Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis

Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis Michael Bowker and Philip R. Davies, Editors

WILEY-VCH Verlag GmbH

Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis

Edited by Michael Bowker and Philip R. Davies

Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis Edited by Michael Bowker and Philip R. Davies

Further Reading K.W. Kolasinski

B.P. Jena, J.K.H. Hoerber (Eds.)

Surface Science

Force Microscopy

Foundations of Catalysis and Nanoscience

Applications in Biology and Medicine

2008

ISBN: 978-0-471-39628-4

2006

ISBN: 978-0-470-03304-3

M. Prutton, M. El Gomati (Eds.) D.G. Brandon, W.D. Kaplan

Microstructural Characterization of Materials 2008

Scanning Auger Electron Microscopy 2006 ISBN: 978-0-470-86677-1

ISBN: 978-0-470-02784-4

D.K. Schroder J.W. Niemantsverdriet

Spectroscopy in Catalysis An Introduction 2007 ISBN: 978-3-527-31651-9

Semiconductor Material and Device Characterization 2006 ISBN: 978-0-471-73906-7

Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis

Edited by Michael Bowker and Philip R. Davies

The Editors Prof. Michael Bowker Cardiff University Wolfson Nanoscience Lab and Cardiff Catalysis Institute School of Chemistry Cardiff, CF10 3AT United Kingdom Dr. Philip R. Davies Cardiff University Wolfson Nanoscience Lab and Cardiff Catalysis Institute School of Chemistry Cardiff, CF10 3AT United Kingdom

Cover illustration The STM images being past of the front cover picture have been kindly provided by the group of D. Wayne Goodman (authos of Chapter 3).

All books published by Wiley-VCH are carefully produced. Nevertheless, authors, editors, and publisher do not warrant the information contained in these books, including this book, to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate. Library of Congress Card No.: applied for British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.d-nb.de # 2010 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Printed in the Federal Republic of Germany Printed on acid-free paper Cover Design Formgeber, Eppelheim Typesetting Thomson Digital, Noida, India Printing Strauss GmbH, Mörlenbach Bookbinding Litges & Dopf Buchbinderei GmbH, Heppenheim ISBN: 978-3-527-31982-4

V

Contents Preface IX List of Contributors

1 1.1 1.1.1 1.1.2 1.2 1.2.1 1.2.2 1.2.3 1.2.4 1.2.5 1.3 1.3.1 1.3.2 1.3.3 1.3.4 1.4

2 2.1 2.2 2.3 2.4 2.5 2.5.1

XIII

Chirality at Metal Surfaces 1 Chris J. Baddeley and Neville V. Richardson Introduction 1 Definition of Chirality 1 Nomenclature of Chirality: The (R),(S) Convention 2 Surface Chirality Following Molecular Adsorption 4 Achiral Molecules on Achiral Surfaces 4 Lattice Matching 8 Chiral Molecules on Achiral Surfaces 12 Chiral Molecules on Chiral Surfaces 15 Chiral Etching 16 Chiral Amplification and Recognition 19 Chiral Amplification in Two Dimensions 19 Chiral Switching 20 Chiral Recognition 21 Prochiral Molecules Interacting with Chiral Surfaces 24 Conclusions 25 References 26 The Template Route to Nanostructured Model Catalysts Conrad Becker and Klaus Wandelt Introduction 29 Surfaces as Two-Dimensional Templates 31 STM Imaging of Oxide Films 34 STM Imaging of Metal Particles on Oxide Films 39 Template-Controlled Growth of Model Catalysts 44 Oxides as Templates 44

29

VI

Contents

2.5.2 2.6

Modified Templates Conclusions 51 References 52

3

In Situ STM Studies of Model Catalysts 55 Fan Yang and D. Wayne Goodman Introduction 55 Instrumentation 56 Visualizing the Pathway of Catalytic Reactions 59 Imaging of Adsorbates and Reaction Intermediates 59 Imaging Chemisorption on Metals 61 Determining the Sites for Chemisorption on Oxide Surfaces 64 Visualizing Reaction Intermediates and the Mechanism of Hydrogen Oxidation 71 Measuring the Reaction Kinetics of CO Oxidation 73 Metal Surfaces at High Pressures 81 In Situ Studies of Supported Model Catalysts 85 Monitoring the Growth Kinetics of Supported Metal Catalysts 85 Studies of the SMSI Effect 88 Sintering Kinetics of Supported Au Clusters 89 Outlook 91 References 92

3.1 3.2 3.3 3.3.1 3.3.2 3.3.3 3.3.4 3.3.5 3.4 3.5 3.5.1 3.5.2 3.5.3 3.6

4

4.1 4.2 4.3 4.4 4.5 4.5.1 4.6

5

5.1 5.2 5.2.1 5.2.2 5.2.2.1 5.2.2.2 5.2.2.3

50

Theory of Scanning Tunneling Microscopy and Applications in Catalysis 97 Gilberto Teobaldi, Haiping Lin, and Werner Hofer Catalysis and Scanning Tunneling Microscopy 97 Image Formation in an STM 98 Simulating Tunneling Currents 99 Simulating Chemical Reactivity 100 Catalytic Water Production 101 TiO2: A Catalytic Model System 106 Outlook 115 References 116 Characterization and Modification of Electrode Surfaces by In Situ STM 119 Dieter M. Kolb and Felice C. Simeone Introduction 119 In Situ STM: Principle, Technical Realization and Limitations Principle Considerations for In Situ Operation 120 Technical Realization 124 Tip Preparation and Isolation 124 Electrochemical Cell 126 Vibration Damping 127

120

Contents

5.2.3 5.3 5.3.1 5.3.2 5.4 5.4.1 5.4.2 5.4.3

6 6.1 6.2 6.3 6.3.1 6.3.2 6.3.3 6.3.4 6.3.5 6.3.6 6.4

7

7.1 7.2 7.3 7.4 7.5 7.5.1 7.5.2 7.5.3 7.6

Limitations 127 Imaging Single-Crystal Surfaces of Catalytically Relevant Systems 128 Preparation and Imaging of Metal Single-Crystal Surfaces 128 Bimetallic Surfaces 130 Strategies for Nanostructuring Surfaces 132 Oxidation–Reduction Cycles for Roughening and Faceting Surfaces 132 Surface Modification by an STM: An Overview 134 Metal Nanocluster Deposition via Jump-to-Contact 139 References 144 STM Imaging of Oxide Nanolayer Model Systems 147 Falko P. Netzer and Svetlozar Surnev Introduction 147 Experimental Aspects and Technical Developments 149 Case Studies: Selected Oxide–Metal Systems 152 Alumina Nanolayers on NiAl Alloys 152 Titanium Oxide Nanolayers 155 Vanadium Oxide Nanolayers 159 Iron Oxides on Pt(1 1 1) 169 Nickel Oxide Nanolayers 173 Ceria Nanolayers on Metal Surfaces 177 Synopsis and Outlook 182 References 183 Surface Mobility of Atoms and Molecules Studied with High-Pressure Scanning Tunneling Microscopy 189 Gabor A. Somorjai, Feng Tao, and Derek Butcher Introduction 189 Characterization of Surface Mobility of Molecules and Atoms High-Pressure STM Technique and Instrumentation 191 Mobility and Flexibility of Catalyst Surfaces at High-Pressure High-Temperature Reaction Conditions 197 Adsorbate Mobility During Catalytic Reactions 206 Ethylene Hydrogenation on Pt(1 1 1) 207 Hydrogenation of C6 Cyclic Hydrocarbons on Pt(1 1 1) 209 CO/NO Coadsorption on Rh(1 1 1) 213 Summary 216 References 216

189

VII

VIII

Contents

8

8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8

Point Defects on Rutile TiO2(1 1 0): Reactivity, Dynamics, and Tunability 219 Chi L. Pang and Geoff Thornton Introduction 219 Methods 220 Water Dissociation at Oxygen Vacancies and the Identification of Point Defects 221 O2 Dissociation at Oxygen Vacancies 229 Alcohol Dissociation at Oxygen Vacancies 229 Diffusion of Oxygen Vacancies and Surface Hydroxy 232 Tuning the Densities of Oxygen Vacancies and Surface Hydroxyl on TiO2(1 1 0) 234 Outlook 236 References 236 Index

239

IX

Preface The objective of this book is to highlight the important strides being made toward a molecular understanding of the processes that occur at surfaces through the unique information provided by the proximal scanning probe family of techniques: this principally involves scanning tunneling microscopy (STM) but some atomic force microscopy (AFM) experiments are also included. The chapters in this book describe several state-of-the-art examples where an atomic understanding of surface processes is developing out of atomically resolved information provided by STM and AFM. The focus of much of the work is on understanding the fundamentals of catalysis, a reflection of the huge significance of heterogeneous catalysis in society today, but the discoveries being made in this field will undoubtedly have a much wider significance in the field of nanoscience/ technology. Reaction equations are derived from the results of global reaction measurements and considerations of stoichiometry. Until recently, the intermediates (and in particular the surface species) involved in the mechanism and their spatial location have remained largely theoretical. The advent of STM has, for the first time, allowed us the possibility of getting direct insight into this area. An example is the molecular identification of the sequence of reaction steps and of the species involved in the reaction of gas-phase methanol with oxygen on a copper surface. This produces methoxy groups as the first step, in which the slightly acidic hydrogen from the alcohol is stripped by surface oxygen leading to water desorption, as shown below, where the subscript ‘‘a’’ refers to an adsorbed species. CH3OHg þ Oa ! CH3Oa þ OHa CH3OHg þ OHa ! CH3Oa þ H2Og In this case, as in many others, STM has enabled us to identify the active sites at which reaction between methanol and adsorbed oxygen takes place, and has also allowed us to verify, at the atomic and the molecular scale, that the reaction does indeed occur in the way the above stoichiometric equations describe. This is invaluable information in the quest to understand and improve catalytic processes.

X

Preface

The principal advantage of the probe methods is their extraordinary spatial resolution (<0.1 nm) that does not, by necessity, require large areas of order. Their potential is further expanded by the ability to study surfaces under conditions that range from ultrahigh vacuum at cryogenic temperatures to industrially relevant pressures and temperatures to aggressive liquid phases. A major aim is to identify surface structures and intermediates present during operation in situ and in operando. There are three principal requirements to achieve these objectives: (i) highpressure operation (already demonstrated by a number of individual groups in the field); (ii) the ability to image at high temperature while at high pressure (to some degree, this has been achieved, although there are limitations to the temperature range that can be studied, especially at above ambient pressure); (iii) fast scanning (of the order of 1–10 images per second) since catalytic turnover rates are often very high (e.g., 10 s1). This is also important for the ultimate aim of determining the statistics of the reaction (e.g., determining directly, in situ, the turnover number for the reaction). The authors believe that these aims will be realized in the next 10 years or so, but we are not quite there yet. There is plenty of room in the field for further developments. It is fitting that this book should include a contribution from the laboratory of Somorjai, one of the principal exponents of the surface science approach to catalysis, who has developed a number of new approaches to the field and who, in recent years, has looked, in particular, at adsorption and surface reactions at high gas pressures. He reviews this work with some considerable and helpful focus on equipment developments. Similarly, Goodman has been a strong advocate of in situ highpressure studies and has been at the forefront of developments in the field for some considerable time. He reviews the state of the art and generously acknowledges the contribution of colleagues in the field, some of whom have contributed to this volume. Kolb and Simeone discuss the application of STM to samples in the liquid environment and deposition of metal from the STM tip itself, via reaction and neutralization of cations from solution, generating reproducible patterns at nanometer scales. The ability to control nanoparticle formation in terms of constant spacing, spatial arrangement, and monosized dispersion is a crucial one for future nanotechnology developments and Becker and Wandelt have also concerned themselves with the synthesis of reproducible patterns of particles at the nanoscale. They illustrate the potential of a bottom-up process, by gas-phase metal deposition, for generating model catalysts that can be studied both by spatially averaging techniques and by scanning tunneling microscopy. Netzer and Surnev consider the problem of generating model catalyst systems for study using STM from a different direction. They examine the growth of thin oxide films on metal substrates, the ‘‘inverse’’ catalyst approach, and show the intrinsic beauty in the geometrical arrangements of thin-layer oxides that can be obtained when high-quality imaging is pursued. They also report the formation of new types of oxide structure in the 2D regime. Importantly, they give consideration to the particular problems of interpretation posed by STM images of thin oxide layers. Since STM images involve a convolution of the substrate, adsorbate, and STM tip

Preface

electronic states, interpretation of STM images is critical to all of the work presented in this book and the contribution by Hofer, Lin, and Teobaldi is particularly welcome since it provides a detailed review of the advances made in the theory of STM imaging, highlighting areas where theory now has a good grasp of the issues and where further development is needed. The ability of STM to image at the atomic scale is particularly exemplified by the two other chapters in the book. Thornton and Pang discuss the identification of point defects at TiO2 surfaces, a material that has played an important role in model catalyst studies to date. Point defects have been suggested to be responsible for much of the activity at oxide surfaces and the ability to identify these features and track their reactions with such species as oxygen and water represents a major advance in our ability to explore surface reactions. Meanwhile, Baddeley and Richardson concentrate on the effects of chirality at surfaces, and on the important field of surface chirality and its effects on adsorption, in a chapter that touches on one of the fundamental questions in the whole of science – the origins of life itself! In recent years, studies using STM have expanded from the use of the atomically flat metal single crystals that have been the mainstay of surface science since the mid-1960s to complex oxide surfaces and to 3D nanoparticles, often grown on representative catalyst supports. The improvement in the technology and interpretation of imaging, and the increasing complexity of the surfaces being studied, ensures a pivotal role in the future for surface science, particularly in the context of the increasing practical importance of nanoscience in technological development. Cardiff University, December 2009

Michael Bowker and Philip R. Davies

XI

XIII

List of Contributors Chris J. Baddeley University of St Andrews EaStCHEM School of Chemistry St Andrews Fife, KY16 9ST UK

D. Wayne Goodman Texas A&M University Department of Chemistry P.O. Box 30012 College Station, TX 77843-3012 USA

Conrad Becker Université de la Méditerranée CINaM – CNRS – UPR3118 Campus de Luminy Case 913 13288 Marseille Cedex 9 France

Werner Hofer The University of Liverpool Surface Science Research Centre Liverpool, L69 3BX UK

Derek Butcher Lawrence Berkeley National Laboratory Materials Science and Chemistry Divisions Berkeley, CA 94720 USA and University of California Department of Chemistry Berkeley, CA 94720 USA

Dieter M. Kolb University of Ulm Institute of Electrochemistry 89069 Ulm Germany Haiping Lin The University of Liverpool Surface Science Research Centre Liverpool, L69 3BX UK Falko P. Netzer Karl Franzens University Graz Institute of Physics, Surface and Interface Physics 8010 Graz Austria

XIV

List of Contributors

Chi L. Pang University College London London Centre for Nanotechnology and Department of Chemistry London, WC1H 0AJ UK Neville V. Richardson University of St Andrews EaStCHEM School of Chemistry St Andrews Fife, KY16 9ST UK Felice C. Simeone University of Ulm Institute of Electrochemistry 89069 Ulm Germany Gabor A. Somorjai Lawrence Berkeley National Laboratory Materials Science and Chemistry Divisions Berkeley, CA 94720 USA and University of California Department of Chemistry Berkeley, CA 94720 USA Svetlozar Surnev Karl Franzens University Graz Institute of Physics, Surface and Interface Physics 8010 Graz Austria

Feng Tao Lawrence Berkeley National Laboratory Materials Science and Chemistry Divisions Berkeley, CA 94720 USA and University of California Department of Chemistry Berkeley, CA 94720 USA Gilberto Teobaldi The University of Liverpool Surface Science Research Centre Liverpool, L69 3BX UK Geoff Thornton University College London London Centre for Nanotechnology and Department of Chemistry London, WC1H 0AJ UK Klaus Wandelt Universität Bonn Institut für Physikalische und Theoretische Chemie Wegelerstrasse 12 53115 Bonn Germany Fan Yang Texas A&M University Department of Chemistry P.O. Box 30012 College Station, TX 77843-3012 USA

j1

1 Chirality at Metal Surfaces Chris J. Baddeley and Neville V. Richardson

1.1 Introduction

Since the mid-1990s, the number of surface science investigations of chirality at surfaces has increased exponentially. Advances in the technique of scanning tunneling microscopy (STM) have been crucial in enabling the visualization of single chiral molecules, clusters, and extended arrays. As such, STM has facilitated dramatic advances in the fundamental understanding of the interactions of chiral molecules with surfaces and the phenomena of chiral amplification and chiral recognition. These issues are of considerable technological importance, for example, in the development of heterogeneous catalysts for the production of chiral pharmaceuticals and in the design of biosensors. In addition, the understanding of chirality at surfaces may be a key to unraveling the complexities of the origin of life. 1.1.1 Definition of Chirality

The word chirality is derived from the Greek kheir meaning “hand.” It is the geometric property of an object that distinguishes a right hand from a left hand. Lord Kelvin provided a definition of chirality in his 1884 Baltimore Lectures, “I call any geometrical figure or group of points ‘chiral’ and say it has ‘chirality’, if its image in a plane mirror, ideally realized, cannot be brought into coincidence with itself.” For an isolated object, for example, a molecule, the above statement can be interpreted as being equivalent to requiring that the object possesses neither a mirror plane of symmetry nor a point of symmetry (center of inversion). If a molecule possesses either one of these symmetry elements, it can be superimposed on its mirror image and is therefore achiral. A chiral molecule and its mirror image are referred to as being a pair of enantiomers. Many organic molecules possess the property of chirality. Chiral centers are most commonly associated with the tetrahedral coordination of four different substituents. However, there are many examples of other rigid

j 1 Chirality at Metal Surfaces

2

structures that have chiral properties where a significant barrier exists to conformational change within the molecule. 1.1.2 Nomenclature of Chirality: The (R),(S) Convention

Most of the physical properties (e.g., boiling and melting point, density, refractive index, etc.) of two enantiomers are identical. Importantly, however, the two enantiomers interact differently with polarized light. When plane polarized light interacts with a sample of chiral molecules, there is a measurable net rotation of the plane of polarization. Such molecules are said to be optically active. If the chiral compound causes the plane of polarization to rotate in a clockwise (positive) direction as viewed by an observer facing the beam, the compound is said to be dextrorotatory. An anticlockwise (negative) rotation is caused by a levorotatory compound. Dextrorotatory chiral compounds are often given the label D or ( þ ) while levorotatory compounds are denoted by L or (). In this chapter, we will use an alternative convention that labels chiral molecules according to their absolute stereochemistry. The (R),(S) convention or Cahn– Ingold–Prelog system was first introduced by Robert S. Cahn and Sir Christopher K. Ingold (University College, London) in 1951 and later modified by Vlado Prelog (Swiss Federal Institute of Technology) [1]. Essentially, the four atomic substituents at a stereocenter are identified and assigned a priority (1 (highest), 2, 3, 4 (lowest)) by atomic mass. If two atomic substituents are the same, their priority is defined by working outward along the chain of atoms until a point of difference is reached. Using the same considerations of atomic mass, the priority is then assigned at the first point of difference. For example, a CH2CH3 substituent has a higher priority than a CH3 substituent. Once the priority has been assigned around the stereocenter, the tetrahedral arrangement is viewed along the bond between the central atom and the lowest priority (4) substituent (often a CH bond) from the opposite side to the substituent (Figure 1.1). If the three other substituents are arranged such that the path from 1 to 2 to 3 involves a clockwise rotation, the stereocenter is labeled (R) (Latin rectus for right). By contrast, if the path involves an anticlockwise rotation, the stereocenter is labeled (S) (Latin sinister for left). It is important to note that the absolute stereochemistry cannot be predicted from the L or D labels and vice versa. In nature, a remarkable, and so far unexplained, fact is that the amino acid building blocks of all proteins are exclusively left-handed and that the sugars contained within the double helix structure of DNA are exclusively right-handed. The consequences of the chirality of living organisms are far reaching. The human sense of smell, for example, is able to distinguish between pure (R)-limonene (smelling of oranges) and (S)-limonene (smelling of lemons). More significantly, two enantiomeric forms of an organic molecule can have different physiological effects on human body. In many cases, one enantiomer is the active component while the opposite enantiomer has no effect (e.g., ibuprofen where the (S)-enantiomer is active). However, often the two enantiomers have dramatically different effects. For example, (S)-methamphetamine

1.1 Introduction

1

1 4

4

2

3

3

2

(R)

(S) 1

2

1

3

3

2

Figure 1.1 Schematic diagram explaining the Cahn–Ingold–Prelog convention for determining the absolute stereochemistry of a chiral molecule.

is a psychostimulant while (R)-methamphetamine is the active ingredient in many nasal decongestants (Figure 1.2). In the pharmaceutical industry, about half of all of the new drugs being tested require the production of exclusively one enantiomeric product. Thermodynamically, this is a challenging problem since the two isolated enantiomers have identical Gibbs energies; the reaction from prochiral reagent to product should therefore result in a 50 : 50 (racemic) mixture at equilibrium. To skew the reaction pathway to form one product with close to 100% enantioselectivity is nontrivial. Knowles [2], Noyori [3], and Sharpless [4] were awarded the Nobel Prize in Chemistry in 2001 for developing enantioselective homogeneous catalysts capable of producing chiral molecules on an industrial scale. Typically, these catalysts consist of organometallic complexes with chiral ligands. Access to the metal center by the reagent is strongly sterically influenced by the chiral ligands resulting in preferential formation of one enantiomeric product. There are many potential advantages of using heterogeneous catalysts, not least the ease of separation of the catalyst from the products. However, despite extensive research over several decades, relatively few successful catalysts have been synthesized on a laboratory scale and the impact on industrial catalysis is essentially negligible. One of the primary motivations behind surface science studies of chirality at surfaces is to understand the surface chemistry underpinning chiral catalysis and to develop methodologies for the rational design of chiral catalysts. Similarly, those interested in issues related to the origin of life are investigating the possibility that surfaces were responsible for the initial seeding of the chiral building blocks of life and that, presumably via some chiral amplification effects, this led to the overwhelming dominance of left-handed amino acids and right-handed sugars in

j3

j 1 Chirality at Metal Surfaces

4

Figure 1.2 The two mirror equivalent forms of the drug methamphetamine. On the right is shown the (S)-form of the molecule; on the left is the (R)-enantiomer.

biological systems on Earth. As such, the surface chemistry of chiral solids, chiral amplification, and chiral recognition are all important subtopics of chiral surface science. STM has proved to be the single most important tool of researchers in this field.

1.2 Surface Chirality Following Molecular Adsorption 1.2.1 Achiral Molecules on Achiral Surfaces

When a molecule is adsorbed on a surface, the symmetry of the combined adsorbate–substrate system is very likely to be reduced compared to that of the isolated gas-phase species or the bare adsorption site. This raises the possibility that, if mirror planes present in the isolated achiral molecule and those at the relevant

1.2 Surface Chirality Following Molecular Adsorption

adsorption site of the clean surface are not coincident, then the combined system of a single adsorbed molecule and the substrate will be locally chiral; that is, mirror planes of the isolated molecule are lost on adsorption and chirality is induced by the adsorption process. Note that a center of symmetry, also capable of ensuring superimposability of an object and its mirror image, is necessarily incompatible with the presence of a nearby surface [5]. A commonly observed case of such adsorption-induced chirality is that of a planar molecule with Cs symmetry (a single mirror plane) in the gas phase adsorbing on a surface such that the molecular plane is parallel to the substrate, as favored, for example, by van der Waals (vdW) interactions, thereby destroying the mirror plane symmetry. The molecule can then exist in two enantiomeric forms, although necessarily as a racemic mixture in the absence of any other influences that might lead to a preference of one rather than the other. Figure 1.3 illustrates this possibility for 4-[trans-2-(pyrid-4-yl-vinyl)]benzoic acid (PVBA) adsorbed parallel to an idealized, unstructured surface [6] Interconversion of the two enantiomers is possible only if the molecule is removed from the surface and rotated by 180 around an axis parallel to the substrate surface. In the case of PVBA adsorbed on Ag{1 1 1}, hydrogen bonding leads to a preference for homochiral double chains based on head-to-tail NHO bonds and a C2 axis relating the two strands of the chains. The chirality of the chain can be recognized in the STM images by the stagger of one strand relative to the other that arises from CHO bonds, as shown in Figure 1.3 [6]. The example described above is that of the separation of enantiomers into 1D chains following adsorption-induced chirality. In addition to forming chirally seg-

Figure 1.3 Molecular models showing the two enantiomers resulting from the loss of mirror plane symmetry on adsorption with the molecular plane parallel to the substrate. The separation of enantiomers observed by STM is identified by the relative displacement of adjacent monomers within the double chain. (Reprinted with permission from Ref. [6]. Copyright 2001, American Physical Society.)

j5

j 1 Chirality at Metal Surfaces

6

Figure 1.4 Low- and high-resolution STM images of a decamer of 1-nitronaphthalene, together with the minimum structure optimized from a force model, showing individual enantiomers in a 6 : 4 ratio. (Reprinted with permission from Ref. [7]. Copyright 1999, American Physical Society.)

regated chains, 1-nitronaphthalene is able to form chiral decamers [7]. Figure 1.4 shows a cluster of ten 1-nitronapthalene molecules [8]. The adsorption process on Au {1 1 1} imposes chirality on the molecule and the clusters can be seen to have a pinwheel, chiral conformation, although within the cluster not all the individual molecules have the same handedness. Each cluster contains six molecules of one enantiomer and four of the other. The overall surface is expected to be racemic as regard to both molecules and clusters. A particularly elegant example of cluster formation involving chiral recognition and retention of chirality through an increasingly complex hierarchical series of clusters is that of rubrene on Au{1 1 1} [9] illustrated in Figure 1.5 The above discussion refers to the loss of mirror symmetry on adsorption leading to chirality at the level of the individual molecule. It is also common for oblique lattices to be formed following molecular adsorption, hence global chirality, even

Figure 1.5 (a) Hierarchy of clusters of rubrene on Au{1 1 1}, showing the evolution from trimers to pentamers of trimers and eventually 150 molecules per cluster as a decamer of the pentamers. (b) Illustration of the preservation of chirality through the hierarchy. (Adapted with permission from Ref. [9]).

1.2 Surface Chirality Following Molecular Adsorption

when the local site retains one or more mirror planes. A specific example of the relationship between local and organizational chirality for a highly symmetric molecule is discussed in some detail in Section 1.2.2. It is relevant at this point to note that chemistry frequently employs a rather weaker, arguably less precise, definition of chirality than the more “mathematical” definition put forward by Lord Kelvin. A species, which in its most stable conformation has no mirror plane or center of symmetry, is formally chiral but, if there were a low-energy pathway to the enantiomer, for example, by a low-frequency vibrational mode, then, in chemical terminology, this would not normally be considered to be chiral. However, if adsorption of such a species raises the frequency of the vibration substantially, then the energy barrier between the two “enantiomers” may become chemically significant such that the adsorbed molecule is meaningfully described as chiral. An early example of this is the case of the deprotonated glycine species adsorbed on copper surfaces. An isolated glycinate anion, although lacking any mirror plane or center of symmetry, is nevertheless readily converted to its enantiomer principally by a rotation around the CN bond, with an energy barrier of approximately 35 kJ mol1, which might readily be overcome at room temperature, such that glycine or glycinate are not generally considered chiral. However, on Cu{1 1 0}, for example, adsorption takes place through both O atoms and the N atom in a tridentate interaction with the copper surface, each atom in an approximately atop site [10, 11]. This inhibits the interconversion of enantiomers, and surfaceinduced chirality leads to distinct mirror image species on the surface [12]. Nevertheless, unlike the examples discussed above, segregation of enantiomers into clusters, chains, or arrays does not occur. Instead, one molecule of each enantiomer gives rise to a heterochiral (3  2) unit cell and is interrelated by glide lines as shown in Figure 1.6. This proposal based on LEED, STM, and IR data [10] has been confirmed by photoelectron diffraction [11] and by DFT calculations [13]. A suggestion that a second phase consists of homochiral unit cells [12] has not been confirmed by photoelectron diffraction [11, 14] or theory, although the energy difference of this

Figure 1.6 The left-hand panel shows a molecular model of the glycinate/Cu{1 1 0} structure with both enantiomers present in the heterochiral (3  2) unit cell, superimposed on an STM image of this surface. (Adapted with permission from Ref. [12]. Copyright 2002,

Elsevier.) The right-hand panel shows the confirmation of this structure calculated by DFT, clearly indicating the atop adsorption sites occupied by the N and both O atoms in this system. (Reprinted with permission from Ref. [13]. Copyright 2004, Elsevier.)

j7

j 1 Chirality at Metal Surfaces

8

phase is calculated to be small (6 kJ mol1) [13]. It is likely that the different “phases” imaged by STM [12] result from the two rotational domains of the heterochiral structure appearing distinct because of anisotropy in the tip. Interestingly, intrinsically chiral amino acids such as alanine or phenylglycine can adsorb on the Cu{1 1 0} surface also in a (3  2) structure with an apparent glide line indicated by the LEED pattern, although only a single enantiomer is present [15–17]. 1.2.2 Lattice Matching

It would seem inherently unlikely that a highly symmetric (D6h) molecule such as coronene, C24H12, could give rise to chiral surfaces and indeed diastereoisomeric interactions, particularly when adsorbed on a hexagonal substrate such as graphite or an fcc{1 1 1} face. Nevertheless, we show in this section that lattice matching between an adsorbate overlayer and the substrate can readily give rise to surface chirality, and while this might be unsurprising in systems of lower symmetry, it is still distinctly likely in adsorbate/substrate systems in which both components have inherently high symmetry. To emphasize this aspect, we choose coronene and a related derivative to illustrate how these effects arise from simple interadsorbate interactions and their simple geometric consequences. To simplify matters further, we restrict the adsorption of coronene and its analogues to atop adsorption sites, where symmetry matching with a hexagonal fcc substrate (also locally D6h) would seem to be optimized. Nondissociative adsorption of coronene on a late transition or coinage metal is likely to be dominated by van der Waals interactions and rather weak p–d interactions, both of which favor a flat-lying and probably atop adsorption geometry again optimizing the symmetry matching. Although relatively weak, these interactions are strong enough to permit stable monolayers to be formed in UHV at room temperature. Interactions between adsorbed coronene molecules are highly isotropic and again dominated by vdW terms. These, therefore, favor hexagonal close packing in an isolated (no substrate) monolayer of planar coronene molecules. Nevertheless, despite all the apparent symmetry matching, it is the subtle energy balance between interadsorbate interactions and those favoring a specific adsorption site, even an atop one, that gives rise to chiral structures and diastereoisomeric effects.  Coronene (Figure 1.7a), considered as a circle, has a vdW diameter of 11.6 A, which  corresponds to a molecular area of 105.7 A2 and leads, with hexagonal but non-space  filling close packing, to a unit cell area of 116.5 A2. However, coronene on some  hexagonal surfaces has an intermolecular separation somewhat less than 11.6 A, for   example, 11.27 A [18] on graphite and 11.18 A on MoS2 [19], suggesting that it is better  considered as having a hexagonal, space filling shape with a vdW width of 11.26 A. Even this, however, is insufficient to rationalize the intermolecular separations found on other, admittedly nonhexagonal, surfaces such as Cu{1 0 0} [19] and Cu{1 1 0} [20].  In such systems, intermolecular spacings significantly less than 11 A can be found. The explanation, while retaining a flat-lying coronene molecule, since there is no evidence to the contrary, lies in recognizing that the 12 H atoms are almost equally

1.2 Surface Chirality Following Molecular Adsorption

Figure 1.7 (a) Molecular model of coronene; (b) hexagonal close packing at the van der Waals diameter; parts (c) and (d) illustrate the packing advantage, which can be obtained by a concerted rotation, counterclockwise or clockwise, of all molecules to allow

interdigitation of the CH bonds on adjacent molecules; part (e) illustrates the (4  4) model of coronene on Au{1 1 1} where the adsorption site dominated separation of molecules is such that interdigitation is unnecessary.

spaced around the periphery of the molecule and confer C12 rotational symmetry on the molecule. A concerted rotation of all molecules on the hexagonal lattice by 8.4 about their centers then allows interdigitation of the H atoms on neighboring molecules (Figure 1.7c and d). This permits a 3% reduction of the intermolecular  spacing to around 10.9 A. Herein lies one element of the surface chirality of this molecule. When the molecules on an isolated hexagonal lattice are rotated in concert away from their initial positions to allow interdigitation and closer packing, the 2D site symmetry is reduced, all mirror planes are lost, and the molecule becomes chiral through the lack of mirror symmetry in the interactions with its neighbors. Rotation to the left or right gives energetically equivalent enantiomers. There is also a second source of chirality when the adsorbate hexagonal lattice is matched with that of the substrate. For a hexagonal substrate, characterized by unit cell vectors a1 and a2 aligned along close-packed directions, at 120 to each other and of length a, there are larger hexagonal unit cells defined by unit cell vectors b1 and b2, where b1 ¼ ma1 þ na2 and b2 ¼ na1 þ (m  n)a2 with m and n integers. These have lengths b ¼ aH(m2  mn þ n2) and are rotated q ¼ tan1(H3n/(2m  n)) relative to the substrate unit cell vectors. Many of the smaller ones, based on m and n values up to 6, are familiar overlayers for atomic and molecular adsorbates on fcc{1 1 1} substrates, for example, (H3  H3)R30 , (H7  H7)R19.1 , and so on. For those overlayers where m ¼ 0, n, or 2n, corresponding to rotations of 0 , 60 , and 30 ,

j9

j 1 Chirality at Metal Surfaces

10

respectively, the structure is achiral since a mirror plane is retained along either the h1 1 0i or the h2 1 1i direction. Conversely, if this condition is not met, there is no coincidence of mirror planes between the substrate and overlayer lattices: enantiomeric structures will exist, for example, based on m, n being 3, 1 or 3, 2, that is, (H7  H7)R19.1 and (H7  H7)R40.9 , respectively, or perhaps more helpfully described as (H7  H7)R  19.1 . Lattice matching of this type giving rise to chiral lattices is common in overlayers on hexagonal substrates, such as the pinwheel structure found for Pd on (1  2) reconstructed TiO2{1 1 0} surfaces [21]. In the case of coronene adsorbed on either Ag{1 1 1} [22] or Au{1 1 1} [23, 24], an achiral (4  4) structure is observed (Figure 1.7e). This is perhaps unsurprising since  this is the hexagonal superlattice that, with a lattice vector of approximately 11.5 A, is the closest match to the coronene dimensions. Although this lattice is achiral, it demonstrates that the balance between interadsorbate interactions and those favoring a specific adsorption site and hence a commensurate overlayer is important. In contrast, for adsorption of coronene on Cu{1 1 1}, a chiral lattice is predicted based on either (H19  H19)R  23.4 or (H21  H21)R  10.9 lattices. The latter with a unit  cell vector of length 11.7 A might be favored if site preference is strong relative to intermolecular close packing but would not require the concerted rotation of coronene molecules to reduce the intermolecular separation since this is below  even the circular diameter of coronene (11.6 A). Chirality would be limited to that derived solely from the lattice matching and molecules would be free to adopt whatever rotation optimized the energy based on an atop local site geometry. Of course, a twist away from a high-symmetry azimuthal orientation, which might be clockwise or anticlockwise, introduces a second chiral element and hence the need to consider diastereoisomerism. There are four possible choices of lattice/molecular twist that might conveniently be designated þ / þ , / for one pair of enantiomers and þ /, / þ for the other pair. In principle, a particular sense of rotation could favor a particular lattice orientation such that one pair is energetically more favorable than the other. However, since intermolecular interactions are likely to be weak at this separation, the energy difference is likely to be small. This contrasts with the situation if the former lattice, (H19  H19)R  23.4 , were preferred because of the importance of intermolecular interactions. In this case, since the substrate imposed  lattice dimension is only 11.14 A, molecular rotation imposed within the 2D adsorbate lattice is required, with CH interdigitation to achieve this reduced separation as shown in Figure 1.8. The second element of chirality is again a molecular rotation but one that has its origin in the intermolecular interactions rather than molecule–substrate site interactions. The energy preference between the two diastereoisomer pairs is now dictated by which pair leads to the more favorable orientation of the molecule on the atop adsorption site. Notable, perhaps, is that for one diastereoisomer pair the azimuthal orientation of the molecule with respect to the substrate is such that a local high symmetry is recovered because the lattice rotation of 23 , combined with the optimum interdigitation rotation of 8 , realigns the mirror planes of the molecule very closely (<2 ) with those of the substrate. To our knowledge, coronene adsorption on Cu{1 1 1} has not been studied, but clearly this system would provide an interesting model for investigating the subtle energy

1.2 Surface Chirality Following Molecular Adsorption

Figure 1.8 Diastereoisomeric effects predicted to arise for coronene adsorption on Cu{1 1 1} from a combination of molecular rotation (curved arrows) within the 2D adsorbate lattice to allow CH bond interdigitation and the (H19  H19)R  23.4 lattice. The black arrows indicate a high symmetry within the molecule bisecting CH bonds.

Figure 1.9 UHV-STM image of coronene adsorbed on Cu{1 1 0} showing the two mirror domains A and B. On the right, the LEED pattern showing the contribution of the mirror domains. (Adapted with permission from Ref. [20]. Copyright 2007, IOP Publishing Ltd and Deutsche Physikalische Gesellschaft.)

balance between coronene/coronene interactions and those determining coronene orientation on an atop site. On fcc {1 1 0} and {1 0 0} surfaces, pseudohexagonal lattices can be found [25]. Coronene adsorption on Cu{1 1 0} leads to a h3 2 | 1 3i structure and its enantiomer h3 2 | 13i, shown in Figure 1.9, corresponding to a pseudohexagonal lattice with

j11

j 1 Chirality at Metal Surfaces

12

nearest-neighbor distances of H17a, H18a, and H19a. Interestingly, it is predicted that the distortion to a pseudohexagonal lattice creates a second element of chirality in the isolated adsorbate monolayer along with the rotation of the molecule that allows interdigitation. Two diastereoisomer pairs therefore exist within this monolayer even in the absence of the substrate due to the coupling of the two possible directions of rotation and the sense of the sequence of distortions around the hexagon, although of course it is not strictly independent of the substrate since it is the mapping onto the substrate that determines this sequence and the mirror lattice has the mirror sequence of distortion. On Cu{1 0 0}, the structure of adsorbed coronene was originally suggested to be rotational, achiral domains of p(4  7), but that would  require an extremely short intermolecular separation of 10.2 A and a very small unit  cell of only 91.4 A [2, 19]. A more likely interpretation of the LEED is that the structure corresponds to the pseudohexagonal, but still achiral, lattice with nearest neighbors at H17a, H17a, and H18a [25]. The most well-characterized example of the interplay between these various chiral elements, which can arise in molecular adsorption in a constrained pseudohexagonal lattice, is that of 2,5,8,11,14,17-hexa-tert-butylhexabenzo[bc,ef,hi,kl,no,qr]coronene (HtB-HBC), on Cu{1 1 0} [26]. HtB-HBC is a larger derivative of coronene with a further sequence of aromatic rings around a coronene core and six t-butyl substituents instead of hydrogen on the outer periphery. The molecule has a shape close to that of a six-pointed star and gives greater scope for close packing through interdigitation of the t-butyl groups by rotation on a hexagonal or pseudohexagonal lattice. Elegant, high-resolution STM studies by Schrock et al. [26] reveal a h7 2 | 1 5i termed L lattice and mirror image h7 2| 1 5i R lattice, which are pseudohexagonal and exactly H3 times larger than the coronene lattice at H51a, H54a, and H57a as the distorted hexagonal nearest neighbors. These dimensions demand a clockwise or anticlockwise rotation of the molecules on a 2D isolated pseudohexagonal lattice to avoid overlap of the vdWenvelopes and, when this is mapped onto the substrate R or L lattices, diastereoisomerism results (see Figure 1.10). For the observed pair of enantiomers, the molecules find themselves rotated by 5 relative to the close-packed h1 1 0i direction of the substrate, while for the alternative pairing, which is not favored, a rotation by an equivalent amount but in the opposite sense relative to the hexagonal unit cell results in 21 rotation relative to the h1 1 0i direction. Studies of an isolated molecule adsorbed on Cu{1 1 1} favor the 5 rotation rationalizing the observed behavior [26]. More detailed discussion of chirality and diastereoisomerism in this system can be found in the paper by Schrock et al. [26] and in the work of Richardson [25], where consideration is also given to an alternative chiral structure for HtB-HBC/Cu{1 1 0}, which differs from that observed by Schrock et al. [26] only in the orientation of the pseudohexagonal lattice to the substrate. 1.2.3 Chiral Molecules on Achiral Surfaces

An isolated molecule, which is chiral in the gas phases, will necessarily be chiral on adsorption if the basic structure and conformation of the molecule are retained.

1.2 Surface Chirality Following Molecular Adsorption

Figure 1.10 UHV-STM images of HtB-HBC adsorption on Cu{1 1 0} showing the correlation between the orientation of the adsorbate lattice vectors and the local rotation of the molecule away from its high symmetry azimuthal orientation atop a copper atom.

The 5 , L ( þ 5 , R) gives an improved interdigitation of t-butyl groups compared to the diastereoisomers þ 5 , L (5 , R). (Reprinted with permission from Ref. [25]. Copyright 2006, American Chemical Society.

Adsorption of the opposite gas-phase enantiomer is necessary to generate the mirror image adsorbate system. STM has been widely used to investigate how substratemediated interactions and intermolecular hydrogen bonding influence the growth of 1D and 2D clusters and long-range ordered structures. In many cases, if chiral molecules form ordered structures on metal surfaces, the adsorbate forms an oblique unit cell such that the ordered adsorbate structure itself is chiral. In this case, the surface possesses both local chirality (determined by the molecule–surface complex) and global chirality (determined by the chirality of the ordered adsorbate domains). One of the most extensively studied examples of the adsorption of a simple chiral molecule on an achiral metal surface involves the adsorption of tartaric acid onto Cu {1 1 0}. Tartaric acid (H2TA) (HOOCCHOHCHOHCOOH) can exist in the (R,R), (S,S), and (R,S) forms. The initial work was motivated by a desire to understand why (R,R)-tartaric acid is the most successful chiral modifier in the Ni-catalyzed enantioselective hydrogenation of b-ketoesters. Work from the catalysis community had proposed that ordered, nanoporous 2D arrays of chiral molecules may be important in defining the active site for chiral catalytic reactions [27]. The shape of the chiral nanopores could favor the adsorption of a reactant molecule in a geometry favoring the production of one enantiomeric product. Alternatively, it was proposed that a direct interaction between a prochiral reagent and a single chiral modifier may be

j13

j 1 Chirality at Metal Surfaces

14

sufficient to direct the reaction along one enantiomeric route [28]. Hence, it was important to investigate how tartaric acid binds to a metal surface and the extent to which it forms ordered 2D arrays. On Cu{1 1 0}, a range of ordered structures were identified with STM following (R,R)-tartaric acid adsorption [29] as functions of tartaric acid coverage and temperature. At 300 K and above, tartaric acid adsorption occurred via deprotonation of either one or both COOH functionalities to produce monotartrate (HTA) or bitartrate (TA) species. Some of the ordered structures gave unit cells such as c(4  6) that would be indistinguishable from that produced by (S, S)-tartaric acid. The h9 0 | 1 2i structure was particularly significant from the point of view of surface chirality. This structure was observed exclusively, with no evidence being found for the mirror image structure h9 0 | 1 2i. By contrast, the adsorption of (S,S)-tartaric acid gave only the h9 0 | 1 2i structure under similar preparation conditions (Figure 1.11) [30]. In these structures, tartaric acid is adsorbed across the troughs of the Cu{1 1 0} surface in the doubly deprotonated bitartrate form. Barbosa and Sautet used DFT calculations to examine the preference by one enantiomer to form one of the two mirror equivalent domains [31]. It was found that there is an approximately 10 kJ mol1 preference for one ordered arrangement over the other. The energy preference is believed to be derived from an optimization of intramolecular H-bonding interactions involving the two OH groups at the chiral centers and was not believed to be related to intermolecular H-bonding interactions since adjacent molecular species are too far apart for any significant H-bonding interactions to occur. Fasel et al. carried out a detailed XPD characterization of the adsorption

Figure 1.11 STM images (13.5 nm  11.5 nm) of the h9 0 | 1 2i (left) and h9 0 | 1 2i (right) phases of (R,R)- and (S,S)-tartaric acid, respectively, on Cu{1 1 0}. (Adapted with permission from Ref. [29]. Copyright 2000, Macmillan Publishers Ltd.)

1.2 Surface Chirality Following Molecular Adsorption

geometry of (R,R)- and (S,S)-tartaric acid in the h9 0 | 1 2i phase and concluded that individual TA species were adsorbed with the planes defined by the two carboxylate OCO planes of each TA species being distorted away from the h1 1 0i azimuth. The distortion observed by (R,R)-TA was exactly mirrored for (S,S)-TA [32]. An interesting feature of the h9 0 | 1 2i structure is the tendency for clusters with three molecular features to be observed in the STM images. It is implicitly assumed in the proposed structural models that the adsorption site of each TA species is essentially equivalent. If this was the case, then it is not obvious why TA species from clusters of three species separated by channels in the surface. Under certain tip conditions, the three features of the cluster appear to give different z-contrast. This may suggest that the three species are in slightly different adsorption sites and that it is energetically more favorable to have an “empty” channel between rows of clusters than to accommodate an additional TA species in a less favorable adsorption site. There is some evidence from STM images of the Ni{1 1 0}/tartaric acid system that the TA species influence the electronic structure of the underlying Ni in the vicinity of the adsorbed TA species perhaps via some local restructuring of Ni atoms [33, 34]. The formation of clusters and channels in the Cu{1 1 0} experiments may be related to a release of strain in the surface copper atoms. This proposed mechanism is supported by a combined DFT and kinetic Monte Carlo study by Hermse et al. [35]. 1.2.4 Chiral Molecules on Chiral Surfaces

One of the central features of many geochemical models for the origin of life is the proposal that abiotic processes that select left-handed molecules versus right-handed molecules could occur on the surfaces of chiral minerals [36]. There are many examples of minerals whose bulk structures are intrinsically chiral. The most naturally abundant chiral mineral is quartz (SiO2) that belongs to the trigonal space group P3221. The structure of quartz contains a helical arrangement of corner-linked SiO4 tetrahedra. The sense of the helix determines left- or right-handed quartz. In addition, more than 200 chiral metal oxide structures are known [37]. Face-centered cubic metallic elements (e.g., Cu, Ni, Pt, etc.) are intrinsically achiral. However, Gellman and coworkers [38] highlighted that certain surfaces of fcc single crystals are chiral (Figure 1.12). Those surfaces displaying kinked steps with uneven step lengths either side of the kink are not superimposable on their mirror image. They proposed that surfaces could be denoted (R) or (S) by assessing whether the sequence of microfacets in order of decreasing atomic density {1 1 1} > {1 0 0} > {1 1 0} is clockwise (R) or anticlockwise (S) about the kink atom. STM imaging of step–kink surfaces such as Cu{6 4 3} show a high degree of atom mobility at the step edges [39]. Sholl et al. used DFT simulations to show that naturally chiral metal surfaces retain their net chirality even after their local structure is disrupted by thermal step roughening [40]. More recently, Jenkins and Pratt showed that stepped bcc and hcp surfaces may be chiral in the absence of kinks [41]. It was realized at an early stage that the adsorption of two enantiomers at chiral step–kink sites was likely to occur with slightly different adsorption energies. In

j15

j 1 Chirality at Metal Surfaces

16

Figure 1.12 Schematic diagram showing the mirror equivalent step–kink arrangements of the fcc {6 4 3}R and fcc{6 4 3}S surfaces (Adapted with permission from Ref. [37]. Copyright 1996, American Chemical Society.)

catalysis, such small modifications to the reaction pathway can be amplified to make significant changes in selectivity. Attard, with an elegant series of cyclic voltammetry experiments, displayed not only a difference between the behavior of D- and L-glucose at the (R)-Pt{6 4 3} surface but also an equivalence between the behavior of D-glucose/ (R)-Pt{6 4 3} and L-glucose/(S)-Pt{6 4 3} [42]. Temperature-programmed desorption (TPD) has been used to identify subtle differences in adsorption energy for enantiomers at chiral surfaces. For example, Gellman and coworkers reported enantiospecific behavior of (R)- and (S)-propylene oxide on Cu{6 4 3} [43]. However, the adsorption of (R)- and (S)-butanol (the simplest chiral alcohols) produced TPD data that were indistinguishable on Ag{6 4 3} [38]. STM studies of the adsorption of chiral molecules on chiral surfaces are surprisingly sparse. Zhao and Perry showed that (R)-3-methylcyclohexanone forms ordered structures on Cu{6 4 3} with a molecular spacing consistent with the spacing of kinks on the ideal Cu{6 4 3} surface [39]. Kuhnle et al. were able to probe, with atomic resolution, the interaction of chiral molecules with kink sites in the case of cysteine on Au{1 1 0} [44]. Although the surface is achiral, it displays both (S)- and (R)-kinks in approximately equal numbers. Kuhnle et al. showed that dimers formed from (R)cysteine adopt different adsorption geometries at (S)-kinks from (S)-cysteine dimers, demonstrating enantiospecific adsorption at these chiral centers. Furthermore, dense, homochiral cysteine islands are found to preferentially grow from kink sites of a specific chirality [44]. 1.2.5 Chiral Etching

For an fcc crystal, the low-index faces (e.g., {1 1 1}, {1 0 0}, and {1 1 0}) are thermodynamically the most stable, having the lowest surface free energies.

1.2 Surface Chirality Following Molecular Adsorption

Chemisorption can lead to large changes in surface free energies. There are many examples where chemisorption of organic molecules on a low-index crystal face results in faceting of a metal surface. A number of factors influence the formation of facets including face-specific adsorption energies, the energy difference between kinks, steps, and terraces, substrate-mediated intermolecular interactions, and surface diffusion barriers. Recent studies of organic molecules adsorbed on lowindex surfaces have found that high-index facets can be formed with complex organic molecular adsorbates containing electronegative elements such as O and N atoms in their functional groups. In these systems, the energy gain, which drives the morphology change, could originate from the molecule–substrate interactions and substrate-mediated interadsorbate interactions, which stabilize the steps and kinks of the substrates. Organic molecules with carboxylic acid functionalities commonly exhibit faceting on metal surfaces. For example, STM investigations have revealed that formic acid [45], benzoic acid [46], and p-aminobenzoic acid [47] all exhibit faceting behavior on Cu{1 1 0}. It has been well established that at room temperature the carboxylic acid group is deprotonated to the carboxylate. A preferential alignment of step edges along the [1 1 2] directions can be easily identified for both formate and acetate. It seems likely that the driving force for the formation of this orientation of step edge is the ordering of the molecular species into c(2  2) arrangements. Surface structures formed by the adsorption of benzoic acid are much more complicated [46]. Benzoate species can adopt either flat-lying or upright geometries and may form several different periodic structures depending on coverage and annealing temperature. The related molecule p-aminobenzoic acid also displays extensive faceting on the Cu{1 1 0} surface as shown in Figure 1.13 [47]. In these cases, it is possible to identify two symmetrically equivalent (11 13 1) facets giving the characteristic sawtooth arrangement of facets. The fact that similar facets are observed for both benzoic and p-aminobenzoic acid leads to the conclusion that the formation of facets is directed by the flat-lying carboxylate units. In the case of formate and acetate, where vibrational spectroscopy reveals upright carboxylate units, step bunching is not observed leading to the proposal that the adsorbate-mediated step–step interaction required for step bunching is at best only weakly attractive when the carboxylate is perpendicular to the surface [48]. Pascual et al. [49] investigated the adsorption of the prochiral carboxylic acid 4[trans-2-(pyrid-4-yl-vinyl)]benzoic acid on Ag{1 1 0}. Following exposure to submonolayer coverages of PVBA and thermal processing, similar sawtooth facets were observed as for benzoic acid on Cu{1 1 0} (Figure 1.14). It was proposed that the formation of facets was driven by the interaction between the carboxylate and the {1 0 0} microfacets at step edges. The microfacets then act as chiral templates nucleating the growth of supramolecular PVBA structures. The chirality of the PVBA species at the microfacet determined the structure of the first four assembled rows of molecules. It is perhaps unsurprising that when a chiral adsorbate is used containing the carboxylate functionality, the distribution of facets produced becomes chiral. Zhao and coworkers carried out studies of the adsorption of a range of amino acids on

j17

j 1 Chirality at Metal Surfaces

18

Figure 1.13 High-resolution STM image (50 nm  50 nm, bias 1.14 V, tunneling current 6.1 nA) showing the faceted structure of p-aminobenzoic acid on Cu(1 1 0). (Reprinted with permission from Ref. [47]. Copyright 2003, Elsevier.)

Cu{0 0 1} [49–52]. In the case of the achiral glycine molecule, a tendency was found for the formation of (3 1 17) facets. Since there is neither rotational nor reflectional symmetry within individual facets, eight symmetry-related facets should be expected, that is, (1 3 17), (1 3 17), (3 1 17), (3 1 17), (3 1 17), (3 1 17), (1 3 17), and (1 3 17) The first four facets are rotationally equivalent to each other as are the final four. The two sets are related by reflectional symmetry to each other. When a chiral adsorbate, for example, S-lysine, is used, the reflectional symmetry is no longer valid and only rotationally equivalent facets should be formed. This was demonstrated elegantly by Zhao with STM [53]. The driving force for facet formation is proposed to be a “three-point interaction” involving the carboxylate group, the a-amino group, and the amino-terminated side chain. The simultaneous optimization of adsorbate–adsorbate and adsorbate–substrate interactions determines the stereochemistry of the facet. Surface faceting may be particularly significant in chiral heterogeneous catalysis, particularly in the Ni/b-ketoester system. The adsorption of tartaric acid and glutamic acid onto Ni is known to be corrosive and it is also established that modifiers are leached into solution during both the modification and the catalytic reaction [28]. The preferential formation of chiral step–kink arrangements by corrosive adsorption could lead to catalytically active and enantioselective sites at step–kinks with no requirement for the chiral modifier to be present on the surface.

1.3 Chiral Amplification and Recognition

Figure 1.14 (a) STM image (10 nm  10 nm, tip bias þ 0.52 V, tunneling current 0.5 nA) of a PVBA-induced “sawtooth blade” in a restructured Ag(1 1 0) surface terrace. (b) Structural model of the chiral kink arrangements induced by lateral interaction of molecular carboxylate end groups with Ag{1 0 0} microfacets. (Reprinted with permission from Ref. [48]. Copyright 2004, American Institute of Physics.)

1.3 Chiral Amplification and Recognition 1.3.1 Chiral Amplification in Two Dimensions

In Section 1.2.1, we discussed the phenomenon of adsorbate-induced chirality whereby the adsorption of achiral species (e.g., glycine) results in the formation of two mirror equivalent domains on the surface. It has recently been shown that the presence of relatively small mole fractions of chiral dopants can result in the exclusive

j19

j 1 Chirality at Metal Surfaces

20

formation of one of the two mirror equivalent domains of the achiral species. For example, succinic acid (HOOCCH2CH2COOH), an achiral molecule, forms two mirror equivalent domains h9 0 | 2 2i and h9 0| 2 2i on Cu{1 1 0} [54]. The doubly deprotonated succinate species are bound via both carboxylate groups to the Cu surface – the mirror relationship between the two domains is thought to arise from the twist of the carbon backbone of succinate with respect to the [0 0 1] surface direction. When as little as 2 mol% (R,R)-tartaric acid is coadsorbed with succinic acid, LEED beams associated with the h9 0 | 2 2i structure are extinguished. The opposite behavior is observed when the dopant is (S,S)-tartaric acid [55]. This behavior is analogous to the “sergeants and soldiers” principle observed for helical polyisocyanate copolymers [56]. The mechanism for this effect is proposed to be substratemediated. Succinate species are unable to form intermolecular H-bonds, so a chiral footprint imposed on the surface by a tartrate species is thought to control the adsorption geometry of the surrounding complex creating an effect that is amplified over 30–50 molecules in a given domain [55]. 1.3.2 Chiral Switching

In discussing adsorption-induced chirality, it is generally assumed that interconversion of enantiomers is highly unlikely since such an interconversion, for example, under thermal excitation, requires a reduced interaction with the surface and desorption is a more probable outcome. However, if the molecule is relatively large and the chiral center does not make a large contribution to the molecule–substrate binding, then there is the possibility that a low-frequency mode can be excited sufficiently on heating that a bond rotation is possible leading to a chiral switching even without the molecule fully leaving the surface. This effect has been observed by Linderoth and coworkers [57]. The molecule consists of a linear backbone formed out of three benzene rings connected by ethynylene spokes and is functionalized at each end with an aldehyde, a hydroxyl, and a tert-butyl group. The molecule is achiral in the gas phase but adsorption on Au{1 1 1}, with the main molecular backbone parallel to the surface, creates, by restricting rotation around the ethynylene spokes, two equivalent chiral centers such that molecules can be classified as LL, RR, or RL, the latter being internally racemic. The tert-butyl group can be readily identified in the STM image and its position relative to the molecular backbone determined, thereby permitting the chirality properties of the molecule to be determined. Thermal switching of conformations was noted and ascribed to a partial loss of binding at one end of the molecule. Detailed temperature-dependent studies allowed the barrier to switching to be determined as approximately 0.3 eV. More subtly, it is found that the internally racemic LR/RL conformation has a relatively low probability on the surface and that it is more likely to switch than the LL or RR enantiomers: the difference in barrier heights being 0.04 eV. This is related to the interadsorbate interactions, which favor the LL/RR molecules over LR/RL. The possibility for chiral switching of this type provides a new mechanism for the growth of large homochiral domains as an alternative to separation relying on interdiffusion.

1.3 Chiral Amplification and Recognition

1.3.3 Chiral Recognition

Perhaps the simplest form of chiral recognition is that in which one enantiomer, for example, A, of a chiral object displays a stronger interaction with a particular enantiomer of a second chiral object, for example, B, rather than its mirror image, B. Of the four possible diastereoisomeric interactions AB, AB; AB, AB, the first two form a mirror equivalent, enantiomeric pair as do the latter. However, the crossrelationships are inequivalent, nonmirror images, for example, AB and AB, and in a chemical system, there would be an energetic preference for one pair of enantiomers over the other. This is the key to the significance of chirality in biology and, therefore, in the need to develop chiral products in the pharmaceutical and agrochemical industries. Studying chiral recognition processes at surfaces is therefore relevant to a better understanding of the separation of enantiomers, for example, following their preparation in an insufficiently enantioselective reaction. It is also relevant to the development of biosensors and biocompatible materials. We have already covered the interaction of chiral molecules with chiral surfaces, which is an important example of chiral recognition and diastereoisomerism. In this section, attention is focused on chiral recognition between molecules adsorbed on surfaces and it is useful to distinguish between self-recognition processes and those involving different molecular species. The latter can be described by the AB system introduced above while extending the analogy to self-recognition; it is the energy differences between the species AA, AA; AA, AA that is of interest. Here, AA is the mirror image of AA, so these form a pair of enantiomers. Similarly, AA and AA are also enantiomers but somewhat trivially since they are also equivalent and might be described as internally racemic. We have already discussed examples of what is effectively chiral self-recognition, when we described the formation of chiral clusters, chains, and arrays following adsorption involving induced chirality in otherwise achiral species in Section 1.2.1. Now, we show examples of self-recognition between intrinsically chiral molecules adsorbed as a racemic mixture on achiral surfaces leading to segregation of enantiomers if the homochiral (AA/AA) pairing is preferred over the heterochiral interaction (AA). A nice example of this is revealed in the work of Besenbacher and coworkers [58] on the adsorption of a racemic mixture of D- and L-cysteine on Au{1 1 0}. At low coverages, STM shows the molecules are present in pairs and, on the basis of the alignment of any given pair with respect to the h1 1 0i direction, it can be identified as being either DD or LL. Notably, DL heterochiral pairs are not observed (Figure 1.15). The reason for the homochiral preference lies in the orientation of the cysteine molecules on the gold surface determined by AuS and AuN interactions. The carboxylic acid functionality is not involved in any significant interaction with the gold substrate but rather dominates the pairing interaction between enantiomers. This “three-point” bonding of each molecule, AuS, AuN, and OHO, drives the self-recognition preference for homochiral pairs [58]. A more subtle example of homochiral preference, which draws attention to the conformational changes in the molecules needed to achieve self-recognition, is that

j21

j 1 Chirality at Metal Surfaces

22

Figure 1.15 Adsorption of cysteine on Au{1 1 0}. Molecular model shows the deprotonated thiolate surface species. (a) Model of the reconstructed (1  2)-Au{1 1 0} surface; (b–d) show, respectively, dimers of L-cysteine, D-cysteine, and the two together characteristically rotated relative to the h1 1 0i azimuth. (Adapted with permission from Ref. [57]. Copyright 2002, Macmillan Publishers Ltd.)

of the dipeptide D-phenylalanine-D-phenylalanine (D-Phe-D-Phe) and its enantiomer L-Phe-L-Phe adsorption on Cu{1 1 0} [59]. Following adsorption of a racemic mixture at low coverage, isolated species are recognizable by the orientation of their principal axis with respect to the h1 1 0i azimuth of the substrate; the LL (DD) enantiomer is rotated 34 (counter) clockwise as shown in Figure 1.16. Density functional and molecular dynamics calculations support an interpretation that the molecule adopts a conformation similar to the gas phase, in which the amine and carboxylic acid functionalities lie on the same side of the principal molecular axis, the peptide backbone. In contrast, homochiral chains (D-Phe-D-Phe)n and (L-Phe-L-Phe)n are observed by STM to have the principal axis rotated by 74 and heterochiral chains are not observed. Calculations suggest that the conformation of each molecule in a chain is dramatically changed relative to the isolated molecules with, inter alia, the amine and carboxylic functionalities now lying on opposite sides of the backbone to optimize intermolecular zwitterion formation between the amine of one molecule and the carboxylic acid of its neighbor. The need to consider the dynamic nature rather than simple lock–key models of chiral recognition is thereby emphasized. Adenine as an isolated molecule has no symmetry elements and therefore might “mathematically” be considered chiral; however, as in the case of glycine (Section 1.2.1), this description is not useful in chemistry since the enantiomers differ only by inversion through the weakly pyramidal nitrogen atom of the amine functionality, the main body of the molecule being planar. The inversion corresponds to a low-frequency vibration and a low-energy barrier such that single enantiomers

1.3 Chiral Amplification and Recognition

Figure 1.16 Comparison of the structure of an isolated (L-Phe-L-Phe) on Cu{1 1 0} rotated 34 clockwise with respect to the h1 1 0i azimuth (a, c, e) and that of the molecules found in rows that are rotated by 74 (b, d, f) based on STM

images. The superimposed models indicate that the change in rotation is linked to a major change of conformation to enable strong intermolecular bonding. (Adapted with permission from Ref. [58].)

cannot be realized. However, adsorption of adenine on a Cu{1 1 0} surface gives rise to flat-lying molecules, which then have a high barrier to interchange of enantiomers; that is, chirality is induced by adsorption [60]. At coverages up to one monolayer, adenine forms homochiral dimers that link into homochiral chains, whose direction on the Cu{1 1 0} substrate is correlated with their chirality [60] as shown in the left panel of Figure 1.17. Subsequent adsorption of one enantiomer of phenylglycine leads to an intermolecular recognition process that favors the decoration of chains running in the (1, 2) direction by S-phenylglycine (right-hand panel of Figure 1.17) while R-phenylglycine decorates the mirror image (1, 2) adenine chains [61]. The origin of the strong interaction between the amino acid, which adsorbs on Cu{1 1 0} as the anion, and the nucleic acid base is electrostatic favoring the close approach of the carboxylate functionality of phenylglycine to the nitrogen of the adenine’s amine group, which

j23

j 1 Chirality at Metal Surfaces

24

Figure 1.17 The left-hand STM image shows homochiral adenine rows aligned in low symmetry but mirror image azimuths on a Cu(1 1 0) surface. On the right, adenine rows in the (1,2) direction are decorated with double rows of S-phenylglycine molecules, while no such interaction occurs with (1, 2) rows. (Adapted with permission from Ref. [60]. Copyright 2000, Macmillan Publishers Ltd.)

lies on the periphery of the chain [62]. Chiral recognition occurs because there is also a repulsive interaction between the amine groups of the two molecules and this is less for the favored enantiomer than for the other [62]. 1.3.4 Prochiral Molecules Interacting with Chiral Surfaces

In enantioselective catalysis, the key problem to overcome is the fact that the Gibbs energy change from gas-phase prochiral reagent to gas-phase product is identical for each enantiomeric product molecule. Hence, in the absence of any chiral influence on the reaction, a racemic mixture of products should always ensue. By providing a reaction pathway to one product that has a much lower activation barrier, the selectivity can be skewed to give an enantiomeric excess of one product. One of the most heavily researched examples of heterogeneous enantioselective catalysis is the hydrogenation of b-ketoesters over Ni catalysts [28]. The simplest b-ketoester is methylacetoacetate (MAA). This molecule is approximately planar and can adsorb via either molecular face with equal probability. The hydrogenation reaction is believed to occur via dissociative adsorption of H2 on the metal surface and attack by H from underneath the molecular plane of MAA. The stereochemistry of the chiral center thus produced would be determined by which face of the prochiral reagent lies down on the surface. To skew the reaction in an enantioselective direction, a clear requirement seems to be to restrict the adsorption geometry to exclusively one enantioface. In this respect, the coverage of chiral modifier is thought to be crucial. If the coverage is too low, the formation of 1 : 1 complexes between modifier and reactant could induce some enantioselectivity, but the adsorption of MAA on bare metal sites would be expected to occur racemically. If the coverage is too high, there

1.4 Conclusions

Figure 1.18 STM image (4 nm  4 nm) showing the 2D “cocrystalline” structure consisting of an ordered array of 1 : 1 H-bonded complexes of (R,R)-tartrate and methylacetoacetate species on Ni{1 1 1} giving a chiral h3 1 | 3 4i structure. (Adapted with permission from Ref. [62]. Copyright 2002, Elsevier.)

may be insufficient space for MAA to adsorb on the surface. Indeed, the sticking probability of MAA on Ni{1 1 1} covered by high coverages of (R,R)-tartaric acid [63] or (S)-glutamic acid [64] is essentially zero. At intermediate tartaric acid coverages, Jones and Baddeley showed that the adsorption of MAA caused a restructuring of the tartrate adlayer and the formation of an ordered array of 1 : 1 tartrate:MAA complexes (Figure 1.18). Interestingly, the geometry of each MAA molecule in the array appeared identical and corresponding to the geometry required for the formation of (R)-methyl-3-hydroxybutyrate – the product observed in excess in the catalytic reactions [63].

1.4 Conclusions

Chirality at surfaces can be manifested in a number of forms including the intrinsic chirality of the surface structure and even the induction of chirality via the adsorption of achiral molecules onto achiral surfaces. The ability of STM to probe surfaces on a local scale with atomic/molecular resolution has revolutionized the understanding of these phenomena. Surfaces that are globally chiral either due to their intrinsic structure or due to the adsorption of chiral molecules have been shown by STM to establish control over the adsorption behavior of prochiral species. This could have profound consequences for the understanding of the origin of homochirality in life on Earth and in the development of new generations of heterogeneous chiral catalysts that may, finally, make a substantial impact on the pharmaceutical industry.

j25

j 1 Chirality at Metal Surfaces

26

References 1 Cahn, R.S., Ingold, C.K., and Prelog, V. (1966) Angew. Chem. Int. Ed. Engl., 78, 385. 2 Knowles, W.S. (2002) Angew. Chem. Int. Ed., 41, 1999. 3 Noyori, R. (2002) Angew. Chem. Int. Ed., 41, 2008. 4 Sharpless, K.B. (2002) Angew. Chem. Int. Ed., 41, 2024. 5 Bradshaw, A.M. and Richardson, N.V. (1996) Pure Appl. Chem., 68, 457. 6 Weckesser, J., De Vita, A., Barth, J.V., Cai, C., and Kern, K. (2001) Phys. Rev. Lett., 87, 096101. 7 Bohringer, M., Morgenstern, K., Schneider, W.D., Berndt, R., Mauri, F., De Vita, A., and Car, R. (1999) Phys. Rev. Lett., 83, 324. 8 Bohringer, M., Morgenstern, K., Schneider, W.D., and Berndt, R. (1999) Angew. Chem. Int. Ed., 38, 821. 9 Blum, M.C., Cavar, E., Pivetta, M., Patthey, F., and Schneider, W.D. (2005) Angew. Chem. Int. Ed., 44, 5334. 10 Barlow, S.M., Kitching, K.J., Haq, S., and Richardson, N.V. (1998) Surf. Sci., 401, 322. 11 Booth, N.A., Woodruff, D.P., Schaff, O., Giessel, T., Lindsay, R., Baumgartel, P., and Bradshaw, A.M. (1998) Surf. Sci., 397, 258. 12 Chen, Q., Frankel, D.J., and Richardson, N.V. (2002) Surf. Sci., 497, 37. 13 Rankin, R.B. and Sholl, D.S. (2004) Surf. Sci., 548, 301. 14 Toomes, R.L., Kang, J.H., Woodruff, D.P., Polcik, M., Kittel, M., and Hoeft, J.T. (2003) Surf. Sci., 522, L9. 15 Haq, S., Massey, A., Moslemzadeh, N., Robin, A., Barlow, S.M., and Raval, R. (2007) Langmuir, 23, 10694. 16 Sayago, D.I., Polcik, M., Nisbet, G., Lamont, C.L.A., and Woodruff, D.P. (2005) Surf. Sci., 590, 76. 17 Rankin, R.B. and Sholl, D.S. (2005) Surf. Sci., 574, L1. 18 Lackinger, M., Griessl, S., Heckl, W.M., and Hietschold, M. (2002) Anal. Bioanal. Chem., 374, 685.

19 England, C.D., Collins, G.E., Schuerlein, T.J., and Armstrong, N.R. (1994) Interfac. Des. Chem. Sensing, 561, 202. 20 Richardson, N.V. (2007) New J. Phys., 9, 395. 21 Bennett, R.A., Pang, C.L., Perkins, N., Smith, R.D., Morrall, P., Kvon, R.I., and Bowker, M. (2002) J. Phys. Chem. B, 106, 4688. 22 Lackinger, M., Griessl, S., Heckl, W.M., and Hietschold, M. (2002) J. Phys. Chem. B, 106, 4482. 23 Gyarfas, B.J., Wiggins, B., Zosel, M., and Hipps, K.W. (2005) Langmuir, 21, 919. 24 Yoshimoto, S., Narita, R., and Itaya, K. (2002) Chem. Lett., 356. 25 Richardson, N.V. (2007) New J. Phys., 9, 395. 26 Schrock, M., Otero, R., Stojkovic, S., Hummelink, F., Gourdon, A., Laegsgaard, E., Stensgaard, I., Joachim, C., and Besenbacher, F. (2006) J. Phys. Chem. B, 110, 12835. 27 Webb, G. and Wells, P.B. (1992) Catal. Today, 12, 319. 28 Izumi, Y. (1983) Adv. Catal., 32, 215. 29 Lorenzo, M.O., Haq, S., Bertrams, T., Murray, P., Raval, R., and Baddeley, C.J. (1999) J. Phys. Chem. B, 103, 10661. 30 Lorenzo, M.O., Baddeley, C.J., Muryn, C., and Raval, R. (2000) Nature, 404, 376. 31 Barbosa, L. and Sautet, P. (2001) J. Am. Chem. Soc., 123, 6639. 32 Fasel, R., Wider, J., Quitmann, C., Ernst, K.H., and Greber, T. (2004) Angew. Chem. Int. Ed., 43, 2853. 33 Humblot, V., Haq, S., Muryn, C., Hofer, W.A., and Raval, R. (2002) J. Am. Chem. Soc., 124, 503. 34 Hofer, W.A., Humblot, V., and Raval, R. (2004) Surf. Sci., 554, 141. 35 Hermse, C.G.M., van Bavel, A.P., Jansen, A.P.J., Barbosa, L., Sautet, P., and van Santen, R.A. (2004) J. Phys. Chem. B, 108, 11035.

References 36 Hazen, R.M. and Sholl, D.S. (2003) Nat. Mater., 2, 367. 37 Halasyamani, P.S. and Poeppelmeier, K.R. (1998) Chem. Mater., 10, 2753. 38 McFadden, C.F., Cremer, P.S., and Gellman, A.J. (1996) Langmuir, 12, 2483. 39 Zhao, X.Y. and Perry, S.S. (2004) J. Mol. Catal. A – Chem., 216, 257. 40 Sholl, D.S., Asthagiri, A., and Power, T.D. (2001) J. Phys. Chem. B, 105, 4771. 41 Jenkins, S.J. and Pratt, S.J. (2007) Surf. Sci. Rep., 62, 373. 42 Attard, G.A. (2001) J. Phys. Chem. B, 105, 3158. 43 Horvath, J.D. and Gellman, A.J. (2001) J. Am. Chem. Soc., 123, 7953. 44 Kuhnle, A., Linderoth, T.R., and Besenbacher, F. (2006) J. Am. Chem. Soc., 128, 1076. 45 Bowker, M., Poulston, S., Bennett, R.A., and Stone, P. (1998) J. Phys: Condens. Matter, 10, 7713. 46 Chen, Q., Perry, C.C., Frederick, B.G., Murray, P.W., Haq, S., and Richardson, N.V. (2000) Surf. Sci., 446, 63. 47 Chen, Q., Frankel, D.J., and Richardson, N.V. (2001) Langmuir, 17, 8276. 48 Chen, Q. and Richardson, N.V. (2003) Prog. Surf. Sci., 73, 59. 49 Pascual, J.I., Barth, J.V., Ceballos, G., Trimarchi, G., De Vita, A., Kern, K., and Rust, H.P. (2004) J. Chem. Phys., 120, 11367. 50 Zhao, X.Y., Gai, Z., Zhao, R.G., Yang, W.S., and Sakurai, T. (1999) Surf. Sci., 424, L347.

51 Wang, H., Zhao, X.Y., and Yang, W.S. (2000) Acta Phys. Sin., 49, 1316. 52 Wang, H., Zhao, X.Y., Zhao, R.G., and Yang, W.S. (2001) Chin. Phys. Lett., 18, 445. 53 Zhao, X.Y. (2000) J. Am. Chem. Soc., 122, 12584. 54 Humblot, V., Lorenzo, M.O., Baddeley, C.J., Haq, S., and Raval, R. (2004) J. Am. Chem. Soc., 126, 6460. 55 Parschau, M., Romer, S., and Ernst, K.H. (2004) J. Am. Chem. Soc., 126, 15398. 56 Green, M.M., Reidy, M.P., Johnson, R.J., Darling, G., Oleary, D.J., and Willson, G. (1989) J. Am. Chem. Soc., 111, 6452. 57 Weigelt, S., Busse, C., Petersen, L., Rauls, E., Hammer, B., Gothelf, K.V., Besenbacher, F., and Linderoth, T.R. (2006) Nat. Mater., 5, 112. 58 Kuhnle, A., Linderoth, T.R., Hammer, B., and Besenbacher, F. (2002) Nature, 415, 891. 59 Lingenfelder, M., Tomba, G., Costantini, G., Ciacchi, L.C., De Vita, A., and Kern, K. (2007) Angew. Chem. Int. Ed., 46, 4492. 60 Chen, Q., Frankel, D.J., and Richardson, N.V. (2002) Langmuir, 18, 3219. 61 Chen, Q. and Richardson, N.V. (2003) Nat. Mater., 2, 324. 62 Blankenburg, S. and Schmidt, W.G. (2007) Phys. Rev. Lett., 99, 196107. 63 Jones, T.E. and Baddeley, C.J. (2002) Surf. Sci., 519, 237. 64 Jones, T.E., Urquhart, M.E., and Baddeley, C.J. (2005) Surf. Sci., 587, 69.

j27

j29

2 The Template Route to Nanostructured Model Catalysts Conrad Becker and Klaus Wandelt

2.1 Introduction

Even though the majority of industrially produced chemicals are based on processes that include at least one heterogeneously catalyzed step, the basic physical and chemical properties that govern the action of a particular catalyst are far from being understood. Surface science can play a major role in the elucidation of the molecular principles that govern reactions on catalyst surfaces. This approach was brought to attention of a wider public when the Nobel Prize for Chemistry in 2007 was awarded to Gerhard Ertl “for his studies of chemical processes on solid surfaces.” The investigation of the basic physical and chemical properties of heterogeneous catalysts down to the molecular and atomic level is, indeed, one of the prime goals of surface science. On the route to ultimate understanding of heterogeneously catalyzed reactions, two major obstacles need to be overcome, and these are commonly referred to as the pressure gap and the materials gap. The pressure gap stems from the fact that industrial catalysts are usually run at rather high pressures, whereas surface science studies are generally undertaken in ultrahigh vacuum. In the recent past, however, considerable progress has been made to close this gap by making standard surface science techniques work at elevated pressures (see Refs [1–3]). Therefore, it seems likely that the pressure gap is not the major obstacle for understanding heterogeneously catalyzed reactions any more provided the structure of the catalytic surface is well known. Here the materials gap enters the scene. It is related to the fact that industrial catalysts possess a highly complex structure and topology, which makes them inaccessible for surface science tools. This problem can be overcome by the use of model catalysts of reduced complexity, that is, designed model catalysts that mimic the nanoscopic structure of the industrial catalyst. For many years, metal single-crystal surfaces have been – and are still – used for this purpose. However, since industrial catalysts are often composed of oxidesupported metal particles, extended metal surfaces are – even though of very low complexity – rather unrealistic models. More realistic model catalysts are provided by

j 2 The Template Route to Nanostructured Model Catalysts

30

metal-on-oxide systems of low complexity. In this respect, it has turned out that thin oxide films on metal substrates can be used as support for the preparation of such systems. By depositing metal vapor on these films, model catalysts can be prepared, which possess the big advantage over bulk oxide materials that they can be investigated by surface science methods that rely on charged particles [4]. The big drawback of such systems is, however, that in general their complexity is still rather high. Owing to the nucleation process of metals on oxide surfaces, which is in general controlled by the defects of the surface, it is quite difficult to reach a uniform particle size and distribution since the defects are randomly distributed over the surface. This renders investigation of particle size effects difficult because only the ensemble average over a particular size distribution is experimentally accessible. The statistical arrangement of defects on the surface not only leads to different capture zones, and, as a consequence, to metal particles of different sizes (Figure 2.1a), but, in turn, also results in a different reverse spillover from the support to the individual particles during a catalytic reaction. This influences the reaction kinetics, which is closely connected to the arrival rate of reactants on the particle [5]. One way to overcome all these limitations is to use an ordered nanostructured model catalyst. The schematic representation of such a model catalyst is provided in Figure 2.1b. It will provide a uniform particle size and distribution, thus, enabling measurements on an ensemble of identical particles. The fabrication of such a system can be accomplished only by nanofabrication, and different routes can be imagined in this context. We will focus in the following section on the template-controlled growth of metal clusters on thin oxide films, which has proven to give excellent results in terms of low complexity. This approach has been successfully employed for metal-on-metal systems (for a comprehensive review see [6]) and has recently been extended to metal growth on oxide films. The key factor in this approach is the provision of suitable growth templates. Before we turn to this point, we will discuss in which way surfaces may be thought of as templates.

(a)

(b)

Figure 2.1 Schematic representation of a random model catalyst surface (a) and an idealized model catalyst (b). The black dots correspond to metal particles, dotted circles to their capture zones.

2.2 Surfaces as Two-Dimensional Templates

2.2 Surfaces as Two-Dimensional Templates

The concept of templates is well known in our contemporary life. Templates are generally used to shape a product. This can be, for example, the mold of a church bell or style sheet in office applications in computing. One common feature of templates is that the information, which is used to shape a product, is intrinsically encoded in the template. A good example in this context is a strand of DNA that sets the genetic sequence of new strands. If we want to use surfaces to act as templates, for example, in nanostructuring, we have to find a way to encode the information into the surface structure. This approach is commonly used in the fabrication of semiconducting devices, where lithography is employed to pattern the surface of a wafer. The pattern thus created is then used as a template for the following manufacturing steps. This approach, generally referred to as “top-down approach,” relies on the active patterning of the structure of the surface by external intervention. Even though this approach is frequently and successfully applied in a variety of areas, it is subject to a number of limitations in the context of model catalyst fabrication. First, if a lithographic process is used, spatial resolution is limited by either the physical properties of the radiation source employed for patterning (e.g., light, electron beams) or the chemical properties of the resist, which contains the pattern. In general, these limitations impose a resolution limited to a few tenths of nanometers. However, very well ordered model catalysts have been prepared in the past using this approach (see Figure 2.2). To fully exploit the nanoscopic properties of materials, for example, in catalysis, this structure size is much too large since it corresponds to a regime where the bulk properties of materials still dominate. An alternative approach can be the patterning of a surface by direct manipulation of atoms or molecules with the scanning tunneling microscopy (STM) [8], which has been successfully employed in the past

Figure 2.2 STM image of platinum particles on an oxidized silicon wafer. (Reproduced with permission from Ref. [7].)

j31

j 2 The Template Route to Nanostructured Model Catalysts

32

in many cases. In this case, only the size of the objects, which are manipulated, determines the resolution limit. However, it is still a top-down approach, which requires the intervention of the operator. Furthermore, the speed of the assembling process is rather slow and the patterning of surface areas on the micrometer scale may be tedious or even impossible. We, therefore, have to conceive an alternative pathway for creating nanopatterned templates. Fortunately, nature points the way for that providing us with nanopatterned surfaces that are derived from the intrinsic physical and chemical properties of the system itself. One possibility is to use the self-organization of colloidal particles to form regular nanostructured arrays [9, 10]. If applied to model catalyst design, this approach has, however, the disadvantage that the ligand shell, which stabilizes the particles, has to be removed without destroying the particles [11, 12]. Nanostructuring by self-organization is also found to occur in many metal surfaces. A particularly beautiful example is the “herringbone” or p “chevron” reconstruction of the Au(1 1 1) surface. This Au(1 1 1)-( 3  22) reconstruction, which has been observed and characterized by a large number of surface science techniques, provides us with a regular array of specific sites – the elbows of the reconstruction – that can act as nucleation sites [6]. The fabrication of a template using self-organization can be referred to as “bottom-up approach” since it relies on the inherent physical and chemical properties of the system. The beauty of this approach is that “nature is doing the job”; by choosing the components of the system, the self-organization process is controlled by the intrinsic properties of the system. The major obstacle that we are facing in this context is to set the controlling parameters of the self-assembly process in such a way that a desired nanopattern is generated. In the case of a pure metal surface such as Au(1 1 1), this of course cannot be done since it is a one-component system with no control parameters. However, other routes to nanostructured surfaces can be envisioned that provide an experimental control on the size and symmetry of the pattern. Among these, thin films provide the most promising route. This has been shown in a number of cases for metal films, oxide films, and molecular films. In the case of metal films, the difference in lattice parameters of substrate and film leads in many cases to the formation of strain relief pattern, to compensate for the lattice mismatch. This has been found for systems such as Ag/Pt(1 1 1) [13–15], Cu/Ru(0 0 0 1) [16], Ag/Cu(1 1 1) [17–19], and many others [20]. Likewise, oxide films can form regular nanopattern depending on the accommodation of the film to the substrate, and we will come back to this later. Finally, molecular films provide access to a whole new class of nanopatterned surfaces since in this case the intermolecular interactions of the molecules along with the molecule–substrate interaction play a decisive role in the pattern formation [21]. The latter case will thus provide us with interesting properties for the growth of nanostructures on surfaces but will have only a limited application for the fabrication of model catalysts because these should be, in general, metal-on-oxide systems. The nanopatterning of a surface is, however, only the first step toward a surface template. We will further need a specific property of the surface, which is encoded into the pattern. Only this will assure that the template will control the subsequent nucleation or growth processes. In many cases, this specific property will be

2.2 Surfaces as Two-Dimensional Templates

a particularly strong interaction (adsorption energy) at certain points of the surface [22]. This will provide us with a given density of nucleation sites that is determined by the nanostructure of the template. If we want to use these sites as a growth template, we have to control the kinetics of the process to reach nucleation on the template sites. Besides the adsorption energy of the particles on the traps, the flux of the impinging particles and the temperature of the substrate play a decisive role [23]. The latter has an important impact on the mobility of the particles and, thus, their ability to reach the traps. This has been illustrated in the work of Pr
evot et al. on the nucleation and growth of Au on the reconstructed N/Cu(1 0 0) surface [24]. This surface provides a square lattice of nucleation sites, which constitute a template for the ordered growth of Au clusters [25]. The kinetic Monte-Carlo (KMC) simulations shown in Figure 2.3 illustrate the importance of growth temperature for the action of the template. Only at intermediate temperatures (e.g., 240 K) is the growth perfectly template controlled. At this temperature, the diffusion of Au atoms on the surface is fast enough so that the nucleation sites are actually reached. At higher temperature also, the template is operative, all islands are found on template sites, but not all template sites are covered by islands, which is due to higher mobility of Au atoms. At lower temperature, the number density of islands exceeds that of the template sites indicating that besides the template-controlled (heterogeneous) nucleation, homogeneous nucleation also takes place. A more detailed picture of the temperature dependence of the growth is given in Figure 2.4, where the island density is plotted as a function of temperature. It can be seen that only in the temperature range from 207 to 288 K the growth is perfectly template controlled and the number of islands matches the number of available nucleation sites. This illustrates the importance of kinetic control for the creation of ordered model catalysts by a template-controlled process. Obviously, there has to be a subtle balance between the adatom mobility on the surface and the density of template sites (traps) to allow a template-controlled growth. We will show more examples of this phenomenon below. That not only an increased interaction energy at the traps can be responsible for a template-controlled growth but also an anisotropy of the surface diffusion

Figure 2.3 Kinetic Monte-Carlo simulations of Au growth on N/Cu(1 0 0) for three different temperatures. From left to right, T ¼ 180 K, T ¼ 240 K, and T ¼ 300 K. The coverage is 0.11 ML. (Reproduced with permission from Ref. [24].)

j33

j 2 The Template Route to Nanostructured Model Catalysts

34

Figure 2.4 Evolution of the island density with growth temperature for Au/N/Cu(1 0 0). (Reproduced with permission from Ref. [24].)

barrier may just be sufficient to control a growth process has been shown in several cases [26, 27]. Indeed, this leads to a template-controlled growth without specific interaction but by spatially confining the movement of the deposited particles. The nucleation is then homogeneous but controlled by the template. This approach is quite similar to the one used in lithography where the spatial confinement is provided by the lithographically produced pattern on the surface. Since surface diffusion in this case is the crucial parameter, the process will be subject to the same limitations in terms of temperature as in the case of template-controlled heterogeneous nucleation. Before we discuss the template-controlled growth of model catalysts in more detail, we will have to consider a few aspects of STM imaging of these systems. This will be crucial for the characterization of the model catalyst surfaces.

2.3 STM Imaging of Oxide Films

Even though scanning tunneling microscopy is nowadays a standard tool for surface characterization, special care has to be taken when using it for the investigation of thin oxide films. On clean metal surfaces, a simple metal–vacuum–metal tunneling gap is usually found [28]. In this case, the tunneling current can be to a first approximation described by the Tersoff–Hamann model [29], which postulates that the main contribution to the tunneling current is due to electrons coming from the Fermi edge. If we introduce a thin oxide film, the tunneling gap can be characterized by a metal–oxide–vacuum–metal junction (Figure 2.5). As a consequence, the band structure of the oxide film – in particular, the gap between the valence band (VB) and

2.3 STM Imaging of Oxide Films

(a) EVac φt

CB

φs

EF

EF

VB df

dg

(b)

CB

EF

EF

VB df

substrate

dg’

oxide

tip

Figure 2.5 Schematic representation of the gap of a metaloxide–vacuum–metal tunneling junction in the case of a low (a) and a high (b) sample bias voltage.

the conduction band (CB) – will play a crucial role in the tunneling process. If we disregard local effects – such as a local variation of the barrier height – and take only the average electronic structure of the film into account, we can distinguish two limiting cases. For small bias voltages, the oxide film will only weakly perturb the tunneling process. The current passes through the oxide film because no states for the tunneling electrons are available (Figure 2.5a). It will thus act much like an additional vacuum barrier. As a consequence, the distance from the tip to the sample will be the sum of the film thickness df and the distance between tip and film dg. This does not necessarily imply that the oxide film is invisible in STM. If the lateral structure of the oxide film is not homogeneous, the tunneling barrier will change as a function of the lateral position of the tip and consequently the z-position of the tip will also change, which will result in a contrast in the STM images. If, at small bias voltage, the tunneling current is increased, the tip will touch and eventually penetrate the film, leading to a direct overlap of the electronic states of the tip and the sample. This will cause variations in the tunneling current as a function of the lateral tip position, which can – in favorable cases – be used to image the surface of the oxide film with atomic resolution [30]. Since at moderate tunneling currents (<1 nA) the film behaves like an additional vacuum barrier, it is possible to image the substrate

j35

j 2 The Template Route to Nanostructured Model Catalysts

36

surface with atomic resolution through the oxide film with STM [31]. This can be very helpful for the analysis of the structural relation (commensurability) of oxide film and substrate. In the case of the high bias voltages – depending on the sign of the bias voltage – either the conduction band or the valence band of the oxide will participate in the tunneling process (Figure 2.5b), and it is expected that the topography of the film in the STM images is dramatically changed compared to the case of low bias voltages. At positive sample bias, the main contribution to the tunneling current stems from electrons tunneling into the conduction band of the oxide film, from which they are further transported into the substrate. If the film is sufficiently thin, that is, thinner than the mean free path of electrons (10 nm), a charging of the oxide film due to the electron current is not expected [32]. It should also be mentioned that, as shown in Figure 2.5, to a first approximation no bias voltage drop is encountered in the film if its thickness is only around 0.5 nm. The significant differences in the tunneling process at low and high bias voltage will result in a strong bias dependence of the apparent topography of the oxide film. Such a bias-dependent tunneling for thin Al2O3 islands on NiAl(1 1 0) was first observed by Bertrams et al. [33]. In this case, the islands were hardly visible at a sample bias voltage of þ 0.4 V, which corresponded to an energy in the bandgap of the alumina islands. At a bias of þ 4 V, however, the electrons tunnel into the conduction band of the Al2O3 islands and these are imaged with an apparent height of 0.3 nm. For the similar system of Al2O3 islands on Ni3Al(1 1 1), a systematic study of the apparent height of the islands as a function of the bias voltage allowed the estimation of the energetic position of the bottom of the Al2O3 conduction band [34]. As can be seen in Figure 2.6a, the measured apparent height of the Al2O3 islands rapidly increases in the bias voltage ranging from 2.1 to 3.0 V from a value of 0.25 to 0.6 nm, which is close to the actual thickness of 0.5 nm [35]. This increase is due to

Figure 2.6 (a) Variation in the apparent height of alumina islands on Ni3Al(1 1 1) as a function of sample bias voltage [34]. (b) STS dI/dV spectra taken at different points of the alumina film surface [36].

2.3 STM Imaging of Oxide Films

the lowered tunneling resistance caused by the participation of the conduction band of the oxide in the tunneling process. This situation is similar to the one shown in Figure 2.5b. The determination of the band edge using this method is, however, not very precise and based on the average properties (e.g., average height) of the islands. However, it provides a first indication of the variations in the apparent topography of alumina films as a function of the bias voltage encountered in STM. Furthermore, a combination of measurements like the one shown in Figure 2.6a with UPS valence band spectra of the respective oxide film provides a rough value of the bandgap of the film [34]. The bandgap of the oxide film can be determined using tunneling spectroscopy, which is shown in Figure 2.6b. These dI/dV spectra clearly show the top of the valence band at a negative sample bias of about 5 V and the bottom of conduction band at roughly þ 3 V [36] indicating a bandgap of approximately 8 eV. This variation in the electronic structure of the tunneling gap is reflected in the pronounced variation in the apparent topography of Al2O3 films on Ni3Al(1 1 1) as a function of bias voltage (Figure 2.7). Here, two effects are clearly visible: First, a contrast reversal between images (c) and (d) can be seen. The formerly hexagonal arrangement of bright dots (2.0 V) is imaged as holes at a bias voltage of 3.2 V. This important bias dependence of the apparent topography of insulating films has also been found in several other cases, for example, CaF2/Si(1 1 1) [32], CoO/ Ag(0 0 1) [37, 38], NiO/Ag(0 0 1) [39], and ZrO2/Pt(1 1 1) and can be attributed to the band structure and more precisely to the bandgap of the insulating films. Second, the apparent corrugation of the film changes drastically. For a low bias voltage of 0.5 V, a corrugation on the order of only 0.02 nm is found, which roughly corresponds to the corrugation found on metal surfaces, whereas at higher bias a corrugation of up to 0.2 nm is visible (Figure 2.8). It is obvious that the sharp increase in the measured corrugation takes place in the same energy interval where the increase in the average island height has been found for this type of oxide film (Figure 2.6a). It must, therefore, be correlated with the electronic structure of the film in the region close to the bottom of the conduction band. Again a closer look at the STS spectra shown in Figure 2.9 reveals that the dots are, indeed, closely related to the electronic structure of the film. They appear in an energy range between 2.0 and 2.8 V, where a localized electronic state in the bandgap can be found at the corners of the unit cell (points labeled “A”). This state, which is found

Figure 2.7 STM images of an alumina film on Ni3Al(1 1 1) at different sample bias voltages. (a) Ub ¼ 0.5 V, (b) 1.5 V, (c) 2.0 V, (d) 3.2 V, and (e) 4.2 V. The unit cell of the dot structure is marked in the images.

j37

j 2 The Template Route to Nanostructured Model Catalysts

38

0.4 UT = +0.5 V UT = +1.5 V U T = +2.0 V U T = +3.2 V

height [nm]

0.3

0.2

0.1

0.0 0

2

4

6 distance [nm]

8

10

Figure 2.8 Line profiles of the alumina film on Ni3Al(1 1 1) for different sample bias voltages [36].

Figure 2.9 (a) STM image of Al2O3/Ni3Al(1 1 1). The unit cell of the dot structure is shown as dashed line. (b) STS spectra taken at the points indicated in the STM image. The feedback loop was opened at a bias voltage of þ 3.5 and þ 3.0 V (inset of b).

2.4 STM Imaging of Metal Particles on Oxide Films

only at the corners of the unit cell, increases the tunneling probability at these locations in the respective bias voltage range resulting in the observed contrast (see Figure 2.7). STS spectra taken at the points labeled “B” do not show this state. It is, thus, clearly localized at the corners of the unit cell. This particular example shows that special care has to be taken when the topography of thin oxide layers on metallic substrates is measured by STM. Electronic effects at or near the band edges can dramatically influence the tunneling probabilities and, thus, lead to rather large changes in the measured corrugation. Obviously, this phenomenon can also influence the imaging of clusters on oxide films, since corrugation changes of about 0.2 nm, which have been encountered here, are on the order of the interlayer distance in metals and can considerably influence the cluster height determination. However, other effects may also be encountered when imaging clusters on oxide films, which will be discussed in the following section.

2.4 STM Imaging of Metal Particles on Oxide Films

As we have seen in the previous chapter, the apparent topography and corrugation of thin oxide films as imaged by STM may vary drastically as a function of the sample bias. This will of course play an important role in the determination of cluster sizes with STM, which will be discussed in the following section. The determination of the size of the metallic nanoparticles on oxide films is a crucial issue in the investigation of model catalysts since the reactivity of the particles may be closely related to their size. Therefore, the investigation of reactions on model catalysts calls for a precise determination of the particle size. If the sizes of the metal particles on an oxidic support are measured by STM, two different effects, which distort the size measurement, have to be taken into account. First, convolution effects between the tip and the cluster, both in general of comparable size, must be considered. This has already been pointed out by Reis et al. [40] in 1989 in the context of roughness measurement [40] and Barbet et al. in 1993 for imaging colloidal gold beads [41] using STM. The situation one encounters for small metal particles on an oxide film is depicted in Figure 2.10. If we disregard electronic effects, which will be discussed below, the trajectory of an STM tip above a cluster will follow the dotted path depicted in Figure 2.10. Two consequences  become immediately obvious: (i) The apparent cluster height h , which can be derived, should be identical to the actual cluster height h. Thus, from this point of view the cluster height can be correctly measured by STM. Since the height is in general an integer multiple of a step height of the corresponding cluster material, it is rather easy to judge on the number of atomic layers, which constitute the cluster.  (ii) The measured cluster diameter d will be larger than the actual cluster diameter d.  Moreover, the actual value of d will depend not only on the cluster diameter but also on the morphology and radius of the tip. Even if we define the diameter of the cluster at the FWHM in a line scan of the particle (d0 ), the measurement will still not reflect

j39

j 2 The Template Route to Nanostructured Model Catalysts

40

Figure 2.10 Trajectory of an STM tip above an oxide-supported metal particle.

the actual diameter of the cluster. To account for this, a detailed analysis of the tip shape is necessary, as it has been described in detail by Bowker et al. [5] as well as Klyachko and Chen [42], for the case of C60 molecules on Ge(1 0 0) and Si(1 0 0). But even if we were able to overcome these geometric distortions by estimating the tip shape and explicitly including it in the particle size calculations, a second and more important problem will come into play. The complex electronic structure of the tunneling contact in these systems, which is schematically depicted in Figure 2.11, renders the tunneling process rather difficult. We are now facing a junction of metal–oxide–metal–vacuum–metal type and a number of effects have to be considered. First, the electronic structure of the clusters will play an important role. Here, we have to consider again two cases: very small particles, which still have – much like molecules – discrete electronic states [43], and larger particles, which possess a band structure and can be regarded as metals. We will confine our discussion to the latter case since most of the arguments found there are also applicable to the former. For a metallic particle, the tunneling current

EVac CB

φs

φt

φc

EF

EF

VB df

substrate

oxide

dc

cluster

dg

tip

Figure 2.11 Schematic representation of the tunneling gap for an oxide-supported metal particle.

2.4 STM Imaging of Metal Particles on Oxide Films

from the tip can pass into the particle at all bias voltages much as it is the case on extended metal surfaces because of the continuous band structure of the metal. As a consequence, the tunneling resistivity will be to a first approximation related to the electronic density of states of the cluster. However, once the tunneling electron has passed the cluster, the further transport of the electron through the oxide film to the substrate strongly depends on the band structure of the oxide film. If the bias voltage is large, then the energy of the electron corresponds to that of the conduction band of the oxide and tunneling is facilitated. If, however, the bias voltage is low, the oxide film will act as an additional tunneling barrier (see also Figure 2.5) and the electron may not as easily relax into the substrate. In the latter case, charging effects of the particle may be encountered, which will in turn influence the tunneling current and thus the apparent topography of the particle. This will, however, depend on the thickness of the oxide film and thus the second tunneling gap. We expect therefore that much like the apparent topography of the oxide film, the apparent height and size of clusters on oxide films will show strong bias dependence. This is shown in Figure 2.12 for a single Fe cluster on Al2O3/Ni3Al(1 1 1). In Figure 2.12a, which was taken at a bias of 3.2 V, that is, in the range of the oxide conduction band, the cluster appears rather large and the surrounding oxide film is also imaged with large corrugation. At small bias voltage (Figure 2.12b), however, the cluster appears to be much smaller and the oxide film is not visible any more. The hexagonal pattern surrounding the cluster is due to the atomic structure of the Ni3Al(1 1 1) substrate, which can be resolved in this case through the oxide film. By comparing these images, the whole problem of particle size measurements on such complex systems becomes very apparent. To systematically investigate this phenomenon, a collection of 12 Fe clusters on Al2O3/Ni3Al(1 1 1) has been imaged by STM in the bias voltage range from 0 to 5 V. In Figure 2.13, it is clearly visible that not only the apparent structure of the oxide film undergoes a strong bias dependence, as it has already been shown in the previous section, but also the appearance of the clusters changes dramatically as a function of bias voltage. A thorough analysis of the cluster heights and diameters as a function of the bias voltages leads to the curves presented in Figure 2.14. It is visible at first glance

Figure 2.12 STM images of a single Fe cluster on Al2O3/ Ni3Al(1 1 1): (a) 3.2 V and (b) 0.19 V, 6.2 nm  6.2 nm [44].

j41

j 2 The Template Route to Nanostructured Model Catalysts

42

Figure 2.13 STM images of Fe clusters on Al2O3/Ni3Al(1 1 1) taken for different sample bias voltages. The image size is 48 nm  48 nm. The labeled particles were used for the evaluation of the particle height and diameter [44].

2.4 STM Imaging of Metal Particles on Oxide Films

Figure 2.14 Apparent cluster height (a) and diameter (b) as a function of bias voltage. Data were obtained from the Fe clusters shown in Figure 2.13 [44].

that the variations of cluster height and size are significant in the investigated range of bias voltages. There are, however, three distinct regions, which show a particular behavior. The range from 0 to 1 V is characterized only by a slight variation in the cluster height and diameter. The same is true for the bias range above 3 V. However, in between (1–3 V), an increase in the apparent height and diameter is encountered. To explain this, we turn back to Figure 2.11. Region I is clearly in the bandgap of the oxide so that alumina does not contribute to the tunneling process. Region III is related to the conduction band of the oxide so that the oxide is taking part in the tunneling process, but the contribution of the oxide can be regarded as constant to a first approximation. This implies that in these two regions the appearance of the clusters is largely determined by the electronic structure of the cluster itself. Apparently, this structure is not a strong function of the bias voltage in the present case. In region II, however, we find a steep increase in both values. This must be due to the electronic structure of the tunneling gap. To explain this variation, it is necessary to have a detailed look at electronic states, which participate in the tunneling process. Here, contributions of both the cluster and the oxide film have to be considered. First of all we note that above 2 V a contrast reversal and an increase in the measured corrugation in the STM images of the pure oxide film was found, which has been discussed above. This is correlated with the appearance of the dots in the STM images exactly at the positions where the clusters are located in the present case. This contrast reversal, which is due to the electronic structure of the oxide film, can therefore explain the increased values of the cluster height and diameter in the range from 2 to 3 V. For bias voltages between 1 and 2 V, however, the increase in cluster height and diameter cannot be attributed to the alumina band structure. Thus, it must be related to the band structure of the Fe clusters themselves. Theoretical data from literature can shed some light on the question whether or not the electronic structure of the clusters is responsible for this sudden increase in Figure 2.14. The calculated density of states for an Fe cluster containing 7 Fe atoms shows an intense empty sp-state that is localized in the range from 1 to 2 eV [45]. This state will facilitate tunneling into the cluster and, thus, lead to a decrease in the tunneling resistance. This in turn will, under constant current

j43

j 2 The Template Route to Nanostructured Model Catalysts

44

conditions, lead to a retraction of the STM tip, which results in a larger diameter and height in the STM images. A very similar electronic state is found in the same energy range for bulk iron [46] proving that the presence of this state does not dramatically change with cluster size. We can therefore explain the apparent changes in the STM images in region II by a combination of electronic effects stemming from the Fe clusters (1–2 V) and the alumina film (2–3 V). The observed changes in cluster size and diameter can clearly be related to the complex electronic structure of the tunneling gap. The real cluster height in the present case is most likely represented by the limiting values of the bias voltage around 0 and 5 V and corresponds roughly to 0.2–0.3 nm (see Figure 2.14), thus to a single atomic layer. As a result, we want to stress that extreme care has to be exercised when using STM as a tool for determining the size of oxide-supported metal clusters. Reliable results can be expected only if the electronic structure of the system is well known and bias voltage ranges can be avoided where localized electronic states determine the tunneling process. It seems that the most reliable results can be obtained at rather high bias voltage corresponding to the conduction band of the oxide film and that bias voltages in the vicinity of the lower conduction band edge of the oxide film have to be avoided.

2.5 Template-Controlled Growth of Model Catalysts 2.5.1 Oxides as Templates

In the previous chapter, we have explored the application of STM for the characterization of thin oxide films and supported clusters. We now focus our attention on the template-controlled growth of metal particles on oxide films. As we have already mentioned in Section 2.1, lithographic techniques can be employed to create a regular array of nucleation sites for cluster growth but only at the expense of a limited lateral resolution. This does not limit the minimum size of the particles, which depends only on the amount of metal that is deposited per nucleation site. However, the resolution limitation is important for the number density of particles on the support that can be reached. If the unit all size of the cluster array is, for example, 100 nm  100 nm, a number density of only ns ¼ 1010 cm2 is obtained, which is approximately five orders of magnitude lower than the number density of atoms in a metallic plane. This number density is much lower than the typical sensitivity of standard surface science techniques, which is on the order of 1% of a monolayer (1013 cm2). Consequently, such techniques cannot be applied. It may, however, be large enough to conduct kinetic measurements of the catalytic reactions on such a surface at higher pressures. Since our prime goal is – as stated in the introduction – to study the catalyst surface with atomic precision and correlate this with the reactivity, we have to conceive systems of much higher island density. The nucleation and growth of metal particles on random defects on bulk oxides [47] and oxide films [4] can

2.5 Template-Controlled Growth of Model Catalysts

Figure 2.15 High-resolution STM image (a) [30] and AFM image (b) [49] of the alumina film on Ni3Al(1 1 1). The high-symmetry sites marked by triangles (circles) and the hexagons correspond to the network and dot structure, respectively.

provide such large number density of clusters, however, at the expense of increased complexity. This can be overcome only by using ordered nanoscopic reconstructions of oxide surfaces and films as growth templates, which provide a periodicity of a few nanometers at most. Fortunately, an increasing number of such surfaces have been found over the years and promising results in terms of metal particle growth have been achieved. One of the best examples of template-controlled growth is the thin alumina film on Ni3Al(1 1 1), which has already been discussed in Section 2.3. This film is characterized by a hexagonal superstructure of 4.16 nm lattice constant, which can be related to the presence of an electronic state in the bandgap of the oxide (Figure 2.9). High-resolution STM images of this film by Schmid et al. suggest that this structure is caused by oxygen vacancies (holes) in the film (Figure 2.15) [30]. Previous AFM measurements did not show these holes [48, 49] so that the origin of the superstructure is still under debate. However, the STM measurements proved, in agreement with the AFM measurements, that the corners of the unit cell of the super lattice formed by the films are high-symmetry points and these points can be related to the so-called dot structure (bright dots in Figure 2.7c) [50]. In addition, a second type of high-symmetry points was found on the film with a periodicity of 2.4 nm, which has previously been referred to as network structure (dark holes in Figures 7d and 15) [50]. Interestingly, both structures can be employed as templates for the growth of metallic nanostructures. These high-symmetry points are obviously high binding energy sites for metals on the alumina film due to high coordination and/or specific electronic properties. The growth of a large variety of metals on the alumina film on Ni3Al(1 1 1) has been studied. Initial studies were conducted for the coinage metals Cu, Ag, and Au [51]. For these metals, a rather strong influence of the dot structure on the nucleation and growth of clusters was observed at low coverage. As can be seen in Figure 2.16a, the few Ag clusters that are visible are located on the dot structure, which is also imaged as protrusions. However, as the coverage is increased, the template control does not

j45

j 2 The Template Route to Nanostructured Model Catalysts

46

Figure 2.16 STM images of Ag clusters (a, b) and Cu clusters on Al2O3/Ni3Al(1 1 1) (c). The image sizes and bias voltages are (a) 60 nm  60 nm, 2.1 V; (b) 80 nm  80 nm, 0.7 V; and (c) 65 nm  65 nm, 0.7 V. (Reproduced with permission from Ref. [51].)

work to that perfection any more. Au (not shown here) tends to grow to granular films [51] and Ag shows large agglomerates (Figure 2.16b). Only Cu seems to keep a rather good order showing signs of a regular arrangement with a lattice constant corresponding to that of the dot structure (Figure 2.16c). In these cases, the growth temperature was 300 K, and it can be expected that different temperatures, which lead to different growth kinetics, could significantly change the order. Furthermore, it seems that a stronger interaction of the metal with the substrate, evidenced, for example, by a higher energy of adhesion [53], leads to a better order of the film at high coverage. This assumption was in part verified by the growth of Pd clusters on the alumina film on Ni3Al(1 1 1). In this case, the dot structure of the alumina film acts as an efficient template for the growth at 300 K (Figure 2.17). A nearly perfect array of Pd clusters can, thus, be obtained by deposition of Pd vapor onto the Al2O3/Ni3Al(1 1 1). Moreover, the Pd clusters show a narrow size distribution, which suggests that this surface is close to the ideal model catalyst of low complexity we are striving for. Unfortunately, for the reasons given in Section 2.4, it is not easy to determine the actual cluster size by STM in this case. It can also be seen in Figure 2.17 that the growth kinetics, as explained in this chapter, plays an important role for the order. By raising the substrate temperature during growth, the order of the Pd clusters is gradually lost and the

Figure 2.17 STM images of Pd clusters grown on Al2O3/Ni3Al(1 1 1) at different substrate temperatures. The image size is 100 nm  100 nm and the bias voltage is 0.7 V. (Reproduced with permission from Ref. [54].)

2.5 Template-Controlled Growth of Model Catalysts

0.05

1 2 3 4

-2

number density [nm ]

0.04

1

1

1

1

1

2

0.03

layer clusters layer clusters layer clusters layer clusters total

2 2

1 0.02

2

2

2

2

2

1

1

2

0.01

0.00 0

3 4 50

1 3

3 3

2 3 4

1

3 4 100

3 4

3 4

3 4

150

4 200

3 4 4 250

4 300

deposition time [s]

Figure 2.18 Plot of the number density of clusters of different height as a function of deposition time. The line is a fit to the nucleation equation given in the text [54].

number density of the Pd islands decreases as it is expected for heterogeneous nucleation on the template sites (compare Figure 2.4). The growth of the clusters at 300 K is indeed best described by heterogeneous nucleation and, thus, templatecontrolled, as can be rationalized by looking at Figure 2.18. Here, the number density of islands n(t) is plotted as a function of the deposition time t (deposited Pd amount). The experimental points are nicely described by the curve for heterogeneous nucleation kinetics n(t) ¼ ns tanh(t/t), which was introduced by Robinson and Robins [52]. In this equation, t is a typical timescale, which accounts for the influence of substrate temperature and flux of the impinging atoms, and ns is the number density of the nucleation sites. From Figure 2.18, it can also be seen that initially only one- and two-layer islands are found. Only after the saturation number density is reached – that is, all template sites are covered – do these islands start to increase in height. This indicates that the interaction of the template with the Pd is very specific and leads to the almost perfect nucleation behavior. As can be seen from Figure 2.17, the favorable temperature range for perfect heterogeneous nucleation does not extend far beyond 300 K for Pd. Already at 400 K substrate temperature, some of the template control is lost and the island size starts to become inhomogeneous. A further increase in the growth temperature leads to an additional loss of order and a reduction in the number density of the Pd clusters. Finally, at 600 K it can be seen that a number of the smaller Pd clusters are still positioned on the dot structure (template sites) and that the morphology is dominated by the presence of large flat islands. The former observation, again, indicates that for this particular system the interaction between the nucleation sites and the Pd clusters is relatively strong.

j47

j 2 The Template Route to Nanostructured Model Catalysts

48

Figure 2.19 STM images of V clusters grown on Al2O3/Ni3Al(1 1 1) at different substrate temperatures. The image size is 80 nm  80 nm and bias voltage is 0.7 V. (Reproduced with permission from Ref. [51].)

As was mentioned before, not only the dot structure corresponds to high-symmetry sites of the oxide film but also the network structure may serve as a template for nanostructured growth for a different choice of metal. Indeed, it has been found that metals, which show a stronger interaction with the alumina film than the metals discussed above, possess a stronger tendency to nucleate and grow on the sites of the network structure. This has been observed for Mn [51, 55] and Fe [56], but neither of these metals shows a regular arrangement of the clusters at higher coverage much like the case of Ag and Au on the dot structure. By contrast, for Va nearly perfect array of clusters has been found for a deposition temperature of 550 K. The lattice constant of this array is 2.4 nm, which corresponds to the periodicity of the network structure and, thus, indicates that the network structure can also be used as a growth template. As can be seen in Figure 2.19, even at 300 K the order of the V cluster is already very good; however, their size distribution is not as narrow as for a deposition temperature of 550 K. The results for Pd and V lead to the conclusion that the high-symmetry sites of the alumina film on Ni3Al(1 1 1) can act as template for the growth of nanostructured model catalysts. They also prove that kinetic control of the growth is of utmost importance. The two cases of Pd and V show the enormous potential of the Al2O3/Ni3Al(1 1 1) system for the template-controlled growth of model catalysts. However, a number of other nanostructured oxide surfaces have also been found to act as growth templates. Among these examples, we find the TiO2(1 1 0)-(1  2) cross-linked reconstruction, which provides a rectangular arrangement of nucleation sites. On this surface, a nanostructured growth of Pd clusters is obtained by deposition at 373 K (Figure 2.21) [57–59]. As in the case of Al2O3/Ni3Al(1 1 1), the deposition of Au does not lead to ordered cluster arrays on the TiO2(1 1 0)-(1  2) cross-linked surface [60]. Gold nanoparticles can, however, be grown in ordered arrays on the so-called z0 -TiO2 films on Pt(1 1 1) (Figure 2.20). Again, the template-controlled growth leads to nicely ordered arrays of Au clusters. In this case, a particular intriguing phenomenon is encountered. Postannealing of these films leads to a change in the symmetry of the

2.5 Template-Controlled Growth of Model Catalysts

Figure 2.20 (a) Au on z0 -TiO2 and (b) Au on w-TiO2 after annealing at 600 K [61]. The image sizes are (a) 100 nm  100 nm and (b) 110 nm  80 nm.

system, which is interpreted in terms of a transformation of the structure of the TiO2 film from the zigzag-like z0 -phase to the hexagonal w-phase (wagon wheel phase) [61]. Iron oxide surfaces can also be used for template-controlled growth as has been demonstrated in the case of the Fe3O4(1 1 1) and FeO films on Pt(1 1 1), which have a common reconstructed close-packed oxygen surface layer with long-range order. On Fe3O4 films on MgO(1 0 0), nicely ordered arrays of Fe clusters (Figure 2.22) have been grown, and similar results were obtained for Fe on FeO/Pt(1 1 1) [62]. Furthermore, the template-controlled nucleation of Cr on Fe3O4(1 1 1) [62] and V on FeO/Pt(1 1 1) [62] was demonstrated.

Figure 2.21 STM image of Pd nanoparticles on the cross-linked (1  2) TiO2(1 1 0) surface [57, 59]. The image size is 100 nm  100 nm.

j49

j 2 The Template Route to Nanostructured Model Catalysts

50

Figure 2.22 STM of Fe nanoparticles on an Fe3O4 film on MgO(1 0 0) [62].

Most of these examples show that it is apparently easier to grow ordered arrays of more reactive metals, for example, Pd, than unreactive metals such as Au and Ag. This is most likely due to the low interaction energy of the latter metals with the oxide surfaces and, consequently, a lower specificity of the traps. This calls for a modification of the template to increase the interaction energy of the template sites (traps) with the deposited metals. 2.5.2 Modified Templates

As discussed in the previous section, the huge potential of alumina films on Ni3Al(1 1 1) for the growth of model catalysts has been exploited. However, only in the case of Pd and V reasonably well-ordered nanostructures have been found. For the other metals that were investigated, either the growth conditions chosen were not such that nanostructuring was accomplished or the interaction between the metal and the nucleation sites was not sufficient. To overcome these limitations, an approach can be used that relies on the modification of the alumina film on Ni3Al(1 1 1) in such a way that the interaction of the deposited metal and the nucleation sites is increased. This can be done by taking advantage of the perfect nucleation behavior of Pd on this film: The deposition of a small amount of Pd, which nucleates on the dot structure, will create an array of small Pd clusters. We will call this novel surface (Pd/Al2O3/Ni3Al(1 1 1)) a second-generation template since it was already produced using a template. On this modified template, ordered structures of metals could, indeed, be grown that cannot be produced by nucleation of the metal on the bare alumina film. So far, two cases have been found that illustrate the potential of this new second-generation template. Hamm et al.

2.6 Conclusions

Figure 2.23 STM images of bimetallic AuPd [63] (a) and FePd [30] (b) clusters grown on the Pd/Al2O3/Ni3Al(1 1 1) secondgeneration template. The image sizes are (a) 100 nm  100 nm and (b) 80 nm  69 nm.

reported on the room-temperature growth of Au clusters [63] and Schmid et al. on the growth of Fe clusters [30] on Pd/Al2O3/Ni3Al(1 1 1). In both cases, nicely ordered cluster arrays have been observed on the dot structure (Figure 2.23). One should, however, bear in mind that by this route only bimetallic cluster arrays can be grown but this does not necessarily represent a big limitation since bimetallic surfaces are often interesting model catalysts. It should also be mentioned that Au and Fe show a different nucleation behavior on the bare alumina film. Only the action of the second-generation template leads to the similar nucleation behavior observed. The situation that is encountered in these two systems is closely related to the reasoning of Venables [23] who claimed that in order to overcome coarsening a high chemical specificity of the nucleation sites and low diffusion lengths are required. The former is apparently provided by the modification of the original template with Pd atoms, which increases the interaction strength with the second metal. This suggests that the second-generation template will most probably work for the nucleation of many other metals as well. This has to be exploited in the near future and will pave the way for a large variety of nanostructured bimetallic model catalysts.

2.6 Conclusions

Oxide surfaces, and in particular oxide films, are versatile substrates for the preparation of model catalysts. Quite a few of these systems show nanoscale reconstructions, which can be employed as templates for the growth of ordered model catalysts of reduced complexity. In order to efficiently control the growth of nanostructured metal particle arrays, two conditions have to be met. First, the template must provide sites of high interaction energy that trap the deposited metals. Second, the kinetics of the growth process must be carefully controlled by choosing

j51

j 2 The Template Route to Nanostructured Model Catalysts

52

the right substrate temperature and flux of impinging atoms during deposition. While the latter point is in general easy to control by the experimentalist, the former is essentially a function of the materials employed. This limitation can, however, be overcome by modifying the template such that the interaction energy of the traps is increased, thereby creating a second-generation template.

References 1 Vang, R.T., Lægsgaard, E., and Besenbacher, F. (2007) Phys. Chem. Chem. Phys., 9, 3460. 2 Somorjai, G.A., York, R.L., Butcher, D., and Park, J.Y. (2007) Phys. Chem. Chem. Phys., 9, 3500. 3 B€aumer, M., Libuda, J., Neyman, K.M., R€ osch, N., Rupprechter, G., and Freund, H.-J. (2007) Phys. Chem. Chem. Phys., 9, 3541. 4 B€aumer, M. and Freund, H.-J. (1999) Prog. Surf. Sci., 61, 127. 5 Bowker, M., Bowker, L., Bennett, R., Stone, P., and Ramirez-Cuesta, A. (2000) J. Mol. Catal. A: Chem., 163, 221. 6 Rousset, S., Croset, B., Girard, Y., Pr
evot, G., Repain, V., and Rohart, S. (2005) C. R. Physique, 6, 33. 7 Jacobs, P.W., Ribero, F.H., Somorjai, G.A., and Wind, S.J. (1996) Catal. Lett., 37, 131. 8 Hla, S.-W. (2005) J. Vac. Sci. Technol. B, 23, 1351. 9 Dutta, J. and Hofmann, H. (2003) Encyclopedia of Nanoscience and Nanotechnology, vol. X (ed. H.S. Nalwa), American Scientific Publishers, pp. 1–23. 10 F€ orster, S. (2003) Top. Curr. Chem., 226, 1. 11 Kung, M.C. and Kung, H.H. (2005) Top. Catal., 34, 77. 12 Gehl, B., Fr€omsdorf, A., Aleksandrovic, V., Schmidt, T., Pretorius, A., Flege, J.-I., Bernstorff, S., Rosenauer, A., Falta, J., Weller, H., and B€aumer, M. (2008) Adv. Funct. Mater., 18, 2398. 13 Brune, H., R€oder, H., Boragno, C., and Kern, K. (1994) Phys. Rev. B, 49, 2997. 14 Gestermann, S., Nohlen, M., Schmidt, M., and Wandelt, K. (1997) Surf. Rev. Lett., 6, 1179.

15 A€ıt-Mansour, K., Ruffieux, P., Xiao, W., Gr€oning, P., Fasel, R., and Gr€oning, O. (2006) Phys. Rev. B, 74, 195418. 16 G€ unther, C., Vrijmoeth, J., Hwang, R.Q., and Behm, R.J. (1995) Phys. Rev. Lett., 74, 754. 17 Meunier, I., Tr
eglia, G., Legrand, B., T
etot, R., Aufray, B., and Gay, J.M. (2000) Appl. Surf. Sci., 162, 219. 18 Bendounan, A., Cercellier, H., FagotRevurat, Y., Kierren, B., Yu Yurov, V., and Malterre, D. (2003) Appl. Surf. Sci., 212, 33. 19 Bendounan, A., Cercellier, H., FagotRevurat, Y., Kierren, B., Yu Yurov, V., and Malterre, D. (2003) Phys. Rev. B, 67, 165412. 20 Brune, H. and Kern, K. (1997) Growth and Properties of Ultrathin Epitaxial Layers (eds D.A. King and D.P. Wopdruff), Elsevier, pp. 149–206. 21 Wang, C. and Bai, C. (2006) Applied Scanning Probe Methods IV (eds B. Bhushan and H. Fuchs), Springer, pp. 159–181. 22 Venables, J.A., Giordano, L., and Harding, J.H. (2006) J. Phys.: Condens. Matter, 18, S411. 23 Venables, J.A. (1997) Physica A, 239, 35. 24 Pr
evot, G., Guesmi, H., and Croset, B. (2007) Surf. Sci., 601, 2017. 25 Ellmer, H., Repain, V., Sotto, M., and Rousset, S. (2002) Surf. Sci., 511, 183. 26 Sabiryanov, R.F., Larsson, M.I., Cho, K.J., Nix, W.D., and Clemens, B.M. (2003) Phys. Rev. B, 67, 125412. 27 Bromann, K., Giovannini, M., Brune, H., and Kern, K. (1999) Eur. Phys. J. D, 9, 25. 28 Besenbacher, F. (1996) Rep. Prog. Phys., 59, 1737.

References 29 Tersoff, J. and Hamann, D.R. (1983) Phys. Rev. Lett., 50, 1998. 30 Schmid, M., Kresse, G., Buchsbaum, A., Napetschnig, E., Gritschneder, S., Reichling, M., and Varga, P. (2007) Phys. Rev. Lett., 99, 196104. 31 Degen, S., Krupski, A., Kralj, M., Langner, A., Becker, C., Sokolowski, M., and Wandelt, K. (2005) Surf. Sci., 576, L57. 32 Viernow, J., Petrovykh, D.Y., Kirakosian, A., Lin, J.-L., Men, F.K., Henzler, M., and Himpsel, F.J. (1999) Phys. Rev. B, 59, 10356. 33 Bertrams, T., Brodde, A., and Neddermeyer, H. (1994) J. Vac. Sci. Technol. B, 12, 2122. 34 Rosenhahn, A., Schneider, J., Kandler, J., Becker, C., and Wandelt, K. (1999) Surf. Sci., 433–435, 705. 35 Bardi, U. and Rovida, G. (1990) Surf. Sci., 239, L511. 36 Maroutian, T., Degen, S., Becker, C., Wandelt, K., and Berndt, R. (2003) Phys. Rev. B, 68, 154414. 37 Sebastian, I. and Neddermeyer, H. (2000) Surf. Sci., 454, 771. 38 Hagendorf, C., Shantyr, R., Meinel, K., Schindler, K.-M., and Neddermeyer, H. (2003) Surf. Sci., 532, 346. 39 Sebastian, I., Bertrams, T., Meinel, K., and Neddermeyer, H. (1999) Faraday Discuss., 114, 129. 40 Reiss, G., Vancea, J., Wiltmann, H., Zweck, J., and Hoffmann, H. (1989) J. Appl. Phys., 67, 1156. 41 Barbet, J., Garvin, A., Thimonier, J., Chauvin, J.-P., and Rocca-Serra, J. (1993) Ultramicroscopy, 50, 355. 42 Klyachko, D. and Chen, D.M. (2000) Surf. Sci., 446, 98. 43 Nilius, N., Kulawik, M., Rust, H.-P., and Freund, H.-J. (2004) Surf. Sci., 572, 347. 44 Degen, S. (2005) Ph.D. thesis, Universit€at Bonn. 45 Zein, N.E. and Antropov, V.P. (2002) Phys. Rev. Lett., 89, 126402. 46 Castro, M. (1997) Int. J. Quant. Chem., 64, 223.

47 Henry, C.R. (1998) Surf. Sci. Rep., 31, 231. 48 Hamm, G., Barth, C., Becker, C., Wandelt, K., and Henry, C.R. (2006) Phys. Rev. Lett., 97, 126106. 49 Gritschneder, S., Becker, C., Wandelt, K., and Reichling, M. (2007) J. Am. Chem. Soc., 129, 492. 50 Wiltner, A., Rosenhahn, A., Becker, C., Pervan, P., Milun, M., Kralj, M., and Wandelt, K. (2001) Thin Solid Films, 400, 71. 51 Becker, C., Rosenhahn, A., Wiltner, A., von Bergmann, K., Schneider, J., Pervan, P., Ilun, M., Kralj, M., and Wandelt, K. (2002) New J. Phys., 4, 75.1. 52 Robinson, V.N.E. and Robins, J.L. (1974) Thin Solid Films, 20, 155. 53 Campbell, C.T. (1997) Surf. Sci. Rep., 27, 1. 54 Degen, S., Becker, C., and Wandelt, K. (2004) J. Chem. Soc., Faraday Discuss., 125, 343. 55 Becker, C., von Bergmann, K., Rosenhahn, A., and Wandelt, K. (2001) Surf. Sci., 486, L443. 56 Lehnert, A., Krupski, A., Degen, S., Franke, K., Decker, R., Rusponi, S., Kralj, M., Becker, C., Brune, H., and Wandelt, K. (2006) Surf. Sci., 600, 1804. 57 Bennett, R.A., Newton, M.A., Smith, R.D., Evans, J., and Bowker, M. (2002) Mater. Sci. Technol., 18, 710. 58 Bennett, R.A., Tarr, D.M., and Mulheran, P.A. (2003) J. Phys: Condens. Matter, 15, S3139. 59 Bowker, M. (2007) Phys. Chem. Chem. Phys., 9, 3514. 60 Maeda, Y., Fuitani, T., Tsubota, S., and Haruta, M. (2004) Surf. Sci., 562, 1. 61 Sedona, F., Agnoli, S., Fanetti, M., Kholomanov, I., Cavaliere, E., Gavioli, L., and Granozzi, G. (2007) J. Phys. Chem. C, 111, 8024. 62 Berdunov, N., Mariotto, G., Balakrishnan, K., Murphy, S., and Shvets, I.V. (2006) Surf. Sci., 600, L287. 63 Hamm, G., Becker, C., and Henry, C.R. (2006) Nanotechnology, 17, 1943.

j53

j55

3 In Situ STM Studies of Model Catalysts Fan Yang and D. Wayne Goodman

3.1 Introduction

The surface science approach to studying heterogeneous catalysis dates back to the pioneering work of Langmuir [1] in the 1910s that addressed the adsorption of gases on catalyst surfaces. Since then surface science studies of catalytic processes have played a central role in our understanding of catalysis and have aided in the design and improvement of catalysts for energy and environmental uses. The goal of surface science investigations has been to provide structural and spectroscopic information of catalyst surfaces at the spatial and temporal limit. Scanning tunneling microscopy (STM), with the capacity to reach the spatial limit at the atomic level, has ignited considerable interest since its discovery and has become a widely used tool in catalytic science. By following a selected area or a molecule at a model catalyst surface, in situ STM can provide temporal measurements regarding the elementary steps of catalytic transformations. The capabilities of in situ STM allow one to follow the dynamic change of surface species and identify what Taylor described as the “active site” in catalytic reactions. Such studies also provide kinetic measurements at the atomic scale, enabling the most precise modeling of macroscopic reactions. With the development of STM techniques, a large body of in situ STM work has emerged in the past decade regarding catalytic processes such as adsorption and diffusion, surface reaction, and catalyst deactivation. These experiments provide invaluable insights into the fundamental issues of catalysis. The timescale of catalytically important processes ranges from 10
12 to 104 s, with the chemisorptions and reactions taking place within picoseconds whereas catalyst deactivation occurs in seconds or minutes. Although the timescale of the fundamental process of adsorption and reaction is beyond the time resolution of STM, information about the pathway and energetics of adsorption/reaction can be acquired with STM by monitoring the change in the spatial distribution of surface adsorbates. Indeed, in situ STM can be combined with femtolasers to probe surface processes at both the spatial and the temporal limits. The feasibility of combining these two

j 3 In Situ STM Studies of Model Catalysts

56

techniques has been demonstrated in a recent study by Bartels et al. [2]. Such studies are extremely promising with respect to our understanding of surface chemical processes. In this chapter, we show how in situ STM has helped to visualize elementary steps of chemical reaction and to elucidate mechanisms of catalyzed processes. Like most surface science techniques, conventional in situ STM studies have been carried out in UHV on model catalysts consisting of extended planar surfaces. When extrapolating the information obtained in UHV surface science studies to real-world catalysis, two issues have generally concerned the catalysis community, namely, the pressure and material gaps. The pressure gap refers to the fact that surface science studies are conducted under UHVconditions (10
10–10
14 bar), whereas industrial catalytic reactions typically are carried out at high pressures (1–1000 bar). Over 10 orders of magnitude difference in the pressure of reactant gases can drastically change the interaction of reactants on the catalyst surface, a process essential to a catalytic reaction. The material gap refers to the gap between the surface structure of metal single crystals often studied in surface science and that of technical catalysts. “Real-world” catalysts usually consist of small metal clusters ranging from 1 to 100 nm in size, finely dispersed onto a highsurface-area oxide support. These metal clusters can have structures and properties that are quite different from the bulk metal. In catalytic research, it is well documented that the reactivity and selectivity of catalysts often depend on the size and shape of supported metal clusters [3]. Furthermore, the presence of the oxide support can modify the structure and properties of supported metal clusters. The effect of cluster size and metal support interaction cannot be addressed in surface science studies on well-defined single crystal metal surfaces. As a local structural probe, STM has the advantage of addressing these two issues. Although the operational range of STM extends from UHV to high pressures, there are challenges in maintaining the stability of STM at catalytically realistic operating temperatures and pressures. Nevertheless, it is possible to apply STM to study catalytic reactions under realistic conditions. To bridge the material gap, supported model catalysts, consisting of small metal clusters supported on planar oxide surfaces, have been introduced into surface science studies. STM is ideally suited to precisely characterize the structure of these supported model catalysts. In this chapter, we show the recent progress in bridging the pressure and material gaps by applying in situ STM to the study of model catalysts under realistic reaction conditions.

3.2 Instrumentation

To visualize the fundamental steps of chemisorptions and reactions that occur at surfaces, in situ STM investigations typically monitor the diffusion or transformation of adsorbed molecules. A series of snapshots of preselected surface regions, compiled into a STM movie, can reveal the evolution of surface phenomena. On metal surfaces, the surface diffusion of adsorbates is usually so rapid that the

3.2 Instrumentation

surface temperature must be lowered below room temperature (RT) for successful STM viewing. Low-temperature (LT) STM, developed in the mid-1990s, not only potentially allows the determination of reaction intermediates and pathways but also enables the precise control and measurement of the bond activation processes using the STM tip. To measure reaction kinetics, STM should have the capability to resolve adsorbates at temperatures relevant to catalytic reactions. For this purpose, a variable temperature (VT) STM is required, as well as capabilities for rapid scanning. VT STM with a typical scan rate of one frame per minute was developed in the mid-1990s. Considering the scanning probe is a mechanical probe driven by electronics, the acquisition time of STM images is typically restricted by the mechanical behavior of the scanning components and the performance of the electronics. In the mid-1990s, a few STM groups achieved a fast scan rate of approximately 20 frames/s on extended model catalyst surfaces [4–6]. Working on the compact design of the scanner probe and using high performance electronics, Frenken and coworkers [7] have recently pushed the scan rate above the video rate (50 frame/s) on a graphite surface, with atomic resolution and an image with 256  256 pixels. Recently, a few groups have taken up the challenge to extend the in situ STM investigations to high pressures. A major challenge in imaging surfaces with STM over a wide pressure range is the sensitivity of the tunneling current to extremely small changes at the tunneling junction resulting from induced instabilities by the ambient gas. Efforts have emphasized the design of a STM that can work at high temperatures and pressures with greater stabilities [8–11]. For in situ STM studies at high temperature and pressures, the inability of being able to track a preselected surface area is often the limiting factor given the tunnel junction instabilities and sample drifts. To overcome this challenge and to maintain contact with a specific surface region, it is important to develop experimental approaches that pattern the surface without influencing the kinetics and dynamics of the particular areas under study. A “shadowing” technique (Figure 3.1a) has been developed where metal atoms are dosed with the STM tip in the tunneling position with the collimated metal flux creating a shadow of the tip on the substrate [12, 13]. For metal clusters supported on an oxide surface, tip manipulation is another method of choice (Figure 3.1c). This technique removes clusters from a specific area through aggressive scanning. Using the STM tip to pattern the surface, it is now possible to monitor a preselected surface area at elevated temperatures while changing the gas pressure over 12 orders of magnitude. In addition to the instrumental performance, the STM tip is of primary importance for in situ STM measurements. Methods for the preparation of STM tips have been extensively studied with a goal of preparing an atomically sharp tip [14–19]. The STM tip is important for high-pressure studies with respect to two aspects, tip selection and in situ tip regeneration. To ensure a continuous DOS near the Fermi level, transition metals are usually selected to prepare STM tips. In the presence of reactant gases, especially under high-pressure and high-temperature conditions, the chemical and thermal stability of the STM tip becomes the ultimate limit for reaction studies and thus the major concern in tip selection. Tungsten tips are very stable in

j57

j 3 In Situ STM Studies of Model Catalysts

58

Figure 3.1 Methods of patterning the surface for in situ STM studies. (a) Schematics of “shadow” technique. (Reprinted with permission from Ref. [12]. Copyright 2002, Wiley, Inc.) (b) STM image of the surface created by “shadow” technique. The shadow area uncovered by metal clusters is distinguished from the area covered with metal clusters by the

white dash line. (c) Schematics of the tip manipulation. (d) STM image of the surface created by tip manipulation. The dash rectangle in (d) shows the area where most clusters are picked up by the STM tip. This area with lower cluster densities can be distinguished from the rest of the surface and serves as a nanomarker for in situ STM studies.

CO but perform poorly in the presence of O2 or mixtures of CO and O2. Platinum or platinum alloy tips are stable in O2 but suffer from adsorption of CO, especially when the sample surface temperature is above 550 K [20]. Gold is stable in both CO and O2 but unstable at high temperatures, especially in the presence of water. In addition to tip selection, in situ tip regeneration or cleaning is also critical for STM studies in the presence of high-pressure reactant gases because the STM tip is susceptible to picking up poorly conducting components during extended measurements at elevated pressures. In situ tip regeneration refers to the method of applying a large voltage pulse (from a few to hundreds of volts) between tip and sample while the tip is in tunneling range. This method induces field emission, which cleans and regenerates the STM tip. Wintterlin and coworkers [21] recently reported a high-pressure STM study, in which tungsten tips were used to study ethylene oxidation on Ag(1 1 1); the tip could be recovered by applying high voltages to the tip (e.g., þ 300 V).

3.3 Visualizing the Pathway of Catalytic Reactions

Gas purification is another important issue for high-pressure STM studies. When the surface is exposed to high pressures, even highly diluted impurities may completely contaminate the surface. Purification of all reactant gases using liquid N2 can greatly improve the operating pressure range for high-pressure reaction studies and prevent the electrical breakdown often induced by humidity when backfilling the STM chamber [13]. For a flow-reactor system, where purification of large volume of gases is required, a heated zeolite filter is effective in removing carbonyl contaminants from the gas flow [22, 23].

3.3 Visualizing the Pathway of Catalytic Reactions

With the introduction of LT and VT STM, it is now possible to monitor the fundamental steps of chemical reactions, that is, reactant chemisorption, diffusion, and catalytic transformation. A detailed review covering this subject was published by Wintterlin in 2000 [24]. Since then, in situ STM studies have flourished and expanded to the visualization of the reaction pathway and kinetics of surface processes. In the following section, we highlight selected examples of recent progress in using in situ STM for studying fundamental catalytic processes. 3.3.1 Imaging of Adsorbates and Reaction Intermediates

STM can induce adsorption–desorption and dissociation processes nonthermally and with spatial control. At low temperatures, with limited surface diffusion, the motion of adsorbates can be controlled and allows the determination of surface reaction intermediates. The catalytic oxidation of CO on precious metal surfaces is one of the most important model reactions in heterogeneous catalysis. Hahn and Ho [25] have recently used in situ STM to visualize the reaction pathway of CO oxidation on Ag(1 1 0) at 4 K. Figure 3.2 depicts the pathway believed to be operative: the Langmuir–Hinshelwood mechanism. Figure 3.2a and b shows a CO molecule and a pair of oxygen atoms adsorbed on the Ag(1 1 0) surface, respectively. The oxygen atom pair was formed by placing the tip over a molecularly adsorbed O2 and raising the sample bias to 0.47 V. Subsequently, the bond between two oxygen atoms is broken and the two oxygen atoms adsorbed at the nearest fourfold sites of the Ag(1 1 0) surface show slight elongation along the [1  1 0] direction in Figure 3.2b. The STM tip was then placed over the CO molecule with a þ 0.24 V bias applied repeatedly, causing the molecule to diffuse across the surface. Eventually, the CO molecule moved close to the pair of O atoms (Figure 3.2c) and then joined them to form the O
CO
O complex (Figure 3.2e). With an additional pulse of the sample bias over the CO molecule, the O
CO
O complex is decomposed, leaving an oxygen atom on the surface with the CO2 desorbing from the Ag surface. A second reaction pathway was also illustrated by moving a CO molecule toward the absorbed oxygen atoms. In Figure 3.3a, an STM tip with a CO molecule adsorbed at the

j59

j 3 In Situ STM Studies of Model Catalysts

60

Figure 3.2 STM images obtained with a CO-terminated tip, Vt ¼ 70 mV and It ¼ 1 nA. (a) Isolated CO molecule, (b) two O atoms (adsorbed on the nearest fourfold hollow sites along the [1 1 0] direction), (c) CO and two O atoms separated by 6.1 Å along the [0 0 1] direction, and (e) O
CO
O complex. Grid lines are drawn through the silver surface atoms. Scan area of (a–c) and (e) is 25 Å  25 Å.

Schematic diagrams for adsorption geometries of (c) and (e) are shown in (d) and (f ), respectively; a linear atop and a tilted off-site CO are implicated. The black (red) circles represent carbon (oxygen) atoms and the large gray circles are silver atoms. The sizes of the circles are scaled to the atomic covalent radii. (Reprinted with permission from Ref. [25]. Copyright 2001, The American Physical Society.)

tip end was positioned over an oxygen atom, which lined up with another oxygen atom from the O2 dissociation. The two oxygen atoms were separated from each other by two lattice spacings. With a pulse of þ 0.47 V sample bias, the CO molecule is detached from the STM tip and reacted with the oxygen below the tip to form CO2. Figure 3.3b shows the two pulses of tunneling current, corresponding to the CO molecule, impinging on the surface and reacting with the adsorbed oxygen, and to CO2 desorption, respectively. Figure 3.3c shows the oxygen atom left on the Ag(1 1 0) surface. The combined imaging, manipulation, and spectroscopic capabilities of the STM provide direct visualization of reaction pathways at the single molecule level. At low temperatures, tip manipulation has been regularly used to promote surface diffusion, to activate bonds, and to synthesize molecules. Recent progress along these lines is described by Hla and Rieder [26, 27].

3.3 Visualizing the Pathway of Catalytic Reactions

Figure 3.3 Reaction of a CO molecule released from a CO-terminated tip with an O atom adsorbed on the surface. (a) STM image, taken with a CO-terminated tip, of two O atoms separated by two lattice spacings (2  2.89 Å) along the ½1  1 0 direction. Grid lines are drawn through the silver surface atoms. (b) Tunneling current during a 1470 mV sample bias pulse with the CO-terminated tip over one of the two O atoms (denoted by “ ”). Two current rises

(at 250 and 310 ms) indicate the moments of desorption and reaction of CO from the tip and the moment of desorption of CO2 into vacuum. (c) STM image of the same area rescanned after the pulse, showing CO on the tip has reacted away. Scan area of (a) and (c) is 25 Å  25Å. (d–f) are the schematic diagrams for (a–c), respectively. (Reprinted with permission from Ref. [25]. Copyright 2001, The American Physical Society.)

3.3.2 Imaging Chemisorption on Metals

Under catalytic reaction conditions, adsorbates are usually mobile on the surface. The diffusion of adsorbates has been studied both on metal surfaces and on oxide surfaces by in situ STM. On metal surfaces, it has been shown that at low surface coverage, adsorbates are mobile and are distributed randomly on the surface. As the surface coverage is increased, the interaction between adsorbates also changes such that an attractive interaction begins to appear, leading to the formation of adsorbate islands. These adsorbate islands are in equilibrium with the diffusing adsorbates (2D gas) at the surface. Eventually, with an increase in the adsorbate coverage, the adsorbate islands grow into an adsorbate overlayer. In situ STM studies by Wintterlin et al. [4] and by Wong et al. [28] both illustrate that the diffusion rate of adsorbates on metal surfaces depend on their coverage and/or nearest neighbors. The surface diffusion coefficient measured for adsorbates on metal surfaces is meaningful only at very low surface coverages where the adsorbate–adsorbate interaction has a negligible influence on adsorbate diffusion. The structure of the adsorbate layer formed at high surface coverage is more relevant to catalytic reactions at high pressures. Besenbacher and coworkers [29–33] found for CO on Pt(1 1 0) and Pt(1 1 1) and NO on Pd(1 1 1) that the structure of high

j61

j 3 In Situ STM Studies of Model Catalysts

62

adsorbate coverages, formed at low-temperature and low-pressure conditions, is identical to the structure formed at room temperature and high pressures. This finding suggests that on metal surfaces, reaction studies at high surface coverage conducted at low temperature likely connect with real catalytic processes at high pressures. By studying the diffusion of surface adsorbates (or adsorbate vacancies), in situ STM can be used to determine the active site for chemisorptions. Mitsui et al. [34–36] studied the process of hydrogen dissociation on Pd(1 1 1) using in situ STM. Pd is a catalyst widely used in hydrogenation and dehydrogenation reactions. Exposure of Pd to H2 leads to dissociative adsorption. At approximately 65 K and in the presence of 2  10
7 Torr of H2, Pd(1 1 1) is nearly saturated with H atoms leaving only a few vacancies as sites for the dissociation of adsorbed H2 molecules (Figure 3.4). Due to the inversion of the image contrast caused by the adsorption of H atoms on the STM tip, the empty surface sites are imaged as protrusions. This surface can be used to model the Pd(1 1 1) surface under high-pressure H2 at room temperature. Figure 3.5 shows a sequence of STM images acquired for the same surface region. Figure 3.5a depicts a number of isolated vacancies (bright spots) separated by more than one Pd lattice, as well as two aggregates of vacancies (marked by dashed circles). Each vacancy aggregate consists of a pair of vacancies (dimer) occupying neighboring fcc sites. The dimers in Figure 3.5 are always imaged as a three-lobed object because of the fast diffusion of a neighbor H atom, which can hop over bridging sites to occupy the vacancy pairs without getting close to other H atoms. The vacancy pairs or dimers are most frequently encountered in the STM study. Isolated vacancies can hop

Figure 3.4 The 6.5 nm  6.5 nm STM image of Pd(1 1 1) with a H coverage near one monolayer. Numerous H vacancies, visible as bright protrusions, are present. Vt ¼ 45 mV and It ¼ 2.7 nA. Streaks and fractional protrusions are due to vacancies moving while the tip is scanning over them. (Reprinted with permission from Ref. [35]. Copyright 2005, Springer.)

3.3 Visualizing the Pathway of Catalytic Reactions

Figure 3.5 STM images from a movie showing the formation, separation, and annihilation of H-vacancy clusters. The images on the left (3 nm  2.5 nm) are repeated on the right with annotations. (a) Five vacancies near the center are labeled (A–E) and two triangular vacancy pairs (2V) are marked with dashed triangles for reference. (b) Vacancies A and B have formed a 2V cluster indicated by the triangle containing the number “2”. Vacancies C–E have formed a

three-vacancy (3V) cluster, indicated by the larger triangle containing the number “3”. (c) The 2V pair has separated into isolated vacancies A and B, while the 3V cluster has been annihilated by dissociative adsorption of a H2 molecule, leaving a single remaining vacancy C. The other 2V clusters separated a few frames later. (Reprinted with permission from Ref. [34]. Copyright 2003, Nature Publishing Group.)

randomly on the Pd surface and occasionally coalesce to form aggregates, as shown in Figure 3.5b. Vacancies A and B aggregate to form a dimer while vacancies C, D, and E coalesce to form a three-vacancy aggregate. The vacancy dimer remains together for several minutes and eventually disintegrates back to isolated vacancies. However, the three-vacancy aggregate disappears and leaves only one vacancy on the surface (Figure 3.5c). The authors concluded then that two vacancies in the three-vacancy aggregate are occupied by H atoms from the dissociation of adsorbed H2. The authors

j63

j 3 In Situ STM Studies of Model Catalysts

64

also found, in the presence of 2  10
7 Torr H2 at 65 K, vacancy aggregates with four or more vacancies are also filled by H atoms from the dissociation of H2 within the aggregates and transformed into a single vacancy or totally annihilated. Through a series of STM movies on the diffusion of surface hydrogen vacancies, the authors found that only an aggregate with three or more vacancies could be annihilated by H2 dissociation. The dimers or isolated vacancies are never occupied by H atoms. Instead, the dimers always dissociate creating isolated vacancies with an average lifetime of 10 min at 65 K. From these data, the authors concluded that three or more empty palladium sites are necessary for the dissociation of H2 molecules. This finding is rather surprising since it has traditionally been assumed that two neighboring empty sites are sufficient for the dissociation of a diatomic molecule. The discovery of the active sites for H2 dissociation on Pd(1 1 1) illustrates the power of in situ STM in addressing the elementary steps of surface reactions and in testing the conventional assumptions in catalysis. 3.3.3 Determining the Sites for Chemisorption on Oxide Surfaces

On reduced oxide surfaces, the diffusion of an adsorbate is often limited by the localized bonding, either ionic or covalent, between the adsorbate and the oxide substrate. The relatively slow diffusion of adsorbates allows chemisorption and diffusion on oxides to be studied by STM at elevated temperatures. For example, TiO2 is an excellent photocatalyst for dissociation of water and decomposition of organic molecules, critical to pollution control and the hydrogen economy. Studies of the adsorption and diffusion of water, oxygen, and organic molecules on TiO2 are of primary importance to our understanding of photocatalysis by TiO2. Being the most stable phase of TiO2, the rutile TiO2(1 1 0) crystal has been studied most extensively by STM and other surface science techniques [37]. Figure 3.6 shows a structural model of the rutile TiO2(1 1 0)-(1  1) surface. The surface contains two types of titanium atoms that form rows along the [0 0 1] direction. Rows of sixcoordinated Ti atoms alternate with five-coordinated terminal Ti atoms, which miss a single O atom perpendicular to the surface. The surface also contains two kinds of oxygen atoms, that is, three-coordinated oxygen atoms, sitting in the surface plane, and bridging oxygen atoms, sitting above the surface plane and bonded to two sixcoordinated Ti atoms. Undersaturated bridging oxygen atoms can be easily removed from the surface by annealing, electron bombardment, or ion sputtering to form bridging oxygen vacancies. Bridging oxygen vacancies are the most common and well-defined defects on the TiO2(1 1 0) surface. STM studies on the adsorption and diffusion of small molecules on the TiO2(1 1 0) surface began in the late 1990s [37–39] and have provided a general understanding of the important role of bridging oxygen vacancies in the dissociation of water, oxygen, and small organic molecules. However, due to the difficulties in distinguishing active surface sites and dissociation products, only recently have the fundamental steps of adsorption and dissociation processes been understood with the help of in situ STM.

3.3 Visualizing the Pathway of Catalytic Reactions

Figure 3.6 The ball model of the TiO2(1 1 0) surface. Large gray (red) balls represent O atoms, small light gray (gray) balls fivecoordinated surface Ti atoms (5f-Ti), and small black balls sixcoordinated Ti atoms (6f-Ti). The bridging oxygen atoms (Obr), single oxygen vacancies (VO), and O atoms adsorbed on top of the 5f-Ti row (Oot) are indicated. (Reprinted with permission from Ref. [40]. Copyright 2005, Elsevier.)

Wendt et al. [40] and Bikondoa et al. [41] studied the adsorption and dissociation of water and O2 on TiO2(1 1 0). In situ STM studies, in parallel with the use of DFT methods, allow surface features such as bridging oxygen vacancies, adsorbed oxygen atoms, surface hydroxyls, and adsorbed water to be distinguished. Figure 3.7a illustrates the difference between a bridging oxygen vacancy, a surface hydroxyl, and adsorbed water in their appearance in STM images. Due to electronic effects, STM resolves the five-coordinated terminal Ti rows as bright rows whereas the bridging oxygen rows are imaged as dark rows. The nuances in the appearance of oxygen vacancies and hydroxyls in STM images are distinguished by visualizing oxygen vacancies being transformed into OH species in situ. The authors have also found that applying voltage pulse (3 V) over the hydroxyls desorbs individual H atoms from the hydroxyl groups while leaving the oxygen vacancies intact. Using in situ STM, Wendt et al. [42] demonstrated that water dissociation takes place at bridging oxygen vacancies of the TiO2(1 1 0) surface at 187 K and form paired hydroxyl groups, with one positioned at the oxygen vacancy site and the other formed at the nearest bridging oxygen site. The diffusion of these pairs is inhibited at 187 K but can be initiated in the presence of neighboring water molecules adsorbed in the five-coordinated Ti trough. Through the interaction with neighboring water molecules, hydrogen atoms from the paired hydroxyl groups are transferred to adjacent bridging oxygen rows, causing the cross-row diffusion of hydroxyl groups. Dohnalek and coworkers [43, 44] have further measured the diffusion kinetics of hydroxyl groups (or H atom) at room temperature and above. In situ STM images (Figure 3.8) have confirmed that H2O dissociates at the bridging oxygen

j65

j 3 In Situ STM Studies of Model Catalysts

66

Figure 3.7 STM images (16 nm  16 nm) of clean, reduced TiO2(1 1 0) samples showing the difference between bridging oxygen vacancies, surface hydroxyls, and adsorbed water. The sample in (a) was less reduced than the sample in (b). (c) STM height profiles along the ½1 1 0 direction of species indicated in (b). (Reprinted with permission from Ref. [40]. Copyright 2005, Elsevier.)

Figure 3.8 STM images of the same area on TiO2(1 1 0) at 357 K (Vt ¼ 1.5 V, It ¼ 0.1 nA) as a function of time (Dt ¼ 60 s): (a) clean TiO2(1 1 0) with bridging oxygen (BBO) vacancies; (b) TiO2 (1 1 0) with a geminate hydroxyl pair formed by adsorption and dissociation of a water molecule. HV marks the

OH hydrogen and HB the hydrogen that split off from the OH; (c) same area after a single hop of HB; and (d) after subsequent hop of HV. Insets exhibit the ball models illustrating the corresponding processes. (Reprinted with permission from Ref. [44]. Copyright 2008, The American Chemical Society.)

3.3 Visualizing the Pathway of Catalytic Reactions

vacancies, producing paired hydroxyl groups. Hydrogen atoms readily diffuse at room temperature along the bridging oxygen row. However, the two hydrogen atoms in the paired hydroxyl groups exhibit inequivalent diffusivity. The hydrogen atom positioned at the healed oxygen vacancy site (HV) diffuses much slower than the one adsorbed at neighbor bridging oxygen sites (HB). The different diffusion rates of HV and HB were measured between 300 and 410 K and the activation barrier of HB estimated to be approximately 0.22 eV lower than HV. The diffusion barrier of these hydrogen atoms increases with the separation between hydroxyl groups, suggesting a repulsive OH
OH interaction. The authors speculated that a long-lived polaronic state is responsible for the inequivalent diffusion rates of HV and HB. However, the measured kinetic parameters (prefactors and diffusion barriers) could not be reproduced by DFT calculations, suggesting a rather complex diffusion mechanism. The pathway of dissociation and diffusion discovered in the above water adsorption experiment also applies to the adsorption of alcohols on TiO2(1 1 0). In situ STM has been used to study the adsorption of methanol [45] and butanol on TiO2(1 1 0) [46]. Figure 3.9 shows a series of STM images obtained on the same area following exposure to methanol. Figure 3.9a depicts the clean surface before exposure, with bridging oxygen vacancies marked as yellow circles. The exposure of 0.06 ML methanol led to the dissociative adsorption of methanol at the bridging oxygen vacancies of TiO2(1 1 0) (Figure 3.9b). The dissociation of methanol forms a methoxy group at the bridging oxygen vacancy, resolved as bright features in Figure 3.9b, and a hydroxyl group at the nearest neighbor. The methoxy group has a similar appearance to the bridging oxygen vacancy in STM images, except that the methoxy group is 0.8 Å higher than the bridging oxygen vacancy. The hydroxyl group could not be differentiated from the bright features of neighboring methoxy groups in Figure 3.9b. However, with time, the hydrogen from the hydroxyl group (red dots in Figure 3.9c and d) diffuses along the bridging oxygen row and across the bridging oxygen rows through interactions with methanol molecules weakly bounded to the Ti trough. The diffusing hydrogen atoms were identified by their apparent height in STM images and by a tip desorption experiment (Figure 3.9e) proposed by Bikondoa et al. [41]. Figure 3.9f gives a graphic illustration of the dissociation and diffusion pathway of methanol on TiO2(1 1 0), which was also observed in the adsorption experiment of 2-butanol (CH3CH2CH(OH)CH3) on TiO2(1 1 0) at room temperature [45]. The adsorption and dissociation of O2 on TiO2(1 1 0) have been investigated by Wendt et al. [40] and Du et al. [47]. Both observed that O2 molecules adsorb and dissociate at the bridging oxygen vacancies, with one O adatom healing the vacancy and the other O adatom bounded to the neighboring five-coordinated Ti site. Du et al. also analyzed the lateral distribution of the O adatoms upon dissociation and discovered a transient mobility of O adatoms along the Ti trough in the [0 0 1] direction. Unlike the dissociative adsorption of O2 on metal surfaces where both adatoms have equal diffusivity, the diffusivity of O adatoms on TiO2(1 1 0) was found to be inequivalent. While the O adatoms filling the vacancy are locked in the bridging oxygen row, O adatoms in the Ti trough are relatively free to move. A majority of O adatoms on the Ti trough (81%) were found separated from the O adatoms in the previous vacancy sites by two lattice constants.

j67

j 3 In Situ STM Studies of Model Catalysts

68

Figure 3.9 STM images of same area before and after adsorption of methanol on reduced TiO2(1 1 0) at 300 K (Vt ¼ 1.0  0.3 V and It ¼ <0.1 nA): (a) bare surface; (b) after 80 s exposure to methanol; (c) after 110 s exposure to methanol; (d) taken on (c) after spontaneous tip change; (e) after high bias (3.0 V) sweep of (c); (f) schematic model of the adsorption

process. Insets show magnified areas marked by squares. Yellow circles show the position of bridging oxygen vacancies. Blue circles show the methoxy groups on oxygen vacancies. Red squares show H atoms diffusing on bridging oxygen rows. (Reprinted with permission from Ref. [45]. Copyright 2006, The American Chemical Society.)

3.3 Visualizing the Pathway of Catalytic Reactions

Zhang et al. [48] have recently measured the stability of bridging oxygen vacancies on TiO2(1 1 0) using in situ STM. Sequences of STM images between 340 and 420 K suggest that bridging oxygen vacancies migrate along the bridging oxygen row via the slow diffusion of bridging oxygen atoms with a diffusion barrier of 1.15 eV, in agreement with DFT calculations. All the above studies suggest that the surface chemistry of TiO2(1 1 0) is dictated by bridging oxygen vacancies, which can account for approximately10% of the bridging oxygen sites. However, there are disagreements. Lyubinetsky et al. [49] studied the adsorption of trimethylacetic acid ((CH3)3CCOOH, TMAA), a photoreactive molecule, on TiO2(1 1 0) at room temperature. In situ STM found that the deprotonation of TMAA to form TMA does not necessarily occur at bridging oxygen vacancies. None of the hydroxyl groups was found during the adsorption of TMAA. Instead, the hydrogen atom was bound to a pair of bridging oxygen atoms and stabilized by the adjacent TMA groups sitting on the five-coordinated Ti trough. At saturation coverage, TMAA formed a (2  1) overlayer on the TiO2(1 1 0) surface. Wendt et al. [50] recently studied the interaction between O2 and TiO2(1 1 0) surface in detail and suggested that bridging oxygen vacancies are only the minor sites that account for O2 dissociation. Even though bridging oxygen vacancies account only for approximately 10% of surface bridging oxygen sites, exposing the clean TiO2(1 1 0) to a few Langmuirs of O2 could not fully remove all bridging oxygen vacancies. To isolate the influence of bridging oxygen vacancies in O2 dissociation, the authors created a “perfect” TiO2(1 1 0) surface by exposing the TiO2(1 1 0) surface to water at room temperature. Hydroxyl groups, formed via water dissociation, covered all bridging oxygen vacancies, yielding a vacancy-free TiO2(1 1 0) surface (Figure 3.10a). Figure 3.10c and d illustrates that O2 exposure can fully remove surface hydroxyl groups and create a TiO2(1 1 0) surface with perfect bridging oxygen rows, as previously suggested in TPD studies [51]. With the titration of hydroxyl groups, oxygen adatoms on the five-coordinated Ti row (Oot) also increase (Figure 3.10b). However, the increase in oxygen adatoms does not seem to stop even after all the hydroxyl groups have been replaced with oxygen (Figure 3.10c and d). Paired Oot atoms start to appear on the five-coordinated Ti row during extended O2 exposure. On the basis of these observations, the authors showed that a second and primary O2 dissociation channel is operative on the five-coordinated Ti row. STM results combined with photoelectron spectroscopy (PES) on the valence state of TiO2(1 1 0) further show that the removal of all hydroxyl groups by oxygen, leading to a perfect TiO2(1 1 0) surface, only slightly attenuates the Ti 3d defect state (Figure 3.10e). The full attenuation of Ti 3d state requires 420 L of O2. Figure 3.9f plots the evolutions of the Ti 3d defect state and the OH 3s state over O2 exposure and suggests the Ti 3d defect state is not mainly caused by bridging oxygen vacancies. The authors suggest that other types of defects, Ti3 þ interstitials that form during the reduction of TiO2(1 1 0) and are hidden beneath the surface, are primarily responsible for the formation of the Ti 3d defect state and the dissociative adsorption of O2. Indeed, the importance of Ti3 þ interstitials has also been realized in previous in situ STM studies of the reoxidation of TiO2(1 1 0) [52–54]. It is noted that Ti3 þ

j69

j 3 In Situ STM Studies of Model Catalysts

70

Figure 3.10 (a–d) STM images (105 Å  105 Å) of the TiO2 (1 1 0) surface covered with hydroxyls [h-TiO2 (1 1 0)] and then exposed to increasing amounts of O2 at room temperature. (e) Selected PES valence-band spectra recorded on an h-TiO2(1 1 0) surface that was exposed to O2 at RT. Arrows indicate the representative STM images. (f) Normalized integrated

intensities of the OH 3s (red) and Ti 3d (blue) features for O2 exposures up to 420 L from PES spectra; circles indicate intensity values that were obtained from the spectra shown in (e). (Reprinted with permission from Ref. [50]. Copyright 2008, The American Association for the Advancement of Science.)

interstitials diffuse to the surface in the presence of O2 and form TiOx species, which serve as the building blocks for the regrowth of TiO2(1 1 0) plane. Using in situ STM, Bowker and coworkers [52] measured the reoxidation kinetics of reduced TiO2(1 1 0) surface at temperatures between 573 and 1000 K and oxygen pressures of 5  10
8 to 2  10
6 mbar. By monitoring the TiOx species that diffused and coalesced on the TiO2(1 1 0) surface, the surface growth rate could be measured through the change of island morphology. This growth rate was found to be linear with respect to the oxygen partial pressure. The regrowth of TiO2(1 1 0) has a low activation energy of approximately 25 kJ/mol, suggesting a high mobility of Ti3 þ interstitials in the presence of oxygen. Previous studies on the reduction of TiO2(1 1 0) [37] have shown the rutile TiO2(1 1 0) bulk serves as a huge reservoir for Ti3 þ interstitials during reduction, so that the surface is maintained near-stoichiometry and is thermodynamically stable.

3.3 Visualizing the Pathway of Catalytic Reactions

Ti3 þ interstitials are essentially oxygen vacancies within the bulk whereas bridging oxygen vacancies are basically undercoordinated Ti ions at the surface. It is not surprising to expect that Ti3 þ interstitials in the subsurface play a role in the dissociation of adsorbed molecules. In the above STM study by Wendt et al., it is worth noting that subsurface Ti3 þ interstitials were neither visualized nor seen to diffuse to the TiO2(1 1 0) surface during O2 adsorption at room temperature. Considering PES usually probes the top few layers at the surface, the complete attenuation of the Ti 3d defect state suggests that Ti3 þ interstitials within those layers have been oxidized and therefore quenched. Details of how the excess charge of Ti3 þ interstitials involves in the bond breaking of O2 molecules remain to be elucidated. Nevertheless, the above studies demonstrate the power of in situ STM in tracing the active sites, hidden or unhidden. The finding of mobile defects in a rigid TiO2(1 1 0) will stimulate more investigations on other reducible oxides and encourage revisiting their surface chemistry. 3.3.4 Visualizing Reaction Intermediates and the Mechanism of Hydrogen Oxidation

In situ STM has also been used to study the pathway of surface reactions and to measure their kinetics. The hydrogen oxidation reaction, catalyzed by Pt group metals to produce water, was the first catalytic reaction discovered in 1823. This reaction is still a core catalytic reaction at the heart of fuel cell technologies. Although the oldest of catalytic reactions, the mechanism of catalytic hydrogen oxidation is still unclear, especially for its surprising reactivity at or below the water desorption temperature (170 K). It has been proposed that the reaction proceeds via the combination of dissociatively adsorbed O atoms (Oad) and H atoms (Had), forming hydroxyl groups that subsequently bind with another Had to form water. The formation of hydroxyl groups has been postulated as the rate-limiting step. However, the hydroxyl group, as a reaction intermediate, could not be confirmed in early surface spectroscopic studies. It has not been until the past decade, with the help of in situ STM, that the reaction mechanism has become apparent for the hydrogen oxidation reaction on metal surfaces. Wintterlin and coworkers [55, 56] studied the hydrogen oxidation reaction on Pt(1 1 1) using in situ STM. The study was conducted as a titration experiment, in which the surface was precovered with Oad and then the O-terminated surface subsequently exposed to H2 molecules. The O-terminated Pt(1 1 1) surface was prepared by exposing the clean Pt(1 1 1) surface to 10 L of O2, followed by annealing at 225 K to dissociate O2. The surface was then exposed to 8  10
9 mbar H2 at 131 K and monitored by STM, as shown in Figure 3.11. The Pt(1 1 1) surface was precovered by a (2  2)-Oad layer, where Oad atoms were imaged as dark dots (Figure 3.11a). Upon exposure to H2, a few bright islands form on top of the (2  2)-Oad layer (Figure 3.11b). These islands continue to grow and eventually develop into an ordered layer with hexagonal and honeycomb phases (Figure 3.11c). These two phases are characterized as a surface hydroxyl (OHad) overlayer, formed by hydrogen bonding. Additional diffusing islands seen in Figure 3.11c are attributed to adsorbed

j71

j 3 In Situ STM Studies of Model Catalysts

72

Figure 3.11 Series of successive STM images, recorded during dosing of the O-covered Pt(1 1 1) surface with H2. (a–c) Frames (17 nm  17 nm) from an experiment at 131 K [P(H2) ¼ 8  10
9 mbar]. The hexagonal pattern in (a) is the (2  2)-O structure; O atoms appear as dark dots and bright features are the initial OH islands. In (c), the area is mostly covered by OH, which forms ordered structures. The white, fuzzy features are H2O-covered areas.

(d–f) Frames (220 nm  220 nm) from an experiment at 112 K [P(H2) ¼ 2  10
8 mbar]. In (d), the surface is mostly O-covered (not resolved). The bright dots are small OH islands, most of which are concentrated in the expanding ring. H2O in the interior of the ring is not resolved here. Thin, mostly vertical lines are atomic steps. (Reprinted with permission from Ref. [55]. Copyright 2001, The American Association for the Advancement of Science.)

H2O islands (H2Oad). Figure 3.11a–c shows the formation of hydroxyl group as the reaction intermediate and reveal the atomic details of hydrogen oxidation catalyzed by Pt(1 1 1). Figure 3.11d–f presents snapshots of STM images of the Oad-terminated Pt(1 1 1) surface exposed to 2  10
8 mbar H2 at 112 K. The imaged area includes several surface terraces covered with numerous small bright dots and a bright ring that expands with time. These small bright dots have been assigned to small OHad islands that appear after H2 exposure. The white ring, termed as the reaction front, is concentrated with small OHad islands and continues to grow as the reaction progresses, suggesting a fast reaction at the boundaries between Oad atoms and the diffusing H2Oad. The fast reaction between Oad and H2Oad produces two OHad, which in turn forms H2Oad through a rapid reaction with Had atom. Since the combination of Oad and Had to form OHad is the rate-limiting step, the presence of H2Oad removes this kinetic limit and promotes an autocatalytic cycle until depletion of Oad atoms. STM images thus illustrate an autocatalytic reaction mechanism that accounts for the low activation barrier and high reactivity of hydrogen oxidation at or

3.3 Visualizing the Pathway of Catalytic Reactions

below the water desorption temperature. At temperatures higher than 170 K, where H2O starts to desorb, the shortened lifetime of H2Oad breaks down the autocatalytic cycle by stopping the fast reaction between H2Oad and Oad to form OHad. The autocatalytic reaction mechanism apparent at low temperatures is expected to apply to catalytic hydrogen oxidation at high pressures. In addition, the above study is the first to use STM to observe the formation of dynamic surface patterns at the mesoscopic level, which had previously been observed by other imaging techniques in surface reactions with nonlinear kinetics [57]. This study illustrates the ability of in situ STM to visualize reaction intermediates and to reveal the reaction pathway with atomic resolution. 3.3.5 Measuring the Reaction Kinetics of CO Oxidation

Another important catalytic reaction that has been most extensively studied is CO oxidation catalyzed by noble metals. In situ STM studies of CO oxidation have focused on measuring the kinetic parameters of this surface reaction. Similar to the above study of hydrogen oxidation, in situ STM studies of CO oxidation are often conducted as a titration experiment. Metal surfaces are precovered with oxygen atoms that are then removed by exposure to a constant CO pressure. In the titration experiment, the kinetics of surface reaction can be simplified and the reaction rate directly measured from STM images. Wintterlin et al. [58] investigated the catalytic oxidation of carbon monoxide on Pt(1 1 1) using in situ STM. Oxygen atoms were preadsorbed on the Pt(1 1 1) surface by exposing the surface to 3 L O2 at 96 K, followed by a short anneal at 293 K to dissociate O2. The oxygen-covered Pt(1 1 1) surface was then cooled to 247 K and exposed to 5  10
8 Torr CO. STM was used to follow the change of surface adsorbate structures as a function of the CO exposure time (Figure 3.12). At 247 K, Oad atoms form an ordered (2  2) overlayer, imaged as dark dots. At t ¼ 0, the Pt(1 1 1) surface is mainly covered by the (2  2)-O layer together with empty Pt sites, imaged as bright islands, scattered on the surface. The addition of CO molecules lowers the mobility of surface oxygen atoms and slowly compresses the (2  2)-O layer into large islands. The adsorbed CO molecules form ordered c(4  2) domains on the Pt(1 1 1) surface. As time progresses, the areas of c(4  2) CO domains continue to grow at the expense of the (2  2)-O islands. From the series of in situ STM images, the rate of CO oxidation can be estimated based on the reduction rate of the surface areas of the (2  2)-O islands. Figure 3.13 plots the dependence of the reaction rate on the surface area or perimeter of oxygen domains, as a function of time. Approximately, the reaction rate is linear with respect to the perimeter of the surface oxygen domains, suggesting CO oxidation mainly occurs along the boundary between the oxygen and the CO domains on the Pt(1 1 1) surface. The titration experiments were repeated at various temperatures between 237 and 274 K. An Arrhenius plot gives an activation energy of 0.49 eV and a prefactor of 3  1021 cm
2 s
1, in good agreement with the kinetic parameters obtained from macroscopic measurements.

j73

j 3 In Situ STM Studies of Model Catalysts

74

Figure 3.12 Series of STM images, recorded during reaction of adsorbed oxygen atoms with coadsorbed CO molecules at 247 K, all from the same area of a Pt(1 1 1) crystal. Before the experiment, a submonolayer of oxygen atoms was prepared and CO was continuously supplied from the gas phase

(PCO ¼ 5  10
8 mbar). The times refer to the start of the CO exposure. The structure at the upper left corner is an atomic step of the Pt surface. Image sizes, 180 Å 170 Å; Vt ¼ 0.5 V; It ¼ 0.8 nA. (Reprinted with permission from Ref. [58]. Copyright 1997, The American Association for the Advancement of Science.)

On the Pd(1 1 1) surface, Wintterlin and coworkers [59] have shown that CO oxidation goes through a different reaction pathway at low temperatures. Similar titration experiments were performed by exposing the Pd(1 1 1) surface precovered by (2  2)-O overlayer to 2  10
8 Torr CO at 143 K (Figure 3.14). Using STM to follow the same area of the Pd(1 1 1) surface, these authors found that CO does not react with surface oxygen at this temperature. Instead, CO molecules slowly occupy the surface sites of Pd(1 1 1) and compresses the (2  2) oxygen domains into the (2  1)-O phase. This phase was imaged with a stripe pattern and exhibited an oxygen density twice that of the (2  2)-O structure. The reaction kinetics of CO titration was then measured on these (2  1)-O islands between 144 and 185 K. The (2  1)-O phase shows a superior reactivity over the (2  2)-O phase that does not

3.3 Visualizing the Pathway of Catalytic Reactions

Figure 3.13 Reaction rates, determined from the change in the size of the (2  2) area between successive panels of the data of Figure 3.12, normalized to (squares) the length of the boundary between oxygen and CO domains (the full line is a linear fit) and

(crosses) divided by qO (1
qO), which is equal to qOqCO if q ¼ 1 implies maximum coverage of the respective phase (the broken line is only to guide the eye). (Reprinted with permission from Ref. [58]. Copyright 1997, The American Association for the Advancement of Science.)

react with CO up to 180 K. More interestingly, the removal of oxygen islands accelerated when the surface coverage of (2  1)-O islands decreased to below 0.3 ML (Figure 3.15a). Below 0.3 ML, unlike the previous study on the Pt(1 1 1) surface, the titration reaction rate with the (2  1)-O islands on Pd(1 1 1) shows a linear relation with the surface area of oxygen, instead of the perimeter of oxygen islands, as shown in Figure 3.15b and c. The authors speculated that a transient occupation of CO on the oxygen island causes all O atoms to be accessible for the reaction. There was no direct evidence for the existence of this kind of mixed O/CO phase based on the STM images or other spectroscopic studies. Nonetheless, this study unambiguously illustrated the superior reactivity of compressed oxygen islands, especially when they become very small. The adsorption of oxygen atoms often induces the reconstruction of metal surfaces as is the case of (1 1 0) surfaces of fcc metals. It is expected that CO titration on such surfaces would also involve the local transformation of metal substrates. Indeed, accompanying the oxygen-induced reconstruction, the mobility of surface oxygen is considerably reduced so that they can be resolved by STM at room temperature. For this reason, the CO titration experiments using STM were initiated on (1 1 0) surfaces of fcc metals in the early 1990s. CO oxidation was first visualized on a Rh(1 1 0) surface by Leibsle et al. [60] where a pronounced reaction anisotropy was observed. The experiments were carried out by titrating the oxygen precovered Rh(1 1 0) surface with CO. Chemisorption of oxygen on Rh(1 1 0) forms several reconstructed phases, which, in turn, were imaged as striped patterns along the ½1  1 0direction. By monitoring the surface changes during CO exposure, STM images revealed that oxygen was removed on the elongated stripes of the added rows in the ½1 1 0direction. Later, similar one-dimensional reactivity was also found on other fcc(1 1 0) systems, such as Cu [61–63], Ni [64], and Ag [65, 66].

j75

j 3 In Situ STM Studies of Model Catalysts

76

Figure 3.14 Series of STM images recorded during CO dosing on the (2  2)-O-covered Pd(1 1 1)surface. T ¼ 143 K, PCO ¼ 2  10
8 Torr, all images are from the same area. Indicated is the time elapsed since the start of the CO

dosing. Vt ¼ 0.3 V, It ¼ 2.2 nA, 240 Å  240 Å. The two close-ups in (d) show details from the marked areas in frames (a) and (c). (Reprinted with permission from Ref. [59]. Copyright 2005, The American Physical Society.)

While most in situ studies on the one-dimensional reactivity of fcc(1 1 0) metals remain qualitative, recent studies on the reactivity of oxygen-induced added rows of Ag(1 1 0) have provided quantitative measurement of one-dimensional reactivity. Nakagoe et al. [65, 66] conducted CO titration experiments on the added rowreconstructed Ag(1 1 0)(n  1)-O surfaces, where one-dimensional
Ag
O
chains arrange periodically along the [0 0 1] direction. Figure 3.16 shows two series of in situ STM images, where Ag(1 1 0)(2  1)-O surfaces were exposed to 1  10
8 Torr CO at room temperature. As a function of time, Figure 3.16a–g depicts the continuous segmentation of AgO chains on the clean or carbon-containing Ag(1 1 0)(2  1)-O surfaces. Figure 3.16h plots the remaining coverage of surface oxygen as a function of CO exposure. Clearly, the reaction rate accompanying the segmentation of AgO chains is significantly accelerated. The authors have gone further to study the structure fluctuation at various temperatures. Below 230 K, the AgO chains were found to be straight while the removal of AgO chains only occurs at the end of the

3.3 Visualizing the Pathway of Catalytic Reactions

Figure 3.15 (a) Time evolution of the (2  1)-O coverage; (b) plot of ln[
dq(21)/dt] versus ln q(21) for the data from (a) for t  0; (c) plot of ln q(21) versus t for the data from (a) for t  0. (Reprinted with permission from Ref. [59]. Copyright 2005, The American Physical Society.)

j77

j 3 In Situ STM Studies of Model Catalysts

78

Figure 3.16 Two series of STM images of 37 nm  27 nm continuously taken at RT under a nominal CO pressure of 1  10
8 Torr for clean (a–d) and C-containing (e–g) Ag(1 1 0) (2  1)-O surfaces (It ¼ 0.2 nA, Vtip ¼ 1.4 V). Schematic models of the regions are also shown for (a–d). (h) Titration curves obtained for both clean (red solid circles) and C-containing

(empty circles) Ag(1 1 0)(2  1)-O surfaces. Thick red and black curves are the least square fits obtained by assuming second-order kinetics. The relative number of segments for the clean surface is also plotted (blue triangles and curve). (Reprinted with permission from Ref. [65]. Copyright 2003, The American Physical Society.)

chains and exhibits zero-order kinetics to CO exposure. The Arrhenius plot gives an activation barrier of 41 kJ mol
1 and a prefactor of 1.7  103 cm
2 s
1. The results below 230 K agree with the previous study by Wintterlin et al. on Pt(1 1 1), where the reaction takes place at the periphery of oxygen domains. On the contrary, at room temperature, the reaction rate is drastically accelerated as the AgO chains become segmented and the shape of the AgO chains begins to fluctuate. It is clear there is a direct correlation between the surface structure and the reaction kinetics, although such correlation cannot be quantitatively described by the first- or second-order kinetic models. The nonlinear kinetics of CO titration on oxygen precovered surface has been found to take place not only on the oxygen-reconstructed one-dimensional wires but also on the two-dimensional surface oxides. Klust and Madix [67] have recently studied the reduction of the Ag(1 1 1)-p(4  4)-O surface by CO titration. The Ag (1 1 1)-p(4  4)-O surface was prepared by exposing Ag(1 1 1) to NO2 at 500 K. The reduction of this surface was monitored at room temperature by STM in the presence of 10
8 mbar CO (Figure 3.17). With time, the surface areas covered by p(4  4)-O continue to shrink while the bright (1  1) islands continue to grow on top of the p (4  4)-O overlayer. Figure 3.18a–c illustrates the atomic structures of p(4  4)-O phase, the oxygen-free (1  1) Ag islands, and the remnant dots of Ag surface oxide. The authors found that the reaction rate does not correlate with the perimeter of the

3.3 Visualizing the Pathway of Catalytic Reactions

Figure 3.17 Evolution of the Ag(1 1 1)-p(4  4)O surface during exposure to CO at room temperature. The images show the same surface area exposed to increasing amounts of CO during imaging. (a) Shows the surface before CO exposure and (b), (c), and (d) at 15,

30, and 45 min after exposure start, respectively. Image (d) shows the final state of the surface, no changes were observed after 45 min exposure to CO. (Reprinted with permission from Ref. [67]. Copyright 2007, The American Institute of Physics.)

boundary layers but increases more rapidly with CO exposure (Figure 3.18d). The nonlinear increase in the reaction rate approximately scales with the reacted area and led the authors to speculate that CO reacts with undercoordinated oxygen atoms, either at the boundary between the Ag surface and the p(4  4)-O phase or with oxygen atoms released onto the Ag surface. Due to the invisibility of such species, speculation of this kind is difficult to verify. In situ CO titration experiments have also been conducted on multicomposition systems, that is, inverse model catalyst. Schoiswohl et al. [68] in their studies compared the CO titration reaction on three surfaces: clean Rh(1 1 1) surface, Rh (1 1 1) surface covered with large 2D V3O9 islands (mean size >50 nm), and Rh(1 1 1) surface covered with small 2D V3O9 islands (mean size <15 nm). Prior to CO titration, the three surfaces were exposed to 10
7 mbar O2 to form a (2  1)-O phase at room temperature. In situ STM was used to follow the titration reaction in the presence of 10
8–10
7 mbar CO. CO titration on the clean Rh(1 1 1) surface or the Rh(1 1 1) surface with large V3O9 islands exhibits similar reaction kinetics. Figure 3.19 shows

j79

j 3 In Situ STM Studies of Model Catalysts

80

Figure 3.18 (a) Ag(1 1 1) islands and pits surrounded by the Ag(1 1 1)-p(4  4)-O structure. The white arrow points to the remnants of the surface oxide that are occasionally observed in the pits. The atomically resolved STM images show (b) the Ag(1 1 1)-p(4  4)-O surface and (c) a small area of the (1  1) structure obtained on the island shown in (a) that appeared during CO oxidation. The black

square on the island shown in (a) marks the approximate scan area of image (c). (d) Development of the reacted surface area during the titration reaction. The curve shows an exponential function fitted to the data. (Reprinted with permission from Ref. [67]. Copyright 2007, The American Institute of Physics.)

the adsorption of CO on the Rh(1 1 1)-(2  1)-O surface occupies the on-top sites and reacts with half of the oxygen in the (2  1)-O phase, leading to the formation of a coadsorbed (2  2) O þ CO phase. The islands of (2  2) O þ CO phase grow at the expense of the (2  1)-O layer upon CO exposure. The titration reaction stops at approximately 30 L of CO and removes half of the surface oxygen atoms. Further removal of the adsorbed oxygen is kinetically inhibited on these two surfaces at room temperature. In contrast, CO titration reaction on the Rh(1 1 1) surface covered with small V3O9 islands is significantly accelerated and could proceed further to remove all chemisorbed oxygen atoms on the Rh(1 1 1) surface (Figure 3.20). The fuzzy edges of the V3O9 islands in Figure 3.20 suggest participation of the periphery of small V3O9 islands in CO oxidation via promotion of the CO oxidation reaction at the metal-oxide phase boundary. In summary, in situ STM studies of CO titration on the oxygen precovered metal surfaces have demonstrated atomic details of CO oxidation on metal surfaces and have shown excellent agreement with macroscopic kinetic measurements. Moreover, in situ studies have revealed an interesting but not well-understood, nonlinear behavior of reaction kinetics. The accelerated reaction rate observed takes place only when surface oxygen islands, either compressed oxygen islands or surface oxide islands, are reduced to the nanometer size. The nonlinear reactivity of these nanoislands is in stark contrast with the large adsorbate layer and requires further investigations.

3.4 Metal Surfaces at High Pressures

Figure 3.19 Series of STM images (25 nm  25 nm, Vt ¼ 1.5 V, It ¼ 0.1 nA) recorded during dosing CO on the (2  1)O-Rh(1 1 1) surface covered with large (mean size of 50 nm) V3O9 islands: (a) 1 L; (b) 8 L; (c) 15 L; (d) 22 L; (e) 30 L; (f) 600 L. A fraction of a large V3O9 island is seen in the upper right-hand side of the image. (Reprinted with permission from Ref. [68]. Copyright 2005, Elsevier.)

3.4 Metal Surfaces at High Pressures

The above studies show that the chemisorptions on metals could often alter the composition and structure of metal surfaces. To bridge the pressure gap, in situ STM has played a critical role in observing the dynamic behavior of catalytic surfaces from UHV to atmospheric pressures. The pioneering high-pressure STM study by McIntyre et al. [69] shows the Pt(1 1 0) surface restructures in single-component gases of H2, O2, and CO at atmospheric pressures and at 425 K. Hendriksen et al. [22, 23] took one step further to view this surface using a flow-reactor STM under high-pressure CO or a CO/O2 mixture. The surface of Pt(1 1 0) exhibits a (1  2) missing-row reconstruction in UHV. The exposure of CO lifts this reconstruction even at low pressures. In the presence of 1 bar CO, in situ STM shows the exposure of high-pressure CO not only lifts the (1  2) reconstruction to a bulk-like (1  1) phase but also causes the coarsening of Pt(1 1 0) surface. Figure 3.21 shows sequences of snapshots on the Pt(1 1 0) surface in 1.25 bar CO at 425 K. The local rearrangement caused by the transition from the (1  2) to (1  1) phase upon CO exposure leads to the fragmentation of Pt(1 1 0) terraces and a high

j81

j 3 In Situ STM Studies of Model Catalysts

82

Figure 3.20 Series of STM images (25 nm  25 nm, Vt ¼ 1.5 V, It ¼ 0.1 nA) recorded during dosing CO on the (2  1)O-Rh (1 1 1) surface covered with small (mean size of 15 nm) V3O9 islands: (a) 1 L; (b) 21 L; (c) 31 L; (d) 108 L; (e) 204 L; and (f) 270 L. Several small

irregular shaped V3O9 islands phase are visible. The areas labeled (2  2)-A, (2  2)-B, and COsat correspond to the (2  2)-O þ CO, (2  2)-CO, and CO saturation layer phases, respectively. (Reprinted with permission from Ref. [68]. Copyright 2005, Elsevier.)

density of surface steps (Figure 3.21a). To reduce the total surface energy, coarsening of the surface steps takes place (Figure 3.21b–f) whereas a much slower (1  1) phase transition occurs upon CO exposure. With time the curved Pt islands are smoothed, forming rounded large islands covered by CO on the Pt(1 1 0) surface. The CO-covered Pt(1 1 0) surface was then exposed to a mixture of CO/O2 gases, with the ratio of CO/O2 adjusted by flow meters. The pressure changes of CO, O2, and the reaction product, CO2, were monitored by leaking the gases from the flow reactor to a quadrupole mass spectrometer (QMS) attached to the flow-reactor STM. The surface structure and reactivity of Pt(1 1 0) could be measured simultaneously with the combination of STM and QMS. Figure 3.22 plots the real-time pressure of CO, O2, and CO2, as well as the corresponding snapshots of in situ STM images on Pt(1 1 0). The production of CO2 starts right away with the presence of both CO and O2 in the reactor. As the reaction proceeds, two stages of reaction rates exist, as evidenced by the step increase in CO2 pressure in Figure 3.22. At the stage of low reaction rate (see B, F, and H in Figure 3.22), the corresponding STM images show no apparent change in the surface structure, indicating the Pt(1 1 0) surface remains metallic. At the stage of high reaction rate (see D and G in Figure 3.22, about three times higher than the low reaction rate), the corresponding STM images suggest a roughened surface,

3.4 Metal Surfaces at High Pressures

Figure 3.21 Series of STM snapshots (140 nm  140 nm) taken on Pt(1 1 0), starting immediately after introduction of 1.25 bar CO in the reactor-STM at 425 K. The “tiger skin” pattern in the first panel shows that the (1  2) to (1  1) transition has divided the surface in two levels, each 50%, and a high density of steps. Subsequent images show the progressive

reduction of the step density by coarsening of the step pattern. The elapsed time in minutes is indicated in each panel. The two ball models indicate the atomic-scale geometries characteristic for the starting and end situations. (Reprinted with permission from Ref. [22]. Copyright 2005, Springer.)

indicating the formation of surface Pt oxide. At the point of the step increase in reaction rate (Figure 3.22C), the STM image suggests a modest and uniform increase in surface roughness, which the authors assigned as a commensurate Pt oxide film. Subsequent high-pressure surface X-ray diffraction (SXRD) studies by the same group [70] verified the formation of this commensurate oxide film, which was assumed to be responsible for the increased reaction rate. Further SXRD results suggest that the thickness of this oxide film is one monolayer with a (1  2) periodicity. It is worth noting that roughening of the oxide surface was observed only during the CO oxidation reaction but not in 1 bar pure O2. Combined with their kinetic measurements, the authors proposed CO from the gas phase could directly react with oxygen atoms in the surface oxides, accounting for relatively high reactivity of this phase for CO oxidation. This mechanism, termed as “Mars-Van Krevelen mechanism,” challenges the general concept that CO oxidation on Pt group metals is dominated by the Langmuir–Hinshelwood mechanism, which proceeds via (1) the adsorption of CO and the dissociative adsorption of O2 and (2) surface diffusion of COad and Oad atoms to ultimately form CO2. The authors further tested the Pt(1 1 1) and Pd(1 1 0) surfaces [71, 72] using in situ STM and SXRD. All these single crystals show a similar kinetic behavior in CO oxidation. The gradual roughening of the surface corresponds to the formation of surface oxides and a higher CO oxidation rate. The structure insensitivity observed at high pressure is in contrast with the results obtained in UHV, where the reactivity shows a strong orientational dependence.

j83

j 3 In Situ STM Studies of Model Catalysts

84

Figure 3.22 QMS signals and STM images simultaneously measured during CO oxidation on the Pt(1 1 0) surface at a temperature of 425 K in a 3.0 ml min
1 flow of mixtures of CO and/or O2 at 0.5 bar. (Upper panel) QMS signals of O2, CO, and CO2, measured directly from the reactor cell. Labels (A)–(H) correspond to the STM images in the lower panel. Rlow and Rhigh denote the low and high CO2 production rate branches. Pth indicates the threshold value of the CO pressure at which the rate switched from Rlow to Rhigh and the surface changed simultaneously from smooth to rough. (Lower panel) STM images (210 nm  210 nm) from an

STMmovie(65 s/image,It ¼ 0.2 nA,Vt ¼ 80 mV). The images were differentiated to enhance the contrast. Images (A), (B), (E), (F), and (H) show flat terraces separated by steps of the Pt lattice. This corresponds to the metallic, CO-covered surface. Image (C) shows the change in the surface accompanying the step in activity at t ¼ 2109 s (the image was built up from bottom to top). Images (D) and (G) show the rough surface consisting of protrusions, with heights of 0.2–0.4 nm, and pits (see inset). (Reprinted with permission from Ref. [23]. Copyright 2002, The American Physical Society.)

Due to the increased complexity at high pressure (e.g., the impinging reactant gases are no longer under molecular flow, causing the gas environment near the surface to be markedly different from the ambient gases), much more work is required before establishing a firm correlation between the surface structure and the surface reactivity. The difficulty in using STM to measure the reaction rate directly

3.5 In Situ Studies of Supported Model Catalysts

can lead to uncertainties in correlating local surface structure with the measured global reaction rate. STM measurements during high-pressure reaction present challenges for resolving the surface structures at high resolution. Rapid diffusion of surface adsorbates further prevents precise characterization of surface-active species. In this case, the application and development of the fast-scanning technique is essential and could provide more precise surface information during the high-pressure reaction. Nonetheless, the above study demonstrates the necessity of expanding high-pressure STM studies in model catalysis. The effect of high pressure intuitively is equivalent to the effect of lowering the temperature. Low-temperature UHV experiments might reproduce the same adsorbate structures as do high pressures. However, due to the restricted diffusion of metal atoms at low temperature, the dynamic change of surface structures under catalytically realistic conditions could not be observed in the UHV experiments.

3.5 In Situ Studies of Supported Model Catalysts

In an effort to bridge the material gap, in situ STM studies of supported model catalysts have emerged in recent years. Among supported model catalysts, Au clusters supported on TiO2(1 1 0) are the most investigated model system by in situ STM. The tremendous interest in studying this model system originates from the remarkable catalytic reactivity of supported Au clusters discovered a few decades ago [73–80]. It was generally believed that Au, the noblest metal, is inert as a catalyst. The TiO2 support alone also exhibits limited catalytic reactivity. In contrast, Au clusters in the 2–4 nm size range supported on TiO2 exhibit a remarkable catalytic reactivity for low-temperature CO oxidation and selective hydrogenation reactions. Although the cause of the extraordinary activity of Au clusters is still under debate, supported model Au catalysts are playing a key role in the development of our understanding of the relative importance of cluster morphology and the cluster support. In this section, we highlight a few recent in situ STM studies in monitoring the synthesis of supported model catalysts, revealing the mechanism of the “Strong metal support interaction” (SMSI) effect and measuring the sintering kinetics of supported Au catalysts. 3.5.1 Monitoring the Growth Kinetics of Supported Metal Catalysts

Supported model catalysts are frequently prepared by thermally evaporating metal atoms onto a planar oxide surface in UHV. The morphology and growth of supported metal clusters depend on a number of factors such as substrate morphology, the deposition rate, and the surface temperature. For a controlled synthesis of supported model catalysts, it is necessary to monitor the growth kinetics of supported metal

j85

j 3 In Situ STM Studies of Model Catalysts

86

Figure 3.23 STM images of the same area of a TiO2(1 1 0) surface after the deposition of (a) 0.17 ML Au; (b) 0.34 ML Au; (c) 0.51 ML Au; (d) 0.69 ML Au; (e) 0.86 ML Au; and (f) 1.3 ML Au. All images have the same dimensions of 100 nm  100 nm. The circle highlighted in white in each image indicates the identical area. (Reprinted with permission from Ref. [81]. Copyright 2003, The Japan Society of Applied Physics.)

clusters. Figure 3.23 shows a series of STM images as a function of increased Au coverage obtained on the same area of a TiO2(1 1 0) surface at room temperature [81]. At the very early stages of growth, Au clusters preferentially decorate the step edges (Figure 3.23a). With an increase in Au coverage, the prevailing role of the step edges as major nucleation sites decreases, whereas new clusters begin to nucleate on terraces (Figure 3.23b–f). Au clusters grown at step edges grow much faster than those at terraces. A quantitative analysis of Au cluster distribution suggests a bimodal size distribution for the growth of Au clusters at step edges and terraces. In situ STM studies allow the growth mode of Au clusters to be followed on a cluster-to-cluster basis. The results demonstrate that step edges play a dominant role in the growth of Au clusters on TiO2(1 1 0) at room temperature. Using in situ STM also makes it possible to monitor the growth of supported alloy model catalysts. The simplest way to synthesize supported alloy model catalysts is to

3.5 In Situ Studies of Supported Model Catalysts

Figure 3.24 STM images (100 nm  100 nm) during Au deposition on an Ag precovered (0.08 ML) TiO2(1 1 0) surface (a) after deposition of (b) 0.17 ML Au, (c) 0.85 ML Au; or on an Ag precovered (0.033 ML) TiO2(1 1 0) surface (d) after deposition of (e) 0.17 ML Au, (f) 0.51 ML Au. The white circles in image (d) and (e) show the appearance of new Au clusters. (Reprinted with permission from Ref. [82]. Copyright 2004, Elsevier B.V.)

evaporate both metal atoms in UHV directly onto an oxide support surface. However, this method does not give information on whether the two metals mix and form alloy clusters or separate to form two types of monometallic clusters. In situ STM allows this problem to be addressed by investigating the growth kinetics of supported alloy catalysts. In our study, Ag–Au alloy clusters of varying ratios were synthesized on TiO2(1 1 0) and the growth kinetics for the clusters determined [82]. To understand the growth of Au clusters in the presence of Ag clusters, sequential in situ STM measurements were carried out as a function of Au coverage on a Ag precovered TiO2(1 1 0) surface. A particular precoverage of 0.08 ML Ag (Figure 3.24a) was chosen to essentially saturate the step edge sites with Ag clusters, allowing the investigation of Au growth in the absence of their preferred nucleation sites. The series of in situ STM images

j87

j 3 In Situ STM Studies of Model Catalysts

88

(Figure 3.24a–c) show a preferential growth of Au on the existing Ag clusters. As the Au coverage is increased, new Au clusters begin to appear on the surface. In another set of experiments, the TiO2(1 1 0) surface was precovered with less Ag so that a fraction of empty step edge sites remain open for the nucleation of new Au clusters (Figure 3.24d). The series of in situ STM images (Figure 3.24d–f) show even a small amount of Au deposition (0.17 ML) leads to the nucleation of new Au clusters. These systematic experiments of Au growth on TiO2(1 1 0) precovered with Ag clusters show the versatility of in situ STM experiments in controlling the synthesis of supported alloy model catalysts. Simply by blocking step edge sites to a varying extent, either monometallic clusters or bimetallic clusters could be selectively synthesized and monitored. By changing the sequence of metal deposition, the surface structure of alloy clusters can also be altered. Adopting the same in situ approach, the synthesis of supported alloy model catalysts has been extended to Pd/Au clusters on TiO2(1 1 0) [83] and Pt/Rh clusters on TiO2(1 1 0) [84]. It is expected that this method will be generally adopted for the preparation and characterization of supported model catalysts. 3.5.2 Studies of the SMSI Effect

In the late 1970s, Tauster et al. [85–87] discovered the unusual properties of group VIII metal, for example, Pt and Ir, when supported on TiO2 and reduced at relatively high temperatures. The Pt and Ir clusters supported on TiO2 show a suppressed CO and H2 chemisorption and an increased methanation catalytic activity. “Strong metal support interaction” has since been introduced to describe the unusual properties of group VIII metal supported on reducible oxides such as TiO2. A few reasons for the SMSI effect have been postulated, including alloy formation, altered electronic effects at the interface, or encapsulation of supported metal clusters. Bowker et al. [88, 89] have investigated the SMSI effect using in situ STM. As shown in Figure 3.25, Pd clusters with a mean size of approximately 4 nm were supported on TiO2(1 1 0) and heated to 673 K. The surface was then exposed to 5  10
8 mbar O2 and monitored by STM as a function of time. In the presence of oxygen, the reoxidation of the TiO2(1 1 0) surface proceeds. Since the dissociation probability of O2 on Pd is about three magnitudes higher than on TiO2(1 1 0), O2 dissociation mainly occurs on supported Pd clusters. The dissociated oxygen atoms subsequently diffuse onto the adjacent surface, that is, spillover, and coalesce with Ti interstitials from the bulk, causing the enrichment of TiOx species surrounding the Pd clusters. The regrowth of TiO2 layers thus accelerates at the periphery of Pd clusters. Eventually, the growth of TiO2 layers surpass the height of the Pd clusters and fully encapsulates them within the TiO2 overlayers, that is, in situ STM study verifies that encapsulation is a major contributor to the SMSI effect. This is also the first study to have directly observed in situ the spillover, an elementary step in catalytic process.

3.5 In Situ Studies of Supported Model Catalysts

Figure 3.25 A sequence of STM images taken at 673 K of oxygen spillover from Pd clusters in an ambient pressure of 5  10
8 mbar of O2 (doubled before image (f)). The total exposures to oxygen from image (a)–(f) were (in units of L): 114, 178, 237, 282, 344, and 531. The spillover begins with the formation of one layer

around the Pd clusters, which then becomes multilayer growth, eventually burying the Pd cluster with TiO2. In image (f), a total of seven layers of TiO2 have grown around the Pd on top of the original TiO2 surface. (Reprinted with permission from Ref. [89]. Copyright 2000, Elsevier)

3.5.3 Sintering Kinetics of Supported Au Clusters

Although supported Au catalysts have great potential in technological applications, such as removing CO selectively from hydrogen fuel for fuel cells, a major problem for the commercialization of supported Au catalysts is that these catalysts deactivate rapidly [78, 90, 91]. Similar deactivation has been observed for model catalysts consisting of Au clusters deposited on a reduced TiO2(1 1 0) where rapid deactivation has been shown to correlate with sintering of the Au clusters [91, 92]. To understand the sintering of supported Au catalysts, we have studied the sintering kinetics of Au/ TiO2(1 1 0) in UHV and in the presence of reactant gases. The first issue in sintering kinetics is to determine the primary mass transport process. The sintering of supported nanoclusters can occur in either one or two modes: (a) the migration and coalescence of whole clusters or (b) the migration of monomers (single metal atoms or metal complexes). The first mode, known as cluster migration, entails the migration of whole clusters on the surface that then coalesce with neighboring clusters. In the second mode, known as Ostwald ripening, monomers dissociate from small clusters, diffuse to, and coalesce with large clusters [93, 94].

j89

j 3 In Situ STM Studies of Model Catalysts

90

In situ STM allows one to monitor individual clusters under realistic conditions and to directly determine the mass transport mode controlling the kinetics. Mitchell et al. studied Au clusters supported on TiO2(1 1 0) at approximately 750 K as a function of time [95]. Their STM results show Au clusters are mostly static while some small clusters with size below 3 nm diffuse along the surface. The diffusion of small clusters does not behave in a continuous manner. Instead, the clusters stick for a relatively long time before taking a sudden jump. This unconventional diffusion of clusters, termed as Levy flight, leads to the speculation that Au clusters sinter in the mode of cluster migration. Such results are in agreement with the in situ STM study by Kolmakov et al. [12], who examined the stability of Au/TiO2(1 1 0) after thermal treatment at 950 K. Dramatic morphological changes were observed. Accompanying the decomposition of surface steps, the density of Au clusters decreased by 20%, whereas the volume of the Au clusters increased. Most Au clusters changed their positions on the TiO2(1 1 0) surface. Clearly, cluster migration has played a role in the thermally induced sintering of supported Au clusters under UHV. Interestingly, the above studies were both conducted at the temperature under which the TiO2(1 1 0) substrate becomes unstable and starts to decompose. To fully explore the sintering mode induced by thermal treatment, it is necessary to conduct such in situ experiments at intermediate temperatures where the TiO2(1 1 0) lattice remains stable. The sintering of Au/TiO2(1 1 0) was also studied by our group in the presence of reactant gases [96]. Figure 3.26 shows the morphological changes of 0.5 ML Au clusters supported on TiO2(1 1 0) in the presence of CO/O2 at 300 K. Under UHV conditions, the Au clusters remain immobile and show no detectable change of either cluster size or shape for more than 4 h. Sintering of Au clusters began immediately upon the introduction of a 0.1 Torr CO/O2 gas mixture (1 : 1 ratio). Figure 3.26 shows no apparent change of cluster positions before and after gas exposure; however, the small Au clusters gradually decay whereas the large Au clusters grow in size. The gradual change in the cluster size and static cluster positions are consistent with an Ostwald ripening process. Most Au clusters with a height around 4.6 Å or less (assuming the height of Au clusters of two-layer thickness is 4.6 Å [97]) disappear within 2 h of CO oxidation reaction. Gold clusters supported on TiO2(1 1 0) were also exposed to O2 and CO separately to investigate the influence of the chemisorbed reactants. Minimal changes were observed for 0.5 ML Au clusters supported on TiO2(1 1 0) after an exposure of 0.1 Torr O2 or 3 Torr CO for more than 1.5 h at 300 K. Therefore, a synergetic effect of CO/O2 was found to induce and accelerate the sintering of Au clusters on TiO2(1 1 0). The synergetic effect of a CO and O2 mixture can be explained by a reactioninduced mechanism, which might involve hot electrons generated by CO oxidation [98–100], inducing the detachment of Au monomers from supported Au clusters and initiating the sintering process. On the basis of the sintering model of Ostwald ripening, the activation energy for Au clusters with diameters around 3 nm is estimated to be approximately 10 kJ mol
1, a value that closely tracks the activation energy of CO oxidation on supported Au clusters. The large difference in comparing this number with the activation energy of sintering for supported Au clusters in UHV, approximately 280 kJ mol
1 [101, 102], demonstrates that CO oxidation itself does

3.6 Outlook

Figure 3.26 75 nm  75 nm in situ STM images of 0.5 ML Auclusters supported on the TiO2(1 1 0) surface in the presence of 0.1 Torr CO and O2 mixture at 300 K. (a–f ) are consequently taken at the same surface area. The time intervals are

(a) 0 min; (b) 28 min; (c) 42 min; (d) 63 min; (e) 120 min; and (f) 280 min. Tunneling parameters are Vt ¼ 2 V and It ¼ 0.1 nA. (Reprinted with permission from Ref. [96]. Copyright 2009, The American Chemical Society.)

indeed influence cluster sintering. This effect may very well be general for CO oxidation over other supported metal catalysts.

3.6 Outlook

This chapter reviews the recent progress in in situ STM studies of model catalysts. From revealing reaction pathways to delineating active sites, in situ STM studies in UHV and on extended surfaces have demonstrated their power to solve fundamental questions in catalysis and enhance our understanding of the elementary steps of

j91

j 3 In Situ STM Studies of Model Catalysts

92

surface catalytic processes. Recent efforts in expanding in situ STM studies to high pressures and to supported model catalysts illustrate the need to study model catalysts under realistic conditions. The studies discussed in this chapter have shown surface structures are extremely dynamic under reaction conditions, consistent with the hypothesis promoted by Somorjai [103–107]. The hypothesis, termed as the “flexible surface,” supposes that metal surfaces with low coordinated metal atoms or curved cluster surfaces are highly fluxional during reaction. Those atoms with high diffusivity are responsible for high activity of metal surfaces. To this end, it is critical to compare the timescale and energetics of metal atom diffusion with their catalytic reactivity. With the development of fast scan techniques, in situ STM now provides a great opportunity to test this hypothesis. The studies to date have shown that small can be quite different and less can sometimes mean more. In the coming years, there is no doubt that in situ STM studies on well-defined nanostructures will exponentially increase and contribute invaluably to an atomic-level understanding of reactions catalyzed by solid surfaces.

Acknowledgments

We gratefully acknowledge the support for this work by the Department of Energy (DOE), Office of Basic Energy Sciences, Division of Chemical Sciences, and the Robert A. Welch Foundation.

References 1 Langmuir, I. (1915) Phys. Rev., 6, 79. 2 Bartels, L., Wang, F., Moller, D., Knoesel, E., and Heinz, T.F. (2004) Science, 305, 648. 3 Che, M. and Bennett, C.O. (1989) Adv. Catal., 36, 55. 4 Wintterlin, J., Trost, J., Renisch, S., Schuster, R., Zambelli, T., and Ertl, G. (1997) Surf. Sci., 394, 159. 5 Kuipers, L., Hoogeman, M.S., and Frenken, J.W.M. (1993) Phys. Rev. Lett., 71, 3517. 6 Ludwig, C., Gompf, B., Glatz, W., Petersen, J., Eisenmenger, W., Mobus, M., Zimmermann, U., and Karl, N. (1992) Z. Phys. B, 86, 397. 7 Rost, M.J., Crama, L., Schakel, P., Van Tol, E., Van Velzen-Williams, G.B.E.M., Overgauw, C.F., Horst, H., Dekker, H., Okhuijsen, B., Seynen, M., Vijftigschild, A., Han, P., Katan, A.J., Schoots, K.,

8 9

10

11

12 13

Schumm, R., Van Loo, W., Oosterkamp, T.H., and Frenken, J.W.M. (2005) Rev. Sci. Instrum., 76, 053710. Rossler, M., Geng, P., and Wintterlin, J. (2005) Rev. Sci. Instrum., 76, 023705. Laegsgaard, E., Osterlund, L., Thostrup, P., Rasmussen, P.B., Stensgaard, I., and Besenbacher, F. (2001) Rev. Sci. Instrum., 72, 3537. Rasmussen, P.B., Hendriksen, B.L.M., Zeijlemaker, H., Ficke, H.G., and Frenken, J.W.M. (1998) Rev. Sci. Instrum., 69, 3879. Mcintyre, B.J., Salmeron, M., and Somorjai, G.A. (1993) Rev. Sci. Instrum., 64, 687. Kolmakov, A. and Goodman, D.W. (2002) Chem. Rec., 2, 446. Kolmakov, A. and Goodman, D.W. (2003) Rev. Sci. Instrum., 74, 2444.

References 14 Melmed, A.J. (1991) J. Vac. Sci. Technol. B, 9, 601. 15 Oliva, A.I., Romero, A., Pena, J.L., Anguiano, E., and Aguilar, M. (1996) Rev. Sci. Instrum., 67, 1917. 16 Weinstein, V., Slutzky, M., Arenshtam, A., and Benjacob, E. (1995) Rev. Sci. Instrum., 66, 3075. 17 Ren, B., Picardi, G., and Pettinger, B. (2004) Rev. Sci. Instrum., 75, 837. 18 Fried, G.A., Wang, X.D., and Hipps, K.W. (1993) Rev. Sci. Instrum., 64, 1495. 19 Iwami, M., Uehara, Y., and Ushioda, S. (1998) Rev. Sci. Instrum., 69, 4010. 20 Schroder, U., Mcintyre, B.J., Salmeron, M., and Somorjai, G.A. (1995) Surf. Sci., 333, 337. 21 Reichelt, R., Gunther, S., Rossler, M., Wintterlin, J., Kubias, B., Jakobi, B., and Schlogl, R. (2007) Phys. Chem. Chem. Phys., 9, 3590. 22 Hendriksen, B.L.M., Bobaru, S.C., and Frenken, J.W.M. (2005) Top. Catal., 36, 43. 23 Hendriksen, B.L.M. and Frenken, J.W.M. (2002) Phys. Rev. Lett., 89, 046101. 24 Wintterlin, J. (2000) Adv. Catal., 45, 131. 25 Hahn, J.R. and Ho, W. (2001) Phys. Rev. Lett., 87, 166102. 26 Hla, S.W. and Rieder, K.H. (2003) Annu. Rev. Phys. Chem., 54, 307. 27 Rieder, K.H., Meyer, G., Hla, S.W., Moresco, F., Braun, K.F., Morgenstern, K., Repp, J., Foelsch, S., and Bartels, L. (2004) Philos. Trans. Roy. Soc., 362, 1207. 28 Wong, K.L., Rao, B.V., Pawin, G., Ulin-Avila, E., and Bartels, L. (2005) J. Chem. Phys., 123, 201102. 29 Longwitz, S.R., Schnadt, J., Vestergaard, E.K., Vang, R.T., Laegsgaard, E., Stensgaard, I., Brune, H., and Besenbacher, F. (2004) J. Phys. Chem. B, 108, 14497. 30 Vestergaard, E.K., Thostrup, P., An, T., Laegsgaard, E., Stensgaard, I., Hammer, B., and Besenbacher, F. (2002) Phys. Rev. Lett., 88, 259601. 31 Osterlund, L., Rasmussen, P.B., Thostrup, P., Laegsgaard, E., Stensgaard,

32

33

34

35 36

37 38

39

40

41

42

43

44

45

I., and Besenbacher, F. (2001) Phys. Rev. Lett., 86, 460. Thostrup, P., Vestergaard, E.K., An, T., Laegsgaard, E., and Besenbacher, F. (2003) J. Chem. Phys., 118, 3724. Vang, R.T., Wang, J.G., Knudsen, J., Schnadt, J., Laegsgaard, E., Stensgaard, I., and Besenbacher, F. (2005) J. Phys. Chem. B, 109, 14262. Mitsui, T., Rose, M.K., Fomin, E., Ogletree, D.F., and Salmeron, M. (2003) Nature, 422, 705. Salmeron, M. (2005) Top. Catal., 36, 55. Mitsui, T., Rose, M.K., Fomin, E., Ogletree, D.F., and Salmeron, M. (2003) Surf. Sci., 540, 5. Diebold, U. (2003) Surf. Sci. Rep., 48, 53. Diebold, U., Lehman, J., Mahmoud, T., Kuhn, M., Leonardelli, G., Hebenstreit, W., Schmid, M., and Varga, P. (1998) Surf. Sci., 411, 137. Schaub, R., Thostrup, R., Lopez, N., Laegsgaard, E., Stensgaard, I., Norskov, J.K., and Besenbacher, F. (2001) Phys. Rev. Lett., 87, 266104. Wendt, S., Schaub, R., Matthiesen, J., Vestergaard, E.K., Wahlstrom, E., Rasmussen, M.D., Thostrup, P., Molina, L.M., Laegsgaard, E., Stensgaard, I., Hammer, B., and Besenbacher, F. (2005) Surf. Sci., 598, 226. Bikondoa, O., Pang, C.L., Ithnin, R., Muryn, C.A., Onishi, H., and Thornton, G. (2006) Nat. Mater., 5, 189. Wendt, S., Matthiesen, J., Schaub, R., Vestergaard, E.K., Laegsgaard, E., Besenbacher, F., and Hammer, B. (2006) Phys. Rev. Lett., 96, 066107. Zhang, Z., Bondarchuk, O., Kay, B.D., White, J.M., and Dohnalek, Z. (2006) J. Phys. Chem. B, 110, 21840. Li, S.C., Zhang, Z., Sheppard, D., Kay, B.D., White, J.M., Du, Y., Lyubinetsky, I., Henkelman, G., and Dohnalek, Z. (2008) J. Am. Chem. Soc., 130, 9080. Zhang, Z.R., Bondarchuk, O., White, J.M., Kay, B.D., and Dohnalek, Z. (2006) J. Am. Chem. Soc., 128, 4198.

j93

j 3 In Situ STM Studies of Model Catalysts

94

46 Zhang, Z.R., Bondarchuk, E., Kay, B.D., White, J.M., and Dohnalek, Z. (2007) J. Phys. Chem. C, 111, 3021. 47 Du, Y.G., Dohnalek, Z., and Lyubinetsky, I. (2008) J. Phys. Chem. C, 112, 2649. 48 Zhang, Z., Ge, Q., Li, S.C., Kay, B.D., White, J.M., and Dohnalek, Z. (2007) Phys. Rev. Lett., 99, 126105. 49 Lyubinetsky, I., Yu, Z.Q., and Henderson, M.A. (2007) J. Phys. Chem. C, 111, 4342. 50 Wendt, S., Sprunger, P.T., Lira, E., Madsen, G.K.H., Li, Z.S., Hansen, J.O., Matthiesen, J., Blekinge-Rasmussen, A., Laegsgaard, E., Hammer, B., and Besenbacher, F. (2008) Science, 320, 1755. 51 Henderson, M.A., Epling, W.S., Peden, C.H.F., and Perkins, C.L. (2003) J. Phys. Chem. B, 107, 534. 52 Smith, R.D., Bennett, R.A., and Bowker, M. (2002) Phys. Rev. B, 66, 7. 53 Park, K.T., Pan, M., Meunier, V., and Plummer, E.W. (2007) Phys. Rev. B, 75, 245415. 54 Onishi, H. and Iwasawa, Y. (1996) Phys. Rev. Lett., 76, 791. 55 Sachs, C., Hildebrand, M., Volkening, S., Wintterlin, J., and Ertl, G. (2001) Science, 293, 1635. 56 Volkening, S., Bedurftig, K., Jacobi, K., Wintterlin, J., and Ertl, G. (1999) Phys. Rev. Lett., 83, 2672. 57 Imbihl, R. and Ertl, G. (1995) Chem. Rev., 95, 697. 58 Wintterlin, J., Volkening, S., Janssens, T.V.W., Zambelli, T., and Ertl, G. (1997) Science, 278, 1931. 59 Kim, S.H., Mendez, J., Wintterlin, J., and Ertl, G. (2005) Phys. Rev. B, 72, 155414. 60 Leibsle, F.M., Murray, P.W., Francis, S.M., Thornton, G., and Bowker, M. (1993) Nature, 363, 706. 61 Leibsle, F.M., Francis, S.M., Haq, S., and Bowker, M. (1994) Surf. Sci., 318, 46. 62 Leibsle, F.M., Francis, S.M., Davis, R., Xiang, N., Haq, S., and Bowker, M. (1994) Phys. Rev. Lett., 72, 2569. 63 Crew, W.W. and Madix, R.J. (1996) Surf. Sci., 349, 275.

64 Ruan, L., Stensgaard, I., Laegsgaard, E., and Besenbacher, F. (1994) Surf. Sci., 314, L873. 65 Nakagoe, O., Watanabe, K., Takagi, N., and Matsumoto, Y. (2003) Phys. Rev. Lett., 90, 226105. 66 Nakagoe, O., Watanabe, K., Takagi, N., and Matsumoto, Y. (2005) J. Phys. Chem. B, 109, 14536. 67 Klust, A. and Madix, R.J. (2007) J. Chem. Phys., 126, 084707. 68 Schoiswohl, J., Eck, S., Ramsey, M.G., Andersen, J.N., Surnev, S., and Netzer, F.P. (2005) Surf. Sci., 580, 122. 69 Mcintyre, B.J., Salmeron, M., and Somorjai, G.A. (1993) J. Vac. Sci. Tech., 11, 1964. 70 Ackermann, M.D., Pedersen, T.M., Hendriksen, B.L.M., Robach, O., Bobaru, S.C., Popa, I., Quiros, C., Kim, H., Hammer, B., Ferrer, S., and Frenken, J.W.M. (2005) Phys. Rev. Lett., 95, 255505. 71 Hendriksen, B.L.M., Bobaru, S.C., and Frenken, J.W.M. (2004) Surf. Sci., 552, 229. 72 Hendriksen, B.L.M., Bobaru, S.C., and Frenken, J.W.M. (2005) Catal. Today, 105, 234. 73 Freund, H.J., Baumer, M., and Kuhlenbeck, H. (2000) Adv. Catal., 45, 333. 74 Henry, C.R. (1998) Surf. Sci. Rep., 31, 235. 75 Campbell, C.T. (1997) Surf. Sci. Rep., 27, 1. 76 Rainer, D.R. and Goodman, D.W. (1998) J. Mol. Catal. A, 131, 259. 77 Campbell, C.T., Grant, A.W., Starr, D.E., Parker, S.C., and Bondzie, V.A. (2001) Top. Catal., 14, 43. 78 Haruta, M. and Date, M. (2001) Appl. Catal. A, 222, 427. 79 Corti, C.W., Holliday, R.J., and Thompson, D.T. (2007) Top. Catal., 44, 331. 80 Bond, G.C. and Thompson, D.T. (1999) Catal. Rev., 41, 319. 81 Santra, A.K., Kolmakov, A., Yang, F., and Goodman, D.W. (2003) Jpn. J. Appl. Phys. Part 1, 42, 4795.

References 82 Santra, A.K., Yang, F., and Goodman, D.W. (2004) Surf. Sci., 548, 324. 83 Han, P. and Goodman, D.W. (2008) J. Phys. Chem. C, 112, 6390. 84 Park, J.B., Ratliff, J.S., Ma, S., and Chen, D.A. (2006) Surf. Sci., 600, 2913. 85 Tauster,S.J. (1987) Acc.Chem. Res., 20, 389. 86 Tauster, S.J., Fung, S.C., Baker, R.T.K., and Horsley, J.A. (1981) Science, 211, 1121. 87 Tauster, S.J., Fung, S.C., and Garten, R.L. (1978) J. Am. Chem. Soc., 100, 170. 88 Bowker, M. (2007) Phys. Chem. Chem. Phys., 9, 3514. 89 Bowker, M., Bowker, L.J., Bennett, R.A., Stone, P., and Ramirez-Cuesta, A. (2000) J. Mol. Catal. A, 163, 221. 90 Haruta, M. (1997) Catal. Today, 36, 153. 91 Valden, M., Lai, X., and Goodman, D.W. (1998) Science, 281, 1647. 92 Lai, X., St Clair, T.P., Valden, M., and Goodman, D.W. (1998) Prog. Surf. Sci., 59, 25. 93 Chakraverty, B.K. (1967) J. Phys. Chem. Sol., 28, 2401. 94 Wynblatt, P. and Gjostein, N.A. (1975) Prog. Solid State Chem., 9, 21.

95 Mitchell, C.E.J., Howard, A., Carney, M., and Egdell, R.G. (2001) Surf. Sci., 490, 196. 96 Yang, F., Chen, M.S., and Goodman, D.W. (2008) J. Phys. Chem. C, 113, 254. 97 Cosandey, F. and Madey, T.E. (2001) Surf. Rev. Lett., 8, 73. 98 Ji, X.Z., Zuppero, A., Gidwani, J.M., and Somorjai, G.A. (2005) Nano Lett., 5, 753. 99 Ji, X.Z., Zuppero, A., Gidwani, J.M., and Somorjai, G.A. (2005) J. Am. Chem. Soc., 127, 5792. 100 Ji, X.Z. and Somorjai, G.A. (2005) J. Phys. Chem. B, 109, 22530. 101 Parker, S.C. and Campbell, C.T. (2007) Phys. Rev. B, 75, 035430. 102 Campbell, C.T., Parker, S.C., and Starr, D.E. (2002) Science, 298, 811. 103 Somorjai, G.A. (1991) Langmuir, 7, 3176. 104 Somorjai, G.A. (1992) Catal. Lett., 12, 17. 105 Somorjai, G.A. (1994) Annu. Rev. Phys. Chem., 45, 721. 106 Somorjai, G.A. (1996) J. Mol. Catal. A, 107, 39. 107 Somorjai, G.A. and Rupprechter, G. (1998) J. Chem. Educ., 75, 161.

j95

j97

4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis Gilberto Teobaldi, Haiping Lin, and Werner Hofer

4.1 Catalysis and Scanning Tunneling Microscopy

Scanning tunneling microscopy (STM) has developed into one of the most ubiquitous tools in surface science and the range of phenomena studied by this technique is nowadays remarkably broad. These include surface topography, electronic and vibrational properties, film growth, contact charging, molecular manipulation, and many other phenomena ranging from the micrometer scale down to the subnanometer scale. Concerning surface chemical reactivity, the use of STM and scanning tunneling spectroscopy (STS) has proved fundamental for probing local surface modifications associated with chemical reactivity at the solid–vacuum interface and to follow, when possible, the mobility of reaction intermediates and final products on surfaces in real time. Owing to the complexity of the studied systems in terms of both surface morphology and variety of chemical reactivity, collaboration between experiment and theory has emerged as a necessary tool in order to arrive at sustainable models and a successful rationalization of experimental findings [1–4]. The aim of theory, when dealing with chemical reactivity at surfaces as imaged by STM is twofold: on the one hand, simulations can provide insight into the energetic landscape governing the chemical reaction from the reactants to the final products; on the other hand, STM simulations are of utmost importance for an understanding of the origins of image contrast and interpreting particular experimental images. Modeling surface reactivity as imaged by STM – the first major challenge – is the actual capacity of ab initio approaches to correctly describe the surface electronic structure and the ensuing chemical reactants in the presence of other chemical species dosed on the reaction substrate. Here, we consider the main practical limitations that affect ab initio modeling of surface chemical reactivity such as the need to work with simulated slabs of finite size and thickness and the approximated description to exchange and correlation (XC) terms. In this chapter, we review for a few selected applications how the modeled reactivity and the simulated STM appearance of reactants and products are affected by the chosen trade-off between the enforced level of approximations and the ensuing accuracy of the calculations.

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

98

4.2 Image Formation in an STM

The control parameter in an STM, the current in the tunneling junction, is always due to the same physical process. An electron in one lead of the junction has a nonvanishing probability to pass the potential barrier between the two sides and to tunnel into the other lead. However, this process is highly influenced by (i) the distance between the leads, (ii) the chemical composition of the surface and tip, (iii) the electronic structure of both the systems, (iv) the chemical interactions between the surface and the tip atoms, (v) the electrostatic interactions of the sample and tip. The main problem, from a theoretical point of view, is that the order of importance of all these effects depends generally on the distance and therefore on the tunneling conditions [5–8]. Apart from the tunneling current, the other parameter required to define the tunneling conditions (for a given tip) is the bias voltage, that is, the electrostatic potential difference enforced between the probed surface and the probing tip. In general, the change of bias voltage has two effects: (i) it shifts the Fermi level of one lead (sample or tip) relative to the other lead and (ii) it compensates for this effect by creating a surface dipole. Both of these effects are understood in principle, even though no detailed simulations actually exist. The reason is that a simulation requires treating the coupled system of two leads and the interaction range in a nonequilibrium situation. Even though the theoretical framework for such a treatment in the STM context, the nonequilibrium Green function formalism of Keldysh, has been known for some 40 years [9], its numerical implementation became possible only recently [10]. Within an STM, the method is generally not applicable due to the different lateral symmetry of surface and tip systems. However, provided the electronic density of states (DOS) of the surface and tip structure are fairly smooth, the current increases approximately linearly with the applied bias voltage. This is due to an increase in the energy difference between the Fermi level of the sample and the tip, which increases the number of electron states provided by one lead, the number of empty states provided by the other, and to the condition of resonant tunneling. The differences in Fermi level due to bias voltage, combined with the general differences between the Fermi level of a sample and the tip, introduces electrostatic effects into the interfaces. A thorough theoretical investigation of this issue for differently conductive substrates, carried out at different level of theory [11–14] has pointed out that although electrostatic interactions are present between the tip and the substrate, for actual measurements under normal experimental conditions the effect is not strong enough to affect the STM image in a substantial way [5]. Unlike electrostatic forces, chemical forces between the probing tip and the probed surface have been shown to profoundly affect the tunneling current from a certain onset. Owing to the advent of first-principle methods and powerful computers, it could finally be resolved by a calculation of the combined tip–sample system [15]. The point of onset for chemical bonding on metals was found to be at a distance of 4–5 Å. As the tip approaches the surface, chemical forces rapidly become large enough to

4.3 Simulating Tunneling Currents

destroy the stability of a tip–sample system in an STM due to ions jumping from the tip to the surface and vice versa. The point of destruction is at about 4 Å. From the onset of chemical forces to the limit of mechanical stability, the two main effects are (i) an increase in the tunneling current beyond exponential growth and (ii) an increase in surface corrugation by a factor of 2–3. Both effects are observed in experiments [16, 17]. They show, unambiguously, that high-resolution STM operates within the range of chemical forces, and that the main effect, omitted in any straightforward perturbation model, is the relaxation of surface atoms due to chemical interactions. Modeling STM images involves several stages: (i) establishing a realistic model of the surface from experiment and theory; (ii) establishing a realistic model of an STM tip using properties known experimentally and inferred from theory; (iii) explicitly modeling the interactions between the tip and the surface; (iv) calculating the tunneling current for the applied bias; (v) comparing with the experimental data. A fundamental requirement of any STM simulation is the demand that the electronic states of the surface, and ideally of the tip, is precisely simulated because these states determine the obtainable contrast and ultimately the simulated STM image. Here, the level of accuracy of the simulation critically depends on the accuracy of the simulation in modeling both the energetic localization of the electronic states with respect to the surface (tip) Fermi level (i.e., chemical potential at 0 K) and the real space distribution outside the slab, that is, their decay in the vacuum region. The first stage of the modeling process is, thus, establishing a good representation of the surface being studied. We note that on the basis of recent results [3, 18–20], an accurate description of the energetics associated with a given reactivity scenario and/or the electronic structure of the reaction intermediates may not be sufficient to accurately account for the experimentally detected STM appearance as will be shortly reviewed. Concerning the probe tip, its correct description is indeed another key factor determining the final level of agreement with the experiment. Owing to the complexity of the subject, we refer the reader to detailed reviews existing on the subject [5, 6]. Here, we only note that since no experimental atomic information is ultimately known about the probing tip to date, the most accurate solution of the problem relies on empirical comparison of modeled images simulated with different tips and available experimental data. In this perspective, the recently reported experimental approaches toward simultaneous recording of STM and atomic force microscopy (AFM) images represent an extremely valuable method of comparison for explicit modeling of the tip influence on the image and in turn the identification of a library of “most likely” tip geometries for certain experimental conditions [4].

4.3 Simulating Tunneling Currents

A complete survey of all methods to calculate the tunneling current, developed over the past few decades, is given in Ref. [21]. In increasing order of theoretical difficulty,

j99

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

100

the four main approaches are (i) the Tersoff–Hamann approach [22, 23], where constant current contours are modeled from the electronic structure of the surface alone; (ii) the transfer Hamiltonian or the Bardeen approach [24], where the tip electronic structure is explicitly included in the calculation; (iii) the scattering or Landauer–B€ utticker approach [25], which includes multiple pathways of tunneling electrons from their initial to their final crystal states; and (iv) the Keldysh or nonequilibrium Greens function approach [26], which also considers inelastic effects such as electron–electron or electron–phonon scattering. For technical details of the various levels of approximation involved, the reader is referred to the literature [28–33]. In the following section, we shall base STM simulations mainly on the multiple scattering approach developed to first order [27].

4.4 Simulating Chemical Reactivity

Disregarding the implementation details with respect to the wave function approximation [34], two critical aspects when modeling the chemical reactivity at surfaces are (i) the evaluation of the XC-terms and (ii) the use of simulation cells of finite size. With respect to (i) we limit the review to a limited number of studies addressing the role of the XC description with respect to the modeled electronic structure and the ensuing simulated STM appearance for a given system (see below and Refs [3, 20]). The main reason for the small numbers of such studies is the remarkably high computational cost associated with the explicit inclusion of Hartree–Fock (HF) terms in the description of XC potentials [35, 36]. This increased computational cost inevitably demands larger computational facilities that at the moment are not generally available for academic research. So far, the most notable improvements by enhanced XC descriptions (as those provided by hybrid DFT approaches such as B3LYP [35] or HSE03 [36]) have been found for ionic semiconductors [37–40]. Unfortunately, the larger computational requirements associated with hybrid-DFT approaches is hardly compatible with another aspect characterizing ionic semiconductors, that is, the role of long-range electrostatic effects. In fact, for these substrates, the use of large simulation cells and thick slabs is a fundamental requirement to obtain converged results about basic reactivity indicators, for instance, the adsorption energy [3, 41, 42]. Just to quote an example, it has been reported that the adsorption energy of molecularly adsorbed O2 on hydroxylated TiO2(1 1 0) [41] or H2O on clean TiO2(1 1 0) [1, 3, 42] dramatically depends on the simulated coverage (see Figure 4.1). Thus, the advantages of an improved XC description might be balanced by the simulation biases associated with the reduction of the size for the simulation cell or, equivalently, with an overestimated coverage compared to the actual situation imaged in the experiments. This aspect is even more delicate for simulations aiming at assessing the relative stability of different intermediates characterized by different long-range interactions (for instance, O2-related products on hydroxylated TiO2(1 1 0) [43]). In fact, within the same modeled extent of the simulation cell, in principle, nothing prevents this finite size bias changing as the polarity of the

4.5 Catalytic Water Production

2.2 2.0 Adsorption Energies (eV)

1.8

O Vacancy

1.6 OH

1.4 1.2 1.0 0.8 0.6 0.4 perfect

0.2 0.0 1/8

1/4

1/2 OH Coverage (ML)

3/4

1

Figure 4.1 Calculated O2 adsorption energies with respect to OH coverages. For comparison, the O2 adsorption energies on the perfect surface and in the vicinity of the O vacancy are also shown. (Reprinted with permission from Ref. [41].)

intermediates change, negatively affecting the reliability of final computed stabilities. In our view, this aspect may represent one of the main (indirect) limitations of the use of hybrid-DFT approaches in modeling the chemical reactivity of ionic semiconductors. Extra care should, therefore, be taken when choosing the simulated system size.

4.5 Catalytic Water Production

Catalytic oxidation of hydrogen on transition metal surfaces plays a crucial role in various subjects such as electrochemistry or heterogeneous catalysis, and it has important applications, for example, for making hydrogen fuel cells. The first report of this phenomenon dates back to the early nineteenth century, when hydrogen was observed to react with oxygen at room temperature to produce water on a Pt surface. The remarkable ease of the reaction led to the definition of the term “catalysis” [44]. In the next 100 years, these reactions were intensively investigated on various metals, such as silver, gold, rhodium, and palladium [45–49]. In these studies, the experimental techniques limited the observations to the macroscopic or mesoscopic level. Although it was found that the catalytic production of water from hydrogen and oxygen is always accompanied by the simultaneous oxidation of the metal surface, the proposed reaction mechanisms could not be firmly proven [50, 51]. Since 1982, the STM has been offering the possibility of direct real-space determination of surface structures at the microscopic scale [52, 53]. Over the past two decades, the

j101

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

102

reaction transition states, that is, the formation of hydrogen bonding networks have been repeatedly imaged in a number of STM measurements [54–56]. However, since STM images are determined by the electronic states rather than the atomic positions at surfaces, the chemical identification of the intermediate states remained somewhat unclear. Thus, only DFT calculations and STM simulations in comparison with experimental data can attain a complete and atomic-scale understanding of the reaction mechanism of catalytic water production on transition metal surfaces. Recently, STM measurements of a water production reaction were undertaken on a well-designed (2  1)p2mg-O/Rh(1 1 0) surface under ultra high vacuum (UHV) conditions [18]. In the initial structure, a Rh(1 1 0) surface is covered by an oxygen overlayer, in which O atoms adsorb in a zigzag fashion at the threefold coordinated sites. When the surface is exposed to molecular hydrogen, the reaction propagates on the whole surface in a wave front, after a short induction time. The whole process can be characterized by three distinct steps: in the first step, half of the oxygen atoms are vaporized, the Rh surface is gradually covered by paired features, and the whole surface shows a c(2  4) symmetry (see Figure 4.2a); in the second step, a close packed c(2  2) island shows up and grows at the expense of the c(2  4) structure; it expands to about half the surface. In the third step, this c(2  2) island gradually shrinks and finally disappears, leaving behind a clean Rh(1 1 0) surface. The interpretation of the two intermediate structures, that is, the c(2  4) and the c(2  2) structures, is obviously of prime importance for an understanding of the mechanism of this catalytic oxidation reaction. A detailed investigation of the c(2  4)

Figure 4.2 Intermediate stage of the reaction with c(2  4)-O structure on Rh(1 1 0): (a) zoom on the surface behind the reaction front. Dimensions: 11.5 nm  7.5 nm. I ¼ 1 nA, VB ¼ þ 0.13 V. (b) STM image after dosing 0.6 l oxygen and annealing at 270 K. Dimensions: 20 nm  20 nm. I ¼ 1 nA, VB ¼ þ 0.42 V. (c) Structural model of the c(2  4)-O. (d) Simulated STM image (VB ¼ þ 0.13 V, I ¼ 0.05 nA).

4.5 Catalytic Water Production

structure shows that the height of the paired features is 0.3–0.4 Å, with an internal distance of about 2.77 Å. All these values are in agreement with the experimental observation in Ref. [57], where, on a Rh(1 1 0) surface at low coverage, pairs of oxygen atoms align along the [1 1 0] direction, the internal distance varying from 3.3 Å (at the lowest coverage) to 2.5 Å (at the highest coverage). The c(2  4) paired structures are, thus, very likely due to the presence of adsorbed oxygen atoms. On the basis of this assumption, first-principle calculations were performed to determine the groundstate adsorption configuration. The structure optimization is carried out with the Vienna ab initio simulation package (VASP) [58–61], a density functional theorybased code for systems with periodic boundary conditions. The ion–electron interactions are described by the projector augmented wave (PAW) [62] method and the exchange correlation potentials are calculated by the generalized gradient approximation (GGA) of Perdew and Wang (PW91) [63]. The metal surface is modeled by a supercell containing a five-layer slab, in which the top two layers of Rh atoms are allowed to relax in three dimensions; the vacuum range between the slabs is about 19 Å. In total, six configurations with c(2  4) symmetry are analyzed for the structure search. The lowest energy adsorption sites are found to be short-bridge sites, in which the adsorption energy per O atom is at least 0.1 eV larger than for the models. This structure model is shown in Figure 4.2c. However, we note that the OO distance in a pair is about 3.8 Å rather than the 2.7 Å observed in the STM experiments. In general, this can be the result of a contaminated tip, strong tip–adsorbate interaction and corresponding atomic relaxations, or electronic effects of the reacted surface. If the tip is contaminated, its apex is most likely attached to a hydrogen molecule or H atoms. As a consequence, the conductance of this tip should be much lower than that of a clean tungsten tip. Since this conductance change has not been reported, it can be concluded that the reduced OO distance is not the effect of a contaminated tip. Surface–tip interactions are evaluated by calculating the interaction between the reacted surface and a tungsten cluster at low distance. Here, the calculations indicate that there is no substantial relaxation due to interactions between the two leads. Consequently, the only possibility left is that the electronic surface structure somehow changes the appearance of the oxygen positions. Subsequently, we calculated constant current contours using the multiple scattering approach and a clean tungsten tip. The results of these simulations are shown in Figure 4.2. It turns out that although the actual distance between oxygen atoms is 3.8 Å, the imaged distance is substantially smaller at 2.7 Å. The reason for this discrepancy is the large size (more than 4 Å) of the tip apex: in this case the contributions from the two electronic states at the two oxygen atoms are not completely separate, as the tip scans across the surface, and the protrusions, therefore, appear to be closer together than in reality. Clearly, this is a generic effect of the surface electronic structure that should also be present in other experiments and thus quite general. The agreement of experimental results and theoretical simulations suggests that the c(2  4) structure is formed by the oxygen atoms at the short bridge site of the Rh(1 1 0) surface. After dosing hydrogen, a different structure with c(2  2) symmetry is formed. The more condensed c(2  2) island is characterized by a hexagonal arrangement of

j103

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

104

Figure 4.3 c(2  2) structure: (a) zoom on the reaction intermediate structure. Dimensions: 6.0 nm  2.5 nm. Parameters: I ¼ 0.6 nA, VB ¼ þ 0.6 V. (b) Line profile on the hexagonal structure along [0 0 1]. (c) STM image of the surface obtained by dosing water up to

saturation at 170 K on 0.35 l O2 adsorbed at 200 K and annealing the surface at 223 K. Dimensions: 15 nm  15 nm. Parameters: I ¼ 1 nA, VB ¼ þ 0.12 V. (Reprinted with permission from Ref. [18].)

“double peak” features (see Figure 4.3, in particular the line scan frame (b)). The two peaks in each feature are reported of the same height and are separated by an internal distance of about 2.8 Å [18]. Extensive DFTcalculations, varying not only the positions of atoms but also the chemical stoichiometry of the units, have been performed for the structure search. The possibility of compressed oxygen is clearly excluded by the decrease in the adsorption energy when the density is increased. Assuming an identical number of H and O in one unit cell, the only possible candidate of a “double peak” feature is the OH group. However, the simulated STM images of the OH-Rh (1 1 0) surface are in strong disagreement with the experimental results. Changing the stoichiometry of hydrogen and oxygen to 3 : 2, two stabilized c(2  2) structures are obtained. In the first structure, shown in Figure 4.4a, all oxygen atoms adsorbed at the threefold coordinated sites of the Rh surface, showing a zigzag pattern along the [1 1 0] direction. In the [0 0 1] direction, the oxygen atoms are in a paired arrangement, with an OO distance of 2.77 Å. The hydrogen atoms are between two adjacent O atoms, resulting in a network of planar hexagonal rings above and parallel to the Rh surface. We call this structure c(2  2)A. The STM simulations on c(2  2)A are limited by the low current values obtained [64, 65]. In order to make the simulations comparable between the test configurations, we generated simulated current maps at identical tip–surface distances. As shown in Figure 4.4b and c, a clear double peak feature emerges, as the tip approaches the surface. The internal OO distance in these simulations is in very good agreement with experimental results. The second stable configuration is shown in Figure 4.5a. Here, the hexagonal network is formed by OH groups and water molecules. The hydrogen atoms belong either to OH or to H2O but are not equally shared by two adjacent oxygen atoms as in the structure c(2  2)A. In each of the OH groups, the oxygen atom is close to the short bridge site of the surface, whereas the hydrogen atom points upward forming a hydrogen bond with the oxygen atom of the water molecule. In the layer of water molecules, the oxygen atoms are roughly above the threefold sites of the

4.5 Catalytic Water Production

Figure 4.4 c(2  2)A structure. Left panel: structural model. Right top panel: corresponding simulated STM image (VB ¼ þ 1.30 V, I ¼ 0.04 nA). The protrusions correspond to oxygen couples, whereas the depressions are the hollow sites surrounded by OH complexes. Right bottom panel: simulated current profiles along [0 0 1] at decreasing (light blue to red) tip–surface distances. (Reprinted with permission from Ref. [18].)

Figure 4.5 c(2  2)B structure. Left panel: structural model. Right top panel: corresponding simulated STM image (VB ¼ þ 1.30 V, I ¼ 0.002 nA). The protrusion maxima correspond to water molecules, whereas the depressions correspond to Rh atoms. Right bottom panel: simulated current profiles along [0 0 1] at decreasing (blue to red) tip–surface distances. (Reprinted with permission from Ref. [18].)

Rh surface, whereas the hydrogen atoms point downward to the on-top sites. We call this configuration c(2  2)B. The constant height current maps are shown in Figure 4.5b and c. It can been seen that it is not possible to image the H2O þ OH unit as a double peak.

j105

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

106

Comparing the two c(2  2) structure models, we find that the adsorption energy per unit cell of c(2  2)B is about 1.5 eV higher than that of c(2  2)A. This means that under realistic experimental conditions one should always see c(2  2)B, and never c(2  2)A. However, it is well known that this reaction process inevitably includes hydrogen atoms or molecules on the Rh surface itself. In general, these hydrogen atoms are very mobile on transition metal surfaces under ambient conditions. Although these hydrogen atoms are usually invisible in STM measurements [66, 67], they may play an important role in the stabilization of surface structures. Even though this assumption certainly requires additional experimental and theoretical analysis of the problem, it could solve the riddle why the structure seen in the STM experiments is not, according to DFT simulations, the structure of lowest total energy. On the basis of the experimental details in Ref. [18] and our theoretical investigations, we propose a possible reaction mechanism. In the first step, the adsorption of hydrogen occurs on the (2  1)p2mg-O/Rh(1 1 0) surface. Half of the oxygen is removed due to the adsorption of hydrogen molecules. Because of its small size, these hydrogen molecules are invisible in the STM measurements. In the second step, the remaining oxygen atoms are “pushed” away from the threefold coordinated sites and moved close to the short-bridge sites, resulting in the less condensed c(2  4) structure, shown in Figure 4.2a. Since the oxygen atoms are chemically bonded to the Rh surface, the driving force of the formation of the closepacked c(2  2) structure should also be the “invisible” H atoms. Very likely, the c (2  2)A structure is initiated at some surface defect. The surrounding O atoms are then “pushed” close to the c(2  2) island and stabilized by hydrogen bonds. The c (2  2) island expands until all O atoms on the Rh surface are reacted. In the third step, the H atoms attack the molecular units in the c(2  2) island, leading to water formation. The produced water molecules then evaporate, leaving only a clean Rh (1 1 0) surface. 4.5.1 TiO2: A Catalytic Model System

In the past few years, a large number of experimental and theoretical studies have focused on metal oxide surfaces with the aim of gaining insight into their catalytic, photocatalytic, and gas-sensing activity [68]. Owing to its thermodynamic stability and relatively easy preparation, the rutile TiO2(1 1 0) surface has evolved into one of the key models for metal oxide surfaces. For example, it has been extensively used in the research of biocompatible materials, gas sensors, and photocatalysts [69]. The rutile TiO2(1 1 0) surface is characterized by alternate rows of fivefold coordinated Ti (Ti5c) and twofold coordinated bridging O atoms (Obr or, equivalently, O2c) along the [0 0 1] direction (Figure 4.6). Its detailed surface relaxation, determined from quantitative low-energy electron diffraction (LEED-IV) studies [70], has proved a very severe test for DFT approaches aiming at reproducing the experimental data. Only recently has it been possible to clarify the apparent discrepancies between LEED-IV and first-principles data. In a very detailed DFT study, using films of up to

4.5 Catalytic Water Production

Figure 4.6 Left: STM image of a stoichiometric 1  1 TiO2(1 1 0) surface, 14 Å  14 Å. Sample bias þ 1.6 V, tunneling current 0.38 nA. The inset shows a ball-and-stick model of the unrelaxed 1  1 TiO2(1 1 0) surface. Rows of bridging oxygen atoms are labeled “A” and rows of fivefold coordinated titaniums “B”. Right: contour plots of [0 1 1]-averaged charge densities associated with electron states within

2 eV of the conduction band minimum for the relaxed 1  1 stoichiometric surface. Contour levels correspond to a geometric progression of charge density, with a factor of 0.56 separating neighboring contours. The “noise” in the topmost contours is an artifact of the periodically repeated slab geometry. (Reprinted with permission from Ref. [72].)

13 OTiO2O trilayers and reducing computational approximation errors, Thompson and Lewis [71] have been able to achieve excellent quantitative agreement with the experimental data. In the same paper, it has also been pointed out that a minimum of five trilayers slab is required to recover the experimental TiO bond lengths [71]. The STM contrast for TiO2(1 1 0) is governed by electronic rather than geometric properties (see Figure 4.6) [72]. Thus, contrary to expectations from geometrical considerations, the deeper lying Ti5c rows are imaged as the bright rows by STM, while the bridging oxygen atoms (Obr) are imaged as the dark rows for physically meaningful tip–surface separations (see Figure 4.6). However, despite the qualitative agreement between experiment and constant density contours, the explicit inclusion of W tips as in the experiment does not yield a quantitative match with the experiments for the simulated tunneling conditions (see Figure 4.7) [4]. Figure 4.7b shows a close-up of Figure 4.7a, a 300 pA constant It contour, which has a corrugation of approximately 100 pm and is located approximately 300 pm from the O2c surface atoms. These values disagree quantitatively with experimental STM results at the same tunneling conditions on two accounts. First, a set point of It ¼ 300 pA is not a particularly large value for constant current STM imaging, and

j107

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

108

Figure 4.7 (a) Calculated It map for the W[1 1 0] tip (tip 1) taken across the Ti(5c) and O(2c) rows. (c) Same as (a) only scaled by a factor of 200. Below (c), a ball model indicates the position of the different types of atoms in topmost surface layer. Both (a) and (c) show

five logarithmically equidistant space constant It contours. (b) and (d) Close-up of (a) and (c), respectively, showing the position of the 300 pA constant It contour. (Reprinted with permission from Ref. [4].)

thus one would assume that the tip should be probing the surface at considerably larger distances than the approximately 300 pm indicated in Figure 4.7b. Second, the resulting constant It corrugation for the 300 pA contour is much larger than experimental results for the same tunneling conditions (see Figure 4.8 [3]), where a 30 pm corrugation over the Ti(5c) and O(2c) rows was reported for an It set point of 300 pA. If the initially calculated It values for the atom-ended W(1 1 0) tip are scaled up by a factor of 200, the agreement is, however, almost perfect as shown in Figure 4.7c and d. The close-up on the resulting constant contour of 300 pA (Figure 4.7d) shows that not only has it moved further from the surface and is now located at a more meaningful 620 pm distance from the surface but also the resulting corrugation of 30 pm matches the reported corrugation in Ref. [3] almost perfectly. The origin of the discrepancy can be traced back to several sources. First, the bSKAN code [6, 33] used for the tunneling current calculation does not allow the atoms in the tip and surface to relax prior to the evaluation of the It when the surface

4.5 Catalytic Water Production

Figure 4.8 Left: experimental STM images (1.5 V; top: 0.3 nA, bottom: 0.59 nA) of TiO2(1 1 0). Features corresponding to Ovac, OHbr, 2OHbr, and an adsorbed water molecule are indicated by blue, green, black, and pink lines, respectively (clean surface: yellow line). The top image (120 Å2) was recorded at room temperature, whereas the bottom image (50 Å  20 Å) was recorded at 150 K. Right: simulated STM images (1.5 V), scan lines and corresponding ball-and-stick models for optimized TiO2(1 1 0) in the presence of Ovac, molecularly adsorbed water molecule, its pseudodissociated state and final dissociation

product. O, Ti, and H atoms are drawn red, yellow, and white, respectively. A: Ovac; B: OHbr; C: molecularly adsorbed water molecule on 5f-Ti; D: pseudodissociated state; E: 2OHbr; F: clean surface. The picture has been prepared merging all the individual simulated images together (evaluated at the same density contour value, i.e., 0.359  107 e Å3) and allowing an exponential decay of half a simulation cell toward the clean surface value whenever the simulation cell has been found too small to allow a full recovery of the clean surface baseline. (Reprinted with permission from Ref. [3].)

model consists of more than one type of atoms [15]. This effect may significantly increase the measured It. Second, the STM experiments are performed on a reduced TiO2 crystal, and states introduced in the bandgap, for example, by Ti bulk interstitials [73], have been shown to increase the overall conductivity of the TiO2 crystal [74–76]. The theoretical calculations are performed on a stoichiometric surface model, and it is, therefore, expected that the simulations should result in underestimated It values. We note also that, experimentally, the STM contrast is found to change as much as 0.5 Å depending on the applied (positive) bias and specific tips. On these grounds, it is reasonable to expect that different tips would result in scaling factors different from the ones reported in [4]. In this respect, simultaneous AFM and STM of the surface acquire additional importance for the identification of experimental tip structures, and consequently, the scaling factor required to gain quantitative agreement between

j109

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

110

theory and experiment [4]. It should be stressed that this feature of STM simulations on TiO2 is quite unique and originates not in a problem of transport theory or the description of tunneling but in problems associated with the description of the electronic structure of doped semiconductors. Owing to the surface preparation protocol, which relies on cycles of Ar-ion sputtering and vacuum annealing to temperatures as high as 1100 K [73], ultra high vacuum prepared TiO2(1 1 0) is inevitably characterized by the presence of oxygen vacancies Ovac. Even in very clean UHV conditions, bridging hydroxyls OHbr are also present on the surface because of the dissociation of residual water in Ovac sites [1, 2, 77]. Despite some variations in the thickness and lateral extent of the simulated slabs, a general consensus has been established concerning the exothermic dissociation of water molecule in Ovac sites [1–3, 42]. Interestingly, it turns out that most of the surface processes at TiO2(1 1 0) are indeed induced and governed by the presence of these pristine defects [69]. The STM appearance of these pristine defects, namely, Ovac and OHbr, is closely related to the contrast assignment between Ti5c and Obr rows of TiO2(1 1 0). This issue has been a matter of long debate within the experimental and theoretical community. Initially, bright protrusions on dark rows were assigned to Ovac [78]. Unfortunately, not only Ovac but OHbr are also imaged as bright protrusions on dark rows [79]. The main difference in their appearance is their relative extent and apparent height with respect to Ti5c and Obr rows. Comparison between STM measurements and simulated images led to the brighter and larger spots being initially assigned to Ovac, whereas the smaller and darker spots were associated with the presence of OHbr on the surface [19]. However, recent STM and DFT data [1–3, 77] have provided compelling evidence that the protrusions associated with Ovac are less bright and extended than those associated with OHb, an aspect which simulations on a sufficiently thick slab are capable to recover [3] (see Figure 4.8). Despite the qualitative agreement between theory and experiment concerning the simulated STM appearance of Ovac and OHbr, a major disagreement of theory with respect to the experimentally detected appearance of H2O-related products on TiO2(1 1 0) is the overestimation of the simulated apparent height for H2O and the dissociation precursor (OHbr–OH) [2] with respect to the simulated appearance of Obr (see Figure 4.8). Another major flaw of any DFT study of TiO2(1 1 0) relying on a semilocal (GGA) approximation to the exchange correlation terms (as in Refs [3, 4]) is represented by the well-known underestimation with respect to the experimental bandgap of wide gap semiconductors [34] and an incorrect description of the defect states associated with pristine surface defects [3]. In fact, different spectroscopic studies suggest that both Ovac and OHbr introduce defect states in the bandgap that are found 1 eV below the conduction band (CB) onset [80, 81]. It has been recently shown that to correctly represent the energy localization of these defect states in the bandgap, some exact HF exchange contribution is required [40] (see Figure 4.9). In fact, hybrid DFT B3LYP results clearly localize the electronic states associated with Ovac and OHbr in the (slightly overestimated) bandgap in line with the experiment (Figure 4.9). Conversely, at odds with the experiment and hybrid DFT descriptions, semilocal (GGA)

4.5 Catalytic Water Production

Figure 4.9 Total and projected density of states for the hydroxylated (top) and reduced (bottom) TiO2(1 1 0) surface, calculated using the B3LYP hybrid functional. The Ti3 þ states are localized on (a) the Ti ion between the two bridging OH groups, Ti3brþ- OH ; (d) the Ti ion nearest to the oxygen vacancy, Ti3brþ- v ; (b), (c) on a five-

coordinated Ti ion, Ti35cþ of the surface. Spin density plots are reported: (a), (d) Ti3brþ , side view; (b), (c) Ti35cþ , top view. The vertical dotted line in the PDOS denotes the position of the Fermi energy. Only the majority spin component is reported. (Reprinted with permission from Ref. [39].)

approaches energetically localize the defect states at the bottom of the conduction band (see Ref. [3] and references therein). Concerning the real-space distribution, B3LYP provides highly localized solutions for Ovac and 2OHbr (Figure 4.9), fully supporting the proposed electron trap affinity for these defect states [39]. The effects of the incorrect GGA description for the defect states on the simulated STM appearance has been recently addressed by comparing the simulated appearance for clean TiO2(1 1 0), Ovac, and OHbr both at GGA and at B3LYP levels [20]. Interestingly, once limitations of the localized atomic basis set with respect to the

j111

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

112

Figure 4.10 Simulated STM images (sample bias 1 V) at constant density (2.5  106 e/B3) of the (a) hydroxylated and (b) reduced (1 1 0) rutile TiO2 surface (one OH group and one oxygen vacancy, respectively) obtained with B3LYP localized basis set calculation. (Reprinted with permission from Ref. [20].)

vacuum decay of the surface states are suitably accounted for [20], it turns out that the simulated STM appearance of the surface (for positive biases) does not depend critically on the relative position of the defect states with respect to the (imaged) bottom of the conduction band. Consequently, Hybrid DFT results are also fully consistent with the experimentally detected larger STM appearance for OHbr with respect to Ovac (see Figures 4.8 and 4.10). For completeness, we report that on the basis of resonant photoelectron diffraction (PED) measurements, it has been recently suggested that the real-space charge distribution for the defect states associated with Ovac is delocalized over several subsurface sites around Ovac [82]. While challenging the B3LYP real-space distribution of the defect states associated with Ovac (Figure 4.9), the qualitative agreement between experiment and B3LYP-based simulated STM appearance for positive biases has to be considered unaffected. This in turn further suggests that the bottom of the conduction band is ultimately only marginally affected by the description of the bandgap electronic states associated with pristine defects, that is, Ovac and Obr. The same conclusions would also be valid in the presence of Ti-interstitials immediately below the surface topmost layer, a possibility that has been recently ruled out on the basis of [82] and, at the same time, also suggested to explain hydroxylated TiO2(1 1 0) reactivity toward molecular oxygen [83]. Thus, the puzzling scenario that emerges from the theory side is that although GGA-based approaches fail in reproducing the experimental position in the bandgap for the electronic states associated with the pristine defects, the onset of the CB edge is unequivocally well described in terms of real-space localization of the electronic distribution far from the surface. At the same time, despite excellent agreement with the experiment about the energy position in the bandgap for the defect states, some (system-specific [20]) extra tuning is required to account for the experimentally detected STM appearance on a localized atomic basis set B3LYP.

4.5 Catalytic Water Production

Figure 4.11 Bottom: optimized ions HSE03 total density of states and integrated number of defect states (Dn) for Ovac. The integrated charge density corresponding to the defect states is shown in the top panel from two different perspectives for the same isocontour value (green: 106 e Å3). O: red, Ti: cyan (unpublished work).

A substantial improvement in the theoretical modeling of this substrate, able to overcome the drawbacks of both former approaches (standard GGA and localized atomic basis set B3LYP) is the implementation of the HSE03 [36] hybrid functionals in a plane-wave formalism. Interestingly, within such an approach, the simulated bandgap of TiO2(1 1 0) is modeled to be in good agreement with experimental value of 3.2 eV [69], and in line with B3LYP results, both Ovac and OHbr are modeled to induce defect states in the bandgap roughly 1 eV below the onset of the CB (Figure 4.11). As for B3LYP [39], the most stable modeled solution for Ovac is a paramagnetic triplet state (Figure 4.11). For completeness, we note that the modeled HSE03 subsurface delocalization (as for B3LYP results) is clearly at odds with recent PED data [82] concerning the rather delocalized nature of the defect-induced electronic states associated with Ovac. Concerning the HSE03 simulated STM appearance, it is found to be in agreement with the experiment, that is, the simulated contrast is modeled to increase going from Ovac to H2O and OHbr (Figure 4.12). Although still overestimated with respect to the experiment, with maximum deviations smaller than 50 pm with respect to the corresponding experimental value, the overestimation effect on the simulated contrast is found to be reduced for HSE03-based results by comparison to the analogous GGA-based data (Figure 4.8). Leveraging on this positive outcome, we next explicitly consider the effect of the experimentally used W tips in the experiment. As in Refs [3, 4], the constant current topographies were simulated by integrating the states in the CB edge from its

j113

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

114

Figure 4.12 Left: experimental line profile for TiO2(1 1 0) along ½1 1 0 in the presence of pristine defects (Ovac, OHbr) and undissociated water molecule for V ¼ 1.5 V, I ¼ 0.08 nA (80 K). Right: calculated density contour (top, 1.5 V,

4  108 e Å3) and current contour (bottom: W tip, V ¼ 1.5 V, I ¼ 0.08 nA) HSE03 scan lines. The modeled baseline for the clean TiO2(1 1 0) surface is at approximately 5 Å for both density and current contours (unpublished work).

onset onward. Owing to a substantially improved description in terms of real-space distribution of the electronic states associated with the CB onset, the agreement with the experiment concerning the simulated corrugation and the tip–surface separation results is notably improved. In fact, the simulations for the clean TiO2(1 1 0) surface at the considered experimental tunneling conditions of I ¼ 0.08 nA, VB ¼ 1.5 V, suggest a tip–surface distance of roughly 480 pm that is physically significant even in the absence of any scaling parameter. Of course, the consideration previously reported on the causes of the underestimated tunneling currents [4] also applies in these cases. On these grounds, the agreement can be considered as semiquantitative. Interestingly, because of the improved description of the surface electronic structure, HSE03 simulations with explicit inclusion of the tips unequivocally assign to H2O a smaller STM protrusion with respect to the spot associated with OHbr, in close agreement with experimental data for the same tunneling conditions (Figure 4.12). To conclude, TiO2(1 1 0) indeed represents a key model for understanding the surface chemical reactivity of metal oxides. So far, it has posed major challenges to theory in terms of both chemical reactivity and STM appearance. Despite an intrinsic incapacity to correctly account for the energy localization of defect-induced states in the bandgap, GGA-based approaches to the surface reactivity have so far been able to account for most of the experimental findings related to chemical reactivity of pristine defect sites of TiO2(1 1 0). At the present stage, the main challenge is

4.6 Outlook

represented by the surface chemical reactivity in the presence of molecular oxygen, O2(g) [43], although some recent results [83] suggest that interstitial Ti atoms in the immediate subsurface region (a possibility that was previously not considered by any ab initio investigation [42, 44]) should play a major role with respect to O2(g) reactivity on TiO2(1 1 0). Concerning the simulated appearance of TiO2(1 1 0) in the presence of pristine defects, also the tendency of GGA to excessively delocalize electronic charge, turns out not to be a major problem when properties related to the onset of the CB edge are investigated such as in the vast majority of positive bias STM applications. The use of hybrid-DFT functionals such as B3LYP and HSE03 is found to correctly account for the energy localization of the electronic states associated with pristine defects. In addition, the simulated STM appearance is notably improved when relying on a HSE03 description of the surface electronic structure. Despite these encouraging findings, the increased computational cost associated with hybrid-DFT descriptions of the surface poses major practical problems to viable modeling of multistep reactions on TiO2(1 1 0), an aspect that urges the theoretical community to explore and/or suitably develop the use of nimbler alternative approaches capable of correctly accounting for the surface properties. In this perspective, we note that to the best of our knowledge, at the present stage the actual capacities of linear scaling N-order DFT methods [84, 85] to model the surface reactivity of TiO2(1 1 0) are still unexplored.

4.6 Outlook

In this chapter, we have reviewed some of the key processes and systems in catalysis research, which have been analyzed by STM. If one message is relevant for all considered systems, it is that electronic structure is usually dominant in the images: what one measures, with an STM, has therefore frequently no immediate bearing on the position of atoms. Given that a key feature of catalytic processes is the balance between electronic structure changes from adsorption and electronic structure changes correlating with reactions, it is thus quite often impossible to attain a thorough understanding of the images and reactions without high-level theory. Here, we pointed to the problem of theoretical representation, in particular, in two aspects of theory: (i) the existence of highly mobile atoms at the surface such as hydrogen, which are usually not considered in the atomistic models and (ii) the importance of bandgaps and relative energy levels of electronic states, which is often distorted in local density approximations. In both respects, a quick fix to the problem is not very likely. However, as both theory and experiment continue to be developed and applied in common research projects, it can be expected that the actual understanding of the processes involved in reaction on model catalysts will substantially improve over the next 10 years. After all, the ability to trace reactions and to account for the position and charge state of each reactant is already a realization of what seemed 20 years ago a fiction rather than fact.

j115

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

116

Acknowledgments

We are grateful to EPSRC UK for the financial support (GT: EP/C541898/1; HL: EP/E062490/1). WAH acknowledges support from the Royal Society.

References 1 Wendt, S., Schaub, R., Matthiesen, J., Vestergaard, E.K., Wahlstr€om, E., Rasmussen, M.D., Thostrup, P., Molina, L.M., Laegsgaard, E., Stensgaard, I., Hammer, B., and Besenbacher, F. (2005) Surf. Sci., 598, 226. 2 Wendt, S., Matthiesen, J., Schaub, R., Vestergaard, E.K., Laegsgaard, M.D.E., Stensgaard, I., Besenbacher, F., and Hammer, B. (2006) Phys. Rev. Lett., 96, 066107. 3 Teobaldi, G. et al. (2007) Chem. Phys. Lett., 437, 73. 4 Enevoldsen, Georg H. et al. (2008) Phys. Rev. B, 78, 045416. 5 Hofer, W.A., Foster, A.S., and Sluger, A.L. (2003) Rev. Mod. Phys., 75, 1287. 6 Hofer, W.A. (2003) Prog. Surf. Sci., 71, 147. 7 Hofer, W.A. and Fisher, A.J. (2003) Phys. Rev. Lett., 91, 036803. 8 Hofer, W.A., Garcia-Lekue, A., and Brune, H. (2004) Chem. Phys. Lett., 397, 354. 9 Keldysh, L.V. (1965) Sov. Phys. JETP, 20, 1018. 10 Taylor, J., Guo, H., and Wang, J. (2001) Phys. Rev. B, 63, 245407. 11 Ness, H. and Fisher, A. (1997) Phys. Rev. B, 55; 56, 10081; 12469. 12 Stokbro, K., Thirstrup, C., Sakura, M., Quaade, U., Hu, B.Y.-K., Perez-Murano, F., and Grey, F. (1998) Phys. Rev. Lett., 80, 2618. 13 Hofer, W.A., Biedermann, A., Redinger, J., and Varga, P. (2000) Surf. Sci. Lett., 466, L795. 14 Hansen, O., Ravnkilde, S., Quaade, U., Stokbro, K., and Grey, F. (1998) Phys. Rev. Lett., 81, 5572.

15 Hofer, W.A., Fisher, A.J., Wolkow, R.A., and Gr€ utter, P. (2001) Phys. Rev. Lett., 87, 236104. 16 Gimzewski, J.K. and Moeller, R. (1987) Phys. Rev. B, 36, 1284. 17 Hallmark, V.M., Chiang, S., Raboli, J.F., Swalen, J.D., and Wilson, R.J. (1987) Phys. Rev. Lett., 59, 2879. 18 Africh, C., Lin, H., Corso, M., Esch, F., Rosei, R., Hofer, W.A., and Comelli, G. (2005) J. Am. Chem. Soc., 127, 11454. 19 Schaub, R., Thorstrup, P., Lopez, N., Laegsgaard, E., Stensgaard, I., Norskov, J.K., and Besenbacher, F. (2001) Phys. Rev. Lett., 87, 266104. 20 Di Valentin, C. (2007) J. Chem. Phys., 127, 154705. 21 Briggs, G. and Fisher, A. (1999) Surf. Sci. Rep., 33, 1. 22 Tersoff, J. and Hamann, D.R. (1985) Phys. Rev. Lett., 50, 1998. 23 Tersoff, J. and Hamann, D.R. (1985) Phys. Rev. B, 31, 805. 24 Bardeen, J. (1961) Phys. Rev. Lett., 6, 57. 25 B€ utticker, M., Imry, Y., Landauer, R., and Pinhas, S. (1985) Phys. Rev. B, 31, 6207. 26 Meir, Y. and Wingreen, N.S. (1992) Phys. Rev. Lett., 68, 2512. 27 Palotas, K. and Hofer, W.A. (2005) J. Phys.: Condens. Matter, 17, 2705. 28 Cerda, J., van Hove, M., Sautet, P., and Salmeron, M. (1997) Phys. Rev. B, 56, 15885. 29 Reuter, K., de Andres, P.L., Garcia-Vidal, F.J., and Flores, F. (2001) Phys. Rev. B, 63, 205325. 30 Feuchtwang, T.E. (1974) Phys. Rev. B, 10, 4135. 31 Feuchtwang, T.E. (1975) Phys. Rev. B, 12, 3979.

References 32 Feuchtwang, T.E. (1976) Phys. Rev. B, 13, 517. 33 Hofer, W.A. and Redinger, J. (2000) Surf. Sci., 447, 51. 34 Martin, R.M. (2004) Electronic Structure. Basic Theory and Practical Methods, Cambridge University Press, Cambridge. 35 Becke, A.D. (1993) J. Chem. Phys., 98, 5648. 36 Heyd, J., Scuseria, G.E., and Ernzerhof, M. (2003) J. Chem. Phys., 118, 8207. 37 Munoz Ramo, D. et al. (2007) Phys. Rev. B, 75, 205336. 38 Munoz Ramo, D. et al. (2007) Phys. Rev. Lett., 99, 155504. 39 Di Valentin, C. et al. (2006) Phys. Rev. Lett., 97, 166803. 40 Cora, F. et al. (2004) Struct. Bond., 113, 171. 41 Liu, L.M., McAllister, B., Ye, H.Q., and Hu, P. (2006) J. Am. Chem. Soc., 128, 4017. 42 Lindan, P.J.D. and Zhang, C. (2005) Phys. Rev. B, 72, 075439. 43 Tilocca, A., Di Valentin, C., and Selloni, A. (2005) J. Phys. Chem. B, 109, 20963. 44 D€ obereiner, J.W. (1823) Schwigg. J., 39, 1. 45 Larson, A.T. and Emett, P.H. (1925) J. Am. Chem. Soc., 47, 346. 46 Benton, A.F. and Emett, P.H. (1926) J. Am. Chem. Soc., 48, 632. 47 Benton, A.F. and Elgin, J.C. (1926) J. Am. Chem. Soc., 48, 3027. 48 Benton, A.F. and Elgin, J.C. (1927) J. Am. Chem. Soc., 49, 2426. 49 Smith, F.E. (1928) J. Phys. Chem., 32, 719. 50 Svintsova, L.G., Boldyreva, N.A., and Korniechuk, G.P. (1977) React. Kinet. Catal. Lett., 6, 357. 51 Domen, K., Naito, S., Soma, M., Onishi, T., and Tamaru, K. (1982) J. Chem. Soc., Faraday Trans. 1, 78, 845. 52 Binning, G., Rohrer, H., Gerber, Ch., and Weibel, E. (1982) Phys. Rev. Lett., 49, 57. 53 Binnig, G., Rohrer, H., Gerber, Ch., and Weibel, E. (1983) Phys. Rev. Lett., 50, 120. 54 Sachs, C., Hildebrand, M., V€olkening, S., Wintterlin, J., and Ertl, G. (2002) J. Chem. Phys., 116, 5759. 55 Bed€ urftig, K., V€olkening, S., Wang, Y., Wintterlin, J., Jacobi, K., and Ertl, G. (1999) J. Chem. Phys., 111, 11147.

56 V€olkening, S., Bed€ urftig, K., Jacobi, K., Wintterlin, J., and Ertl, G. (1999) Phys. Rev. Lett., 83, 2672. 57 Hla, S.W., Lacovig, P., Comelli, G. et al. (1999) Phys. Rev. B, 60, 7800. 58 Kresse, G. and Hafner, J. (1993) Phys. Rev. B, 47, 558. 59 Kresse, G. and Hafner, J. (1994) Phys. Rev. B, 49, 14251. 60 Kresse, G. and Furthm€ uller, J. (1996) Comput. Mater. Sci., 6, 15. 61 Kresse, G. and Furthm€ uller, J. (1996) Phys. Rev. B, 54, 11169. 62 Bl€ochl, P.E. (1994) Phys. Rev. B, 50, 17953. 63 Perdew, J.P., Chevary, V.A., Vosko, S.H., Jackson, K.A. et al. (1992) Phys. Rev. B, 46, 6671. 64 Calleja, F., Arnau, A., Hinarejos, J.J., Vazquez de Parga, A.L., Hofer, W.A., Echenique, P.M., and Miranda,1, R. (2004) Phys. Rev. Lett., 92, 206101. 65 Corriol, C., Calleja, F., Arnau, A., Hinarejos, J.J., Vazquez de Parga, A.L., Hofer, W.A., and Miranda, R. (2005) Chem. Phys. Lett., 405, 131. 66 Pivetta, M. et al. (2007) Phys. Rev. Lett., 99, 126104. 67 Hofer, W.A. et al. (2008) Nanotechnology, 19, 305701. 68 Henrich, V.E. and Cox, P.A. (1996) The Surface Science of Metal Oxides, Cambridge University Press, Cambridge. 69 Diebold, U. (2003) Surf. Sci. Rep., 48, 53. 70 Lindsay, R., Wander, A., Ernst, A., Montanari, B., Thornton, G., and Harrison, N.M. (2005) Phys. Rev. Lett., 94, 246102. 71 Thompson, S.J. and Lewis, S.P. (2006) Phys. Rev. B, 73, 073403. 72 Diebold, U., Anderson, J.F., Ng, K.-O., and Vanderbilt, D. (1996) Phys. Rev. Lett., 77, 1322. 73 Henderson, M.A. (1999) Surf. Sci., 419, 174. 74 Becker, J.H. and Hosler, W.R. (1965) Phys. Rev., 137, A1872. 75 Yagi, E., Hasiguti, R.R., and Aono, M. (1996) Phys. Rev. B, 54, 7945. 76 Yagi, E., Hasiguti, R.R., and Aono, M. (1996) Phys. Rev. B, 54, 7945.

j117

j 4 Theory of Scanning Tunneling Microscopy and Applications in Catalysis

118

77 Bikondoa, O., Pang, C.L., Ithnin, R., Muryn, C.A., Onishi, H., and Thornton, G. (2006) Nat. Mater., 5, 189. 78 Diebold, U., Lehman, J., Mhmoud, T., Kuhn, M., Leonardelli, G., Hebenstreit, W., Schmid, M., and Varga, P. (1998) Surf. Sci., 411, 137. 79 Suzuki, S., Fukui, K., Onishi, H., and Iwasawa, Y. (2000) Phys. Rev. Lett., 84, 2156. 80 Henderson, M.A., Epling, W.S., Peden, C.H.F., and Perkins, C.L. (2003) J. Phys. Chem. B, 107, 534.

81 Batzill, M. and Diebold, U. (2007) Phys. Chem. Chem. Phys., 9, 2307. 82 Kr€ uger, P. et al. (2008) Phys. Rev. Lett., 100, 055501. 83 Wendt, S., Sprunger, P.T., Lira, E., Madsen, G.K.H., Li, Z., Hansen, J.Ø., Matthiesen, J., Blekinge-Rasmussen, A., Laegsgaard, E., Hammer, B., and Besenbacher, F. (2008) Science, 320, 1755. 84 de Pablo, P.J. et al. (2000) Phys. Rev. Lett., 85, 4992. 85 Ordejon, P. et al. (1996) Phys. Rev. B, 153, R10441.

j119

5 Characterization and Modification of Electrode Surfaces by In Situ STM Dieter M. Kolb and Felice C. Simeone

5.1 Introduction

Structure–reactivity relations are among the key issues both in electrocatalysis and in heterogeneous catalysis [1, 2]. Hence, determining the structure of an electrode surface, which shows a given catalytic activity, is an important goal in electrochemistry, and likewise, tailoring the structure of electrode surfaces for maximum catalytic output has been a long-lasting desire. The routine use of single-crystal electrodes with structurally well-characterized surfaces laid the basis of structure–reactivity studies in electrochemistry [3]. However, while well-prepared flat single-crystal surfaces, preferably with the three low-index crystallographic orientations, tremendously increased our understanding of structure effects in adsorption reactions, it was quite clear that truly catalytic reactions will occur preferentially on surface defects. The latter, however, will escape detection by diffraction techniques, the commonly employed method to determine the structure of single-crystal surfaces. In this respect, scanning tunneling microscopy (STM) like its related scanning probe techniques plays an important role for catalysis because of their inherent ability to image surfaces in real space rather than in k-space. Under very special conditions, surfaces may be imaged while the reaction of interest is ongoing. This is limited to reactions where at least one partner gives sufficient contrast to be detected by STM [4]. A typical example of this is metal deposition. In the following section, we focus on imaging single-crystal electrode surfaces that are of relevance to electrocatalysis. We will first deal with flat, defect-free terraces as well as with more real surfaces with monoatomic high steps as the most common active sites. We will then explore various strategies for nanostructuring surfaces, for example, by repetitive oxidation–reduction cycles (ORCs). Soon after the invention of the STM as a tool for imaging surfaces in real space, it was discovered that the microscope could also be used (or misused) for surface manipulations, that is, for nanostructuring of surfaces [5]. The extremely close vicinity of the STM tip and the sample surface required by the tunnel process

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

120

inherently leads to an overlap of both electric double layers and hence to a tip influence on the surface processes under study. This seemingly undesirable drawback of tunnel microscopy is frequently employed for nanostructuring with hitherto unprecedented precision. While imaging bare and adsorbate-covered electrode surfaces, even with atomic resolution, has become a routine procedure, measuring electrochemical activities, that is, reaction currents, with a similarly high lateral resolution, still needs to be achieved. In this respect, scanning electrochemical microscopy (SECM) shows promise in accomplishing this goal in the near future. Since SECM is not the subject of this chapter, the interested reader is referred to a recent review [6].

5.2 In Situ STM: Principle, Technical Realization and Limitations 5.2.1 Principle Considerations for In Situ Operation

The principle of STM is well described in this book as well as in many other monographs [7–9]. In brief, a fine metal tip is brought into close proximity of the surface under study, typically 0.5–2.0 nm, so that the electrons can tunnel from one side to the other when a voltage UT is applied between the tip and the sample (Figure 5.1). Then, the tip is scanned across the surface with either the tunnel current kept constant via a feedback circuit (“constant current mode”) or the height of the tip kept constant (“constant height mode”) [8]. In the first case, which is the commonly used one, the surface topography is reflected in the voltage applied to the z-piezo

Figure 5.1 Schematic diagram showing the principle of STM.

5.2 In Situ STM: Principle, Technical Realization and Limitations

(z being the direction normal to the surface) in order to keep the tunnel distance s constant. In the second case, the surface topography can be obtained from the variation of the tunnel current. The exponential dependence of the tunnel current on the tunnel gap s is the reason for the extreme height sensitivity of the STM, which allows detecting height variations in the 0.01 nm range and below. The lateral resolution, of course, is considerably lower, the exact value depending on the tip shape. The image of single (nonperiodic) events, such as monoatomic high steps, is usually smeared out by tip shape convolution over 1–2 nm. Nevertheless, periodic structures with 0.3 nm distances like that from individual atoms of a single-crystal surface can be clearly imaged [10]. One has to keep in mind that STM images show contours of constant tunnel probability rather than height contours directly. Nevertheless, for simple cases such as a metal surface, both quantities are closely related to each other. The dependence of the tunnel current IT on the tunnel voltage UT and on other parameters is given for the ideal case in Eq. (5.1): pffiffiffiffiffi IT ¼ R1 ð5:1Þ 0 UT expðA fT sÞ; where fT is the tunnel barrier, R0 ¼ 12.90 kW is the resistance of a point contact, and A ¼ 10.25 eV1/2 nm1 [11]. The joint density of states (JDOS) for the electrons in the tip and the sample enter the equation for IT, and hence in principle, STM images contain some chemical information about the imaged species. For all practical purposes, however, it is fair to state that STM does not yield chemical information because such information is very indirect and often heavily masked by topographic effects. This statement holds primarily for electrochemical systems, and also for a great deal of the UHV studies, although there are a few very beautiful examples presented in the literature that demonstrate a clear chemical contrast for different atomic species of bimetallic surfaces [12–14]. Since for electrochemical systems, potential limitations (e.g., due to hydrogen evolution or tip oxidation) severely restrict the application of scanning tunneling spectroscopy (STS), chemical information can hardly be derived using an STM. Additional methods, such as cyclic voltammetry, in situ SXRD or even ex situ XPS, and Auger electron spectroscopy, are often crucial for a safe assignment of features in the STM image. Although the tunnel barrier fT varies with the tip–sample distance s, it resembles the local work function of the surface for the large s, but the numbers for an electrochemical environment are substantially different from those for UHV conditions. For metal electrodes in aqueous solution and s > 1 nm, tunnel barriers range between 1.0 and 2.0 eV, fT ¼ 1.5 eV being a typical value [15]. However, while for metals in UHV or in air and s < 0.5 nm, the tunnel barrier increases almost linearly from zero at point contact to a constant maximum value [16], fT in situ shows a very characteristic, potential-dependent variation with s, from which structure information about the electrochemical interface normal to the metal surface can be extracted [17]. A result from the so-called distance tunneling spectroscopy, IT ¼ f(s) at UT ¼const, is given in Figures 5.2 and 5.3. From a simple visual inspection of the various exponential decays of IT with increasing tip–sample distance s for

j121

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

122

Figure 5.2 Tunnel current as a function of tip–sample distance for Au(1 1 1) in 0.1 M H2SO4 for three different electrode potentials (vs. SCE): Positive of (    ), negative of (–    –), and at the pzc (—). s0 ¼ 0 refers to the tip position at the closest approach that could be experimentally achieved. (Reproduced with permission from Ref. [19].)

different electrode potentials, it becomes evident that the tunnel barrier markedly depends on the electrode potential. A quantitative evaluation of fT via Eq. (5.2)  
h2 d ln IT 2 ð5:2Þ fT ðsÞ ¼ 8m ds for three different electrode potentials is shown in Figure 5.3 [18, 19]. The three cases refer to potential values, positive and negative of the potential of zero charge (pzc) and at the pzc. An oscillatory behavior of fT is observed on either side of the pzc, that is, when the ions of the supporting electrolyte build up the solution side of the electric double layer. With the help of ab initio DFTcalculations, the maxima and minima of the tunnel barrier could be related to positive and negative excess charge densities and hence to the positions of ions in the double layer [18, 19]. At the pzc, that is, in the absence of any excess charge, and hence in essence in the sole presence of water, the barrier height varies with distance from the surface like for metal/air interfaces. These results emphasize once more the importance of ions in determining the electrochemical interface properties. For in situ investigations of electrode surfaces, that is, for the study of electrodes in an electrochemical environment and under potential control, the metal tip inevitably also becomes immersed into the electrolyte, commonly an aqueous solution. As a consequence, electrochemical processes will occur at the tip/solution interface as well, giving rise to an electric current at the tip that is superimposed on the tunnel current and hence will cause the feedback circuit and therefore the imaging process to malfunction. The STM tip nolens volens becomes a fourth electrode in our system that needs to be potential controlled like our sample by a bipotentiostat. A schematic diagram of such an electric circuit, employed to combine electrochemical studies with electron tunneling between tip and sample, is provided in Figure 5.4. To reduce the electrochemical current at the tip/solution

5.2 In Situ STM: Principle, Technical Realization and Limitations

Figure 5.3 Tunnel barrier fT as a function of tip–sample distance for Au(1 1 1) in 0.1 M H2SO4 for three different potentials (vs. SCE). s ¼ 0 refers to the surface plane of Au(1 1 1). fT (s) for Au(1 1 1) in air is also shown for comparison. For details see Refs [18, 19].

interface sufficiently enough to let the tunnel current control the feedback circuit of the microscope, two different actions have to be taken [20]: (a) the tip potential may be chosen such that it is close to its rest potential and (b) the area of the tip exposed to the electrolyte has to be reduced as much as possible by an appropriate isolation (see Section 5.2.2.1).

j123

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

124

Figure 5.4 Electric circuit for in situ STM, which allows sample and tip potentials to be controlled independent from each other.

5.2.2 Technical Realization

While the first STM studies of electrode surfaces were performed with self-built instruments, scanning tunneling microscopes for electrochemical use are nowadays commercially available at a price that hardly justifies the effort of homemade equipment. Nevertheless, new instrumental designs are now and then discussed in the literature, which are still worthwhile to be considered for special applications. There is, however, additional equipment required for the operation of an electrochemical STM, for which homemade designs may be advantageous over commercially available ones and hence is briefly mentioned here in terms of tip preparation and isolation, the electrochemical cell, and vibration damping. 5.2.2.1 Tip Preparation and Isolation Very fine tips are required for high lateral resolution. The most commonly used tip materials are tungsten and a platinum–iridium alloy (80 : 20). Tips are manufactured by electrochemical etching of a 0.25 mm thick wire in a lamella of solution (Figure 5.5) [21]. For W tips, the solution consists of 2 M NaOH (etching at 2.4 V DC), for Pt:Ir tips 3.4 M NaCN is used (and 4.2 V AC). For tungsten, both parts can be used as tips, while with Pt:Ir only the lower part is suitable for high-quality imaging. While from an electrochemical point of view, Pt:Ir tips are easier to handle, their potential range of stability being clearly larger than that for tungsten, W tips are sharper and yield better images. Atomically resolved images are preferably obtained from tungsten tips. We mention in passing that according to literature, Pt:Ir tips were frequently produced by simply cutting the wire with a pair of pliers, with reasonably good success as far as imaging is concerned. According to our experience, tips produced in this way are not very suitable for electrochemical studies, but we have used such tips for imaging surfaces in air. It seems important to cut the wire while it is being pulled.

5.2 In Situ STM: Principle, Technical Realization and Limitations

Figure 5.5 Setup for the tip production by electrochemical etching of a tungsten wire. (Reproduced with permission from Ref. [21].)

For in situ STM measurements, the tip is inevitably immersed into the electrolyte and acts as a fourth electrode with reactions occurring at the tip–electrolyte interface. To reduce the electrochemical current at the tip to a size well below the tunnel current at the tip, the area in contact with the solution must be reduced by coating the largest portion of the STM tip with an insulating layer. In the literature, various ways of insulating the tip have been described. In the past, Apiezon , a chemically very inert thermoplast, has been used with success [22, 23], but at present, electrophoretic paints are widely employed for tip insulation [24]. In both cases, the uncoated part of the tip is about 1 mm or less, leaving an area in contact with solution of the order of 108–107 cm2. The remaining electrochemical currents are generally smaller than 50 pA (which is below the detection limit of commercial STM potentiostats), and they no longer interfere with the imaging process. Besides the reduction of the electrochemically active area of the tip, the proper choice of the tip potential can also help in minimizing Faradaic currents through the tip/electrolyte interface. This requires the use of a bipotentiostat, which allows one to choose the tip potential independent of the sample potential with respect to a common reference electrode. Such a bipotentiostat is supplied by most STM manufacturers. It enables one to select a tip potential close to the rest potential of the tip, where by definition no Faradaic currents should flow. While this precaution was indeed necessary a few years ago, the tip insulation has meanwhile progressed to a point, where restrictions of the tip potential to values close to the rest potential are no longer necessary. This has been an important advancement because with a freely chosen tip potential, scanning tunneling spectroscopy becomes feasible, albeit in a very limited potential region dictated by the decomposition of water or the stability of the tip material against anodic oxidation.

j125

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

126

Figure 5.6 Electrochemical cell design as used in the authors’ laboratory for STM.

5.2.2.2 Electrochemical Cell The design of the electrochemical cell is largely determined by (a) the shape of the single-crystal electrodes to be studied and (b) the stringent requirements for a thorough cleaning before its use. The design of our electrochemical STM cells is shown in Figure 5.6. All parts are made of Kel-F, which is easy to clean and which resists strongly oxidizing agents, such as caroic acid (conc. H2SO4 þ 30% H2O2), and the cell is designed for single-crystal disks of about 10 mm diameter and 2 mm thickness. Since only the polished face of the single-crystal disk is in contact with the electrolyte, electrochemical experiments can in principle be performed in the STM cell, for example, for surface characterization by cyclic voltammetry. However, one has to keep in mind that the cell has been optimized for STM use rather than for electrochemical experiments and accordingly two major deficiencies prevent one from obtaining high-quality cyclic voltammograms, routinely recorded in normal electrochemical cells: (a) the STM cell is usually open to air, hence oxygen reduction distorts the current–potential curves and (b) it normally takes minutes to assemble the STM cell, which creates contamination problems. Consequently in most cases, STM images and the corresponding electrochemical characterization are obtained in different cells and different experiments. The electrolyte volume of the STM cells is usually very small (of the order of a 100 ml in the above-described case) and evaporation of the solution can create problems in long-term experiments. Miniature reference electrodes, mostly saturated calomel electrodes (SCE), have been described in the literature [25], although they are hardly used anymore in our laboratory for practical reasons: Cleaning the glassware in caroic acid becomes cumbersome. For most studies, a simple Pt wire, immersed directly into solution, is a convenient, low-noise quasireference electrode. The Pt wire is readily cleaned by holding it into a Bunsen flame, and it provides a fairly constant reference potential of EPt ¼ þ 0.55  0.05 V versus SCE for 0.1 M sulfuric or perchloric acid solutions ( þ 0.67  0.05 V for 0.1 M nitric acid), which has to be checked from time to time and for different solutions.

5.2 In Situ STM: Principle, Technical Realization and Limitations

5.2.2.3 Vibration Damping It is obvious from the principle of STM that the microscope has to be shielded from mechanical and acoustic vibrations of the outside world as much as possible to achieve good imaging quality, particularly if atomic resolution is required. After all, there are two macroscopic parts – tip and sample – that are only fractions of a nanometer apart, and this distance needs to be controlled within hundredths of a nanometer. Experience has shown that vibrations of the building with frequencies below 10 Hz are especially critical, that is, difficult to eliminate. A simple, yet very effective construction for vibration damping is described in Ref. [26]. It consists in essence of two platforms, a very heavy stone plate (about 200 kg) and a light one (e.g., a wooden board, onto which the STM rests), suspended on metal frames with springs that have vastly different force constants [27]. Needless to say that the preferred location for setting up an STM is the basement rather than the top floor of a building. The microscope is placed in a little Faraday cage, lined with foam rubber for damping acoustic waves. 5.2.3 Limitations

Possible limitations in the use of STM arise from the close proximity of the tip to that part of the sample that is imaged. Under normal imaging conditions, for example, IT ¼ 2 nA and UT ¼ 50 mV, the tip–substrate distance s can be estimated from Eq. (5.1) to be around 0.6 nm (with fT ¼ 1.5 eV [15]). Considering the fact that the electric double layer of a metal electrode in concentrated solution is about 0.3 nm thick [28, 29], the double layers of tip and substrate begin to merge and the ideal picture of a noninteracting tip is no longer valid under these conditions. For example, contact with the reference electrode for the imaged area right underneath the tip may be lost because the bulk electrolyte that carries the reference potential has been squeezed out. It has been shown that a Cu surface can be locally corroded right underneath the tip if a positive potential is applied to the tip rather than to the sample [30]. Another disturbance brought about the STM tip is the so-called tip shielding [31]. Considering a typical tip radius of a few tens of nanometers, tip and sample constitute an extreme example of a thin-layer cell with restricted diffusion of reactants to the imaged area (e.g., metal ions in metal deposition studies) and with iR-drops distorting the externally applied electrode potential. Hence, great care must be exercised when treating kinetic data acquired by an STM as absolute; the mere presence of the tip under tunneling conditions can strongly affect the kinetics of a reaction. Finally, some requirements with respect to the substrates under study should be mentioned. One may notice that practically all STM studies are performed with single-crystal electrodes and not with (industrially more relevant) polycrystalline samples. For one, this certainly has something to do with the high lateral resolution that the STM offers and the researcher wants to make use of. Rough surfaces would be too demanding for a feedback circuit, capable of reacting to atomic heights. Since mechanistic interpretations of electrochemical reactions require well-defined surface structures and atomically resolved images of bare and adsorbate-covered

j127

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

128

electrodes, one has to retreat to single-crystal electrodes. However, the STM-derived structure information stems from a tiny area of the electrode, typically 100 nm  100 nm, and needs to be compared with electrochemical data that inevitably represent the whole macroscopic electrode surface. Such a relation will be meaningful only if the structure information holds for the whole electrode surface. This is the case only for high-quality single-crystal surfaces.

5.3 Imaging Single-Crystal Surfaces of Catalytically Relevant Systems 5.3.1 Preparation and Imaging of Metal Single-Crystal Surfaces

Mechanistic interpretations of electrochemical processes, which involve adsorption of reaction intermediates or products, require in general the use of single-crystal electrodes with structurally well-defined surfaces. Classical examples are the oxidation of small organic molecules such as formic acid [32] or the underpotential deposition of metals [33, 34]. In the 1970s, right at the beginning of “electrochemical surface science,” single-crystal surfaces were prepared in a UHV chamber by sputtering and annealing, and their structure and cleanliness checked by electron diffraction and AES [35–37]. This was a rather cumbersome approach for electrochemists and limited the use of well-characterized electrodes to those groups that had access to surface science equipment. A significant advancement of single-crystal electrochemistry came with the so-called flame-annealing technique, which required in essence only a Bunsen burner to prepare clean and well-ordered surfaces, as first demonstrated by Clavilier et al. for platinum [38] and later by Hamelin for gold [39]. Although the initial advice of the French school, to quench rapidly the still hot crystal in water to reduce the danger of surface contamination as much as possible, had to be abandoned because the heat shock turned out to be detrimental to the bulk crystallinity, the resulting surface quality in retrospect has to be considered high. For platinum, which is particularly sensitive to contamination from air, cooling in an iodine [40] or CO [41] atmosphere was advocated, the adsorbed layer protecting the surface extremely well during transfer to the electrochemical cell and being finally removed from the surface by oxidative desorption. Ultimately, inductive heating in a reducing atmosphere turned out to be the best choice as this technique allows the preparation of clean and well-ordered surfaces of reactive metals such as Cu, Ag, Pd, Rh, and Ru for which the by now classical flame-annealing in ambient atmosphere has failed. This is particularly true for large single-crystal electrodes, commonly employed for spectroscopic studies, which due to their higher heat capacity require longer cooling times. Details of the technique can be found in Ref. [42]; a schematic diagram is given in Figure 5.7. The devastating influence of trace amounts of oxygen during cooling on the quality of a Pt single-crystal surface is demonstrated in Figure 5.8, where the STM images of Pt(1 1 1) in 0.1 M H2SO4 after cooling down the crystal in air and in hydrogen

5.3 Imaging Single-Crystal Surfaces of Catalytically Relevant Systems

Figure 5.7 Setup for the inductive heating of single-crystal electrodes in controlled atmosphere.

Figure 5.8 STM images of Pt(1 1 1) in 0.1 M H2SO4 at þ 0.35 V versus SCE, after cooling the sample in air (a) and in H2 (b). (Reproduced with permission from L.A. Kibler, personal communication.)

j129

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

130

Figure 5.9 STM images of (a) Pd(1 1 1) in 0.01 M H2SO4 and (b) Rh(1 1 1) in 0.1 M H2SO4. Both crystals were annealed in a H2-flame and cooled in H2. (Reproduced with permission from Ref. [43].)

are compared (L.A. Kibler, personal communication). While cooling in air leads to a rough surface, cooling in a reducing atmosphere such as H2 or H2/Ar mixtures yields large, atomically flat terraces. Other examples of well-prepared single-crystal surfaces are given in Figure 5.9, which shows Pd(1 1 1) and Rh(1 1 1) in 0.1 M H2SO4 [43]. Quite often atomically resolved in situ images of single-crystal surfaces are desirable because they would allow a precise length calibration of the piezos. However, zooming with the microscope into the terraces, one frequently images anion adlayers rather than the metal surface proper. Although interesting in their own right, these adlayers prevent direct viewing of the substrate. Sulfate, chloride, and metal chloro complexes are well known to form ordered adlayers [44], at least at high coverages, which are easy to image by STM with molecular resolution. Examples thereof are given in Figure 5.10 [45–48]. 5.3.2 Bimetallic Surfaces

Bimetallic surfaces, either alloys or a metal A onto which a metal B was deposited in submonolayer amounts, play an important role in electrocatalysis. For their structural characterization, a chemical contrast in the STM images would be highly desirable. So far, however, the number of such examples is vanishingly small, and in almost all cases, one has to rely on morphological (height) contrast. A rare example of a system showing chemical contrast is Pd and Au as has been demonstrated for Pd deposits on Au(1 1 1) [49] as well as for Pd–Au alloy surfaces [50]. When Pd is deposited from aqueous solution onto Au(1 1 1), nucleation starts exclusively at the monoatomic high steps of the substrate, followed by a two-dimensional growth of the Pd onto the lower terrace. Figure 5.11 shows the growth of a Pd layer that had nucleated at the rim of a monoatomic high gold island. Although the monolayers of both metals should have about the same height (the Pd layer being slightly lower, if at all), the gold island appears darker in the STM image than the surrounding Pd.

5.3 Imaging Single-Crystal Surfaces of Catalytically Relevant Systems

Figure 5.10 STM images of ordered anionic adlayers. (a) PdCl42 on Au(1 0 0) in 0.1 M H2SO4 þ 0.1 mM H2PdCl4 þ 0.6 mM HCl [45]; (b) PtCl42 on Au(1 0 0) in 0.1 M H2SO4 þ 0.1 mM K2PtCl4 [46]; (c) sulfate on Au(1 0 0) in 0.1 M H2SO4 [47]; (d) sulfate on Ag(1 0 0) in 0.1 M H2SO4 [48].

This chemical contrast may be due to electronic effects or caused by differences in anion adsorption on both metals. Evidence for electronic effects as possible origin of the chemical contrast between Pd and Au has been presented in STM images of Pd/Au alloy surfaces with atomic resolution, which allowed an identification of Pd or Au atoms on the basis of their brightness [50]. A similar picture is presented in Figure 5.12 showing the surface of a Pt50Ru50 alloy [43]. There are atoms that appear clearly brighter (about 0.04 nm higher), which in accordance with UHV–STM investigations [51, 52] could be assigned to Ru because of its higher electron density at the Fermi level. Their number, however, is much smaller than that of the Pt atoms, indicating a marked difference between bulk and surface composition of the alloy. Indeed, the corresponding cyclic voltammograms recorded in 0.1 M H2SO4 reveals a surface that is almost Pt(1 1 1)-like. From the image in Figure 5.12 it is concluded that the Ru atoms are more or less uniformly distributed over the surface and only small assemblies are formed. We mention in passing that cooling the Pt50Ru50 singlecrystal alloy after inductive heating in a reducing atmosphere yields the Pt-rich

j131

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

132

Figure 5.11 STM image of a growing Pd monolayer, which had nucleated on Au(1 1 1) at the rim of a gold island. Although practically equal in height, the Pd layer appears brighter than the gold island in the center. (Reproduced with permission from Ref. [49].)

surface, whereas cooling in an inert (Ar) atmosphere with traces of oxygen leads to a Ru-rich surface [53, 54]. From the image in Figure 5.12 one may guess how difficult the measurements and how limited the systems are that show a true chemical contrast. In most cases, one has to retreat to the morphological information in order to assign features to metal A or metal B. This is routinely done in metal deposition studies that start with an image of the bare surface, followed by the ones with the metal deposit at various stages, that is, various amounts. Numerous examples are given in the literature [26, 55, 56], one being reproduced in Figure 5.13. It shows Pt electrodeposited onto Au(1 1 1), the little hillocks on a flat substrate being easily identified as the Pt clusters [46]. With this image another well-established observation is confirmed: Nucleation starts preferentially at surface defects, the growing nuclei decorating the substrate’s defect structure. 5.4 Strategies for Nanostructuring Surfaces 5.4.1 Oxidation–Reduction Cycles for Roughening and Faceting Surfaces

The dominance of surface defects over terrace sites in catalysis and electrocatalysis had been recognized already in the early stages of surface science. For example,

5.4 Strategies for Nanostructuring Surfaces

Figure 5.12 Atomically resolved STM image of a Pt50Ru50(1 1 1) alloy electrode in 0.01 M NaF after annealing and cooling in H2/Ar. The arrows mark bright spots that are assigned to Ru atoms. (Reproduced with permission from Ref. [43].)

Figure 5.13 STM image of Pt clusters electrodeposited onto Au(1 1 1) in 0.1 M H2SO4 þ 0.1 mM K2PtCl4. E ¼ þ 0.1 V versus SCE. (Reproduced with permission from Ref. [46].)

j133

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

134

stepped single-crystal surfaces were used to make the role of monoatomic high steps in the substrate visible [57]. Likewise, surfaces with a regular roughness, that is, surfaces covered with islands or clusters of a narrow size distribution, may serve as model systems, for which size-reactivity relations can be derived. The application of oxidation–reduction cycles, repetitively applied to an electrode to create rough or facetted surfaces, has a long tradition in electrochemistry [58–62]. Particularly worth noting are the works of Arvia and his group [59, 60], in which faceting of polycrystalline Pt by ultrafast potential cycling has been described. It was shown that cycling in the kHz-region for an extended period of time (typically for about 1 h) caused either (1 1 1)- or (1 0 0)-type of facets to grow, depending on the negative and positive potential limits. In a systematic study on the influence of conventional ORCs, that is, with scan rates on the order of 10–100 mV s1, on the surface structure of Au(1 1 1), it was demonstrated that slow potential cycling from the oxide formation region back to the reduced state caused monoatomic deep holes in the surface, whereas fast cycling or potential stepping led to clusters on the surface in addition to the holes [63]. The place exchange between metal and oxygen during oxide formation leads to metal adatoms on the surface upon oxide reduction. In the first case (slow potential cycling), the adatoms apparently are given enough time to be incorporated at nearby monoatomic high steps of the substrate (the Ehrlich–Schwoebel barrier would prevent them to fall into the advacancies), whereas the advacancies condense to vacancy islands (holes that are visible in STM images). Fast oxide reduction, for example, due to potential stepping leads to cluster formation on flat terraces because of the quickly established large supersaturation potential. Repetitive potential cycling or stepping enhances the above-described effect and surface roughnesses emerge that should be of interest in the study of model catalysts. It has been demonstrated for gold that specifically adsorbing ions such as Cl drastically enhance surface diffusion, which is the basis of the so-called electrochemical annealing [64, 65]. Hence, by selecting the appropriate parameters for the ORC and choosing the right electrolyte composition, a tailoring of surface roughness seems feasible. Figure 5.14 shows the STM image of an originally flat Au(1 1 1) terrace, which was subjected to 100 potential cycles at 100 mV s1 between 0.7 and 1.3 V versus SCE in 0.1 M H2SO4. The clusters have an average height of six–eight layers (K€ontje et al., in preparation). 5.4.2 Surface Modification by an STM: An Overview

Inspired by the amazing successes of surface scientists in nanostructuring surfaces with the tip of an STM, albeit at UHV conditions and often at low temperatures [66–68], electrochemists began to use an STM or AFM as a tool for nanostructuring electrode surfaces, mostly by spatially confined metal deposition. Figure 5.15 summarizes the various routes, which are currently employed in the community for electrochemical nanostructuring. In the following, we shall briefly address seven of them, and devote a separate chapter to the case sketched in

5.4 Strategies for Nanostructuring Surfaces

Figure 5.14 STM image of a Au(1 1 1) electrode, roughened by about 100 oxidation–reduction cycles at 100 mV s1 in 0.1 M H2SO4. Cycling between 0.7 and 1.3 V versus SCE. Image taken at þ 0.05 V versus SCE. (Reproduced with permission from K€ ontje et al., in preparation.)

Figure 5.15h because this approach is intensively pursued in the authors’ laboratory. The first successful attempts of electrochemical nanostructuring, pioneered by Penner et al. [69], involved the generation of surface defects by the tip at predetermined positions, which were created either by a mechanical contact between tip and substrate (tip crash) or by some sort of sputtering process, initiated by high-voltage

Figure 5.15 Various approaches to electrochemical nanostructuring with an STM, currently employed by the community.

j135

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

136

Figure 5.16 Tapping-mode AFM image of a 25 Cu nanodot array on H-terminated p-Si(1 0 0), formed by nanoindentation. (Reproduced with permission from Ref. [72].)

pulses applied to the tip [70]. These artificially created defects then acted as very effective nucleation centers for metal deposition, which allowed the decoration of electrode surfaces by metal clusters on a nanometer scale (Figure 5.15a). While in the beginning, this technique had been applied almost exclusively to metal substrates, studies were extended more recently to semiconductor surfaces. Impressive examples of patterned nanostructures made of metals such as Cu, Ag, or Au on silicon wafers were given by Homma et al. [71, 72]. In those cases, however, the defects that acted as nucleation centers were generally made by a nanoindentation process via an AFM tip (Figure 5.16). A slightly different approach to spatially confined metal deposition, which is less harmful to the substrate, is sketched in Figure 5.15b. It is the local removal of an overlayer that causes a high overpotential for metal deposition. By choosing an electrode potential slightly negative of the Nernst potential, where no metal deposition will take place on top of the overlayer, deposition will immediately set in upon removal of the tarnishing film by the tip of an STM or AFM sliding across the surface, at the freed portion of the surface only. The applicability of this approach has been demonstrated in an AFM study for Cu deposition onto an oxide-covered Cu surface [73] and in an STM study for Cu deposition onto Au(1 1 1) covered by a monolayer of sodium dodecyl sulfate (SDS) [74]. Although the precision of the metal nanostructures generated in such a way was far from being satisfactory [74], this method again reveals the potential of decorating semiconductor surfaces with metal nanostructures, while so far the studies have been restricted to metal on

5.4 Strategies for Nanostructuring Surfaces

metal. This, however, will require suitable, preferably organic molecules, which will adsorb strongly enough on the semiconductor electrode to form a dense monolayer with sufficient inhibition for metal deposition, but which can be removed by the tip without damaging the substrate. An obvious way of generating metal structures of nanometer dimensions via an STM tip is sketched in Figure 5.15g: It is the burst-like dissolution of metal from the tip, onto which it had been deposited from solution, and the redeposition onto the substrate within a narrow region directly underneath the tip [75]. In a systematic study by Schindler et al., it was demonstrated how to achieve redeposition of the metal dissolved from the tip and at the same time prevent metal deposition from solution onto the substrate directly [76]. The key lies in the momentarily high metal ion concentration after the sudden metal dissolution at the tip that causes a more positive Nernst potential for the surface region underneath the tip. Figure 5.17 shows the STM image of two Pb clusters, about 3 nm in height, generated by the burstlike dissolution of Pb from the STM tip and by redeposition onto H-terminated n-Si(1 1 1) [77]. The potential of this technique lies in the ability to decorate semiconductor surfaces with metal clusters. A conceptually different approach to nanostructuring electrode surfaces by tipgenerated metal clusters is sketched in Figure 5.15h. This approach, which facilitates a so-called jump-to-contact between tip and substrate for generating metal clusters, has been developed by our group and will be described in more detail in Section 5.4.3.

Figure 5.17 STM image of H-terminated n-Si(1 1 1) in 0.1 M HClO4 þ 1 mM Pb(ClO4)2, onto which two Pb clusters have been deposited by a burst-like dissolution of Pb from the STM tip. (Reproduced with permission from Ref. [77].)

j137

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

138

The remaining methods sketched in Figure 5.15 either deal with spatially confined oxidation/dissolution of the substrate or describe means of studying electrochemical reactions on a nanometer scale. We will first refer to what could be termed as double-layer crosstalk (Figure 5.15c). As mentioned already in Section 5.2.3, commonly employed tunnel parameters for imaging (e.g., IT ¼ 2 nA and UT ¼ 50 mV) lead to tip–substrate distances around 0.6 nm. This implies that the double layers of tip and substrate begin to merge, and the assumption of a noninteracting tip is no longer valid. For example, the close proximity of the tip may cause a change in potential for the imaged area directly underneath the tip because direct contact with the reference electrode is lost. It was demonstrated that spatially confined copper dissolution directly underneath the tip can be achieved by applying to the tip a potential positive of the Cu/Cu2 þ reversible potential E0, despite the fact that the sample potential was held clearly negative of E0 [30]. Hence, copper was oxidatively dissolved by a tip–sample interaction underneath the tip and there only, although this process should not be possible at that sample potential. Later, this double-layer crosstalk was used to selectively dissolve Ag overlayers [78], demonstrating that this tip-induced metal dissolution is by no means restricted to Cu only as has been claimed in the literature [79]. It was also shown [30] that there is actually a smooth transition between imaging without much tip interference and tip-induced surface processing. Depending on the potentials of tip and sample, the following three regimes could be distinguished: (1) tip-enhanced copper deposition; (2) mere surface imaging; and (3) tip-induced copper dissolution. In several publications, Schuster and coworkers have shown the use of STM tips (or other thin metal wires) as tools for electrochemical machining of electrode surfaces on a micrometer scale [80–82]. Spatially confined etching was achieved by applying nanosecond voltage pulses between tool and sample making use of the vastly different time constants for double-layer charging for different parts of the tool. As is sketched in Figure 5.15d, the variation of the time constant t ¼ RC for doublelayer charging is solely due to the electrolyte resistance R, which increases tremendously when comparing that part of the tip (or tool) next to the sample surface with those higher up parts. The voltage pulse duration is now chosen in such a way that only for the forefront of the tool, that is, in a spatially very defined region, an electrode potential is established at the sample surface large enough for oxidative dissolution. In a series of impressive images, the viability of this route to electrochemical microand nanostructuring has been demonstrated (Figure 5.18). A clever design for local oxide formation on silicon surfaces is depicted in Figure 5.15e. Operation of an STM in humid air leads to a neck of liquid due to capillary forces. Applying a voltage between tip and sample will trigger simple electrochemical processes in such a miniature electrochemical cell. Avouris et al. have used this method for pattering a Si surface with oxide [83]. The creation of nanostructured surfaces is one thing, the study of electrochemical reactions on such nanostructures is another one. Especially in electrocatalysis, where size effects on reactivity are often discussed, there have been attempts to use the tip of an STM as a detector electrode for reaction products from, say, catalytically active metal nanoclusters [84]. However, such ring-disk-type approaches are questionable,

5.4 Strategies for Nanostructuring Surfaces

Figure 5.18 Scanning electron microscopy image of a microcantilever, electromachined into a stainless steel sheet by ultrashort voltage pulses (100 ns, 2 V, 1 MHz repetition rate) in 3 M HCl þ 6 M HF. The tool electrode was a tiny loop of a 10 mm thick Pt wire. (Reproduced with permission from Ref. [80].)

when it comes to a quantitative analysis because of the ill-defined (if not to say unknown) geometry of the tip, which does not allow reliable mass transport calculations. On the other hand, scanning electrochemical microscopy has been demonstrated to ideally fulfill all these requirements, albeit on a micrometer scale [85, 86]. A major breakthrough in applying SECM for nanostructuring was achieved by Heinze and coworkers [87, 88], who developed Pt nanodes with active diameters down to 20 nm and glass insulation around them that ensure defined diffusion conditions, which are essential for a quantitative evaluation of reaction rates (Figure 5.15f). The potential of SECM for electrocatalytic studies on a nanoscale may even exceed that of the STM, provided the miniaturization of the electrodes will routinely reach the nanometer length scale. For a recent review, see Ref. [6]. 5.4.3 Metal Nanocluster Deposition via Jump-to-Contact

Most of the work on nanostructuring electrode surfaces, which can be found in the literature, deals with the deposition of small metal clusters at predetermined positions. Over the years, we have developed a technique that is based on the “jump-to-contact” between tip and substrate [89] (Figure 5.15h) and that allows the formation of metal clusters in quick succession and without destroying the single crystallinity of the substrate. The principle behind this method is sketched in Figure 5.19 [90, 92]: By applying an electrode potential to the STM tip that is slightly

j139

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

140

Figure 5.19 Schematic diagram for continuous cluster generation by the tip of an STM via jump-to-contact.

negative of the bulk deposition potential for the metal ions in solution, metal is deposited from the electrolyte onto the tip. Then, the metal-loaded tip is made to approach the surface close enough so that the jump-to-contact can occur. This leads to the formation of a connective neck, a metal bridge between tip and substrate, which will break upon the subsequent retreat of the tip, leaving a small metal cluster on the substrate surface. The tip is automatically “reloaded” because of the ongoing metal deposition and hence is ready for the next cluster formation. The jump-to-contact requires an approach of the tip down to about 0.3 nm tunnel gap, which must be externally controlled. Actually, in our case all three spatial coordinates of the tip are externally controlled by a microprocessor, which makes the nanodecoration of an electrode surface with metal clusters a fully automated process, allowing even complex patterns to be fabricated rapidly and reproducibly. Two examples of tip-induced cluster formation are given in Figure 5.20, both referring to Cu on Au(1 1 1) in sulfuric acid solution [93]. Image (a) shows a circle of 12 Cu clusters on Au(1 1 1), all 0.8 nm in height. The pattern in image (b) not only proves the feasibility of complex structure formation but also demonstrates that monoatomic high steps in the substrate surface are no obstacles for nanostructuring, as the feedback control of the STM is not switched off. Although we have described various aspects of this method in a number of publications [91, 92], some technical details are briefly mentioned again for the sake of convenience:

5.4 Strategies for Nanostructuring Surfaces

Figure 5.20 Two examples for the nanodecoration of a Au(1 1 1) electrode by tip-generated Cu clusters. Electrolyte: 0.05 M H2SO4 þ 1 mM CuSO4. (Reproduced with permission from Refs [90, 93].)

.

If metal deposition is fast (as in the case of Cu in sulfuric acid solution), cluster generation can be performed at kHz rates. Obtaining an array of 10 000 Cu clusters on Au(1 1 1) takes a couple of minutes [15]. Typical parameters are 10–20 ms pulses at a rate of 50–80 Hz.

.

Despite cluster formation via the STM tip, the imaging quality of the latter surprisingly remains high. Hence, writing and reading is possible with one and the same tip.

.

The cluster size can be varied at will within a given range by changing the tip approach, the latter being controlled externally. Variation of cluster size with tip approach has been demonstrated for several metals on Au(1 1 1) [92, 94, 95].

.

The high stability of the metal clusters allows one to hold the sample potential slightly positive of the Nernst potential, typically at þ 10 mV versus Cu/Cu2 þ in the case of copper. Thus, “normal” electrodeposition onto the sample directly from solution is prevented, whereas the tip-generated Cu clusters remain on the surface [96].

.

Depending on the cohesive energies of cluster and substrate material, the jumpto-contact occurs from the tip to the substrate (e.g., for Cu and a gold electrode) or from the substrate to the tip (e.g., for Ni on the tip and a gold electrode) [93].

.

So far, more than a dozen systems have been investigated and tested for nanostructuring [97]. While in the beginning most studies dealt with Cu clusters for testing and developing the method, our more recent work focused on Pd clusters for electrocatalytic investigations. A Au(1 1 1) surface with 12 arrays, each containing 2500 Pd clusters, is shown in Figure 5.21. Although such seemingly very large number of clusters are not sufficient for characterization by ordinary cyclic voltammetry, there may be a good chance to do so with an SECM, using a nanode that matches the cluster field in dimension.

j141

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

142

Figure 5.21 STM image of 12 cluster fields on Au(1 1 1), each field containing 2500 Pd clusters. Electrolyte: 0.1 M H2SO4 þ 1 mM PdSO4.

Two aspects deserve particular mention as they are of great practical relevance. The first deals with the unexpectedly high stability of the tip-generated clusters against anodic dissolution, which was briefly addressed above [96]. This is again demonstrated for Cu clusters on Au(1 1 1) in Figure 5.22, where the height of a tipgenerated cluster is shown as a function of potential, as the latter is scanned from þ 10 mV versus Cu/Cu2 þ to þ 250 mV. While bulk Cu would be quickly dissolved at an overpotential of 10 mV, and Cu upd is completely desorbed at þ 250 mV, the Cu cluster is still seen to exist, albeit at reduced height. Note that this information was obtained by scanning the tip in x-direction at constant y-position. Hence, a falsification of the potential values caused by possible tip-shielding effects can be ruled out. Although a reasonable explanation of the high cluster stability still has to be found, alloy formation as an obvious cause can safely be ruled out: Cluster dissolution brings back a perfectly flat, bare surface (see STM images in Figure 5.22 before and after anodic cluster dissolution [93]), whereas monoatomic deep holes are generally found on the surface for those cases where clusters did form an alloy with the substrate. The second aspect deals with room-temperature salt melts, the so-called ionic liquids (ILs), which open up many new and interesting perspectives for electrochemical studies [98]. The main advantage of ILs over aqueous solutions is their extremely wide stability range of almost 4 V, which has to be compared with the 1.23 V for water. Hence, electrodeposition of very unnoble, that is, reactive metals

5.4 Strategies for Nanostructuring Surfaces

Figure 5.22 (a) STM image of an array of 225 tip-generated Cu clusters on Au(1 1 1) in 0.05 M H2SO4 þ 0.1 mM CuSO4. (b) Same area, but after dissolution of the Cu clusters at þ 300 mV versus SCE. (c) Height of a single tip-generated Cu cluster as a function of potential and time, demonstrating the unusually high stability of the cluster against anodic dissolution. (Reproduced with permission from Ref. [93].)

Figure 5.23 Ring of 48 tip-generated Fe clusters on Au(1 1 1) in 1-butyl-3-methyl-imidazolium BF4 þ approximately 50 mM FeCl3. (Reproduced with permission from Ref. [95].)

j143

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

144

such as Fe or Ta, becomes feasible. Even deposition of semiconductors such as Ge from ILs has been successfully demonstrated [99]. Tip-induced nanostructuring has also been performed in ILs [94, 95]. We end this chapter by showing a ring of Fe clusters generated on Au(1 1 1) in 1-butyl-3-methyl-imidazolium tetrafluoroborate, a typical and widely used IL, by jump-to-contact (Figure 5.23). Here again, the unusually high stability of the Fe clusters, which even exceeds that of upd Fe is noteworthy [95].

Acknowledgments

Parts of this work were supported by the Deutsche Forschungsgemeinschaft through SFB 569 and the Fonds der Chemischen Industrie. One of us (FS) gratefully acknowledges a stipend of SFB 569.

References 1 Trasatti, S. (ed.) (2003) Electrocatalysis: from theory to industrial applications. Electrochim. Acta, 48, 3727–3974. 2 Ertl, G. (2008) Angew. Chem. Int. Ed., 47, 3524. 3 Markovic, N.M. and Ross, P.N. (2002) Surf. Sci. Rep., 45, 117. 4 Wintterlin, J., V€olkening, S., Janssens, T.V.W., Zambelli, T., and Ertl, G. (1997) Science, 278, 1931. 5 Eigler, D.M. and Schweizer, E.K. (1990) Nature, 344, 524. 6 Pust, S.E., Maier, W., and Wittstock, G. (2008) Z. Phys. Chem., 222, 1463. 7 Bonnell, D.A. (1993) Scanning Tunneling Microscopy and Spectroscopy, VCH, Weinheim. 8 Wiesendanger, R. (1994) Scanning Probe Microscopy and Spectroscopy, Cambridge University Press, Cambridge. 9 G€ untherodt, H-.J. and Wiesendanger, R. (eds) (1994) Scanning Tunneling Microscopy I–III, Springer, Berlin. 10 Wintterlin, J., Wiechers, J., Brune, H., Gritsch, T., H€ofer, H., and Behm, R.J. (1989) Phys. Rev. Lett., 62, 59. 11 Simmons, J.G. (1963) J. Appl. Phys., 34, 1793.

12 Schmid, M., Stadler, H., and Varga, P. (1993) Phys. Rev. Lett., 70, 1441. 13 Wouda, P.T., Nieuwenhuys, B.E., Schmid, M., and Varga, P. (1996) Surf. Sci., 359, 17. 14 Ruff, M., Takehiro, N., Liu, P., Nørskov, J.K., and Behm, R.J. (2007) ChemPhysChem., 8, 2068. 15 Engelmann, G.E., Ziegler, J.C., and Kolb, D.M. (1998) Surf. Sci., 401, L420. 16 Kuk, Y. and Silverman, P.J. (1990) J. Vac. Sci. Technol., A8, 289. 17 (a) Hugelmann, M. and Schindler, W. (2003) Surf. Sci., 541, L643; (b) (2004) J. Electrochem. Soc., 151, E97. 18 Simeone, F.C., Kolb, D.M., Venkatachalam, S., and Jacob, T. (2007) Angew. Chem. Int. Ed., 46, 8903. 19 Simeone, F.C., Kolb, D.M., Venkatachalam, S., and Jacob, T. (2008) Surf. Sci., 602, 1401. 20 Kolb, D.M., Nichols, R.J., and Behm, R.J. (1992) Electrified Interfaces in Physics, Chemistry and Biology, NATO ASI, Vol. C 355 (ed. R. Guidelli), Kluwer, Dordrecht, p. 275. 21 Wintterlin, J. (1988) PhD Thesis, Free University Berlin.

References 22 Wiechers, J., Twomey, T., Kolb, D.M., and Behm, R.J. (1988) J. Electroanal. Chem., 248, 451. 23 Nagahara, L.A., Thundat, T., and Lindsay, S.M. (1989) Rev. Sci. Instrum., 60, 3128. 24 Bach, C.E., Nichols, R.J., Beckmann, W., Meyer, H., Schulte, A., Besenhard, J.O., and Jannakoudakis, P.D. (1993) J. Electrochem. Soc., 140, 1281. 25 See, for example, Dakkouri, A.S. (1996) PhD Thesis, University Ulm. 26 Kolb, D.M. (2002) Advances in Electrochemical Science and Engineering., Vol. 7 (eds R.C. Alkire and D.M. Kolb), WileyVCH Verlag, Weinheim, p. 107. 27 Pohl, D.W. (1986) IBM J. Res. Dev., 30, 417. 28 Bockris, J.O’M., Devanathan, M.A.V., and M€ uller, K. (1963) Proc. Roy. Soc. Lond., A274, 55. 29 Gerischer, H. (1997) The CRC Handbook of Solid State Electrochemistry, Chapter 2 (eds P.J. Gellings and H.J.M. Bouwmeester), CRC Press, Boca Raton. 30 Xie, Z.-X. and Kolb, D.M. (2000) J. Electroanal. Chem., 481, 177. 31 Stimming, U., Vogel, R., Kolb, D.M., and Will, T. (1993) J. Power Sources, 43/44, 169. 32 Adzic, R.R., Tripkovic, A.V., and O’Grady, W.E. (1982) Nature, 296, 137. 33 Schultze, J.W. and Dickertmann, D. (1976) Surf. Sci., 54, 489. 34 J€ uttner, K. and Lorenz, W.J. (1980) Z. Phys. Chem. N.F., 122, 163. 35 Hubbard, A.T. (1973) Crit. Rev. Anal. Chem., 3, 201. 36 O’Grady, W.E., Woo, M.Y.C., Hagans, P.L., and Yeager, E. (1977) J. Vac. Sci. Technol., 14, 365. 37 Kolb, D.M. (1987) Z. Phys. Chem. N.F., 154, 179. 38 Clavilier, J., Faure, R., Guinet, G., and Durand, R. (1980) J. Electroanal. Chem., 107, 205, 211. 39 Hamelin, A., Doubova, L., Wagner, D., and Schirmer, H. (1987) J. Electroanal. Chem., 220, 155. 40 Wasberg, M., Palaikis, L., Wallen, S., Kamrath, M., and Wieckowski, A. (1988) J. Electroanal. Chem., 256, 51.

41 Kibler, L.A., Cuesta, A., Kleinert, M., and Kolb, D.M. (2000) J. Electroanal. Chem., 484, 73. 42 Kibler, L.A. Preparation and Characterization of Noble Metal Single-Crystal Electrodes, http://www.uni-ulm.de/ echem/index.html?id¼1023001. 43 Hoyer, R. (2004) PhD Thesis, University Ulm. 44 Magnussen, O.M. (2002) Chem. Rev., 102, 679. 45 Kibler, L.A., Kleinert, M., and Kolb, D.M. (2000) Surf. Sci., 461, 155. 46 Waibel, H.-F., Kleinert, M., Kibler, L.A., and Kolb, D.M. (2002) Electrochim. Acta, 47, 1461. 47 Cuesta, A., Kleinert, M., and Kolb, D.M. (2000) Phys. Chem. Chem. Phys., 2, 5684. 48 Schweizer, M. and Kolb, D.M. (2003) Surf. Sci., 544, 93. 49 Kibler, L.A., Kleinert, M., Randler, R., and Kolb, D.M. (1999) Surf. Sci., 443, 19. 50 Maroun, F., Ozanam, F., Magnussen, O., and Behm, R.J. (2001) Science, 293, 1811. 51 Varga, P., Schmid, M., and Redinger, J. (2000) Physik in Unserer Zeit, 31, 215. 52 Diemant, T., Hager, T., Hoster, H.E., Rauscher, H., and Behm, R.J. (2003) Surf. Sci., 541, 137. 53 Kibler, L.A. and Kolb, D.M. (2003) Z. Phys. Chem., 217, 1265. 54 El-Aziz, A.M. and Kibler, L.A., in preparation. 55 Nichols, R.J., Bunge, E., Meyer, H., and Baumg€artel, H. (1995) Surf. Sci., 335, 110. 56 Hoyer, R., Kibler, L.A., and Kolb, D.M. (2003) Electrochim. Acta, 49, 63. 57 Somorjai, G.A. (1977) Adv. Catal., 26, 1. 58 Lehmpfuhl, G., Uchida, Y., Zei, M.S., and Kolb, D.M. (1999) Imaging of Surfaces and Interfaces (eds J. Lipkowski and P.N. Ross), Wiley-VCH Verlag, Weinheim. 59 Cerviño, A.M., Triaca, W.A., and Arvia, A.J. (1985) J. Electrochem. Soc., 132, 266. 60 Canullo, J.C., Triaca, W.E., and Arvia, A.J. (1984) J. Electroanal. Chem., 175, 337. 61 Canullo, J., Uchida, Y., Lehmpfuhl, G., Twomey, T., and Kolb, D.M. (1987) Surf. Sci., 188, 350.

j145

j 5 Characterization and Modification of Electrode Surfaces by In Situ STM

146

62 Twomey, T.A. (1989) J. Electroanal. Chem., 270, 465. 63 Schneeweiss, M.A. and Kolb, D.M. (1997) Solid State Ionics, 94, 171. 64 Stickney, J.L., Villegas, I., and Ehlers, C.B. (1989) J. Am. Chem. Soc., 111, 6473. 65 Will, T., Dietterle, M., and Kolb, D.M. (1995) Nanoscale Probes of the Solid/Liquid Interface, NATO ASI, Vol. E288 (eds A.A. Gewirth and H. Siegenthaler), Kluwer, Dordrecht, p. 137. 66 Crommie, M.F., Lutz, C.P., and Eigler, D.M. (1992) Science, 262, 218. 67 Meyer, G., Bartels, L., Z€ophel, S., and Rieder, K.H. (1999) Appl. Phys. A, 68, 125. 68 Cuberes, M.T., Schlitter, R.R., and Gimzewski, J.K. (1997) Surf. Sci., 371, L231. 69 Li, W., Virtanen, J.A., and Penner, R.M. (1992) Appl. Phys. Lett., 60, 1181. 70 Nyffenegger, R.M. and Penner, R.M. (1997) Chem. Rev., 97, 1195. 71 Homma, T., Kubo, N., and Osaka, T. (2003) Electrochim. Acta, 48, 3115. 72 Kubo, N., Homma, T., Hondo, Y., and Osaka, T. (2005) Electrochim. Acta, 51, 834. 73 LaGraff, J.R. and Gewirth, A.A. (1994) J. Phys. Chem., 98, 11246. 74 Petri, M. and Kolb, D.M. (2002) Phys. Chem. Chem. Phys., 4, 1211. 75 Schindler, W., Hofmann, D., and Kirschner, J. (2001) J. Electrochem. Soc., 148, C124. 76 Schindler, W., Hofmann, D., and Kirschner, J. (2000) J. Appl. Phys., 87, 7007. 77 Schindler, W., Hugelmann, P., Hugelmann, M., and K€artner, F.X. (2002) J. Electroanal. Chem., 522, 49. 78 Garcia, S.G., Salinas, D.R., Mayer, C.E., Lorenz, W.J., and Staikov, G. (2003) Electrochim. Acta, 48, 1279. 79 Lang, G., Bakos, I., and Horanyi, G. (2000) J. Electroanal. Chem., 493, 141. 80 Kirchner, V., Cagnon, L., Schuster, R., and Ertl, G. (2001) Appl. Phys. Lett., 79, 1721. 81 Kirchner, V., Xia, X.H., and Schuster, R. (2001) Acc. Chem. Res., 34, 371.

82 Schuster, R. and Ertl, G. (2003) Catalysis and Electrocatalysis at Nanoparticle Surfaces (eds E. Savinova C. and Vayenas), Marcel Dekker, New York. 83 Avouris, P., Hertel, T., and Martel, R. (1997) Appl. Phys. Lett., 71, 285. 84 Meier, J., Friedrich, K.A., and Stimming, U. (2002) Faraday Discuss., 121, 365. 85 Bard, A.J. and Mirkin, M.V. (eds) (2001) Scanning Electrochemical Microscopy, Marcel Dekker, New York. 86 Zhou, J., Zu, Y., and Bard, A.J. (2000) J. Electroanal. Chem., 491, 22. 87 Baltes, N., Thouin, L., Amatore, C., and Heinze, J. (2004) Angew. Chem. Int. Ed., 43, 1431. 88 Ufheil, J., Hess, C., Borgwarth, K., and Heinze, J. (2005) Phys. Chem. Chem. Phys., 7, 3185. 89 Landman, U., Luedtke, W.D., Burnham, N.A., and Colton, R.J. (1990) Science, 248, 454. 90 Kolb, D.M., Ullmann, R., and Will, T. (1997) Science, 275, 1097. 91 Kolb, D.M. (2001) Angew. Chem. Int. Ed., 40, 1162. 92 Kolb, D.M. and Simeone, F.C. (2005) Electrochim. Acta, 50, 2989. 93 Kolb, D.M., Engelmann, G.E., and Ziegler, J.C. (2000) Solid State Ionics, 131, 69. 94 Wang, J.G., Tang, J., Fu, Y.C., Wei, Y.M., Chen, Z.B., and Mao, B.W. (2007) Electrochem. Commun., 9, 633. 95 Wei, Y.M., Zhou, X.S., Wang, J.G., Tang, J., Mao, B.W., and Kolb, D.M. (2008) Small, 4, 1355. 96 Kolb, D.M., Engelmann, G.E., and Ziegler, J.C. (2000) Angew. Chem. Int. Ed., 39, 1123. 97 Engelmann, G.E. (1997) PhD Thesis, University Ulm. 98 Endres, F., Abbott, A.P., and MacFarlane, D.R. (eds) (2008) Electrodeposition from Ionic Liquids, Wiley-VCH Verlag, Weinheim. 99 Endres, F. (2004) Z. Phys. Chem., 218, 255.

j147

6 STM Imaging of Oxide Nanolayer Model Systems Falko P. Netzer and Svetlozar Surnev

6.1 Introduction

Scanning tunneling microscopy (STM) has become an invaluable tool for the study of atomic processes at surfaces. Amongst other advantages, such as resolution at the atomic level, high sensitivity, and providing information on a local scale, “seeing” the atomic structures directly in real space at surfaces is a significant benefit for fueling the inspiration of scientific intuition. The application of STM to problems in heterogeneous catalysis has thus been widely recognized since it allows one to gain insight into the nanoscience aspects behind catalysis, with the ultimate goal of imaging surface reactions with atomic or molecular resolution. The application of STM to catalysis requires suitable model systems, preferably of planar geometry, which are often based on oxide model surfaces that are covered in this chapter. Metal oxides are ubiquitous in heterogeneous catalysis as support surfaces in oxide-supported metal catalysts or as active surface phases in oxide catalysis. Epitaxially grown thin films of oxides on metal substrates constitute the most promising route to generate clean, well-ordered oxide surfaces, which exhibit a number of technical advantages over their corresponding bulk-terminated oxide single-crystal surfaces. Oxide single crystals of suitable size are often difficult to obtain, sample handling including heating and cooling is difficult, and the charging problem of insulating oxide samples upon exposure to charged particle probes (such as electrons) is experimentally difficult to overcome. These problems can be avoided by using thin oxide films grown on metallic substrates for the investigation of surface properties [1, 2], and thin films have therefore become the preferred oxide phases employed in model studies of fundamental catalytic research. Originally, it was thought that thin films of oxides mimic closely the respective properties of the corresponding bulk-terminated surfaces [1]; however, more recently it has become clear that this is true only if the films are beyond a certain critical thickness [3]. Oxide films on metal supports thinner than this critical thickness – which is typically of the order of 2–3 nm – display novel properties due to confinement and interfacial

j 6 STM Imaging of Oxide Nanolayer Model Systems

148

proximity effects, which make them very different from their respective bulk counterparts [3, 4]. These ultrathin films – for convenience we will call them oxide nanolayers – may be regarded as new kinds of material, which do not occur in nature but are characterized by novel emergent properties of structure, electronic and magnetic behavior, and chemical reactivity. The nanometer scale in the thickness dimension means that the oxide nanolayers are only a few unit cell thick and that the interfacial bonding to the metal substrate and charge transfer effects across the metal–oxide interface may be important ingredients in determining their structure and stability. Moreover, low-dimensionality aspects may play a role in the advent of emergent phenomena. Oxide nanolayers may be classified into two categories: (i) surface oxides, which are generated by the oxidation of the corresponding bulk metal samples and (ii) nanolayers of particular oxides supported on a different metal. Surface oxides have attracted considerable interest recently as intermediate phases between the chemisorbed oxygen layers on metal surfaces and the corresponding bulk oxides [5–7], and their particular significance in some catalytic reactions has been pointed out [8, 9]. Oxide nanolayers supported on different metals constitute a different kind of systems, which allow one to design nanostructured materials with tunable properties via appropriate choices of metal substrate and oxide overlayer [10, 11]. They thus form versatile and novel catalytic model systems, both as supports for the deposition of metal nanoparticles and as active catalytic phases themselves. Of particular interest are the epitaxially ordered oxide nanolayers that grow on noble metal single-crystal surfaces. It is this class of oxide nanolayers and their emergent physical and chemical properties as probed by STM that are addressed in this chapter. A special case of nanostructured oxide-metal systems can be created by submonolayer coverages of oxides on metal single crystals, where the surfaces expose areas of both bare metal substrate and oxide islands. This forms the basis of the so-called “inverse model catalyst” concept [12]. An “inverse model catalyst” surface is a planar system consisting of a metal single-crystal surface that is decorated with oxide island structures, that is, it is the reverse of a real catalyst where the metal component is oxide supported. This inverse catalyst system contains both metal and oxide surface sites and sites at the metal-oxide phase boundary; it is therefore an attractive model system to study the role of the metal-oxide interface in catalytic reactions, as involved, for example, in the promotor effect of oxide minority components that leads to the enhancement of the reactivity of the underlying metal [13]. It has been shown that the length of the metal-oxide phase boundary can be controlled by careful tuning of deposition parameters of the oxide, thus allowing the assessment of the kinetics of the promotion effect [14]. Since the main topic of this review is STM imaging, growth properties, surface morphology, and atomic structures of oxide nanosystems are the central themes. Oxide nanolayers on noble metal surfaces often display very complex structural arrangements, as illustrated in the following sections. The determination of the surface structure of a complex oxide nanophase by STM methods is, however, by no means trivial; resolution at the atomic scale in STM is a necessary but not sufficient condition for elucidating the atomic structure of an oxide nanophase. The problem

6.2 Experimental Aspects and Technical Developments

with interpreting the STM images of oxide surfaces is that the physical and chemical nature of the (bright or dark) contrast in the images is not always clear: bright protrusions may be due to metal or oxygen atoms, they may represent single atoms or groups of atoms. Since novel metal–oxygen coordination spheres and building blocks may be present in oxide nanostructures that do not occur in the known bulk oxide compounds, it is necessary to combine the STM data with as much structural and spectroscopic information as possible; results from other experimental techniques are necessary to obtain complementary information on long-range order and lattice parameter (LEED), oxidation state, oxygen–metal coordination and stoichiometry (XPS, NEXAFS) of the oxide phases. Moreover, the experimental data have to be combined with high-level theoretical modeling, including both oxide overlayer and support surfaces, to resolve the complicated oxide phases. Since very large unit cells are frequently involved containing several tens of atoms, advanced DFT methodology is necessary. Indeed, the progress of the last decade in resolving complex oxide structures has been made possible by this close combination of high-level experimental and theoretical methods [15]. The structure of the review is organized as follows. In Section 6.2, we will address experimental aspects concerning apparatus developments and oxide nanolayer preparation methods, and briefly comment on the interplay between experimental and theoretical results. Section 6.3 constitutes the main body of this chapter, where we present case studies of selected oxide-metal systems. They have been chosen according to their prototypical oxide nanosystem behavior and because of their importance in catalysis. We conclude with a synopsis and a brief outlook speculating on future developments.

6.2 Experimental Aspects and Technical Developments

Since the invention of the STM by Binnig and Rohrer about 25 years ago [16], this technique has become one of the most powerful tools for surface structure observation, in real space and at atomic length scales. The ongoing technical development and improvement of STM instruments has largely extended the scope of their application in a number of different disciplines – physics, chemistry, biology, and engineering. One of the major apparatus developments was the building of variabletemperature STMs, which allow imaging at temperatures well above or below room temperature. In many surface phenomena, for example, surface phase transitions, surface diffusion, or crystal growth, the temperature plays a key role. However, STM instruments are susceptible to temperature variations of the sample–tip unit, causing large thermal drifts (both parallel and perpendicular to the sample surface), which prevent from keeping the same surface area within the microscope’s view. Recently, several variable-temperature STMs have been designed by various research groups and commercial companies (e.g., Omicron, JEOL, RHK Technology, WA Technology), where the thermal drift has been minimized in one or the other way (the number of the various technical modifications has grown enormously over the past

j149

j 6 STM Imaging of Oxide Nanolayer Model Systems

150

decade, and their review is beyond the scope of this paper). The strengths of employing a variable-temperature STM instrument in the field of the oxide nanolayer growth are obvious and have been demonstrated recently in the study of oxidation– reduction processes [17], structural phase transitions [18], and surface diffusion [19] of vanadium oxide nanostructures on Pd(1 1 1) and Rh(1 1 1) surfaces, as well as for monitoring in situ the growth of nickel oxide films on a Pt(1 1 1) substrate at elevated temperature [20]. At the other extreme, the use of STM at liquid He temperatures allows one to immobilize the surface species and thus investigate their local electronic density of states (LDOS) by scanning tunneling spectroscopy (STS). This technique has recently been applied to study the homogeneity of the surface potential [21] or the appearance of Kondo-like resonances [22] on FeO bilayers on Pt(1 1 1), as well as for identifying the electronic signatures of different building units of vanadium oxide phases on Rh(1 1 1) [23]. Because the tunneling effect is not restricted to a vacuum environment, the STM has the potential of overcoming the so-called pressure gap in heterogeneous catalysis. This has been realized in diverse high-pressure STM designs [9, 24–27], which enable one to achieve atomic resolution during adsorption/desorption and reaction experiments on single-crystal surfaces up to atmospheric pressures. In close relation to the topic of the present review, Frenken and coworkers [9, 28] have performed experiments in a high-pressure flow reactor and shown that surface oxides, forming on the Pt(1 1 0) and Pt(1 0 0) surfaces under oxygen-rich conditions, have a dramatic effect on the CO oxidation rate; these surface oxides have been suggested to be the catalytically active phase. High-pressure STMs, when combined with the recent development of the control electronics allowing high-speed scanning [29], will make the STM an invaluable tool for monitoring surface dynamic processes or catalytic reactions in real time. Typically, ultrathin oxide films have been prepared by physical vapor deposition of metal atoms onto a dissimilar single-crystal metal substrate in an oxygen atmosphere: this is the so-called reactive evaporation (RE) method. Alternatively, the metal atoms can be deposited first in UHV onto the metal substrate and subsequently oxidized; this is the so-called postoxidation (PO) method. Which of these two preparation techniques is applied with success in a specific case depends on the interfacial chemistry, where factors such as the rate of oxygen dissociation on the metal substrate, oxidation reaction, and bulk diffusion need to be carefully considered. In most of the cases, molecular oxygen has been used as an oxidant in both the RE and the PO methods, and the oxygen content in the oxide nanolayer has been controlled by the oxygen pressure and the substrate temperature. Although the reactivity of many metals (e.g., Mg, Al, or Ni) is sufficient to form fully oxidized films using molecular oxygen in a high-vacuum environment (106 mbar), this is not the case for other metals (e.g., Fe). The alternative is to use atomic oxygen sources, such as an electron cyclotron resonance (ECR) oxygen plasma generator or gaseous NO2 [30]. More recently, a high-pressure cell has been employed by Kuhlenbeck and coworkers [31, 32], where oxygen pressures up to 50 mbar have been applied to fully oxidize ultrathin V-oxide films to a V2O5 stoichiometry. A modification of the traditional PO approach has been reported recently by Matsumoto et al. [33, 34], who

6.2 Experimental Aspects and Technical Developments

grew TiOx nanolayers on a Pt(1 0 0) surface, by forming an ordered Pt3Ti surface alloy prior to the oxidation step with ozone. This procedure is to some extent similar to the method used to form Al-oxide nanolayers via oxidation of NiAl(1 1 0) and Ni3Al(1 1 1) bulk alloy crystal surfaces, as mentioned in Section 6.3.1, and was found to be beneficial for the growth of well-ordered Ti-oxide overlayers, with an atomically flat surface morphology, compared to films prepared by the conventional RE method with molecular oxygen. An original approach for growing two-dimensional (2D) oxide overlayers has been developed by Niehus et al. [35–37]. Here, a Cu3Au(1 1 0) alloy surface has been used as a substrate, which when bombarded by O þ ions and subsequently annealed in UHV becomes oxygen-terminated. This smooth Cu3Au(1 1 0)-O surface, on the one hand, prevents the intermixing between the evaporated metal atoms and the substrate, and, on the other hand, significantly acts as an oxygen reservoir with subsurface oxygen stored near the surface, which can be released in a controlled fashion by thermal treatment. This preparation technique has been successfully applied for growing atomically flat VOx [35, 36], NbOx, and MoOx [37] overlayers. As mentioned in the introduction, the determination of oxide overlayer structures beyond the simplest cases requires the combination of a high-level experimental and theoretical methodology. As experimentalists, we refrain from a discussion of the theoretical aspects, but some words on how the models derived by the theoretical calculations can be checked against the experiments may be appropriate. For a given structure model, which has been found to be stable on total energy grounds, simulations of the STM images, calculation of core-level binding energies and X-ray absorption profiles, and of phonon frequencies give suitable data for comparison with experimental values. Core-level binding energies of ultrathin oxide overlayers on metal surfaces as obtained by X-ray photoelectron spectroscopy (XPS) are problematic to compare with corresponding values from bulk oxide samples because of the proximity of the underlying metal surfaces. The latter gives rise to interfacial bonding and screening effects of the photoionized final state that modify the measured binding energy; typically, lower core-level binding energies of metal cations than in the corresponding bulk oxides are obtained, making the oxidation state difficult to evaluate. Here, calculations are often necessary to make progress and, conversely, the agreement between theoretical predictions and experimental findings give credence to a derived model. The phonon frequencies of oxide nanolayers have been found to be particularly diagnostic [18] because they can be interpreted to good approximation in a local building block picture. The agreement between calculated and measured phonon frequencies, the latter obtained in, for example, high-resolution electron energy loss spectroscopy experiments, is a stringent test on the validity of a structural model. The combination of state-of-the-art first-principles calculations of the electronic structure with the Tersoff–Hamann method [38] to simulate STM images provides a successful approach to interpret the STM images from oxide surfaces at the atomic scale. Typically, the local energy-resolved density of states (DOS) is evaluated and isosurfaces of constant charge density are determined. The comparison between simulated and measured high-resolution STM images at different tunneling

j151

j 6 STM Imaging of Oxide Nanolayer Model Systems

152

voltages, therefore corresponding to the imaging of the spatial distribution of different electronic states, is a very rigorous test for a given structural model.

6.3 Case Studies: Selected Oxide–Metal Systems 6.3.1 Alumina Nanolayers on NiAl Alloys

Ultrathin alumina layers on metallic substrates play a key role in many technologies, for example, as dielectric and diffusion barrier layers in electronic devices and as corrosion resistant coatings; and in model studies of catalytic processes, they are widely used as stable supports for metal cluster growth [39]. The selective oxidation of alloys provides a very promising route to synthesize highly ordered thin alumina films [40], and in 1991 the Freund group has published a comprehensive paper on the formation of a well-ordered aluminum oxide overlayer by high-temperature oxidation of NiAl(1 1 0) [41], taking up earlier reports in the literature [42, 43]. Jaeger et al. [41], applying a variety of electron spectroscopic techniques including LEED, concluded that the oxide film is about 5 Å thick and most likely consists of two Al layers and two quasi-hexagonal oxygen layers with oxygen surface termination. A distorted hexagonal Al2O3-type surface such as a-Al2O3 (0 0 0 1) or g-Al2O3(1 1 1) has been proposed. Further progress has been made by Libuda et al. by the application of spot profile analysis low-energy electron diffraction (SPA-LEED) and STM [44], where two reflection domains inclined by 24 with respect to the [1 1 0] direction of the NiAl(1 1 0) substrate have been identified. The alumina overlayer structure is commensurate along the NiAl [1 1 0] direction but incommensurate along [0 0 1]. The first atomically resolved images of AlOx on NiAl(1 1 0) have been recorded with a low-temperature STM by Kulawik et al. [45] using a sample bias of 0.5 V, that is, a voltage that is within the bandgap of the oxide overlayer (see Figure 6.1). The tunneling current is thus mediated by the substrate, with the oxide overlayer exerting only a modulating influence. It was suggested that mainly the topmost oxygen layer atoms are being imaged, whereas the contributions from the Al3 þ sublattice are very small, but a possible influence of the properties of the tip has been emphasized [45]. Nevertheless, the atomic structure of the alumina overlayer on NiAl(1 1 0) remained obscure and could not be resolved in this study. The structure of alumina on NiAl(1 1 0) was the subject of a surface X-ray diffraction study by Stierle et al. [46]. The model derived by Stierle et al. from the analysis of the X-ray diffraction data was based on a strongly distorted double layer of hexagonal oxygen ions, where the Al ions are hosted with equal probability on octahedral- and tetrahedral-coordinated sites; the resulting film structure was closely related to bulk k-Al2O3. An attractive feature of Stierle’s model was that it provided a natural explanation of the domain structure of the alumina overlayer, which is induced by a periodic row matching between film and substrate lattices. However, as pointed out recently by Kresse et al. [47], this structure model has two bonds with

6.3 Case Studies: Selected Oxide–Metal Systems

Figure 6.1 Atomically resolved STM image of a thin Al2O3 film on NiAl(1 1 0), taken at 4 K with U ¼ 500 mV, I ¼ 1 nA (58 Å  40 Å). All atoms exhibit a quasi-hexagonal coordination, where the three characteristic directions are marked with arrows. The unit cell was determined from the repetition of atoms. Here, gray-filled circles mark the most prominent atomic features,

thus characterizing six positions of the cell. Furthermore, a zigzagged pattern is observed, where the atoms appear alternately larger and smaller. Differences in the apparent height are emphasized with black and gray circles, respectively. (Reproduced with permission from Ref. [45].)

unphysically short AlAl and AlO lengths, and was found to be unstable in ab initio DFT calculations. The riddle of the structure of alumina on NiAl(1 1 0) was finally solved by Kresse and collaborators [47] with the help of STM data and extensive theoretical DFT modeling calculations. The essential breakthrough was that the idea of hexagonal oxygen planes, taken previously from the structure of practically all known Al2O3 bulk phases, has been given up and has been replaced by a square arrangement of oxygen atoms. The latter has been suggested after a close inspection of STM images. Figure 6.2 shows the STM results and the resulting DFT-derived model. As suggested before by Kulawik et al. [45], the positions of the oxygen atoms in the surface layer have been taken from the room-temperature STM images (Figure 6.2b and c), whereas the Al surface atoms, almost coplanar with these O atoms, have been associated with bright protrusions in low-temperature STM images under certain tunneling conditions (Figure 6.2e). It was also pointed out by Kresse [47] that the atomic resolution in the low-temperature STM images at moderately high tunneling voltages may not be simply because of LDOS effects but rather caused by an adsorbate at the tip interacting with surface Al atoms. The surface thus consists of oxygen atoms arranged in squares and triangles and Al atoms in between slightly below but almost coplanar with the oxygen layer. The oxide film also has another oxygen and Al layer at the interface with the substrate. The stoichiometry of the film is Al4O6Al6O7 in

j153

j 6 STM Imaging of Oxide Nanolayer Model Systems

154

Figure 6.2 (a) Top and (b) side view of the DFT- and STM-based model for the ultrathin aluminum oxide film on NiAl(1 1 0). (c–e) Experimental STM images of the film at (c) and (d) room temperature and at (e) low temperature. Sample bias voltage and tunneling current values are (c) 2.5 mV/1.4 nA, (d) 0.2 V/ 0.9 nA, and (e) 0.5 V/0.3 nA. Two oxide unit cells are marked by white rectangles, the diagonal

along which the oxide is commensurate is yellow, and the parallelogram enclosed by the yellow and white lines delimits the simulation cell. Green rectangles and squares highlight oxygen atoms in a square arrangement. Circles indicate the Al and O positions from (a) and (b) in the corresponding colors. (f) Close-up of the structure. (Reproduced with permission from Ref. [47].)

contrast to the usual Al2O3 stoichiometry. The registry of the substrate and oxide is provided by the interfacial Al atoms, which are bonded to the Ni rows of the NiAl part of the interface. This leads to interfacial row matching and to the stress along the [1 1 0] direction of the oxide, which causes the incorporation of line defects and the observed antiphase domain boundaries [45]. The latter have also been modeled in terms of an oxygen-deficient defect structure in the recent work of Kresse and coworkers [48]. The NiAl(1 1 0) surface is not the only Al-containing alloy surface that can support a highly ordered oxide overlayer. The formation of a well-ordered alumina film on Ni3Al(1 1 1) by high-temperature oxidation has been reported first by Rovida et al. [49, 50]. The Wandelt group has taken up the idea subsequently and has refined the preparation procedure [51]. Rosenhahn et al. recognized from STM images [52] that two long-range superstructures in the nanometer range can be identified: the so-called “network” structure with a periodicity of approximately 2.4 nm and the “dot” structure with a periodicity of approximately 4.2 nm, the dot structure being a (H3  H3)R30 subset of the network structure. SPA-LEED and more extensive STM investigations of Degen et al. [53] revealed that only the dot structure is a real

6.3 Case Studies: Selected Oxide–Metal Systems

superstructure of the alumina film, and they refined the unit cell of the (H67  H67) R48 superstructure to 4.16 nm. Since atomic resolution of the alumina film could not be achieved in this latter work, the building blocks of the oxide layer and its structural relation at the atomic scale to the substrate could not be established. The full structural complexity of the alumina on Ni3Al(1 1 1) has been unraveled by the noncontact atomic force microscopy (AFM) study by Gritschneder et al. [54], who found that the main structural element of the oxide film is a lattice of hexagons that is pinned to the periodicity of the substrate. The surface unit cell is defined by the dot structure, with the network structure being formed by a honeycomb-like topographic modulation as a result of the substrate-overlayer pinning. Recently, Kresse and coworkers [55] have applied their experience with the alumina model on NiAl(1 1 0) to the determination of the structure of alumina on Ni3Al(1 1 1). Combining experimental STM images with extensive DFTmodeling, they found a structural similarity between the alumina films on the two NiAl alloy substrates, with an AlOAlO stacking sequence, square and triangular arrangements of atoms at the surface, and a hole in the unit cell reaching down to the Ni3Al substrate. In simulated STM images, these holes have been identified with the bright contrast of dots in the “dot” structure [55]. It is these holes that provide the anchoring centers for the growth of monodisperse metal clusters [56] and make this alumina overlayer an excellent nanotemplate for growing regular arrays of nanoparticles. In summary, the alumina nanolayers formed by the high-temperature oxidation on NiAl alloy surfaces are structurally and chemically very different from the bulkterminated surfaces of the various Al2O3 phases, and they thus provide very prototypical examples of oxide phases with novel emergent properties because of interfacial bonding and thickness confinement effects. 6.3.2 Titanium Oxide Nanolayers

There is considerable interest in titanium oxides owing to their importance in many areas, such as photo-assisted oxidation, heterogeneous catalysis and gas sensors [57–59]. Titanium dioxide (TiO2), in particular, is one of the most prominent materials in the industrial catalysis used for the selective reduction of NOx [60, 61], photocatalysis for pollutant elimination [62] or organic synthesis [63], photovoltaic devices [64], sensors [65], and paints [66]. Additional applications include its use as a food additive [67], in cosmetics [68], and as a potential tool in cancer treatment [69]. Titanium oxide is one of the best-characterized model systems in surface science [70]. Most of the studies were performed on single-crystal surfaces, but at present, ultrathin Ti-oxide films grown on metallic substrates are a subject of intensive investigation. Self-assembled, low-dimensional Ti-oxide nanostructures could be of potential interest for applications in electronic devices, nanocatalysts, and gas sensors. Controlling the oxidation state and the dimensions of these nanostructures may allow the production of a new class of technologically important materials. Ultrathin Ti-oxide films have been grown on various metal surfaces, Pt(1 1 1) [71–75], Pt(1 0 0) [33, 34], Pt(1 1 0) [76], Mo(1 0 0) [77, 78], Mo(1 1 0) [79, 80],

j155

j 6 STM Imaging of Oxide Nanolayer Model Systems

156

Mo(1 1 2) [81], W(1 0 0) [82], Ni(1 1 0) [83–86], and Ru(0 0 0 1) [87]. The group of Somorjai was first to report the growth of ordered TiOx overlayers on a Pt(1 1 1) substrate [71]. The as-grown films were disordered but annealing between 500 and 700  C in an oxygen background caused the formation of a hexagonal (H43  H43) R7.6 superstructure, which was observed in LEED and STM for coverages ranging between 1 and 5 ML, and was tentatively ascribed to a reconstructed TiO2(1 1 1) surface. Annealing at higher temperatures in UHV produced a second ordered structure, with a reduced Ti4O7 stoichiometry, according to XPS data [71]. Ten years after this work, Granozzi and coworkers [72] reexamined this system, focusing their investigation on the submonolayer to monolayer coverage range and revealed the complexity of the TiOx-Pt(1 1 1) phase diagram in great detail. These researchers evaporated Ti reactively onto a clean Pt(1 1 1) surface, kept at room temperature, in an oxygen pressure of 1  106 mbar. By varying the latter in a postoxidation step between UHV and 1  106 mbar, they observed in LEED and STM six different oxide structures in the coverage range up to 1.2 ML; these are summarized in Figure 6.3. At low Ti coverage (0.4 ML) and after annealing at 400  C in 107 mbar O2, a hexagonal overlayer with a lattice constant of 6 Å has been obtained, which is incommensurate with the Pt(1 1 1) substrate (Figure 6.3a). For this kagom
e-phase (k-TiOx), the authors have tentatively proposed a structure model with a formal Ti2O3 stoichiometry, where the Ti and O ions are arranged in a honeycomb lattice, identical to that of surface-V2O3 layers on Pd(1 1 1) [88]. At higher Ti coverages, between 0.8 and 2 ML, and annealing temperature of 700  C and oxygen pressure of 5  106 mbar, a rectangular structure (Figure 6.3b) grows in the form of rectangular-shaped islands. On the basis of XPS results and because of the similarity of the STM images to those reported for a rect-VO2 phase on Pd(1 1 1) by Surnev et al. [89], this rectangular structure has been ascribed to an interfacestabilized phase with a TiO2 stoichiometry. Striking similarities to the structure of V-oxide overlayers on Pd(1 1 1) [15] and Rh(1 1 1) [18, 90] surfaces have also been recognized for Ti-oxide phases with a “zigzag”- (Figure 6.3c and d) and “wagonwheel”-like appearance (Figure 6.3e and f) in the STM images [72]. More recent work of the same group, combined with DFT calculations [74, 75], has yielded structural models for the “zigzag” TiOx phases in terms of layers with Pt–Ti–O stacking and stoichiometries varying between TiO1.33 (z-TiOx) and TiO1.2 (z0 -TiOx). The most reduced TiOx phases on Pt(1 1 1) comprise the commensurate “wagonwheel” structures, w-TiOx (corresponding to the hexagonal (H43  H43)R7.6 superstructure, reported initially by Boffa et al. [71]) and w0 -TiOx (with a (7  7) R21.8 periodicity), which have been interpreted as TiO(1 1 1) bilayers because of their characteristic contrast in the STM images, in close similarity to bilayers of FeO(1 1 1) on Pt(1 1 1) [91] and VO(1 1 1) on Rh(1 1 1) [18] surfaces. Ultrathin Ti-oxide films grown on Pt(1 0 0) by Matsumoto et al. [33, 34] show also a rich phase diagram of interface-stabilized oxide structures as a function of the Ti coverage and the oxide preparation conditions. An interesting contrast to this general growth behavior of oxide nanolayers display the titania bilayers on a (1  2)-Pt(1 1 0) surface, investigated by Sambi and coworkers [76]. Here, reactive evaporation of approximately 2 MLTi in an oxygen atmosphere of 1  106 mbar was found to result

6.3 Case Studies: Selected Oxide–Metal Systems

Figure 6.3 STM images of various TiOx structures on Pt(1 1 1): (a) k-TiOx (30 Å  30 Å, 0.4 V, 1.06 nA); (b) rect-TiO2 (90 Å  90 Å, þ 0.8 V, 1.8 nA); (c) z-TiOx (60 Å  60 Å, þ 0.1 V, 1.5 nA); (d) z0 -TiOx (90 Å  90 Å, þ 0.8 V, 1.5 nA); (e) w-TiOx (75 Å  75 Å, þ 1.3 V, 1.9 nA); (f) w0 -TiOx (126 Å  126 Å, þ 0.2 V, 1.0 nA). (Reproduced with permission from Ref. [72].)

j157

j 6 STM Imaging of Oxide Nanolayer Model Systems

158

in the formation of a wetting Ti-oxide layer on the Pt(1 1 0) surface, which exhibits a regular array of dark stripes in STM images, corresponding to a (14  4) superstructure (Figure 6.4a, c, and d). The high-resolution image (Figure 6.4b) reveals that a rectangular overlayer unit cell with dimensions 3.0 Å  3.9 Å is at the root of this coincidence structure, and it is compatible with the unit cell of 2D titania nanosheets with the lepidocrocite structure, which can be obtained by delamination of layered titanates (for a model see the inset in Figure 6.4b). DFTcalculations performed in the same work have revealed that such a lepidocrocite nanosheet is the most stable 2D titania phase, irrespective of the presence of the substrate, and it can be considered as a distorted anatase (0 0 1) bilayer obtained by sliding the upper atomic layer with respect to the lower one by half a unit cell in the [1 0 0] direction. By this transformation, the resulting lepidocrocite nanosheet is stabilized by 1.24 eV with respect to the undistorted anatase (0 0 1) bilayer. The authors have concluded that it is the reduced dimensionality of the oxide overlayer, rather than the interaction

Figure 6.4 (a) Large-area STM image (620 Å  620 Å, þ 0.48 V, 1.4 nA) of a singledomain lepidocrocite nanosheet on (1  2)Pt(1 1 0). The central brighter area is separated from the lower terrace by a substrate monoatomic step. Inset: (14  4) LEED pattern. (b) High-resolution STM image (137 Å  137 Å , þ 0.28 V, 1.65 nA) of the nanosheet. The overlayer (superstructure) unit cell is indicated by the small (large) rectangle. Inset:

Tersoff–Hamann simulation of the STM image with superimposed solid sphere model (oxygen: dark gray/red; titanium: light gray/blue). Dark stripes along [0 0 1] are due to overlayer bending toward the substrate (see the interface side view in the [0 0 1] plane, lower left corner; Pt: large, gray). (c),(d) High-resolution STM images of the nanosheet (136 Å  136 Å, þ 0.42 V, 0.9 nA and 180 Å  180 Å, þ 1.80 V, 1.0 nA). (Reproduced with permission from Ref. [76].)

6.3 Case Studies: Selected Oxide–Metal Systems

with the substrate, that stabilizes the lepidocrocite structure, in contrast to what has been typically observed for other ultrathin oxide phases on metal substrates, reviewed in the present work. These findings are also supported by results reported for TiO2 layers on Pt(1 1 1) [72] and Ni(1 1 0) [85], which exhibit a structure very similar to that of the lepidocrocite nanosheets. The group of Thornton has used a Ni(1 1 0) substrate with the aim to grow epitaxially TiO2 layers, which can mimic the behavior of rutile TiO2(1 1 0) singlecrystal surfaces. Postoxidation in an oxygen pressure of 1  107 mbar at 800 K of Ti films with a coverage exceeding 1 ML was reported to result in 3D Ti-oxide islands with a structure resembling the surface of the bulk rutile TiO2(1 1 0) surface [83–85]. For a Ti coverage between 1 and 2 ML, the unit cell dimensions of the rutile-like islands were 2.9 Å  6.2 Å, that is, slightly contracted from single-crystal TiO2(1 1 0) (2.96 Å  6.5 Å), but thicker films display the relaxed rutile surface lattice constant. The question to what extent such titania layers can resemble bulk samples is of significant relevance for many technological applications and was treated in a recent study by this group [86]. Here, titania films with a thickness of up to 4 ML have been subjected to reductive treatments at elevated temperatures and an oxygen pressure of 1  107 mbar. Annealing at temperature between 873 and 1023 K causes a 1  2 reconstruction of the surface of the rutile TiO2(1 1 0) islands, in close similarity to the well-documented reduction behavior of bulk TiO2(1 1 0) crystals [92–94]. Higher annealing temperatures of up to 1100 K result in a completely changed film morphology, with the original rutile islands being replaced by a continuous film displaying step edges and facets in STM images (Figure 6.5). Here, regularly spaced lines with a periodicity of approximately 34–35 Å can be seen, which correspond to the intersections of {1 3 2} and {1 2 1} families of crystallographic shear planes with the (1 1 0) surface and the facets, as previously reported for single-crystal titania surfaces [95]. As the crystallographic shear planes are bulk defects rather than surface defects, the authors conclude that ultrathin TiO2 films on Ni(1 1 0) behave at least structurally in a bulk-like manner. The question to what extent this also holds for other physical and chemical properties of TiO2 nanolayers remains to be answered by future experiments. 6.3.3 Vanadium Oxide Nanolayers

In the bulk form, vanadium oxides display different oxidation states and VO coordination spheres and exhibit a broad variety of electronic, magnetic, and structural properties [96, 97], which make these materials attractive for many industrial applications. Prominent examples range from the area of catalysis, where V-oxides are used as components of important industrial catalysts for oxidation reactions [98] and environment pollution control [99], to optoelectronics, for the construction of light-induced electrical switching devices [100] and smart thermochromic windows. In view of the importance of vanadium oxides in different technological applications, the fabrication of this material in nanostructured form is a particularly attractive goal.

j159

j 6 STM Imaging of Oxide Nanolayer Model Systems

160

Figure 6.5 STM images of the crystallographically sheared film: (a) 1700 Å  2100 Å, þ 0.30 V, 0.90 nA; (b) 1500 Å  1800 Å, þ 0.15 V, 0.80 nA. The arrows indicate both the principal azimuths of the TiO2(1 1 0) surface and the directions of the CS plane intersections at

the (1 1 0) surface. (c) Line profile along the line indicated in (a). (d) Line profile along the line indicated in (b). The image in (b) has a shadow effect applied, and the line profile in (d) is taken from an image without the shadow applied. (Reproduced with permission from Ref. [86].)

We refrain here from giving an extensive overview of studies on the surface structure of vanadium oxide nanolayers, as this has already been done for up to year 2003 in our recent review [97]. Instead, we would like to focus on prototypical examples, selected from the V-oxide–Rh(1 1 1) phase diagram, which demonstrate the power of STM measurements, when combined with state-of-the-art DFT calculations, to resolve complex oxide nanostructures. Other examples will highlight the usefulness of combining STM and STS data on a local scale, as well as data from STM measurements, and sample area-averaging spectroscopic techniques, such as XPS and NEXAFS, to derive as complete a picture as possible of the investigated system. The surface phase diagram of vanadium oxides on Rh(1 1 1) has been investigated in a series of papers of our group [4, 18, 19, 90, 101–103]. It is characterized by pronounced polymorphism and many different oxide structures have been detected as a function of coverage and growth temperature. The vanadium oxide structures for coverages up to the completion of the first monolayer formed on Rh(1 1 1) under the different preparation conditions may be subdivided into highly oxidized phases

6.3 Case Studies: Selected Oxide–Metal Systems

containing vanadium atoms in formal oxidation states of around þ 5, into more reduced oxides with V2 þ or V3 þ species, and into mixed valency vanadium oxides [18, 90]. The latter are a peculiarity of the vanadium oxide – Rh(1 1 1) phase diagram – and occur only in a narrow range of the chemical potential of oxygen. The various vanadium oxide structures have been investigated with a combination of experimental techniques, including STM, LEED, high-resolution photoelectron spectroscopy (HR-XPS, UPS) with the use of synchrotron radiation, and HREELS (vibrational spectroscopy of phonon modes). The results have been combined with ab initio DFT calculations to derive the structural models for the vanadium oxide surface phases. We emphasize that the synergy of the different experimental techniques in combination with a high-level theory was absolutely necessary to achieve the success obtained in the analysis of these very complex oxide structures. In the following section, we illustrate the complexity of the structures observed with just a few examples. Under highly oxidative conditions, that is, p(O2) 2  107 mbar, Tsubstrate ¼ 400  C, a V-oxide layer with a (H7  H7)R19.1 structure, here called simply H7, grows at low submonolayer coverages (Q < 0.3 MLE) as two-dimensional islands, as seen in the STM image in Figure 6.6a [4, 18]. The high-resolution STM image in Figure 6.6b reveals the structural details of the H7 phase at the atomic scale, with a hexagonal lattice that consists of three bright protrusions per unit cell. The DFT calculations [18] have established a model of the (H7  H7)R19.1 structure (Figure 6.6c and d): it corresponds to a V3O9 oxide phase that contains identical pyramidal O4V¼O building blocks (marked squares in Figure 6.6c). Each pyramid (inset of Figure 6.6d) has the V atom (in green) in the center, four bridging O atoms (in red) in the basal plane, and a vanadyl-type O atom at the apex. The presence of the latter O species in the H7 overlayer has been inferred from HREELS spectra [18], which contain a characteristic loss at approximately 130 meV because of the stretching vibrations of vanadyl (V¼O) groups. The pyramids are linked together via the four bridging O atoms at the interface, which results in three VO3 units per H7 unit cell, or in an overall V3O9 stoichiometry. Owing to the fact that the basal O atoms are shared with the Rh substrate, the formal V3O9 stoichiometry is not incompatible with the maximum oxidation state of þ 5 of the V atoms in the bulk V2O5 structure. The structure model of the H7-V3O9 phase is confirmed by the good agreement between the experimental and simulated STM image (inset of Figure 6.6b) and the calculated phonon spectrum, which reproduces the vanadyl stretching vibrations observed in the experiment [18]. Within the “inverse model catalyst” approach, the H7-V3O9–Rh(1 1 1) nanostructures have been used to visualize surface processes in the STM with atomic-level precision [104]. The promoting effect of the V-oxide boundary regions on the oxidation of CO on Rh(1 1 1) has been established by STM and XPS by comparing the reaction on two differently prepared H7-V3O9–Rh(1 1 1) inverse catalyst surfaces, which consist of large and small two-dimensional oxide islands and bare Rh areas in between [105]. A reduction of the V-oxide islands at their perimeter by CO has been observed, which has been suggested to be the reason for the promotion of the CO oxidation near the metal-oxide phase boundary.

j161

j 6 STM Imaging of Oxide Nanolayer Model Systems

162

Figure 6.6 The (H7  H7)R19.1 oxide phase: (a) large-scale STM image (1000 Å  1000 Å, þ 2.0 V, 0.1 nA) of 2D oxide islands on the Rh(1 1 1) surface; (b) high-resolution STM image (5 Å  5 Å, þ 0.75 V, 0.2 nA). The (H7  H7)R19.1 unit cell and the Rh(1 1 1) substrate direction are indicated. The inset

shows a DFT-simulated STM image; (c) DFT model of the H7-V3O9 phase. Unit cell and structural units are indicated (V green, O red, Rh gray); (d) Side view of the V3O9 model. The inset shows a detailed view of the pyramidal O4V¼O unit. (Reproduced with permission from Ref. [4].)

Exposing the H7-phase to a reducing environment, for example, by annealing in UHV or in a hydrogen atmosphere, results in the formation of various reduced 2D V-oxide phases [18, 101]. In particular, annealing at 650–750 K in vacuum produces an oxide structure with a rectangular (5  3H3) unit cell, presented in Figure 6.7. The large-scale STM image of Figure 6.7a shows well-ordered oxide islands with rectangular shapes, elongated along the principal h1 1 0i azimuthal directions of the Rh(1 1 1) substrate. The inset of Figure 6.7a displays the (5  3H3)-rect structure at a higher magnification, also revealing details on the island boundary; the latter will be discussed in the next paragraph. The synergy of various experimental and theoretical methods has allowed us to derive the structure model for the (5  3H3)-rect phase [18], represented in Figure 6.7b. Accordingly, the structure consists of two types of building units, the O4V¼O tetragonal pyramid and a planar V6O12 hexagon (indicated on the figure), which are linked together to give the observed rectangular

6.3 Case Studies: Selected Oxide–Metal Systems

Figure 6.7 (a) STM image of (5  H3)-rect vanadium oxide islands on Rh(1 1 1) (1000 Å  1000 Å, þ 1.5 V, 0.1 nA). Inset: enlarged section of an (5  Ö3)-rect island (70 Å  70 Å, þ 0.5 V, 0.1 nA); (b) DFT-derived model of the (5  H3)-rect structure, unit cell and structural units are indicated (V green, O red, Rh gray). Inset: simulated STM image. (Reproduced with permission from Refs [18, 101].)

structure. The stoichiometry per unit cell is V13O21, and thus the (5  3H3)-rect structure corresponds to a novel mixed-valence vanadium oxide phase, which occurs only in ultrathin layer form. The simulated STM image (inset of Figure 6.7b) is in excellent agreement with the experimental one and thus supports the structure model. Further annealing the (5  3H3)-rect structure at 750–800 K leads to the transformation into a (9  9) phase [4, 18], which grows in the form of compact branched 2D islands with rounded boundary shapes, as shown in the large-scale STM image in Figure 6.8a. The high-resolution STM image (inset of Figure 6.8a) reveals a complex contrast pattern of this phase, with the (9  9) unit cell indicated. In the center of the unit cell, seven hexagonally arranged depressions are visible, showing a local (2  2) periodicity. The structure model for the (9  9) phase, derived from the DFT calculations is presented in Figure 6.8b and involves only one type of building unit, the planar V6O12 hexagons (encircled). The latter are connected in a complex way, with eightfold and fivefold rings in between, to give the observed (9  9) periodicity. Each unit cell contains 36 vanadium atoms and 54 oxygen atoms, yielding basically a V2O3 stoichiometry; the structure thus contains only V atoms in a formal þ 3 oxidation state. The DFT model is strongly supported by the simulated STM image, shown in the inset of Figure 6.8b; note that every single detail is reproduced in the simulation. Interestingly, the presence of locally ordered (2  2) areas within the (9  9) unit cell and the overall V2O3 stoichiometry bear a strong similarity to the (2  2) surface-V2O3 phase observed on the Pd(1 1 1) surface [88]. We recently suggested that the absence of a long-range ordered (2  2) structure on Rh(1 1 1) is due to interfacial strain [23]; while the (2  2)-V2O3 superstructure fits perfectly on the Pd(1 1 1) lattice, there is a 2.3% lattice mismatch between the relaxed (2  2)-V2O3

j163

j 6 STM Imaging of Oxide Nanolayer Model Systems

164

Figure 6.8 The (9  9) V-oxide phase: (a) large-scale STM image (2000 Å  2000 Å, þ 2.0 V, 0.05 nA). The inset shows a highresolution STM image (60 Å  60 Å, þ 2.0 V, 0.1 nA); (b) DFT model of the (9  9) V36O54 phase. The (9  9) and (2  2) unit cells are indicated with solid and dashed lines, respectively (V green, O red, Rh gray). The inset shows a DFT-simulated image. (c) STM images

of the Rh(15 15 13) surface covered with a V2O3-type overlayer: (center) large-scale image (500 Å  500 Å, þ 2.0 V, 0.1 nA), (left) magnified view of the (9  9) structure on the highlighted wide terrace area (45 Å  45 Å, þ 2.0 V, 0.1 nA), and (right) magnified view of the (2  2) V-oxide superstructure on the highlighted narrow terraces (75 Å  75 Å, þ 1.0 V, 0.1 nA). (Reproduced with permission from Refs [4, 14].)

overlayer and the Rh(1 1 1) surface. It has been proposed in this study that the regions of local (2  2) symmetry are connected in the (9  9) structure via the larger eightfold rings in order to release the interfacial strain. Support for this conjecture has been obtained very recently by growth experiments of V-oxide on a Rh(15 15 13) surface, which is vicinal to the Rh(1 1 1) [14]; the results are summarized in Figure 6.8c. Here, the (9  9) structure has been found on wide Rh(1 1 1) terraces, whereas a (2  2) structure was observed on small terraces, where the step edges may provide the strain relief. As a particular highlight in the V-oxide–Rh(1 1 1) phase diagram, we mention the spontaneous formation of quasi-zero-dimensional V-oxide clusters, which are illustrated in Figure 6.9a and b [4, 19]. These star-like clusters form at very low V coverages

6.3 Case Studies: Selected Oxide–Metal Systems

Figure 6.9 The V6O12 cluster molecules on Rh(1 1 1): (a) STM image (63 Å  63 Å, þ 0.5 V, 0.1 nA). The inset shows a DFT-simulated image; (b) Relaxed DFT model geometry of the V6O12 clusters on Rh(1 1 1) in top and side view. (Reproduced with permission from Ref. [4].)

(<0.1 ML) at intermediate values of the chemical potential of oxygen in a narrow range of the surface phase diagram. With the help of DFT calculations, the stars have been identified as planar V6O12 cluster molecules, which are unstable in the gas phase but are stabilized by the interactions at the interface with the Rh substrate [19]. They thus constitute a novel form of oxide cluster material. The V6O12 clusters become mobile at elevated temperature, which has allowed us to study their diffusion behavior by variable-temperature STM measurements. The clusters hop and diffuse across the surface as entire molecular units, which provides an easy means for oxide mass transport across a metal surface [19]. The V6O12 clusters self-assemble into twodimensional oxide phases of mixed valent character, the (5  5) V11O23 structure and the (5  3H3)-rect V13O21 structure (see the inset of Figure 6.7a), depending on the gradient of the chemical potential of oxygen [101]. This self-organization of spontaneously formed V6O12 cluster building blocks at a metal surface into different two-dimensional V-oxide nanostructures, directed by the chemical potential of oxygen, is a process that might be of interest for the controlled design of ultrathin oxide nanolayers [101]. Recently, we have also employed scanning tunneling spectroscopy measurements to probe the local electronic structure of the various vanadium oxide surface phases on Rh(1 1 1) [23]. Figure 6.10 summarizes differential conductance dI/dU spectra taken at 77 K in constant height (left, blue curves) and constant current (right, black curves) modes from the (H7  H7)R19.1 -V3O9 (a), (5  3H3)-rect-V13O21 (b), (9  9)-V36O54 phases (c), and from the “star” V6O12 clusters (d), discussed above. The constant current spectra display at higher voltages (>5 V) field emission resonances that were found to be of less diagnostic value for structural details. Constant height STS spectra from the (H7  H7)R19.1 and (5  3H3)-rect (in the vanadyl position) V-oxide overlayers show peaks at around þ 0.9 and þ 2.7 V in the

j165

j 6 STM Imaging of Oxide Nanolayer Model Systems

166

Figure 6.10 STS differential conductance spectra taken in constant height (left, blue curves) and constant current (right, black curves) modes from the (a) (H7  H7)R19.1 , (b) (5 H3)rect, (c) (9  9) structures, and (d) from the “star” cluster. The DFT-calculated DOS for the oxygen (red) and vanadium (green) atoms in the “star” cluster are shown in (d) for comparison. (Reproduced with permission from Ref. [23].)

empty states above the Fermi energy (U ¼ 0 V), which are characteristic of the VO5 pyramidal building units. Constant height spectra recorded above the eightfold hollow rings, present in the (5  3H3)-rect and (9  9) phases (see Figures 6.7 and 6.8), also display a similar structure, containing a characteristic peak at þ 0.3 V that has been attributed to vanadium atoms coordinated to bridging oxygen atoms. STS spectra of the V6O12 clusters (Figure 6.10d) shows peaks at þ 0.6, þ 1.6, and þ 2.7 V that approximately coincide with peaks of DFT calculated atom-projected density of states for the V (green) and O (red) atoms. The latter reveal a close correspondence, as a result of strong hybridization between V and O atoms in the molecule-type oxide clusters. We emphasize that STS is the only experimental technique that allows one to measure the electronic states of these surface supported nanoobjects. The growth of vanadium oxide overlayers on Rh(1 1 1) converges after a number of intermediate stages to the formation of a three-dimensional bulk-like epitaxial V2O3 film [90], which is oriented with the (0 0 0 1) plane of its corundum structure parallel to the Rh(1 1 1) substrate surface. The V2O3 phase is the thermodynamically stable

6.3 Case Studies: Selected Oxide–Metal Systems

phase of V-oxide under the given experimental conditions of ultrahigh vacuum experiments. A pertinent question of interest in relation to the physical and chemical properties of the V2O3 (0 0 0 1) surface is its surface termination. We have addressed this question and have found, by a combined experiment–theory effort [102, 103], that the most stable surface termination is a vanadyl V¼O terminated (1  1) surface. This is remarkable, since V¼O groups are not VO coordination units contained in the bulk V2O3 corundum structure; they do occur, however, in the layered structure of the V2O5 bulk phase. The occurrence of vanadyl units as the most stable surface termination of V2O3 is clearly unexpected and of interest to the large field of V-oxide catalysis. Apart from the vanadyl-terminated (1  1) surface, two other surface terminations could be produced, a (H3  H3)R30 structure, containing V¼O vacancies, and a V-terminated (1/H3  1/H3) surface, which requires highly reducing preparation conditions [102, 103]. The lack of bulk-type V-oxide phases on Rh(1 1 1) with an oxidation state higher than þ 3 can be ascribed to kinetic effects, that is, the limited speed of oxidation under the employed oxidation conditions (oxygen pressures of max 106 mbar). This limitation has recently been circumvented by Kuhlenbeck and coworkers [31, 32], who reported that highly oxidized V-oxide layers with V6O13 and V2O5 stoichiometries can be prepared on a Au(1 1 1) substrate, using a high-pressure cell, where oxygen pressures of up to 50 mbar can be applied for in situ oxidation of V atoms. At submonolayer to monolayer coverages, two well-ordered interface-stabilized V-oxide structures have been observed in STM and LEED, with a V oxidation state estimated to be close to þ 5, according to NEXAFS spectra taken in the V L-edge and O K-edge region [31]. By comparing the polarization dependence and the shape of NEXAFS spectra from these monolayer films and from bulk V2O5(0 0 1) single crystals, the authors concluded that the monolayer V2O5x structures should contain V¼O groups and suggested that these are part of VO5 building units, in close similarity to the highly oxidized V3O9 structures on Rh(1 1 1) [18]. After the completion of the first layer, a two-layer V6O13(0 0 1)-like wetting film was observed in STM [31], which is the precursor for the bulk-type V2O5(0 0 1) growth. The V2O5(0 0 1) phase appears at a V coverage of approximately 1.5 ML in the form of elongated 3D islands, rotationally aligned with the V6O13(0 0 1) wetting layer [31], which eventually merge to form a continuous film [32]. Kuhlenbeck and coworkers [32] have established that atomically flat and nearly defect-free V2O5(0 0 1) overlayers can be prepared by cycles of successive V deposition and postoxidation, as illustrated in the STM images in Figure 6.11a and b. The corresponding LEED image (Figure 6.11c) shows a ring pattern, indicating that the V2O5(0 0 1) domains have a random azimuthal orientation, suggesting a weak film–substrate interaction. The V2O5 stoichiometry has been confirmed by XPS and angle-dependent NEXAFS. The application of these techniques was found, however, to lead to the formation of X-ray-induced O vacancies in the V2O5 film, as revealed by STM. Accordingly, O vacancies were not randomly distributed on the surface, but as pairs or rows, as a result of a concerted reduction process. The V2O5 film was found to be thermally stable upon annealing in UHV up to 500  C, starting to sublime above this temperature, as revealed by temperature-programmed desorption (TPD) spectra.

j167

j 6 STM Imaging of Oxide Nanolayer Model Systems

168

Figure 6.11 STM images (a) 3000 Å  3000 Å, 3.5 V, 0.2 nA and (b) 1000 Å  758 Å, 3 V, 0.2 nA (differentiated), and LEED pattern (c) obtained for a film formed by three successive expositions of 2.6 MLE V and oxidation cycles. (Reproduced with permission from Ref. [32].)

6.3 Case Studies: Selected Oxide–Metal Systems

6.3.4 Iron Oxides on Pt(1 1 1)

Iron oxides, which are active materials in the catalysis of dehydrogenation reactions [106, 107], can be grown with good epitaxial order on Pt(1 1 1) surfaces. This was demonstrated by the work of the Berkeley group, Weiss et al. [108] and Barbieri et al. [109], who performed a careful LEED analysis of iron oxide films grown on Pt(1 1 1) up to a thickness of 10 ML. The 10 ML oxide was identified by a dynamical LEED analysis to be magnetite Fe3O4, terminated by an unreconstructed polar (1 1 1) surface that exposes 1/4 monolayer of Fe atoms over a distorted hexagonal closepacked oxygen layer. It was also noted in this work that at 1 ML coverage the Fe oxide is of a different kind and forms a hexagonal coincidence structure with respect to the Pt substrate. Virtually all subsequent surface science-related studies on Fe oxide films have been performed using the Pt(1 1 1) surface as a metallic substrate. The established preparation procedure for well-ordered Fe oxide films on Pt(1 1 1) involves PVD of Fe in 1–2 ML quantities onto clean Pt(1 1 1), followed by annealing in oxygen at elevated temperature; this process can be repeated until oxide layers of the desired thickness have been formed. The preparation of Fe oxides on Pt(1 1 1) and the morphology of the resulting films as a function of the preparation parameters as well as the properties of Fe oxides in relation to catalysis have been comprehensively reviewed by Weiss and Ranke [106]. Iron oxides grow with a Stranski–Krastanov mode on Pt(1 1 1) surfaces. The first phase that forms, up to coverages of approximately 2.5 ML, is a FeO(1 1 1)-type structure that grows in a layer-by-layer fashion. Subsequent layers of Fe oxide rapidly converge to Fe3O4, which nucleates in the form of 3D islands. Thick films of Fe3O4 are formed by coalescence of the 3D islands [106]. The Fe3O4(1 1 1) surface is made up by Fe atoms in a (2  2) arrangement with respect to the hexagonal oxygen layer of the subsurface plane [108, 109]. At higher oxygen pressures, Fe3O4 is transformed into Fe2O3(0 0 0 1) with a (H3  H3)R30 structural arrangement. This sequence of phases is roughly in accordance with the one found in the bulk-phase diagram of Fe oxides, but differences occur in the detailed temperatures and pressures, pointing toward the importance of kinetic effects in the stabilization of the thin film phases. All Fe oxide films on Pt have strongly relaxed, unreconstructed bulk-terminated surfaces, but while the Fe3O4 and Fe2O3 oxide layers are similar to their respective bulk compounds, the ultrathin FeO layers are true 2D oxide phases that are different from the FeO bulk and stabilized by the metal–oxide interface. Galloway et al. [110] were first to apply STM to investigate the growth of Fe oxides on Pt(1 1 1). For coverages 1ML, atomically resolved STM images have been obtained and interpreted in terms of a FeO(1 1 1) bilayer with an approximate (9  9) periodicity. The overlayer can be described with a coincidence structure, where approximately eight lattice spacings of the FeO fit approximately nine of the Pt(1 1 1) surface; this gives rise to a Moir
e pattern in STM as shown in Figure 6.12. The structure contains an atomic spacing of approximately 3.1 Å, a longer range Moir
e periodicity of approximately 26 Å and is slightly rotated against the Pt substrate.

j169

j 6 STM Imaging of Oxide Nanolayer Model Systems

170

Figure 6.12 Current mode, (60 Å  60 Å) STM image of a monolayer of FeO on Pt(1 1 1). The Moir
e structure is due to the mismatch between the 3.1 Å periodicity of the FeO small maxima in the image and the Pt lattice dimensions (2.77 Å). The small circular images at the

bottom are calculated images using the ESQC method for the various high-symmetry epitaxial configurations associated with the circled regions in the image. (Reproduced with permission from Ref. [110].)

At higher coverages, (2  2) and (H3  H3)R30 structures have been observed, which have been ascribed as before [108, 109] to Fe-terminated Fe3O4(1 1 1) and a-Fe2O3(0 0 0 1) phases, respectively. As mentioned several times in this review, the interpretation of STM images of oxide surfaces is a nontrivial problem because the relationship between images and crystallography of the surface is not straightforward. Sautet et al. [111] have successfully attempted to determine the origin of the STM image contrast and its relationship with the atomic positions of the FeO monolayer on Pt using electron scattering quantum chemistry (ESQC) theory. In their analysis, the Moir
e pattern has been decomposed into simple high-symmetry FeO/Pt configurations containing Fe in on-top, bridge, and hollow sites. In the real image, there is a continuous transition between them, but major aspects of the pattern can be understood by considering the high-symmetry configurations shown in Figure 6.12, bottom part. As an interesting result of the analysis, it turned out that the highest topographic points in the images occurred in the bridge and hcp hollow regions, depending on the tip conditions, but never in the on-top positions, which support the largest corrugation,

6.3 Case Studies: Selected Oxide–Metal Systems

since there the O atoms are furthest out. This indicates that the electronic contributions to the contrast dominate over the geometric contributions. The ESQC method also allowed the analysis of the contributions to the tunneling current of the various electronic states; both O and Fe orbitals contribute to the tunnel current, thus the contrast in STM images of FeO layers is due to the interplay of various electronic effects. Moreover, it was found that the terminating atom of the tip is important – Pt and O atoms have been used in the simulations – and that simpler treatments of the STM contrast neglecting tip–surface interactions are insufficient to correctly model the STM images. The structural parameters of the FeO(1 1 1) monolayer on Pt(1 1 1) have been examined by X-ray photoelectron diffraction by Kim et al. [112]; the previously proposed structure model of Galloway et al. [110] could be confirmed and specified in terms of its oxygen surface termination. The FeO interlayer distance was found to be highly compressed by about 50% compared to bulk FeO, 0.68 Å versus 1.25 Å. This vertical lattice compression is balanced by a lateral lattice expansion as revealed from the STM images [110, 111], namely, 3.09 Å for the FeO monolayer versus 3.04 Å in the bulk; the former is necessary to maintain the FeO bond distances at a reasonable length. As discussed below, this structural rearrangement is driven by the tendency to minimize the electrostatic surface energy. Kim et al. [112] found only one of the two possible rotational domains of the oxide monolayer on their surface, which they ascribed to an OFePt interlayer interaction. This manifestation of interlayer effects in the growth mode is an interesting aspect in epitaxial growth of oxide nanolayers. However, Ritter et al. [113] seem to have observed both rotational domains on a formally identical FeO monolayer on Pt(1 1 1), and the presence of only one domain in the experiments of Kim et al. may be the effect of some particular kinetic conditions during surface preparation. Ritter et al. [113] performed a very detailed study of the initial stages of iron oxide growth on Pt(1 1 1) and found four different coincidence structures of FeO(1 1 1)-type phases up to approximately 2.5 ML film thickness. With increasing coverage several structural changes occur in the FeO layers resulting in coincidence structures with slightly different lateral lattice constants and rotation misfit angles; all of them give rise to Moir
e patterns with periodicities between 22 and 38 Å. Large-scale and atomic resolution STM images of the four coincidence structures found by Ritter et al. [91, 113] are shown in Figure 6.13 gives a rigid sphere model of the structures illustrating the different coincidence sites. The FeO(1 1 1)-type overlayer on Pt(1 1 1) thus exhibits a remarkable variety in terms of lattice parameter and interfacial registry. The physical origin of this structural flexibility of the FeO overlayer is still unclear, the more so since no clear trend is observable in the sequence of lattice parameters of the coincidence structures. The FeO(1 1 1) phase forming up to coverages of 2–3 ML is clearly stabilized by the interactions with the Pt substrate since FeO is thermodynamically metastable with respect to the higher iron oxides [106, 114]. FeO has the rock salt structure and the (1 1 1) plane yields a polar surface with a high surface energy [115], which requires stabilization by internal reconstruction or external compensation. The structural relaxation observed in the form of the reduced FeO

j171

j 6 STM Imaging of Oxide Nanolayer Model Systems

172

Figure 6.13 Left panel: Large-scale (left) and atomic resolution (right) STM images of epitaxial FeO(1 1 1) films on Pt(1 1 1). Four different coincidence structures 1–4 are formed sequentially as the coverage increases. They exhibit different contrasts in the large-scale images as indicated by the numbers. Right panel: Rigid model of the four FeO(1 1 1)

structures formed on Pt(1 1 1). FeO(1 1 1) bilayers with different lattice constants aFeO and rotated against the platinum surface lattice by different angles a lead to coincidence sites 1–4. Here, coincidence structure 2 with its superstructure cell is shown. (Reproduced with permission from Ref. [91].)

interlayer spacing and the expanded lateral lattice constant is the result of reducing the distance between the positively charged Fe and the negatively charged O layers, thereby reducing the surface dipole. Moreover, external compensation of the FeO(1 1 1) surface dipoles by the image dipoles in the underlying Pt metal substrate add a further contribution to the stabilization of the FeO overlayer. However, instead of converging to the FeO bulk lattice with increasing film thickness, the lateral expansion increases from 3.09 Å at 1 ML to 3.15 Å at >2 ML oxide coverage [91, 113]. Ranke et al. [91] have attempted a qualitative interpretation of the different coincidence structures based on an ionic model of FeO and the decrease of the electrostatic

6.3 Case Studies: Selected Oxide–Metal Systems

energy of the polar FeO stack by reducing interlayer distances and increasing lateral lattice constants. While catching the essence of the overall lattice relaxations in the FeO(1 1 1) nanolayers, a purely ionic description of FeO is questionable, and the model is too simplistic to explain the sequence of coincidence structures observed as a function of coverage. High-level ab initio calculations of the various interfacial configurations will be necessary to shed light on the origin of the observed coincidence structures of FeO on Pt(1 1 1); this will be no easy task in view of the large unit cells involved. 6.3.5 Nickel Oxide Nanolayers

Thin nickel oxide (NiO) films are used in many oxidation catalysts and industrial chemical processes including steam reforming of methane to syngas [116], the oxidation of methane [117], hydrocarbons [118], methanol [119], CO [120], and other organic compounds [121, 122]. NiO is also employed in alkaline electrochemical systems, such as alkaline fuel cells [123, 124]. Owing to its defect structure, nonstoichiometric nickel oxide is a good p-type semiconductor and is therefore used as resistive sensors for reducing gases, such as H2 [125], CO [126], NH3 [127], and also NO2 [128], formaldehyde [129], and methanol [130]. Epitaxial growth of nickel oxide nanolayers and the characterization of their structure at the atomic level is thus of considerable fundamental and applied interest. Historically, ultrathin Ni-oxide layers have been first grown by oxidation of singlecrystal Ni(1 0 0) and Ni(1 1 1) surfaces (see Ref. [131] and references therein). The lattice parameter of NiO (4.176 Å) is by nearly 20% larger than that of metallic Ni (3.524 Å), which is not a favorable prerequisite for epitaxial NiO growth. B€aumer et al. [132] have reported a (1 0 0)-oriented NiO film growth by exposing a Ni(1 0 0) surface to 104 l O2 at 570 K. According to their STM and SPA-LEED data, the film consists of NiO(1 0 0) crystallites, which are tilted by approximately 8% with respect to Ni(1 0 0) surface normal. The authors concluded that by means of this tilt the surface strain, caused by the large misfit between Ni and NiO lattices, may be partially relieved [132]. On the other hand, a much more favorable epitaxial relationship for a NiO(1 0 0) layer growth has been established on the Ni(1 1 1) substrate, that is, NiO(1 0 0)||Ni(1 1 1) and NiO[0 1 0]||Ni[110] with a mismatch of only 1.4%, where locally ordered (2  2), (2  3), (3  2), and (3  3) superstructure cells of NiO(1 0 0) have been detected in LEED and STM [133, 134]. In particular, Hildebrandt et al. [134] used a variable-temperature STM to investigate in situ and in real time the oxidation of Ni(1 1 1) at elevated temperature. They found that the oxidation starts at the Ni step edges, which is followed by the transformation of the Ni(1 1 1) terraces into a threedomain NiO(0 0 1)-like layer, containing a mixture of (2  2), (2  3), and (3  3) NiO (0 0 1) unit cells, as shown in Figure 6.14. At higher oxygen exposures (>25 l), the formation of triangular bulk NiO(1 1 1) islands has been observed. More favorable lattice matching conditions have been achieved by growing epitaxial NiO layers on dissimilar metal surfaces, such as Au(1 1 1) [135, 136], Ag(1 0 0) [137–142], Cu(1 0 0) [143], Cu(1 1 1) [144, 145], and more recently

j173

j 6 STM Imaging of Oxide Nanolayer Model Systems

174

Figure 6.14 Room temperature STM image (400 Å  270 Å, þ 1.0 V, 0.5 nA) after oxidation at 400 K. An enlarged image detail shows the structure of the (2  2), (2  3), and (3  3) cells in the NiO(0 0 1) regions. (Reproduced with permission from Ref. [134].)

Pd(1 0 0) [146–150] and Pt(1 1 1) [20]. The Neddermeyer group was first to demonstrate that owing to the small lattice mismatch (2%), well-ordered ultrathin NiO films can be prepared on a Au(1 1 1) surface by reactive evaporation of Ni metal in an oxygen atmosphere [135, 136]. Their STM images revealed tripoid-like ridge structures with a (2  2) periodicity, corresponding to a p(2  2) reconstructed NiO(1 1 1) surface, which has been interpreted in terms of Wolf’s octopolar model [151]. The latter consists of {1 0 0} nanofacets, obtained by removing three-fourth of the surface and one-fourth of the subsurface ions, which results in a compensation of the polarity of the NiO(1 1 1) surface. More recent glancing incidence X-ray diffraction (GIXD) study has shown that the (2  2)-reconstructed NiO(1 1 1) films can be grown on the Au(1 1 1) surface up to 8 ML in a layer-by-layer fashion, and such films exhibit both possible terminations (O and Ni), separated by single-atom steps [152]. The Ag(1 0 0) surface was considered as an ideal candidate for the epitaxial growth of NiO films oriented along the (1 0 0) direction due to the low overlayer/substrate lattice mismatch ( þ 2.2%) and the low reactivity of silver toward oxygen. Interestingly, despite this good structural matching no pseudomorphic NiO(1 0 0) layer forms for submonolayer coverages. Instead, a 2D oxidic phase with a (2  1) periodicity grows in the form of two orthogonal domains on the Ag(1 0 0) surface up to a coverage close to 1 ML, as shown by STM images obtained by Bertrams and Neddermeyer [138]. On the basis of quantitative LEED analysis, Caffio et al. [142] have

6.3 Case Studies: Selected Oxide–Metal Systems

proposed a structure model of the (2  1) phase similar to the (1 1 1) surface of NiO, which consists of a buckled NiO bilayer (O-terminated) with a distorted hexagonal structure. It has been argued that this polar NiO(1 1 1)-like nanolayer is stabilized by the electrostatic interaction with the metal substrate and is kinetically preferred over the NiO(1 0 0) structure in the early stages of oxide formation [142]. As the oxygen dose and/or the film thickness increases the (2  1) phase transforms into a pseudomorphic (1  1)-NiO(1 0 0) phase, which grows in a layer-by-layer mode up to 5 ML [139]. For higher coverages, as demonstrated by SPA-LEED analysis [139], the strain is relieved by the introduction of misfit dislocations with {1 1 0} glide planes, which leads to the formation of mosaics on the oxide film surface. Recently, we investigated in a series of joint studies [146–150] with the group of Granozzi in Padua the structure of NiO nanolayers supported on a Pd(1 0 0) surface. The Pd(1 0 0) surface differs from Ag(1 0 0) by a significantly larger lattice mismatch ( þ 7.3%) to NiO(1 0 0) and by the predominantly 4d character of the valence band near the Fermi edge, in contrast to the 5sp character of the silver top valence band. This difference in the electronic structure of Pd and Ag is considered to play an important role in the hybridization of states at the metal–oxide interface and in the substrate reactivity toward oxygen. The latter was found to strongly affect the morphology and degree of structural order of the Ni-oxide overlayer in the submonolayer coverage range via the oxide preparation procedure: reactive deposition (RD) versus postoxidation [148]. Irrespective of the oxide preparation route, an interface-stabilized c(4  2) phase forms in the early stages of Ni-oxide growth, as detected by LEED and STM [147, 148]. The PO procedure was shown to be beneficial for obtaining a well-ordered wetting 2D layer, which has been imaged with atomic resolution in STM (Figure 6.15a and inset). A structure model of this c(4  2) monolayer has been derived by quantitative LEED I–V analysis [147] and is presented in Figure 6.15b. It is based on a NiO(1 0 0) surface with O atoms sitting in on-top Pd positions and with one-fourth of the Ni atoms missing in a rhombic c(4  2) unit cell, which corresponds to a formal Ni3O4 stoichiometry. Hybrid-exchange density functional theory calculations of Pisani and coworkers [149] have confirmed the LEED I–V model, and it has been argued that the rhombic geometry of Ni vacancies is electrostatically preferred over a squared one due to the higher average distance between Ni vacancies in the former structure. We close this section with highlighting our recent results on the fabrication of quasi-one-dimensional (1D) Ni-oxide structures on vicinal Rh(1 1 1) surfaces via decorating their regular step arrays [153, 154]. Figure 6.16 shows STM images of the 0.3 ML Ni/Rh(15 15 13) surface after exposure to 15 l O2 at 300  C. Figure 6.16a and b show straight steps decorated by bright lines of protrusions, which display a linear 1D-(1) and 1D-(2) periodicity in terms of the Rh lattice parameter parallel to the step edges (Figure 6.16c and d). The oxide-free Rh terraces are covered by (2  1) domains of chemisorbed oxygen. The structure of Ni-oxide wires at the step edges has been rationalized in the DFTmodel of Mittendorfer and coworkers [153] for a 0.6 ML Ni-oxide decorated Rh(5 5 3) surface (Figure 6.16e). Accordingly, the first row of Ni atoms at the step edges is coordinated to four O atoms each, yielding the 1D-(1) periodicity at the upper and lower step edges. Additional O adatoms are adsorbed in

j175

j 6 STM Imaging of Oxide Nanolayer Model Systems

176

Figure 6.15 (a) STM image of 1.25 ML Ni-oxide on Pd(1 0 0) (500 Å  500 Å, þ 2.0 V, 0.1 nA). The inset shows a highresolution STM image of the c(4  2) structure (35 Å  35 Å, þ 1.0 V, 1.0 nA). (b) Top and side views of the NiO(1 0 0)-based model for the c(4  2) overlayer. (Reproduced with permission from Refs [147, 148].)

a (2  1) lattice on the other Ni rows, resulting in the 1D-(2) lines of protrusions in the STM images and reproduced well in the simulated STM image in Figure 6.16f. Although the formal stoichiometry of the NiO stripes on the vicinal Rh(1 1 1) surfaces is NiO, it has no relation to the bulk NiO, since a fraction of the O atoms is shared between the Ni and the Rh surface atoms.

6.3 Case Studies: Selected Oxide–Metal Systems

Figure 6.16 (a–d) STM images of 0.3 ML Ni-oxide layer on the Rh(15 15 13) surface: (a) 1000 Å  1000 Å, þ 1 V, 0.1 nA; (b) 150 Å  150 Å, þ 0.13 V, 1.0 nA; (c) 18 Å  18 Å, þ 0.13 V, 1.0 nA; (d) 45 Å  37 Å, þ 0.14 V, 1.0 nA; (e) DFT model of a 0.6 ML Ni-oxide on Rh(5 5 3); and (f) corresponding simulated STM image. (Reproduced with permission from Ref. [154].)

6.3.6 Ceria Nanolayers on Metal Surfaces

Cerium oxides are outstanding oxide materials for catalytic purposes, and they are used in many catalytic applications, for example, for the oxidation of CO, the removal of SOx from fluid catalytic cracking flue gases, the water gas shift reaction, or in the oxidative coupling reaction of methane [155, 156]. Ceria is also widely used as an active component in the three-way catalyst for automotive exhaust pollution control,

j177

j 6 STM Imaging of Oxide Nanolayer Model Systems

178

where it is believed to play a crucial role as an “oxygen storage-release” agent to control the oxygen concentration on the catalyst surface during the different cycles of the operating engine. The latter involves oxygen transport from and to the surface, which is determined to a large extent by the formation, concentration, and mobility of oxygen vacancy defects at the surface and in the bulk. The ease of formation and annihilation of oxygen vacancy defects in cerium oxides is of course intimately connected to the propensity of Ce ions to undergo valence changes from þ 4 to þ 3 and vice versa. Therefore, it is not surprising that there has been and still is a strong interest in the atomic structure of cerium oxide surfaces and of the defects encountered thereon. There has been considerable effort to elucidate the surface structures of the low-index single-crystal surfaces of CeO2, which crystallizes in the cubic fluorite structure. According to theory, the most stable low-index surface is the (1 1 1) surface with an oxygen-terminating outer layer [157]. Note that in the (1 1 1) orientation, the CeO2 crystal may be regarded as a stack of hexagonal OCeO trilayer planes. N€ orenberg and Briggs [158] were the first to apply STM to the study of CeO2(1 1 1) surfaces, and they indeed observed images with atomically resolved features in a hexagonal arrangement. Because CeO2 is a wide-band insulator with a bandgap of approximately 6 eV, the experiments had to be performed on partially reduced CeO2x surfaces to enhance the electron conductivity for the STM experiments; the reduced CeO2x surfaces were obtained by annealing the samples in vacuum at high temperatures (1000  C). On the basis of their tunneling conditions (negative sample bias of several V, thus tunneling out of the filled states of the surface), N€ orenberg and Briggs concluded that they are imaging the oxygen atoms of the CeO2(1 1 1)-1  1 surface. The typical defect structures observed at the partially reduced CeO2x surfaces displayed triangular and linear dark contrast features, which were interpreted by N€orenberg and Briggs in terms of multiple oxygen vacancies [158]. The defect formation process on CeO2(1 1 1) single-crystal surfaces has recently been addressed in more detail using STM by Esch et al. [159], again on partially reduced samples. In combination with DFTcalculations, the local structure of surface defects has been ascribed to multiple surface and subsurface oxygen vacancies in various combinations. A better approach than using STM to study defects at CeO2 single-crystal surfaces, independent of the overall reduction state, is to use atomically resolved atomic force microscopy. Iwasawa and coworkers [156, 160, 161] have imaged the oxygen atoms on CeO2(1 1 1) surfaces with atomic resolution with noncontact AFM and have observed several kinds of defects, both point defects and multiple defects such as the line and triangular structures, which were also observed in STM before [158]. The creation of oxygen vacancies leads to the formation of reduced Ce 3 þ ions in the vicinity of the defect, which take up the two negative charges left behind per oxygen vacancy. This leads to local reconstructions of the atoms around the defect that can be visualized in the AFM images and that contribute to the stability of the defect. By means of successive AFM measurements of the same area on slightly reduced CeO2x surfaces, the dynamics of surface O atoms and the healing of defects by exposure to O2 from the gas phase has been observed directly by Namai et al. [161].

6.3 Case Studies: Selected Oxide–Metal Systems

High-quality resolved images and atomic details of surface features on CeO2(1 1 1) in various oxidation states have also been obtained recently by dynamic AFM measurements by Gritschneder et al. [162]. Thin films of CeO2x with (1 1 1) surface orientation have been grown epitaxially on a variety of metal substrates. Mullins et al. [163] have prepared highly ordered CeO2x(1 1 1)-type films, up to 10 ML thick, by reactive evaporation of Ce in oxygen atmosphere on Ru(0 0 0 1) and Ni(1 1 1) surfaces. The oxide overlayers were aligned parallel to the principal azimuths of the substrates as judged from ion scattering spectroscopy and LEED measurements, and the degree of oxidation (i.e., the x in CeO2x) could be adjusted to some extent by the oxygen pressure during evaporation. No direct imaging information on the morphology and structure of these films was available in the early work of Mullins et al. [163]. However, Lu et al. [164] recently took up and modified the recipe of Mullins for the preparation of CeO2(1 1 1) films on Ru(0 0 0 1) and have shown that higher oxidation temperatures lead to flat ceria films with a low defect density. Schierbaum and Berner [165–167] have used a different approach to generate epitaxial ceria nanolayers on Pt(1 1 1), namely, the oxidation of Pt–Ce alloy surfaces; the latter were formed by depositing Ce metal onto Pt(1 1 1) surfaces followed by a high-temperature (1000 K) annealing treatment. The surface structures of the oxide nanolayers formed were consistent with CeO2(1 1 1) as revealed by STM, but heating of the oxide phase in vacuum at 900 K lead to a different structure, which was attributed to a Ce2O3 surface phase [167]. Epitaxial cerium oxide nanolayers have been deposited by reactive evaporation on Rh(1 1 1) by Eck et al. [168]. STM and LEED indicated that the ceria grows in the form of ordered CeO2 double-layer islands, with (1 1 1) faces parallel to the surface and orientationally aligned to the main azimuthal directions of the substrate. While the nanolayer films contained significant amounts of reduced Ce 3 þ species, for the thicker films (>6 ML) stoichiometric CeO2 was detected in XPS. Vacuum annealing of the CeO2/Rh(1 1 1) submonolayer nanostructures produced morphologically welldefined hexagonal islands (see Figure 6.17a), which displayed a characteristic Moir
e pattern in STM at their surfaces (Figure 6.17b and c) [169]; the latter is the result of the interfacial lattice mismatch and the interference effects between the CeO2 and the Rh lattices, which give rise to a (5  5) coincidence lattice of a compressed cerium oxide overlayer (5  aCeO2  7  aRh). When subjected to more reducing conditions, for example, by additional vacuum annealing at >600  C, an ordered array of surface defects has been observed on the ceria nanoislands as displayed in the STM images of Figure 6.18. In contrast to the STM images obtained from bulk-terminated CeO2 surfaces [158, 159], the STM images from the CeO2 nanolayers on Rh (Figures 6.17 and 6.18) have been recorded with positive bias voltages [169], where the empty states are imaged. Ab initio DFT calculations have revealed that the density of states within 1 eV above the Fermi level in the first CeO2 trilayer of the ceria–Rh system is made up by Ce 4f states, which are hybridized to a certain extent with O 2p states; according to the calculations, the bright maxima in the corresponding STM images correspond, therefore, to the cerium sites [169]. Using STM images with atomic resolution in combination with DFT calculations, the defects seen in Figure 6.18 have been

j179

j 6 STM Imaging of Oxide Nanolayer Model Systems

180

Figure 6.17 Constant current topographic STM images of 0.5 monolayer (ML) of CeO2 on Rh(1 1 1). (a) 1850 Å  1850 Å, þ 1.99 V, 1 nA; (b) 200 Å  200 Å, þ 0.89 V, 0.83 nA; (c) 93 Å  57 Å, þ 0.78 V, 0.83 nA. A CeO2(1 1 1)1  1 unit cell (u) is indicated on the image (c). The inset to (a) shows a corresponding LEED pattern (electron energy ¼77 eV). (Reproduced with permission from Ref. [169].)

6.3 Case Studies: Selected Oxide–Metal Systems

Figure 6.18 STM images of the 0.5 ML CeO2x–Rh(1 1 1) surface after annealing to 625  C. (a) 200 Å  200 Å, þ 0.93 V, 0.86 nA; (b) 100 Å  100 Å, þ 0.80 V, 1.05 nA. The grid of thin lines illustrates the superlattice of defects. (Reproduced with permission from Ref. [169].)

associated with oxygen vacancies. It has been suggested that this self-assembly of oxygen vacancies is facilitated by the lattice mismatch between the oxide overlayer and the metal substrate and that it may be ascribed to a strain-related reduction in the vacancy formation energy [169].

j181

j 6 STM Imaging of Oxide Nanolayer Model Systems

182

An interesting point is the fact that the epitaxial growth of CeO2 overlayers has been observed on a number of noble metal surfaces [163, 165, 168], all of which display a large lattice mismatch of the order of at least 30% at the respective interfaces. It has been proposed that coincidence lattices with relatively low interfacial strain are formed in all these cases, in a manner similar to the one described above for the CeO2–Rh(1 1 1) interface, thus creating ordered low-energy interfaces that mediate the heteroepitaxial growth further [169].

6.4 Synopsis and Outlook

As in other fields of nanoscience, the application of STM techniques to the study of ultrathin oxide layers has opened up a new era of oxide materials’ research. New emergent phenomena of structure, stoichiometry, and associated physical and chemical properties have been observed and new oxide phases, hitherto unknown in the form of bulk material, have been detected in nanolayer form and have been elucidated with the help of the STM. Some of these oxide nanolayers are and will be of paramount interest to the field of advanced catalysis, as active and passive layers in catalytic model studies, on the one hand, and perhaps even as components in real nanocatalytic applications, on the other hand. We have illustrated with the help of prototypical examples the growth and the structural variety of oxide nanolayers on metal surfaces as seen from the perspective of the STM. The selection of the particular oxide systems presented here reflects in part their relevance in catalysis and is also related to our own scientific experience. We believe that the field of oxide nanolayer characterization is still in its early stages and that many novel oxide phases will be detected in the years to come. While the way to their structural characterization is essentially paved, the elucidation of other properties such as catalytic chemistry or physical properties including magnetic behavior is largely unexplored. The commercial availability of high-quality low-temperature STM instruments has opened the way to the observation of chemical reactions on a single-atom basis – keyword: “single-atom catalysis” – that is of fundamental scientific interest. The local spectroscopy capability of STS in probing the electronic structure at the single-atom level is central to this single-atom chemistry approach. Bond formations induced by the STM tip [170] or STM-induced catalysis [171] are directions of chemical nanostructure research, where increasing activities can be expected in the future. The reduction of the dimensionality of metal-supported oxide nanostructures from three- and two-dimensional to onedimensional (monoatomic oxide line structures [153]) or quasi-zero-dimensional (oxide quantum dots), thus creating low-dimensional oxide–metal nanoscale hybrid structures, is another area of scientific endeavor with promise of finding novel chemical and physical behavior. These latter systems will allow us to study, in a controlled way and at the atomic scale, the chemical reactivity of low-coordinated surface sites, which are abundantly present in practical catalyst systems but which escape direct scientific characterization due to the inherent lack of control at the

References

atomic level. In all these areas, STM techniques will play a central role, and the future of STM applications in the field of surface chemistry and catalysis, and oxide catalysis in particular, is bright and wide open.

Acknowledgments

This work has been supported by the Austrian Science Funds within the National Research Network “Nanoscience on Surfaces.” FPN acknowledges with gratitude the excellent hospitalities of Professor Wolf-Dieter Schneider, EPFL Lausanne, and Professor Charlie Campbell, University of Washington, during his sabbatical stays in Lausanne and Seattle in 2008.

References 1 Freund, H.-J., Kuhlenbeck, H., and Staemmler, V. (1996) Rep. Prog. Phys., 59, 283. 2 Street, S.C., Xu, C., and Goodman, D.W. (1997) Annu. Rev. Phys. Chem., 48, 43. 3 Freund, H.-J. (2007) Surf. Sci., 601, 1438. 4 Schoiswohl, J., Surnev, S., Netzer, F.P., and Kresse, G. (2006) J. Phys.: Condens. Matter, 18, R1. 5 Lundgren, E., Kresse, G., Klein, C., Borg, M., Andersen, J.N., De Santis, M., Gauthier, Y., Konvicka, C., Schmid, M., and Varga, P. (2002) Phys. Rev. Lett., 88, 246103. 6 Gustafson, J., Mikkelsen, A., Borg, M., Lundgren, E., K€ohler, L., Kresse, G., Schmid, M., Varga, P., Yuhara, J., Torrelles, X., Quiros, C., and Andersen, J.N. (2004) Phys. Rev. Lett., 92, 126102. 7 Lundgren, E., Mikkelsen, A., Andersen, J.N., Kresse, G., Schmid, M., and Varga, P. (2006) J. Phys.: Condens. Matter, 18, R481. 8 Over, H., Kim, Y.D., Seitsonen, A.P., Wendt, S., Lundgren, E., Schmid, M., Varga, P., Morgante, A., and Ertl, G. (2000) Science, 287, 1474. 9 Hendriksen, B.L.M. and Frenken, J.W.M. (2002) Phys. Rev. Lett., 89, 046101. 10 Giordano, L., Cinquini, F., and Pacchioni, G. (2005) Phys. Rev., B73, 045414.

11 Giordano, L. and Pacchioni, G. (2006) Phys. Chem. Chem. Phys., 8, 3335. 12 Sock, M., Surnev, S., Ramsey, M.G., and Netzer, F.P. (2001) Top. Catal., 14, 15. 13 Boffa, A., Lin, C., Bell, A.T., and Somorjai, G.A. (1994) J. Catal., 149, 149. 14 Schoiswohl, J., Surnev, S., and Netzer, F.P. (2007) J. Phys. Chem. C, 111, 10503. 15 Surnev, S., Kresse, G., Sock, M., Ramsey, M.G., and Netzer, F.P. (2001) Surf. Sci., 495, 91. 16 Binnig, G., Rohrer, H., Gerber, Ch., and Weibel, E. (1982) Phys. Rev. Lett., 49, 57. 17 Surnev, S., Schoiswohl, J., Kresse, G., Ramsey, M.G., and Netzer, F.P. (2002) Phys. Rev. Lett., 89, 246101. 18 Schoiswohl, J., Surnev, S., Sock, M., Eck, S., Ramsey, M.G., Netzer, F.P., and Kresse, G. (2005) Phys. Rev., B71, 165437. 19 Schoiswohl, J., Kresse, G., Surnev, S., Sock, M., Ramsey, M.G., and Netzer, F.P. (2004) Phys. Rev. Lett., 92, 206103. 20 Hagendorf, Ch., Shantyr, R., Neddermeyer, H., and Widdra, W. (2006) Phys. Chem. Chem. Phys., 8, 1575. 21 Rienks, E.D.L., Nilius, N., Rust, H.-P., and Freund, H.-J. (2005) Phys. Rev. B, 71, 241404(R). 22 Rienks, E.D.L., Nilius, N., Giordano, L., Goniakowski, J., Pacchioni, G., Felicissimo, M.P., Risse, T., Rust, H.-P.,

j183

j 6 STM Imaging of Oxide Nanolayer Model Systems

184

23

24

25

26 27 28 29

30 31

32

33 34 35 36 37 38

and Freund, H.-J. (2007) Phys. Rev. B, 75, 205443. Schoiswohl, J., Sock, M., Chen, Q., Thornton, G., Kresse, G., Ramsey, M.G., Surnev, S., and Netzer, F.P. (2007) Top. Catal., 46, 137. Jensen, J.A., Rider, K.B., Chen, Y., and Salmeron, M. (1999) J. Vac. Sci. Technol. B, 17, 1080. Lægsgaard, E., Österlund, L., Thostrup, P., Rasmussen, P.B., Stensgaard, I., and Besenbacher, F. (2001) Rev. Sci. Instrum., 72, 3537. Kolmakov, A. and Goodman, D.W. (2003) Rev. Sci. Instrum., 74, 2444. R€ oßler, M., Geng, P., and Winterlin, J. (2005) Rev. Sci. Instrum., 76, 023705. Hendriksen, B.L.M., Bobaru, S.C., and Frenken, J.W.M. (2005) Top. Catal., 36, 43. Rost, M.J., Crama, L., Schakel, P., van Tol, E., van Velzen-Williams, G.B.E.M., Overgauw, C.F., ter Horst, H., Dekker, H., Okhuijsen, B., Seynen, M., Vijftigschild, A., Han, P., Katan, A.J., Schoots, K., Schumm, R., van Loo, W., Oosterkamp, T.H., and Frenken, J.W.M. (2005) Rev. Sci. Instrum., 76, 053710. Chambers, S. (2000) Surf. Sci. Rep., 48, 105 and references therein. Guimond, S., Sturm, J.M., G€obke, D., Romanyshyn, Y., Naschitzki, M., Kuhlenbeck, H., and Freund, H.J. (2008) J. Phys. Chem. C, 112, 12363. Guimond, S., Sturm, J.M., G€obke, D., Romanyshyn, Y., Naschitzki, M., Kuhlenbeck, H., and Freund, H.J. (2008) J. Phys. Chem. C, 112, 11835. Matsumoto, T., Batzill, M., Hsieh, S., and Koel, B.E. (2004) Surf. Sci., 572, 127. Matsumoto, T., Batzill, M., Hsieh, S., and Koel, B.E. (2004) Surf. Sci., 572, 141. Niehus, H., Blum, R.P., and Ahlbehrendt, D. (2001) Phys. Stat. Sol. (a), 187, 151. Niehus, H., Blum, R.P., and Ahlbehrendt, D. (2003) Surf. Rev. Lett., 10, 353. Middeke, J., Blum, R.P., Hafemeister, M., and Niehus, H. (2005) Surf. Sci., 587, 219. Tersoff, J. and Hamann, D.R. (1985) Phys. Rev. Lett., 50, 805.

39 B€aumer, M., Freund, H.-J., and Henry, C.R. (1998) Surf. Sci. Rep., 31, 235. 40 Franchy, R. (2000) Surf. Sci. Rep., 38, 195. 41 Jaeger, R.M., Kuhlenbeck, H., Freund, H.-J., Wuttig, M., Hoffmann, W., Franchy, R., and Ibach, H. (1991) Surf. Sci., 259, 235. 42 Doychak, J., Smialek, J.L., and Mitchell, T.E. (1989) Metall. Trans., A20, 499. 43 Isern, H. and Castro, G.R. (1989) Surf. Sci., 211/212, 865. 44 Libuda, J., Winkelmann, F., B€aumer, M., Freund, H.-J., Bertrams, Th., Neddermeyer, H., and M€ uller, K. (1994) Surf. Sci., 318, 61. 45 Kulawik, M., Nilius, N., Rust, H.-P., and Freund, H.-J. (2003) Phys. Rev. Lett., 91, 256101. 46 Stierle, A., Renner, F., Streitel, R., Dosch, H., Drube, W., and Cowie, B.C. (2004) Science, 303, 1652. 47 Kresse, G., Schmid, M., Napetschnig, E., Shishkin, M., K€ohler, L., and Varga, P. (2005) Science, 308, 1440. 48 Schmid, M., Shishkin, M., Kresse, G., Napetschnig, E., Varga, P., Kulawik, M., Nilius, N., Rust, H.-P., and Freund, H.-J. (2006) Phys. Rev. Lett., 97, 046101. 49 Bardi, U., Atrei, A., and Rovida, G. (1990) Surf. Sci., 239, L511. 50 Bardi, U., Atrei, A., and Rovida, G. (1992) Surf. Sci., 286, 87. 51 Becker, C., Kandler, J., Raaf, H., Linke, R., Pelster, T., Tanemura, M., and Wandelt, K. (1998) J. Vac. Sci. Technol., A16, 1000. 52 Rosenhahn, A., Schneider, J., Becker, C., and Wandelt, K. (2000) J. Vac. Sci. Technol., A18, 1923. 53 Degen, S., Krupski, A., Kralj, M., Langner, A., Becker, C., Sokolowski, M., and Wandelt, K. (2005) Surf. Sci., 576, L57. 54 Gritschneder, S., Becker, C., Wandelt, K., and Reichling, M. (2007) J. Am. Chem. Soc., 129, 4925. 55 Schmid, M., Kresse, G., Buchsbaum, A., Napetschnig, E., Gritschneder, S., Reichling, M., and Varga, P. (2007) Phys. Rev. Lett., 99, 196104.

References 56 Becker, C., Rosenhahn, A., Wiltner, A., von Bergmann, K., Schneider, J., Pervan, P., Milun, M., Kralj, M., and Wandelt, K. (2002) New J. Phys., 4, 75. 57 Linsebigler, A.L., Lu, G., and Jates, J.T., Jr. (1995) Chem. Rev., 95, 735. 58 Fox, M.A. and Dulay, M.T. (1993) Chem. Rev., 93, 341. 59 Kamat, P.V. (1993) Chem. Rev., 93, 267. 60 Bosh, H. and Janssen, F. (1988) Catal. Today, 2, 369. 61 Forzatti, P. (2000) Catal. Today, 62, 51. 62 Hoffman, M.R., Martin, S.T., Choi, W., and Wahneman, D.W. (1995) Chem. Rev., 95, 69. 63 Maldoti, A., Molnari, A., and Amareni, R. (2002) Chem. Rev., 102, 3811. 64 Kalyanasendavan, K. and Gratzel, M. (1999) Optoelectronic Properties of Inorganic Compounds (eds D.M. Roundhill and J.P. Fackler), Plenum Press, New York, pp. 169–194. 65 Sheveglieri, G. (ed.) (1992) Gas Sensors, Kluwer, Dordrecht. 66 Johnson, R.W., Thieles, E.S., and French, R.H. (1997) Tappi J., 80, 233. 67 Phillips, L.G. and Barbeno, D.M. (1997) J. Dairy Sci., 80, 2726. 68 Selhofer, H. and M€ uller, R. (1999) Thin Solid Films, 351, 180. 69 Fujishima, A., Rao, T.N., and Tryk, D.A. (2000) J. Photochem. Photobiol. C: Photochem. Rev., 1, 1. 70 Diebold, U. (2003) Surf. Sci. Rep., 48, 53. 71 Boffa, A.B., Galloway, H.C., Jacobs, P.W., Ben
ıtez, J.J., Batteas, J.D., Salmeron, M., Bell, A.T., and Somorjai, G.A. (1995) Surf. Sci., 326, 80. 72 Sedona, F., Rizzi, G.A., Agnoli, S., Llabr
es i Xamena, F.X., Papageorgiou, A., Ostermann, D., Sambi, M., Finetti, P., Schierbaum, K., and Granozzi, G. (2005) J. Phys. Chem. B, 109, 24411. 73 Sedona, F., Agnoli, S., and Granozzi, G. (2006) J. Phys. Chem. B, 110, 15359. 74 Barbaro, G., Sedona, F., Fortunelli, A., and Granozzi, G. (2007) J. Phys. Chem. C, 111, 6095.

75 Sedona, F., Granozzi, G., Barbaro, G., and Fortunelli, A. (2008) Phys. Rev. B, 77, 115417. 76 Orzali, T., Casarin, M., Granozzi, G., Sambi, M., and Cittadini, A. (2006) Phys. Rev. Lett., 97, 156101. 77 Herman, G.S., Gallagher, M.C., Joyce, S.A., and Peden, C.H.F. (1996) J. Vac. Sci. Technol. B, 14, 1126. 78 Oh, W.S., Xu, C., Kim, D.Y., and Goodman, D.W. (1997) J. Vac. Sci. Technol. A, 15, 1710. 79 Guo, Q., Oh, W.S., and Goodman, D.W. (1999) Surf. Sci., 437, 49. 80 Lai, X., Guo, Q., Min, B.K., and Goodman, D.W. (2001) Surf. Sci., 487, 1. 81 Chen, M.S., Wallace, W.T., Kumar, D., Yan, Z., Gath, K.K., Cai, Y., Kuroda, Y., and Goodman, D.W. (2005) Surf. Sci., 581, L115. 82 McCavish, N.D. and Bennett, R.A. (2003) Surf. Sci., 546, 47. 83 Ashworth, T.V. and Thornton, G. (2001) Thin Solid Films, 400, 43. 84 Ashworth, T.V., Muryn, C.A., and Thornton, G. (2005) Nanotechnology, 16, 3041. 85 Papageorgiou, A.C., Cabailh, G., Chen, Q., Resta, A., Lundgren, E., Andersen, J.N., and Thornton, G. (2007) J. Phys. Chem. C, 111, 7704. 86 Papageorgiou, A.C., Pang, C.L., Chen, Q., and Thornton, G. (2007) ACS Nano, 1, 409. 87 M€annig, A., Zhao, Z., Rosenthal, D., Christmann, K., Hoster, H., Rauscher, H., and Behm, R.J. (2005) Surf. Sci., 576, 29. 88 Surnev, S., Vitali, L., Ramsey, M.G., Netzer, F.P., Kresse, G., and Hafner, J. (2000) Phys. Rev. B, 61, 13945. 89 Surnev, S., Kresse, G., Ramsey, M.G., and Netzer, F.P. (2001) Phys. Rev. Lett., 87, 086102. 90 Schoiswohl, J., Sock, M., Eck, S., Surnev, S., Ramsey, M.G., and Netzer, F.P. (2004) Phys. Rev. B, 69, 155403. 91 Ranke, W., Ritter, M., and Weiss, W. (1999) Phys. Rev. B, 60, 1527.

j185

j 6 STM Imaging of Oxide Nanolayer Model Systems

186

92 Stone, P., Bennett, R.A., and Bowker, M. (1999) New J. Phys., 1, 8. 93 Bennett, R.A., Stone, P., Price, N., and Bowker, M. (1999) Phys. Rev. Lett., 82, 3831. 94 Takakusagi, S., Fukui, K., Nariyuki, F., and Iwasawa, Y. (2003) Surf. Sci., 523, L41. 95 Bursil, L.A. and Hyde, B.G. (1972) Prog. Solid State Chem., 7, 177. 96 Hermann, K. and Witko, M. (2001) Chapter 4, in The Chemical Physics of Solid Surfaces: Oxide Surfaces, vol. 9 (ed. D.P. Woodruff), Elsevier Science, p. 136 and references therein. 97 Surnev, S., Ramsey, M.G., and Netzer, F.P. (2003) Prog. Surf. Sci., 73, 117. 98 Kung, H.H. (1989) Transition Metal Oxides, Surface Chemistry and Catalysis, Elsevier, New York. 99 Weckhuysen, B.M. and Keller, D.E. (2003) Catal. Today, 78, 25. 100 Stefanovich, G., Pergament, A., and Stefanovich, D. (2000) J. Phys.: Condens. Matter, 12, 8837. 101 Schoiswohl, J., Surnev, S., Sock, M., Ramsey, M.G., Kresse, G., and Netzer, F.P. (2004) Angew. Chem. Int. Ed., 43, 5546. 102 Schoiswohl, J., Sock, M., Surnev, S., Ramsey, M.G., Netzer, F.P., Kresse, G., and Andersen, J.N. (2004) Surf. Sci., 555, 101. 103 Kresse, G., Surnev, S., Schoiswohl, J., and Netzer, F.P. (2004) Surf. Sci., 555, 118. 104 Schoiswohl, J., Surnev, S., and Netzer, F.P. (2005) Top. Catal., 36, 91. 105 Schoiswohl, S., Eck, S., Ramsey, M.G., Andersen, J.N., Surnev, S., and Netzer, F.P. (2005) Surf. Sci., 580, 122. 106 Weiss, W. and Ranke, W. (2002) Prog. Surf. Sci., 70, 1. 107 Ritter, M., Ranke, W., and Weiss, W. (1998) Phys. Rev., B57, 7240. 108 Weiss, W., Barbieri, A., Van Hove, M.A., and Somorjai, G.A. (1993) Phys. Rev. Lett., 71, 1848. 109 Barbieri, A., Weiss, W., Van Hove, M.A., and Somorjai, G.A. (1994) Surf. Sci., 302, 259.

110 Galloway, H.C., Benitez, J.J., and Salmeron, M. (1993) Surf. Sci., 298, 127; (1994) J. Vac. Sci. Technol., A12, 2302. 111 Galloway, H.C., Sautet, P., and Salmeron, M. (1996) Phys. Rev., B54, R11145. 112 Kim, Y.J., Westphal, C., Ynzunza, R.X., Galloway, H.C., Salmeron, M., Van Hove, M.A., and Fadley, C.S. (1997) Phys. Rev., B55, R13448. 113 Ritter, M., Ranke, W., and Weiss, W. (1998) Phys. Rev., B57, 7240. 114 Massalski, T.B., Murray, J.L., Bennett, L.H., and Baker, H. (1986) Binary Alloy Phase Diagrams, American Society for Metals, Metals Park, OH, p. 1807. 115 Tasker, P.W. (1979) J. Phys., C12, 4977. 116 Choudhary, V.R., Rajput, A.M., and Mamman, A.S. (1998) J. Catal., 178, 576. 117 Takeguchi, T., Furukawa, S., and Inove, M. (2001) J. Catal., 202, 14. 118 Zaera, F., Gleason, N.R., Klingenberg, B., and Ali, A.H. (1999) J. Mol. Catal. A: Chem., 146, 1323. 119 Christoskova, S.G., Stoyanova, M., Georgieva, M., and Mehandzhiev, D. (2002) Appl. Catal. A, 173, 95101. 120 Ruckenstein, E. and Hu, Y.H. (2002) Appl. Catal. A, 183, 85. 121 Christoskova, S.G., Danova, N., Georgieva, M., Argirov, O.K., and Mehandzhiev, D. (1995) Appl. Catal. A, 128, 219. 122 Duma, V., Poppa, K.E., Kung, M.C., Zhou, H., Nguyen, S., Ohyamac, S., Kung, H.H., and Marshall, C.L. (2004) Chem. Eng. J., 99, 227–236. 123 Kordesh, K. and Simadar, G. (1996) Fuel Cells and Their Applications, Wiley-VCH Verlag, Weinheim. 124 Escudero, M.J., Rodrigo, T., and Daza, L. (2005) Catal. Today, 107–108, 377–387. 125 Hotovy, I., Huran, J., and Siciliano, P. (2004) Sens. Actuators B, 103, 300. 126 Canalini, C., Post, M., Buso, D., Guglielmi, M., and Martucci, A. (2005) Sens. Actuators B, 108, 184–192. 127 Hotovy, I., Rehacek, V., and Siciliano, P. (2002) Thin Solid Films, 418, 9.

References 128 Dirksen, J., Duval, K., and Ring, T.A. (2001) Sens. Actuators B, 80, 106. 129 Shim, H.S., Ahn, H.J., Seong, T.Y., and Park, K.W. (2005) Electrochem. Solid-State Lett., 8, A277. 130 Fereira, F.F. and Fantini, M.C.A. (2004) Solid State Ionics, 175, 517–520. 131 Kikatsu, N., Maurice, V., and Markus, P. (1998) Surf. Sci., 411, 215. 132 B€aumer, M., Cappus, D., Kuhlenbeck, H., Freund, H.-J., Wilhelmi, G., Brodde, A., and Neddermeyer, H. (1991) Surf. Sci., 253, 116. 133 Kikatsu, N., Maurice, V., Hinnen, C., and Markus, P. (1998) Surf. Sci., 407, 36. 134 Hildebrandt, S., Hagendorf, Ch., Doege, T., Jeckstiess, Ch., Kulla, R., and Neddermeyer, H. (2000) J. Vac. Sci. Technol., A18, 1010. 135 Ventrice, C.A., Jr., Bertrams, Th., Hannemann, H., Brodde, A., and Neddermeyer, H. (1994) Phys. Rev., B49, 5773. 136 Hannemann, H., Ventrice, C.A., Jr., Bertrams, Th., Brodde, A., and Neddermeyer, H. (1994) Phys. Rev. Stat. Sol. A, 146, 289. 137 Marre, K. and Neddermeyer, H. (1993) Surf. Sci., 287–288, 995. 138 Bertrams, Th. and Neddermeyer, H. (1996) J. Vac. Sci. Technol., B14, 1141. 139 Wollschl€ager, J., Erdoes, D., Goldbach, H., Hopken, R., and Schr€oder, K.M. (2001) Thin Solid Films, 400, 1. 140 Lamberti, C., Gropo, E., Prestipino, C., Cassasa, S., Ferrari, A.M., and Pisani, C. (2003) Phys. Rev. Lett., 91, 46101. 141 Caffio, M., Rovida, G., Atrei, A., Giovanardi, C., di Bona, A., and Valeri, S. (2003) Surf. Sci., 531, 368. 142 Caffio, M., Cortigiani, B., Rovida, G., Atrei, A., and Giovanardi, C. (2004) J. Phys. Chem. B, 108, 9919. 143 May, F., Tisher, M., Arvantis, D., Russo, M., Hunter Dunn, J., Henneken, H., Wende, H., Clauvistr, R., Mårtensson, N., and Baberschke, K. (1996) Phys. Rev., B53, 1076.

144 Barbier, A., Stanescu, S., Boeglin, C., and Deville, J.-P. (2003) Phys. Rev. B, 68, 245418. 145 Stanescu, S., Boeglin, C., Barbier, A., and Deville, J.-P. (2004) Surf. Sci., 549, 172. 146 Agnoli, S., Orzali, T., Sambi, M., Granozzi, G., Schoiswohl, J., Surnev, S., and Netzer, F.P. (2005) J. Electron Spectrosc. Relat. Phenom., 144–147, 465. 147 Agnoli, S., Sambi, M., Granozzi, G., Atrei, A., Caffio, M., and Rovida, G. (2005) Surf. Sci., 576, 1. 148 Schoiswohl, J., Agnoli, S., Xu, B., Surnev, S., Sambi, M., Ramsey, M.G., Granozzi, G., and Netzer, F.P. (2005) Surf. Sci., 599, 1. 149 Agnoli, S., Sambi, M., Granozzi, G., Schoiswohl, J., Surnev, S., Netzer, F.P., Ferraro, M., Ferrari, A.M., and Pisani, C. (2005) J. Phys. Chem., B108, 17197. 150 Schoiswohl, J., Zheng, W., Surnev, S., Ramsey, M.G., Granozzi, G., Agnoli, S., and Netzer, F.P. (2006) Surf. Sci., 600, 1099. 151 Wolf, D. (1992) Phys. Rev. Lett., 68, 3315. 152 Barbier, A., Mocuta, C., Kuhlenbeck, H., Peters, K.F., Richter, B., and Renaud, G. (2000) Phys. Rev. Lett., 84, 2897. 153 Schoiswohl, J., Mittendorfer, F., Surnev, S., Ramsey, M.G., Andersen, J.N., and Netzer, F.P. (2006) Phys. Rev. Lett., 97, 126102. 154 Parteder, G., Allegretti, F., Wagner, M., Ramsey, M.G., Surnev, S., and Netzer, F.P. (2008) J. Phys. Chem. C, 112, 19272. 155 Trovarelli, A. (ed.) (2002) Catalysis by Ceria and Related Materials, Imperial College Press, London. 156 Fukui, K., Takakusagi, S., Tero, R., Aizawa, M., Namai, Y., and Iwasawa, Y. (2003) Phys. Chem. Chem. Phys., 5, 1. 157 Conesa, J.C. (1995) Surf. Sci., 339, 337. 158 N€orenberg, H. and Briggs, G.A.D. (1997) Phys. Rev. Lett., 79, 4222. 159 Esch, F., Fabris, S., Zhou, L., Montini, T., Africh, C., Fornasiero, P., Comelli, G., and Rosei, R. (2005) Science, 309, 752. 160 Fukui, K., Namai, Y., and Iwasawa, Y. (2002) Appl. Surf. Sci., 188, 252.

j187

j 6 STM Imaging of Oxide Nanolayer Model Systems

188

161 Namai, Y., Fukui, K., and Iwasawa, Y. (2003) J. Phys. Chem., B107, 11666. 162 Gritschneder, S., Namai, Y., Iwasawa, Y., and Reichling, M. (2005) Nanotechnology, 16, S41. 163 Mullins, D.R., Radulovic, P.V., and Overbury, S.H. (1999) Surf. Sci., 429, 186. 164 Lu, J.-L., Gao, H.-J., Shaikhutdinov, S., and Freund, H.-J. (2006) Surf. Sci., 600, 5004. 165 Schierbaum, K.-D. (1998) Surf. Sci., 399, 29. 166 Berner, U. and Schierbaum, K. (2001) Thin Solid Films, 400, 46. 167 Berner, U. and Schierbaum, K.-D. (2002) Phys. Rev., B65, 235404.

168 Eck, S., Castellarin-Cudia, C., Surnev, S., Ramsey, M.G., and Netzer, F.P. (2002) Surf. Sci., 520, 173. 169 Castellarin-Cudia, C., Surnev, S., Schneider, G., Podloucky, R., Ramsey, M.G., and Netzer, F.P. (2004) Surf. Sci., 554, L120. 170 Hla, S.-W., Bartels, L., Meyer, G., and Rieder, K.H. (2000) Phys. Rev. Lett., 85, 2777. 171 Chen, P. and Somorjai, G.A (2002) In situ catalysis and surface science methods, In-Situ Spectroscopy in Heterogeneous Catalysis (ed. J.F. Haw), Wiley-VCH Verlag, Weinheim, pp. 15–52.

j189

7 Surface Mobility of Atoms and Molecules Studied with High-Pressure Scanning Tunneling Microscopy Gabor A. Somorjai, Feng Tao, and Derek Butcher

7.1 Introduction

Atomic and molecular studies of solid surfaces have been an intensive research area over the past few decades. Most of these investigations were carried out in vacuum to take advantage of electron, ion, and molecular beam scattering techniques that provided surface sensitivity because of their high scattering cross sections. However, most surface phenomena occur at solid–high-pressure gas and solid–liquid interfaces that we call “buried interfaces,” including heterogeneous catalysis, electrochemistry, and corrosion, to mention a few. The lack of techniques for studies of buried interfaces was commonly called the “pressure gap” in surface science. There has been a great deal of effort in our laboratory and elsewhere to develop surface techniques that provide molecular information at buried interfaces. Our laboratory has focused on developing three of these techniques: high-pressure scanning tunneling microscopy (HPSTM), sum frequency generation (SFG) vibrational spectroscopy, and ambient pressure X-ray photoelectric spectroscopy (APXPS). This paper reviews the development of the HPSTM technique and our research results, including the determination of surface mobility, surface structure on the atomic and molecular scales, and reaction kinetics.

7.2 Characterization of Surface Mobility of Molecules and Atoms

Since the most active catalytic sites are usually steps, kinks, and surface defects, atomically resolved structural information including atomic distribution and surface structure at low pressure, possible surface restructuring, and the mobility of adsorbate molecules and of the atoms of the catalyst surface at high temperature and high pressure is crucial to understanding catalytic mechanisms on transition metal surfaces. The importance of studying the structural evolution of both adsorbates

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

190

and catalyst surfaces at high pressure has driven the development of new surface analytical techniques to determine surface structure and mobility of molecules and atoms in the harsh conditions of industrial catalysis. Scanning tunneling microscopy (STM) has the capability to examine molecules and atoms of adsorbate and catalyst surface to address several crucial issues in heterogeneous catalysis, including the rearrangement of surface atoms, restructuring of the catalyst surface, and mass transfer. Since the advent of STM in the 1980s [1], it has been one of the dominant tools to study catalytic reactions with a vacuum surface science approach. This is because it has the unique ability to investigate adsorption and surface structure with atomic resolution and in some cases the electronic structure as well. This technique can be applied to a pressure range from ultrahigh vacuum (UHV) to atmospheric or high pressure since the tunneling process between the sample and tip occurs only in a very close range of 5–50 Å. We first applied STM to study adsorbate and surface structure at high pressure by filling reactant gases into an STM chamber and keeping them at the high-pressure during STM scanning [2, 3]. After the first demonstration, it has been applied to catalysis studies under a condition of a high pressure by a few groups [4–11]. Recently, a new high-pressure high-temperature STM reactor/UHV system was designed and homebuilt by our group in 2006 and 2007 [12]. In this design, the STM body, sample, and tip are placed in a small high-pressure reactor (19 cm3) installed in a UHV chamber. A sealable port on the wall of the reactor separates the highpressure environment of reactant gases in the reactor from the vacuum environment of the UHV chamber and permits one to change sample and tip in UHV. The combination of a sample transfer arm, wobble stick, and sample load-lock system allows convenient transfer of samples and tips between the preparation chamber, high-pressure reactor, and ambient environment. Experiments performed on inert samples such as HOPG and reactive samples such as hex-Pt(1 0 0) both in vacuum and in high-pressure conditions demonstrate the ability to perform in situ investigations of heterogeneous catalysis and surface chemistry at atomic resolution at a wide pressure range from UHV to a pressure higher than that of atmosphere and a temperature range from 300 to 700 K. Other than the STM techniques for catalysis studies at high-pressure conditions in our group, sum frequency generation vibrational spectroscopy and ambient pressure X-ray photoelectron spectroscopy were also developed for catalysis studies at high pressures. Early on we developed SFG vibrational spectroscopy to study surface chemistry at high-pressure conditions with an emphasis on catalysis [13, 14]. SFG is a surface-specific spectroscopy that can be used to measure vibrational spectra under realistic catalytic conditions to allow us to identify the reaction intermediates and products and therefore reveal the catalytic mechanism [15–17]. Recently, a synchrotron-based ambient pressure XPS was developed [18–20]. Since the XPS signal is measured through the collection of electrons ejected from the sample, the instrument performance is highly susceptible to scattering from gas molecules between the sample and the detector. The APXPS was developed with the idea of bringing a sample close enough to a differentially

7.3 High-Pressure STM Technique and Instrumentation

pumped skimmer cone with a small aperture diameter of <0.3 mm to decrease the electron collision rate. Taking advantage of the tunability of X-ray energy from a synchrotron source, this technique is highly desirable for fundamental study of catalysis in realistic conditions. It is particularly useful for in situ exploration of the evolution of composition and restructuring, the catalyst surface, and the atomic layers under the catalyst surface during catalysis.

7.3 High-Pressure STM Technique and Instrumentation

The application of STM to catalysis studies at high-pressure and high-temperature condition has a great challenge in instrumentation. In general, a homebuilt STM body combined into a high-pressure reactor or a flowing cell is needed; or at least a major modification for a commercial STM system is required. To study catalysis at a high pressure of reactant gases and a high-temperature catalyst surface, a specially designed STM system should meet the following requirements: (1) a reactor is needed for maintaining a high-pressure environment of reactant gases; (2) the volume of the reactor or flowing cell should be as small as possible for several reasons including maintaining a high concentration of reaction products, low gas consumption, easily maintaining a constant pressure of reactants during an experiment, quickly restoring a UHV environment in the reactor after each experiment; (3) in situ heating of the sample must be available for studying catalysis at high temperature; (4) it must be possible to isolate the high-pressure environment in a batch or flow reactor from the vacuum environment of a UHV system to remain at high pressure during the catalytic reaction; (5) the reactor must be accessible to the UHV system for sample cleaning and tip treatment and analysis of the catalyst surface before and after the catalytic reaction; (6) there must be a reliable gas introduction system; and (7) the system must include product measurement. A simple method to carry out high-pressure STM studies is to fill the STM chamber with reactant gases while the reactant gases remain in the chamber. However, this method has disadvantages such as a large volume of reactant gases and limits in sample heating and reactant gas pressure. To overcome these difficulties, we designed and homebuilt a novel high-pressure–high-temperature STM reactor/UHV system [12]. The new system overcomes these limitations and allows catalytic studies under a wide pressure range of reactant gases (1010–5000 Torr) and a wide temperature range (300–700 K). Figure 7.1 shows photographs of the STM reactor/UHV system and electronic control units. The system is schematically presented in Figure 7.2. The STM chamber is separated from the sample preparation chamber by an 8 in. gate valve. A quadrupole mass spectrometer is installed in this chamber for monitoring reactants and products during STM scanning. The STM chamber contains the high-pressure reactor, a docking scaffold assembled on a customdesigned sample manipulator, and a wobble stick for sample and STM tip transfer between the transfer rod, high-pressure reactor, and docking disk.

j191

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

192

Figure 7.1 Pictures of the high-pressure high-temperature STM reactor/UHV system.

The high-pressure reactor (Figure 7.3) is a special vessel designed for assembling the STM body and sample, and providing a high-pressure environment of reactant gases. It houses a homebuilt STM body. It has a volume of approximately 19 cm3 and is placed inside the vacuum environment of a UHV chamber by a special docking scaffold and mounting framework (2, 3, and 4 of Figure 7.2). It is vibrationally isolated from the UHV chamber with three springs, offering the capability of imaging the surface with atomic resolution. The reactor consists of a cell lid, cell neck, cell top stage, and cell bottom stage. There are recesses on the cell neck, cell top stage, cell bottom stage, and port of the reactor, which host Kalrez o-rings, forming gas-tight seals. The sealing of the Kalrez o-rings allows pressurization of the reactor while

7.3 High-Pressure STM Technique and Instrumentation

Figure 7.2 Schematic diagram of the highpressure high-temperature STM reactor/UHV system. (1) View window, (2) mounting framework, (3) docking scaffold, (4) docking disk, (5) high-pressure reactor (STM body

housed within), (6) bayonet seal, (7) guide rod of docking scaffold, (8) sample/tip load-lock system, (9) transfer rod, (10) gate valve, (11) four-finger sample stage, and (12) sputtering ion gun.

maintaining a high vacuum in the surrounding chamber. All four sections of the reactor are assembled together with four venting screws and sealed by these o-rings. The reactor was plated with a layer of gold (5 mm thick) to avoid possible reactions between the materials of the high-pressure reactor and reactant gases. The cell lid has a set of holes precisely designed to glue a set of pin-socket contacts (set II in Figure 7.3a) for assembling a set of male contacts (set I) from the docking scaffold and another set of male contacts (set III in Figure 7.3c) for the wiring connections from the shear piezoelectric plates and the scanning tube. These pin-socket contacts provide convenient detachable wiring connections for the high-pressure reactor. The contacts (set I) from the docking scaffold can be inserted into the vacuum side of the interfacial contacts (set II) glued on the cell lid. Another set of contacts (set III) is glued onto a set of holes on the STM body (Figure 7.3b) that have the exact same size and arrangement as the holes on the cell lid (Figure 7.3a). This convenient pin-socket wiring strategy (Figure 7.3c) makes dismantling and assembling the STM convenient when maintenance is required on the high-pressure reactor and STM body. The STM body (Figure 7.3b) is the key component of the high-pressure reactor. It includes a coarse approach system, a scanning tube, a receiver for the tip holder,

j193

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

194

Figure 7.3 High-pressure reactor and STM body. (a) A view of the whole reactor: (1) cell lid, (2) cell neck, (3) middle stage, (4) bottom stage, (5) pin-socket contacts for wiring of STM body at the interface of high-vacuum and high-pressure environment, (6) a pin on the lever of the cell lid for mounting, (7) Swagelok fitting welded on the tube of the cell lid for gas introduction, (8) Swagelok fitting welded on the button stage of reactor for gas exit, (9) port of reactor for sample and tip transfer. (b) Side view of the STM body: (1) Wall of STM body coated with a layer of gold, (2) hexagonal sapphire, (3) CuBe spring plates, (4) receiver of tip holder, (5) screw to adjust the pressure

applied to hexagonal sapphire, (6) hole to assemble sample stage to STM body. (c) Scheme of pin-socket alignment for wiring. Set III is the contacts glued to the holes on top of STM body; Set II is the contacts glued to the holes on top of cell lid of the reactor; Set I is the contacts glued to wires from docking scaffold. (d) Scheme showing the assembly of three packs of shear piezoelectric plates between the hexagonal sapphire and the wall of the STM body and the assembly of the scanning tube: (1) a set of piezoelectric plates, (2) wall of STM body, (3) hexagonal sapphire, (4) scanning tube. (e) Scheme showing how the four shear piezoelectric plates of one set work for coarse approach.

7.3 High-Pressure STM Technique and Instrumentation

a sample stage assembled at the end of the STM body, and wire connections to these parts. The coarse approach is carried out by six sets of shear piezoelectric plates (three of them are schematically shown in a bottom view of STM body in Figure 7.3d) located between a hexagonal sapphire (Figure 7.3b2) and the wall of the STM body. One side of each shear piezoelectric set is glued to the internal wall of the STM body through an alumina plate for thermal isolation while the other side is glued to another alumina plate that contacts the surface of the hexagonal sapphire (Figure 7.3d and e). By applying negative or positive voltages to the first/third and the second/fourth piezoelectric plates, the lateral force moves the hexagonal sapphire forward and backward (Figure 7.3e). A single piezoelectric scanning tube is glued to an alumina disk that is in turn glued to one end of the hexagonal sapphire. Five Kapton wires are glued to the five components ( þ x, x, þ y, y, z) of the scanning tube through holes on the alumina disk. Another alumina disk is glued to the other end of the scanning tube onto which a bowl-shaped tip receiver is glued (Figure 7.3b4). The central part of this tip receiver is a SmCo magnet. The tip change mechanism is described below. A flexible coaxial wire is glued to this tip receiver for transmitting the tunneling current. A sample assembly stage is screwed to the end of the STM body (Figure 7.3b6), which is thermally and electrically insulated from it by three precisely aligned sapphire balls and insulating washers. For sample transfer and tip change, a port is opened on the wall of the bottom stage of the high-pressure reactor (Figure 7.3a9). A bayonet seal (inset of Figure 7.3a) was fabricated to seal the port on the reactor. This sealing keeps the pressure in vacuum chamber lower than 5  107 Torr when the high-pressure reactor is filled with 1000 Torr N2. The naturally leaked gases from the sealing interface between different sections of the reactor can be analyzed in situ with a mass spectrometer during reaction. The in situ sample heating during reaction is carried out by a light source outside the high-pressure reactor to avoid heating elements in the high-pressure environment. It consists of a halogen lamp with an elliptical reflector that focuses the radiation onto the sample through a sapphire window welded at the bottom of the reactor (Figure 7.4). The distance between the lamp and the reactor can be adjusted to focus the light on the back of the sample for efficient heating. The heating rate can be controlled by adjusting the power supplied to the lamp. A K-type thermocouple was spot-welded to the sample stage for both sample bias and temperature measurements. Another thermocouple was attached to the STM body to monitor the temperature of the shear piezoelectric plates during sample heating. Thus, thermal diffusion and possible increases in the temperature of the STM body can be simultaneously monitored when the sample in the high-pressure reactor is heated. Replacement of the STM tip is accomplished using a magnetic tip exchanger with the same geometry as the sample holder (Figure 7.5a) and a tip holder (Figure 7.5b). The tip exchanger can be easily transferred to and from the high-pressure reactor (Figure 7.5c), the storage disk, and the load-lock system. Figure 7.6 schematically shows the setup of gas introduction for the high-pressure reactor. For gas introduction, the male part of a Swagelok fitting (4 in Figure 7.6) is welded on a 1/8 in. tube of the cell lid (3 in Figure 7.6). A 1/32 in. PEEK tube

j195

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

196

Figure 7.4 STM chamber in situ sample heating system. (1) Halogen lamp, (2) elliptical reflector, (3) sapphire window welded at the center of the bottom of the high-pressure reactor, (4) the assembled sample, (5) high-pressure reactor. Turquoise dashed line shows the alignment of light beam and sample center.

(7 in Figure 7.6) capable of supporting high pressure was glued to a 1/16 in. stainless steel tube (6 in Figure 7.6) assembled into a female part of the Swagelok fitting (5 in Figure 7.6). Another end of the PEEK tube is also glued and assembled to another Swagelok fitting welded to a hole in a double-sided CF flange. This design isolates the reactant gases in the high-pressure reactor from the vacuum environment of the STM chamber. The high-pressure gases in the reactor can be pumped down by a turbo molecular pump to obtain a UHV environment after completion of a high-pressure experiment, therefore quickly restoring a UHVenvironment. Thus, this high-pressure reaction system can work under both UHV and high pressure, offering the capability of studying catalysts over the entire pressure range from 1  1010 to 5000 Torr. In addition, the reactions can be carried out using batch or flow reactor mode. Atomically resolved images of inert HOPG under 700 Torr N2 and of hex-Pt(1 0 0) prepared in UHV can be routinely obtained with this homebuilt high-pressure hightemperature STM reactor/UHV system.

7.4 Mobility and Flexibility of Catalyst Surfaces

Figure 7.5 Parts used for a convenient tip exchange. (a) Tip exchanger: (1) the central slot that has a size between the stopping disk and outer diameter of the tube of the tip holder, (2) magnets. (b) Tip holder with

an empty tube for placing a tip: (1) empty tube for placing a STM tip, (2) stopping disk. (c) Magnetic bowl glued at the end of the scanning tube. The arrow shows the hidden bowl at the end of the scanning tube.

7.4 Mobility and Flexibility of Catalyst Surfaces at High-Pressure High-Temperature Reaction Conditions

The exploration of the surface of single-crystal model catalysts has a long history. Electron microscopy and field-emission microscopy developed in the 1950s revealed the existence of surface steps and kinks, which led to the development of the heterogeneous rigid surface model (Figure 7.7) proposed in 1960s [21]. In this model, a surface is heterogeneous and the atoms of the surface layer are arranged at the exactly same sites as those in the bulk. That is, the location of surface atoms can be exactly provided by the structural parameters of a bulk layer. This model ignored

j197

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

198

Figure 7.6 Scheme showing the introduction of high-pressure reactant gases to reactor when the reactor is under UHV environment of the STM chamber. (1) STM chamber, (2) highpressure reactor, (3) stainless steel tube (ID ¼ 1/8 in.), (4) and (10) male part of a Swagelok fitting, (5) and (9) female part of a Swagelok fitting, (6) and (8) silica-coated stainless steel tube (ID ¼ 1/16 in.), (7) PEEK tubing (OD ¼ 1/16 in., ID ¼ 1/32 in.), (11)

double-sided CF flange, (12) and (13) angle valves, (14) angle valve for pumping reactant gases after completion of a batch mode highpressure experiments, (15) vessel for mixing different reactant gases, (16) and (17) variable leak valves, (18) Baratron capacitance manometer for pressure measurements, (19) and (20) gas filters, (21) and (22) gas cylinders, and (23) angle valve.

the surface effects of relaxation and reconstruction. But it was reasonable to some extent at that time because it could rationalize the kinetics of crystal growth and the site-dependent performance of chemisorption and charge distribution. The new results obtained with the surface analytic techniques developed in 1980s showed that this heterogeneous rigid model was not correct. We put forward the concept of the flexible surface in 1991 [13, 22]. One structural evidence of the flexibility of surfaces is the inward relaxation of the first layer of a clean metal surface to produce a short

7.4 Mobility and Flexibility of Catalyst Surfaces

Figure 7.7 Heterogeneous rigid surface model.

interlayer distance in contrast to the distance between two adjacent bulk layers (Figure 7.8). In fact, extensive studies of surface structure confirm the general phenomena of surface relaxation as shown in Figure 7.9. Moreover, we discovered two rules that the flexibility of metal surfaces generally follow [22]: (a) the closer packed and lower Miller index surfaces on which atoms have large coordination are generally less flexible and (b) , more importantly, the flexibility of chemisorbed surfaces results in possible surface reactions induced by several factors, including adsorption and thermal-induced restructuring. This model gave a clearer picture of the catalyst surface [13, 22]. The proposed flexibility of the surface on the basis of vacuum surface science approach is definitely applicable to the catalyst surface under an environment of high pressure of reactive gases. As mentioned above, entropy could significantly contribute to the surface structure and flexibility at high-pressure conditions. Our recent studies showed that the surface mobility of catalyst surfaces is significantly increased at high pressures. Thus, catalyst surfaces under high pressures of reactant gases could be considered as mobile surfaces in some cases. Here, we use the

Figure 7.8 Restructuring of surface atoms at step sites of a clean surface.

j199

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

200

-40

410

411 -35 510 Δ(dz)12/Δ(dz)bulk (%)

-30 -25

110 311

-20 -15

320 311 332 211 511 322 221 210 332 321 221 331 321 320 210 310

331

111

-10

bcc

211 001

-5

322

111

fcc

bcc

} expt.

fcc

} theory

001

0 +5

110 1

2

3

4 5 ROUGHNESS

6

7

Figure 7.9 Experimental and theoretical first-layer relaxation (in percentage) as a function of roughness (¼ l/packing density) for several bcc and fcc surfaces.

pressure-dependent adsorption and surface structure of hex-Pt(1 0 0) under an environment of CO as an example to illustrate the mobility of the metal atoms of catalyst surfaces. Figure 7.10 is one image of the clean hex-Pt(1 0 0) surface prepared in UHV. The surface structures and the nanoclusters formed at different pressures of CO are shown in Figures 7.11 to 7.13. Overall, the surface of hex-Pt(1 0 0)

Figure 7.10 STM image of a clean hex-Pt(1 0 0) surface.

7.4 Mobility and Flexibility of Catalyst Surfaces

Figure 7.11 STM image of a lifted Pt(1 0 0) surface at a condition of 5  109 Torr CO.

undergoes a phase change and forms nanoclusters under certain pressures of CO at room temperature. More importantly, the size and shape of the nanoclusters formed are pressure-dependent. At a low pressure of 5  109 Torr (Figure 7.11), a portion of the surface is preferentially lifted along the [0 1 1] direction. However, when a clean hex-Pt(1 0 0) surface is exposed to approximately 105 Torr CO, islands with a rectangular shape and a size of 1.5–4.0 nm are formed (Figure 7.12). There is no preferential growth along any direction. At a high pressure of 1 Torr CO,

Figure 7.12 STM image of a lifted Pt(1 0 0) surface at a condition of 5  107 Torr CO.

j201

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

202

Figure 7.13 STM image of Pt(1 0 0) surface at a condition of 1 Torr CO.

nanoclusters with a round or ellipse shape are formed on the whole surface. Figure 7.14 shows the evolution of the surface structure formed at approximately 5  107 Torr CO in a time interval of 28 min. The landmarks in Figure 7.14 are used to identify the same location between two adjacent images since thermal drift is present at 300 K. This figure clearly shows the shape and size of the formed nanoclusters. During this time interval, the shape of islands changes quickly and a portion of the atoms in adjacent nanoclusters move together and then form new nanoclusters, indicating significant surface mobility for the lifted Pt atoms at this pressure. In fact, under 1 Torr CO, clean hex-Pt(1 0 0) is dramatically roughened and forms highly dense nanoclusters with a size of 1–3 nm on the whole surface (Figure 7.13). It suggests that a dramatic restructuring of the catalyst surface is induced by the presence of a high pressure of reactant gas. Notably, no atomically resolved image can be obtained from the catalyst surface in an environment with CO pressure higher than 104 Torr, whereas atomically resolved images can be obtained at a pressure of 1010–105 Torr CO. This suggests higher mobility of Pt atoms on the catalyst surface under higher pressure of CO. Temperature is another factor in surface mobility as evidenced by the aggregation of surface islands when the lifted islands are heated from the original 25–150  C (Figure 7.15). Figure 7.15a is the STM of hex-Pt(1 0 0) exposed to 1 Torr CO at room temperature. Clearly, the surface formed at this pressure of CO consists of Pt nanoclusters with a size of 2–3 nm. When this surface is heated in situ to 150  C, the lifted surface structure significantly changes (Figure 7.15b). The clusters formed at room temperature largely aggregate together at 150  C, forming clusters with a size of 3–7 nm. This aggregation illustrates again the high flexibility and mobility of a catalyst surface. Surface mobility and flexibility were evidenced by dramatic reconstruction of the Pt(1 0 0) surface upon annealing this surface at 150  C for 4–5 h in 1–1.6 atm of

7.4 Mobility and Flexibility of Catalyst Surfaces

Figure 7.14 STM images of a lifted Pt(1 0 0) surface at condition of 5  107 Torr CO: (a) t ¼ t0 min, (b) t ¼ t0 þ 10 min, (c) t ¼ t0 þ 19 min, and (d) t ¼ t0 þ 28 min. All the images have the same size of 14.1 nm  14.1 nm. They

were obtained with the same tunneling condition and scanning speed. Landmarks 1, 2, and 3 represent the same cluster between (a) and (b), between (b) and (c), and between (c) and (d), respectively.

reactant gases (H2, O2, or CO) [13, 23]. After exposing a clean Pt(1 1 0) surface to 1.6 atm H2 and annealing at 150  C for 5 h in this hydrogen environment, the surface is dominated by parallel rows that are the formed (1  n) missing rows randomly nested (Figure 7.16a). This clearly demonstrates reconstruction of a catalyst surface induced by a reactant gas at high temperature and high pressure. Compared to the restructuring driven by hydrogen, oxygen induces the faceting of the Pt(1 1 0) surface. After heating in 1 atm O2 for 5 h, this catalyst surface consists of microfaceted (1 1 1) with a width of 30–60 Å separated by steps (Figure 7.16b). The reconstruction of Pt(1 1 0) into numerous microfaceted Pt(1 1 1) surfaces could be driven by selective adsorption of oxygen atom on face-centered cubic threefold hollow sites [24]. Thus, the transformation from Pt(1 1 0) to small Pt(1 1 1) surfaces is definitely energetically favorable. When a clean Pt(1 1 0) surface is

j203

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

204

Figure 7.15 STM images of a lifted Pt(1 0 0) surface under 1 Torr CO at 300 K (a) and 420 K (b).

exposed to 1 atm CO and annealed at 150  C in this environment for 4 h, the restructured surface is smooth compared to the rough surface formed in H2 (Figure 7.16a). The surface restructured in CO (Figure 7.16c) has a width of approximately 1–30 Å and does not show missing row reconstruction. In fact, the high-pressure CO lifts the missing row reconstruction and thus forms a flat 1  1 structure separated by steps. Another example of the flexibility of the Pt catalyst is the reconstruction of a stepped Pt(1 1 1) crystal with adsorbed sulfur upon exposure to CO [25]. Single-crystal Pt(1 1 1) cut at an angle of approximately 5 from the (1 1 1) direction consists of numerous terraces with a width of 20–60 Å separated by steps with single-atom height. The adsorption of sulfur atoms restructures the clean stepped Pt(1 1 1) surface with single-atom steps into a sulfur-adsorbed surface with double-atom

7.4 Mobility and Flexibility of Catalyst Surfaces

Figure 7.16 (a) Topographic image of Pt(1 1 0) annealed at 150  C in 1.6 atm H2 for 5 h, showing (1  n) missing row reconstruction randomly nested. Image size: 730 Å  700 Å vertical range Dz ¼ 10 Å. (b) Topographic

image of Pt(1 1 0) annealed at 150  C in 1 atm O2 for 5 h. Image size: 900 Å  780 Å vertical range Dz ¼ 25 Å. (c) Topographic image of Pt (1 1 0) annealed in 1 atm CO for 4 h. Image size: 770 Å  740 Å vertical range Dz ¼ 42 Å.

height. The width of the terraces of the double-step surface is twice that of the original sample: 40–120 Å (Figure 7.17). A similar doubling of step height and terrace width was also observed on other metal surfaces [26]. This phenomenon is rationalized by a decrease in the step edge–step edge repulsive interaction that prevents the energetically favorable coalescence of steps. Surprisingly, the exposure of this double-atom

Figure 7.17 (a) A 250 Å  100 Å STM image of the stepped Pt(1 1 1) surface after depositing 0.25 LM of S and then annealing at 600  C. (b) Magnified image showing the p(2  2)-S

structure covering the terraces. (c) Line-profile analysis for the section marked with a line in (a). This analysis clearly shows the double-step heights formed by S chemisorption.

j205

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

206

Figure 7.18 (a) A 250  100 Å STM image of the S-covered Pt(1 1 1) after adsorption of CO from a background pressure of 1  106 Torr that is maintained during imaging. The double steps visible in Figure 7.17 before CO coadsorption are split into monoatomic steps separated by new terraces that contain exclusively CO. These new terraces appear as smooth bands in the image. Sulfur is compressed in the remaining alternating terraces and appears as disordered maxima in

the image. (b) A close view of the alternating CO- and S-covered areas. (c) A cross section of eight averaged scan lines along the line marked in (a), showing the single-atom step heights. The difference in heights of the S- and COcovered terraces is due to the difference in tunneling probability, which is lower through CO than through S. The large circles represent S atoms and the small lines represent CO molecules.

step surface to a background of CO (1  106 Torr) causes the step with double-atom height to split into two monoatomic height steps (Figure 7.18). This suggests the flexibility of the Pt catalyst surface under a relatively high pressure of reactant gas. In fact, the restructuring of the adsorbate-covered surface by coadsorption of a second reactant gas stems from three factors: a repulsive electrostatic interaction upon CO adsorption at the step edge, the mechanistic interaction, and an effective repulsive interaction resulting from entropy [25].

7.5 Adsorbate Mobility During Catalytic Reactions

Heterogeneous catalysis is by nature a surface phenomenon. An efficient catalytic system under reactive conditions demands a delicate energetic balance. For a catalytic reaction to occur, reactant molecules must adsorb on a surface, react, and

7.5 Adsorbate Mobility During Catalytic Reactions

then desorb in a cyclical process, leaving the active site of the catalyst free for the next set of reactants. Depending on the specific reaction, this may involve multiple species with differing adsorption energetics and mechanisms. There are also a wide variety of reaction mechanisms. For bimolecular reactions, these can be classified into two types: a reaction between two adsorbed species, which is known as the Langmuir–Hinshelwood mechanism, and a reaction between an adsorbed species and a colliding gas-phase species, which is known as the Eley–Rideal mechanism. The collision frequency, the sticking coefficient, the residence time, the diffusion rates of the surface species, the concentration of active sites, and the presence of poisoning species all play an important role in the rate of reaction. For reactions involving two adsorbed species, the density of surface vacancies and their role in the diffusion rates of the adsorbed species will have a large effect on the turnover rate. This complex interplay of energetic factors is the central issue of all catalytic behavior [13]. There are a wide variety of techniques that can provide molecular-level information about a surface [13]. The majority of these techniques give ensemble averages of a large area and many are restricted to a working condition of high vacuum. STM can provide a valuable complement to surface studies on well-defined extremely flat systems because it is capable of atomic resolution imaging of the surface structure and morphology under a wide range of pressures and temperatures. We first applied STM technique to study catalytic reactions in situ [2]. This technique combines the ability to prepare well-ordered atomically clean samples with the capability to study those samples under realistic catalytic conditions. This section will discuss the application of STM to catalytically active systems and the effect it has had on the characterization of the adsorbate layer. Much attention has been given to the pressure gap in heterogeneous catalysis [27], which refers to the energetic disparity in systems at UHV and at industrially relevant pressures. Under UHV conditions, a sample might average 104 gas-phase collisions per surface atom per second compared to 109 collisions per surface atom per second at atmospheric pressure. The energy transfer from these gas-phase collisions as well as from adsorption, desorption, and reaction events could result in a significantly higher occurrence of high-energy barrier processes compared to the low-pressure system. These factors underline the advantages of studying catalytic systems in situ with high-pressure STM. 7.5.1 Ethylene Hydrogenation on Pt(1 1 1)

CO poisoning of the ethylene hydrogenation reaction on Pt(1 1 1) at high pressure was studied by our group [28]. A clean Pt(1 1 1) surface was prepared with sputtering and annealing cycles in the sample preparation chamber and then transferred to the STM chamber. Surface morphology of this single-crystal model catalyst was examined after the reactant gas was introduced into the STM chamber. At both 20 mTorr H2 (Figure 7.19a) and with a mixture of 20 mTorr H2 and 20 mTorr C2H4 (Figure 7.19b), there was no ordered structure detected in the STM images. When

j207

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

208

Figure 7.19 (100 Å  100 Å) STM images of the Pt(1 1 1) surface under different pressures: (a) 20 mTorr H2, (b) 20 mTorr H2 and 20 mTorr ethylene, and (c) 20 mTorr H2, 20 mTorr ethylene, and 2.5 mTorr CO.

The presence of CO induced the formation of a (H19  H19)R23.4 structure on the surface. (d) 200 Å  200 Å STM image showing two rotational domains of (H19  H19)R23.4 .

the H2/C2H4 mixture was dosed into the chamber, the hydrogenation reaction was monitored by the production of ethane in the mass spectra. The absence of ordering was attributed to diffusion of the adsorbate species on timescales much faster than the scan speed of the STM tip, which was estimated to be 10 mm s1. This phenomenon has been observed by Dunphy et al., who were able to show an ordered structure through a correlation analysis of an adsorbed layer diffusing slightly faster than the acquisition time of the STM images [29]. When CO was added to the H2/C2H4 mixture, the reaction was poisoned and an ordered structure, a Moire interference pattern, was observed by STM (Figure 7.17c and d). This Moire pattern stems from the interference between the hexagonal surface lattice of the Pt atoms of Pt(1 1 1) and a different hexagonal lattice for the adsorbed CO molecules. The adsorption of hydrogen, ethylene, and CO on Pt(1 1 1) was extensively studied. Molecular hydrogen dissociatively adsorbs on the catalytic Pt(1 1 1) surface

7.5 Adsorbate Mobility During Catalytic Reactions

and the dissociated hydrogen atoms bind preferentially to threefold hollow sites [30]. The ethylene on Pt(1 1 1) exhibits obvious temperature dependence. Below 50 K, ethylene physisorbs on the surface. Above 60 K, it forms a di-s-bonded species at two adjacent hollow sites [31]. At room temperature, this species undergoes proton transfer to form ethylidyne (CCH3), which binds most strongly at hollow sites with the CC bond perpendicular to the surface. At high pressure and 300 K, another adsorbate is formed. This adsorbate, a p-bonded ethylene molecule, weakly binds to an on-top site through p-stacking with the C¼C bond parallel to the surface. Ethylidyne promotes the formation of the p-bonded species by binding to the threefold sites and leaving on-top sites open. The weakly p-bonded ethylene molecule then acts as the reactive surface intermediate for ethylene hydrogenation. CO on Pt(1 1 1) is one of the most extensively studied model systems in heterogeneous catalysis. It was studied using several surface techniques such as high-pressure STM [32], ambient pressure XPS [33], and surface-specific sum frequency generation vibrational spectroscopy [34]. The structural evolution of adsorbed CO in the entire pressure range from UHV to one atmosphere was studied with STM [32]. The highpressure structure forms a moire interference pattern resulting from the hexagonal symmetry of the adsorbed CO overlaid on the hexagonal packed Pt(1 1 1) surface. The CO layer is incommensurate with the Pt atoms until the pressure reaches 760 Torr, when it forms the commensurate (H19  H19)R23.4 -13CO phase. The incommensurate phase results from the electrostatic repulsion of the neighboring CO adsorbate molecules [32]. The absence of a visible ordered structure of hydrogen or ethylene adsorbates under STM suggests a high mobility of the adsorbed reactant species. The measurement of ethane under the above reaction condition suggests that molecular mobility is necessary for product formation. The formation of an ordered structure after CO is introduced into the mixture of high-pressure hydrogen and ethylene confirmed the mobility of reactant species and the existence of surface vacancies. More importantly, the comparison of structural information obtained before and after the introduction of CO demonstrates a method to use high-pressure STM to study the mobility of molecular adsorbates on catalyst surfaces. This approach can definitely be applied to studies of the molecular mobility of other catalysis reactions under realistic conditions. Particularly, the combination of HPSTM with other highpressure techniques that identify chemical composition and vibrational signatures will provide a clear picture of the catalytic mechanism of reactions performed under realistic conditions. 7.5.2 Hydrogenation of C6 Cyclic Hydrocarbons on Pt(1 1 1)

The adsorption of cyclohexene, cyclohexane, 1,3-cyclohexadiene, 1,4-cyclohexadiene, and benzene on Pt(1 1 1) was studied with STM [35, 36]. Figure 7.20a shows an STM image of 2  106 Torr cyclohexene on Pt(1 1 1). The low-pressure structure shows a hexagonal symmetry with a periodicity of approximately 7 Å that is rotated approximately 18–20 with respect to the [1 1 0] direction of the Pt crystal face. From prior

j209

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

210

Figure 7.20 (a) 60 Å  60 Å STM image of Pt(1 1 1) in the presence of 2  106 Torr cyclohexene at 300 K. Periodicity of approximately 6.5 Å. Spectroscopy studies have found surface species to be p-allyl.

Lines represent [1 1 0] directions of underlying platinum. A unit cell of the adsorbate structure is drawn. (b) Diagram of the proposed (H7  H7) R19.1 model for the p-allyl structure in Figure 7.18a. Inset shows bonding structure.

studies of this system using SFG, high-resolution electron energy loss spectroscopy (HREELS), and thermal desorption spectroscopy (TDS) [37, 38], this structure can be assigned as a (H7  H7)R19.1 C6H9 p-allyl structure, which preferentially binds in a chair configuration to threefold hollow sites (inset of Figure 7.20b). The partially dehydrogenated p-allyl adsorbate structure is similar to that formed from the adsorption of low-pressure cyclohexane (Figure 7.21). Molecular cyclohexane is not the most stable adsorbate species, which leads to a similar dehydrogenation to the pallyl covered surface [38]. The adsorbed structure of the further dehydrogenated species was also imaged.

7.5 Adsorbate Mobility During Catalytic Reactions

Figure 7.21 75 Å  75 Å STM image of Pt(1 1 1) in the presence of 2  106 Torr cyclohexane at 300 K. Periodicity of approximately 7 Å. Spectroscopic studies suggest that the surface species is p-allyl.

At a pressure of 1  105 Torr 1,4-cyclohexadiene, the surface shows ordered domains of hexagonal rings approximately 18 Å in diameter made up of six adsorbed molecules with an intermolecular distance of approximately 10 Å (Figure 7.22a). A (H43  H43)R7.6 structure was proposed as schematically shown in Figure 7.22b. 1,4-Cyclohexadiene adsorbs in a boat configuration on bridge or hollow sites on the Pt (1 1 1) surface with comparable adsorption energies of 145.6 and 141.6 kJ mol1, respectively [39]. 1,3-Cyclohexadiene and benzene form identical structures on Pt(1 1 1) at low pressures (Figures 7.23 and 7.24). 1,3-Cyclohexadiene dehydrogenates to form benzene on the surface, while benzene adsorbs molecularly. Figure 7.24b schematically shows the adsorbed benzene structure at low pressure. The STM images of the C6 cyclic hydrocarbons show three different adsorbed structures on Pt(1 1 1). Cyclohexene and cyclohexane partially dehydrogenate to form p-allyl, 1,4-cyclohexadiene adsorbs in a boat configuration, and both 1,3-cylohexadiene and benzene adsorb as molecular benzene on the surface. The hydrogenation of unsaturated C6 cyclic hydrocarbon molecules on Pt(1 1 1) was studied with STM at high pressure with the same approach as the hydrogenation of ethylene in Section 7.5.1. At 300 K, with 200 mTorr hydrogen and 20 mTorr cyclohexene, the hydrogenation reaction forms cyclohexane. The absence of an ordered structural feature of the reactant species in the STM images indicates a mobile adsorbed layer under high pressure. When 5 mTorr CO was then added to the chamber in addition to the 200 mTorr hydrogen and 20 mTorr cyclohexene already present, the reaction was poisoned and the Moire interference pattern indicative of high-pressure CO structure was observed. The experiment on Pt(1 1 1) with

j211

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

212

Figure 7.22 (a) 50 Å  50 Å STM image showing the structure formed by 1,4-cyclohexadiene on Pt(1 1 1) at 1  105 Torr and 300 K. Lines in the [1 1 0]-type directions of the underlying platinum lattice are drawn. The surface species form hexagonal units in domains containing a few unit cells and in

antiphase relationship with each other, as shown in the image on the right (dashed hexagons are extrapolated cells). The periodicity and rotation suggest a (H43  H43)R7.6 structure. (b) Diagram of the proposed (H43  H43)R7.6 1,4-cyclohexadiene structure. Inset shows bonding orientation.

200 mTorr hydrogen and 20 mTorr cyclohexene was performed at 353 K. The increase in temperature favors the dehydrogenation reaction pathway. Thus, both cyclohexane and benzene products were measured and the rate of cyclohexane formation was found to be twice that of the benzene. When 5 mTorr CO was introduced into the chamber, the reaction was poisoned but no ordered structure was observed in STM images. Since CO desorbs at 410 K in UHV [40] and the mobility of reactant species increases at elevated temperature, it is not surprising that the more weakly bound high-pressure structure becomes disordered at 353 K. Despite the mobile adsorbate layer, the reaction was poisoned, which indicates that the rate of vacancy generation was still not sufficient to activate the reaction. When the sample was subsequently cooled, the STM images showed an ordered CO structure on the surface. This study

7.5 Adsorbate Mobility During Catalytic Reactions

Figure 7.23 100 Å  100 Å image of Pt in the presence of 1  105 Torr 1,3-cyclohexadiene at 300 K. The surface structure, (2H3  2H3)R30.0 , is similar to that formed by benzene.

was able to identify the low-pressure adsorbate structure of all of the C6 hydrocarbon species and correlate the mobility of the adsorbed layer with the reactivity of the catalytic versus the poisoned surface. These systematic studies suggest that an intrinsic connection between the adsorbate structure, mobility, and the formation of product can be established with the aid of structural information obtained from high-pressure STM. It further demonstrated the importance of STM in studies of heterogeneous catalysis at high pressure. 7.5.3 CO/NO Coadsorption on Rh(1 1 1)

The coadsorption of CO and NO on Rh(1 1 1) was studied by high-pressure STM [41]. CO and NO conversion on rhodium catalysis is part of the catalytic converter reaction for automobile emissions and has been extensively studied. After cleaning the Rh(1 1 1) crystal with sputtering and annealing cycles, the sample was isolated in the highpressure cell, which was then filled with 500 mTorr CO. CO was imaged as a (2  2)-3 structure binding on Rh top sites. When 150 mTorr NO was introduced into the chamber, individual bright spots began to accumulate within the (2  2)-3 structure. The bright spots were observed with an apparent height 0.3 Å more than CO and were assigned to molecularly adsorbed NO. When the NO pressure was increased to 700 mTorr, adsorbed NO accounted for roughly one-quarter of the surface layer on the CO precovered surface at room temperature. By counting the mole fraction of top site NO as a function of the molar fraction of gas-phase NO, an Arrhenius relation was plotted yielding the relative binding energies of CO and NO on top sites [41]. CO was determined to bind 66  5 meV more strongly than NO at top sites. Figure 7.25 shows

j213

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

214

Figure 7.24 (a) 80 Å  80 Å STM image of Pt in the presence of 1.2  105 Torr benzene at 300 K. Species is molecular benzene. Lines in the [1 1 0]-type directions of the underlying

platinum lattice and a unit cell of the adsorbate structure have been drawn. (b) Diagram of proposed (2H3  2H3)R30.0 benzene structure. Inset shows bonding structure.

two consecutive STM images of the coadsorbed NO/CO structure. The images were taken 1 min apart, during which time a small number of rearrangement events occurred (two are marked by arrows). In the analysis of 10 images with approximately 400 top sites per image, 64 events were recorded. Fifty-six of these events were a single event of a top site NO adsorbate moving to an adjacent top site. The other events were adsorbates that hopped two top sites or the appearance or disappearance of a top site. Figure 7.25 includes a schematic for vacancy-mediated diffusion of a high-coverage overlayer. The study also shows that a phase transition occurred

7.5 Adsorbate Mobility During Catalytic Reactions

Figure 7.25 STM image: sequential (55 s delay) 200 Å  115 Å images of Rh(1 1 1) with 0.50 Torr CO and 0.92 Torr NO (I ¼ 260 pA, V ¼ 50 mV). The top arrow shows one top site occupied by NO in the top image and by CO in the bottom one. The bottom arrow shows the opposite.

through a phase boundary that had previously been observed during Torr pressure STM studies of NO adsorption [42]. The high-pressure structure has not yet been conclusively assigned [43–45]. The study of CO and NO conversion on Rh(1 1 1) shows that high-pressure STM can be used not only to study the adsorbate structure and its mobility but also to study the reaction kinetics of single-crystal model catalysis under realistic conditions.

j215

j 7 Surface Mobility of Atoms and Molecules Studied with High-Pressure STM

216

7.6 Summary

The possible pressure gap of adsorbates of reactant molecules and catalyst surface structure between low-pressure and realistic high-pressure industrial catalysis conditions drives the study of adsorption and surface structure at high pressures of reactant gases. One of the most valuable techniques is high-pressure high-temperature STM, with the capability of working at the entire pressure range, from UHV to several bars of reactant gases, and at variable temperature. This technique and instrumentation was presented on the basis of the recently developed high-pressure high-temperature STM reactor/UHV system at Berkeley. Using high-pressure STM, we revealed the pressure-dependent surface structure of several systems, including hex-Pt(1 0 0) under an environment of CO at 300 K. The flexibility of the surface structure and the mobility of the metal atoms of the catalyst surface were demonstrated at high pressure. Our high-pressure high-temperature STM studies of catalytic systems also revealed the mobility of reactant species at high pressure. We also demonstrated that high-pressure STM can be used to study reaction kinetics for single-crystal model catalysts.

Acknowledgments

This work was supported by the Director, Office of Science, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231.

References 1 Binnig, G. and Rohrer, H. (1983) Scanning Electron Microsc., 3, 1079–1082. 2 Mcintyre, B.J., Salmeron, M.B., and Somorjai, G.A. (1992) Catal. Lett., 14, 263–269. 3 Somorjai, G.A., Tao, F., and Park, J.Y. (2008) Top. Catal., 47, 1–14. 4 Westerstrom, R., Wang, J.G., Ackermann, M.D., Gustafson, J., Resta, A., Mikkelsen, A., Andersen, J.N., Lundgren, E., Balmes, O., Torrelles, X., Frenken, J.W.M., and Hammer, B. (2008) J. Phys.: Condens. Matter, 20, 184018. 5 Laegsgaard, E., Osterlund, L., Thostrup, P., Rasmussen, P.B., Stensgaard, I., and Besenbacher, F. (2001) Rev. Sci. Instrum., 72, 3537–3542. 6 Kolmakov, A. and Goodman, D.W. (2003) Rev. Sci. Instrum., 74, 2444–2450.

7 Rasmussen, P.B., Hendriksen, B.L.M., Zeijlemaker, H., Ficke, H.G., and Frenken, J.W.M. (1998) Rev. Sci. Instrum., 69, 3879–3884. 8 Rossler, M., Geng, P., and Wintterlin, J. (2005) Rev. Sci. Instrum., 76, 023705. 9 Vestergaard, E.K., Vang, R.T., Knudsen, J., Pedersen, T.M., An, T., Laegsgaard, E., Stensgaard, I., Hammer, B., and Besenbacher, F. (2005) Phys. Rev. Lett., 95, 126101. 10 Hendriksen, B.L.M., Bobaru, S.C., and Frenken, J.W.M. (2005) Catal. Today, 105, 234–243. 11 Westerstrom, R., Gustafson, J., Resta, A., Mikkelsen, A., Andersen, J.N., Lundgren, E., Seriani, N., Mittendorfer, F., Schmid, M., Klikovits, J., Varga, P., Ackermann, M.D., Frenken, J.W.M.,

References

12

13

14 15

16 17 18

19 20

21

22 23

24

25

26 27

28

29

Kasper, N., and Stierle, A. (2007) Phys. Rev. B, 76, 155410. Tao, F., Tang, D., Salmeron, M., and Somorjai, G.A. (2008) Rev. Sci. Instrum., 79, 084101. Somorjai, G.A. (1999) Introduction to Surface Chemistry and Catalysis, Wiley-VCH, Weinheim. Somorjai, G.A. and McCrea, K.R. (2000) Adv. Catal., 45, 385–438. Cremer, P.S., Su, X.C., Shen, Y.R., and Somorjai, G.A. (1996) J. Am. Chem. Soc., 118, 2942–2949. Shen, Y.R. (1989) Nature, 337, 519–525. Zhu, X.D., Suhr, H., and Shen, Y.R. (1987) Phys. Rev. B, 35, 3047–3050. Ogletree, D.F., Bluhm, H., Lebedev, G., Fadley, C.S., Hussain, Z., and Salmeron, M. (2002) Rev. Sci. Instrum., 73, 3872–3877. Salmeron, M. and Schlogl, R. (2008) Surf. Sci. Rep., 63, 169–199. Tao, F., Grass, M.E., Zhang, Y.W., Butcher, D.R., Renzas, J.R., Liu, Z., Chung, J.Y., Mun, B.S., Salmeron, M., and Somorjai, G.A. (2008) Science, 322, 932–934. Somorjai, G.A. (1981) Chemistry in Two Dimensions: Surfaces, Cornell University, Ithaca, NY. Somorjai, G.A. (1991) Langmuir, 7, 3176–3182. Mcintyre, B.J., Salmeron, M., and Somorjai, G.A. (1993) J. Vac. Sci. Technol. A, 11, 1964–1968. Mortensen, K., Klink, C., Jensen, F., Besenbacher, F., and Stensgaard, I. (1989) Surf. Sci., 220, L701–L708. Batteas, J.D., Dunphy, J.C., Somorjai, G.A., and Salmeron, M. (1996) Phys. Rev. Lett., 77, 534–537. Blakely, D.W. and Somorjai, G.A. (1977) Surf. Sci., 65, 419–442. Vang, R.T., Laegsgaard, E., and Besenbacher, F. (2007) Phys. Chem. Chem. Phys., 9, 3460–3469. Tang, D.C., Hwang, K.S., Salmeron, M., and Somorjai, G.A. (2004) J. Phys. Chem. B, 108, 13300–13306. Dunphy, J.C., Sautet, P., Ogletree, D.F., Dabbousi, O., and Salmeron, M.B. (1993) Phys. Rev. B, 47, 2320–2328.

30 Koeleman, B.J.J., Dezwart, S.T., Boers, A.L., Poelsema, B., and Verheij, L.K. (1986) Phys. Rev. Lett., 56, 1152–1155. 31 Steininger, H., Ibach, H., and Lehwald, S. (1982) Surf. Sci., 117, 685–698. 32 Longwitz, S.R., Schnadt, J., Vestergaard, E.K., Vang, R.T., Laegsgaard, E., Stensgaard, I., Brune, H., and Besenbacher, F. (2004) J. Phys. Chem. B, 108, 14497–14502. 33 Montano, M., Bratlie, K., Salmeron, M., and Somorjai, G.A. (2006) J. Am. Chem. Soc., 128, 13229–13234. 34 Chen, P., Westerberg, S., Kung, K.Y., Zhu, J., Grunes, J., and Somorjai, G.A. (2002) Appl. Catal. A – Gen., 229, 147–154. 35 Montano, M., Salmeron, M., and Somorjai, G.A. (2006) Surf. Sci., 600, 1809–1816. 36 Bratlie, K.M., Montano, M.O., Flores, L.D., Paajanen, M., and Somorjai, G.A. (2006) J. Am. Chem. Soc., 128, 12810–12816. 37 Yang, M.C., Chou, K.C., and Somorjai, G.A. (2004) J. Phys. Chem. B, 108, 14766–14779. 38 Koel, B.E., Blank, D.A., and Carter, E.A. (1998) J. Mol. Catal. A – Chem., 131, 39–53. 39 Saeys, M., Reyniers, M.F., Marin, G.B., and Neurock, M. (2002) Surf. Sci., 513, 315–327. 40 Ertl, G., Neumann, M., and Streit, K.M. (1977) Surf. Sci., 64, 393–410. 41 Rider, K.B., Hwang, K.S., Salmeron, M., and Somorjai, G.A. (2002) J. Am. Chem. Soc., 124, 5588–5593. 42 Rider, K.B., Hwang, K.S., Salmeron, M., and Somorjai, G.A. (2001) Phys. Rev. Lett., 86, 4330–4333. 43 Wallace, W.T., Cai, Y., Chen, M.S., and Goodman, D.W. (2006) J. Phys. Chem. B, 110, 6245–6249. 44 Popa, C., Flipse, C.F.J., Jansen, A.P.J., van Santen, R.A., and Sautet, P. (2006) Phys. Rev. B, 73, 245408. 45 Requejo, F.G., Hebenstreit, E.L.D., Ogletree, D.F., and Salmeron, M. (2004) J. Catal., 226, 83–87.

j217

j219

8 Point Defects on Rutile TiO2(1 1 0): Reactivity, Dynamics, and Tunability Chi L. Pang and Geoff Thornton

8.1 Introduction

Metal oxides are exploited in a number of technologies including gas sensing, microelectronics, and catalysis [1, 2]. A number of oxides directly act as catalysts while many others are employed as supports on which an active metal is dispersed. As there is evidence that the oxide support influences the reactivity of the dispersed metal [3, 4], understanding the behavior of oxides becomes important both when they are used as supports and when they are the active component in a catalyst. To date, rutile TiO2(1 1 0) is the most studied single-crystal TiO2 surface and its 1
1 termination is considered the prototypical metal oxide surface for fundamental studies [5]. There are a number of reasons why this surface has been studied so extensively, such as the ready availability of high-quality crystals, its thermodynamic stability, its significant electrical conductivity following vacuum annealing, and the relative simplicity of its geometric structure. The vast majority of the literature concerns work carried out in ultrahigh vacuum (UHV) that ensures clean, welldefined surfaces, although there are efforts directed toward working both in ambient pressures and in liquids [6–8]. The chemistry of both TiO2(1 1 0) and metal oxide surfaces in general has long been thought to be dominated by reactions involving defects [9, 10]. Despite this, even on the unreconstructed 1
1 phase of the model oxide – TiO2(1 1 0) – point defects in scanning probe microscopy (SPM) images have not been well understood until very recently. In this chapter, we explore how these point defects were unambiguously identified in SPM images, then review a series of studies made possible by the positive identification of point defects. All the work we describe in this chapter was carried out in UHV on the rutile TiO2(1 1 0)1
1 surface. Exposures to vapors and gases are given in Langmuirs (L) where 1 L ¼ 1.333
106 mbar s. Coverages of defects or molecules adsorbed at the surface will be given in monolayers (ML), where 1 ML corresponds to the density of primitive surface unit cells.

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

220

2

8.2 Methods

We use commercial TiO2 crystals (Pi-Kem) cut and polished to within 0.3 of the (1 1 0) face and we prepare them further with cycles of Ar þ bombardment and UHV annealing to approximately 950–1100 K, typically 5–10 min for each cycle. The samples are mounted onto tantalum back-plates via strips of tantalum spot-welded to the back-plate. Annealing is performed by high-energy electron bombardment of the back-plate from a hot filament. Temperatures are measured from optical pyrometers (Minolta) focused on the back-plate. The temperatures are not measured directly from the samples because they are translucent and get darker with each sputter/anneal cycle. Our laboratories are currently equipped with three UHV Omicron microscopes, a variable-temperature scanning tunneling microscope (STM), a room-temperature atomic force microscope (AFM)/STM, and a low-temperature liquid helium bath cryostat STM, all of which are currently driven by Omicron Scala software and electronics. All the STM results from our group presented in this chapter employed the variable temperature STM, with tips made by electrochemical etching of tungsten wire. For noncontact AFM (NC-AFM), we employ commercial conducting silicon cantilevers with force constants of approximately 2–14 N m1 and resonant frequencies of approximately 60–350 kHz (Nanosensors and Mikromasch). The NC-AFM images we present here were recorded in collaboration with Professor Onishi at Kobe University and employed a UHV JEOL (JSPM-4500A) microscope. To reduce the exposure to residual gas from the vacuum, our samples are sputtered/annealed in separate preparation chambers attached via valves to the SPM chambers. After the final anneal cycle of the cleaning procedure, the sample is transferred to the SPM chamber within a minute or so where the pressure is lower. Whenever possible, we perform experiments in the same part of the surface before and after some “operation.” The operation may involve exposure of the surface to a reactant such as water or O2, for example; or it may be some switch in the functionality of the microscope, for instance, imaging first in AFM mode and then in STM mode. For reactions, we simply stop the scan and retract the tip by up to 1000 nm, which ensures it does not crash when we admit gases or vapors to the chamber via the high-precision leak valves. We have also exploited “multichannel” imaging. As described in Section 8.3, we simultaneously recorded STM and NC-AFM images. Briefly, NC-AFM operates by oscillating a cantilever (which carries the tip) at its resonant frequency. As the tip approaches the surface, a force is felt by the cantilever and the resonant frequency changes, that is, there is a frequency shift (Df ). Typically, a topographic map is then constructed by scanning the surface at a constant Df. Simultaneous measurement of STM can be achieved simply by applying a bias between the tip and the sample during normal NC-AFM imaging and recording the tunneling current.

8.3 Water Dissociation at Oxygen Vacancies and the Identification of Point Defects

8.3 Water Dissociation at Oxygen Vacancies and the Identification of Point Defects

As discussed in the introduction, point defects are important on this surface as they are thought to dominate surface reactivity. As for water, TiO2 photocatalyzes its dissociation and it is this photoactivity that underpins the use of TiO2 in Green Technology such as self-cleaning surfaces, air/water purification devices, and novel solar cells. Apart from the intrinsic importance of water on TiO2, understanding water chemistry is also a crucial issue because water is present in the environment in almost all conceivable applications of TiO2. A model of TiO2(1 1 0) is shown in Figure 8.1. The key features are the fivefold coordinated Ti (Ti5c) rows that run in the [0 0 1] direction and alternate with bridging O (Ob) rows. In almost all published STM images of TiO2(1 1 0), tunneling is into empty substrate states (positive bias) because tunneling from the substrate tends to

Figure 8.1 Schematic depiction of water dissociation on TiO2(1 1 0). Blue and red spheres denote lattice O and Ti, respectively. The light blue spheres are Ob atoms, which lie in the [0 0 1] azimuth of the substrate. Parallel Ti rows that lie between the Ob rows are Ti5c atoms. Green spheres indicate O atoms bonded to H atoms that are shaded pink.

j221

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

222

2

Figure 8.2 175 Å
90 Å STM images of TiO2(1 1 0) exposed to 3.5 L H2O at 150 K, and consecutively heated to the measurement temperatures of (a) 195 K and (b) 290 K. (Modified with permission from Ref. [14].)

be unstable. Of these images, most show alternating bright and dark rows with bright spots (type-A defects) appearing between the bright rows, as shown in Figure 8.2b [5, 11–20]. Images that appear differently, such as those with dark depressions on bright rows [12, 21] are poorly understood and will not be discussed further here. Density functional theory (DFT) simulations of STM images indicate that the Ti5c atoms should appear brighter than the Ob atoms in STM images [11] despite the Ti5c atoms lying roughly 1.5 Å below the Ob atoms [22, 23]. This is due to the dominance of the Ti 3d states in the conduction band of TiO2. Although not explicitly simulated, Diebold et al. [11] argue that bridging oxygen vacancies (Ob-vacs) should appear bright and between bright Ti5c rows because the vacancies expose Ti sites. This was further supported by STM simulations of a missing row of Ob atoms that appear as a bright row. As such, the bright spots seen between bright Ti5c rows in STM images were assigned to Ob-vacs [11, 12]. In contrast, based both on STM images and on electron stimulated desorption experiments, Suzuki et al. [13] suggested that type-A defects originate from surface hydroxyl (OHb). The hydrogen was thought to originate from residual water vapor or dihydrogen in the vacuum chamber. Brookes et al. [14] gave support to Suzuki et al.’s interpretation. STM images were recorded at low temperature (200 K), then warmed to room temperature (290 K) to follow the dissociation of water. At low temperature, bright spots are seen on bright

8.3 Water Dissociation at Oxygen Vacancies and the Identification of Point Defects

Ti5c rows whereas at room temperature, bright spots appear between bright Ti5c rows (see Figure 8.2). At low temperature, the bright spots were assigned to molecular water adsorbed on Ti5c sites. This was corroborated in subsequent scanned energy mode photoelectron diffraction (PhD) measurements, which show that water bonds to Ti5c sites via its oxygen atom [24]. As for the bright spots seen in STM images at room temperature, Brookes et al. [14] assigned these to OHb resulting from water dissociation, on the basis that they have a similar concentration to those observed by Suzuki et al. [13] on sputtered/annealed surfaces. Although Brookes et al. show that OHb do indeed show up as bright spots between the bright Ti5c rows, no indication was given as to the appearance of Ob-vacs in STM images. Schaub et al. [15] resolved this issue, in part, by showing that two kinds of type-A defects coexist (see Figure 8.3). The brighter (Ab) defects were assigned to Obvacs and the darker (Ad) defects to OHb. Schaub et al. gave two reasons for their assignment: (i) their DFTsimulations of STM images suggest that Ob-vacs should be brighter than OHb and (ii) the number of type-A defects after exposing TiO2(1 1 0) to water was counted at twice that of the Ab defects before exposure. This ratio is in line with the expectation from other experimental work [14, 25–27], which suggested that water dissociates in Ob-vacs forming two OHb species. STM experiments carried out over the last few years, however, show that the assignment by Schaub et al. [15] is incorrect. One of the keys to correctly identifying these defects was hinted at in studies in which the STM tip was scanned across the TiO2(1 1 0) surface at a raised bias of þ 3 V. Type-A defects were removed during these high bias scans, the process being interpreted as tip-induced healing of Ob-vacs by Diebold et al. [12] and as tip-induced removal of hydrogen from OHb by Suzuki et al. [13]. Bikondoa et al. [16, 17] applied electrical pulses (approximately þ 3 V) to individual type-A defects to remove them one by one. Although this is almost always successful for Ab defects (an example is shown in Figure 8.3), Ad defects could never be removed in this way. As such, this was used as a way of distinguishing the two types of defects.

Figure 8.3 Sequential (150 Å)2 STM images of sputter/annealed TiO2(1 1 0). The densities of Ab and Ad defects are approximately 1.5 and 3.0% ML, respectively. (a) Before the voltage pulse and (b) after the voltage pulse. The voltage

pulse ( þ 3 V, 0.35 nA, 300 ms) was applied to the Ab defect circled in (a). In (b), Ab, and Ad defects are marked with red and blue crosses, respectively, and the pulsed Ab defect is missing. (Modified with permission from Ref. [16].)

j223

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

224

2

An Ab defect-free region (the “clean-off area”) was created by scanning across the surface at þ 3 V before exposing to water, as shown in Figure 8.4. According to the assignment of Schaub et al. [15] exposure to water should have little effect in the cleanoff area as there would be no Ob-vacs with which to react. However, exposing the surface to water in fact replenishes the Ab defects in the clean-off area and reduces the number of Ad defects (Figure 8.4c). If water dissociates in Ob-vacs then the number of Ob-vacs is expected to decrease and the number of OHb species to increase, thus Ab defects were reassigned as OHb and Ad defects reassigned as Ob-vacs. This also means that approximately þ 3 V scans and tip pulses remove hydrogen from OHb. While the mechanism for this hydrogen desorption has not yet been elucidated, it may be related to an empty state identified by Onda et al. [28] approximately 2.4 eV above EF. Returning to Figure 8.4 because the STM images were recorded from the same area before and after exposure to water, it can be seen that some of the new OHb take the positions of the reacting Ob-vacs. Thus, Bikondoa et al. [16, 17] imaged water dissociating in the Ob-vacs forming one OHb in place of the Ob-vacs and another OHb elsewhere, consistent with the mechanism depicted in Figure 8.1. STM measurements at low temperature show that water dissociates at least down to approximately 187 K [19], a conclusion confirmed by high-resolution electron energy loss spectroscopy [29]. Surprisingly, the number of OHb in positions previously taken by Ob-vacs always appears slightly higher than elsewhere. According to the mechanism in Figure 8.1, an equal number of OHb species should appear at the Ob-vacs and at a distance from them. Wendt et al. [18, 19] suggest that when the water molecules dissociate, the two resulting OHb lie initially in pairs, adjacent in the [0 0 1] direction. As individual OHb from the pair are difficult to resolve with STM, each OHb pair therefore appears as a single feature, which would account for the apparent anomaly. This interpretation is supported by the observation of three apparent sizes of type-A defects in some STM images [17–20, 30]. The largest of these is assigned to OHb pairs, the next largest to isolated OHb, and the smallest to Ob-vacs, an assignment corroborated in recent STM simulations of Ob-vacs and OHb [30, 31]. Further evidence for the OHb pairs is given in an STM movie recorded at approximately 187 K. The movie shows the OHb pairs separating across the Ob rows via proton exchange with water molecules [19]. Key frames from the movie are shown in Figure 8.5. This water-assisted OHb diffusion mechanism is supported by calculations, which show that the barrier to diffusion is lowered by the exchange with water. The same diffusion mechanism is also observed for isolated OHb and because of the misassignment of OHb and Ob-vacs by Schaub et al. [15], this watermediated diffusion of OHb was incorrectly reported as oxygen-mediated diffusion of Ob-vacs [32, 33] via a mechanism inconsistent with subsequent isotope studies [34]. Figure 8.6 summarizes our current knowledge of the appearance of point defects in STM images. The most prevalent point defects on sputtered/annealed TiO2(1 1 0) 1
1 surfaces have been identified as Ob-vacs, OHb, and OHb pairs and these are shown in a ball model together with an STM image decorated with a number of all three types of defects.

8.3 Water Dissociation at Oxygen Vacancies and the Identification of Point Defects

Figure 8.4 Sequential (285 Å
250 Å) STM images of TiO2(1 1 0). (a) Sputtered/annealed TiO2(1 1 0) with approximately 3.5 and 5.5% ML Ab and Ad defects, respectively. The blue crosses denote Ad defects that are also present in (b), the red circles denote Ab defects, and the yellow circles denote Ab defects that are removed to form the image in (b). (b) Following a þ 3 V scan. The yellow lines indicate the approximate boundaries beneath which the þ 3 V scan was applied. Blue crosses and red circles denote Ad

and Ab defects, respectively. (c) Following exposure to 0.1 L water. Blue crosses and filled red circles, respectively, denote Ad and Ab defects that were present in (b). Open red circles indicate positions where Ab defects were present in (b) but not in (c). Red crosses denote new Ab species that reside where Ad defects were positioned in (b) and black crosses denote new Ab species appearing elsewhere. The images are duplicated for clarity in (d), (e), and (f). (Modified with permission from Ref. [16].)

j225

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

226

2

Figure 8.5 Sequential images of TiO2(1 1 0) showing the splitting of an OHb pair mediated by a water molecule at 187 K. OHb groups are labeled with white circles and water on Ti5c sites is labeled with black squares. Solid grid lines intersect the Ob sites whereas dashed gridlines are along Ti5c rows. (Modified with permission from Ref. [19].)

The unambiguous assignment of OHb and Ob-vacs in STM images also paved the way for the same point defects to be identified in NC-AFM images. Most NC-AFM images of TiO2(1 1 0) show bright rows alternating with dark rows and with dark depressions appearing on the bright rows. Early interpretations of the NC-AFM contrast were based on geometrical considerations so that the bright rows were assigned to Ob rows and the depressions to Ob-vacs [35]. The assignment of the bright rows to Ob rows was backed up by adsorbing a probe molecule (formic acid) [36] and by checking the alignment of the bright rows with coexisting strands of the 1
2 reconstruction [37]. Pang et al. exploited information accessible from STM by scanning an area with NC-AFM then rescanning the same area with STM within minutes [38]. The experiment was revisited by simultaneously recording STM and NC-AFM images, an example being shown in Figure 8.7 (Pang et al., unpublished work). In both experiments, the positions of OHb and Ob-vacs in the STM images correlate with dark depressions in the NC-AFM images, thus showing that both OHb and Ob-vacs give rise to dark depressions in NC-AFM. The positions of these OHb species and Ob-vacs indicate that the bright rows in the NC-AFM image must correspond to Ob rows.

8.3 Water Dissociation at Oxygen Vacancies and the Identification of Point Defects

Figure 8.6 (a) Schematic models of the three most prevalent types of point defects on sputtered/annealed TiO2(1 1 0) surfaces. (b) (120 Å)2 perspective STM image showing the same defects. The image is color contoured so that Ti5c rows appear green, Ob rows appear blue, Ob-vacs appear green, OHb yellow, and OHb pairs appear red.

In the true surface topography, OHb are protrusions and Ob-vacs are depressions. As such, the appearance of both OHb and Ob-vacs as depressions in the NC-AFM image is significant because it means that NC-AFM images do not reproduce the true surface topography, as is often assumed. In further support of this, Pang et al. recorded images in the same area with different contrasts, between which there was presumably some change to the tip (Figure 8.8) [38]. In Figure 8.8a, type-A defects appear as dark spots in the bright Ob rows as they do in Figure 8.7b (Contrast I). However, in Figure 8.8b, the same type-A defects appear as bright spots between bright rows, which means the bright rows must be assigned to Ti5c rows (Contrast II). Similar contrast inversions were reported by Lauritsen et al. [39] who also reproduced the tip change in simulations. The tip change was modeled as a switch between a tip with a positive potential and one with a negative potential. The positive tip gives images with Contrast I while the negative tip produces images with Contrast II.

j227

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

228

2

Figure 8.7 Simultaneously recorded 80 Å
150 Å STM and NC-AFM images of TiO2(1 1 0). (a) A current map represents the STM image. Because the tip is oscillating during NC-AFM measurements, the timeaveraged current is recorded. The current map is color contoured so that OHb appear as broad, red spots and Ob-vacs appear as

narrow, yellow spots. Arrowheads point at three Ob-vacs and circles are drawn over three OHb, respectively. (b) The simultaneously recorded NC-AFM image with the arrowheads superimposed in blue and the circles superimposed in red, both indicate dark depressions. (From Pang et al., unpublished work.)

A third contrast regime, which appears more rarely, has also been reported in which bright rows are imaged with bright spots on them [40, 41]. Fukui and Iwasawa [40] assigned the bright rows to Ob rows and the bright spots to OHb based on geometrical considerations. This assignment was supported by simulations that employed a charge-neutral tip [41].

Figure 8.8 115 Å
185 Å NC-AFM images of TiO2(1 1 0). (a) Before the tip change. (b) After the tip change. Purple arrowheads indicate coincident type-A defects. (Modified with permission from Ref. [38].)

8.5 Alcohol Dissociation at Oxygen Vacancies

8.4 O2 Dissociation at Oxygen Vacancies

Like water, the interaction of O2 with TiO2 also has implications for photocatalysis. As with water again, reactions of O2 with TiO2 are also important because O2 will form part of the environment in many TiO2 applications. It has been known for some time that the spectroscopic signature of Ob-vacs can be healed by exposure to O2 [42–46]. In addition, Epling et al. [47] show that temperatureprogrammed desorption (TPD) spectra of water and ammonia are perturbed when the surface is predosed with O2. This implies that oxygen is left on the surface in some form when Ob-vacs are healed by O2, As such, Epling et al. proposed that one Ob-vac is healed per O2 molecule with the other O atom being adsorbed at a Ti5c site (Oad), a dissociation mechanism supported by theoretical calculations [48, 49]. Bikondoa et al. followed the dissociation of O2 using STM [16, 17]. Any convolution of the O2 reaction with Ob-vacs and that with OHb was avoided by removing OHb from an area of the surface by scanning at þ 3 V. The same area was then imaged before and after exposure to approximately 0.6 L O2 at room temperature, as shown in Figure 8.9. Following O2 exposure, the four Ob-vacs highlighted in Figure 8.9a are healed and four bright features appear nearby on the bright Ti5c rows. This suggests that O2 dissociates in the Ob-vacs, one O atom filling the vacancy and the other forming Oad on a nearby Ti5c site. Wendt et al. [18] performed similar experiments at low temperature, showing that O2 dissociates in Ob-vacs at least down to 120 K. Detailed analysis of the Oad positions following dissociation at room temperature shows that most Oad are found one lattice constant (in the [0 0 1] direction) away from the Ob-vac that is filled, the rest being immediately adjacent and two lattice constants away [50]. As there is little thermal diffusion of Oad on TiO2(1 1 0), the separation of Oad from the closest positions to the reacting Ob-vacs was attributed to the energy released during the exothermic dissociation of O2 (calculated at 3.5 eV [18]) in a similar way to that observed, for example, for Cl2 dissociation on TiO2(1 1 0) [51].

8.5 Alcohol Dissociation at Oxygen Vacancies

The reactivity of TiO2 with alcohols is important in a number of technological applications. For example, alcohols can be used as energy carriers in renewable sources and they are also employed as model pollutants so that environmental cleaning strategies can be tested. TPD and static secondary ion mass spectrometry (SSIMS) data suggest that methanol dissociatively adsorbs at Ob-vacs and molecularly at the Ti5c sites [52, 53]. There is also some evidence that methanol also dissociates at other sites apart from Ob-vacs, presumably Ti5c sites [53–55]. Similar conclusions have been reached for a series of short-chain (C2–C8) aliphatic alcohols [56–58].

j229

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

230

2

Figure 8.9 Reaction of O2 with Ob-vacs on TiO2(1 1 0). (a) 100 Å
120 Å STM image of TiO2(1 1 0) following a þ 3 V scan to remove OHb. The Ob-vacs are marked with blue crosses except for four Ob-vacs that do not appear in (b) that are instead marked with purple crosses. (b) STM image with the same size and scan parameters as (a), following exposure to 0.6 L O2. Ob-vacs that coincide with those in (a) are

indicated with blue crosses. Bright spots that appear on the bright rows, close to the position of the Ob-vacs in (a) are indicated with yellow crosses. The images are duplicated for clarity in (c) and (d). (e) Schematic representation of the reaction. Oxygen originating from the exposed O2 is colored yellow. The Oad is positioned diagonally adjacent to the Ob-vac after Du et al. [50]. (Modified with permission from Ref. [16].)

Figure 8.10b shows an STM image of the TiO2(1 1 0) surface after a low exposure to methanol at approximately 300 K [59]. Bright spots are found in place of Ob-vacs. These bright spots were attributed to pairs of methoxy/OHb, by analogy with the OHb pairs formed from water dissociation [18–20]. After dosing more methanol, further bright spots appear together with a number of darker spots (Figure 8.10c). The darker spots are assigned to OHb that have separated from the methoxy/OHb pairs across the Ob rows via proton exchange with molecularly adsorbed methanol, through essentially the same mechanism as described for water in Section 8.3.

8.5 Alcohol Dissociation at Oxygen Vacancies

Figure 8.10 STM images of the same area before and after exposure of methanol on TiO2(1 1 0) at 300 K. The dosing pressure is constant in all images so the dose time is proportional to the exposure. (a) Before exposure to methanol. (b) After 80 s exposure to methanol. (c) After 110 s exposure to

methanol. (d) The same area as (c) following a þ 3 V scan. Yellow circles show the position of Ob-vacs, blue circles show the bright features on Ob-vacs, red squares mark the darker spots. Green arrows point at bright spots on the bright Ti5c rows. (Modified with permission from Ref. [59].)

As it is known that high tip bias (approximately þ 3 V) can remove OHb [16–18], Zhang et al. performed a high bias scan in order to distinguish the dark and bright spots. The darker spots were removed but the brighter spots remained, thus confirming that the darker species are OHb and the brighter spots therefore methoxy. As the methoxy groups take the positions of the Ob-vacs, this indicates that the CH3OH bond is broken rather than the CH3OH bond. A similar series of STM experiments were performed for 2-butanol with almost identical results, evidence again being given for ROH cleavage at Ob-vacs [60]. In addition to this, for both methanol and 2-butanol, bright spots appeared on the bright Ti5c sites. Zhang et al. did not discuss these features in the case of methanol [59] but it seems likely that the bright spots arise from either methanol adsorbed at Ti5c sites or methoxy dissociated at Ti5c sites. For 2-butanol, one of the bright spots was seen

j231

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

232

2

Figure 8.11 Sequential STM images showing a 2-butanol molecule initially adsorbed on a Ti5c site dissociating at an Ob-vac. (a) Before dissociation, the 2-butanol molecule is indicated with a green arrowhead. (b) After the reaction, the 2-butoxy takes the position

of the Ob-vac (red arrowhead) and the OHb sits adjacent (red cross). (c) Schematic representation of the dissociation process. (Reprinted with permission from Ref. [60]. Copyright 2007, American Chemical Society.)

filling an Ob-vac in sequential STM images (Figure 8.11). This sequence was therefore interpreted as a 2-butanol molecule initially adsorbed at a Ti5c site eventually dissociating in an Ob-vac. As the adsorption behavior of ROH on TiO2(1 1 0) is similar, where R ¼ H, CH3, or 2-butyl, Zhang et al. [60] suggest that all alcohols may behave as follows: (i) the ROH bond breaks at the Ob-vac sites with RO filling the Ob-vac and (ii) ROH adsorbs molecularly at Ti5c sites and facilitates diffusion of OHb formed from dissociation at Ob-vacs. 8.6 Diffusion of Oxygen Vacancies and Surface Hydroxy

The surface diffusion of defects and adsorbates is of obvious importance in heterogeneous catalysis, as this process brings the reactants together. Understanding the dynamics of molecules on oxide surfaces is also a key step toward the realization of working molecular electronics. We note here that diffusion of Ob-vacs really means diffusion of Ob into the vacancy, which leaves another Ob-vac in the position vacated by the Ob. Similarly, diffusion of OHb occurs by diffusion of the H atom. Zhang et al. [61] recorded STM images between 350 and 423 K to investigate the diffusion of Ob-vacs. Figure 8.12 shows two sequential STM images taken at 400 K, with a difference image being shown in Figure 8.12c. The difference image clearly shows that the Ob-vacs diffuse along the Ob rows. Diffusion was never observed across the rows in the temperature range investigated. Hopping rates for Ob-vacs along the rows were determined at seven temperatures so that an Arrhenius plot could be performed. This yielded a diffusion barrier of Eb ¼ 1.15 eV that closely matches the value given by their DFT calculations (Eb ¼ 1.03 eV) [61]. Calculations also show that it is energetically unfavorable for Ob-vacs to be positioned adjacent to or close to another Ob-vac along the Ob row [48, 61, 62]. This is reflected in the spatial distribution of Ob-vacs in STM images. If Ob-vacs are positioned randomly over the surface, statistically a 10% density of vacancy sites should lead to 10% of the Ob-vacs existing as pairs, yet Ob-vac pairs have not been reported [61].

8.6 Diffusion of Oxygen Vacancies and Surface Hydroxy

Figure 8.12 64 Å
69 Å STM images of a TiO2(1 1 0) surface recorded at 400 K with Obvacs present. (a) and (b) are sequential images recorded 2 min apart. The Ti5c rows appear red and the Ob rows appear blue. A schematic model of the surface is shown to scale above parts of the image in (a) and (b). Ti atoms are shown red, and oxygen blue with the Ob rows

shown a lighter blue. (c) Difference image, the image in (b) being subtracted from that in (a). Yellow protrusions indicate the original positions of Ob-vacs whereas blue depressions indicate their positions in (b). (Reprinted with permission from Ref. [61]. Copyright 2007, American Physical Society.)

As for OHb diffusion, Zhang et al. [20, 63] also show that diffusion occurs along the [0 0 1] direction, distinct from the [1
1 0] direction water-assisted diffusion reported by Wendt et al. [19]. This is demonstrated in Figure 8.13. Furthermore, by carefully analyzing the positions of the OHb it was shown that the initial hop away from OHb pairs is usually taken by the OHb formed from the hydrogen, which splits off from the OH fragment that fills the Ob-vac, a preference that holds at least from 300 to 372 K. Therefore, in contrast to previous expectations, the two OHb that form from dissociation of a water molecule appear to be inequivalent. Zhang et al. speculate that this is due to a different charge distribution around each OHb; the Ob-vac involves two nominally Ti3 þ ions so that when water dissociates, the OH fragment

Figure 8.13 STM images of a TiO2(1 1 0) surface recorded at 381 K with OHb present. (a) and (b) are sequential images recorded 3 min apart. The white arrows in (a) mark the positions of two remaining Ob-vacs. (c) Difference image, in which (a) is subtracted from (b). The dark

spots represent the initial hydrogen positions in (a) whereas the bright spots show their final positions in (b). The black arrows show the hopping directions. (Reprinted with permission from Ref. [63]. Copyright 2008, American Chemical Society.)

j233

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

234

2

that sits in the vacancy is connected to two Ti3 þ ions. In contrast, the adjacent OH group formed by the hydrogen that splits from HOH is bound to only one Ti3 þ ion and a Ti4 þ ion. It is important to note that this is a simplified description of the charge distribution, since scanned angle PhD measurements clearly show that the charge distribution around Ob-vacs is rather complex [64] and one would expect a similar situation for OHb. Arrhenius analysis of OHb diffusion in the temperature range 300–410 K, for already separated OHb, gives Eb ¼ 0.75 eV [63]. DFT calculations predict a much higher barrier of 1.29 eV. Even when considering an intermediate relay point whereby the OHb diffuses to an in-plane O atom before the next Ob site [65], Eb is reduced only to 1.19 eV [63]. This is still much higher than the experimental value. In light of this large discrepancy, Li et al. tentatively propose a two-stage diffusion mechanism whereby electronic charge associated with the OHb diffuses first followed by diffusion of the OHb itself.

8.7 Tuning the Densities of Oxygen Vacancies and Surface Hydroxyl on TiO2(1 1 0)

The engineering of surfaces is an idea that has come to the fore recently. This is related to the drive to build nanodevices, but such especially prepared surfaces are also valuable in providing controlled surfaces for reactivity tests, as described in Sections 8.3 and 8.4. Experience shows that the density of Ob-vacs can be varied by changing the anneal temperature or the anneal time [66]. Wendt et al. [18] assessed this systematically by demonstrating a linear dependence of the Ob-vac density (measured by STM) on the “sample history,” the latter being defined as the anneal temperature
anneal time. The defect density can be controlled in a more flexible manner by the use of electron bombardment. For some time [44, 45, 67], this method has been used to introduce Ob-vacs at the surface. Until recently, however, evidence for Ob-vac formation has come only from the spectroscopic signature. Pang et al. used STM to assess the atomic-scale surface structure of TiO2(1 1 0) following electron bombardment [17]. Figure 8.14a shows an image of the TiO2(1 1 0) surface following electron bombardment at 75 eV with an emission current of approximately 0.1 mA (0.03 mA cm2) and a duration of 7 s. The image appears similar to those recorded from sputter/annealed TiO2(1 1 0) surfaces but with a rather high Ob-vac concentration of approximately 11% ML and a low OHb concentration of approximately 1.5% ML. Figure 8.14b contains an image of the surface measured prior to electron bombardment that shows the surface was covered with approximately 12% ML OHb and only approximately 0.4% ML Ob-vacs. This clearly shows that electron bombardment does indeed introduce Ob-vacs. It is equally clear that the OHb density is greatly reduced by the electron bombardment. Electron bombardment for longer durations and higher energies led to more severe surface modifications, such as surface pitting and disordered 1
n reconstructed areas [17]. This means that the introduction of Ob-vacs by electron bom-

8.7 Tuning the Densities of Oxygen Vacancies and Surface Hydroxyl on TiO2(1 1 0)

Figure 8.14 (a) (345 Å)2 STM image of TiO2(1 1 0) following electron bombardment at 75 eV and 0.1 mA (0.03 mA cm2) for 7 s. The surface contains 11% Ob-vacs and 1.5% OHb. Some OHb and Ob-vac defects are indicated with red and blue crosses, respectively. The circles surround electron bombardment-induced pits. (b) (220 Å
140 Å)

STM image taken prior to electron bombardment. The surface contains 12% ML OHb and 0.4% ML Ob-vacs. The image is color contoured so that OHb appear yellow, OHb pairs appear red, and Ob-vacs appear green. Squares are drawn around some of the Ob-vacs. (Modified with permission from Ref. [17].)

bardment has to be carefully controlled to avoid the formation of these other structures. While electron bombardment can be used to introduce Ob-vacs, there is also considerable interest in preparing pseudoperfect surfaces where the only defects are step edges. Such pseudoperfect surfaces allow processes on “normal” surface sites to be separated from the often more reactive defect sites.

j235

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

236

2

In Section 8.3, we already hint at how pseudoperfect surfaces can be created locally. By scanning a surface at approximately þ 3 V, hydrogen from OHb can be desorbed leaving Ob-vacs intact but, of course, if the Ob-vacs are already removed by exposure to water, then a pseudoperfect part of the surface can be created. Gan et al. [68] suggest a way to make pseudoperfect surfaces not restricted to local areas. They annealed their sample first in an oxygen plasma at 1073 K, then in UHV at 773 K. High-resolution STM images show no evidence of point defects, although this in itself cannot necessarily be taken as evidence for a pseudoperfect surface. This is because point defects often become “invisible” in STM images depending on the nature of the tip. Valence band photoelectron spectroscopy would be useful here because if the surface is devoid of Ob-vacs, no bandgap states should be detected.

8.8 Outlook

Over the past decade or so, our understanding of oxide surfaces has markedly improved, especially in the case of the model oxide surface – TiO2(1 1 0). In particular, the influence of defects such as oxygen vacancies on the surface reactivity has been demonstrated in exquisite detail and it is clear that they play a crucial role in many surface reactions. As we have shown in this chapter, scanning probes have been instrumental in evaluating the surface reactivity as they allow individual defects, reactants, and products to be imaged on relatively short timescales (seconds or minutes). High-resolution STM and NC-AFM images have appeared both at elevated pressures and in liquids, and one would expect a great deal of attention to be focused in this direction over the coming years. An area of particular topical interest is the interface between light harvesting surfaces (e.g., TiO2) and liquid water in connection with photocatalysis. Despite the enormous impact that scanning probe methods have had on our understanding of reactions at oxide surfaces, both STM and AFM suffer from the lack of chemical specificity. The application of STM–inelastic electron tunneling spectroscopy is a potential solution as it can be used to measure the vibrational spectrum of individual molecules at the surface [69, 70].

References 1 Fujishima, A. and Honda, K. (1972) Nature, 238, 37. 2 Gr€atzel, M. (2001) Nature, 414, 338. 3 Yoon, B., H€akkinen, H., Landman, U., W€ orz, A.S., Antonietti, J.-M., Abbet, S., Judai, K., and Heiz, U. (2005) Science, 307, 403.

4 Tauster, S.J. (1987) Acc. Chem. Res., 20, 389. 5 Diebold, U. (2003) Surf. Sci. Rep., 48, 53. 6 Komiyama, M. and Gu, M. (1997) Appl. Surf. Sci., 120, 125. 7 Sasahara, A., Kitamura, S., Uetsuka, H., and Onishi, H. (2004) J. Phys. Chem. B, 108, 15735.

References 8 Kolmakov, A. and Goodman, D.W. (2003) Rev. Sci. Instrum., 74, 2444. 9 Henrich, V.E. and Cox, P.A. (1994) The Surface Science of Metal Oxides, Cambridge University Press, Cambridge. 10 Freund, H.-J. (1999) Faraday Discuss., 114, 1. 11 Diebold, U., Anderson, J.F., Ng, K.-O., and Vanderbilt, D. (1996) Phys. Rev. Lett., 77, 1322. 12 Diebold, U., Lehman, J., Mahmoud, T., Kuhn, M., Leonardelli, G., Hebenstreit, W., Schmid, M., and Varga, P. (1998) Surf. Sci., 411, 137. 13 Suzuki, S., Fukui, K., Onishi, H., and Iwasawa, Y. (2000) Phys. Rev. Lett., 84, 2156. 14 Brookes, I.M., Muryn, C.A., and Thornton, G. (2001) Phys, Rev. Lett., 87, 266103. 15 Schaub, R., Thostrup, P., Lopez, N., Laegsgaard, E., Stensgaard, I., Nørskov, J.K., and Besenbacher, F. (2001) Phys. Rev. Lett., 87, 266104. 16 Bikondoa, O., Pang, C.L., Ithnin, R., Muryn, C.A., Onishi, H., and Thornton, G. (2006) Nat. Mater., 5, 189. 17 Pang, C.L., Bikondoa, O., Humphrey, D.S., Papageorgiou, A.C., Cabailh, G., Ithnin, R., Chen, Q., Muryn, C.A., Onishi, H., and Thornton, G. (2006) Nanotechnology, 17, 5397. 18 Wendt, S., Schaub, R., Matthiesen, J., Vestergaard, E.K., Wahlstr€om, E., Rasmussen, M.D., Thostrup, P., Molina, L.M., Lægsgaard, E., Stensgaard, I., Hammer, B., and Besenbacher, F. (2005) Surf. Sci., 598, 226. 19 Wendt, S., Matthiesen, J., Schaub, R., Vestergaard, E.K., Lægsgaard, E., Besenbacher, F., and Hammer, B. (2006) Phys. Rev. Lett., 96, 066107. 20 Zhang, Z., Bondarchuk, O., Kay, B.D., White, J.M., and Dohnalek, Z. (2006) J. Phys. Chem. B, 110, 21840. 21 Fischer, S., Munz, A.W., Schierbaum, K.D., and G€ opel, W. (1995) Surf. Sci., 337, 17. 22 Lindsay, R., Wander, A., Ernst, A., Montanari, B., Thornton, G., and Harrison, N.M. (2005) Phys. Rev. Lett., 94, 246102.

23 Cabailh, G., Torrelles, X., Lindsay, R., Bikondoa, O., Joumard, I., Zegenhagen, J., and Thornton, G. (2007) Phys. Rev. B, 75, 241403(R). 24 Allegretti, F., O’Brien, S., Polcik, M., Sayago, D.I., and Woodruff, D.P. (2005) Phys. Rev. Lett., 95, 226104. 25 Hugenschmidt, M.B., Gamble, L., and Campbell, C.T. (1994) Surf. Sci., 302, 329. 26 Henderson, M.A. (1998) Surf. Sci., 400, 203. 27 Henderson, M.A. (2002) Surf. Sci. Rep., 46, 1. 28 Onda, K., Li, B., Zhao, J., Jordan, K.D., Yang, J., and Petek, H. (2005) Science, 308, 1154. 29 Henderson, M.A., Epling, W.S., Peden, C.H.F., and Perkins, C.L. (2003) J. Phys. Chem. B, 107, 534. 30 Teobaldi, G., Hofer, W.A., Bikondoa, O., Pang, C.L., Cabailh, G., and Thornton, G. (2007) Chem. Phys. Lett., 437, 73. 31 Di Valentin, C. (2007) J. Chem. Phys., 127, 154705. 32 Schaub, R., Thostrup, P., Lopez, N., Laegsgaard, E., Stensgaard, I., Nørskov, J.K., and Besenbacher, F. (2001) Phys. Rev. Lett., 87, 266104. 33 Schaub, R. (2006) Science, 314, 5801. 34 Thompson, T.L., Diwald, O., and Yates, J.T., Jr. (2004) Chem. Phys. Lett., 393, 28. 35 Fukui, K., Onishi, H., and Iwasawa, Y. (1997) Phys. Rev. Lett., 79, 4202. 36 Fukui, K., Onishi, H., and Iwasawa, Y. (1997) Chem. Phys. Lett., 280, 296. 37 Pang, C.L., Raza, H., Haycock, S.A., and Thornton, G. (2000) Appl. Surf. Sci., 157, 233. 38 Pang, C.L., Sasahara, A., Onishi, H., Chen, Q., and Thornton, G. (2006) Phys. Rev. B, 74, 073411. 39 Lauritsen, J.V., Foster, A.S., Olesen, G.H., Christensen, M.C., K€ uhnle, A., Helveg, S., Rostrup-Nielsen, J.R., Clausen, B.S., Reichling, M., and Besenbacher, F. (2006) Nanotechnology, 17, 3436. 40 Fukui, K. and Iwasawa, Y. (2002) Noncontact Atomic Force Microscopy (eds S. Morita, E. Meyer, and R. Wiesendanger), Springer, Berlin.

j237

j 8 Point Defects on Rutile TiO (1 1 0): Reactivity, Dynamics, and Tunability

238

2

41 Enevoldsen, G.H., Foster, A.S., Christensen, M.C., Lauritsen, J.V., and Besenbacher, F. (2007) Phys. Rev. B, 76, 205415. 42 Kurtz, R.L., Stockbauer, R., Madey, T.E., Roman, E., and De Segovia, J.L. (1989) Surf. Sci., 218, 178. 43 Pan, J.M., Maschhoff, B.L., Diebold, U., and Madey, T.E. (1992) J. Vac. Sci. Technol. A, 10, 2470. 44 Wang, L.Q., Shultz, A.N., Baer, D.R., and Engelhard, M.H. (1996) J. Vac. Sci. Technol. A, 14, 1532. 45 Onda, K., Li, B., and Petek, H. (2004) Phys. Rev. B, 70, 045415. 46 Perkins, C.L. and Henderson, M.A. (2001) J. Phys. Chem. B, 105, 3856. 47 Epling, W.S., Peden, C.H.F., Henderson, M.A., and Diebold, U. (1998) Surf. Sci., 412/413, 333. 48 Rasmussen, M.D., Molina, L.M., and Hammer, B. (2004) J. Chem. Phys., 120, 988. 49 Wu, X., Selloni, A., Lazzeri, M., and Nayak, S.K. (2003) Phys. Rev. B, 68, 241402(R). 50 Du, Y., Dohnalek, Z., and Lyubinetsky, I. (2008) J. Phys. Chem. B, 112, 2649. 51 Diebold, U., Hebenstreit, W., Leonardelli, G., Schmid, M., and Varga, P. (1998) Phys. Rev. Lett., 81, 405. 52 Henderson, M.A., Otero-Tapia, S., and Castro, M.E. (1999) Faraday Discuss., 114, 313. 53 Henderson, M.A., Otero-Tapia, S., and Castro, M.E. (1998) Surf. Sci., 412/413, 252. 54 Sanchez de Armas, R., Oviedo, J., San Miguel, M.A., and Sanz, J.F. (2007) J. Phys. Chem. C, 111, 10023. 55 Li, B., Zhao, J., Onda, K., Jordan, K.D., Yang, J., and Petek, H. (2006) Science, 311, 1437.

56 Gamble, L., Jung, L.S., and Campbell, C.T. (1996) Surf. Sci., 348, 1. 57 Kim, Y.K., Kay, B.D., White, J.M., and Dohnalek, Z. (2007) J. Phys. Chem. C, 111, 18236. 58 Bondarchuk, O., Kim, Y.K., White, J.M., Kim, J., Kay, B.D., and Dohnalek, Z. (2007) J. Phys. Chem. C, 111, 11059. 59 Zhang, Z., Bondarchuk, O., White, J.M., Kay, B.D., and Dohnalek, Z. (2006) J. Am. Chem. Soc, 128, 4198. 60 Zhang, Z., Bondarchuk, O., Kay, B.D., White, J.M., and Dohnalek, Z. (2007) J. Phys. Chem. C, 111, 3021. 61 Zhang, Z., Ge, Q.F., Li, S.-C., Kay, B.D., White, J.M., and Dohnalek, Z. (2007) Phys. Rev. Lett., 99, 126105. 62 Vijay, A., Mills, G., and Metiu, H. (2003) J. Chem. Phys., 118, 6536. 63 Li, S.-C., Zhang, Z., Sheppard, D., Kay, B.D., White, J.M., Du, Y., Lyubinetsky, I., Henkelman, G., and Dohnalek, Z. (2008) J. Am. Chem. Soc., 130, 9080. 64 Kr€ uger, P., Bourgeois, S., Domenichi, B., Magnan, H., Chandesris, D., Le Fevew, P., Flank, A.M., Jupille, J., Floreano, L., Cossaro, A., Verdini, A., and Morgante, A. (2008) Phys. Rev. Lett., 100, 055501. 65 Kajita, S., Minato, T., Kato, H.S., Kawai, M., and Nakayama, T. (2007) J. Chem. Phys., 127, 104709. 66 Thompson, T.L., Diwald, O., and Yates, J.T. (2003) J. Phys. Chem. B, 107, 11700. 67 Farfan-Arribas, E. and Madix, R.J. (2002) J. Phys. Chem. B, 106, 10680. 68 Gan, S., Liang, Y., and Baer, D.R. (2000) Surf. Sci., 459, L498. 69 Stipe, B.C., Rezaei, M.A., and Ho, W. (1998) Science, 280, 1732. 70 Leung, L., Muryn, C.A., and Thornton, G. (2004) Surf. Sci., 566–568, 671.

j239

Index a ab initio investigation 115ff achiral 4ff – species 19ff – surface 12ff adenine 22ff adsorbate – mobility 206 – overlayer 61 adsorption of cysteine on Au 22 Al2O3/Ni3Al(1 1 1) 38 Al4O6Al6O7 153 alloy model catalyst 86 p-allyl 210 Ambient pressure – pd cluster 89 – X-ray photoelectron – spectroscopy (APXPS) 189 amino acid building blocks 2 anatase (0 0 1) bilayer 158 anion adlayer 130 anisotropy 8 Apiezon 125 Arrhenius relation 213 atom mobility 33 atomic – adsorber 4 – force microscopy (AFM) 99 – layers 39 – resolution 35 – spacing 169 – substituents 2 autocatalyic cycle 72 azimuthal orientation 10

b B3LYP real-space distribution 112 band – edge 37ff

– gap 43 benzene 214 bias – dependence 36 – voltage 35ff bimetallic cluster 51 binding energy 45, 151 biomolecular reactions 207 bipotentiostat 122 p-bonding 209 bSKAN code 108 bulk-oxide 148 bulk-terminated surfaces 148 butanol 67, 231 1-butyl-3-methyl-imidazolium 144

c c(2
2) structure 103ff Cahn-Ingold-Prelog system 2ff cantilevers 220 capacitance manometer 198 capillary forces 138 catalytic – model system 106 – oxidation 73, 103 – reaction 55, 59 – water production 101 cerium oxide 177 charge distribution 112, 198, 233 chemisorption 12, 56ff chiral 1ff – amplification 3–19 – building blocks 3 – dopant 19 – etching 16 – heterogeneous catalysis 18 – ligand 3 – molecules 1 – recognition 21

j Index

240

– switching 20 CO/NO coadsorption on Rh(1 1 1) 213 coalescence 89 cohesive energy 141 conductance 103 conduction band 35f conformation LR/RL 20 constant current mode STM 120 constant current STM image of CeO2 on Rh(1 1 1) 180 contamination, surface 128 contour plots of TiO2[0 1 1]-averaged – charged densities 107f contrast, image 149 convolution effects 39 CO-oxidation 150 coronene 9 corundum structure 167 cross-link reconstruction 48 cross-talk 138 crystallographic shear planes current profile 105 cyclic hydrocarbons 209 cyclic voltammetry 121 cyclohexene 209ff (S)-cysteine dimer 16

d damping 127 defect on TiO2 71 dehydrogenation 169 delamination 158 delineating 91 delocalization 113 density – functional theory (DFT) 222 – functional theory (DFT) calculation 7 – of primitive surface unit cell 219 – of state (DOS) 41ff depletion of Oad atoms 72 deprotonated thiolate surface 22 deprotonation 14ff desorption 224 determining site for chemisorption 64 dextrorotatory 2 diastereoreoisomeric interaction 8ff diffusion barrier 67, 152 dihydrogen 222 dipeptide 22 direct manipulation of atoms 31 dissociative adsorption 63ff distance tunneling microscopy 121 dot structure 45

e Ehrlich-Schwoebel barrier 134 electric circuit in situ STM 124 electrical – breakdown 59 – pulse 223 electrocatalysis 119 electrochemical cell 126 electrochemical nanostructure 135 electrochemistry 101 electrolyte 122 electron – affinity 111 – bombardment 64 – cyclotron resonance (ECR) – plasma oxygen 150 – density 131 – microscopy 197 – scattering quantum chemistry (ESQC) 170 electronic – state 44 – structure 37–44 electrostatic interaction 175 emission current 233 enantiomer 1ff enantioselective homogeneous catalyst 3 encapsulation 88 entropy 199 epitaxtial – film 172 – growth 173 equidistant space constant It contours 108 etching 125 ethylene hydrogenation on Pt(1 1 1) 207

f faceting 132ff reactivity of fcc(1 1 0) metals 76 feedback – control 140 – loop 38 femtolaser 55 FeO bilayer on Pt(1 1 1) 150ff Fermi – edge 34 – energy 111 – level 98ff field emission 165 – microscopy 197 flame annealing 128 flexible surface 92 (S)-form methamphetamine molecule 4

Index

g geometric distortion 40 Gibbs Energy 3 glancing incidence X-ray diffraction (GIXD) 174 glutamic acid 18 growth kinetics supported model catalyst 85 growth temperature 33

h Hartree-Fock terms 100 heterochiral (3
2) unit cell 7 heterogenous – enantioselective catalysis 147 – nucleation kinetics 47 hexagonal closed-pack 169 hexagonal superstructure 45 2,5,8,11, 14,17-hexa-tert-buthylhexabenzo [bc, ef, hi, kl, no, qr]coronene (HtB-HBC) 12 hierarchy of cluster rubene on Au{1 1 1} 6 high-energy – barrier 207 – electron bombardment 220 highly ordered pyrolitic graphite (HOPG) 190 high-pressure – high-temperature STM reactor 192 – scanning tunneling microscopy (STM) 189–191 high-resolution – AFM image of alumina film on Ni3Al 45 – energy loss spectroscopy (HREELS) 161 – image of V3O9 162 – STM-AFM image of alumina film on Ni3Al 45 homogeneous – catalyst 1 – nucleation 46 hopping 232 HSEO3 subsurface delocalization 113 HtB-HBC adsorption on – Cu{1 1 0}, STM-UHV 13 hybrid – DFT 101 – exchange density 175 hybridization 166 hydrogen oxidation 71

i image formation in STM 98 imaging – adsorbates intermediate reaction 59ff

– chemisorption on metals 61 – oxide film 34 – single-crystal surfaces 128 in situ – STM investigation 56 – STM study of model catalyst 55 – study supported model catalyst 85 – tip regeneration 57f inductive heating setup 129 instrumentation 56 interdigitation 9ff interface 88, 148 interfacial – contacts 193 – lattice mismatch 179 – proximity effect 147 – strain 163–182 intermediate stage reaction of c(2
4)-O 102 intermolecular – bonding 23 – interaction 33 – separation 8 interstitial, defects 69ff ion – scattering spectroscopy 179 – sputtering 64, 110 iron oxide 169 island density 34

j jump-to-contact 137

k kagomé-phase 156 kinetic Monte Carlo simulation 33

l lateral – force 195 – interaction of molecular carboxylate 19 – lattice constant 172 – lattice expansion 171 – resolution 121 – tip position 35 lattice – constant 46 – matching 8 – mismatch 163–174 – spacing 169 line profile analysis 205 linear fit, reaction rate 75 liquid helium bath cryostat STM 220 lithography 31ff

j241

j Index

242

local electronic density of states (LDOS) 150 lock-key model 22 long-range order 149 low temperature STM 57ff low-energy electron diffraction (LEED) 7, 106 L-phe-L-phe 23

m map of W[1 1 0] tunneling current 108 Mars-van-Krevelen mechanism 83 material gap 56ff metal surface at high pressure 81 metal-on-oxide system 30ff metal-oxide interface 175 methamphetamine 4 methanol adsorption 67, 231 (R)-methyl-3-hydroxybutyrate 25 methylacetoacetate (MAA) 24 (R)-3-methylcyclohexanone 16 microcantilever 139 microfaceted surface 203 microprocessor 140 Miller index surfaces 199 miniaturization 139 mirror domain 5–11 mobility 70, 73 model – catalyst 30ff – TiO2 surface 65 modeling surface reactivity 97 modified template 50 Moiré pattern 208 molecular – adsorption 4 – beam scattering 189 – model enantiomer 5 – model of glycinate {1 1 0} 7 monoatomic steps 206 monodisperse metal cluster 155 monolayer 130 monomer migration 89

n nanocatalyst 155 nanocluster 139ff nanodecoration 141 nanodot array, Cu 136 nanofabrication 30 nanofacet 174 nanoisland 80 nanolayer 148 nanophase 148

nanoscale hybrid structure 182 nanosheet 158 nanostructure 29 nanostructure by self-organization 32 near-edge X-ray absorption fine structure (NEXAFS) 160 Nernst potential 137 network structure 45 nickel oxide nanolayers 173 nomenclature of chirality 2 noncontact AFM 178ff nucleation – density 44 – site 32ff number density 44

o one dimensional-Ag-O-chain 76 optoelectronics 159 oscillatory 122 Ostwald ripening 89f oxidation-reduction cycle 132 oxide as template 44 oxidizing agent 126 oxygen-mediated diffusion 224

p paramagnetic triple state 113 particle on oxide film 39 Pd hydrogenation 62 periodicity 48 PES spectra 70 phase – boundary 80, 148 – transition 82 phenylglycine 8–23ff phonon spectrum 161 photo-assisted oxidation 155 photocatalysis 64, 229ff photoelectron – diffraction (PED) 112 – spectroscopy (PES) 69 photoreactive molecule 69 piezoelectric – plates 193 – scanning tube 195 platinum particles on oxidized silicon, STM 31 plot number density of cluster 47 point defect 219 polycrystalline 127 polyisocyanate copolymers 20 polymorphism 160 positive bias 221

Index postoxidation 150 pressure gap 29ff prochiral – carboxylic acid 17 – molecules interaction with chiral surfaces 24 – reagent 3 pseudodissociate state 109 pseudomorphic phase 175 pseudoperfect surfaces 235 pseudohexagonal lattice in fcc {1 1 0} 11f p-type semiconductor 173 purification, reactant gas 59 PVBA structure 17

q QMS signal O2 84 quadruple mass spectrometer (QMS) 82 quartz 15 quasi-hexagonal 153 quasi-zero-dimensional V-oxide clusters 164

r racemic mixture 3ff reaction – kinetics of CO oxidation 73 – rate plot CO 75 reactive – evaporation (RE) method 150 – metal 142 real-space distribution, electronic states 114 resolution limitation 44 resonant frequency 220

s saturated calomel electrode (SCE) 126 scanning – electrochemical microscopy 120 – tunneling microscopy (STM) 1, 7ff – tunneling spectroscopy (STS) 97 schematic, continuous cluster generation 140 segmentation of AgO chains 76 self-assembled nanostructures 165 self-recognition 21 semilocal (CGA ) approximation 110 sequential image of TiO2 (1 1 0) – showing the splitting of an Ohb 226 shadowing technique 57 schematic model point defect 227 simulating – chemical reactivity 100 – tunneling current 99 single-crystal surface 29–178

sintering kinetics of Au 89 spatial – confinement 34 – limit 55 spectroscopic signature 233 spin density 111 p-stacking 209 static secondary ion mass spectrometry 229 step edge 86ff step-kink arrangement fcc{6 4 3}R 16 stereocenter 2 stereochemistry 2ff STM image of – Ag(1 1 0)(n·1)-O surface 78, 80 – alumina film on Ni3Al 37 – r-aminobenzoic acid on Cu 18 – anionic adlayer 131 – at oxygen spillover 89 – Au cluster 91 – CO molecule 74 – CO oxidation 84 – cocrystalline 1:1 structure 25 – Cu clusters on Al2O3/Ni3Al(1 1 1), Ag 46 – 1,4 cyclohexadiene 212 – decamer of 1-nitroaphthalene 6 – electrodeposited Pt cluster 133 – Fe nanoparticle on FeO4 film on MgO(1 0 0) 50 – growing Pd monolayer 132 – H-vacancy clusters 63 – lifted Pt(1 0 0) 201 – methanol adsorption 68 – Ni-oxide on Pd(1 0 0) 176 – O-covered Pt(1 1 1) 72 – oxidation cycle 168 – oxidation-reduction 135 – Pd cluster on Al2O3/Ni3Al 46ff – Pd(1 1 1) in H2SO4 130 – Pt(1 1 0) 83 – PVBA-induced 19 – reduction of Ag(1 1 1)-p(4·4) 79 – sequential image Rh(1 1 1) 215 – single Fe cluster 41 – stoichiometric 1x1 TiO2 107 – TiO2 66ff – TiOx structure 157 – V cluster on Al2O3/Ni3Al 48 – V3O9 island 81f – water saturated surface 103 STM image with CO-terminated tip 60f stoichiometry 70 strong metal support interaction effect 88 STS alumina film spectra 36 STS differential conductance spectra 166

j243

j Index

244

submonolayer 130 subnanometer 97 substrate 113 sum frequency generation – vibrational spectroscopy 189 supercell 103 superlattice of defects 181 surface – characterization 34 – defect 106 – diffusion 60, 62ff – dipole 98–72 – flexibility 202 – mobility 189 – morphology 97 – pitting 233 – reactivity 84 – relaxation 199 – science investigation of chirality 1 – tip interaction 103 – topography 97 surface hydroxyl 65 surface X-ray diffraction (SXRD) 83, 152 synchrotron radiation 161 synergetic 90

t tartaric acid 14 temperature-programmed desorption 229 template – controlled growth 44 – model catalyst 29 tetragonal pyramid 162 thermal – desorption spectroscopy 210 – drift 149 – switching of conformation 23 – induced restructuring 199 thermochromic 159 thermocouple 195 time – constant 138 – evolution of y(2·1)-O coverage, plot 77 tip – apex 103 – exchanger 197 – manipulation 58ff

– morphology 39 – preparation 124 – trajectory 40 titanium oxide 155 titration experiment 73 titration curve Ag(1 1 0)(n
1)-O surfaces 78 topographic image of Pt(1 1 0) 205 topography change 36 4-[trans-2-(pyrid-4-yl-vinyl)]benzoic acid (PVBA) 5 tunneling – probability 206 – gap 34ff – barrier 41 – barrier plot 123 – current 34, 40ff – density 43 – junction 98 – process 35f – resistivity 41 – spectroscopy 37 two-dimensional template 31

u UHV-STM image of Coronene on Cu{1 1 0} 11 ultra-high vacuum (UHV) 81 ultrathin – alumina layers 152 – oxide layers 182 unit cell 38

v vacancy generation 212 vacancy-mediated diffusion 214 vacuum barrier 35 valence band 36f van der Waals interaction 8ff vanadium oxide 159 variable temperature STM 57 vertical lattice compression 171

x X-ray photoelectron – diffraction characterization 14 X-ray absorption profile X-ray photoelectron spectroscopy (XPS) 151