density-functional theory [15,24,25] using only the atomic numbers of the alloy components and the crystal structure of the bulk as input parameters, which gives a reliable first insight into possible surface alloy behavior. In this chapter we discuss the general trends for the formation of transition metal surface alloys obtained by first-principles calculations [15,24,25]. We also present some examples where the behavior of the surface alloy appears to be more complicated than expected from the simple considerations based on the knowledge of the surface energy curve of a monolayer of a random alloy and the bulk phase diagram. In all cases considered here we assume that there is no exchange of atoms through the vacuum region due to either evaporation or condensation, since such processes do not affect mass transfer towards the surface region in most metallic systems at ordinary temperatures, where a surface alloy may exist for a sufficiently long time.

2. S U R F A C E E N E R G Y 2.1.Monoatomic solids The surface energy is the energy required to create one unit of surface area. Such a process is schematically illustrated in Fig.l, and its energy is the difference between the total energies of system 2, in which the additional surface area A has been created, and system 1, which is the initial state. Thus the surface energy, u , is 1

2

Y--"~(Eto,-

El

tot)

9

(1)

Here, the surface energy is determined per surface area, and the total energies, i Eto ~ , correspond to the complete systems (extensive quantities). In theoretical calculations another but equivalent definition of the surface energy is usually used, i.e., 1 (E~ury Et, Ulk) Y=-, ,o, - ,o,, 9

(2)

ns

Here, E ~ur tot is the total energy of the surface region, which usually consists of several layers the crystal and electronic structures of which are different from ~ is the total energy of a bulk region equivalent in their bulk counterparts, E tot size to the surface region, and n, the number of atoms at the surface. Thus, the surface energy in (2) is normalized per number of atoms at the surface.

Fig. 1. The surface energy of a monoatomic solid. A is a new surface created due to a change of the form of the crystal under the conservation of the number of atoms. Light grey color schematically indicates the surface region.

Although it is very difficult to measure surface energies, they may today relatively easily and reliably be calculated from first-principles [26,27], even in the cases of quite open surfaces [28]. In Fig. 2 we show the surface energies of metals in the 4d-series of the Periodic Table obtained by first-principles calculations [27]. The energies in the figure exhibit a parabolic-like behavior as a function of the atomic number. Such a behavior is explained in terms of the occupation of the valence d-band by the Friedel model [29,30] in which the surface energies follow the same trend as the corresponding cohesive energies and can be estimated from

11 1

WNd(IO-N a)

u = 2"-0- 1 -

(3)

,

where W is the width of the d-band, N d the number of valence d-electrons, and Zsand zb are the coordination numbers of the atoms at the surface and in the bulk, respectively. It follows from (3) that the transition metals with a half-occupied d-band have the highest surface energies, the magnitude of which increases down the Periodic Table from the 3d to the 5d metals due to a corresponding increase in the d-band width [31 ]. Formula (3) also shows the dependence of the surface

l::: 1.5

~ v

I

>, 1.0 (D (--

(D

o 0.5 '1:: r 0.0

hcp bcc bcc hc:p hcp fcc

Rb Sr

Y

fcc

fcc hcp

Zr Nb Mo Tc Ru Rh Pd A cl Cd

Fig. 2. The energies of the most closed-packed surfaces of the metals in the 4d transition series obtained from first-principles [27].

energy of a transition metal on the surface coordination number z~. With decreasing z,, or with increasing number of "broken" bonds, the surface energy increases, and thus the surface energy of open surfaces may be quite large. The later is a consequence of the localized bonding provided by valence d-states. In contrast, for the "simple" metals the free-electron like contribution to the bonding dominates, making their surface energies much less sensitive to the surface orientation. 2.2. Alloys. In the case of alloys the composition of the surface region may differ from that of the bulk and therefore (2) should be modified to take into account the energy of an exchange of atoms between the bulk and the surface region as sketched in Fig. 3. As usual, we assume that the bulk is infinitely large compared to the surface region and therefore such an exchange does not influence the composition of the bulk. Let us consider a binary A~_cBc alloy. The energy of removing a B atom is minus the chemical potential of the B-component, -/~ B 9At T = 0 K, when there is no contribution from the entropy term, E to t

-UB = - ' - - - - ~

,

(4)

ON B

where NB is the number of B atoms in the bulk. Thus, the surface energy of an alloy is

Fig. 3. Schematic exchange of A and B atoms between the surface region and the bulk.

1

=

--( ns

surl Etot

bulk -Etot

-

Z i = A B, I.IiA N i )

9

(5)

As in the case of a monoatomic solid E `ury is the total energy of the surface tot region having a given composition and configuration, A N i the number of A and B atoms which have been exchanged between the surface region and the bulk. For a binary A,_~Br alloy A N A = - A N B if no vacancies are formed in the surface region, and thus using the concentration variable, c =c B = N B I N (N = NA + NB), (5) can be rewritten in the form: 1

surl

}" = --- ( E tot t/s

bulk

- -

E,o , ) - n l u A c

9

Here, n I is the number of layers in the surface region, A c = c , - c

(6)

the

difference between the concentration c, in the surface region and the concentration in the bulk, and /~=/2 B--/./A the effective chemical potential of the bulk alloy, which may be determined by ,,-.,(O)-bulk

U =

0 lgL-tot Oc

Ebulk ---

1

0 __tot

,

~

(7)

NOc

where the first energy is per atom. At non-zero temperatures one should instead of the total energy of the system consider its free energy by adding the corresponding entropy contribution - T S . In general, it is a quite complicated problem to obtain the chemical potential since the concentration derivative should be taken along the minimal path in the phase space of short and long-range order and other parameters which define the equilibrium alloy configuration and structure at each concentration. However, this problem is greatly simplified in the case of a dilute alloy, where all the configurational effects become negligible, because to lowest order they are proportional to c 2. In this case, which in fact corresponds exactly to the deposition of one element (B) on the surface of another (A), the effective chemical potential is defined as u =

cgE(t~l-bulk(Al_cnc ) Oc

,

(8)

where E(~ l B ) is the total energy (per atom) of a random A~_cB~ alloy and the derivative is taken at c = O. tot

-

c

3. S T A B L E S U R F A C E A L L O Y C O N F I G U R A T I O N S A small amount of material deposited on a pure host crystal will always be metastable at non-zero temperatures, since the gain in entropy by dissolving into the bulk, which is roughly A S = k l n ( N b l N s ) , where Nb and N~are the number of sites in the bulk and at the surface, respectively, will drive the deposited material away from the surface. However, as has already been mentioned, near room temperature bulk diffusion in a metal is extremely slow, and a local equilibrium is usually established in the surface region. The local equilibrium surface alloy configuration and structure may be found by minimization of the surface free energy, or if several different phases may exist, by finding a convex hull of the lowest free energies of different phases at different alloy compositions (at T=0), or more generally by a common-tangent construction which is completely analogous to the usual treatment of the bulk systems. The procedure is illustrated in Fig. 4. Given the surface energy curve in Fig. 4, the surface alloy with an overall concentration Co of atoms deposited at the surface will, instead of forming a homogeneous solution, H, separate into two distinct phases, say S and P, with concentration Cs and Cp, respectively, if Cs < C o < C e . The relative fraction of the S and P phases is determined by the lever rule as c e - c 0 to C o - C s , which also implies that the energy of the phase equilibrium of S and P will be a straight line between the points S and P in the surface energy diagram.

$

H

p

S 't=

I I I I I I !

0

Cs

Co

c

Cp

1

Fig. 4. Sketch of a common-tangent construction for the surface energy of an alloy, c is the coverage of the deposited material.

Such a phase diagram has, for instance, recently been calculated for a Mn/Cu(111) surface alloy [20]. Although these calculations include only the simplest alloy configurations in the limit of an infinitely large pseudomorphic surface it gives a better understanding of the initial stages of surface alloy formation during deposition growth of Mn on Cu(111), and, in particular, the formation of islands of a ~ x ~ Cu2Mn ordered alloy. 4. G E N E R I C CLASSES O F S U R F A C E A L L O Y I N G To categorize in a simple way the behavior of surface alloys we will use the so-called surface energy curve which is the surface energy of a pseudomorphic monolayer of a random AcB~_c alloy on the surface of B Although such a surface alloy is almost never realized in practice, it is quite useful in theoretical considerations. First of all, the surface energy of such an alloy may easily, and quite accurately, be determined by first-principles calculations [15,24]. Secondly, it allows one to categorize the deposition behavior in a simple way, and to predict some general features of real surface alloys. In Fig. 5 we show the surface energy (per substrate atom) for four different 0.8

AgcPt,_JPt(111)

Ag~Cul_JCu(100) !

i

i

i

I

!

I

I

I

0.9 0.7

0.7

0.6

0.5

E

o

0.5 0.00

0.25 0.50

0.75

1.00

0.3

0.00

0.25 0.50 0.75 1.00

>, t._

(i) t(D o 't:: :D r

1.4

RucAul_JAu(111 )

Pt~Cu,JCu(111 ) i

i

i

0.9

1.2

0.8

1.0

0.7

0.8 0.6

0.( 10 0.25

0.50

0.75

1.00

0.6

0.00 0.25 0.50 0.75 1.00

C

Fig. 5. Surface energy curves for a monolayer of a random alloy on surfaces of pure metals.

systems obtained by first-principles calculations [15]. These surface energy curves naturally fall into four distinct generic classes which may clearly be recognized by their curvatures and slopes. For instance, the surface energy curve of Agr has a positive curvature and negative slope, while the surface energy of Ru~Au~_c/Au(lll) exhibits negative curvature and positive slope. In this section we show that these two features of the surface energy curve in fact correspond to the mixing and segregation energies of the deposited element in the surface of a substrate. These energies may easily be obtained by first-principles calculations and thus the general trend of the surface alloying can be established.

4.1. Mixing energy The surface alloy mixing energy is determined similarly to case of bulk alloys as A BI

Cmi x "- y

~

/B

-~

AIB

--C y

A B~ IB

where

y~

-~

--(l--c)

y

B

,

(9)

is the surface energy of a monolayer of a random AcB~-c

alloy on a B substrate,

yA/B the surface energy of an infinite pseudomorphic

monolayer of A on B, and yB the surface energy of B. Thus, the straight line which connects yA/B and yB in Fig. 5 represents the energy of the standard state, which is a mixture of infinitely large islands of B and A on B, given by the last two terms in (9). It is obvious from the consideration in the A B~ IB previous section that if the surface energy curve y ~ -~ goes above the standard line there should be a phase separation of the surface alloy into islands of pure B and A elements in the surface layer. In contrast, if the surface energy curve goes below the standard state line, then alloying will occur on the surface. Since, the surface energy curve is obtained for a pseudomorphic alloy on a fixed lattice of the substrate, its behavior can be related directly to the type of so-called effective interactions which are responsible for the ordering of A and B atoms on the surface. That is, if the multisite interactions are small in the system, which is usually the case for metallic alloys on a fixed lattice, the mixing energy can be written in terms of pair potentials between alloy components, v~AA ,viAB ' and v i BB for each coordination shell i at the surface as

1

Emix

=---c(1-c) 2

where

Zi

E

1

i

z (v~A+v88--2vAB)=----C(1--C) i i 2

E i

ziV

i

'

(lo)

is the coordination number of the i-th coordination shell at the surface

10

and Vi the so-called effective interactions. Since the nearest-neighbor interactions are usually the strongest, the mixing energy is roughly proportional to minus the effective interaction at the first coordination shell. Thus, if the mixing energy is negative, i.e. the surface random alloy is stable against separation into islands of pure A and B elements, the effective interaction at the first coordination shell is positive, which means that a surface alloy has a tendency towards ordering. Such an ordering usually takes place at low temperatures. This is indeed the case for the two systems, presented in Fig. 5: AgcCUl_c on Cu(100) [12,15] and Pt~CUl~on Cu(111) [ 15], while the deposition of Ag on Pt(111) and Ru on Au(111) should lead to the formation of islands of the deposited element and the substrate. Although there appears to be no experimental data for the Ru/Au(111) system, the surface alloy structures of the Ag/Pt(111) have been thoroughly investigated experimentally [32-35], and island formation is wellestablished. In fact Ag islands have a finite size and they may form different, droplet- or stripe-like, structures, exhibiting a quite fascinating behavior with temperature, which unfortunately is beyond the scope of the present considerations. 4.2. S e g r e g a t i o n energy

Another distinctive feature of the surface energy curve is its slope. In fact the slope of yAB,_/B is simply the segregation energy of the deposited element to the surface layer at a given concentration: ~r,,ABj_ /B e segr = ~ C

,

(11)

which is the energy of transfering an atom of the deposited element from the bulk to the surface. On the other hand, esegr'--l.ls--12 i.e., the segregation energy is equal to the difference of the effective chemical potentials in the surface layer and in the bulk, where the chemical potentials in the surface layer are defined by (at T=0): ,

Os=

0 Eto,urr t ( A c B l _ C/B)

Oc

.

(12)

If the segregation energy is negative, as in the case of Ag on Pt(111), the deposited element stays at the surface. If the segregation energy is positive, as in the case of Pt on Cu(111) and Ru on Au(111), the deposited element should go into the deeper layers of the surface region (if the transfer of deposited element into the bulk is kinetically hindered). Usually, the deposited element

11

appears to be capped by a monolayer of the substrate, which is a process that may be observed in deposition experiments due to the quite fast diffusion of atoms between the surface and subsurface layers. The experimental data, a discussion of which may be found in [15], confirm the above mentioned general features of the formation of surface alloys. In general the surface segregation energy is different for different alloy concentrations of the deposited element (as one may see it even changes sign in the case of Ag on Cu(100)). Such a change in the surface segregation energy is in fact related to the alloying behavior, presented by the mixing energy, and therefore, the surface segregation can be characterized by a single parameter, which is the initial slope of the surface energy curve of the segregation energy of a single impurity of the element deposited at the surface of a substrate. 5. GENERAL TRENDS FOR THE SURFACE MIXING ENERGIES IN TRANSITION METAL ALLOYS In Table 1 we present the sign of the curvature of the surface energy curve calculated from first-principles [ 15] for the closed-packed surfaces of the 4d and 5d metals (fcc(111): Rh, Pd, Ag, Ir, Pt, Au; bcc(110): Nb, Mo, Ta, W; hcp(0001): Tc, Ru, Re, Os). Since the sign of the mixing energy is opposite to that of the curvature, a "+" in the table means a negative mixing energy or alloy formation, and a "-" means that alloying of the deposited element in the surface layer of the substrate is energetically unfavorable against island formation. It is clearly seen that 4d-4d, 4d-5d, 5d-4d and 5d-5d combinations exhibit similar patterns of "+" and "-" signs. This is because the bonding in t~ansition metals as well as in transition metal alloys is mainly determined by the valence d-electrons [29,31 ], which form quite localized bonds in contrast to the free-electron like bonding found in the simple metals. As a result the d band occupation is the main parameter for the characterization of the bonding in this case. In general, alloying in the surface alloy cases follows the trends observed in the corresponding bulk systems [31,36]. However, there are exceptions due to several factors. One of these is the crystal structure of the host (or substrate), which may play crucial role in the alloying [37], especially when the substrate is an earlier transition metal. This is, for instance, reflected in the asymmetry of the alloying behavior of A-B and B - A systems (see, for instance, W - M e and M e - W or Ta-Me and Me-Ta). Another factor which may change the alloying at the surface is the epitaxial strain of the surface alloy due to its pseudomorphic attachment to the substrate. This concerns especially systems with elements that differ considerably in size

12

where

the

resulting

epitaxial

surface more favorable of Au on Ni(110) reconstruction

[9]. H o w e v e r ,

energy

usually

makes

alloying

the pseudomorphic

at t h e

if t h e e p i t a x i a l s t r a i n is r e l i e v e d b y a s t r u c t u r a l

of the surface layer, the alloying may disappear,

the case of Ag growth hexagonal

strain

[ 3 8 ] . T h i s is t h e c a s e , f o r i n s t a n c e , i n t h e i n i t i a l g r o w s

on Cu(111)

surface

as observed

[ 12,15] for a higher coverage

alloy separates

into islands

of Cu

in

of Ag. Here, and

a c(2xl)

Ag phase.

Table 1 T h e sign of the curvature of the surface energy curve: "+" corresponds to surface alloy formation, "-" to island formation, and . .=. . to zero curvature. C o l u m n s are labelled by the deposited element and rows by the substrate.

Zr Zr Nb

+

Nb

Mo

Tc

Ru

Rh

Pd

Ag

Hf

Ta

W

Re

Os

Ir

Pt

An

+

+

+

+

+

+

+

+

+

+

+

+

+

+

4-

+

+

+

+

-

-

+

+

+

+

+

+

+

+

+

=

.

+

+

+

-

+

+

+

+

+

+

-

-

+

+

-

+

+

+

=

-

-

+

-

+

+

+

+

-

-

-

4-

+

+ +

Mo

+

+

Tc

+

+

+

.

Ru

+

+

+

.

Rh

+

+

+

+

-

Pd

+

+

+

+

-

Ag

+

.

Hf

=

+

+

+

+

Wa

-

-

+

+

.

.

.

.

=

W

+

+

+

+

.

.

.

.

+

+

Re

+

+

+

.

+

+

+

.

.

.

.

.

.

.

.

.

.

.

. .

-

.

.

.

+

+

+

-

.

-

.

.

+

.

+

+

+

+

+

+

+

+

+

+

+

+

+

-

.

.

. +

+

+

+

+

.

+

+

+

+

+

+

+

+

=

-

-

+

+

+

+

=

Pt

+

+

+

+

+

+

+

-

+

+

+

+

-

Au

+

.

+

+

+

+

.

.

The surface orientation also plays a very important formation,

since

corresponding

cases,

is

coordination

cases, especially periodic

alloying when

determined

numbers,

the substrate

belongs

surface

energy

curve

will

not

.

.

+

interactions

are surface

the may

simple

in Fig. 4 and

described

above. To have even a qualitative understanding

and

the

specific. In some

to the IVb-VIIIb

have

.

-

play an important

behavior

+

role in the surface alloy

presented

need further input.

the alloying

may

.

effective

z~, w h i c h

table, the multisite interactions

the

by

.

.

+

.

.

.

Os

.

.

.

.

Ir

.

.

.

.

group

parabolic

be more

in the

role. In those complex

shape than

one may therefore

13

6. GENERAL TRENDS FOR SURFACE SEGREGATION ENERGIES IN TRANSITION M E T A L ALLOYS In most cases the experimental techniques used to study surface phenomena do not seem to yield consistent values for the surface segregation energies. One important exception is the special case of an atom of atomic number Z+I in a host of atoms of atomic number Z, where the surface segregation energy may in fact be extracted with a high degree of accuracy from X-ray photoemission spectroscopy (XPS) measurements of surface core-level shifts (SCLS) [39]. In contrast they may be calculated quite accurately by modern first-principles methods [ 18,25,40]. In Fig. 6 we show the results of such calculations for transition metal alloys using grey scales for presenting absolute values of the surface segregation energies. To differentiate their sign we use filled circles for the negative segregation energies which correspond to segregation of the impurity (solute or deposited element) to the surface of the host (substrate). If the segregation energy is positive, the impurity prefers to be in the bulk or to be covered by the host. The actual values of the surface segregation energies can be found in Ref. 25. Similar to the case of the mixing energies in Table 1 one observes a pattern of segregation behavior which repeats itself for each combination of transition series. Again, the main feature of this pattern, an hourglass shape formed by the elements of the matrix which correspond to the negative surface segregation energy, is a consequence of the bonding along each transition metal series. The surface segregation energy is roughly proportional to the difference of the surface energies of the alloy components (under the condition, that they are determined for the given surface, structure, and lattice spacing of the alloy or of the host, in the case of an impurity). The main contribution to the surface energy in transition metal alloys is due to the bonds broken by the surface, and the energy involved is a parabolic function of the number of d-electrons as given by the Friedel model [25]. Thus, the surface segregation energy of a d metal impurity in another d-metal host may be estimated from: ._

E segr A~B

Here,

WB

O[WBNaB(IO--NB)--WA_~BNA(IO--NaA)]. d

d

(13)

and WA~ B are the d-band widths of the host (B) and the impurity

(A) in the host, respectively,

N di

the number of d-electrons in the host and

the impurity, and 0 = 0 . 0 5 [ 1-~/z/zb] , where zs and zb are the coordination numbers at the surface and in the bulk, respectively. The dependence on the surface coordination number means that the segregation energy in transition

14 metal alloys may increase dramatically for more open surfaces.

Fig. 6. Surface segregation energies of transition metal impurities (solute) for the closedpacked surfaces of transition metal hosts.

15

The deviations from the hourglass behavior predicted by the Friedel model (13) are due to the crystal structure effects, which originates from the local character of the interatomic bonding and its dependence on the number of valence d-electrons. The later determines the sequence of crystal structures which is the same along each transition metal series, except for the 3d transition metals where magnetic effects occur. Although the structural energy difference (bcc-fcc or bcc-hcp) in the pure transition metals is of the order of 0.2-0.4 eV, the structural energy difference in the segregation energies in some cases reach 1 eV [25], which makes it a very important parameter in the general analysis. 7. ISLAND F O R M A T I O N : GROWTH

MULTILAYER

VERSUS MONOLAYER

In this and the next section we will consider several examples which illustrate the application of the stability analysis based on the surface energy curve. We will start with the deposition of an element which does not form alloys (at low temperatures) with the substrate in the bulk and on the surface. Hence, there should be a formation of islands of the pure, deposited element incorporated in the surface of the substrate. Such structures, for instance, are usually formed during epitaxial growth of Co, Fe, and Cr on different surfaces of Cu: Systems which are well studied experimentally (see, for instance, [22,41-45], and references therein). Since the behavior of all the above mentioned systems is similar, we will consider here the growth of Co on Cu(111) during a submonolayer deposition. This case has been investigated thoroughly and it is found that the deposition of Co at low temperature (150K) leads to the growth of three-layer islands of Co with one subsurface layer, which at room temperature transform into twolayer islands of Co capped by one layer of Cu [22]. Note, that in contrast to the case considered here, most experimental investigations, e.g., Fe on Cu, have been carried out because of the interest in magnetic multilayers and therefore the amount of deposited element has usually been quite large. However, our main interest is the equilibrium structures formed during the initial epitaxial growth with up to one monolayer of the deposited element. Let us first mention, that the formation of islands of pure Co on Cu is an obvious consequence of the bulk phase diagram [36]: Co and Cu do not form alloys up to the melting temperature. Further, the size mismatch of Co and Cu is very small, and thus the alloying behavior will not be altered at the surface. The capping of Co islands by Cu is explained on the basis of surface segregation arguments knowing the fact that the surface energy of Cu is less than that of Co. Therefore it is no surprise that the surface energy curve for a monolayer of a random Co~Cu~_r alloy, obtained in first-principles calculations

16

[22] has the form shown in Fig. 7. The initial slope of the surface energy curve (at small concentrations of Co) in the figure indicates that the surface segregation energy is positive (it is equal 0.33 eV, for Cu(111) surface [25]). Hence, the Co islands which will form during epitaxial growth will be capped by Cu atoms, if diffusion at the surface is sufficiently fast. It is important to notice that the fact that the surface energy decreases when the Co coverage exceeds about a quarter of monolayer, is a consequence of the phase separation of Co and Cu in the bulk and does not mean that Co islands with no Cu on top will be stable (locally) at the surface. Let us demonstrate how the growth mode can be understood and obtained from the surface energy curve. To do so, one needs the surface energy of n layer (pseudomorphic) Co-structures on Cu(111) as a function of n shown in Fig. 8 [22]. In the limit n ~ oo the surface energy of Co,/Cu(111) is

YC~ Here

=

u

]/Co(lll)+yiCtolCu(lll) n E co-~cu sol

and

u

(14)

"

are the surface energies of fcc(111) Co and Cu,

ColCu(lll)

respectively ' u the Co/Cu(111) interface energy, and b u l k solution energy of Co in Cu.

COxCU~_x/Cu(111) ......

"

I

'"

I

"

I

'

I

"

> o.7 E~ L_

E

e

9 0.6

I:= .,

.5

o.o

i

i

i

o.4

i

Fig. 7. Surface energy of a monolayer CocCu~-c on Cu(111).

i

o18

1.0

E sol c~

the

17 It follows from (14) that the generally negative slope of the surface energy curve is due to the positive solution energy of Co in Cu, and it simply reflects the fact that the formation of Co islands is an energetically favorable process, since it "removes" Co from the bulk. One may analyze the stability of an n layer island against separation into islands of different heights by a common tangent construction, or in this particular case simply by the convex hull of the lowest surface energy points. The procedure is shown in Fig. 8. First we draw a line from the point n=0 (the surface energy of Cu(111)) to the surface energy of Co2/Cu(111). This line is below the surface energy of Co~/Cu(lll), and thus monolayer islands of Co are unstable against separation into a pure Cu surface and two-layer islands of Co. If we neglect the effect of island boundaries, the system should gain 0.39 eV per surface atom as a result of such a separation. We can continue this process and find that two-layer islands are unstable against separation into a pure Cu surface and three-layer islands, and so on. However, the energy of the separation is reduced for every step: In the case of the separation of Co2/Cu(111) into pure Cu(111) and Co3/Cu(111) it is only 0.11 eV as shown in Fig. 8. For large n the gain in energy due to the separation into a pure Cu surface and n+ l-layer island is Co / Cu( ~ ~~)

AE,=y

1

------y n+l

1.0

Cu( ~~ ~)

n

--------y n+l

!

!

AE 1 = 0.39 eV

v

0.0

.

--~.~..

"-"Z,-'~ CoJCu(111) ""-~',~ AE 2 = 0.11 eV

e--

'1=

(15)

,

(111)

>

o

Co.+,/ Cu( ~ ~ ~)

-1.0 COxOU,_,/Cu (111 ) Con/Cu (111 )

00 -2.0

I

I

I

1

2

3

Number of layers (n)

4

Fig. 8. The surface energy of Co,/Cu(111) as a function of the number of Co layers. Broken lines correspond to the energy of a mixture of the those structures which they connect.

18

which using (14) can be rewritten as AE n m 1 AEo= 1 ( yCO(lll) _ yCu(ll])+ Y i n tCo/Cu(lll) ) er n+l n+l

(16)

"

Hence, if A E0>0 , the deposited material should constantly undergo "island" separation, during which low islands transform into higher islands and clean surface areas. This is exactly the Volmer-Weber epitaxial growth mode and since the condition is satisfied for the Co/Cu(111) system multilayer epitaxial growth is energetically favorable. One can also see from (16) that the energy gain due to an increasing height of the islands reduces quite fast when n is small. If one includes the effect of the step-edges and the additional microfacets created by the formation of multilayer islands the energy will quite fast become positive. A simple estimate of the neglected effects allows one to explain the stability three-layer islands of Co on Cu during the initial deposition at low temperatures [22]. Next, we consider the capping of Co islands and find the equilibrium height of the capping slab. In Fig. 9 we show the calculated surface energies of

,0

c

C ...... "

>

0.5 -

~

0.0

]

u(111) -.N

C~"%.

co

Cu/Co/

CuZCo/

CuZCo/

]

C o e ' ~ . ...... "\,Cu/Coe' Cue'CoJ . \ " .....- ~ ...................L~...................,'

a~o -0.5 t1:l

o .

o9

.

.

.ocolCu(!! ) .

.

/~ ......- A c u j c o j c u ( 1 1 1 )

_\_ I

..A..

0

1 Number

2

3

of

4

layers (n+m)

5

Fig. 9. The surface energy of different multilayer CUmCOn overlayers on Cu(lll) as a function of n + m . The circles mark the surface energy of Co,/Cu(lll) and the triangles correspond to an additional capping of Co layers by Cu. The dashed-doted line shows the energy of a mixture of a clean Cu surface and two Co layers capped by a Cu monolayer.

19 capped overlayers of Co on Cu(111). First, one can again observe that any, e.g., Cu/Co~/Cu(111), structure is unstable against separation into a pure Cu surface and Cu/Con§ islands. This is schematically shown for Cu/Co/Cu(111), the energy of which is above dot-dashed line, representing the mixture of pure Cu surface and Cu/Co2/islands, by about 0.18 eV. But again, this energy drops quite fast for small n, and for the next island separation of Cu/Co2/Cu(111) into pure Cu(111) and Cu/Co3/Cu(111) it is about 0.05 eV only. In fact, the gain in energy due to such an island separation _ Co/Cu(lll) for Cum/Con/islands is equal to 2/(n + 1 ) Y i n t e r for large n. The interface energy in the case of a phase separated system is usually positive (proportional to the mixing energy) and thus this result simply reflects the ordinary phase separation in the bulk. Now, following the change of the surface energy of Co islands of a fixed height, one finds that there is a substantial gain in energy when Co layers become capped by the a single layer of Cu. In fact, this energy independently of the height of the Co layers is about 0.3 eV, which is simply the value of the segregation energy. A further increase in the of height of the Cu cap does not lead to a corresponding gain in energy, and thus, the capping stops (when the height of the Co islands is greater than one, a one-layer capped configuration is in fact the most stable configuration, although the energy difference between one-layer and multilayer capped configurations is very small). In this section we have considered examples of systems where the alloy behavior on the surface remains the same as in the bulk. As has been mentioned this is basically due to the fact that the size of the alloy components is practically the same. On the other hand, it is now well-known that alloying behavior on surfaces may change due epitaxial strain of the surface alloy [38] when the alloy components have different sizes. Such an alloying in this case is simply a consequence of the release of the epitaxial strain energy, which is positive and reaches its maximal value for an overlayer of a pure deposited element on the substrate. 8. B U L K - T Y P E ORDERED SURFACE ALLOYS A very good initial guess at the structure of a surface alloy may actually be obtained from the bulk phase diagram for the deposited element-substrate system. This is so, simply because, if there are no specific surface effects, the observed structures would have to be those found in the bulk phase diagram. Since the concentration of the deposited element should be considered small (it is actually "almost" zero, but in the case of local equilibria only the substrate atoms close to the surface may participate in the alloy formation, and thus the "effective" concentration of the deposited element could be quite high), the surface alloy will usually have the structure of the first ordered phase in the

20 substrate-rich part of the phase diagram. This kind of surface ordered alloy would be trivial, if the surface did not add some specific features. The simplest surface specific feature of an ordered phase is the fact that there usually are different truncations of the bulk ordered alloy by the same surface orientation. In this case the problem is to find the stable truncation which, as we will show in this section, is usually directly related to the surface segregation energy of the deposited element to the corresponding surface of the substrate. Let us consider the deposition of A1 on a (110) surface of Ni. According to the bulk phase diagram, the addition of A1 to Ni in the limit T=0K must lead to the formation of Ni3A1 in pure Ni. Therefore, the surface alloy formed during such a deposition may have a structure which corresponds to Ni3AI(ll0). Ni3A1 has the L12 structure, and therefore two different truncations are possible for the (110) surface as shown in Fig. 10: The ordered phase can be truncated either by a layer of pure Ni or by an ordered p(2xl)-NiA1 layer, which alternate in the [ 110] direction of ordered Ni3A1.

Fig. 10. Two different truncations of the A3B-LI2(ll0) surface: A pure A layer or an equiatomic p(2xl)-AB layer.

21 The segregation energy of A1 into the first layer of a Ni(110) surface or the surface energy curve can be calculated using first-principles methods [24]. One finds that the energy of segregation to the first layer is approximately -0.3 eV while the energy of segregation to the second and deeper layers is almost zero [24]. This is clearly seen from the initial slope of the surface energy curve of random AlcNi~_c alloys in the first (surface) and in the second (subsurface) layers shown in Fig. 11. In this figure the two squares at c=0.5 correspond to the two different possible truncations of Ni3AI(ll0): a monolayer of p ( 2 x l ) - N i A 1 ordered layer on the surface and a monolayer of p ( 2 x l ) - N i A 1 ordered layer on the surface but capped by Ni atoms. From this result it is clear that the NiA1truncation of the surface alloy is the most stable, and the energy gained by forming this truncation with respect to the Ni-truncation is about 0.15 eV, which is approximately half the segregation energy. Another important result presented in this figure is the behavior of the surface energies of partially ordered p(2xl)-NiA1 alloys in the surface layer. Such partially ordered alloys have the same ordered p ( 2 x l ) structure, but the excess of Ni atoms form partial antisite defects on the A1 sublattice. One can

2.0

oE 1.9 ~ cr

~---~random alloy in the 1st layer O--Orandom alloy in the 2nd layer ~partially ordered alloy in the 1st layer

1.8

..-%

1.7'

0 ~ 1.6

1.5

0.00

Ni/p(2xl)NiAI p(2xl )NiAI/Ni

~

I

0.25

~

I

0.50 O

~

I

0,7'6

a

1.00

Fig. 11. The calculated surface energies of Ni(ll0) with random, partially ordered, and p(2xl) ordered NiA1 layer on the surface and in the subsurface layers (capped by a Ni layer). The dotted line indicates the energy of the two-phase system for a given c: The pure Ni(ll0) surface and the ordered NiA1 alloy in the first layer.

22 see in Fig. 11 that the surface energies of the partially ordered alloys go above the line which connects the surface energy of the pure Ni(110) surface (c=0) and the completely ordered p(2xl)-NiA1 alloys in the surface layer. This is a very general feature, which holds not only in the case of surface alloys, but also in the case of bulk systems. It is connected to the concentration dependence of the ordering energy and means that at low temperature the partially ordered alloys should undergo phase separation if the alloy composition is not stoichiometric. That is, if the A1 coverage is less than half a monolayer, the surface of Ni(110) will be covered by pure Ni and completely ordered p(2x 1)-NiA1 islands. A similar growth of the ordered Ni3A1 alloy is observed experimentally during deposition of A1 on the (100) surface of Ni [46]. Here the formation of a stable c-(2x2) ordered NiA1 alloy was found on the surface while the second layer was an almost entirely pure Ni layer and the third layer was enriched by A1. This type of structure corresponds to the NiA1 termination of the Ni3AI(100) surface, which also has an alternative truncation. The surface segregation energy of A1 on the (100) surface of Ni is only about-0.1 eV, and as has been shown [24], the NiA1 termination is more stable than Ni termination by approximately half of this value.

1,3

E

0 *"~

m

11st random )2nd random Plst ordered 12nd ordered

"

1.2 "i

>

1.1 -

r-

1.0-

0

0.9

'1:: 09

0.8 0.00

I

I

0.25

I

I

0.50 C

I

|

0.75

I

1.00

Fig. 12. The surface energies of random and p(2xl)-ordered Pd~Cul_c alloys in the first (surface) and second (subsurface) layers.

23

A system which exhibits a behavior somewhat different from A1-Ni is PdCu. The first ordered phase in the Cu-rich region of the Cu-Pd bulk phase diagram [47] is L12-Cu3Pd, and therefore it is not a surprise that the growth of Pd on Cu(110) leads to the formation of surface alloy with the corresponding bulk ordered structure [21]. However, in contrast to the growth of A1 on Ni(110), Pd does not segregate to the (110) surface of Cu. This can be seen in Fig. 12 where the first-principles results for different surface alloys of are presented [21 ]. In fact, although the segregation energy of Pd into the first layer is positive, but very small (less than 0.05 eV: It is the initial slope of the surface energy curve for the random alloy in the first layer) the main driving force behind the final surface alloy configuration is the segregation energy of Pd into the second layer, which is -0.23 eV. As a result the energy gain of having the Cu truncation at the surface is about 0.1 eV relative to the CuPd truncation. The reason, why the energy of segregation to the second layer is so large is the fact that the (110) surface is quite open: as one can see from Fig. 9, the second layer is in fact not covered by the surface atoms. In the case of NiPt random alloys, this even leads to a segregation reversal at the (110) surface. Like the case considered above this is directly related to the quite large energy of segregation to the second, subsurface layer [48,49], which is greater than the energy of surface segregation to the first layer. Therefore, in general one should be very careful in making predictions for more open surfaces: simple surface segregation arguments may not work at all. 9. ALTERNATIVE ORDERED STRUCTURES ON THE S U R F A C E

In Fig. 12 we have also shown that partially ordered (2x l) CuPd alloys in the subsurface layer (c < 0.5) are unstable against separation into islands of pure Cu(110) surface and ordered (2x l) CuPd islands capped by Cu atoms. This is indeed observed experimentally [21]. However, at a very low coverage of a few percent, ordered - C u - P d - one dimensional chains aligned along the closed-packed [ 110] direction are formed in the surface layer, see Fig. 13. Although, this may look as a change in the ordering behavior of the surface alloy, the effect is entirely consistent with the ordering behavior in the bulk and is in fact related to the specific features of the structure of the surface itself. Namely, the strongest effective interaction (see (10)) which is responsible for the ordering in CuPd is the effective interaction for the first coordination shell [50]. All the other interactions are rather small. This means that the main gain in the ordering energy is due to CuPd ordering in the closed-packed direction. Since the (110) surface is anisotropic, the Pd atoms first tend to form order in this specific direction, forming thereby ordered CuPd strings at very low Pd coverage. The reason, why such ordered chains

24

Fig.13. (a) STM image following deposition of small amounts of Pd on Cu(ll0). Linear chains are observed, which are aligned along the closed-packed direction (70x70 Ang.). (b) Atomically resolved image of an island of pure Cu at coverage 0.28 ML Pd.

are not covered by Cu atoms, is the fact that the energy gain by this process does not counterbalance the energy cost of creating the steps, which must appear during such a process. At higher Pd coverage when PdCu islands of ordered alloy start to form the perimeter-to-area ratio of the islands drops dramatically and hence the energy balance changes in favor of capping CuPd islands. One may also see in Fig.13 that such islands have a preferential alignment along the [ 110] direction. In the example considered above the ordering of the deposited element and the substrate leads to the formation of distinct long-range structures: chains and islands. However, it may not always exhibit itself as long-range order even below the order-disorder transition temperature. This kind of behavior is observed in the Cu-Pd system, but during a deposition of Pd on Cu(111) in the temperature range between - 8 0 - 250 C which is well below the orderdisorder transition temperature of Cu3Pd in the bulk (about 500 C[47]). In this case similar to the growth of Pd on Cu(110) considered above and on Cu(100) [51,52] one may expect a formation of ordered (2x2) Cu3Pd surface alloy consistent with the (111) surface of L12-Cu3Pd alloy. Nevertheless, a formation of bands of a quite stable random CuPd alloy along the steps at the surface has been observed [53]. The Pd concentration in this alloy depends on the subsequent heat treatment, and varies between 0.18 and 0.31 at.% of Pd. As a matter of fact although the alloy configuration seems to be completely random without any distinct long-range order features, the analysis of the STM image shows that almost all of the Pd atoms are surrounded by Cu atoms

25

0.9

E

0

m

0.8

O-~Orandom

PdcCU~_c alloy

0 C~

,- 0.7

0 tO 0 0

m 0.6

"1:::

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

:3

......- A

CuaPd

O9 0.5

,

0.00

I

0.25

Cu2Pd ,

I

0.50

0.75

Fig. 14. First-principles results for the surface energy of random and ordered surface alloys on Cu(111). CuzPd and Cu3Pd are ~ x qr~ and (2x2) ordered alloys correspondingly. The dotted line is the stability line which is the surface energy of a disordered alloy with the maximal possible value of the SRO parameter for a given concentration.

in the first coordination shell. The energy gain due to such a short-range order (SRO) in the (111) fcc layer can be expressed in terms of the effective pair interactions at the first coordination shell, V~, defined in (10), as [54]"

ESRO= -1- Z l 2

C (1-c)

V~ er

,

(17)

where Zi--" 6 is the coordination number of the first coordination shell for fcc(111), c the concentration of Pd, and oc~ the so-called Warren-Cowley SRO parameter for the first coordination shell. The value of the SRO parameter in the case where all Pd atoms are surrounded only by Cu atoms reaches its minimal value, which is - c / ( 1 - c ) [54], and therefore the energy of the SRO effects is - 3 c 2V 1 " As a result the total mixing energy of an alloy with the maximal SRO is the mixing energy of the random alloy, given by (10) plus the ordering energy (17), which yields - 3 c V 1 . This energy is a linear function of the

26 concentration and therefore the energies of (2x2)-Cu3Pd and x/3 x ~-3 CuzPd, as well as the energies of random alloys with maximal value of the SRO parameter lies practically on the same stability line. This means that all these structures are equally stable and may coexist on the surface. This is shown in Fig. 14 where the results of the first-principles calculations for the surface energies of random and ordered Pd~Cul_~ alloys [53] are presented together with the surface energy of the disordered alloys having the smallest possible value of the SRO parameter at the first coordination shell.

Fig. 15

~f3 x ~

- A2B (a) and (2x2)-A3B (b) structures on the triangle lattice.

27

Such an unusual behavior is in fact a consequence of the highly (infinitely, to be precise) degenerated ground state (at T=0K) of an alloy on a triangle (fcc(111), hcp(0001)) lattice with positive nearest neighbor interactions due to frustration effects [54]. There are, for instance, infinitely many random alloy configurations of A3B alloys the energy of which are equal to the energy of the 2x2-A3B alloy. Such a degenerate ground state for alloy compositions different from A2B (or AB2) leads to the so-called surface induced disorder in the case of the (111) surface of L12-A3B and L10-AB ordered alloys [55]. The only exception is the A2B alloy on the triangle lattice which has a Vr3 x ~ - A2B structure in the ground state. If the alloy composition exceeds 1/3, then again the ground state becomes infinitely degenerate. Nevertheless, its energy will be higher than the stability line connecting the surface energies of A and ~ x ~ - A2B, since 1/3 is the maximal concentration at which atoms of one alloy component can be surrounded exclusively by the atoms of the opposite type on the triangle lattice and at this composition there is only one way to arrange every triangle to be A2B. This makes the ~ x ~r~-A2B surface structure special in the d e p o s i t i o n experiments for the fcc(111) and hcp(0001) surfaces in the case of ordered alloys, and is the reason why it is so frequently observed in the deposition experiments [56-60]. It should be noticed, however, that the L12(lll)-A3B and ~ x ~ A2B ordered structures are equally stable (on the same stability line) only if the effective interactions for more distant coordination shells are zero. If this is not the case, then the relative stability of these structures will depend on the value of the other interactions. The first difference actually appears at the second coordination shell in the surface layer (which corresponds to the third coordination shell in the bulk). The corresponding contribution from V2 to the mixing energy of the L l z ( l l l ) - A 3 B is again - 3 c V 2 , while it is zero in the case of the ~ x ?r~-A2B alloy, that is, the ordering at the second coordination shell will favor the L12(111)-A3B ordered structure. Exactly the opposite situation occurs with the effective interactions at the third (fourth in the bulk) coordination shell, V3, which give zero and - 3 c V3 contributions to the mixing energy of the L12(111)-A3B and phases, respectively.

~r~ x ~

-A2B ordered

ACKNOWLEDGMENT The Center for Atomic-scale Materials Physics is sponsored by the Danish National Research Foundation.

28

REFERENCES [1] D.A. King, D.P. Woodruff (eds.), Growth and Properties of Ultrathin Epitaxial Layers, Elsvier, Amsterdam, 1998. [2] U. Bardi, Rep. Prog. Phys. 57 (1994) 939. [3] J. Wintterlin and R.J. Behm, in Scanning Tunneling Microscopy I, 2nd ed., edited by H.J. Guntheroldt and R. Wiesendanger, Springer-Verlag, Berlin, 1994. [4] R.Q. Hwang, C. Gunter, J. Schroder, S. Gunter, E. Kopatzki, and R.J. Behm, J. Vac. Sci. Technol. A 10 (1992). [5] D.D. Cahmbliss, R.J. Wilson, and S. Chiang, IBM J. Res. Dev. 39 (1995) 639. [6] S.C. Wu, S.H. Lu, Z,Q. Wang, C.K.C. Lok, J. Quinn, Y.S. Li, D. Tian, F. Jona, and P.M. Marcus, Phys. Rev. B 41 (1990) 3353. [7] K.-O. Ng and D. Vanderbilt, Phys. Rev. B 52 (1995) 2177. [8] G.E. Thayer, V. Ozolins, A.K. Schmid, N.C. Barlet, M. Asta, J.J. Hoyt, S. Chiang, and R.Q. Hwang, Phys. Rev. Lett. 86 (2001) 660. [9] L.P. Nielsen, F. Besenbacher, I. Stensgaard, E. Lagsgaard, C. Engdahl, P. Stolze, K.W. Jacobsen, and J.K. Norskov, Phys. Rev. Lett. 71 (1993) 754. [10] C. Nagl, E. Platzgummer, O. Haller, M. Schmid, and P. Varga, Surf. Sci. 331 (1995) 831. I1 l] J.L. Stevens and R.Q. Hwang, Phys. Rev. Lett. 74 (1995) 2078. [12] P.T. Sprunger, E. Lagsgaard, and F. Besenbacher, Phys. Rev. B 54 (1996) 8163. [13] H. Brune, H. Roder, C. Boragno, and K. Kern, Phys. Rev. B 49 (1994) 2997. [14] C. Gunter, J. Vrijmoeth, R.Q. Hwang, and R.J. Behm, Phys. Rev. Lett. 74, (1995) 754. [15] A. Christensen, A.V. Ruban, P. Stolze, K.W. Jacobsen, H.L. Skriver, J.K. Norskov, and F. Besenbacher, Phys. Rev. B 56 (1997) 5852. [16] A. Senhaji, G. Treglia, B. Legrand, N.T. Barret, C. Guillot, and B. Villete, Surf. Sci. 274 (1992) 297. [17] S. Blugel, Appl. Phys. A 63 (1996) 595. [18] T. Asada, S. Bulgel, Physica, B 237-238 (1997) 359. [19] Ch. Ross, S. Schirmer, M. Wuttig, Y. Gauthier, G. Bihlmayer, and S. Blugel, Phys. Rev. B 57 (1998) 2607. [20] G. Bihlmayer, Ph. Kurz, and S. Blugel, Phys. Rev. B 62 (2000) 4726. [21] P.W. Murray, S. Thorshaug, I. Stensgaard, F. Besenbacher, E. Lagsgaard, A.V. Ruban, K.W. Jacobsen, G. Kopidakis, and H.L. Skriver, Phys. Rev. B 55 (1997) 1380. [22] M.O. Pedersen, I.A. Bonicke, E. Lagsgaard, I. Stensgaard, A. Ruban, J.K. Norskov, and F. Besenbacher, Surf. Sci. 387 (1997) 86. [23] L.T. Wille, B. Nonas, P.H. Dederichs, and H. Dreysser, Phil. Mag. B 78 (1998) 643. [24] A.V. Ruban and H.L. Skriver, Comp. Mat. Sci. 15 (1999) 119. [25] A.V. Ruban, H.L. Skriver, and J.K, Norskov, Phys. Rev. B 59 (1999) 15990. [26] M. Methfessel, D. Henning, M. Schefler, Phys. Rev. B 46 (1992) 4816. [27] L. Vitos, A.V. Ruban, H.L. Skriver, and J. Kollar, Surf. Sci., 411 (1998) 186. [28] J. Kollar, L. Vitos, B. Johansson, and H.L. Skriver, Phys. Stat. Sol. (b) 217 (2000) 405. [29] J. Friedel, The physics of metals (ed. J.M. Ziman), p.494 Cambridge University Press, New York (1969). [30] M.C. Desjonqueres and D. Spanjaard, Concepts in Surface Physics, Springer, Berlin, 1996. [31] D. Pettifor, Bonding and structure of molecules and solids, Clarendon Press, Oxford, 1995. [32] A.F. Becker, G. Rosenfeld, B. Poelsema, and G. Comsa, Phys. Rev. Lett., 70 (1993), 477.

29 [33] H. Roder, R. Schuster, H. Brune, and K. Kern, Phys. Rev. Lett. 71 (1993) 2086. [34] U. Struber and K. Kuppers, Surf. Sci. Lett., 294 (1993) L924. [35] P. Zeppenfeld, M.A. Krzyzowski, Ch. Romainczyk, R. David, G. Comsa, H. Roder, K. Bromann, H. Brune, and K. Kern, Surf. Sci. Lett., 342 (1995) L1131. [36] F.R. de Boer, R. Boom, W.C.M. Mattens, A.R. Miedema, and A.K. Niessen, Cohesion in Metals: Transition Metal Alloys, North-Holland, Amsterdam, 1988. [37] A.V. Ruban, H.L. Skriver, and J.K. Norskov, Phys. Rev. Lett., 80 (1998) 1240. [38] J. Tersoff, Phys. Rev. Lett. 74 (1995) 434. [39] B. Johansson, N. Martenson, Phys. Rev. B 21 (1980) 4427. [40] B. Nanos, K. Wildberger, R. Zeller, and P.H. Dederichs, Phys. Rev. Lett. 80 (1998) 4574. [41] M.T. Kief and W.F. Egelhoff, Jr. Phys. Rev. B 47 (1993) 10785. [42] J. Jandelleit, Y. Gauthier, M. Wuttig, Surf. Sci. 319 (1994) 287. [43] J. Giergeil, J. Shen, J. Woltersdorf, A. Kirilyuk, and J. Kirschner, Phys. Rev. B 52 (1995) 8528. [44] J. Shen, J. Giergiel, A.K. Schmid, J. Kirschner, Surf. Sci. 328 (1995) 32. [45] C. Pflitsch, R. David, L.K. Verheij, R. Franchy, Surf. Sci. 468 (2000) 137. [46] D.J.O'Connor, M. Draeger, A.M. Molenbbroek, Y. Shen, Surf. Sci. 357/358 (1996) 202. [47] P.R. Subramanian and D.E. Laughlin, J. Phase Equilibria 12 (1991) 231. [48] I.A. Abrikosov, A.V. Ruban, H. L. Skriver, and B. Johansson, Phys. Rev. B 50 (1994) 2039. [49] L.V. Pourovskii, A.V. Ruban, I.A. Abrikosov, Yu. Kh. Vekilov, and B. Johansson, Phys. Rev. B 64 (2001) 35421. [50] Z.W. Lu, D.B. Laks, S.-H. Wei, and Z. Zunger, Phys. Rev. B 50 (1994) 6642. [51] P.W. Murray, I. Stensgaard, E. Lagsgaard, F. Besenbacher, Phys. Rev. B 52 (1995) R 14404. [52] P.W. Murray, I. Stensgaard, E. Lagsgaard, F. Besenbacher, Surf. Sci., 365 (1996) 591. [53] A.B. Aaen, E. Lagsgaard, A.V. Ruban. and I. Stensgaard, Surf. Sci., 408 (1998) 43. [54] F. Ducastelle, Order and Phase Stability in Alloys, North-Holland, Amsterdam, 1991. [55] J. Neugebauer, M. Scheffler, Phys. Rev. Lett. 71 (1993) 577. [56] W. Schweika, D.P. Landau, K. Binder, Phys. Rev. B 53 (1996) 8937. [57] S. Oppo, V. Fiorentini, and M. Scheffler, Phys. Rev. Lett. 71 (1993) 243. [58] P. Baily, T.C.Q. Noakes, D.P. Woodruff, Surf. Sci. 426 (1999) 358. [59] D. Tian, H. Li, S.C. Wu, F. Jona, and P.M. Marcus, Phys. Rev. B 45 (1992) 3749. [60] D. Tian, A.M. Begley, and F. Jona, Surf. Sci. Lett. 273 (1992) L393.


Surface Alloys, and Alloy Surfaces 30

D.P. Woodruff, (Editor)

Chapter 2

Atomistic modeling of surface alloys Guillermo Bozzolo a'b and Jorge E. Garces a'c

aOhio Aerospace Institute, 22800 Cedar Point Rd., Cleveland, OH 44142, USA bNASA Glenn Research Center, Cleveland, OH 44135, USA CCentro Atomico Bariloche, 8400 Bariloche, Argentina 1. INTRODUCTION Different experimental techniques provide detailed information on the structure and composition of surface alloys and alloy surfaces, sometimes leaving little doubt regarding the often complex patterns that appear, for example, during the process of surface alloy formation [1-13]. There are cases, however, in which some level of modeling is necessary in order to reconcile the sometimes conflicting conclusions that can be drawn from different experiments. Whether it is the often unpredictable alloy surface composition and structure, due to segregation or surface defects, or the several active degrees of freedom during deposition of different types of atoms on an arbitrary substrate (a polycrystalline surface of a monatomic crystal, an alloy surface, etc.), atomistic modeling is essential in identifying isolated features, analyzing correlations, or simply allowing for the study of a wide range of possibilities not accessible via experiment as, for example, the study of metastable structures. Whether it is the analysis of different bulk or surface alloy phases, it is difficult to predict what can be expected during the corresponding process of formation. From the modeling standpoint, the process is extremely complex, and any attempt to develop a useful modeling tool would be almost hopelessly limited by the excessive number of variables that should be taken into account in order to provide a satisfactory description of the actual process. Increasing computer power, alone, is not necessarily the only answer, nor is the development of detailed theories that, regardless of the computer power available, are sometimes difficult to implement. However, the purpose of atomistic modeling is not to exactly reproduce every detail of the real process, but to be able to identify the main features and driving mechanisms of a certain specific behavior. The amount

31

of detail that can be considered satisfactory is clearly dependent on the problem at hand, but in spite of the particular characteristics of any given system, a few fundamental ingredients are necessary: 1) simplicity in the implementation of the physical theory and the ensuing calculations, 2) universality in the set of parameters or potentials used (i.e., complete transferability) and 3) versatility, in terms of a minimum number of restrictions on the type and number of elements and type of structures. Satisfying these minimum requirements is essential for the successful modeling of surface alloys which has been, so far, mostly limited to either a single-element substrate or the deposition of one single element at a time. To extend modeling to more complex systems would require complete freedom from the restrictions imposed by not fulfilling any or all of the above mentioned conditions. While a demanding challenge, the recent trend of combining first-principles methods with quantum approximate methods has resulted in steady progress in this area, allowing for increased understanding on the atomic processes that govern the phenomenon of surface alloy formation. Although different in their formulation, all quantum approximate methods rely on the simplicity introduced by a global description of the interaction between atoms, but at the same time, such generalization often translate into limitations thus failing to provide an ideal framework that would allow for a general and unrestricted application. In this work, we will concentrate on the description and application of one such method, particularly suitable for the study of surface phenomena. The Bozzolo-Ferrante-Smith (BFS) method for alloys [14] satisfies most of the requirements imposed on quantum approximate methods, in terms of simplicity, accuracy, generality and ease of implementation, with the added advantage that its novel interpretation of the alloy formation process is free of constraints that would limit its applicability to arbitrary systems. To a great extent, this lack of restrictions in the general formulation of the BFS method relies on the ability to properly define the parameters of each constituent element. In high symmetry situations (i.e., Cu on Cu(100), where both the adsorbate and the substrate atoms are of the same atomic species), the implementation of the method is generally straightforward. That is also the case when the different participating elements have the same bulk symmetry (i.e., Pd on Cu(100), where both elements form fcc bulk solids) [15]. It is not necessarily true, however, that the bulk symmetry of each element dictates the structure of the alloy, as it is most generally the case that phases of other symmetries can and do form [ 16]. This is even more so in the case of surface alloys, where not even a thorough knowledge of the bulk phases constitutes a sufficient basis for the determination of the structures that form on the surfaces. Depending on the characteristics of the surface, even immiscible metals in the bulk are known to form ordered surface structures [ 17-21].

32

From a theoretical standpoint, the traditional approach for the determination of an alloy structure implies, in principle, a search through any possible configuration until the most energetically favorable is found. While current first-principles methods, coupled with a substantial increase in computational power, have made this approach a standard practice for the calculation of phase diagrams of (mostly binary) bulk alloys, the complexity of surfaces makes quantum approximate methods a necessary tool to supplement the existing techniques and the growing body of experimental data. However, the study of surfaces and surface processes has been a severe test for quantum approximate methods, which usually rely on parameters or potentials determined from bulk properties, thus limiting their effectiveness in the low-symmetry environments represented by surfaces. One way to circumvent this obstacle is by formulating the method on the basis of a one-to-one mapping between any arbitrary bulk or surface environment onto an equivalent, ideal, bulk-like one. If such correspondence is uniquely established, then the parameterization becomes universal and equally applicable to bulk or surface problems. The BFS method for alloys satisfies this requirement by assigning to every atom i, regardless of its location and environment, a set of equivalent bulk crystals whose states of isotropic compression or expansion (and the corresponding difference in energy with their ground states) are taken as a measure of the defect in the real crystal where atom i is located. Three equivalent crystals for a given atom i are thus defined to completely describe the different aspects in the process of alloy formation. The first of these equivalent crystals describes purely structural effects. This is achieved by considering every neighboring atom as if it was of the same atomic species of atom i but retaining the actual positions that they have in the real system. The second equivalent crystal of atom i describes chemical effects, considering every neighboring atom by its true chemical identity, but forced to occupy lattice sites of an equilibrium, otherwise monatomic, crystal of species i. A third equivalent crystal is also defined in order to ensure a complete decoupling between the first two, eliminating any structural information in the calculation of the chemical effects. Each of these equivalent crystals shows some degree of departure from a certain equilibrium reference state R i. The energies associated with these departures represent, respectively, the strain (s), chemical Co (c), and chemical reference (~,) energies which, properly coupled, represent the contribution of atom i to the total energy of formation AH of the actual system. The choice of the reference state R i is, in most cases, a straightforward task. For example, the reference state of a Cu atom in a pure Cu crystal is, clearly, fcc. That is also the case for a Cu atom in a Cu3Pd L 12 ordered alloy. For a Cu atom in an ordered CuPd B2 alloy, however, the reference state is bcc. For general situations, a substantially useful degree of freedom in the methodology is therefore

33

introduced by allowing the reference state R i to have symmetries other than the one corresponding to the ground state of each constituent element. With the appropriate means for computing the parameters corresponding to arbitrary reference states, a complete characterization of R i is not less straightforward. For the sake of simplicity, it is convenient to illustrate this issue with an example. It is well known that A1 deposition on a Ni substrate leads to the formation of an fcc-like Ni3A1 film [22], followed by a transitional region leading to a bcc-like NiA1 pattern [22,23]. A1 atoms adopt the symmetry of the substrate (fcc) for low coverages, slowly transitioning to a different symmetry (bcc) as the A1 coverage increases. The layer-by-layer individual distortions from equilibrium lattice sites translate into what can be seen as a continuous transition from Al-fcc to Al-bcc. Other examples include the deposition of Cu on Ag(100) [24], where Cu layers transition from bct to bcc with increasing coverage. In general, varied situations ranging from grain boundaries, where each individual atom sees a different environment that could be best described by a particular intermediate state between the symmetries describing each grain, to liquids and amorphous materials, could be described with this approach. It is clear that with the novel way of partitioning the individual atomic contributions to the energy of formation, added to the appropriate determination of the reference state for each atom, the method provides a tool that is general enough to tackle equally general situations: a) The representation of arbitrary situations in terms of bulk equivalent crystals removes any distinction between bulk, surface or interface situations, all dealt with the same degree of accuracy, b) the calculation of the strain energy in terms of equivalent monatomic crystals lifts any restriction on the structural characteristics of the system at hand, c) the calculation of the chemical energy in terms of perfect crystals lifts restrictions in terms of the number of different atomic species that can be included and d) the atomby-atom determination of the reference state lifts any restriction on the number and type of phases that can be formed by any group of atoms. From a computational standpoint, the usefulness of the method relies on the simplicity of the calculations needed for the determination of the three equivalent crystals associated with each atom i. This is accomplished by building on the simple concepts of Equivalent Crystal Theory (ECT) [25,26], as will be discussed in detail below. The procedure involves the solution of one simple transcendental equation for the determination of the equilibrium Wigner-Seitz radius (rWSE) of each equivalent crystal. These equations are written in terms of a small number of parameters describing each element in its reference state, and a matrix of perturbative parameters Aji, which describe the changes in the electron density in the vicinity of atom i due to the presence of an atom j (of a different chemical species), in a neighboring site. The determination of parameters for each atom in

34

its reference state, whichever it is, is also a straightforward task, as it can be easily accomplished with first-principles methods when experimental input is not available. While we have restricted the examples shown in this work to systems for which experimental input exists, it is important to note that the possibility of expanding the input database by means of first-principles calculations allows the methodology to be applied to any arbitrary system. In addition, the ability to obtain every necessary parameter from the same source has the additional advantage of giving the BFS method much needed consistency in terms of the parameterization used. The primary set of parameters needed are the equilibrium values for the Wigner-Seitz radius, the cohesive energy, E c, and the bulk modulus, Bt~ Two additional single-element parameters are determined as a function of the parameters in the primary set: a screening length ~., that accounts for the screening of atoms beyond the nearest-neighbor layer, and a parameter tx, which represents a measure of the electron density in the overlap region between neighboring atoms. Moreover, the parameter tx is determined by requiting that the maximum strain energy that can be assigned to a given atom is given by the cohesive energy, thus allowing for a direct and simple calculation that also establishes the boundaries for the range of validity of the method. Thus, the primary set of parameters describing any arbitrary reference state for a given atom is then {rws E, E c, l, ~, o~}, where l, a scaling length, replaces B o in order to allow for a closer correspondence with the universal binding energy relationship (UBER) of Rose et al. [27], which is usually written in terms of I rather than B o. A detailed description of the operational equations and the role of each parameter will be presented in Sec. 2. A second set of parameters for element i, the BFS parameters Aji, account for the changes in the electron density in the vicinity of atom i due to the presence of an atom j. These parameters can be computed with first-principles methods by means of a straightforward calculation based on the energy of formation and equilibrium atomic volume of all the binary systems that can be formed with the participating elements. For more accurate results, the parameters Aji can be computed as a function of the concentration of element j in the vicinity of atom i. In some cases it is possible to fully parametedze a given system from experimental input, as will be done in every example presented in Sec. 3. However, the determination of complete primary parameter sets for every possible element in every possible reference state, as well as the associated secondary sets for the concentration-dependent binary cases, is not an easy task as, in most cases, it would require input that is not available from experiment. Once again, this issue can be properly and systematically dealt with by means of first-principles calculations. For example, reference states can differ from the ground state symmetries of the participating elements in the alloy. In those situations, it is strictly neces-

35

sary to rely on first-principles methods for the determination of the equilibrium properties of such crystals. To illustrate this point, we focus on one particular example, namely, the continuous transformation between a bcc and an fcc structure (Fig. 1). Several intermediate configurations can be singled out and the firstprinciples calculation of the primary set of parameters can be made for each one of these configurations. In doing so, each one of the relevant parameters can be written as a function of a single quantity, properly defined to identify each step in the transformation. For the particular case of the bcc fcc transformation, illustrated in Fig. 1, this parameter is r - -~. This procedure can be generalized to include transformations between any number of elemental crystallographic structures. Once this parameterization procedure is completed, the primary set of single-element parameters is general enough so as to allow for the identification of the appropriate reference state for every atom in the system under consideration. In this work, recent progress on BFS modeling of surface alloys will be summarized, with the main purpose of exploring the possibilities that become available with the synergy between a computationally simple and physically sound methodology, increasing computing power, and a substantial practical foundation based on powerful techniques for experimental analysis. For the problem at hand, surface alloys, the general formulation of BFS is not just convenient, but necessary. While it is true that a large number of applications deal with rather simple systems (i.e., deposition of one element on a monatomic substrate), there is a growing body of experimental evidence dealing with more complex situations, thus imposing challenging demands on any modeling effort. It is therefore important to establish a modeling tool for experimentalists based on an operating procedure with a minimum number of constraints, thus allowing for the systematic analysis and interpretation of specific observed features. a

a

r

Fig. 1' Relation b e t w e e n the fcc (c _ 1 ) and bcc (a -" "T ) structures.

36 2. THE BFS M E T H O D

The BFS method has been applied to a variety of problems, ranging from the determination of bulk properties of solid solution fcc and bcc alloys and the defect structure in ordered bcc alloys [28] to more specific applications including detailed studies of the structure and composition of alloy surfaces [29], ternary [30] and quaternary alloy surfaces and bulk alloys [31,32], and even the determination of the phase structure of a 5-element alloy [33]. Previous applications have focused on fundamental features in monatomic [26] and alloy surfaces [29]: surface energies, reconstructions, surface structure and surface segregation in binary and higher order alloys [34,35] and multilayer relaxations [36,37]. While most of the work deals with metallic systems, the lack of restrictions on the type of system that can be studied translated into the extension of BFS to the study of semiconductors [38]. In what follows, we provide a brief description of the operational equations of BFS. The reader is encouraged to seek further details in Refs. 28-35, where a detailed presentation of the foundation of the method, its basis in perturbation theory, and a discussion of the approximations made are shown [14]. The BFS method provides a simple algorithm for the calculation of the energy of formation of an arbitrary alloy (the difference between the energy of the alloy and that of its individual constituents). In BFS, the energy of formation AH is written as the superposition of elemental contributions e~ of all the atoms in the alloy AH - E ( E ' i - E i ) i

- EEi

(1)

i

where E i' is the energy of atom i in the alloy and E i is the corresponding value in a pure equilibrium monatomic crystal. In principle, the calculation of AH would simply imply computing the energy of each atom in its equilibrium pure crystal and then its energy in the alloy. In BFS, beyond directly computing the difference E~for each atom in the alloy, a two-step approach is introduced for such a calculation in order to identify contributions to the energy due to structural and compositional effects. Therefore, E~ is broken up in three separate contributions" a strain energy (Es), a chemical energy (Ec), and a chemical reference energy (Ec~ While there is a certain level of arbitrariness in how this separation is implemented, it is only meaningful when a good representation of the initial and final states of the actual process is obtained by properly linking all contributions. This is achieved by recoupling the strain, chemical and chemical reference contributions by means of a coupling function, gi, properly defined to provide the correct asymp-

37 totic behavior of the chemical energy contribution. Each individual contribution ~t can therefore be written as

S gi(eC Co -E i ) Ei = E i +

(2)

The BFS strain energy contribution s is defined as the contribution to the energy of formation from an atom in an alloy computed as if all the surrounding atoms were of the same atomic species, while maintaining the original structure of the alloy. To visualize this concept, Fig. 2.a represents the atom in question (identified with an arrow) in an equilibrium position in its reference, ground state crystal (arbitrarily represented by a simple cubic lattice). Fig. 2.b shows the same atom in the alloy being studied (also arbitrarily represented by a different crystallographic symmetry). The reference crystal and the alloy differ in two basic aspects. First, atoms of other species may occupy neighboring sites in the crystal and, second, the crystal lattice may not be equivalent in size or structure to that of the ground state crystal of the reference atom. In Fig. 2.b, the different atomic species are denoted with different symbols from that used for the reference atom, and the differences in size and/or structure are denoted with a schematically different atomic distribution as compared to the ground state crystal shown in Fig. 2.a. The BFS strain energy accounts for the change in energy due only to the

(a)

(b)

(c)

k

d

h

al

Fig. 2: (a) A pure, equilibrium crystal (reference atom denoted by the arrow), (b) a reference atom (denoted by the arrow) in the alloy to be studied (atoms of other species denoted with other shading) and (c) the same reference atom in a monatomic crystal, with the identical structure of the alloy to be studied, but with all the atoms of the same atomic species as the reference atom, for the calculation of the strain energy term for the reference atom. The strain energy is the difference in energy of the reference atom between (c) and (a).

38 change in geometrical environment of the crystal lattice (fromFig. 2.a to 2.b), ignoring the additional degree of freedom introduced by the varying atomic species in the alloy. In this context, Fig. 2.c shows the environment 'seen' by the reference atom when computing its BFS strain energy contribution. The neighboring atoms conserve the sites in the actual alloy (Fig. 2.b), but their chemical identity has changed to that of the reference atom (Fig. 2.a) thus simplifying the calculation to that of a single-element crystal. The BFS strain energy term represents the change in energy of the reference atom in going from the configuration denoted in Fig. 2a to Fig. 2.c. In this sense, the BFS strain energy differs from the commonly defined strain energy in that the actual chemical environment is replaced by that of a monoatomic crystal. Its calculation is then straightforward, even amenable to first-principles techniques. The chemical environment of atom i is considered in the computation of ~c, the first term in the total BFS chemical energy contribution, where the surrounding atoms maintain their identity but are forced to occupy equilibrium lattice sites corresponding to the reference atom i. Following the convention introduced in Fig. 2, Fig. 3.a shows the reference atom in the actual alloy (similar to Fig. 2.b), while Fig. 3.b indicates the atomic distribution used in computing the BFS chemical energy Ec (note that the lattice used in Fig. 3.b corresponds to that of the reference crystal of the reference atom, as shown in Fig. 2.a). The total BFS chemical energy is then the difference between the energy of the reference atom 2.a). in Fig. B.2.b, ec, and its energy in its ground state crystal ~C~ The chemical reference energy ~Co (Fig. 2.a) is included in order to completely free the chemical energy from structural defects, taking into account the possibility that the reference atom is not in a full-coordination environment (as is the case (a)

(b)

Fig. 3: (a) The reference atom (denoted by an arrow) in the actual alloy environment and (b) the reference atom surrounded by a chemical environment equivalent to that in (a) but with the different neighboring atoms occupying equilibrium lattice sites corresponding to the reference, ground state of the reference atom.

39

E

m

~~~li:

9

=

..t_ gi

.

+

_

9

Fig. 4: Schematic representation of the BFS contributions to the total energy of formation. The left hand side represents the reference atom (denoted by an arrow) in an alloy. The different terms on the right hand side indicate the strain energy (atoms in their actual positions but of the same atomic species as the reference atom), the chemical energy term (atoms in ideal lattice sites) and the reference chemical energy (same as before, but with the atoms retaining the original identity of the reference atoms).

in or near a surface). This is accomplished by recomputing the contribution Ec defined before, but once again assuming that all atoms are of the same species as the reference atom. As mentioned above, the BFS strain and chemical energy contributions take into account different effects, i.e., geometry and composition, computing them as isolated effects. A coupling function, gi, restores the relationship between the two terms. This factor is defined in such a way as to properly consider the asymptotic behavior of the chemical energy, where chemical effects are negligible for large separations between dissimilar atoms. Within the framework of this discussion, the total BFS contribution Ezof each atom in the alloy can be graphically depicted by the combination of strain and chemical effects shown in Fig. 4. In what follows, we present the necessary equations and concepts needed for the computation of each energy term.

2.1. Calculation of the BFS strain energy The BFS strain energy can be computed by any method appropriate for the calculation of pure element crystals. Due to its consistency with the determination of the chemical energy contribution, we choose the ECT [25,26] for its computation. ECT is based on an exact relationship between the total energy and atomic locations, and applies to surfaces and defects in both simple and transition metals as in covalent solids [25]. Lattice defects and surface energies are determined via

40

perturbation theory on a fictitious, equivalent single crystal whose lattice parameter is chosen to minimize the perturbation. The energy of the equivalent crystal as a function of its lattice constant, is given by a UBER [27]. The method can be easily applied to calculate surface energies, surface reconstructions and bulk distortions of metals and semiconductors. ECT is based on the concept that there exists for each atom i, a certain perfect equivalent crystal with its lattice parameter fixed at a value so that the energy of atom i in the equivalent crystal is the BFS strain energy contribution Es. This equivalent crystal differs from the actual ground state crystal only in that its lattice constant may be different from the ground state value. We compute Es via perturbation theory, where the perturbation arises from the difference in the ion core electronic potentials of the actual defect solid and those of the effective bulk single crystal. For the sake of simplicity, the formal perturbation series in ECT is approximated by simple, analytic forms which contain a few parameters, which can be obtained from experimental results or first-principles quantum mechanical calculations. The simplified perturbation series for Es is of the form ~"iS --

* Ec' F[al(i)] + ~ F [ a 2 (*i , J)] + ~ F [ a 3 (*i , J, k)]

j

j,k

+

* p, ~_F[a4(i, p,q

q)] l

J

(3)

where F[x]

= 1 - (1 + x ) e - x

(4)

Four different contributions to the energy of atom i, which find their origin in four different perturbations, are singled out. The linear independence attributed between these four terms is consistent with the limit of small perturbations which is assumed in the formulation of ECT. Correspondingly, four different equivalent crystals have to be determined for each atom i [25]. The first term, e[a~(i)], contributes when average neighbor distances are altered via defect or surface formation (i.e., changes in coordination). It can be thought of as representing local atom density changes. In most cases this "volume" term is the leading contribution to E~ and in the case of isotropic volume deformations, it gives Es to the accuracy of the UBER [27], given by Eq. 4. The value of a~(/), the scaled lattice parameter of the first equivalent crystal associated with atom i, is chosen so that the perturbation (the difference in potentials between the solid containing the defect and its bulk, ground state equivalent crystal) vanishes. This requirement translates into the following condition from which a~(i) is determined:

41

gRPe_t~R 1 + MRPe-(a+ ~)R2 =

2

rf e-(a+ S(rj))rj

(5)

j(defect)

where the sum over the defect crystal or surface is over all neighbors within nextnearest-neighbor distance. 1) is the actual distance between the reference atom i and a neighbor atom j, N and M are the number of nearest-neighbor (NN) and next-nearest-neighbors (NNN), respectively, of the equivalent crystal (12 and 6 for fcc, 8 and 6 for bcc). The ECT parameters p, {~ and 9~for all the elements used in this work are listed and described in Table 1. S(r) is a screening function given by

i

S(r)=

(6)

1 - cos ~(r 2 _ rl)_]

for r I < r < r 2, S(r) = 0 for r < r I and S(r) = 1/~, for r > r 2, and R 1 and R 2 are the NN and NNN distances in the equivalent crystal of lattice parameter a~, which is obtained by solving Eq. 5. The equivalent lattice parameter a 1 is related to the scaled quantity al(i) via 9

- rWS E

al =

1

(7)

where rws E is the equilibrium Wigner-Seitz radius, l is a scaling length,

Table 1 Computed input parameters for Ni, Cu, Pd, Pt and Au Experimental results Lattice Parameter (/~)

Cohesive Energy (eV/atom)

Bulk Modulus (GPa)

Ni

3.524

4.435

187.48

6

Cu

3.615

3.50

142.12

Pd

3.89

3.94

Pt

3.92

Au

4.078

ECT parameters

(A-1)

(A)

(A -1)

3.015

0.270

0.759

6

2.935

0.272

0.765

195.83

8

3.612

0.237

0.666

5.85

288.54

10

4.535

0.237

0.666

3.78

180.74

10

4.339

0.236

0.663

42

l =

Ec 12~Borws E ,

(8)

cl is the ratio between the equilibrium lattice constant and rws E and where B o is the bulk modulus. The higher order terms are relevant for the case of anisotropic deformations [25]. The second term, F[a*2(i,j)], is a two-body term which accounts for the increase in energy when N bonds are compressed below their equilibrium value. This effect is also modeled with an equivalent crystal, whose lattice parameter is obtained by solving a perturbation equation given by

NRPle-o~R~- NR~e -~176+ A a R ~ (Rj - Ro) e-~(RJ- Ro)

= 0

(9)

J

where ~ - 4t~ for metals [25], R 1 is the NN distance of the equivalent crystal associated with the deviation of NN bond length Rj from R o, and R o is the bulk NN distance of a pure crystal of lattice parameter a e, at whatever pressure the solid is maintained. A 2 is a constant determined for each metal [25]. The scaled equivalent lattice parameter is then , a2 =

(R-~)-rWSF-,

l

(10)

The third term, F[a~(i,j,k)], accounts for the increase in energy that arises when bond angles deviate from their equilibrium values of the undistorted single crystal, and the fourth term, F[a:(i,p,q)], describes face diagonal anisotropies (see Ref. 25). For the topics of interest for this work, these two terms can be neglected, as typical contributions from these anisotropies are exceedingly small for fcc and bcc metals. When ECT is applied to the study of surfaces of monatomic crystals, all four terms should be included in the calculations. However, when considering rigid surfaces (i.e., no interlayer relaxation) all bond lengths and angles retain their bulk equilibrium values, thus Fta2J = Fta3J -- F [ a 4 ] - 0. The rigid surface energy is therefore obtained by solving for the "volume" term represented by Fta~l only. If we consider a rigid displacement of the surface layer towards the bulk, as is the case in most metallic surfaces, the higher-order terms become finite: some bonds are compressed, contributing to Fta2J, the bond angles near the surface are distorted as well as the difference in length between face diagonals in some cases, generating an increase in energy via Fta~l and Fta:J, respectively. For the cases studied in this work, those additional contributions to ~ are generally small, usu9

,

,

9

43

ally representing 1% to 2% of the total energy. In this approximation (i.e., ignoring the third and fourth term in Eq. 4), the method can be further simplified by avoiding the solution of Eq. 9 and determining the bond-length anisotropy term, , F[a2], with an alternative scheme [26]. In this approximation, which we will call ECT in the rest of this work, the corresponding energy contribution is directly computed using Ns

F~2

=

E ~

Mn v~mn

~_~

n= lm= 1

-~nmnF(amn)

(11)

where N s is the number of atoms in the solid, 0,, = 1 if a*mn0 2

(1)

(uAA, uBB and u AB being the interaction energies between the corresponding atoms), the formation of ordered metallic phases is quite common ("chemical" or "intermetallic" compounds with fixed composition, or alloys with a certain range of concentrations). In such ordered alloys atoms of one element tend to be surrounded by atoms of the other element in periodic crystal sub-lattices (Fig.l). At finite temperatures LRO is never perfect, and in addition, there are local fluctuations in composition, known as short-range order (SRO, see Fig.2),

87

A

o w

a

b

Fig. 2. 2D schematics of a binary ordered alloy crystal with LRO alone (a), and with LRO accompanied by short-ranged compositional fluctuations (b). The probability of finding atom A (or B) at a lattice site is expressed by means of the circle size (and colour): higher probability corresponds to larger (and brighter) circles. The largest white circle in (b) represents 100% probability. which unlike LRO does not vanish at the order-disorder transition temperature. In case V < 0, bonding of like atoms is energetically preferred, leading in principle to separation of the alloy into a mixture of A and B rich solid-solution phases, each with nearly homoatomic SRO clusters, compared to short range AB mixed or heteroatomic clusters in the former case of V >0. In other words, the tendency to order (or phase-separate) is manifested to some, local degree also in most solid solutions, where the distribution of atoms in the crystal is not entirely random, and should be incorporated too in any theoretical quantitative evaluation of surface segregation phenomena. Moreover, many alloys of

88 practical importance are comprised at temperatures below solubility limits of two or more phases in a certain micro- (or nano-) structure with ordered regions or clusters (characterized by LRO and SRO) embedded in solid-solution matrixes (with SRO). Since not only the LRO and SRO contributions are temperature dependent, but also the solid solution bulk compositions and relative amounts of phases (each with its distinct surface segregation behaviour), the segregation characteristics of such a multi-phase alloy surface can be even more complex than the single-phase cases. Surface segregation of an alloy constituent, which is very common in substitutional (and interstitial) solid solutions, is expected to be manifested to a less extent in ordered alloys [ 1]. Thus, the process of segregation in A-B alloys, whereby atoms of one constituent element populate preferentially the surface layer, can be viewed as a sort of near-surface phase separation, which is typically incompatible with ordering tendencies*. Actually, segregation is expected to disrupt order and break energetically favourable A-B bonds, and hence as an endothermic process in strongly ordered systems, it may not occur at all, at least at relatively low temperatures. At higher temperatures this suppression is expected to diminish, as entropy-driven segregation with progressively higher levels prevails, until in case of a transition to a disordered phase, it becomes maximal usually around the range of the transition temperature (Fig.3). Then, in the solid-solution high temperature regime, the segregation level eventually decreases with temperature as an exothermic process. The resultant peaked segregation vs. temperature curve expected under certain conditions in strongly ordered systems has been predicted for alloys with LRO [2] and solid-solutions with strong SRO [3], and observed experimentally in several cases (e.g., Refs.3-5). The interplay of LRO and segregation can lead to other types of segregation curves, as described in section 3. Another complication, worthwhile mentioning in the context of developing insight into phenomena of surface segregation in the presence of ordering tendency, emerges when V is strongly composition dependent [6,7], or even changes sign, as in the case of the Fe-Cr system [8,9]. Describing of the equilibrium state of the macroscopic system by means of a statistical-mechanical approximation or Monte-Carlo (MC) simulations is one of the two main aspects of surface segregation theory, while the second aspect deals with the segregation energetics related to "microscopic" atomic interactions. Early experimental data on surface segregation phenomena in solid solutions were usually analyzed by means of the Langmuir-McLean theory [ 10]. This simplistic approach predicts monolayer segregation that decreases monotonously with temperature, and enabled to derive "segregation enthalpy" * Yet, as discussed in Sec.3, in certain bulk truncated terminations of ordered alloys the two tendencies can be compatible.

89

o

r o

~

,AS 0 < 0

Temperature Fig.3. Schematics of the evolution of equilibrium segregation with temperature in alloys with order-segregation competition: (a) dominant surface segregation tendency (LangmuirMcLean behaviour), (b) dominant ordering tendency. Signs of enthalpy and entropy of segregation are indicated. and "excess entropy" from experimental surface compositions vs. temperature (Fig. 3), but fails to account for the above mentioned complex segregation in alloys with interaction-induced strong ordering tendencies. Hence, together with the development of experimental techniques and the fast increase of relevant data, more elaborate theoretical approaches to surface segregation phenomena became necessary [1]. A better starting point for theoretical studies of LRO/segregation interrelations [11,12] became the Bragg-Williams (BW) statistical-mechanical approximation adapted for multilayer surface segregation while still assuming random distribution of atoms at identical layer and sublattice sites. It is based on Ising type rigid lattice model with constant bond energies, ignoring surrounding-dependent pair bonding and many-body interactions (an Ising type model that does consider composition dependent local interactions was introduced recently [9]). As further steps, basic SRO effects on surface segregation were treated by means of the statisticalmechanical cluster variation method (CVM) [13-18], and the free-energy expansion methods (FCEM, described in the next section) [1,3,9,19,20]. On the other hand, Monte-Carlo simulation methods [21-38] are capable of taking into account such contributions as atomic vibrations and surface atomic relaxation [29]. When combined with the embedded atom method (EAM)[23-25] as an improved energy model, or its modified version (MEAM) [37,39,40], MC simulations overcome several drawbacks of the above Ising type models. Yet,

90

the latter analytical approach can be helpful in predicting basic effects of atomic long-range and short-range order on surface segregation in alloys, including multi-component and dilute systems. This chapter is focused on the most recent theoretical and experimental efforts aiming at unravelling the diverse phenomena of segregation/ordering interplay. The issue was reviewed by us comprehensively about two years ago [ 1], and new topics are addressed here in three separate sections: (i) Theoretical formulation of multi-layer segregation in a multi-element solid-solution alloys (ternary alloys in particular) with emphasis on the role of short range order. It is followed by model calculations for NiA1-Cu solid solution. (ii) Evaluation of surface segregation trends for several classes of ordered alloy surface structures, including case studies, primarily in terms of segregation/ordering energetics. In view of the prominence of LRO effects, they constitute a central topic in this review. (iii) The complex segregation behaviour in a bi-phase system comprising of ordered clusters in a solid solution matrix. 2. S E G R E G A T I O N IN M U L T I - E L E M E N T ALLOYS Compared to numerous studies of surface segregation phenomena in binary alloys [1,41], quite fewer studies have been devoted to the theory of surface segregation in multi-component (in particular, ternary) metallic alloys. Characteristic phenomena as co-segregation and site competition were addressed originally by Guttmann [42] using a regular solution model. Later, Wynblatt and Hoffmann [43] used a monolayer segregation model with more accurately approximated total free energy, and this formalism was modified to include the prediction of possible compositional phase transitions [44]. However, as mentioned above, a more accurate description demands taking into account short-range order (SRO), as well as multilayer segregation. Free-energy approximate expressions that take into account SRO in the bulk of dilute binary [45,46], or multi-component [47] alloys were derived previously. However, their application to alloy surfaces is somewhat problematic, since upon segregation a solute can become a major constituent at the surface, thus violating the assumed low concentration. SRO correction for the binary alloy free energy, which is symmetric with respect to the alloy constituents, and thus overcomes this difficulty, was derived in the Ising model based "free-energy concentration expansion method" (FCEM) [1,3,19]. Being more accurate than the mean-field Bragg-Williams (BW) theory, and simpler to apply compared to the quasi-chemical and cluster variation methods, FCEM agreed quite well with MC simulations of segregation [19], while demanding much less computational efforts. Recently the FCEM approach was extended to

91

the case of alloys with any number of components [20]. An approximate SRO formula for multi-component alloy was constructed by adapting the corresponding binary alloy formula as a boundary case, and by making the multi-component alloy expression symmetric with respect to its components. The FCEM expressions for binary alloys were obtained using the Ising model Hamiltonian and an expansion of the partition function and free energy in terms of solute concentration [1,3,19]. The free energy of a binary alloy (A solute, S - solvent) reads,

A A F - k TE (c A In cA + cS In cS) + E Ahmcinm m 1 + - E VmASI2cA-1)(2c A - 11-

(2)

2 {mn}

- ~ kTc~m(1-cA)cA(1-cA)IexF(-2vAS/kT)+2vAS/kT-1 ) {mn} where

I cm

..l

is the concentration of a constituent I on a lattice site m, Ah~n

IJ

denotes a layer "field" (assuming that the lattice site m ~ p-layer), and Vmn is the effective pair interaction strength (see eq.1) between atoms of constituents I and J on lattice sites m and n. Rearranging the third term gives,

-il ~

AS mn( Acm,)( _

{mn}

cAa)_ _

(Vmn(CmC a .an + cAcB))+ 1 ZVmn {mn} 2 {mn}

The last, constant term can be omitted, and contributions related to the interaction Vm AS in eq.2 can be rewritten in a form symmetric with respect to the constituent concentrations.

F - k T E I cA

lnc A + c S lnc S

) +EAhmc A Am -

m

AS c A cS + c S Ac Vmn (mnmn)

m

+

-E A S A S (exF(_2Vmn AS /kT) + 2Vmn {mn}(+kTcmcmCnCn AS /tc/'-T-

(3) 1)

92

Generalization of this formula to multi-component alloys is straightforward [201, I I 2Ahmc mm,I :/:S

F-kT~2cllncI+ mI

)

CmCn CmCn . +

(4)

- {mn},{IJ}~~kTcmCmCnCn(exF(-2VIJn/kT)+2VIJn/kTJ I J Pair probability of finding atoms of types I and J on lattice site m and n is given by the formula,

plmJn - CmC I nJ + CmCnCmC I I J nJ(l_exp(_2VIJ/kT)) that coincides with the corresponding formula for a binary alloy [1,19]. In case of a ternary alloy (A,B - solutes, S - solvent) the free energy is given by, F - k T ~ ( c A In c A + c B l n c B + c S l n c S ) + m

A BB) +E (A Ahmcm + Ahmcm m r.,AB( A B B A'~ Vmn ~CmCn + CmCn J + VmAC( A S S A + n~CmCn+CmCn) l/maB( B S +CmCn S B 1+ n ~CmCn -

Z

{mn} kTcmcmC A B nA cnB( ex F(- 2 VAnB/kT 1 +2Vmn AB/kT - 1) + A S nA c nS( ex F(-2Vmn AS/kT 1 + 2Vmn AS/kT - 13 + kTcmcmC kTcmcmC n c n

ex

-

2Vmn

+ 2Vmn

(5)

93 Formulas within the BW mean field theory are obtained by omitting the

SROrelatedcontributionscontaining(ex~-2VIJ/kT)+2VIJ/kT-l]

from

eqs.3-5. The method was applied to the elucidation of effects of interatomic interactions and SRO on surface segregation in Ni-8%A1-4%Cu as a model ternary solid solution [20]. The results were then compared quantitatively to mean field calculations, and inspected in terms of the pertinent energetic parameters and effects of temperature. A primary consideration in choosing the Ni-A1-Cu system (Ni solvent) for model calculations was the relatively strong attractive Ni-A1 interactions (which lead to significant SRO effects on surface segregation in Ni-9at%A1 solid solution [ 1,3,19]). These effects are expected to be operative also in alloys containing a third constituent in low concentrations. Copper was chosen since Ni-Cu binary solid solutions (with quite weak repulsive interactions) had been extensively studied earlier and the corresponding energetic parameters are fairly known [48]. The energetics of the model was based on three nearest-neighbor (NN) interactions, V NiCu , V NiAI and V AlCu and two surface fields, Ah A1 and Ah Cu all listed in Table 1 In order to obtain the equilibrium layer compositions the free-energy (eq.5) was minimized numerically [20]. The alloy constituent concentrations calculated in the FCEM approximation for the first three atomic layers of the Ni-8at%A1-4at%Cu(111) surface are shown in Fig.4. A distinct surface phase transition characterized by a sharp jump in surface concentrations appears at 1075 K. Below this temperature the alloy surface is strongly A1 depleted and Cu rich, while at the transition A1 rises and Cu decreases, both reaching rather moderate segregation levels above it. The segregation behavior at all temperatures is indicative of site competition. ~

*

Table 1 Energetic parameters used in the model (in meV) v~C~ r~lc~ I~IN~ ahc~ AhAI -12.5" 31"* 136"* -120" -570*** *Ni-Cu energetic parameters were taken from Ref.48; v NiCu are enhanced at the surface by a factor of 1.5. **Estimation obtained from the heat of mixing [49] ***The surface field Ah Al for Ni-AI(100) has been determined as -680 meV [3], with -450 meV due to the difference in surface tensions [50]. Keeping the same elastic strain contribution, the estimation (-570 meV) takes into account the smaller nunaber of (111) broken bonds per atom (3 vs. 4).

94

Cu = 0.8

(p=O)

0.6

-""

-...

p=l

_

~'~ ~ ~

.~ .~

0.4 '"--..

p=2

"'""-..

i

i

~ 0.2

900

950

t .................................. ..... i. . . . . . . . . F .........

i

1000

1050 1100

r .......

1150 1200

""

1250

Temperature, K

0.16 ml

Surface (p=0)

.~ 0.12

e.~

p=2

0.08

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

./. 0.04

......

......-

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

/

~176176

....

900

"~176176176176176176

950

1000

A

1050

1 1 0 0 1150 1200 1250

Temperature, K

.; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....................... p=2

t

0.8 . -. - -. - - -.

. ." ~

~ .- ." -. ~

o 9

~

0.6

..--"r ~ 0.4 Z 0.2

900

Surface

950

1000

1050

1100

1150 1200

1250

Temperature, K

Fig.4. The near surface concentrations of fee Ni-8at O~AI-4at o~Cu(111) calculated in the FCEM approximation. (p=l and 2 correspond to the first and second under-layer, respectively.)

95

The predicted phenomena can be explained in terms of the different energetics and atomic coordination numbers involved. In particular, because of the reduced surface coordination, the Ni-A1 prominent mixing tendency is stronger in the bulk, leading to diminished surface concentrations of these constituents ("co-desegregation") throughout the temperature range below the transition. Consequently, in spite of its lower surface field (Table 1) the overall effective driving force for Cu segregation in the ternary alloy is significantly stronger than for A1, and even at temperatures above the transition the (diminished) surface enrichment by Cu is higher.

Cu 0.8 0.6 0.4

I FCEM

i

Ni-4at%Cu IlBw

0.2

-r .........

0 900

I

i

950

1000

I

p

i

I

i

1050

1100

1150

1200

1250

Temperature, K

0.2

Ni-$at%A1 FCEM

0.15

A1

I

......

i

r

BW

FCEM

0.1

0.05

0 900

____.~I 950

I

1000

i

~ 1050 ,

I

I

1100

1150

1200

1250

Temperature, K

Fig.5. Calculated solute surface concentrations for Ni-8at%A1-4at%Cu(111): thick solid lines - FCEM, thin solid lines - the BW-type approximation. Dashed-dotted lines - solute surface concentrations for the binary alloy Ni-Sat%Al(111) and Ni-4at%Cu(111) surfaces calculated in the FCEM approximation. Note the enhancement of Cu segregation induced by ternary alloying and short-range order effects.

96

The segregation of copper is further enhanced by short-range order that suppresses surface segregation of the solute (A1) interacting strongly with the solvent (Ni) [3,51]. Actually, SRO amplifies the interaction induced effects on segregation, without changing the general trend. Thus, as can be seen in Fig.5, the sharp transition in Cu surface concentration as predicted by the FCEM calculations occurs at a considerably higher temperature as compared to the results of the mean field (BW) theory that neglect interatomic correlations. Furthermore, it can be expected that the A1-Ni strong mixing tendency which diminishes surface concentration of both these constituents in the ternary Ni-8at%A1-4at%Cu alloy, would promote Cu surface segregation far beyond the driving forces operative in the corresponding binary alloy Ni-4at%Cu. Conversely, A1 segregation should be suppressed relative to its segregation levels in the binary Ni-8at%A1 alloy. The FCEM results for the binary alloys, shown in Fig.5, indeed exhibit below the transition temperature Cu surface concentration much lower (and A1 concentration much higher) than in the ternary alloy. To summarize this section, the multi-layer FCEM calculations predict strong segregation of Cu associated primarily with the Ni-A1 strong mixing tendency (attractive interactions) that effectively repels these constituents from the surface into the alloy bulk in an apparent site competition process. It appears to be operative also following a compositional phase transition, when the surface solute concentrations tend to be slightly below the respective binary alloy moderate segregation levels. Part of the former enhanced Cu surface segregation is associated with short range order effects that shift the transition to a higher temperature. These calculations can be further extended to other nominal compositions of this alloy, and the energetic parameters can be varied as to their general effects on site competition and surface phase transitions in ternary alloys. 3. SURFACE SEGREGATION IN ORDERED ALLOYS

Compared to SRO effects on surface segregation in solid solutions, the role of LRO should be naturally more prominent and common. Its elucidation requires calculations that take into account various factors contributing to the "net" segregation characteristics in ordered alloys including the temperature dependence: the crystal bulk structure and surface orientation, effective bulk and surface interatomic interactions (NN, non-NN) in relation to segregation driving forces, deviation from exact stoichiometry, possible surface relaxation and reconstruction, atomic vibrations, etc. This section attempts to quantify some of these factors and present several possible scenarios of segregation/order interplay.

97

Spatial ordering in the bulk of alloys and "classical" surface segregation in completely random solid solutions (without LRO or SRO) are both exothermic processes, which are enhanced at lower temperatures and accompanied by an entropy decrease. As discussed in our previous review [1] and mentioned in see.l, their interplay in ordered alloys can completely modify the segregation behaviour resulting either in endothermic or exothermic surface segregation, depending primarily on the energy balance of the respective tendencies. In the former case segregation is hampered, and an increase in its equilibrium level with temperature can be expected due to the enhancement of compositional disorder that disrupts the near-surface LRO, and is associated with increased configurational entropy. 3.1 Prediction of order/segregation interplay by means of a simple model As a first step, the interplay of surface segregation and long-range order in a binary alloy A~B~_~can be qualitatively evaluated by comparing the effective interaction strength (V) as a measure of ordering tendency with the "surface field" (Ah) reflecting the segregation basic driving force, similarly to the original approach of Moran-Lopes [2]. In this simple nearest-neighbour (NN)

pair interaction model, as the "segregation/order factor", r

1-71,

gets larger the

I - - I

balance tips more towards segregation. To obtain more quantitative estimation of the effects, r has been used as a parameter in FCEM calculations for two types of ordered structures with ideally equiatomic bulk truncated surface, assuming segregation limited to the three outmost atomic layers.

3.1.1 Equiatomic binary alloys Among possible equiatomic surfaces of equiatomic bulk alloys (e.g., B2(l10), B32(110), and L10(lll), see Fig.l) the calculations focused on bee B2(110). For low or moderate values of r (-~10) full monolayer is formed at low temperatures, and the segregation decreases with temperature monotonously (Langmuir-McLean type behaviour). The role of bulk off-stoichiometry is exemplified in Fig.7. Even slight negative deviations diminish considerably segregation levels, while positive deviations lead to strong enhancement relative to the levels calculated for the exactly equiatomic bulk. These somewhat surprising findings can be understood in terms of the dominant bulk ordering tendency, by which excess atoms (>50%) are effectively pushed out from the bulk (due to its reduced coordination,

98 ordering tendency at the surface is weaker). This strong dependence of segregation on small deviations f r o m the bulk stoichiometry should be taken into account in any analysis of ordered alloy segregation data (see below).

1.0

"

"

".:. :. .:.:.:.;. ~. .~. .~. ". . :. '. .i . . . . . . . . . .

[

12

0.8
Cb, the segregation vs.

temperature curve is not necessarily peaked. Equiatomic termination predicted for r 4.5) (Fig.8), but for lower r values (-~10), the behaviour resembles that of the previous class, namely, a full monolayer is formed at low temperatures, and then the segregation level monotonously decreases. Moreover, the diversity is manifested also by the Table 2 Relationship of surface/bulk transition temperatures calculated for L12(100) Segregation/order factor r r >2 7.8>r>4 8>r>7.8

Surface induced order/disorder or neither Ts < Tb (SID) Ts = Tb Ts > Tb (SIO) Ts = Tb

r>8

Ts < Tb (SID)

100

disordering temperatures of the surface (Ts) vs. bulk (Tb ). Depending on r, they can coincide, Ts can exceed Tb (surface-induced order, SIO), or be lower than Tb (surface-induced disorder, SID). SIO is promoted by surface compositions close to equiatomic (Fig.9), which correspond to intermediate values of the segregation/order factor, while SID occurs for high and low values of r, as shown in Table 2.

1.0 0.8
1). The B32 structure exhibits ordering also in the second coordination sphere, since NN and NNN interactions are comparable [76]. Therefore, it was treated assuming uniform interactions (V = V1 = V2 , Vn - 0 for n > 2 ). As can be seen from Table 3, the heat of formation (and the corresponding effective interaction strength) of alloys with B2 structure, except for FeA1, is considerably higher than of alloys with the B32 and the L10 structures (it is least exothermic for aluminides of metals close to group 6 [77]).

106 Table 3 Energetic parameters o f aluminum ordered alloys

Alloy

Structure/ Surface

SeA1

B2(110)

,n, era ,ion

Heat o f formation*, (kJ/mol)

strengthmevV**,

-84.6

218

Surface field Ah, meV 14

r = 0.064

CoAl

.

.

.

.

.

63.8

165

566

3.4

NiA1

.

.

.

.

.

67.3

174

455

2.6

RuA1

.

.

.

.

.

58.2

150

905

6.0

RhA1

.

.

.

.

.

89.3

231

674

2.9

.

.

.

.

.

FeA1 CrA1

B32(110)

28.6

74

660

8.9

-11.7

24

363

15

MoA1

.

.

.

.

.

22.9

47

746

16

TeA1

.

.

.

.

.

19.2

40

830

21

TiAI

-37.2

96

629

6.5

VA1

L 10(111) .

.

.

.

.

20.7

54

695

12.9

MnA1

.

.

.

.

.

23.7

61

124

2.0

* From Ref.75 ** Interaction strength in B2 and L10 is calculated in the NN approximation, while in B32 equal NN and next nearest neighbor (NNN) interactions are assumed.

I

1.0 o= 0.9 3

0.8

2

=o 0.7 1

0.6 0.5 0.4 600

\

I

I

I

I

800

1000

1200

1400

1600

Temperature, K . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Fig.15. F C E M calculated variations with temperature in the average AI concentration at the (110) surface o f bee aluminides (B2 structure - solid lines, B32 structure - dashed lines). 1 SEA1, RuAI, RhA1, NiA1, CoAl, 2 - FeA1, 3 - CrA1, 4 - MoA1, 5 - TeAl. The surface concentrations were calculated in accordance with data o f Table 3.

107

i i. ................... iiiiii

IJ 0.80 co

I ."

0.75

.~ .....

t-tO

o E g < o t~ 3:

=

O9

Z

0.70

.."

,.....'"!

.+.

0.65 .. ....

0.60

/

"'" ,.

0.55

.,'"

0.50 .... "

0.45 '

'

,tO

~

,'0

200

~.

. . .

~~

.......

, ............

i. . . . . . . . .

400

,

I

I

,

'

X

',~

(Vt3 X~3)R30 o

I

I

'CT~t'~

~.."

., "d order i '

:9~

/

1_= I

.

a

.

.

.

.

.

.

B(30

600

I

I}

I '

1000

Temperature (~

l0~ C

._o

8=

0,80-

T (100) 0,75- ~ (110) - --11--(210) 0,70 - - e - - (310)

o

0,65 -

E:=

0,60

=

0,55

8

o,5o

co

0,45 -

z

0,40

ca" a~

--0--

(111)

(210) (110

. 0

s~ 200

400

600

800

1000

Temperature (*C)

Fig.16. (a) AES determined near-surface (average) concentration of A1 as function of annealing temperature for the FeAI(111) surface (three datasets). The dotted lines estimate the uncertainty introduced by the error in the matrix factor. The phases, which are observed in LEED after quenching the annealed sample to room temperature, are also shown. (b) Comparison of the segregation curves for all investigated surface orientations. Near-surface concentrations corresponding to bulk terminated surfaces are marked by open circles [77]. These energetic parameters were used in F C E M calculations assuming segregation at the three outmost layers only. As shown in Fig.15, the segregation tendency prevails only in the B32 ordered alloys and the surface concentration decreases with temperature (entropy-driven monotonous desegregation). This behavior is associated with the distinctly high segregation/order factor (sec. 3.1). On the other hand, ordered bulk truncation with surface concentration very

108

slightly increasing with temperature is predicted for most of the B2 ordered alloys. Only in FeA1, with relatively low heat of formation and intermediate r value, there is a subtle balance between segregation and order, leading to a peaked segregation curve in experimentally accessible range of temperatures. Indeed, as measured by AES and LEED, the behavior of the FeAI(ll0) [77] differs substantially from the equivalent surfaces of alloys with NiA1 and CoAl like energetic parameters, which exhibit nearly perfect stoichiometry at the top layer (bulk truncation) [78-80]. The predicted surface segregation increase in FeA1 with annealing temperature was observed [77] also for other surface orientations and was accompanied by surface reconstructions (Fig.16). The calculations for the strictly stoichiometric FeA1 alloy predict somewhat higher segregation levels (Fig.15) compared to the reported formation of incommensurate FeA12 surface alloy on the FeAI(ll0) surface [77]. The discrepancy can be due, at least partially, to a slight deviation from stoichiometry in the measured alloy (see Fig.7). As an example for a third class of equiatomic aluminides, calculations done for three fcc L10 alloys are presented in Fig.17. Again, the segregation behavior is governed mainly by the segregation/order interplay, as expressed by means of r (see Table 3). Only in VA1 with relatively high r value (12.9), the segregation tendency prevails.

1.0 3

0.9

\

0

"~ 0.8

2

,D 0 r

0.7

1

\

0.5 0~

i

i

i

1

1

600

800

1000

1200

1400

1600

Temperature, K

Fig.17. F C E M calculated variations with temperature in the average A1 surface concentration at the (111) surface of fcc L10 aluminides. 1 - MnA1, 2 - TiA1, 3 - VA1. The surface concentrations were calculated in accordance with data of Table 3.

109 4. S e g r e g a t i o n in a b i - p h a s e b i n a r y a l l o y

As discussed in previous sections the involvement of ordering effects in binary alloy surface segregation complicates its theoretical treatment. Unraveling segregation phenomena in multi-component alloys is another challenge. But the situation can become even more complex for segregation in multi-phase alloys, when distinct segregation processes from individual bulk phases are coupled to the temperature dependent phase equilibria (Fig.18) In particular, in many binary alloy systems with ordering tendency bi-phase equilibrium exists between a solid-solution and an ordered compound when the bulk concentration exceeds the solubility limit (Fig.18). As discussed below, besides segregation~RO-SRO effects that can be operative in each phase separately, the variations with temperature in the solid-solution bulk composition can have a dominant effect and also lead to peaked segregation curves. Such a behavior, as measured by means of XPS, was reported previously for fcc-based A1-3%Ag alloy equilibrated between 550 and 770 K [82] (Fig.19a). Below the bulk phase transition (680K)hcp-based Ag2Al-like

I

77~

726

..--..

o

v

Ag

@O

611 $

==

~

I--

567

6:1.

76,.5-~..~..

2-phase 2D equilibrium a(surface) < >- 6(surface) Atomic seg.

Cluster~ seg. /

c~(bulk) 9 > 2-phase 3D equilibrium 0

AO

10

20

30

40

50

60

Atomic Percent AI

70

8(bulk)

80

90

100

AI

Fig.18. Phase diagram of A1-Ag [81]. Insert: schematics of processes pertinent to surface segregation in bi-phase alloys ( a - solid solution, 6;-ordered compound).

110

'

'

I

bI

2.50

13UP

2.40

9 DOWN

z 2.30 ,to

BULK TRANSITION

:

[]uP

i

"

2.20

0.50

'~

-

0.40

i

"q

0.20 570

670

770

TEMPERATURE (K)

Fig.19. The XPS bandwidth of the Ag 4d states in A1-3at.%Ag (top) and the Ag concentration (bottom), deduced from the emission intensity, as a function of temperature [82]. The bulk phase transition lies at 680 K. clusters (~-phase) precipitate in the fcc solid solution a (Fig.18). Evidence for the appearance of (111) surface clusters came from secondary electron imaging (SEI), Fig.20. In addition, changes in the Ag 4d linewidth (Fig.19b) were attributed to varying numbers of Ag neighbors of a given Ag atom, and thus were supposed to reflect the relative extent of clustering vs. Ag dissolved in the A1 matrix. Based on these data, the three regions in the segregation curve (Fig.19a) were tentatively attributed to: i) Segregation enhancement of small Ag2Al-like clusters with increasing temperature; ii) Their gradual dissolution (first-order phase transition), without a change in the overall concentration of Ag atoms in the analyzed volume (610-690K), and iii) Ag atom desegregation at higher temperatures. Recently, an attempt was made to analyze the compositional changes in a quantitative manner and so to elucidate the pertinent mechanism in terms of the

111

Fig.20. Secondary-electronimaging (SEI) pattem obtained from epitaxial Ag on AI(111) heat treated at 410 K. The sixfold symmetry verifies the formation of Ag2AI clusters with hcp structure [82]. two-phase bulk equilibrium as well as the involvement of two distinct segregation routes, that of atomic Ag and of Ag2A1 clusters [83]. In principle, since the formation of ordered phase clusters at lower temperatures is accompanied by a reduction in Ag solute concentration in the bulk of the solid solution, surface segregation from the latter is suppressed. As temperature increases, gradual dissolution of bulk clusters results in increase in bulk and surface Ag concentration of the solid solution matrix. Around the phase transition (crossing the solubility line), when the solid-solution composition becomes constant with temperature (3%Ag), the surface concentration starts to decrease monotonously as is common in random solid solutions (McLeanLangmuir entropy driven desegregation). More quantitatively, the Ag concentration c a of the bulk solid solution can be simply evaluated from the relevant portion of the A1-Ag phase diagram (Fig.18). For a given Ag overall atomic concentration ( c ) , c a increases with temperature (concomitantly with decreasing amounts of the d-phase clusters) according to the solubility line approximate formula c a = A exp -

.

(6)

112

0.3

0.04

0.25

0.03

0.2

0.02

0.15

0.01

O

~D O w o

~9

iv'

0.15

1.0

o.lo

C)

E

0.8

0.05

0.6 0.00

100

,

i

200

,

,,

I

300

i

I

400

m

I

500

,

I

600

i

l

700

i

I I

800

i 0

1

900

1000

Temperature(K) Fig. 11: ISS - XPS of Cu(100)-Ir: Amount of Ir as function of the annealing temperature: in the topmost layer (left axis; circles) as determined from ISS in comparison with the amount of Ir in the first few layers as deduced from XPS (ratio of Ir 4f and Cu 3p levels, right axis; squares). Initial nominal Ir coverage: 1.5 ML. (from ref. [93]).

382

Simultaneously, massive structural changes directly show up in the STM data after sample annealing. On a larger scale, the complete disappearance of the small ad-islands (seen in Fig. 10) can be recognized in Fig. 12. Obviously, the intermixing process flattens out the entire Ir-Cu surface. The surface visible on individual terraces gives the impression in STM of being almost structureless, except for special tunneling parameters. By tunneling into occupied states of the sample, weak shady depression features become visible on the entire area (Fig. 12). The onset of ordering might be recognized already in small areas of limited size. Considerably better ordering of these shady features have been obtained by evaporation of Ir directly onto the heated sample rather than by subsequent annealing.

Fig. 12: STM of Cu(100)-Ir: 1.5 ML Ir evaporated at 200 K followed by 30s annealing at 650 K. (Image size: 250 nm x 250 nm, Ut~p= 0.2 V). STM image taken at 300 K.

A careful analysis of the novel depression structures can be performed best, by first studying these features for just a few Ir atoms at the Cu surface. An STM image with atomic resolution is presented in Fig. 13a as measured after 0.05 ML iridium deposition at 200 K followed by post-annealing at 650 K. Two basic features can be emphasized, the appearance of an ordered array of white dots and additional star-like depressions which are irregularly spread over the displayed surface area. The array of dots in Fig. 13a has been identified as the location of first layer Cu atoms. Further on, it was verified by corresponding ab initio calculations by Heinze at al. [102] that no bias voltage dependent corrugation reversal as e.g. predicted for W(110) [ 103] occurs on the Cu(100)

383

surface. The apparent height of the star-like depressions is measured to about 0.03 nm depending strongly on the applied tunneling gap voltage. A closer inspection of Fig. 13a reveals, that the centers of gravity of the 'stars' are not located on regular lattice sites of copper atoms in the first layer, but instead in fourfold-hollow sites. Excluding interstitial positions for Ir, the star-like characteristics have to be caused by iridium atoms located at regular lattice sites below or on top of the surface. The latter has been excluded by the ISS measurements reported above. Another distinct property can be recognized in Fig. 13a; sometimes two iridium atoms in the second layer are coming close in a next neighbor configuration, as a result the imaged overlap of this situation manifests itself in the onset of a stripe formation (visible along the [011] direction e.g. in the lower right comer of Fig. 13a).

Fig. 13: a) STM of Cu(100)-Ir: 0.05 ML deposited at 200 K and subsequently annealed at 650 K (Image size: 5 nm x 5 nm, Ut~p- -0.02 V) b) STM image of the ordered surface alloy: 0.6 ML Ir deposited at 620 K (Image size: 10 nm x 10 nm, UTip- 0.3 V). STM images taken at 300 K. (from ref. [93]). After direct deposition of higher doses of Ir (0.6 ML in Fig. 13b) at elevated temperature, long range ordering occurs which can be seen as well in LEED by exposure of a (2xl) superstructure with two domains. The corresponding STM image (Fig. 13b) exhibits a distinct chain like structure with chains running in the [110] directions. The distance between adjacent chains is measured to 0.5 nm, which is about twice the distance of nearest neighbor Cu atoms and thus in good agreement with the (2xl) LEED superstructure. Successful imaging of the striped structure by STM was just possible in a limited range of tunneling gap

384

voltages indicating electronic effects for the origin to measure the stripes. The best ordering of the chains with domain sizes of about 5 nm (calculated from the full width at half maximum of the LEED superstructure spots) have been reported for 0.5 - 0.6 ML Ir deposition on the 620 K hot Cu(100) surface. By following the discussed experiments, a model for the Ir-Cu(100) surface has been suggested and is presented in Fig. 14. As a matter of fact, a two dimensional epitaxial sub-surface alloy has developed and consists of adjacent chains of Ir and Cu atoms along the [011 ] directions to form an ordered (2xl) periodicity. The Ir-Cu sub-surface layer happens to be buried under a monolayer of copper. Remarkably enough, although the surface crystallography of Cu(100) expresses four-fold symmetry, a two fold symmetric pattern is showing up for the chains of subsurface Ir to resemble the (2xl) superstructure.

Fig. 14: Structure model of the ordered Cu(100)-(2xl)-Ir sub-surface alloy. (from ref.

[93]). Hence, on the first sight the proposed model of an ordered sub-surface alloy might appear somewhat surprising because of three facts: firstly, the large miscibility gap in the bulk phase diagram [90], secondly the formation of a (2xl) periodicity on a quadratic surface lattice and finally the assumed possibility to image the buried layer itself by STM. Stimulated by the experimental findings, the Cu-Ir system has been investigated theoretically by Heinze et al. [102] with the help of ab initio calculations. In a first step, the existence of a Cu-Ir sub-surface alloy has been verified via electronic structure-, total energy- and force- calculations by a full potential augmented plane wave method (FLAPW) in bulk and film geometry [104]. For the determination of the alloy structure the surface near region was modeled by nine layers of Cu and

385

one layer containing Cu and Ir atoms placed on both sides of the film either at the surface, sub-surface or deeper layers. For the low coverage Ir situation, impurities were introduced in a p(2x2) surface cell, whereas for the ordered alloy structure a p(2xl) superstructure was assumed. These surface structures are compared with results from a c(2x2) model [4], which often occurs on fcc (100) surfaces (see also the chapter by Colin Barnes). The theoretical outcome explains well that Ir located at the surface is the most unfavorable configuration. This result is again consistent with the fact that the surface free energy for Ir is higher than for Cu and so the overall energy would be lowered by a Cu surface termination. Additionally, the computation revealed in accordance with the presented STM analysis that Ir located in the sub-surface layer presents indeed the energetically most stable configuration. This has been interpreted in terms of the bonding situation: the bond strength of Cu-Ir is expected to increase with the reduction of nearest neighbors in the CuCu environment. Accordingly, among all Cu atoms, the Cu atoms at the surface form the strongest bonds to Ir atoms and the equilibrium position of Ir is found in the sub-surface layer and thus prevents Ir to segregate into deeper layers. From the calculations it turned out as well, that the p(2xl) chain structure at the sub-surface location is 86 meV per Ir atom more favorable as compared with the c(2x2) array of Ir and Cu atoms, which is basically due to directional forces of the straight d-d hybridization between Ir atoms along the chains. These forces are obviously absent in a c(2x2) situation. For the p(2xl) Cu-Ir structure, an energy increase of 49 meV has been determined before segregation of Ir into deeper layers sets in. Such an energy barrier can evidently be overcome by temperature augmentation. Therefore, the experimentally observed diffusion of Ir at T > 650 K into the bulk (cf. Fig. 11) becomes plausible too. In order to estimate the topographic influence of the Ir-Cu structure on the STM data, additional force calculations have been performed by minimizing the total energy. As a result a buckling Az of the Ir vs. Cu atoms of Az/d = 2.9% of the interlayer distance d has been found, which should give rise to a corrugation amplitude in STM topography of less then 5 pm (protrusions for sub-surface Ir atoms). Evidently, pure topography marked by this small height variation (additionally of wrong direction) cannot explain the measured depressions of 30 pm (Fig. 13). In a next step the possibility to image sub-surface impurities in metal surfaces by STM has been investigated [ 102]. The STM images were calculated for room temperature in the Tersoff and Hamann [105] approximation to determine the tunneling current I(r, U) for a gap voltage U. The local density of states (LDOS) of the sample is expressed in n(rll, z, ~F + ~) [ 106] at the position

386

lateral (rl~) and normal (z) to the surface with the Fermi energy ~F. g U,T (~) represents the difference of the Fermi functions f r at (OF- eU + c) and (eF + ~) [106]. 1(1"11,z, U) oc f g ~v (~) n(rls, z, ~'F + ~) d~" In order to describe the wave functions decaying from a single crystal surface into the vacuum, the FLAPW method has been applied which gave the justification to expand the wave functions into 2D basis functions as symmetrized plane waves parallel to the surface (so called 'star' functions ~b,.) with their corresponding z-dependent 'star' coefficients n~:

n(?'ll, z, ~ = Zrti(z, s ~i(?'ll) By this procedure the corrugation amplitudes Az(z, U, n~, n2) as a function of the tip location above the sample surface has been determined [102, 103]. The corresponding 'STM images' consist basically of the information expressed in the first two coefficients n~ and n2, where ~bl is a constant and does not contribute to the STM corrugation pattern. The height modulation of the probe as a function of the tip position is basically determined by ~b2with the sign and strength being settled by the positive or negative n2 coefficient. Fig. 15 represents the calculated STM images for the impurity (p(2x2)) and chain (p(2xl)) structures for Ir buried by the Cu(100) monolayer surface.

Fig. 15: Calculated STM images at UT~p= 0.6 V, z = 0.5 nm for Ir impurity a) and chain b) located at sub-surface locations. Open (full) circles represent Cu (Ir) atoms, big (small)

circles represent atoms at surface (sub-surface) (from ref. [102]). In Fig. 15, obviously the formation of a star-like structure for the impurity case (Fig. 15a) of a single Ir sub-surface atom and the onset of a chain structure

387 for the (2xt) superstructure (Fig. 15b) can be recognizes as depressions (dark). These theoretical predictions nicely reproduce and rationalize the experimental findings for tunneling in the filled states of the sample. The influence on the sampled LDOS profile by changing the bias voltage U is calculated and the obtained corrugation amplitudes are compiled in Fig. 16. The condition of buried Ir impurities and chains can be recognized in the lower part of Fig. 16. A measurable height variation is expected at bias voltages a r o u n d - 0 . 5 V (occupied states in the sample), the corrugation amplitude of about 0.03 nm manifests itself as a depression and is in excellent agreement with the measured data. On the other hand, from the upper part in Fig. 16 it turns out, that Ir atoms or chains being located in the first surface layer would be imaged as protrusions of comparable amplitude. The experimental STM data obtained at Ir impurities (Fig. 13a) undoubtedly excludes the latter occurrence. As a remainder, all calculations resemble only the influence of the electronic effect on the STM data because it has been established above, that ~the topography is not much altered by the substitution of Cu atoms by Ir.

Fig. 16: Calculated corrugation amplitudes of a tip at z = 0.53 nm, as a function of the applied bias-voltage U for the Ir impurity and the Ir chain. In the insets at the upper and lower right corners, filled (open) circles denote Cu (Ir) atoms. Positive (negative) corrugation amplitudes are defined as imaging the Ir site as a protrusion (depression). (from ref. [102]).

Moreover, the charge density distribution above the surface of the buried Ir atoms could be calculated, and by that inferring a correlation with the actual bond situation. The measured STM corrugation was correlated with the variation of the n2 coefficient in terms of the theory. A charge density contour plot based on the calculation of n2 for the Ir chain in the second layer at an energy of 0.6eV below the Fermi energy of the alloy is given in Fig. 17.

388

The hybridization of the Ir d- states with the Cu sp- states yields in tilted pd- orbitals located at the neighbored Cu atoms. Because of the tilt, the charge density maximum, which is for pure Cu(100) right above the Cu surface atom, shifts to the position above the Cu sub-surface atom. As a consequence, in case of a fcc(100) surface, the charge density depletion above buried Ir atoms in combination with a higher intensity aside of the first layer Cu atoms results in the star-like pattern obtained in the experiment (Fig. 13a) and theory (Fig. 15)

Fig. 17: Cross section along the [100] direction through the charge density of a typical state in the 2D Brillouin zone at E = EF- 0.6 eV for Ir chains at the sub-surface location. White (black) color denotes a high (low) charge density. (from ref. [102]).

In addition, the theoretical investigation of the special situation of Ir on a Cu surface was used to compare with possible sub-surface conditions at other transition metal cases. On the basis of the calculations, the hybridization of the Cu sp- and Ir d-states are expected to be of rather general quality and therefore the prospect to detect buried transition-metal atoms by STM should be valid for other couples too. A number of possible candidates have been suggested [102]. For sure, all of the propositions rely on the assumption that buried layers might be created indeed experimentally, which probably will be difficult because of decent miscibility behavior for some elemental pairs recommended below. With decreasing number of d electrons (Ir, Os, Re, W, Ta) it is expected by theory that the d band energy increases with respect to the Fermi energy and therefore the tunneling barrier of the state seen in STM becomes lower. Therefore the corrugation amplitudes are supposed to increase up to 0.05 nm and the subsurface location of these impurities in Cu should turn up even clearer in STM. Also Rh as an iso-electronic pendant of Ir is expected to yield measurable corrugation. Conversely, larger numbers of d- electrons (Ir, Pt, Au) lead to smaller height differences of less then 0.01 nm. This might be the reason why

389 the STM investigation failed to image the sub-surface growth of Pd (being isoelectronic to Pt) in Cu(ll0) [107]. On the other hand, sub-surface alloy formation has been reported for vanadium on Pd(111) and the position of the sub-surface V atoms forming a (~/3 x ~/3)R30 ~ arrangement could be observed in the STM data as depressions appearing in the Pd layer [ 108].

3.2 Intermixing versus phase separation: Copper on Ir(100)-(5xl) In the preceding section it has been confirmed that intermixing occurs for certain bulk immiscible constituents (A) and (B). The question may arise, whether this phenomenon depends on the preparation sequence to evaporate material (A) on substrate (B) in comparison with (B) on substrate (A). With the materials (A) = Ir and (B) = Cu, the mixing properties have been verified for the first situation and were tested afterwards for the reversed order. Of course intermixing is just one option of the system to react. As an alternative way, a clear-cut (perhaps two dimensional) phase separation between the two elements might happen. Another complication may arise because of the more complex structure of the substrate, to be exact, Ir(100)-(5xl) as compared with Cu(100)(Ix1). Apparently, the aspect of a possible lifting of the surface reconstruction upon Cu deposition has to be considered, also since it is known that even small energy variations in the surface e.g. by temperature increase or gas adsorption already might induce a lifting of the surface reconstruction [ 109].

Fig. 18: a) STM image of the clean Ir(100)-(5xl) surface taken at 300 K (Inset: corresponding LEED pattern, E = 180 eV). Image size: 62.5 nm x 62.5 nm. b) STM of clean Ir(100)-(5x 1) with atomic resolution. Image size: 6.5 nm x 6.5 nm. (from ref. [110]). The characteristic (5xl) reconstruction of the clean surface expresses after careful cleaning [ 110] and has been explained in a model structure by coverage of an fcc(100) surface with a quasi-hexagonal close packed monolayer of the Ir

390 atoms ontop [111, 112]. Due to the quasi-hexagonal packing of the first layer, the density of the top layer has to be 20% higher as compared with the fcc(100)( l x l ) surface. Accordingly, the surface layer is marked by a characteristic height modulation leading to the (5xl) periodicity visible in LEED experiments. In the STM image of Fig. 18a the typical corrugation appears as a stripe pattern with lines running parallel to the [011] directions ('reconstruction lines'). Two quasi-hexagonal domains with an orientation rotated by 90 ~ show up and have been found on terraces as well as separated by step edges at adjacent terraces. In Fig. 18b the stripes are measured with atomic resolution. The typical 'double row' height modulation has been attributed to the two-bridge configuration [110] in agreement with LEED I-V investigations [111, 112] and theoretical predictions [ 113 ].

Fig. 19: a) STM constant-current image after deposition of 0.2 ML Cu on Ir(100)-(5• at 300 K. Image size: 100 nm x 100 nm. b) Side view model of the Ir(100) surface before (upper panel) and after lifting of the (5x l) reconstruction due to deposited Cu atoms. The formation of Ir chains embedded in the Cu layer after lifting of the surface reconstruction is illustrated. (from ref. [ 110]).

Deposition of Cu on the Ir surface leads in ISS immediately to an increase of the peak for He scattering at Cu atoms. It could be concluded, that all of the Cu atoms stay ontop and strict 2D layer growth was found up to a coverage of 0.7 ML [110]. At higher coverage, 3D islands start to grow and can be seen with STM. By knowledge of the surface composition, the growth mechanism was followed up in STM measurements. The initial growth of Cu at room temperature on the reconstructed (5x l) Ir(100) surface appears to be strongly

391

influenced by the reconstruction lines of the (5xl) structure on the individual terraces (Fig. 19a). Cu starts to grow by formation of small chains and islands (marked in Fig. 19a) with a preference along the reconstruction lines. In the STM image of Fig. 19a, essentially three different height levels exhibit, connected with surface areas of Cu islands- (bright) and Ir chains- (bright lines), Ir(100)-(5xl)- (medium gray) and Ir(100)-(lxl)- (dark), respectively. In a more specific examination of the STM topographic images it appears that that the intense chains indeed continue straight on across the bright islands as faint dim lines. Similar depression structures have been already successfully identified above as Ir atoms embedded in the Cu surface (cf. section 3.1). Therefore, the bright lines on the Ir terraces as well as their continuation as darker lines in the Cu islands have been recognized as Ir chains resulting from surplus Ir atoms caused by a lifting of the Ir-(5xl) reconstruction. In a cross section image, a model of the evolution to grow the structure is explained in Fig. 19b. The decomposition proceeds as follows" In a side view the Ir atoms of the reconstructed surface (upper panel) residing in the upper corrugation sites (B, D, H, J, N, P) are marked by gray shading. Indeed among all Ir atoms of the first layer, these marked atoms are expected to have the highest chance to be shifted up in the decomposition mechanism. Consequently, the lifting of the (5xl) reconstruction will lead to the formation of one Ir chain at each (5xl) super cell exposing on a larger scale three distinct distances between the chains of 3, 5 and 7 lattice units (cf. Fig 19b). Actually, Cu deposition on Ir(100)-(5xl) at 300 K induces right away the lifting of the reconstruction. The reconstruction is not only lifted in the Cu islands but the action proceeds onto pure Ir terraces. The 20% Ir surplus atoms pop up onto Ir terraces and into the top Cu layer to remain there as embedded Ir chains. This structure might be viewed as a kind of dilute 2D ordered surface alloy. It turns out that the long Ir atom strings in the 2D Cu matrix behave rather fragile against temperature increase and they mark just a metastable situation of this special alloy layer. Surface annealing at about 1000 K influences not much the surface composition as can be monitored by ISS, AES or XPS. However, the surface structure changes dramatically. Indeed, the different surface energies of Cu and Ir play at this point again a dominant role and may explain the effect: Ir strings in the first layer expose long border lines, hence in order to minimize the length of the rims, the Ir chains start to transform themselves into compact round shaped 2D disks via mass transport within the surface layer. In case of the low coverage deposition of Cu, the conversion can be seen directly in Fig. 20a. Four characteristic surface features (A, B, C, D) show up in Fig. 20a. Upon higher pre-coverage at 300 K some 3D islands of Cu have already developed (cf. fig.9

392 in ref. [110]) and finally after annealing an additional gray level (E) can be noticed in the corresponding STM survey topograph (Fig. 20b).

Fig. 20: STM images of Ir(100)-(5• 1) after deposition of Cu at 300 K and subsequent annealing at 1000 K: a) Cu deposition: 0.3 ML. Image size: 250 nm x 250 nm. b) Cu deposition: 0.9 ML. Image size: 250 nm x 250 nm c) barrier height image, d) corresponding STM topography - Cu deposition: 0.9 ML, image size: 100 nm x 100 nm. (from ref. [110]).

After all, five different surface species have been identified: namely A) as the clean ( 5 x l ) Ir surface layer, B) as the clean ( l x l ) Ir surface layer, C) as the pseudomorph C u ( 1 0 0 ) - ( l x l ) overlayer on unreconstructed Ir(100), D) as embedded Ir islands in a Cu(100)-(1 x 1) matrix and E) as Cu ad-islands on top of such embedded Ir areas. In fact, surface annealing enhanced the effect of phase separation and the weak alloy formation found before in the occurrence of Ir chains in a Cu monolayer at 300 K is completely overruled by the formation of compact separated areas of Cu and Ir content, respectively. In addition, depending on the initial coverage, the bare surfaces of Ir islands

393 introduce the tendency to cover themselves up by a monolayer of Cu, in order to minimize the surface free energy. By performing local barrier height measurements in the usual way [114], evidently a chemical contrast of the Ir and Cu areas has been achieved (Fig. 20c). The island types A) and D) -Ir, C) and E) -Cu are indicated in the barrier height image and can be recognized in the simultaneously recorded STM topograph given in Fig. 20d. A correlation between Ir- (A, D) and Cu- (C, E) areas with the corresponding barrier height image has been established by comparing the parts of high surface barrier (white areas, (A, D)- Ir) and low barrier (dark gray, (C, E)- Cu). Moreover, this association is in good agreement with the trend, that barrier heights of not too small surface areas correlate with the values of the macroscopic work function ~b, (~b CuaOO)= 4.6 eV, ~bI r ( l O O ) - ( l x l ) - 5.5 eV and ~bIraOO)-(5~)= 5.4 eV [91, 115]). Furthermore, from the barrier height image in Fig. 20c it becomes evident, that all Cu islands E) are surrounded by white rings, indicating that the Ir island beneath is not completely covered by the Cu atoms. A possible Smoluchowsky effect at the step edges [116] was excluded by direct comparison with barrier height measurements at Cu islands on Cu(100). The incomplete coverage of Ir islands with copper can be explained by the requirement of energy to generate Cu steps and by surface strain which builds up due to the different lattice parameters of Cu and Ir. Probably, the gain of energy by covering the Ir areas does not completely outweigh the energy expense due to Cu island formation on the strained area in combination with the island border line. Similar depletion rings due to substrate strain have been rePorted for oxygen adsorption on Ru(0001) [ 117]. To recapitulate this part, the effect of intermixing for immiscible constituents depends indeed sensitively on the preparation order. For the described system of Cu on Ir, at room temperature a kind of dilute mixture of Cu and Ir can be assigned, basically as a result from the lifting of the surface reconstruction of the Ir substrate. The related release of 20% Ir surplus atoms is being incorporated as atomic strings in the Cu islands. This surface configuration turns out to be metastable and is completely transformed after sample annealing into phase-separated areas of compact Ir islands in a 2D Cu matrix. Whereas intermixing occurs for Ir on Cu, strict phase separation develops for the reversed system of Cu on Ir. 4. A L L O Y SURFACES AS SUBSTRATES FOR ORDERED SUPERSTRUCTURES

The effect of intermixing and phase separation has been applied in a different approach for the creation of additional ordered heterostructures by

394

adding a third component to the system. Hence, either alloy formation of two materials on a single metal substrate [118] or alternatively, alloy formation ontop of a binary alloy system can be discussed. An easy way to use phase separation has been followed up by oxidation of special alloy surfaces. In particular the NiA1 surfaces are of great popularity for creation of thin A1203 surfaces. These structures serve as substrates for catalytic reactions in form of nanostructured oxide arrays [37, 47, 119] or as thin films for catalyst support [114, 120-127]. Recently Franchy published a comprehensive report on the formation of thin oxide structures on several alloy surfaces [28]. In a further step, the alloy surface itself was taken as a substrate for epitaxial film growth of another metal-film material. Among others, the Cu-Au alloy system turned out to yield promising results. This is primarily because several ordered alloy phases Cu• as function of the composition are known to exist [90] and which can be employed to vary the substrate lattice parameter in the range from 0.3614 nm (pure Cu) to 0.4078 nm (pure Au). Such substrates have been proposed for lattice mismatch tailoring in epitaxial metal film growth to make use of the sequence Cu --~ Cu3Au --~ CuAu --~ CuAu 3 ~ Au. In particular the investigation of magnetism for dimensionally reduced systems has proven that the magnetic properties depend strongly on the film stress and morphology. For example, both, ferromagnetic and antiferromagnetic characteristics have been considered for thin fcc iron films depending on the lattice constant in the film. As a consequence, in order to vary the substrate lattice starting with Cu, also Cu3Au(100 ) surfaces have been utilized for epitaxial film growth of Fe or Ni [71-74]. For Fe deposition, a miscibility gap for Cu and Au occurs for bulk material, whereas Ni seems to be miscible with Cu and Au. As an example for the growth of thin films in the two limits of intermixing or phase separation, the case of vanadium on Cu3Au (100) will be shown below. 4.1 Vanadium on Cu3Au(100 ) For the binary system of bulk vanadium and gold, several ordered alloys are known to be present from the corresponding phase diagram [90]. Obviously, for V on Cu3Au the situation for intermixing is fulfilled. After vanadium evaporation on the Cu3Au (100)-c(2x2) surface at room temperature, both, direct clustering and surface embedding of the vanadium atoms shows up in STM. Simultaneously, the fractional spots in LEED vanish at low coverage and after further exposure the remaining ( l x l ) spots become gradually weaker. In the end, no clear LEED pattern is obtained beyond a vanadium coverage of three monolayers. At this stage a highly disordered and rough V surface can be recognized by STM and ion scattering. On the other hand, annealing initiates

395 directly an ordering process and after appropriate heating at a temperature of about 500 K, the c(2x2) superstructure recovers completely. The induced intermixing and surface segregation has been investigated by ion scattering spectroscopy. The corresponding TOF measurements are presented in Fig. 21. A three-monolayer thick vanadium film was deposited at 190 K. From the ISS data the complete coverage of the substrate by V can be deduced, visible by the lack of the ISS peaks for Cu and Au (upper TOF spectrum in Fig. 21). He~V A _

190K

He ~ Cu I [ ~ I He ~ A u I I I I

220K 270K 320K t-

"*-'

fie

370K

0")

._ '-

420K

0 v.

470K

I

I

I

I

I

I

I I I

I I I

F--t,

..Q

520K 570K

4.0

4.5

I I I 5.0

5.5

6.0

6.5

Time-of-flight [ps]

Fig. 21: Ion scattering of He at a 3 ML thick vanadium film on Cu3Au(100). E0 = 3 keV. V was deposited at 190 K, the film was subsequently annealed for 30s at the indicated temperature. ISS data acquired at 190 K. The positions for single scattering flight times of He at V, Cu and Au are indicated. Scattering angle 180~

After annealing at about 450 K first surface segregation of Au can be detected. Upon annealing around 550 K an alloy of V3Au stoichiometry has formed by vanadium atoms to substitute Cu positions. The related LEED pattern reveals the c(2x2) superstructure indeed demonstrating the development of an ordered V3Au (100)-c(2x2) surface alloy. Annealing at higher temperature, leads successively to the formation of a ternary Cu•215 alloy, with x running

396

from 0 at low temperatures to 3 at a temperature of about 800 K. Annealing the ternary alloy at 800 K for longer periods initiates segregation of the entire vanadium layer into the bulk, by that leaving the bare Cu3Au(100)-c(2x2) surface behind. As a result, V deposition on Cu3Au triggers the formation of a substitutional ordered surface alloy by reason of intermixing and surface segregation which, on the other hand can be stopped completely by precovering the Cu3Au substrate with oxygen as will be shown next. 4.2. Vanadium oxide o n C u 3 A u ( 1 0 0 ) - O An appropriate oxygen treatment of Cu3Au(100)-c(2x2 ) has been already described in section 2.2.2. After O + implantation, a flat C u 3 A u ( 1 0 0 ) - c ( 2 x 2 ) - O surface has been established upon annealing at 650 K. The smooth oxygencopper surface layer acts positively in two ways" firstly, it prevents completely intermixing. Secondly, in contrast to a pure Au or Cu crystal, the C u 3 A u sample may proceed as an oxygen reservoir with sub-surface oxygen stored close to the surface, which might be released in a controlled way via temperature treatment of the sample.

Fig. 22: a) high resolution STM image after deposition of 0.1 ML V on Cu3Au (100)-O at 300 K. Positions of individual V atoms are seen as depressions by chemical contrast. Image size: 10 nm x 10 nm; (UT, = -0.35 V, i = 1.0 nA). b) STM survey for higher coverage of 0.6 ML vanadium. Image size: 100 nm x 100 nm; surface wetting of the film can be seen by strict 2D growth.

Evaporation of small quantities of V onto of the Cu3Au (100)-O surface can be monitored directly in the STM image. As a consequence of a strong chemical contrast, tunneling into the empty states marks the position of V atoms at the surface by dark spots, i.e. strong depressions of an apparent depth of about 0.04 nm are visible in Fig. 22a at a tip voltage o f - 0 . 3 5 V. Evaporation of higher

397

quantities of V leads to the formation of a 2D film resulting in no ordering effect at all, neither visible in STM (Fig. 22b) nor in LEED. Ordering of the V film can be just achieved upon annealing and by that oxidizing the vanadium layer in a controlled way. After the preparation of three different initial Cu3Au-O substrates, distinct VO• layers have been generated. All three initial Cu3Au-O substrates provide the same c(2x2) LEED superstructure, but the modification has been obtained by generating dissimilar contents of sub-surface oxygen (low, medium, large). Depending on the preoxygen contents, three different ordered layers of vanadium oxides have been prepared on Cu3Au-O by oxidizing the room temperature deposited V films through annealing in the suggested manner. Indeed, it was possible to produce flat epitaxial and uniform VOx films [88].

Fig. 23: Vanadium oxide layer obtained by vanadium oxidation at a CuaAu(100)-O substrate with medium oxygen content, a) LEED pattern b) survey STM c) Schematic model of the V203 oxide structure. Large white spheres: oxygen; small dark gray spheres: lower half part of the vanadium double layer; small light gray spheres: upper half part of the vanadium double layer, d) High resolution STM. (from ref. [88]). Starting with the sample of low oxygen content, the vanadium oxide structure that is obtained after vanadium-oxidation by substrate annealing

398 consisted of a quadratic unit cell. A homogeneous film covers the entire surface. By means of SPA-LEED measurements the lattice parameter of the oxide was determined to result with 0.28 nm in a slightly larger distance as compared with the 0.26 nm of the Cu3Au substrate. From the knowledge of the crystallographic structure (STM, SPA-LEED) and the rather low oxygen content (AES), this oxide configuration has been correlated with a layer of vanadium monoxide VO(100) carrying vanadium in the V 2+oxidation state. Another oxidation state (V 3+) of vanadium in the VOx film was produced by employing the Cu3Au-O substrate with medium oxygen content. In variance to the quadratic lattice structure of vanadium monoxide, the LEED superstructure of this specific VOx phase give rise to a ring type diffraction pattern (Fig. 23a), which can be best explained by the occurrence of two domains of a hexagonal lattice structure. Indeed the high resolution STM image of Fig. 23d displays a single type of the domains measured at one of the large flat oxide terraces (Fig. 23b). On the basis of the hexagonal structure in combination with the measured lattice constant of 0.522nm, a good structure fit was obtained under the assumption that a V203 type oxide has developed. As a result of the epitaxial relationship, it was concluded that the oxide film forms a V203(0001) surface plane, indeed quite comparable with epitaxial Cr203(0001 ) [128] films grown on Cr(110).

Fig. 24" Vanadium oxide layer obtained by vanadium oxidation at a Cu3Au(100)-O substrate with high oxygen content, a) STM survey showing the different domains, b) Schematic model of the VO2 oxide structure. Large white spheres: oxygen, small dark spheres: vanadium. Unit cell is indicated in b) and c). High resolution STM in c) and d) with different tip configurations. (from ref. [88]).

399 Vanadium in the V 4+ oxidation state has been created too in a two dimensional film configuration. Here, vanadium has been oxidized on the substrate with high oxygen content. This type of vanadium oxide manifests itself in a smooth thin epitaxial layer where several domains can be recognized in a striped oxide pattern (Fig. 24a). The related LEED pattern had changed from a hexagon to a superstructure consisting of 90 ~ angles between the unit cell vectors. Indeed the high-resolution STM data resemble well the rectangular geometry (Fig. 24c and d). The unit cell with the dimensions of 0.264 nm x 0.529 nm is indicated in the images. A model of the vanadium oxide layer is displayed in Fig. 24b and would be in good agreement with an oxide of VO2 stoichiometry. As can be seen in the STM image, vanadium and oxygen atoms can be imaged with slightly different contrast in the gray scale image, the apparent corrugation varies in fact with different temporary tip configurations (Fig. 24c and d). In conclusion, it can be noted that Cu3Au(100)-0 is preferably suited as a metal alloy substrate for growing metal oxides. Phase separation of the substrate and the grown ordered oxide layer is complete. Depending on the preoxygen content at CuaAu(100), the amount of vanadium deposition and annealing temperature, three different epitaxial layers of vanadium oxides have been prepared on the oxygen treated Cu3Au substrate. Following the order of the oxidation states of vanadium, the production of two dimensionally ordered oxide phases ofVO(100), V203(0001) and g o 2 stoichiometry was reported. 5. S U M M A R Y

Ordered metal alloy systems might expose profound different surface characteristics even though consisting of the same elemental composition in the bulk. Intermixing or phase separation is correlated with the surface composition and structure. Differences appear associated by the influence of the free surface energy with segregation and surface ordering. Some prospects have been illustrated at specific metal alloy surfaces. A number of dissimilar surface compositions and structures develop at the NiA1 ordered bulk alloy by preparation dependent effects. Completely different chemical behavior against oxygen adsorption and dissociation has been found for two Cu3Au surfaces, the (100) and (110) plane, consisting of the same surface composition. The (100) surface with Cu atoms surrounded by Au atoms turns up non-reactive alike Au. On the other hand, the (110) surface with Cu chains in the first layer acts similar to a Cu(110) surface. Intermixing or phase separation can be manipulated at Cu3Au too. Upon vanadium deposition on the bare alloy surface, strong intermixing and alloy formation towards a V3Au

400 surface occurs. On the other hand, oxygen at Cu3Au prevents completely intermixing and ordered VOx layers can be grown on top of the Cu3Au (100)-O surface. Finally intermixing has been demonstrated as well for a bulk immiscible system like Ir deposited at Cu(100). After heat treatment an ordered two-dimensional sub-surface alloy has been produced. The position of subsurface Ir atoms could be imaged by STM via electronic effects. Whereas intermixing occurs for Ir on Cu, strict phase separation takes place for the reversed system of Cu on Ir showing neither intermixing nor segregation. As a matter of fact, opposed to solids with single elemental composition, the class of binary alloys obviously resembles an additional freedom to produce from the same bulk material various new surface configurations marked by different chemical reactive states and even completely new surface alloys.

ACKNOWLEDGEMENTS It is a pleasure to acknowledge the excellent cooperation and helpful discussions with Rail-Peter Blum, Dirk Ahlbehrendt, Gerhard Gilarowski and Ralf Spitzl. The work was financed in part by the German Council of Research DFG through the SFB 290 and SFB 546.

REFERENCES H. Brune, Surf.Sci.Rep. 31 (1998) 121. [1] G.L. Kellogg, Surf.Sci.Rep. 21 (1994) 1. [2] J.A. Rodriges, Surf.Sci.Rep. 24 (1996) 223. [3] U. Bardi, Rep.Progr.Phys. 57 (1994) 939. [4] B. Sch6nfeld, Progr.Mat.Sci. 44 (1999) 435. [5] C. Laubschat, K. Mi~ller, K. Wandelt, Physics and Chemistry of Alloy Surfaces, [6]

[7] [8] [9] [10] [11] [12]

[13] [14] [15] [16] [17]

[18] [19]

Surface Review and Letters, Vol. 3, World Scientific, Singapore 1996. M. Polak, L. Rubinovich, Surf.Sci.Rep. 38 (2000) 127. J. A. Rodriguez, Surf.Sci.Rep. 24 (1996) 223. C. T. Campell, Annu.Rev.Phys.Chem. 41 (1990) 775. Interfacial Electrochemistry, Marcel Dekker, Inc., New York 1999. W. Polewska, M.R. Voigt, O.M. Magnussen, R. J. Behm, J. Phys. Chem. B 103 (1999) 10440. M. Wilms, M. Ki'uft, G. Bermes, K. Wandelt, Rev. Sci. Instrum. 70 (1999) 3641. M. Lennartz, M. Arenz, C. Stuhlmann, K. Wandelt, Surf. Sci. 461 (2000) 98. M. Henzler, in H. Ibach (Ed.): Topics in Current Physics, Vol. 4, Springer, Berlin 1977. G. Binnig, H. Rohrer, Ch. Gerber, E. Weibe, Phys.Rev.Lett. 49 (1982) 57. R.J. Behm, N. Garcia, H. Rohrer, NATO ASI SERIES E: Applied Science, Vol. Vol.184, Kluwer Academic Publishers, Dordrecht NL 1990. F. Besenbacher, I. Stensgaard, in D.A.King, D.P.Woodruff (Eds.): The Chemical Physics of Solid Surfaces, Vol. 7, Elsevier 1994. F. Besenbacher, Rep.Prog.Phys 59 (1996) 1737. H. Neddermeyer, Rep.Prog.Phys. 59 (1996) 701.

401

[20] [21] [22] [23] [24] [25] [26] [27] [28] [29]

[30] [31] [32] [33] [34] [35] [36] [37]

[38] [39]

[40]

[41] [42]

[43] [44]

[45] [46] [47] [48] [49]

[50] [51] [52]

[53] [54]

[55] [56] [57]

[58] [59] [60]

P. Varga, M. Schmid, Appl. Surf. Sci 141 (1999) 287. H. Niehus, W. Heiland, E. Taglauer, Surface Science Reports 17 (1993) 213. J.W. Rabalais, Low energy ion-surface interactions, Wiley 1994. H. Niehus, Surf.Sci. 166 (1986). H. Niehus, R. Spitzl, Surf.Interf.Anal. 17 (1991) 287. H. Niehus, Appl.Phys.A53 388 (1991) 388. A.T. Hanbicki, A.P. Baddorf, E.W. Plummer, B. Hammer, M. Scheffier, Surf.Sci. 331 (1995) 811. R.M. Jaeger, H. Kuhlenbeck, H.-J. Freund, M. Wuttig, W. Hoffmann, R. Franchy, H. Ibach, Surf.Sci. 259 (1991) 235. R. Franchy, Surf.Sci.Report 38 (2000) 195. H.L. Davis, J. R. Noonan, Phys.Rev.Lett. 54 (1985) 566. D.R. Mullins, S. H. Overbury, Surf.Sci. 199 (1988) 141. H.-J. Freund, H. Kuhlenbeck, V. Staemmler, Rep.Prog.Phys. 59 (1996) 283. H. Niehus, J.Vac.Sci.Technol. A5 (1987) 751. Y. G. Shen, D. J. O'Connor, K. Wandelt, Surf.Sci 410 (1998) 1. M. T. Robinson, I. M. Torrens, Phys. Rev. B9 (1974) 5008. V. Bykov, C. Kim, M.M. Sung, K.J. Boyd, S.S. Toderov, J. W. Rabalais, Nucl. Instr. Meth, Phys. Res. B 114 (1996) 371. http'//asp 2.p hysi k. hu-ber l in. de/. R.P. Blum, D. Ahlbehrendt, H. Niehus, Surf.Sci. 366 (1996) 107. J.P. Biersack, L. G. Haggemark, Nucl. Instr. Meth. 174 (1980) 257. D. J. O'Connor, J. P. Biersack, Nucl. Instr. Meth. B 15 (1986) 14. J.F. Ziegler, J.P. Biersack, U. Littmark, The Stopping and Range in Solids, Vol. Vol. 1, Pergamon, New York 1985. H. Niehus, Nucl.Instr.Meth. B 33 (1988) 876. S.H. Overbury, D.R. Mullins, J. F. Wendelken, Surf.Sci. 236 (1990) 122. J.R. Noonan, H. L. Davis, J.Vac.Sci.Technol. A6 (1988) 722. H. Niehus, W. Raunau, K. Besocke, R. Spitzl, G. Comsa, Surf.Sci. 225 (1990) L8. P. Gassman, R. Franchy, H. Ibach, Surf.Sci. 319 (1994) 95. J.P. Roux, H. J. Grabke, Appl.Surf.Sci. 68 (1993) 49. R.P. Blum, H. Niehus, Appl.Phys. A66 (1998) 529. Y. Cao, E. H. Conrad, Phys.Rev.Lett. 65 (1990) 2808. L. Yang, T. S. Rahmann, Phys.Rev.Lett. 67 (1991) 2327. Y. Beaudet, L.J. Lewis, M. Perrson, Phys.Rev. (1993) 4127. P. von Blankenhagen, W. Schommers, V. Voegel, J.Vac.Sci.Technol. A5 (1987) 649. Y. G. Shen, D. J. O'Connor, K. Wandelt, Surf.Sci 406 (1998) 23. U. Schneider, G.R. Castro, K. Wandelt, Surf.Sci 287/288 (1993) 146. S. Schomann, E. Taglauer, Surf.Rev.Lett. 3 (1996) 1823. M. Aschoff, S. Speller, J. Kunze, W. Heiland, E. Platzgummer, M. Schmed, P. Varga, B. Baretzky, Surf.Sci.Lett. 415 (1998) L 1051. Y.G. Chen, D.J. O'Connor, K. Wandelt, R. J. MacDonald, Surf.Sci. 328 (1995) 21. V.S.Sundaram, R.S.Alben, W.D.Robertson, Surf.Sci. 46 (1974) 653. S.F. Alverado, M. Campagna, A. Fattah, W. Uelhoff, Z.Phys. B66 (1987) 103. H. Dosch, L. Mail~inder, H. Reichelt, J. Peisl, R.L. Johnson, Phys.Rev. B43 (1991) 13172. H. Dosch, Appl.Phys. A61 (1995) 475.

402 [61] [62] [63] [64] [65] [66] [67]

[68] [69] [70] [711 [72] [73] [74] [75] [76] [77]

[78] [79]

[80] [81] [82]

[83] [84] [85] [86] [87]

[88] [89] [90] [91] [92] [93] [94] [95] [96] [97]

[98]

H. Reichert, H. Dosch, P.J. Eng, I. K. Robinson, Europhys.Lett. 53 (2001) 570. K.F. Ludwig, G.B. Stephenson, J.L. Jordan-Sweet, Y. J. Mainville, S. Yang, M. Sutton, Phys.Rev.Lett. 61 (1988) 1859. A. Stuck, J. Osterwalder, L. Schlapbach, H.C. Poon, Surf.Sci. 251/252 (1991) 670. D.H. Oh, H.J. Kang, K.H. Chae, C.N. Whang, B.V. King, D.J. O'Connor, D.W. Moon, Surf.Sci. 477 (2001) L289. B. Gans, P.A. Knipp, D.D. Koleske, S.J. Sibener, Surf.Sci. 264 (1992) 81. W.E. Wallace, G. J. Ackland, Surf.Sci.Lett. 275 (1992) L685. R. Courths, H. Wern, U. Hau, B. Cord, V. Bachelier, S. Htifner, J.Phys.:Met.Phys 14 (184) 1559. S.V. Halilov, H.Gollisch, E. Tamura, R. Feder, J.Phys: Condens.Matter 5 (1993) 4711. Ch.E. Lekka, N.I. Papanicolaou, G. A. Evangelakis, Surf.Sci. 479 (2001) 287. J.A. Con Foo, A.P.J. Stampfl, B. Mattem, A. Ziegler, M. Hollering, L. Ley, J.D. Riley, R. C. G. Lecky, Sol.State Comm. 107 (1998) 385. M.T. Lin, J. Shen, W. Kuch, H. Jenniches, M. Klaua, C.M. Schneider, J. Kirschner, Surf.Sci. 410 (1998) 290. W.A.A. Macedo, F. Sirotti, A. Schatz, D. Guarisco, G. Panaccione, G. Rossi, J.Magn.Mat. 177 (1998) 1262. F. Bruno, D. Cvetko, L. Floreano, R. Gotter, C. Mannori, L. Mattera, R. Moroni, S. Prandi, S. Terreni, A. Verdini, M. Canepa, Appl.Surf.Sci. 162 (2000) 340. A. Braun, B. Feldmann, M. Wuttig, J.Magn.Mat. 171 (1997) 16. E.G. McRae, T.M. Buck, R.A. Malic, W. E. Wallace, Surf.Sci.Lett. 238 (1990) L481. H. Niehus, Th. Baumann, M. Voetz, K. Morgenstem, Surf.Rev.Lett. 3 (1996) 1899. Y. Huang, J.M. Cowley, Surf.Sci. 285 (1993) 42. H. Over, G. Gilarowski, H. Niehus, Surf.Sci. 381 (1997) L619. T.M. Buck, G.H. Wheatley, L. Marchut, Phys.Rev.Lett. 51 (1983) 43. T.M. Buck, G.H. Wheatley, D. P. Jackson, Nucl.Instr.Phys. 218 (1983) 257. K. Morgenstern, H. Niehus, G. Comsa, Surf.Sci. 338 (1995) 1. J. B. Pendry, J.Phys. C 17 (1980) 937. A. R. Miedema, Z.Metallk. 69 (1979) 455. D.J. Coulman, J. Wintterlin, R.J. Behm, G. Ertl, Phys.Rev.Lett. 64 (1990) 1761. K. Kern, H. Niehus, A. Schatz, P. Zeppenfeld, J. Goerge, G. Comsa, Phys.Rev.Lett. 67 (1991) 855. H. Niehus, C. Achete, Surf. Sci. 289 (1993) 19. Ch.E. Lekka, N. I. Papanicolaou, G. A. Evangelakis, Surf.Sci. 488 (2001) 269. H. Niehus, R.-P. Blum, D. Ahlbehrendt, Phys. Stat. Sol. a 187 (2001) 151. L. Houssiau, P. Bertrand, Surf. Sci. 352 (1996) 978. T. B. Massalski,, ASM International, Ohio 1990. H.L. Skriver, N. M. Rosengaard, Phys. Rev. B 46 (1992) 7157. E. Bauer, Appl. Surf. Sci. 11/12 (1982) 479. G. Gilarowski, H. Niehus, Surf. Sci. 436 (1999) 107. G. Gilarowski, H. Niehus, Phys. Stat. Sol. a 173 (1999) 159. J. E. P. J. de la Figuera, R. Miranda, Surf. Sci. 307-309 (1994) 538. A. Rabe, N. Memmel, A. Steltenpohl, T. Fauster, Phys. Rev. Lett. 73 (1994) 2728. A. Brodde, H. Neddermeyer, Surf. Sci. 287/288 (1993) 988. Th. Detzel, N. Memmel, T. Fauster, Surf. Sci. 293 (1993) 227.

403 [99] [ 100] [101 ] [102] [103] [104] [105] [106] [107] [108] [109] [110] [111] [112] [113] [ 114] [115] [116] [117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128]

S.-L. Chang, J.-M. Wen, P.A. Thiel, S. Gfinther, J.A. Meyer, R. J. Behm, Phys. Rev. B 53 (1996) 13747. P.J. Schmitz, W.-Y. Leung, G.W. Graham, P. A. Thiel, Phys. Rev. B 40 (1989) 11477. A. Christensen, A.V. Ruban, P. Stoltze, K.W. Jacobson, H.L. Skriver, J.K. Norskov, F. Besenbacher, Phys. Rev. B 56 (1997) 5822. S. Heinze, R. Abt, S. Blt~gel, G. Gilarowski, H. Niehus, Phys.Rev.Lett. 83 (1999) 4808. S. Heinze, S. Blfigel, R. Pascal, M. Bode, R. Wiesendanger, Phys.Rev. B 58 (1998) 16432. E. Wimmer, H. Krakauer, M. Weinert, A.J. Freeman, Phys. Rev. B24 (1981) 864. J. Tersoff, D. R. Hamann, Phys. Rev. Lett. 50 (1983) 1998. S. Heinze, M. Bode, A. Kubetzka, O. Pietzsch, X. Nie, S. Blfigel, R. Wiesendanger, Science 288 (2000) 1805. P.W. Murray, S. Thorshaug, I. Stensgaard, F. Besenbacher, E. Laegsaard, A.V. Ruban, K.W. Jacobsen, G. Kopodakis, H.L. Skriver, Phys.Rev. B 55 (1997) 1380. Ch. Konvicka, Y. Jeanvoine, E. Lundgren, G. Kresse, M. Schmid, J. Hafner, P. Varga, Surf.Sci. 463 (2000) 199. T. Ali, A.V. Walker, B. K16tzer, D. A. King, Surf. Sci. 414 (1998) 204. G. Gilarowski, J. Mendez, H. Niehus, Surf. Sci. 448 (2000) 290. E. Lang, K. Mfiller, K. Heinz, Surf. Sci. 127 (1983) 347. N. Bickel, K. Heinz, Surf. Sci. 163 (1985) 435. Q. Ge, D.A. King, N. Marzari, M. C. Payne, Surf. Sci. 418 (1998)529. J. Mendez, H. Niehus, Appl.Surf.Sci. 142 (1999) 152. T.N. Rhodin, G. Brod6n, Surf. Sci. 60 (1976) 466. J.F. Jia, K. Inoue, Y. Hasegawa, W.S. Yang, T. Sakurai, Phys. Rev. B 58 (1998) 1193. M. Gsell, P. Jakob, D. Menzel, Science 280 (1998) 717. M. Schick, G. Ceballos, Th. Pelzer, J. Sch~fer, G. Rangelov, J. Stober, K. Wandelt, J.Vac.Sci.Technol. A12 (1994) 1795. R.-P. Blum, D. Ahlbehrendt, H. Niehus, Surf.Sci. 396 (1998) 176. M. Che, C. O. Benett, Adv.Catal. 20 (1998) 153. C.R. Henry, Surf.Sci.Rep. 31 (1998) 231. D.W. Goodmann, Surf.Sci. 299/300 (1994) 837. C.T. Campell, Surf.Sci.Rep. 27 (1997) 1. M. B~iumer, H.-J. Freund, Prog.Surf.Sci. 61 (1999) 127. D.A. King, D. P. Woodruff, The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis, Vol. 1 - 4, Elsevier, Amsterdam 1990. J. Libuda, F. Winkelmann, M. Bgumer, H.-J. Freund, Th. Bertrams, H. Neddermeyer, K. Mt~ller, Surf.Sci. 318 (1994) 61. A. Stierle, V. Formoso, F. Comin, G. Schmitz, R. Franchy, ESRF Newslett. 31 (1998) 25. F. Rohr, M. B~iumer, H.-J. Freund, J.A. Mejias, V. Staemmler, S. Mt~ller, L. Hammer, K. Heinz, Surf. Sci. 372 (1997) L291.




Chapter 11

Surface structure and catalytic reactivity of palladium overlayers for 1,3-butadiene hydrogenation J.C. Bertolini and Y. Jugnet Institut de Recherches sur la Catalyse - C N R S 2 avenue Albert Einstein, F-69626 Villeurbanne Cedex, France

1. I N T R O D U C T I O N Transition metals are known to have good catalytic properties for many reactions. Their chemical properties and consequently their catalytic properties (activity, selectivity and stability) can be strongly modified when alloyed to a second element [1, 2]. The behaviour of binary alloys with respect to catalysis is generally interpreted in terms of either the geometric , associated with the number of nearest neighbours of a given element for the catalytic reaction to occur [3-5] or to the electronic

4-

F"

2-

~

0

1

I

2

II 3

Opd (ML)

II

, U

4

20

5

o

0

,

l

1

,

2

l

3

,

4

5

I

6

epd (ML)

Fig. 15. Catalytic activity for the 1,3-butadiene hydrogenation reaction (at RT, 10 Torr hydrogen and PH2/ P d i e n e "- 10) (left part) and surface composition determined by LEIS (fight part) as a function of the Pd amount deposited at RT on Cu(110). at 310K on Cu(110), an ordered (2xl) surface alloy has been also observed by LEED by Bennett et al. [64]. The catalytic activity increases suddenly for Pd deposits higher than about 1 ML, reaches a maximum near 3ML and further decreases towards a value which would be comparable to that of pure Pd(ll0) at large coverage (Fig. 15, left part). 3 ML Pd means only 4 0 - 50 at% Pd atoms in the surface layer (Fig. 15, fight part) but the activity per Pd surface atom, in 1,3 butadiene hydrogenation, is about 10 times higher than that for Pd surface atoms on pure Pd(110). Looking now at the structure, beyond 0.5-1 ML, a 2D-3D transition occurs in the growth mode. 3D islands elongated along the close-packed [110] direction and about 7-8 nm wide are formed at 0.75 ML [60, 64]. These islands increase in size as the Pd coverage is increased, until they cover the whole surface. LEIS results reported in the fight part of Fig. 15, agree well with these STM observations. A mild annealing (near 500K) does not modify strongly the Pd and Cu amounts; only a moderate increase of Cu concentration has been measured by LEIS. Upon annealing at higher temperatures (603K-723K) Bennett et al. [64] have measured by AES a preferential surface segregation of Cu and significant changes in the surface morphology with larger flat terraces that showed ragged step edges aligned bands running across the surface. These bands would be the consequence of strain relaxation between Cu and alloy sublayers. The surface is then similar to those obtained for deposition of Cu on

430

P d ( l l 0 ) at elevated temperatures; it actually moves towards its equilibrium state. The explanation of enhanced activity is not at first sight directly related to the apparition of 3D islands. However, high resolution STM observations, in the 2-4 ML range, revealed a nanostructuration, similar to that observed when 4 ML Pd were deposited on Ni(ll0) ; the period of alignments along the [110] direction is about 5 nm, instead of 2.5 nm for 4 ML Pd on Ni(ll0). As previously discussed for Pd on Ni(ll0) one can suppose that the surface reconstructs to relax the stress, at least partially, with here again creation of a specific structural arrangement, generating very active surface sites. The understanding of the modified chemical reactivity would recquire complementary informations relative to the electronic properties of the considered thin films. Finally, for larger deposits, Pd recovers its ~normal ~ structural and chemical properties, together with its ~ normal ~ reactivity. To summarize, at low coverage Pd deposited on Cu(110) primarily forms a PdCu3 like surface alloy, with tendency for Cu to come out ; such a surface has no catalytic activity for the 1,3-butadiene hydrogenation reaction. When increasing the amount of deposited Pd, highly strained 3D islands, probably constituted by Pd-Cu alloys with a noticeable amount of Pd atoms in the outer layer, are formed. They exhibit a largely modified chemical reactivity as compared to pure Pd(ll0). For larger deposits, Pd surface atoms gradually recover their own structural and chemical properties.

5.2. Pd in tension on A u ( l l 0 ) Gold has a very low fusion temperature and is therefore expected to be very mobile, i.e. to diffuse easily, even at low temperature. Moreover, its surface tension is very low compared to Pd (Table 4), and it is therefore expected to segregate largely to the surface of Pd-Au alloys (Fig. 4). This has been experimentally verified; for example, the topmost layer of Au3Pd(100) and Au3Pd(110) has been found to consist of Au atoms only [65, 66]. Moreover, the (110) surface of the alloy reconstructs in a (lx2) missing row mode as does pure Au(ll0). The growth of Pd on A u ( l l l ) using atomic beam deposition has been studied by Koel et al. [67]. Initial stages of Pd on Au deposited by electrochemistry have also been investigated on Au(11 l) and Au(100) [68, 69]. The analysis of the reactivity of Pd deposits with respect to CO chemisorpfion [70] and to the cyclization of acetylene to benzene [71 ] has been the subject of a few experimental works. A theoretical work has been devoted to the study of electronic factors governing ethylene hydrogenation and dehydrogenation activity of pseudornorphic Pd on Au(111) [21].

431

The results presented here after, relative to the growth of Pd on Au(110), are based on the combined use of LEED, MEED, grazing X-ray diffraction [72], STM [73], and LEIS [74] techniques. The test reaction is, here again, the 1,3butadiene hydrogenation. The LEED pattern of the clean (lx2) reconstructed Au(110) is presented in Fig. 16 [72] together with a schematic representation of the missing row reconstructed surface. The peculiar geometry of the missing row reconstructed surfaces of fcc metals offers a priori a unique way to generate linear structures of adatoms and to measure their specific properties. However, as will be seen below, in the present case, the situation is more complex. For low deposition, the intensity of the LEED 1/2 spots along the [001] direction decreases rapidly as the Pd coverage is increased. It vanishes at 0.5 ML Pd coverage. The intensities of the (0,1/2) spots observed by MEED (Medium Energy Electron Diffraction) and of the (0,1/2) rods in grazing incidence X-ray diffraction experiments are also largely reduced when the Pd coverage increases and approaches 0.5 ML. This indicates that Pd adatoms do not simply fill the missing rows. Actually, Au has migrated on the surface layer, the Pd concentration of the outer layer measured by LEIS being less than about 10 % for 0.5 ML (Fig. 17, fight part). 2D islands, with the formation of pits probably associated to areas for diffusion of Au atoms which cover and/or associate to the Pd deposited atoms to form a surface alloy, have been observed by STM up to nearby 1 ML. This is quite similar to what happens for Pd deposited on Cu(ll0) (w The activity for 1,3-butadiene hydrogenation is then near zero. A moderate annealing does not strongly modify the surface composition. It is necessary to heat above 650 K to induce a complete dissolution of Pd atoms into the bulk.

Fig. 16. LEED diagram observed for the (lx2) reconstructed Au(ll0) surface (130 eV electron energy) and schematic representation of the corresponding missing row reconstruction.

432

7 ~

100

Pd/Au(11 O)

80

6

E _._o 5 0

E 4 v

0 ',-

3

i

2

o
I>

"

o~

Pd(110)

60

"o 40 12. 20

0

1

2

3

4

5

ere (ML)

6

7

8

O ~lr

0

i

II

1

,

II

2

,

I

3

~

I

4

0

I

5

,

I

6

~

I

7

8

ep~ (ML)

Fig. 17. Catalytic activity for the 1,3-butadiene hydrogenation reaction (at RT, 10 Torr hydrogen and PH2 / Pdiene = 10) (left part) and surface composition determined by LEIS (fight part) as a function of the Pd amount deposited at RT on Au(110).

Between 0.5 and 3.5 ML Pd, at room temperature, the intensity of the (0,0) specular beam oscillates with a rather regular period, showing a minimum each time the number of deposited monolayers is an integer (N) and a maximum each time this number is equal to (N+l/2). STM observations reveal a quite smooth surface in this range of Pd coverage. That characterises a pseudomorphic layer by layer growth. In the mean time, the outer layer Pd concentration increases monotonically, up to about 50 % at 3.5 ML (Fig. 17). A recent quantitative analysis of diffraction beams intensities, measured by grazing incidence X-ray diffraction, would agree with the formation of large 2D islands whose top layer is rich in gold while the second layer is essentially Pd [72]. A layer by layer growth has also been put forward in the case of Pd deposited on Au(111) [67] ; nevertheless, the tendency for Au to come out seems to be less pronounced on the close-packed (111) Au face than on the more open (110) face. Au migration in the outer layer would need a higher temperature for Au(111) than for Au(110) as a substrate. The catalytic reactivity in the range of 1 to 3 ML Pd on A u ( l l 0 ) is surprisingly low taking into account the noticeable amount of Pd measured by LEIS on the surface (Fig. 17). This low reactivity could be due either to a low number of active sites (two adjacent Pd atoms) resulting from chemical order or to the tensile stress applied to Pd atoms in pseudomorphic epitaxy with Au substrate. Such a hypothesis would be in agreement with the theoretical predictions of Pallassama and Naurock [21] who propose that Pd stressed in

433

tension would chemisorb more strongly unsaturated hydrocarbons, with, as a consequence, a lower hydrogenation rate. Lastly, for thicker layers (4-8 ML), islands elongated along the [110] direction are clearly observed by STM and the intensity measured in the specular direction by MEED does not show any structuration. New spots appear in the X-ray diffraction patterns corresponding to an atomic distance characteristic of pure Pd. This structural change is accompanied by a sudden and noticeable catalytic activity (Fig. 17). The activity for 7 ML is even 3 times higher than that of pure Pd(110). One can suppose that this rough Pd-rich surface contains a lot of very active low cordinated surface sites. In summary, gold has a great ability to migrate to the surface. However, uncovered Pd atoms which are in registry with the substrate are in tension which makes them unactive for the 1,3-butadiene hydrogenation. This coud be a positive effect for some other reactions.

6. S U M M A R Y AND C O N C L U S I O N What is clear from all these results, is that structural parameters and catalytic activity are intimately related. Starting from one metal, palladium in this case, we have shown that, by alloying with or deposition on other metals, it was possible to generate a lot of distinct local structures where surface Pd could exhibit largely modified chemical reactivity. Some particular systems can show large amplifications of activity for the 1,3-butadiene hydrogenation reaction Pd exhibits structures and chemical reactivities which are similar in segregated alloy surfaces and in surface alloys (obtained by atomic beam deposition) as long as the Pd concentration in the outer layer and the misfit between overlayer and substrate lattice parameters are the same. Palladium overlayers deposited on a metallic substrate of higher surface energy are quite easy to produce and control; such systems, after annealing, may show a good structuration with Pd atoms staying out. On the contrary, when the surface energy of atoms contituting the substrate is lower than that of Pd, the substrate atoms will aim to come out and cover thin Pd layers. The migration towards the surface can be effective even at room temperature. On close-packed fcc (111) oriented surfaces a pseudomorphic adlayer is often formed and the stress is actually retained by the surface atoms. In the case of the association of two elements having similar electronic properties, i.e. very close in the periodic table, the modified chemical properties can be tentatively associated to this stress. On the contrary, on more open (110) oriented surfaces there is a tendency to reconstruction which relax, at least partially, the stress ; strained surfaces show then surface sites having very peculiar geometries and strongly modified

434 chemical reactivities. For example, the catalytic activity for 1,3-butadiene hydrogenation, of a well ordered nanostructuration formed by 3-4 ML Pd deposited on Ni(110), is nearby two orders of magnitude higher than that of pure Pd(110). This system can be considered as the most reactive catalyst known for this reaction. The structural changes together with the modified chemical reactivity are for sure associated to modifications of electronic properties which have not been largely discussed in this paper. However, it has been the object of several experimental and theoretical papers in relation with the chemical reactivity 9see for example [75-77, 19, 21, 20]. The trends with respect to chemisorption of CO [19] and unsaturated hydrocarbons [21] can be summarized as follows : on compressed pseudomorphic overlayers, large interaction and overlap between surrounding atoms will result in a broadening of the d valence band shifting downwards from the Fermi level if the d-band filling is larger than 0.5; the adsorption energy of CO and hydrocarbons [21] is decreased. On the contrary, tensile strain induces a narrowing of the d band shifting upward accompanied by an increase of adsorption energies, with important consequences for catalytic reactions. One can expect that compressive stress will make easier the addition reactions such as hydrogenation, while tensile stress will facilitate bond cleavage reactions such as hydrogenolisis. In conclusion, strained surfaces can show very original structures and new catalytic properties. In order to associate the modified catalytic properties to the peculiar structures generated, one has to asume that these original structures are still present under the reactive mixture, at high pressure. Measurements under pressure of reactants are then necessary to measure both the surface structure and the surface species as reaction intermediates. Up to now, only very few data are available in that field. Recent developments around techniques such as STM [79-80], grazing X-ray Diffraction [81] ... and optical vibrational spectroscopies such as IRRAS[82-83] using a polarized light and SFG [79] have demonstrated the possibility to realise such observations. Finally, a good knowledge of the processes involved at alloy surfaces and surfaces alloys, and the ability to control the stress would enable us to carry out the synthesis of new catalysts having tailor made surface sites specific for a particular reaction.

ACKNOWLEDGEMENTS Acknowledgment is made to R6gion Rh6ne-Alpes for its financial support through a project "nanotechnologies" (# PR97039).

435 The authors would like to thank all those who contributed to the results presented in this work : M. Abel, L. Porte and Y. Robach for their STM observations, P. Delich~re for LEIS investigations, P. Ruiz for his contribution in doing the catalytic reactions, R. Baudoing-Savois, P. Dolle, M.C. Saint-Lager, and M. De Santis for the X-ray diffraction experiments and quantitative analysis of the results, J.S. Filhol, D. Simon, P. Sautet for their theoretical contributions on the Pd/Ni(110) system. We would like also to acknowledge several students and postdoctoral fellows for their participation to this work, among them L. Constant, A. Franquet, J.M. Guigner, P. Hermann, L. Lianos, and A.C. Michel. X A N E S and X-ray Diffraction experiments have been done in L U R E (Orsay) and ESRF (Grenoble) respectively. We would like to thank beam operators and beam line local contacts. Lastly, we would like to thank N.S. Prakash for his help in proof reading.

REFERENCES [ 1] J.H. Sinfelt, Bimetallic Catalysts, Wiley, New-York, 1983. [2] B.E. Nieuwenhuys, The Chemical Properties of Solid Surfaces, D.A. King and D.P. Woodruff eds., Elsevier, Amsterdam, 1993, pp. 185-224. [3] J.H. Sinfelt, J.L. Carter and J.T. Yates, J. Catal. 24 (1972) 283. [4] G.A. Martin, Catal. Rev. Sci. Eng. 30 (1988) 519. [5]C.T. Campbell, J.M. Campbell, P.J. Dalton, F.C. Henn, J.A. Rodriguez and S.G.Seimanides, J. Chem. Phys 93 (1989) 815. [6] P.W. Wynblatt and R.C. Ku, Surf. Sci. 65 (1974) 511. [7] J.L. Rousset, J.C. Bertolini and P. Miegge, Phys. Rev. B 53 (1996) 4947 and references therein. [8] Y. Gauthier, Y. Joly, R. Baudoing and J. Rundgren, Phys. Rev. B 31 (1985) 6216. [9] E. Bauer, Z. Kryst. 110 (1958) 372. [10] E. Bauer and J.H. Van der Merwe, Phys. Rev. B 33 (1986) 3657. [11] R. Koch, The Chemistry of Solid Surfaces, vol. 8, the growth and properties of ultrathin epitaxial layers, D.A. King and D.P. Woodruff eds., Elsevier, Amsterdam, 1997. [12] H. Ibach, Surf. Sci. Rep. 29 (1997) 193 and references therein. [13] M. Schmid, W. Hofer, P. Varga, P. Stoltze, K.W. Jacobsen and J.K. Norskov, Phys. Rev. B 51 (1995) 10937. [14] F. Besenbacher, L. Pleth Nielsen and T. Sprunger, The Chemical Physics of Solid Surfaces and Heteregeneous Catalysis, Surface alloying in heteroepitaxial growth, D.A. King and D.P. Woodruff eds., Elsevier, Amsterdam, 1997. Vol. 8, Chapter 10 [15] See for example Th. Pearsall ed., Strained-layer Superlattices- Materials Science and Technology in Semiconductors and Semimetals, Vol. 32, Academic Press, Boston, 1991. [16] J.M. Heitzinger, S.C. Gebhardt and B.E. Koel, Surf. Sci. 275 (1992) 209. [17] R.A. Campbell, J.A. Rodriguez and D.W. Goodman, Phys. Rev. B 46 (1992) 7077. [18] M. Gsell, P. Jakob and D. Menzel, Science 280 (1998) 717. [19] M. Mavritakis, B. Hammer and J.K. Norskov, Phys. Rev. Lett. 81 (1998) 2819. [20] P. Sautet, Stress and Strain in Epitaxy : theoretical concepts, measurements and applications, M. Hanbucken and J.P. Deville eds., Elsevier, Amsterdam, 2001.

436 [21] V. Pallassama and M. Neurock, J. Catal. 191 (2000) 301. [22] A.W. Denier Van der Gon, R. Cortenraad, W.P.A. Jansen, M.A. Reijme and H.H. Brongersma, Nucl. Instr. and Meth. in Phys. Res. B 161-163 (2000) 56. [23] D.O. Boerma, Nucl. Instr. and Meth. in Phys. Res. B 183 (2001) 73. [24] D. Briggs and M.P. Seah eds., Practical Surface Analysis, 2nd edition, volume2, John Wiley and sons eds, Chichester (1992). [25] D.P. Woodruff, D. Brown, P.D. Quinn, T.C.Q. Noakes and P. Bailey, Nucl. Instr. and Meth. in Phys. Res. B 183 (2001) 128. [26] P. Bailey, T.C.Q. Noakes, C.J. Baddeley, S.P. Tear and D.P. Woodruff, Nucl. Instr. and Meth. in Phys. Res. B 183 (2001) 62. [27] D.P. Woodruff, R. Davis, N.A. Booth, A.M. Bradshaw, C.J. Hirschmugl, K.M. Schindler, O. Schaff, V. Femandez, A. Theobald, Ph. Hofmann and V. Fritzsche, Surf. Sci. 357-358 (1996) 19. [28] Y. Gauthier, R. Baudoing, Y. Joly, J. Rundgren, J.C. Bertolini and J. Massardier, Surf. Sci. 162 (1985) 342. [29] J. Massardier, J.C. Bertolini and A.J. Renouprez, Proceedings of the 9 th International Congress on Catalysis, M.J. Philips and M. Teman eds., Calgary, Vol. 3 (1988) 1222. [30] T. Ouchaib, J. Massardier and A.J. Renouprez, J. Catal. 119 (1989) 517. [31] J.C. Bertolini, A. Cassuto, Y. Jugnet, J. Massardier, B. Tardy and G. Tourillon, Surf. Sci 349 (1996) 88. [32] P.Sautet and J.F. Paul, Catal. Lett. 9 (1991) 245. [33] J.C. Bertolini, J. Massardier, P. Ruiz and B. Tardy, Surf. Sci. 211-212 (1989) 1053. [34] J.B. Darby Jr and K.M. Mills, Metallurgical Transactions 3 (1972) 653. [35] R. Hultgen, P.D. Desai, D.T. Hawkins, M. Gleiser and K.K. Kelley, Selected Values of Thermodynamic Properties of Binary Alloys, American Society of Metals, Metals Park OH, 1973. [36] W.R. Tyson and W.R. Miller, Surf. Sci. 62 (1977) 267. [37] P. Miegge, J.L. Rousset, B. Tardy, J. Massardier and J.C. Bertolini, J. Catal. 149 (1994) 404. [38] J.C. Bertolini, P. Hermann and P. Delich6re, Surf. Interfaces Anal. 24 (1996) 34. [39] I. Moysan, V. Paul-Boncour, S. Thiebault, E. Sciora, J.M. Foumier, R. Cortes, S. Bourgeois and A. Percheron-Gu6gan, J. of Alloys and Compounds 322 (2001) 14. [40] J. Goetz, Y. Murzin and R. Touroude, Ind. Eng. Chem. Res. 35 (1996) 703. [41 ] J. Goetz, M.A. Volpe and R. Touroude, J. Catal. 164 (1996) 369. [42] A.C. Michel, L. Lianos, P. Delich6re, N.S. Prakash, J. Massardier, Y. Jugnet and J.C. Bertolini, Surf. Sci. 416 (1998) 288. [43] M. Abel, Y. Robach, J-C. Bertolini et L. Porte, Surf. Sci. 454-456 (2000) 1. [44] L. Lianos, Y. Debauge, J. Massardier, Y. Jugnet and J.C. Bertolini, Catat. Lett. 44 (1997) 211. [45] A.C. Michel, thesis, Universit6 Claude Bemard Lyonl, November 2000. [46] A. Rochefort, M. Abon, P. Delich6re and J.C. Bertolini, Surf. Sci. 294 (1993) 43. [47] Y. Debauge, M. Abon, J.C. Bertolini, J. Massardier and A. Rochefort, Appl. Surface Sci. 90 (1995) 15. [48] R.S. Rao, A. Bansil, H. Aasonen, and M. Pessa, Phys. Rev.B 29 (1984) 1713. [49] D.J. Holmes, D.A. King and C.J. Barnes, Surf. Sci. 227 (1190) 179. [50] M.A. Newton, S.M. Francis, Y. Li, D. Law and M. Bowker, Surf. Sci. 259 (1991) 45. [51] C. Sousa,V. Bertin and F. Illas, J. Phys. Chem. B 105 (2001) 1817. [52] P. Hermann, B. Tardy, D. Simon, J.M. Guigner, B. Bigot and J.C. Bertolini, Surf. Sci. 307-309 (1994) 422.

437 [53] S. Terada, T. Yokoyama, N. Saito, Y. Okamoto and T.Ohta, Surf. Sci. 433-435 (1999) 657. [54] X.D. Xiao, X.D. Zhu, W. Daum, Y.R. Shen, Phys. Rev. Lett. 66 (1991) 2352. [55] L. Porte, M. Phaner-Goutorbe, J.M. Guigner and J.C. Bertolini, Surf. Sci. 424 (1999) 262. [56] M.C. Saint-Lager, private communication. [57] J.S. Filhol, D. Simon and P. Sautet, Surf. Sci. Lett. 472 (2001) 139. [58] J.M. Guigner, thesis, Universit6 Claude Bernard, Lyonl, January 1997. [59] J.S. Filhol, thesis, Ecole Normale Sup6rieure, Lyon, July 2001. [60] L. Constant, P. Ruiz, M. Abel, Y. Robach, L. Porte and J.C. Bertolini, Topics in Catalysis 14 (2001) 125. [61] W. Murray, S. Thorshaug, I. Stensgaard, F Besenbacher, E. Laegsgaard, A.V. Ruban, K.W. Jacobsen, G. Kopidakis, and H.L. Dkiver, Phys. Rev. B 55 (1997) 1380. [62] Y. Fujinaga, Surf. Sci. 86 (1979) 581. [63] J.E. Garces, G.H. Bozzolo, P.Abeland H.O. Mosca, Appl. Surface Sci. 167 (2000) 18. [64] R.A. Bennett, S. Poulston, N.J. Price, J.P. Reilly, P Stone, C.J. Barnes and M. Bowker, Surf. Sci. 433-435 (1999) 816. [65] J. Kuntze, S. Speller, W Heiland, A. Atrei, G. Rovida and U. Bardi, Phys. Rev. B 60 (1999) 1535. [66] J. Kuntze, S. Speller, W. Heiland, P. Deurinck, C. Creemers, A. Atrei and U. Bardi, Phys. Rev. B 60 (1999) 9010. [67] B.E. Koel, A Sellidj and M.T. Paffett, Phys. Rev. B 46 (1992) 7846. [68] L.A. Kibler, M. Kleinert, R. Kandler, D.M. Kolb, Surf. Sci. 443 (1999) 19. [69] L.A. Kibler, M. Kleinert, R. Kandler, D.M. Kolb, Surf. Sci. 461 (2000) 155. [70] A Sellidj and B.E. Koel, Phys. Rev. B 49 (1994) 8367. [71] C.J. Baddeley, R.M. Omerod, A.W. Stephenson and R.M. Lambert, J. Phys. Chem. 99 (1995) 146. [72] P. Dolle, M.C. Saint-Lager, M. de Santis, R. Baudoing-Savois, M. Abel, J.C. Bertolini and P. Delich6re, Surf. Sci., submitted. [73] M. Abel, thesis, Ecole Centrale de Lyon, September 2001. [74] P. Delich6re and J.C. Bertolini, Surf. Interface Anal., submitted. [75] J.A. Rodriguez and D.W. Goodman, J. Phys. Chem. 95 (1991) 4196. [76] J.A. Rodriguez, Surf. Sci. Rep. 24 (1996) 233. [77] J.H. Larsen and Ib Chorkendorff, Surface Sci. Rep. 35 (1999) 163. [78] A. Ruban, B. Hammer, P. Stoltze, H.L. Skriver and J.K. Norskov, J. Molec. Catalysis A : Chemical 115 (1997) 421. [79] G.A. Somorjai, Appl. Surf. Sci. 121-122 (1997) 1. [80] L. Osterlund, P.B. Rasmussen, P. Thostrup, E. Laegsgaard, I. Stensgaard and F. Besenbacher, Phys. Rev. Lett. 86 (2001) 460. [81] H. Isem, K. Peters, P. Steadman, O. Robach, J. Alvarez, E. Lundgren and S. Ferrer, Surf. Sci. 482-485 (2001) 101. [82] M. Endo, T. Matsumoto, J. Kubota, K. Domen and C. Hirose, Surf. Sci. Lett. 441 (1999) 931. [83] Y. Jugnet, L.J. Shorthouse, J.C. Bertolini and P. Colin, Silicon for the Chemical IndustryV, H.A. Oye, H.M. Rong, L. Nygaard, G. Schussler and J. Kr. Tuset eds., Tromso, 2000, pp 331-345.




Chapter 12

Electronic and chemical properties of palladium in bimetallic systems: How much do we know about heteronuclear metalmetal bonding? Jos~ A. Rodriguez Department of Chemistry, Brookhaven National Laboratory, Upton, New York 11973, USA

1. INTRODUCTION In many industrial applications, bimetallic systems are superior to their single metal-metal counterparts in terms of catalytic activity and/or selectivity [ 1-4]. For a long time it has been known that a bimetallic surface can exhibit chemical and catalytic properties that are very different from those of the surfaces of the individual metals. Systematic research on alloy catalysts started in the late 1940s [5-7] with the purpose of establishing links between the electronic and catalytic properties of a surface, a knowledge necessary for a scientific design of catalysts. However, due to the lack of adequate techniques for the preparation and characterization of the surface alloys, no real progress was made at an experimental level. In the 1960s and 1970s the development of bimetallic catalysts for hydrocarbon reforming in the petrochemical industry increased the need for a fundamental understanding of the behaviour of bimetallic surfaces, and renewed the interest in catalysis by alloys [2,3,8,9]. This effort provided the basis for the concepts of "ensemble" and "ligand" effects [3,9], which are frequently used to rationalize the superior performance of bimetallic catalysts. "Ensemble" effects are defmed in terms of the number of surface atoms needed for a catalytic process to occur. Changes in catalyst composition modify the ensembles of available active sites. "Ligand" effects refer to those modifications in catalytic activity or selectivity that are the product of electronic interactions between the components of the bimetallic system. Over the years, it has become clear that it is difficult to find pure "ensemble" or "electronic" effects [ 10]. In the last two decades, the development of new experimental techniques [11,12] and reliable theoretical methods [13,14] have

439 made it feasible to study in detail electronic and chemical properties of bimetallic surfaces. Thus, many phenomena responsible for the behaviour of bimetallic surfaces have been identified [ 14-16]. Yet, several important issues associated with heteronuclear metal-metal bonding remain mysterious or badly understood [15,17,181. In this chapter, an overview is presented of studies that deal with the electronic and chemical properties of Pd in bimetallic systems. We will focus on palladium for three main reasons. First, bimetallic catalysts that contain Pd or other Group-10 metals have many uses: isomerization of hydrocarbons, olefin hydrogenation, CO oxidation, alcohol synthesis, acetylene trimerization, etc. [8,10,19-21]. Second, palladium is very sensitive to the formation of bimetallic bonds [22-24]. And third, there is a vast number of experimental and theoretical articles in the literature that examine the properties of Pd in bimetallic systems [14,15,19-23,25-44]. From this large volume of work, one can get a general idea of how deep is our knowledge about the basic nature of bimetallic bonding and how it affects the properties of a metal. The chapter is organized as follows. It starts with a description ofphotoemission and thermal desorption experiments for Pd overlayers on different types of metal substrates. General trends in the experimental data are examined and bonding models that explain them are discussed. Then, the validity of the bonding models is tested through ab initio or first-principles quantum mechanical calculations. From the combination of experiments and theory, a complete picture of the nature of bimetallic bonding is beginning to emerge. 2. PHOTOEMISSION STUDIES Pd atoms bonded to surfaces of early-transition metals exhibit large electronic perturbations in their valence and core levels [15]. The valence photoemission spectra shown in Figure 1 for Pd/Nb(ll0) and Pd/W(100) illustrate this phenomenon [26,43,44]. In early studies examining the interaction of Pd with a Nb(100) surface [43], it was found that the supported Pd monolayer (ML) had a relatively narrow 4d band which exhibited a low density of states (DOS) around the Fermi level (EF). In contrast, Pd multilayers and bulk palladium show emission spectra characterized by a large DOS at EF. More recent photoemission studies for a Pd layer in contact with Wa(ll0) [25], W(100) [26], W(ll0) [26] and Mo(ll0) [15,45] also show a narrow Pd(4d) band with a centroid shifted toward higher binding energy. Thus, it appears that the bonding interactions between Pd and earlytransition metals are quite strong. This will be confmned below by results of

440

Pd/W(100)

Pd/Nb(; 10i '

(~

~,z, / I |

/!i I

~

e(10,5/cm2)

P"

---3.4

t

\! 09

/0=1 ~

0 lo;8;i BINDING ENERGY (eV) J,

--0.0

I,

8

6

4

2

0

BINDING ENERGY ( e V )

Fig. 1 (a) Ultraviolet photoelectron spectra for monolayer (dashed curve) and greater than monolayer coverages of Pd on Nb(ll0). (b) UPS spectra of various coverages (0) of Pd on W(100). Reprinted from ref. [44]. thermal desorption mass spectroscopy. Figure 2 displays photoemission spectra for the valence region of Pd/Rh(111) as a function of admetal coverage [32]. The Pdo.9/Rh(lll) system exhibits a band structure that is very similar to that of Rh(111) or Pd multilayers. Difference spectra showed only minor electronic perturbations for supported palladium near the Fermi level [32].

-

Valence

o.9

~

-

1).4

~

1

1

l

1

1

I

12

10

8

6

4

2

0

Binding Energy (eV) Fig. 2 Valence spectra for the bonding of Pd to Rh(111). Reprinted from ref. [32].

441 The magnitude of the binding-energy shift in the Pd 4d band depends on the position of the metal substrate in the Periodic Table. Figure 3 shows the electronic perturbations observed for Pd in surface alloys (PdTi [46] and PdA1 [47,48]) and Pd monolayers supported on several metals (Ta(ll0) [25], M o ( l l 0 ) [15,45], W ( l l 0 ) [26], Re(0001) [27], Ru(0001) [27] and A I ( l l l ) [49]). The experimental results are ordered according to the group in the Periodic Table of the metal bonded to Pd. One finds that the electronic perturbations for the bonding of Pd to s,p metals like A1 [47-49] or Zn [35,50] are as large as those seen for Pd bonded to early-transition metals, and much bigger than those found when Pd is bonded to late-transition metals. In general, the magnitude of the shill in the Pd valence levels increases when the fraction of empty states in the valence band of the metal substrate rises [23,48]. This phenomenon could result from a simple hybridisation of the admetal and substrate valence bands [14,25]. In addition, a substrate induced Pd(4d)--Pd(5s,5p) rehybridization could contribute to it [23,51,52]. It is interesting that the systems with the largest shifts reported for the centroid of the Pd 4d band (Pd/A1, Pd/Zn, Pd/Ti) also undergo alloy formation [46-50]. Indeed, results to be shown below show a correlation between the strength of the bimetallic bond and the size of the electronic perturbations in Pd. The core levels of Pd are also very sensitive to the formation of bimetallic bonds. Figure 4 shows Pd 3d XPS spectra for different coverages of palladium on

,~__%2.7 "I=t 2.4 ,-~ 2.1 1.8 1.5

a~ 1.2 0.9

.~

o.6

~ o.3 r~

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

I

I

I

I

\\\ \\\ \\\ \\\ \\\

-0.0

Pd/A1 PdTi Pd/Ta Pd/Mo Pd/W Pd/Re Pd/RuPd(100) Fig. 3 Effects of bimetallic bonding on the properties of Pd: Shift in the first peak of the Pd 4d band, the one closer to the Fermi level, as a function of metal substrate. Reprinted from ref. [15].

442

XPS

Pd

3d3/2

03 I-Z

izi re

,a: i

z I,LI I.Z

X8.6

x5.1

3.5

xl

.41

344

342

340

BINDING

338 ENERGY,

336

334

eV

Fig. 4 Palladium3d core-level spectra for the Pd/W(110) system. Reprinted from ref. [53]. W ( l l 0 ) [53]. There is no observable change with coverage in the separation of the palladium 3d core levels. In the top panel of Figure 5, one can observe that the supported palladium monolayer has a Pd 3d~/2 binding energy substantially larger than that seen for palladium multilayers. Photoemission studies indicate that the Pd 3d5/2 binding energy of the surface atoms of Pd(100) is - 0.4 eV smaller than that of bulk Pd [54]. When this is taken into consideration [53], one fmds that palladium atoms bonded to W ( l l 0 ) have 3d core levels shitted - 0.85 eV toward higher binding energy with respect to those of the surface atoms of pure palladium. The perturbations induced by tungsten on the palladium core levels affect not only the first layer in direct contact with the substrate but also subsequent layers (see Figure 5) [53]. This phenomenon has been also observed on Re(0001) [27] and Mo(ll0) [44]. It tracks changes in the structural and chemical properties of the palladium adatoms [27,53].

443 ,I/----

.20 .10 .90 .80

Pd/W(110) .

335.70 IJJ Z

Ill

=,o

.10

(oool)

336.00 Z

.so

t'~

.so

Z

.--- 335.70 rn

U3

"o ~

"13

,

,

,

,

,

,

,

~

.20

,o

,

~:;,~______.

1 o)

o

336.00

.90.80

. . . .

,, i " "

335.70

,, 0

1

2

3

4

5

1

6

i

7

t

8

I

9

10

j/-"

40

Pd COVERAGE, ML Fig. 5 Palladium 3d5/2 core-level positions for Pd/W(ll0) , top, Pd/Re(0001), center, and Pd/Mo(110), bottom, as a function of Pd coverage. Reprintedfrom refs. [27,44]. Figure 6 compares 3ds/2 core-level binding-energy shifts for the deposition of palladium on several metal substrates: A I ( l l l ) [48], Ti [15,46], Ta(ll0) [25,27], Mo(ll0) [15], W(ll0) [53], Re(0001) [27] and Ru(0001) [27]. Alloying takes place in the Pd/AI(lll) and Pd/Ti systems with very big core-level shifts. In all cases, bimetallic bonding shifts the Pd core levels towards higher binding energy. The electronic perturbations in palladium are larger when the element is bonded to a s,p metal or to a transition metal with a valence band almost empty. The case of Pd/Re(0001) is particularly interesting because the palladium adlayer is pseudomorphic to the rhenium substrate, with an atomic density and structure that are very similar to those of the surface atoms in Pd(111) [27]. Yet, the admetal atoms in Pd/Re(0001) are electronically and chemically perturbed due to the effects of bimetallic bonding [27]. The trends in Figures 3 and 6 are identical. In fact, one can say that the shifts in the Pd core levels track shifts in the centroid of the Pd

444

1.8 1.6 1.4 1.2 ,~ ~.o

"~ 0.8 "~ 0.6 "~ 0.4 0.2 0.0

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

\

Pd/AI

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

4\ ..\ \ \ ~,,\\ b,.\\ N.\\ ~_ .. x

PdTi

\[\\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \

j\\\

\ \ \

x_ \ \ \ \ \ \\\ \ \.\. \ \ \ \ \ \ \ \ \ x.x.x \ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \

\ \ \ \ \ \ \ \ \ \ \ \

Pd/Ta PdRVlo Pd/W

Pd/Re

Pd/RuPd(100)

Fig. 6 Effects of bimetallic bonding on the properties of Pd: Shift in the Pd 3d5/2 core level as a function of metal substrate. Reprinted from ref. [15]. 4d band [15,17], although the magnitude of the shifts in the core and valence levels is different in many cases. The type of perturbations seen for Pd in the bimetallic surfaces are similar in many aspects to those found in bulk alloys, where heteronuclear metal-metal bonding induces an increase in the binding energy of the core levels and valence band of Pd [56-59]. Results of x-ray absorption spectroscopy indicate that this phenomenon is accompanied by a reduction in the d electron population of Pd [51,52]. Core-level shifts have been detected in many bimetallic surfaces [15]. Other admetals also exhibit well defined trends as seen for palladium [15]. For example, Figure 7 displays core-level shitts measured after depositing a monolayer of platinum or nickel on a series of metal substrates [23]. Two clear trends can be observed in these experimental data. First, the magnitude of the core-level shitt for an admetal increases when the fraction of empty states in the valence band of the metal substrate rises: ruthenium < titanium < aluminum. And second, the larger the occupancy of the admetal d band, the bigger the core level shift in the admetal: nickel < platinum < palladium [23]. The largest electronic perturbations are found in systems that combine and admetal with an electron-rich d band and a substrate with an electron-poor valence band. A priori, it is not clear what causes the core level shifts seen in Figures 4-7 and, in particular, if these shifts come from initial state effects [60,61]. In principle,

445 2.0 Ni 1.5 1.0 r~

.~

0.5 0.0

~

/

1

Ni/AI Ni/Ti

/ Ni/Ta

Ni/W

Ni/Mo Ni/Ru

Pt

O

r,,)

"O

Ni(100)

2.0

1.5 1.0

1 ML, but no (or little) hydrogen adsorbs when 0pd= 1 ML [43,81]. A similar behaviour is seen for the interaction of H2 with Pdffa(ll0) [82] and Pd/Mo(100) [83]. Electronic perturbations reduce the adsorption energy of ethylene on a Pd monolayer supported on Mo(100) [84]. Ethylene is weakly chemisorbed on the Pd monolayer (desorption temperature ~ 250 K against-~ 290 K on pure Pd), and the adsorbed species is much less rehybridized from sp2 in the gas phase toward sp3 on this surface compared to C2H4 chemisorbed on the (100) face of pure palladium [84]. 5. MODELS FOR BIMETALLIC BONDING The experimental results in Figures 9, 13, 14 and 15 show strong correlations between the electronic and chemical properties of an dement in a bimetallic surface. In the early 1990s, it became clear that the electronic perturbations induced by bimetallic bonding are associated with the strength of the heteronuclear metal-metal bond [27], and that these perturbations can determine the chemical reactivity of a bimetallic surface [22,44]. To explain the correlations in Figures 9, 13, 14 and 15 a model for bimetallic bonding was proposed [22,27,44]. There were three basic assumptions in the model. First, on the basis of the correlations in Figures 9 and 14, it was assumed that the shifts in the core levels reflected real changes in the initial state of the Pd electrons. Second, since the largest electronic perturbations were found in systems that involved "electron-rich + electron-poor" metal combinations (i.e. Pd/Ta, Pd/W, etc) with an admetal-induced reduction in the work function of the metal substrate, it was thought that bimetallic bonding produced some transfer of electrons (Pd ~) which eventually led to positive shifts in the core and valence levels of palladium. And third, it was proposed that the electronic perturbations in Pd reduced the strength of the Pd-CO bond by weakening z back-bonding. On metal surfaces the CO chemisorption bond is dominated by interactions between the occupied valence levels of the metal and the LUMO (2~ orbital) of the adsorbate (~ back-bonding) [71,72]. For supported Pd the 4d valence band is more stable than in pure Pd, probably weakening ~ back-bonding and leading to smaller CO adsorption energies [44,85]. At the time, this model for metal-metal bonding offered a logical and consistent explanation for the experimental facts [22,27,44]. Its three basic assumptions had to be validated by additional experimental and/or theoretical work. Photoemission studies have shown that in many cases the formation of a bimetallic bond induces positive core-level shills for both metals [17,86,87,88,]. This, obviously, is not consistent with a simple metal--metal charge transfer [60,90]. The phenomenon could be a consequence of combining inter- and intra-atomic charge redistributions (for example, d-.sp rehybridization) induced by bimetallic

455 bonding [23,51,60,90]. Thus, the bond between two different metals can be quite complex [17]. Theoretical studies have been useful for clarifying this issue and other aspects associated with heteronuclear metal-metal bonding. 6. T H E O R E T I C A L STUDIES

6.1 Charge redistribution in bimetallic bonding The nature of the bond between Pd and surfaces of transition or s,p metals has been the subject of a large series of theoretical works [23,33,34,35-42,89-91]. From these studies, it is clear that the Pd-substrate bond is best described as metallic with a small degree of ionic character. The direction of the net charge transfer (i.e. Pd~substrate or substrate-.Pd) varies from one calculation to the other. This discrepancy can be attributed to the lack of charge self-consistency in some of the calculations, and to the intrinsic difficulties associated with determining charge transfer, especially when the net amount of electron density transferred is small. The different schemes used for partitioning the electron population of each atom are more or less arbitrary [90,92,93], and in practice, the results of a given type of analysis can only be justified by comparing against the trends or predictions of experimental measurements. A reasonable approach is to plot the electron density around a metal atom and observe any possible change in the spatial distribution of the electrons [33,34,40,94]. The calculated electron density for a Pd monolayer supported on Ta(ll0) is plotted in Figure 16. These results are from first-principles density-functional calculations with the full-potential linearized augmented plane-wave (FPLAPW) method [33,34]. A strong Pd-Ta bonding interaction can be seen in the charge density difference shown in Figure 16c, where electrons deplete from both the interfacial Ta and Pd sites and accumulate in the region between them [33,34]. The more significant charge redistribution occurs around the Pd atoms, with the average center of electrons shifting away from the plane of Pd nuclei toward the substrate. The complex nature of the bimetallic bond in the Pd/Ta(ll0) system leads to positive core-level shifts for Pd and Ta [27,33,90,95,96]. The Pd-Ta bond cannot be classified as a simple "metallic" or "ionic" bond [33,34]. It involves and important shift of electrons from the Pd atom toward the Ta substrate, as the work function and Pd core levels suggest [48,97], and a simultaneous electron depletion around Ta, as the Ta core-level shifts and a simple Pd-Ta "covalent" interaction imply [87,90,96,97]. The FPLAPW method has also been used to study bimetallic bonding in Pd/W(ll0), Pd/Re(0001) and Pd/Ru(0001) [34]. In general, electron density plots show an important shift of electrons from the Pd layer toward the metal-metal

456

(a)

(b)

(c)

Pd

Ta(I)

Ta(I-1)

Ta(I-2)

Ta(C) Fig. 16 (a) Calculated valence charge density for a Pd monolayer (top) and clean Ta(ll0). (b) Calculated valence charge density for the Pd/Ta(110) system. (c) Charge density difference obtained by subtracting the superposition of the charge densities of the Pd monolayer and Ta(110) from that of Pd/Ta(110). Dashed lines indicate a decrease in the electron density. Reprinted from ref. [33]. interface. A similar result has been found in first-principles density-functional slab calculations for Pd/Mo(110) [40,98]. The larger the movement of electrons from around Pd to the metal-metal interface, the stronger the bimetallic bond [34,98]. The charge redistribution around Pd is in part caused by a Pd(4d)-. Pd(5s,5p) rehybridization that accumulates electrons in the bimetallic bonds [23,98]. Such a rehybridization has been observed in many theoretical studies, using different levels of theory and cluster or slab models [23,37-39,41,48,98]. In general, this redistribution of electrons is more significant than the net charge transfer between the Pd overlayers and metal substrates. From studies of x-ray absorption spectroscopy [51,52], it is known that Pd has a tendency to lose d electrons when forming bulk intermetallic compounds. Figure 17

457 shows the calculated 4d electron population for a Pd atom bonded to clusters that model hollow sites o f A I ( l l l ) , W ( l l 0 ) , R h ( l l l ) and P d ( l l l ) [23,32]. After comparing the results for Pd/Rh9 and Pd/Pdg, o n e can conclude that Rh induces minor changes in the electron distribution around the Pd atoms. This is consistent with the photoemission results in Figures 2 and Figure 9. For a Pd/Rh system the loss in the Pd 4d population, as a consequence o f a d-.s,p rehybridization and a Pd--substrate shift o f electrons, is smaller than for Pd/A1 and Pd/W systems [32]. The qualitative trends in Figures 3, 6 and 17 are identical: as the fraction o f empty states in the valence band o f the substrate rises, there is an increase in the magnitude o f the electronic perturbations in palladium. A similar correlation is observed in DF slab calculations for the bimetallic systems [34,98]. I0.0

Pd/X9 0 ,p,4

"K"K"X'I \ \ \ .,_ .,_ ,_ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ xxx x \ \

0

o

9.5

\ \ \ \ \ \ \ \ \ \ \ \ xxx \ \ \

9.0

X= AI

Cluster I (1:6:8)

W

Rh

,.x,~ ,. - . - . ,

\\\"

~\'?" ~\-,
2, 0S= 0.74}/Mo(110) surfaces show trends and a zero-order line shape that match those of gold multilayers [22]. No signal is seen for gold attached to molybdenum. A plot of the natural logarithm of the gold desorption rate

Fig 4 STM images (300x300 nm) of a strained Au monolayeron Ru(0001) before (left) and after (fight) coadsorptionwith sulfur [39].

473

@

Au-TDS Au/S/M o(110) 0s=0.74

i

03 . m

t--

_ci t._

>., . m

03 c--

{(33 03 03

0.08 '

I

1000

'

1100

I

'

I

1200

'

I

1300

1400

1500

Temperature (K)

a~

|

Au/S/Mo(110)

n,"

----

.

r

.o {3.. k..

o (D

s

I

0.84

'

i

'

0.86

I

0.88

'

I

0.90

1000/T (K -1)

Fig 5 (A) Au-TDS spectra acquired after dosing Au to a S0.74/Mo(110) surface at 300 K. (13) Activation energy for the desorption of gold. Reprinted from ref. [22]. against 1/T yields a straight line with a desorption activation energy of ~ 80 kcal/mol. This value is close to the heat of vaporization of metallic gold [22,42]. Results of STM for the S/Au/Mo(100) system again show segregation of Au and S into separate areas of the surface [14].

474 Ab initio self-consistent-field calculations and cluster models have been used to

study the bonding of sulphur and gold to Mo(110) [22]. Both adsorbates behave as electron acceptors and, therefore, compete for the electrons of molybdenum. The electronic interactions between sulphur and the metal substrate are considerably stronger than those seen for the adsorption of gold. In addition to withdrawing charge from molybdenum atoms, sulphur substantially reduces the density-of-states (DOS) that the metal atoms exhibited around the Fermi level (or highest-occupied molecular orbital, HOMO). This is illustrated in Figure 6, has been proven by photoemission spectra [43], and mainly arises from hybridisation of the Mo(4d,5s) and S(3s,3p) orbitals. Such a phenomenon considerably weakens the strength of the Mo-Au bonds [22]. From the studies described above, one can expect that sulphur alters (or poisons) the properties of catalysts that combine gold and transition metals by inducing a reduction in the degree of "wetting" of the surface of the transition metal by gold. This effect can explain changes induced by sulphur on the activity and selectivity of bimetallic catalysts used for hydrocarbon reforming [7,22,30]. !

|

........ i

!

S)

(x4)

o -

, -6.0

i HOMOr -5.6

...... ] -5.2

..................................................... ] -6.0

-5.6

-5.2

L

-6.0

1

~

!

-5.6

!

-"L2

Energy (eV) Fig 6 Calculated density-of-states (DOS) for MO13and S/Mo13 dusters. Only occupied states are included and the energies are reported with respect to the vacuum level. The left-side panel shows results for clean Mo13, whereas the center panel shows the corresponding values for S/Mo13. MoB refers to the contribution of a metal atom in a site where sulphur adsorbs. The right-side panel compares the DOS of this metal atom before and after bonding to S. Reprinted from ref. [22].

475 3. INTERACTION OF SULPHUR W I T H Ag/Ru(0001) AND Cu/Ru(0001) Silver and copper are also used as "inert" site blockers when preparing hydrocarbon reforming catalysts [4,7,31,33]. With respect to sulphur, they are more reactive than gold and can form bulk sulphides [42]. Thus, when sulphur is dosed to Ag/Ru(0001) and Cu~u(0001) [13], it weakens Ru-Ag and Ru-Cu interactions at low coverages, but at large sulphur coverages AgSx and CuSx are formed. Figure 7 displays Ag-TDS spectra recorded after depositing silver at 300 K on Ru(0001) surfaces with different coverages of sulphur (0, 0.12, 0.21 and 0.44) [13]. The silver atoms bonded directly to clean Ru(0001) desorb near 1000 K. In the presence of sulphur there is a significant weakening of the Ru-Ag bonds. For {0Ag > 0, 0S > 0.5 } systems, the results of Auger spectroscopy suggest the formation of AgSx on the ruthenium substrate at room temperature [ 13]. The top part of Figure 8 shows Agand S2-TDS spectra acquired during the thermal decomposition of a Ag2S film generated by adsorption of $2 on a Ag/Ru(0001) surface [13]. Desorption of a small amount of $2 is observed between 350 and 450 K, with most of the sulphur evolving into gas phase at temperatures from 750 to 900 K. The $2 desorption peak at high temperatures exhibits a line shape that is characteristic of zero-order desorption kinetics. For this peak, a plot of the desorption rate against 1/T yields a straight line (see Fig 8B), with an apparent activation energy of 48.8 kcal/mol. This value is very close to the enthalpy of decomposition of bulk silver sulphide (2Ag2Ssolid" Ag-TDS: Ag/S/Ru(001 ) 0.44

0.00

o.s8

o_321_

..O v (/) r O E r",.O (/) r

0.62

0.44.~.....~. .'/

k

I

1

800

900

,,

I

1000

1100

Temperature (K)

Fig 7 Ag-TDS spectra acquired after depositing silver at 300 K on clean Ru(0001) and on surfaces precovered with 0.12, 0.21, and 0.44 ML of sulphur. Reprinted from ref. [13].

476

|

TDS: S/Ag/Ru(001 ) ..m

t-:D

0~=5.45

~d

iit

t~

--- - mass 64, S2 ' 9 mass 107, Ag

>, t-

1I

t-

I l

I,,,.

/I /I

E O 1..

o

09

/

r

/

P ,,

I

400

300

,

I

500

1

,,

600

/

.

I

I

700

800

1

.

.

.

.

.

I

.

900

.

.

.

1

1000

1100

T e m p e r a t u r e (K) mass 64, S 2

(~

mass 107, Ag

(~

t~

n- 6

-1

to

~.5 O (D

o

4

E==48.8 Kcal rnol"1

3

1.15

1.20

1.25

1000/T (K")

1.30

I

105

1

1.10

1000/T (K")

Fig 8 (A) S2" and Ag-TDS spectra acquired during the decomposition of a Ag2S film generated by reaction of sulphur with a Ag/Ru(0001) surface at 300 K. At 1100 K, after the thermal desorption experiment, only 0.45 ML of S were lel~ on the Ru(0001) surface. (B and C) Apparent activation energies for the main desorption peaks in part A. From ref. [13]. 4Agsolia + S2,ga~, AH = + 46.4 kcal/mol [13]). Alter decomposition of the silver sulphide at 800-900 K, a substantial amount of sulphur ( - 0.45 ML) remained

477 bonded to the Ru(0001) surface and the Ag adatoms formed three-dimensional clusters or particles. In Figure 8A, the position and shape of the silver desorption peak match those observed for desorption of silver multilayers from Ru(0001) [36]. The graph in Figure 8C indicates that the desorption of silver in Figure 8A follows zero-order kinetics with an apparent activation energy of 63.4 kcal/mol. This desorption activation energy is close to the heat of vaporization of metallic silver and the desorption activation energy for silver multilayers from Ru(0001) or other metal substrates [ 13,36]. Figure 9 shows an STM image recorded after adsorbing - 0.1 ML of sulphur on 0.8 ML of silver supported on Ru(0001) [44]. Initially, a mismatch between the lattice parameters of silver and ruthenium produced misfit dislocations in the structure of the metal overlayer (not shown) [24,44]. The sulphur adatoms attack preferentially these positions. Ag atoms are displaced from the Ru interface and their positions are occupied by sulphur atoms. Within the structure of the metal overlayer a highly ordered triangular lattice of silver vacancy islands forms (Figures

Fig 9 (a) 2000x2000/t~k2 image of a S/Ag/Ru(0001) system. Three ruthenium terraces are shown (stepping down from the bottom left to the upper right comer). The inset shows the Fourier transform of the image. (b) A 700x640/~2 z o o m On the STM image in (a). (c) Size distribution of the vacancy islands induced by sulphur adsorption. (d) Trajectories of the center-of-mass of four neareast-neighbor vacancy islands; the positions were measured every 20 seconds. Reprinted from ref. [44].

478

Fig 10 Atomicallyresolved STM image (115xl 15 ]k) of a large silver vacancy island, about 50 ]k in diameter. The island step edges of this and the smaller islands move much faster than the acquisition rate of the STM images, and thus appear "blurred". The cluster of nearly 50 sulphur adatoms inside the large island exhibits p(2x2) order. Reprinted from ref. [44]. 9a and 9b). The average area of these islands is -- 462 ,/k, with an standard deviation of-- 117 ,/k. Figure 10 displays an STM image for a typical silver vacancy island, where one can see sulphur atoms accommodated in a p(2x2) array. In summary, the results of TDS [13], photoemission [13,45] and scanning tunnelling

microscopy

[24,45]

indicate that at low sulphur coverages the

interactions between S and Ag on Ru(0001) can be classified as repulsive, in the sense that there is weakening of the Ru-Ag bond and no mixing of the adsorbates. Once the ruthenium substrate becomes saturated with sulphur, then attractive interactions between silver and sulphur are possible and AgSx is formed [13,45]. Very similar trends are observed for the coadsorption of sulphur and copper on Ru(0001) [ 13,23]. Figure 11 shows Cu- and S2-TDS spectra for the decomposition of a CuSx film on Ru(0001) [13]. The copper sulphide was formed after the adsorption of sulphur on a supported copper multilayer at 300 K. The initial stoichiometry of the sulphide was CUl.3S. An increase in temperature from 300 to 800 K produced desorption of a significant amount of $2. Photoemission spectra taken after heating the sample to 800 K revealed that at this point a film of Cu2S was present on top of the Ru substrate. This film decomposed at temperatures between 900 and 1100 K, producing evolution of $2 and Cu into gas phase (see Figure 11). After the crystal was heated to 1250 K, only a small amount of sulphur remained on the Ru(0001) surface (-- 0.4 ML) [13]. On Ru(0001), the first copper layer adopts a pseudomorphic structure that reflects the lattice constant of the underlying ruthenium [46]. Because the lattice

479

S/Cu/Ru(O001)

mass 63

fl!

mass 64 JC} v

/

_= (/) r

(D I..

E 0

i-.

/ [ \ J"-"

~

(D {3_ CO (/) (/)

I

[

400

600

800

I

I

1000

1200

Temperature (K) Fig 11 Cu- and S2-TDS spectra acquired during the heating of a Cux.3S film to 600, 800 and 1250 K. The film was prepared by dosing sulphur to a Cu multilayer (0c~= 4.55) at 300 K. Reprinted from ref. [13].

constant of copper is 5.5% smaller than that of ruthenium, the first Cu layer is under tensile strength. Part of the stress is relieved upon addition of more copper to the Cu~.0/Ru(0001) system. A sequence of strain-relieved structures develops for thicker copper films [46,47]. An anisotropically relaxed second Cu layer, consisting of three domains of double stripes is shown in Figure 12 [23]. The bright stripes are misfit dislocations buried at the Cu-Ru interface separating regions of fee and hcp stacking [23]. Figure 13 displays S 2p core level spectra recorded after exposure

Fig 12 STM image for a Cu second layer on Ru(0001). Reprinted from ref.[23].

480

BINDING E N E R G Y (e V) 166

165

164

163

162

161

160

,-p, &

Z;

Z;

.e .....

__._..//.,,

. . . . . .

.,,-. - ~

0.3

/

v

A

x

,

m sulphided o total

-

~

-

. . . . . . -__- i 0 . ~

_

~

0.2

0

0.1 0.0 0.01

'

'

'

'

. . . .

,

0.10

.

.

.

.

.

.

.

,

.

.

.

.

.

.

.

1.00

.

i

'

10.00

EXPOS URE I l A N G M U I R S i

Fig 13 S 2p core-level spectra for the adsorption of sulphur at 300 K on a striped Cu layer (Ocu~ 2 ML) supported on a Ru(0001) surface. Reprintedfrom ref.[23]. of the - 2ML thick copper layer to sulphur. The first spectra in the set display a well-defmed S 2p3/zl/2 doublet with the 2p3/2 component at a binding energy of 161.85 eV, an energy characteristic of adorbed atomic sulphur. The inset at the bottom of the figure shows the sulphur uptake curve based on the curve fitted and integrated experimental data. After the initial adsorption of-~ 0.2 ML of sulphur, a weak shoulder appears in the high binding energy side of the photoemission curve. This new feature is well defined at a sulphur coverage of 0.37 ML. Curve fitting of this spectrum (top of Fig 13) indicates that a sulphide is now present on the surface [23]. The intensity of the sulfide peak grows with increasing sulphur dose, while that of the adsorbed S levels off and even decreases.

481 The long induction period seen in Figure 13 for the formation of the sulphide is unusual. To determine the cause, STM was used to visualize structural changes of the surface [23]. The corresponding images are shown in Figure 14. At very low S coverages (0.001 ML), sulphur adsorbs mainly at the edge dislocations and one sees straight lines that contain 4 to 8 atoms (Figure 14A). As the coverage of sulphur increases, big morphological changes are seen in the Cu overlayer and new dislocations are induced by the adsorbate. At a sulphur coverage of 0.03 ML, Figures 14D and 15, the adsorbate self-organizes into a network of hexagons and close-packed equilateral S-triangles made of 18 atoms that bound the hcp stacking areas (top of Figure 15). This self-organizing network fluctuates in time (bottom of Figure 15). It disappears upon additional dosing of sulphur (not shown), well before the formation of a copper sulphide. The image quality at these higher sulphur coverages degrades and the final details of the conversion cannot be ascertained experimentally with STM [23]. Nevertheless, the results in Figures 14 and 15 illustrate quite clearly the magnitude of the structural perturbations that sulphur can induce in a bimetallic surface. Active sites for catalytic reactions can be completely destroyed in the presence of sulphur.

Fig 14 (A) Early stages of sulphur adsorption on the stripped Cu layer. Individual sulphur adatoms images as black dots are arranged in short rows and are found at the edge dislocations and less frequently on stripes. Estimated sulphur coverage < 0.01 ML. (B-D) Development of sulphur features with increasing sulphur coverage: sulphur adatoms self-organize in rows, hexagons, and equilateral triangles. Sulphur rows can be imaged as dark or bright lines depending on the tip status. Reprinted from ref. [23].

482

Fig 15 An image (7.3 nm x 6.9 nm) of sulphur self-organized in hexagons and equilateral triangles made of 18 sulphur adatoms. At room temperature and fixed S/Cu stoichiometry (0Cu -0.03 ML for this image) the observed structural patterns fluctuate for hours. Lower two time-lapse images (3.5 nm x 3.3 nm) taken 50 s apart show formation of new equilateral triangles. Reprinted from ref. [23].

4. ADMETAL PROMOTED SULPHIDATION OF Pt(111) AND Mo(110) A large number of studies described in this book indicate that the formation of a heteronuclear metal-metal bond can lead to important changes in properties of the bonded elements. large redistribution

of charge

the chemical

In many cases, bimetallic bonding induces a

around

the metals

[48-50].

In principle,

this

redistribution of charge could affect the reactivity of a metal toward sulphur. A very

483

interesting situation is found when silver or copper are added to Pt(111) [ 15,17,41]. Figure 16 compares Pt 4f core-level spectra acquired before and after dosing $2 to Pt(111) and a Ag/Pt(111) system with 2.26 ML of the admetal [ 17]. The exposure of P t ( l l l ) to large amounts of $2 produces only a chemisorbed layer of S, without forming bulk-like sulphides which are thermodynamically very stable (PtS2, AGe= 109 kJ/mol [42]). For the S/Pt(lll) system, two factors make difficult the penetration of S into the bulk of the metal. First, the surface free energy of sulphur (0.08 J m "2 [51]) is much lower than that ofplatinum (2.69 J rn"2 [51]). And second, the cohesive energy of metallic Pt is relatively large (564 kJ/mol [52]). If the influence of these two factors is somehow suppressed, then, the formation of platinum sulphides should take place. In Figure 16, new Pt 4f features for platinum sulphide are detected atter dosing $2 to a Ag/Pt(111) surface. The relative large intensity of these features indicates that a big amount (> 1 ML) of PtSx is formed [17]. Silver has a relatively low surface free energy (1.30 J rn a [51]), and its presence on the Pt surface probably frees sulphur for migration into the bulk of the sample. In the Ag-Pt(lll) bond there is a significant shift of electrons from the admetal toward the metal substrate [50,53] that favors the formation of Pt-.S dative bonds. In addition, silver sulphides could promote the formation of platinum sulphides by inducing changes in the structural geometry that enhance the diffusion of sulphur into the lattice ofmetallic Pt [17]. Figure 17 displays photoemission Pt 4f

.--=

~,

-

S/Pt(111)

.

i._~_~o~_m~.,.----, ~ r--- . . . . .

82

80

78

J

/

76

i.

"-, i >.,,.j

74

i

72

i t \ o.oo

',._x,~ . o.oo

70

68

Binding Energy (eV)

Fig 16 Pt 4fXPS spectra acquired after doing S2 to Pt(111), bottom, and Ag/Pt(111) surfaces, top. Reprinted from ref. [17].

484 Valence: S/Ag/Pt(111 )

,~

"12 (I) .N

i .s (1)

12

....

I

10

i

8

I

6

I

4

I

2

I

0

Binding Energy (eV)

Fig 17 Valence photoemission data for Pt(lll), Ag/Pt(lll) and S/Ag~t(lll) surfaces. Initially, 0.21 ML of silver were vapor-deposited at ~ 300 K, and the Ag/Pt(111) surface was annealed to 550 K before dosing S2 at this temperature. Reprinted from ref. [41]. data for the valence region of Pt(111), Ag/Pt(111) and S/Ag/Pt(111) [41 ]. Pt(111) and Ag/Pt(111) exhibit a substantial DOS near the Fermi level and are chemically and catalytically active. The silver-induced formation of PtSx in S/Ag/Pt(111) leads to a very large drop in the DOS around the Fermi level, hindering the ability of the system to respond to the presence of adsorbates. Thus silver, ideally added as an inert site blocker to reduce C-C hydrogenolysis on Pt reforming catalysts [31-33], can actually accelerate the rate of sulphur poisoning. Copper also promotes the rate of sulphidation of platinum [15], but not all the admetals used as site blockers (Zn, AI, Sn) in Pt-based reforming catalysts behave in this way [15,25,26,29]. Figure 18 shows Pt 4f core-level spectra acquired after adsorbing sulphur on Pt(111) and several bimetallic systems. Strong peaks are seen for PtS~ in S/Ag/Pt(111) [17] and S/Cu/Pt(111) [15]. No platinum sulphide formation is observed for S/Zn/Pt(lll) [15] and S/AIIPt(lll) [25]. In Zn-Pt and A1-Pt bonds there is a net charge transfer toward platinum [54-56] that should facilitate the formation of Pt-* S dative bonds. In addition, zinc and aluminium (like silver and copper) have a smaller surface free energy than platinum [51]. However, the Zn-Pt (or AI-Pt) bonds break apart in the presence of sulphur and the Pt.*ZnSx (or Pt.-A1Sx) interactions are weak. After analysing the results in Figure 18, one can conclude that an admetal~Pt charge transfer and a low surface-free energy for the

485 admetal may be necessary, but insufficient conditions for seeing a promotional effect of the admetal on the formation of platinum sulphides [25]. On the other hand, the relative stabilities of the admetal sulphides may have a direct impact on whether or not sulphidation of the Pt substrate will occur. In bimetallic system~ where the admetals form sulphides of higher stability than those formexl by platinum (Zn/Pt and A1/Pt) [42], the adsorption of sulphur stops once the admetal is saturated with sulphur and no PtS~ is formed. Not included in Figure 18 are data for the S / S n ~ t ( l l l ) system [26]. In this special system, bimetallic bonding acamlly reduces the reactivity of both metals toward sulphur [26-28]. This can be usefifl for the prevention of sulphur poisoning and will be the subject of section 5. Pt 4f: S 2 at 550 K

PtSx

i

Pt

I

-AHfof admetal sulfide

Ag2.2e/Pt(111 )

Cu-Pt surface alio,

Zn-Pt surface

AI-Pt surface alloy

Pt(111)

'

82

I .............. '

80

1

78

'"

I

76

'

I

74

'

I

72

'

I

70

'

68

Binding Energy (eV)

Fig 18 Pt 4fXPS results comparingthe effect of dosing S2 at 550 K to clean Pt(111) and a series of X/Pt systems {X= Ag, Cu, Zn and A1}. The heat of formation for the sulphides of the admetals increases (more exothermic) when going from the top to the bottom of the figure. Reprinted from rcf. [25].

486 The sulphidation of Mo(ll0) is promoted by a series of admetals (Fe, Co, Ni, Cu, Ag and Zn) [19,20,57,58] that form sulphides that are less stable than those formed by molybdenum [42]. Figure 19 displays Mo 3d XPS spectra acquired upon dosing $2 to Mo(110) and Mo(110) surfaces with similar coverages (-~ 1.5 ML monolayers ) of nickel [19], copper [16], zinc [20] and silver [16]. These and other results [57,58] indicate that the amount of MoSx formed depends strongly on the nature of the admetal. Specifically, nickel and cobalt have a unique ability to promote Mo.*S interactions and the formation of molybdenum sulphide [19,57,58]. Results for the reaction of $2 with a series of X/Mo(ll0) surfaces (X=admetal) indicate that the "promotional effect" of an admetal increases following the sequence: Ag = Zn < Cu < Fe < Co < Ni [19,57,58]. Figure 20 compares trends observed in the activity of a series of XSy/MoS2

Mo 3d XPS S2 on X/Mo(110) T=700 K

Mo Or) ~

r

v

S on pure Mo

tr

0s

Ag, 1.4 ML Zn, 1.4ML Cu, 1.3 ML Ni, 1.5 ML

I

MOSy '

I

234

'

I

232

'

i

230

~ '

i

228

'

] I

226

Binding Energy (eV) Fig 19 Mo 3d XPS spectra acquired after dosing large amounts of $2 to clean Mo(110) and X/Mo(110) surfaces (X= Ag, Zn, Cu or Ni) at 700 K. The spectra correspond to systems in which the rate of $2 adsorption has become zero under UHV conditions. Reprinted from ref. [19].

487

E ID T-

e r0 o O

Ni

oo/ "

4

X

m X ~> 3 ~.O

c0

.>__m ~ 2 '6 o if)

Zrl

o

r" -5 o 9 0

Fe Ir

Cu

Mo

go

I

0.00

0.25

'

I

0.50

'

I

0.75

'

I

1.00

1.25

MoSy / Mo 3d5/2 XPS Area Ratio in S/X/Mo(110)

Fig 20 X axis: relative amount of MoSyformed after exposing X1.5/Mo(110) surfaces (X= Zn, Cu, Fe, Co and Ni, with 0 x 1.5 ML) to $2 at 700 K_ Y axis: activity of MoS2 and XSy/MoS2 catalysts for the hydrodesulphurization of dibenzothiophene (DBT). Reprinted from ref. [61]. catalysts (X = Zn, Cu, Fe, Co or Ni) during the hydrodesulphurizafion of dibenzothiophene [59,60] with trends found for the sulphidation of molybdenum in X/Mo(ll0) surfaces [19,58,61]. In general, a good correlation is observed between the changes in the two chemical properties. The presence of Ni leads to a significant enhancement in the Mo~S interactions and a very large HDS activity. In contrast, the effects of Zn, Cu, and Fe on the Mo*.S interactions and HDS activity are less pronounced. From the correlation in Figure 20, it is clear that the effects of bimetallic bonding can be useful in HDS catalysis. Three factors probably contribute to the large HDS activity of NiMoSx catalysts [19,58,61]: (1) the existence of Ni centers that have S-free sites on which a S-containing molecule can adsorb; (2) the presence of Ni-Mo sites that are very reactive for the desulphurization of the adsorbed molecule; and (3) on the S-free Ni sites hydrogen molecules can dissociate, producing in this way a source of hydrogen atoms that helps to remove sulphur from the surface and keeps a large number of unsaturated Mo and Ni sites. Ag/Mo(110) and Zn/Mo(110) are very useful for the synthesis of MoSx films under UHV conditions [16,20,57,61]. The dosing of S: to Ag/Mo(ll0) and Zn/Mo(110) produces bimetallic sulphides, but upon heating to 1000-1100 K the silver and zinc desorb, leaving films of pure MoSx on top of the Mo(110) substrate. Following this methodology, films that have between 2 and 6 sulphide monolayers

488 can be prepared. The films exhibit Mo 3d and S 2p XPS spectra that are very similar to those of MoS2. They show no reactivity toward CO, 02 or H2 at 100-400 K. But they can be activated aider the creation of S vacancies by reaction with atomic H [62], providing convenient surfaces for examining the chemistry of desulphurization reactions on molybdenum sulphide [63]. 5. B I M E T A L L I C BONDING AND THE PREVENTION OF SULPHUR POISONING In the previous section we have discussed several cases in which bimetallic bonding increases the overall reactivity of a system towards sulphur. If the opposite occurs, such a phenomenon could be useful for the prevention of sulphur poisoning. In practical terms, the idea is to fmd bimetallic systems that have a good catalytic activity and are less sensitive to the presence of sulphur-containing molecules in the feedstream than pure metals. Sn/Pt and Pd/Rh satisfy these requirements [26-29]. Pt-Sn bimetallic catalysts are widely used for hydrocarbon reforming or dehydrogenation reactions [4,5,64-66]. In Sn/Pt alloys, there is a redistribution of charge and both metals accumulate electrons around the Pt-Sn bonds [26,67-69]. The effects of bimetallic bonding on the chemical properties are very dramatic in the case of fin [26,27]. In the presence of $2, tin does not get fully sulphided as other metals (A1, Zn, Cu, Ag) do when they are supported on Pt(111) [15,17,25]. The formation of Sn-Pt bonds reduces the electron density of tin and the metal has difficulties responding to the presence of sulphur-containing molecules [26,27,29]. The bottom of Figure 21 compares the uptake of sulphur and SOx species after dosing SO2 to polycrystalline Sn, P t ( l l l ) , and a (~3x-/'3)R30~ surface alloy [27]. The top of the figure shows the structural geometry of the Sn/Pt alloy. Sn atoms are present only in the top layer and protrude 0.22 ]k out of the plane of Pt atoms [70,71]. Each Pt atom present in the surface has the same number of Pt and Sn neighbours (3 and 3). In the alloy, there are plenty of a-top and bridge Pt sites that can adsorb and dissociate a small molecule like SO2. Figure 21 indicates that pure tin is much more reactive than pure platinum. In fact, photoemission studies indicate that even at temperatures as low as 100 K, tin reacts vigorously with SO2 [27]. Therefore, one could expect that Sn adatoms would enhance the ability of the P t ( l l l ) surface to adsorb and dissociate SO2. However, the (-/'3x~3)R30 ~ Sl~t(111) surface alloy exhibits a reactivity smaller than that of pure Sn or Pt(111). It may be argued that the low reactivity of the alloy with respect to tin is due to the fact that the bimetallic system does not have adsorption sites with two or three adjacent tin atoms ("ensemble effects" [31,32]). But the differences in reactivity

489

OSn=0.33 ML

(4"3x~3)R30~

Sulphur Uptake

0.4

300-310 K polycrystalline Sn

0.3

Pt(111)

,~ 0.2

_____--&

Sn/Pt(111)

.~ 0.1

0.0 0

-

I

I

I

I

I

2

4

6

8

10

802 Exposure (L) Fig 21 Top: Structural geometry for a (-]'3x4"3)R30~ surface alloy. The dark and white circles represent Sn and Pt atoms, respectively. The Sn atoms are present only in the top layer. Bottom: Total sulphur uptake (SOx + S) for the adsorption of SO2 on polycrystalline Sn, Pt(111), and a (~f3xC'3)P,30~ alloy. Reprinted from ref. [27].

490 between P t ( l l l ) and ((3x(3)R30~ can only be explained invoking "electronic effects", since in the surface alloy there are plenty of adsorption sites with two or three adjacent Pt atoms and some Pt atoms are being replaced with Sn atoms which, in principle, should be more reactive. The importance of "electronic effects" has been confirmed by theoretical calculations [27,29]. Ab initio SCF calculations indicate that the Pt atoms in ((3x(3)R30~ interact poorly with the LUMO of SO2, leading to a small adsorption energy and hindering the dissociation of S-O bonds [27]. Density-functional slab calculations for the adsorption of atomic sulphur on a p(2x2)-Sn/Pt(lll) surface give adsorption energies on the pure Pt hollow sites that are 7-9 kcal/mol smaller than on Pt(111) [29]. Thus, "electronic effects" probably play an important role in the low chemical affinity of Sn/Pt alloys for sulphur-containing molecules ($2, H2S, SO2, thiophene, etc) [26-29,72]. This does not imply that "ensemble" [32,72] or "geometrical effects" [73] are negligible. For example, in the case of thiophene on p(2x2)SnfPt(111) and (4"3x(3)R30~ one is dealing with a bulky adsorbate and small ensembles of Pt atoms [32,72] or geometrical blocking of Pt.-adsorbate interactions by tin [72,73] help to prevent the decomposition of the sulphurcontaining molecule. Cu, Ag and Sn are added to Pt catalysts as site blockers to improve their selectivity for the reforming of hydrocarbons [4,31,33,64-66]. In this respect the effects of the admetals are more or less similar. From the trends discussed above, it is clear that tin is a much better choice than Cu or Ag when trying to minimize the sensitivity of Pt reforming catalysts toward sulphur poisoning. Palladium has a high catalytic activity for the selective hydrogenation of olefms, the oxidation of alcohols, cyclotrimerization of acetylene, and the removal of CO and NO from automobile exhaust gases [3-5,7]. One of the major limitations in the use of Pd in industrial catalysis is its extreme sensitive to sulphur poisoning [6,74]. Experimental and theoretical studies indicate that bimetallic bonding can reduce the reactivity of palladium toward sulphur-containing molecules [28,72,75-77]. The interaction of SO2 with Pd in bimetallic systems has been studied in detail using a combination of photoemission and theoretical (ab initio SCF, density functional) calculations [28,72,77]. On pure palladium surfaces, SO2 adsorbs molecularly at 100 K and dissociates (60-70%) at temperatures between 200 and 400 K leaving large coverages (> 0.5 ML) of S and O on the surface [28]. A very different behaviour is found for the adsorption of SO2 on a palladium monolayer supported on R h ( l l l ) [28]. At 100 K, SO2 chemisorbs molecularly on a Pdl.dRh(111) surface and heating to 300 K produces the desorption of almost 80% of the adsorbed SO2, leaving a few S adatoms and no SO• species on the surface. In this respect, the Pdl.0/Rh(lll)

491

t

9 ~176

~ O

6 9

E

0eV

Pd site 4d o r b i t a l - ' l t ' ~ E~

9

SO 2 3b 1 LUMO

~

.

~

".

1'1 1r

Q ~ 132/(ELuMO - E,~ )

Fig 22 Bonding interactions between the LUMO of SO2 and an occupied Pd 4d orbital. Reprinted from ref. [28]. system is less chemically active than polycrystalline Pd, Pd(100), or R h ( l l l ) [28]. The results of theoretical studies clearly indicate that bimetallic bonding weakens the Pd'*SO2 bonding interactions [28,77]. In the bond between SO/and palladium, a transfer of electrons from the metal into the LUMO of SO2 (see Figure 22) plays a dominant role in the bonding energy of the molecule [77,78]. This g back donation leads to a weakening of the S-O bonds, since the LUMO of SO2 is S-O antibonding. On the Pdl.0/Rh(lll) surface, the Pd--Rh interactions reduce the electron donor ability of palladium producing weaker Pd-SO2 bonds and stronger S-O bonds than on Pd(111) [28,77]. Even much weaker adsorption bonds are found when Pd is supported on surfaces of s,p or early transition metals [28,72,77]. For example, in Pdl.0/Mo(ll0) and Pdl.0/Al(lll), bimetallic bonding largely shifts the Pd 4d band toward higher binding energy [48] preventing effective interactions with the LUMO of SO2 (i.e. very large End to ELtn~o separations in the diagram of Figure 22) [77]. A similar principle is useful for reducing the rate of thiophene dissociation on Pd/Mo(ll0) [72,78]. When following this approach one has to fmd a good balance between the decrease in the overall catalytic activity of Pd and its affinity for sulphur [77]. Such a balance has been observed in the case ofPd/Rh, Pd/Mn and Pd/Ni catalysts [75,76,79]. All these results together indicate that bimetallic bonding is a viable route for increasing the sulphur tolerance of metal catalysts.

492 6.

CONCLUSION

In recent years, several new interesting phenomena have been discovered when studying the interaction of sulphur with bimetallic surfaces using the modem techniques of surface science. Very small amounts of sulphur can induce dramatic changes in the morphology of bimetallic surfaces. The electronic perturbations associated with the formation of a heteronuclear metal-metal bond affect the reactivity of the bonded metals toward sulphur. This can be a very important issue to consider when trying to minimize the negative effects of sulphur poisoning or dealing with the design of desulphurization catalysts. ACKNOWLEDGEMENT Many of the studies described above were done in collaboration with M. Kuhn, S. Chaturvedi, T. Jirsak, S.Y. Li, J. Dvorak and R.Q. Hwang. Special thanks to all of them for their superb contributions. This work was carried out at Brookhaven National Laboratory under Contract DE-AC02-98CH10086 with the US Department of Energy (Division of Chemical Sciences).

REFERENCES [1] J.G. Speight, The Chemistry and Technology of Petroleum, 2nd ed, Dekker, New York, 1991. ' [2] A.C. Stern, R.W. Boubel, D.B. Turner, and D.L. Fox, Fundamentals of Air Pollution, 2nd ed, Academic Press, Orlando, 1984. [3] K.C. Taylor, Catal. Rev. Sci. Eng. 35 (1993)457. [4] J.M. Thomas and W.J. Thomas, Principles and Practice of Heterogeneous Catalysis, VCH, New York, 1997. [5] B.C. Gates, Catalytic Chemistry, Wiley, New York, 1992. [6] C.H. Bartholomew, P.K. Agrawal and J.IL Katzer, Adv. Catal. 31 (1982) 135. [7] G. Ertl, H. KnSzinger, and J. Weitkamp (eds.), Handbook of Heterogeneous Catalysis, Wiley-VCH, New York, 1997. [8] R.R. Chianelli, M. Daage, and M.J. Ledoux, Adv. Catal. 40 (1994) 177. [9] C.C. Knight and G.A. Somorjai, Surf. Sci. 240 (1990) 101. [10] D.A. Chela, C.M. Friend, and H. Xu, Langmuir 12 (1996) 1528. [ 11] M. Kuhn and J.A. Rodriguez, Chem. Phys. Lett. 231 (1994) 199. [12] M. Kuhn, J.A. Rodriguez and J. Hrbek, Surf. Sci. 314 (1994) L897. [13] M. Kuhn and J.A. Rodriguez, J. Phys. Chem. 98 (1994) 12059. [14] J.C. Dunphy, C. Chapelier, D.F. Ogletree and M.B. Salmeron, J. Vac. Sci. Technol. B, 12 (1994) 1742. [15] M. Kuhn and J.A. Rodriguez, Catal. Lett. 32 (1995) 345. [16] J.A. Rodriguez and M. Kuhn, J. Phys. Chem. 99 (1995) 9567. [17] M. Kuhn and J.A. Rodriguez, J. Catal. 154 (1995) 355. [18] F.H. Ribeiro, A.L. Bonivardi, C. Kim and G.A. Somorjai, J. Catal. 150 (1994) 186. [19] M. Kuhn and J.A. Rodriguez, Surf. Sei. 355 (1996) 85. [20] M. Kuhn and J.A. Rodriguez, Surf. Sci. 336 (1995) 1.

493 [21] W.K. K-uhn, J.-H. He, and D.W. Goodman, J. Vac. Sci. Technol. A, 10 (1992) 2477. [22] J.A. Rodriguez, M. Kuhn and J. Hrbek, J. Phys. Chem. 100 (1996) 3799. [23] J. Hrbek, J. de la Figuera~ K. Pohl, T. Jirsak, J.A. Rodriguez, A.K. Schmid, N.C. BarteR, and R.Q. Hwang, J. Phys. Chem. B, 103 (1999) 10557. [24] K. Pohl, M.C. Bartelt, J. de la Figuera, N.C. BarteR, J. Hrbek, and R.Q. Hwang, Nature, 397 (1999) 238. [25] J.A. Rodriguez and M. Kuhn, J. Vac. Sci. Technol. A, 15 (1997) 1608. [26] J.A. Rodriguez, S. Chaturvedi, T. Jirsak, and J. Hrbek, J. Chem. Phys. 109 (1998) 4052. [27] J.A. Rodriguez, T. Jirsak, S. Chaturvedi, and J. Hrbek, J. Ant Chem. Soc. 120 (1998) 11149. [28] J.A. Rodriguez, T. Jirsak and S. Chaturvedi, J. Chem_ Phys. 110 (1999) 3138. [29] J.A. Rodriguez, J. Hrbek, M. Kuhn, T. Jirsak, S. Chaturvedi and A. Maiti, J. Chem. Phys. 113 (2000) 11284. [30] J.A. Rodriguez and J. Hrbek, Accounts of Chem. Research, 32 (1999) 719. [31] J.H. Sinfelt, Bimetallic Catalysts, Wiley, New York, 1983. [32] W.H.M. Sachtler, Faraday Disc. Chem. Soc. 72 (1981) 7. [33] V. Ponce, Adv. Catal. 32 (1983) 149. [34] J.A. Rodriguez and D.W. Goodman, Surf. Sci. Reports 14 (1991) 1. [35] S. Galvagno et al, J. Catal. 69 (1981) 283; 61 (1980) 223. [36] J.W. Niemantsverdriet, P. Dolle, K. Markert and K. WandeR, J. Vacuum Sci. Technol. A, 5 (1987) 875. [37] S.R. Kelemen and T.E. Fisher, Surf. Sci. 87 (1979) 53. [38] J.A. Rodriguez, J. Dvorak, T. Jirsak and J. Hrbek, Surf. Sci. 490 (2001) 315. [39] J. Hrbek, J. de la Figuera, K. Pohl, A.K. Schmid, N.C. Barter and 1LQ. Hwang, to be published. [40] R.Q. Hwang, J. Schroder, C. Gunther and R.J. Behm, Phys. Rev. Lett. 67 (1991) 3279. [41] J.A. Rodriguez, M. Kuhn and J. Hrbek, J. Phys. Chem. 100 (1996) 15494. [42] Lange's Handbook of Chemistry, 13th ed, McGraw-Hill, New York, 1985 [43] J.A. Rodriguez, J. Dvorak and T. Jirsak, Surf. Sci. 457 (2000) IA13. [44] K. Pohl, J. de la Figuera, M.C. BarteR, N.C. BarteR, J. Hrbek and R.Q. Hwang, Surf. Sci. 433-435 (1999)506. [45] J. Hrbek, M. Kuhn and J.A. Rodriguez, Surf. Sci. 356 (1996) L423. [46] G.O. Potshke and R.J. Behm, Phys. Rev. B, 44 (1991) 1442. [47] C. Gtmther, J. Vrijmoeth, R.Q. Hwang, and ILL Behm, Phys. Rev. Lett. 74 (1995) 754. [48] J.A. Rodriguez, Surf. Sci. Reports, 24 (1996) 223. [49] R. Wu and A.J. Freeman, Phys. Rev. B, 52 (1995) 12419. [50] P.J. Feibelman, Surf. Sci. 313 (1994) L801. [51] L.Z. Mezey and J. Giber, Jpn. J. Appl. Phys. 21 (1982) 1569. [52] C. Kittel, Introduction to Solid State Physics, 6th ed, Wiley, New York, 1986. [53] J.A. Rodriguez and M. Kuhn, J. Phys. Chem. 98 (1994) 11251. [54] J.A. Rodriguez and M. Kuhn, J. Chem. Phys. 102 (1995) 4279. [55] R.E. Watson and L.H. Bennett, Phys. Rev. B, 15 (1977) 5136. [56] J.A. Rodriguez and M. Kuhn, Chem. Phys. Lett. 240 (1995) 435. [57] J.A. Rodriguez, S.Y. Li, J. Hrbek, H.H. Huang and G.-Q. Xu, J. Phys. Chem. 100 (1996) 14476. [58] J.A. Rodriguez, S.Y. Li, J. Hrbek, H.H. Huang and G.-Q. Xu, Surf. Sci. 370 (1997) 85. [59] S. Harris and R.R. Chianelli, J. Catal. 98 (1986) 17. [60] R.R. Chianelli, T.A. Pecoraro, T.R. Halbert, W.-H. Pan, and E.I. Stiefel, J. Catal. 86 (1984) 226. [61] J.A. Rodriguez, Polyhedron, 16 (1997) 3177.

494 [62] S.Y. Li, J.A. Rodriguez, J. Hrbek, H.H. Huang, and G.Q. Xu, Surf. Sei. 366 (1996) 29. [63] J.A. Rodriguez, J. Dvorak, T. Jirsak, S.Y. Li, J. I-Irbek, A.T. Capitano, A.M. Gabelnick, and J.L. Gland, J. Phys. Chem. B, 103 (1999) 8310. [64] C. Xu, J.W. Peck and B.E. Koel, J. Am. Chem. Soc. 115 (1993) 751. [65] O.A. Barias, A. Holmen, and E.A~ Blekkan, J. Catal. 158 (1996) 1. [66] J. Szanyi and M.T. Paffett, J. Am. Chem. Soe. 117 (1995) 1034. [67] S. Pick, Surf. Sci. 436 (1999) 220. [68] Y. Jeon, J. Chen, and M. Croli, Phys. Rev. B, 50 (1994) 6555. [69] P. Ross, J. Vac. Sci. Technol. A, 10 (1992) 2546. [70] S.H. Overbury, D.R. Mullins, M.T. Paffett and B.E. Koel, Surf. Sci. 254 (1991) 45. [71] S.H. Overbury and Y.-S. Ku, Phys. Rev. B, 46 (1992) 7868. [72] J.A. Rodriguez, J. Dvorak and T. Jirsak, to bo published. [73] C. Xu and B.E. Koel, Surf. Sci. 327 (1995) 38. [74] J.A. Rodriguez, S. Chaturvedi and T. Jirsak, Chem. Phys. Lett. 296 (1998) 421. [75] P.C. L'Argentiere, M.M. Cation, N.S. Figoli and J. Ferron, Appl. Surf. Sci. 68 (1993) 41. [76] P.C. L'Argentiere, M.M. Cation and N.S. Figoli, Appl. Surf. Sci. 89 (1995) 63. [77] J.A. Rodriguez and L. Gonzalez, to be published. [78] H. Sellers and E. Shustorovich, Surf. Sci. 346 (1996) 322. [79] D.M. DiCicco, A.A. Adamczyk, and K.S. Patel, Book of Abstracts for the 210 th American Chemical Society National Meeting (Chicago, August 1995), Fuel-145.


Surface Alloys, and Alloy Surfaces D.P. Woodruff, (Editor)

495

Chapter 14

Adsorbate induced segregation at bimetallic surfaces C.J. Baddeley

School of Chemistry, University of St Andrews, St Andrews, Fife, KY16 9ST, United Kingdom. 1. I N T R O D U C T I O N The surface and interface chemistry of bimetallic surfaces is an important subject for a variety of technological reasons including corrosion resistance and hardening, metal-metal interfaces, microelectronics fabrication, electrochemistry, surface magnetic films and heterogeneous catalysis [1]. Bearing in mind, the huge economical importance of heterogeneous catalysis, it can be argued that this aspect of bimetallic surface chemistry can be regarded as the most important. This chapter concentrates primarily on issues of heterogeneous catalysis. The thermodynamic and kinetic factors are outlined that are important in defining the surface chemistry of bimetallic surfaces. In addition, the various approaches will be introduced that are utilised by surface scientists in an attempt to measure the composition of bimetallic surfaces under the influence of adsorbates. Furthermore, the chapter will investigate the difficulties encountered when attempting to obtain accurate measurements on nanoscale bimetallic particles under environments typically encountered in a catalytic reaction. By way of contrast, the relevance of much more accurate measurements on well-defined surfaces under idealised ultrahigh vacuum (UHV) conditions will be questioned. 1.1. Bimetallic surface chemistry - traditional ideas Bimetallic catalysts have often been shown to outperform their monometallic counterparts in terms of both activity and selectivity [2]. There are now many examples of catalytic reactions which have been studied over bimetallic systems some of which are summarised in Table 1.

496 Table 1: Some reactions catalysed by bimetallic systems (adapted from [ 1]). Reaction CO oxidation

Bimetallic system Cu/Cr Cu/Pd Pt,Pd and Rh alloys

Reference [3] [4] [5]

dehydrogenation

Ni/Cu Ni/W Ni/Sn Pt/Co

[6] [7] [8] [9]

acetylene cyclotrimerisation

Pd/Au Pd/Sn other Pd alloys

[10] [11] [12]

Fischer-Tropsch synthesis

Ru-Group IB alloys Fe/Ru, Fe/Ni, Co alloys CufPd

[13] [5] [141

Exhaust emission conversion

P t ~ h or Pd

[15]

Olefin hydrogenation

Pd/T1 Pd/Cu, Sn or Fe Pd/Fe Pd/Co

[161 [ 17] [18] [19]

Hydrocarbon reactions Reforming

Ni, Pt, Pd, Ru based alloys Pt/Re, Pt/Ir, Pt/Au Pt/Sn, multimetallic systems

[20] [21-24] [5]

CO methanation

W/Ni, W/Co, Ru/Cu Ce based intermetallics

[25] [26]

alkane hydrogenolysis

W/Ni, Pt/Ni, Pt/Re, Ru/Cu Cu~d

[25]

hydrodesulphurisation

Co/Mo, Ni/Mo

[28-29]

hydrodenitrogenation

Ni/Mo

[30]

hydrogenation of edible oils and fats

Ni and Ni based alloys

[31]

[27]

Traditionally, there have been two reasons proposed for the enhanced performance of a bimetallic catalyst over each monometallic counterpart. These are k n o w n as ensemble (structural) effects and electronic (ligand) effects.

497

1.1.1. Ensemble effects The idea of the importance of surface structure in the chemistry of bimetallic surfaces relies largely on the concept of the "active site" for a particular surface chemical reaction. If one reaction requires the presence of three-fold hollow sites on an fcc (111) facet, while a competing reaction requires only single atomic sites, then the random dilution of the active metal by an inert second metal would rapidly deplete the number of active three-fold ensembles. By contrast, the number of single atom sites available would be depleted much less dramatically as a function of composition. Thus, the selectivity of the catalyst would vary with surface composition. An example where this effect is observed dramatically is in the trimerisation of ethyne (C2H2) to benzene (C6H6) over Au/Pd surfaces [32]. Detailed investigations by Lambert and co-workers [33] proposed that, over Pd(111), the trimerisation reaction requires a relatively large ensemble of Pd atoms constituting a central atom surrounded by a hexagonal array of Pd atoms - i.e. a Pd7 cluster [33]. On alloying with Au, the activity of the AuxPdl00_x(lll) surfaces, as measured by TPD experiments, varies in a dramatic way as a function of surface composition as shown in Fig. 1.

0

20

40

60

80

100

Composition/atom % Pd

Figure 1: The yield of benzene from random PdAu alloys on Pd(lll) as a function of composition (white circles). Also shown is the theoretical fit (filled black squares) which is formed by summing the contribution from Pd7 ensembles (open squares) and AuPd6 ensembles (white squares) [32].

498 The pronounced maximum in activity at a composition of-~Au15Pd85 was attributed to the presence of a second active ensemble, AuPd6. The number density of AuPd6 ensembles was considered to vary as 06(1-0) where 0 is the mole fraction of Pd in the surface layer. This function reaches a maximum at 0 = 6/7 and is clearly zero at 0 - 0 and 1. It was also known that the selectivity to benzene formation approaches 100% over PdAu surface alloys on Au(111) [34]. The experimental data show that there is considerable activity at 0 - 1, this activity being that of a clean P d ( l l l ) surface. Using the fact that the trimerisation reaction is only 25% selective over Pd(111) due to competing hydrogenation and decomposition reactions, and assuming that the number density of Pd7 ensembles varies as 07, the measured activity of the surface as a function of composition could be modelled closely as a linear superposition of the statistically derived number of AuPd6 ensembles (100% active for the cyclisation process) and Pd7 ensembles (25% active) assuming the surface to consist of a random arrangement of Au and Pd atoms. This conclusion was supported by measurements on AuPd colloidal catalysts where the selectivity of the catalysts showed a strong correlation with the extent of alloying in the particles [10]. X-ray diffraction data showed that the bimetallic colloids consisted of a (-7.5 nm) Au core with a thin (

The Chemical Physics of Solid Surfaces, Vol. 10: Surface alloys and alloy surfaces

Oxide Surfaces (The Chemical Physics of Solid Surfaces Vol. 9)

Solid state physics: surfaces

Solid Surfaces and the Gas-Solid Interface

Physics of Surfaces and Interfaces

Physics of Surfaces and Interfaces

Physics of Surfaces and Interfaces

Methods of Experimental Physics - Solid State Physics - Surfaces

Photochemistry on Solid Surfaces (Studies in Surface Science and Catalysis)

Surfaces

Surfaces

Surfaces

Surfaces

Surfaces

The Chemical Physics of Solid Surfaces : Growth and Properties of Ultrathin Epitaxial Layers

Surfaces

Solid surfaces, interfaces and thin films

Solid Surfaces, Interfaces And Thin Films

Chemical Bonding at Surfaces and Interfaces

Chemical bonding at surfaces and interfaces

Chemical Bonding at Surfaces and Interfaces

Reactions at Solid Surfaces (Baker Lecture Series)

Physics and Chemistry at Oxide Surfaces

Physics and chemistry at oxide surfaces

Equilibria and Dynamics of Gas Adsorption on Heterogeneous Solid Surfaces (Studies in Surface Science and Catalysis)

Surfaces Algebriques

Rough Surfaces

Rough Surfaces

Spillover and Mobility of Species on Solid Surfaces ,Volume 138 (Studies in Surface Science and Catalysis)

Equilibria and Dynamics of Gas Adsorption on Heterogeneous Solid Surfaces (Studies in Surface Science and Catalysis)

Riemann surfaces

The Chemical Physics of Solid Surfaces, Vol. 10: Surface alloys and alloy surfaces

Oxide Surfaces (The Chemical Physics of Solid Surfaces Vol. 9)

Solid state physics: surfaces

Solid Surfaces and the Gas-Solid Interface

Physics of Surfaces and Interfaces

Physics of Surfaces and Interfaces

Physics of Surfaces and Interfaces

Methods of Experimental Physics - Solid State Physics - Surfaces

Photochemistry on Solid Surfaces (Studies in Surface Science and Catalysis)

Surfaces

Surfaces

Surfaces

Surfaces

Surfaces

The Chemical Physics of Solid Surfaces : Growth and Properties of Ultrathin Epitaxial Layers

Surfaces

Solid surfaces, interfaces and thin films

Solid Surfaces, Interfaces And Thin Films

Chemical Bonding at Surfaces and Interfaces

Chemical bonding at surfaces and interfaces

Chemical Bonding at Surfaces and Interfaces

Reactions at Solid Surfaces (Baker Lecture Series)

Physics and Chemistry at Oxide Surfaces

Physics and chemistry at oxide surfaces

Equilibria and Dynamics of Gas Adsorption on Heterogeneous Solid Surfaces (Studies in Surface Science and Catalysis)

Surfaces Algebriques

Rough Surfaces

Rough Surfaces

Spillover and Mobility of Species on Solid Surfaces ,Volume 138 (Studies in Surface Science and Catalysis)

Equilibria and Dynamics of Gas Adsorption on Heterogeneous Solid Surfaces (Studies in Surface Science and Catalysis)

Riemann surfaces

Recommend Documents