Scanning tunneling microscopy and its applications

SPRINGER SERIES IN SURFACE SCIENCES 32 Springer Berlin Heidelberg New York Barcelona Hong Kong London Milan Paris Sing...

Author: Chunli Bai

160 downloads 1829 Views 125MB Size Report

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

DOWNLOAD PDF

SPRINGER SERIES IN SURFACE SCIENCES

32

Springer Berlin Heidelberg New York Barcelona Hong Kong London Milan Paris Singapore Tokyo

ONLINE LIBRARY

Physics and Astronomy '---J

http://www.springer.de/phys/

SPRINGER SERIES IN SURFACE SCIENCES

Chunli Bai

Series Editors: G. Ert!, R. Gomer, H. Liith and D.L. Mills This series covers the whole spectrum ofsurface sciences, including structure and dynamics of clean and adsorbate-covered surfaces, thin films, basic surface effects, analytical methods and also the physics and chemistry of interfaces. Written by leading researchers in the field, the books are intended primarily for researchers in academia and industry and for graduate students. 38

Progress in Transmission Electron Microscopy I Concepts and Techniques Editors: X.-F. Zhang, Z. Zhang

39

Progress in Transmission Electron Microscopy II Applications in Materials Science Editors: X.-F. Zhang, Z. Zhang

Scanning Tunneling Microscopy and Its Applications Second, Revised Edition With 208 Figures

Series homepage - http://www.springer.de/phys/books/ssss/

Springer Volume 1-37 are listed at the end of the book

SHANGHAI SCIENTIFIC &TECHNICAL PUBLISHERS

Preface

Professor Dr. Chunli Bai Institute of Chemistry The Chinese Academy of Sciences Beijing 10080, People's Republic of China

Series Editors: Professor Dr. Gerhard Ertl Fritz-Haber-Institute der Max-Planck-Gesellschaft, Faradayweg 4- 6 , 14195 Berlin, Germany

Professor Robert Gomer, Ph.D. The James Franck Institute, The University of Chicago, 5640 Ellis Avenue, Chicago, IL 60637, USA

Professor Dr. Hans Liith lnstitut fur Schicht- und lonentechnik Forschungszentrum Jiilich GmbH, 52425 Jiilich, Germany

Professor Douglas L. Mills, Ph.D. Department of Physics, University of California, Irvine, CA 92717, USA Library of Congress Cataloging-in-Publication Data. Bai, Chunli, Scanning tunneling microscopy and its applications I Chunli Bai. - 2nd rev. ed. p. cm. - (Springer series in surface sciences, ISSN 0931-5195 ; 32) Includes bibliographical references and index. ISBN 3-540-65715-0 (alk. paper) I. Scanning tunneling microscopy. 2. Surfaces (Physics). 3. Surface chemistry. I. Title, II. Series. QC 173.4.S94B35 1999 99-34035 502'.8'25-dc21

Revised translation of the original Chinese edition: BAI Chunli: Saomiao suidao xianweishu ji qi yingyong © Shanghai Scientific and Technical Publishers, 1992

ISSN 0931-5195 ISBN 3-54°-65715-0 Second Edition Springer-Verlag Berlin Heidelberg New York ISBN 3-540-59346-2 First Edition Springer-Verlag Berlin Heidelberg New York ISBN 7-5323-2787-6/0'161 Shanghai Scientific & Technical Publishers This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German Copyright Law. Springer-Verlag Berlin Heidelberg New York a member of BertelsmannSpringer Science+ Business Media GmbH © Shanghai Scientific & Technical Publishers and Springer-Verlag Berlin Heidelberg 1995, 2000

Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: PS™ Technical Word Processor Cover concept: eStudio Calamar Steinen Cover production: design & production GmbH, Heidelberg Printed on acid-free paper

SPIN: 10715754

~

.... .. _-

We have witnessed an amazingly fast growth of research related to Scanning Tunneling Microscopy (STM). The present monograph written nearly five years can clearly not be considered sufficient for researchers in this field. This observation motivated the author to come up with a new edition to include the recent developments. The revisions are focussed on recent advances (mainly since the time progress the first edition has been published) in the field of STM, stressing \ on methodology and impacts on relevant areas. I chose to add the recent results, while largely keeping the original content since it still remains valid in these days. Some of the additions are very brief, as I deem them necessary to reflect on the current status of a subfield; detail introductions can be found there. There are indeed many superb achievements which should be discussed in this new edition, but the author is only able to select a limited number of them. The revised content is increased more than twenty five percent compared to the first edition, and about one hundred new references have been added. A few new sections have been prepared to reflect the progress of the respective topics. I greatly appreciate the efficient and careful editing of this new edition by Dr. Helmut K.V. Lotsch, who was also responsible for the first edition of the book, and Dr. Claus Ascheron for his support. My colleagues at the Institute of Chemistry, CAS, have been of great help in completing this new edition. The effort of Dr. Chen Wang is highly appreciated. Ms. N.X. Wang, Mr. M.Z. Liu. Dr. LJ. Yang have contributed to collecting and organizing the references, and to typing the draft version of the added material. The author also wishes to express sincere gratitude to many fellow researchers in the STM field for supporting the second edition; to name some of them: Drs. Ph. Avouris and N.D. Lang (IBM TJ. Watson Research Center), Dr. RJ. Colton (Naval Research Center), Prof. A. de Lozanne (University of Texas at Austin), Dr. G. Poirier (Nat'l Institute of Standards), Dr. D.M. Chen (Rowland Institute of Science), N.J. Tao (Florida Int'l University), Dr. R.A. Wolkow (Nat'l Research Council of Canada), Prof. S. Morita (Osaka University), Prof. M. Edidin (Johns Hopkins University), Prof.H. van Kempen (University of Nijmegen), Dr. H. von Kaenel ( ETH Zuerich), Prof. R. Berndt (RWTH

57/3144/Xo - 5 4 3 210

v

Aachen), Prof. A.L.V. de Parga (Universidad Autonoma de Madrid), Prof. H.E. Gaub (TU Munich), Prof. c.L. Lieber (Harvard University). They have generously provided original data and granted permission to use them in this new edition. I enjoyed my efforts to prepare the first edition of this book, and I sincerely hope that was able in this second edition to improve the presentation to better serve the readers. Beijing January 2000

Chunli Bai

Preface to the First Edition

This book evolved out of the Chinese version which was published in 1992 by the Shanghai Science & Technology Publishers. The English version was drafted in 1992 when I was worked as a visiting professor at the Institute for Materials Research, Tohoku University, Japan. Since then some new achievements have constantly been added to the English version. In the middle of 1993, I received the computer printouts of the manuscripts of this book from Springer-Verlag, when I was organizing the 7th International Conference on Scanning Tunneling Microscopy in Beijing, China. Later on I participated in the preparation of the First Asian Conference on Scanning Tunneling Microscopy (STM) with Prof. T. Sakurai (Tohoku University, Japan) and Prof. Y. Kuk (Seoul National University, Korea). By the turn of the year, I was again involved in the academic activities of the Third International Conference on Nanometer-Scale Science and Technology. In order to include some of the latest achievements reported to the above three conferences, I have made a great amount of modifications and revisions to the English version. The additions and updates amount to almost one third of the original Chinese version. This book is divided into nine chapters, with the first chapter discussing the basic principles and concepts of STM, and comparing it with electron microscopy and field-ion microscopy. The second chapter talks about the theoretic background and related concept. Experimental modes and applications of tunneling spectroscopy and spectroscopic imaging are presented in the third chapter. The fourth chapter deals with STM instrumentation and STM tip preparations; the fifth chapter includes other related scanning probe microscopes, such as atomic force microscope (AFM), the lateral force microscope (LFM), the magnetic force microscope (MFM), the electrostatic force microscope (EFM), ballistic-electron-emission microscopy (BEEM), the scanning ion-conductance microscope, the scanning thermal microscope, scanning tunneling potentiometry, the photon scanning tunneling microscope, and the near-field scanning optical microscope. Chapters 6 to 9 present some typical examples introducing STM studies of clean metal and semiconductor surfaces (Chap. 6), surface adsorbate and surface chemistry (Chap.7), biological applications (Chap.8), and surface modifications (Chap. 9). VII

VI

My obligations in this endeavor are numerous. I want first to thank my colleagues from the STM LAB of the Institute of Chemistry, the Chinese Academy of Sciences. Their support, help and encouragement made this book a reality. I want to express my gratitude to the following for allowing me to use their high quality pictures and related materials: Prof. P.K. Hansma (University of California at Santa Barbara), Dr. Phaedon Avouris, Dr. C. Julian Chen and Dr. Y.W. Mo (IBM Yorktown Laboratory), Prof. F. Besenbacher and Dr. L. Ruan (Aarhus University, Denmark), Prof. P.S. Weiss (Penn State University), Dr. N.J. Tao (Florida International University), Dr. Y. Strausser (Digital Instrum. Inc.), Prof. Y. Kuk (Seoul National University), Dr. R.J. Colton (Naval Research Laboratory) and Prof. T. Sakurai (Tohoku University, Japan). My special thanks to Dr. R. J. Colton for his carefull reviewing of specific parts of the book and providing valuable comments and suggestions. I would also like to thank Dr. H. Lotsch, senior editor in physical sciences at Springer-Verlag for assisting in the publication of this book. Due to various reasons, I was not able to provide the preliminary camera-ready manuscript of the book. This together with several quite extensive changes and modifications of the manuscripts burdened him with a very time consuming workload. Scanning tunneling microscopy is a fast developing field, with new concepts and new applications appearing everyday. Though I have tried my best to cover all possible major trends and developments of the areas, I still could hardly avoid omissions and uncovered subjects. I welcome any comments and suggestions from readers which will help towards a wellrounded, truthful and more accurate version should this book be reprinted in the future. Beijing February 1995

Chunli Bai

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . ..

'"

1.1 Advantages of STM Compared with Other Techniques '" 1.2 From Optical Microscopy to Scanning Tunneling Microscopy . . . . . . . . . . . 1.2.1 Electron Microscopes . 1.2.2 Field Ion Microscope " . 1.2.3 Scanning Tunneling Microscope.. 1.3 Overview . . . . . . . . . . . . . . . . . . . . . . .. . '

. . .

3 3 5 5 7

2. The Tunneling Effect . 2.1 Historical Remarks . 2.2 Tunneling-Current Theory. . . . . . . . . . . . . . . . . .. 2.2.1 Tunneling Current . 2.2.2 Practical Tip and Surface Wave Functions . 2.3 Tip-Surface Interaction Model . 2.3.1 Tunneling Current . . . . . . . . . . . . . . . . . . . . . . .. 2.3.2 Tunnel Conductance . 2.3.3 Tunneling Active Orbital at the Tip.. . . 2.3.4 Double-Tip and Interference Effects .

11 12 17 18 22 25 25 29 30 32

3. Spectroscopy, and Spectroscopic Imaging. . . . . . . . . . . . . 3.1 Concepts of Tunneling Spectroscopies . . . . . . . . . . . . . . 3.1.1 Solid-State-Barrier Tunneling . . . . . . . . . . . . . . . 3.1.2 Metal-Vacuum-Metal Tunneling. . . . . .. 3.2 Experimental Modes . . . . . . . . . . . . . . . . .. 3.2.1 Current-Voltage Characteristics. . . . . . . . . . . .. .. 3.2.2 Current-Separation and Separation-Voltage Characteristics . . . . . . . . .... 3.2.3 Constant-Current Topography. . . . . . .. 3.2.4 Current-Imaging Tunneling Spectroscopy. . . . . . . 3.3 Energy Resolution. . . . . . . . . . . . . . . . . . . . . . . 3.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4.1 Surface States " 3.4.2 Adsorbate-Covered Surfaces . . . . . . . . . . . . . . . . . . 3.4.3 Superconductivity....... 3.4.4 Outlook. . . . . .

37 38 38 40 42 43 44 46 46 47 49 50 56 57 60

VIII IX

a) Influence of the Tip , b) Interpretation of Spectroscopy Results . . . . . . . 4. STM Instrumentation. . . . . . . . . . . . . . . . . . . . . . . . . . .. 4.1 The Vibration Isolation System. . . . . . . . . . . . . . . . . . .. 4.2 Mechanical Designs. . . . . . . . . . . . . . . . . . . .. 4.2.1 Piezoelectric Ceramics 4.2.2 Three-Dimensional Scanners. . . . . . . . . . . . . . . .. 4.2.3 Coarse Sample Positioning. .. . 4.2.4 STMs for Operation in Various Environments , 4.3 Tip Preparation . . . . . . . . . . . . . . .. 4.3.1 Preparation of Tungsten Tips . .. 4.3.2 Preparation of Pt-Ir Tips , . . .. 4.3.3 Other Ways to Prepare STM Tips. 4.3.4 Tip Treatment. . . . . . . . . . . .. . 4.4 Electronics . . . . . . . . .. . . 4.5 Computer Automation . . . . . . . . .. 4.5.1 Hardware...... 4.5.2 Software . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 4.5.3 Image Processing. . . . . . . . . . . . . . . . .. a) Histogram Equalization b) Convolution Filter. . . . . . . . . . . . . . . . . . . c) Statistical Differencing. . . . . . . . . . . . . . . . d) Three-Dimensional Representation , ..

60 60 63 63 67 70 71 74 76 80 80 84 87 89 . 92 96 97 98 100 101 101 102 102

5. Other Related Scanning Probe Microscopes. . . . . . . . . . . .. 5.1 Atomic Force Microscope . . . . . . , 5.1.1 The Force Sensor. .. . 5.1.2 Deflection Detection. . . . . . . . . . . . . . . . . . . . 5.1.3 Illustrating AFM Applications . . . . . . . . . . . . . . .. 5.2 Other Scanning-Force Microscopies , 5.2.1 Lateral Force Microscope. . . . . . . . . . . .. 5.2.2 Force Microscope Operating in the Noncontact Mode. 5.2.3 Force Microscope Operating in the Tapping Mode .. , 5.2.4 Magnetic Force Microscope . . . . . . . . . 5.2.5 Electrostatic Force Microscope. 5.3 Ballistic-Electron-Emission Microscopy . . . . . . . . . . 5.3.1 The Principle of BEEM . . . . .. 5.3.2 The BEEM Experiment. . . . . . . . . . . . . . . . . . .. 5.3.3 The Application of BEEM . . . . . . . . . . . . . . . .. 5.3.4 Ballistic-Hole Spectroscopy of Interfaces , , 5.3.5 Interfacial Modification with BEEM 5.4 Scanning Ion-Conductance Microscope. . . . . . .. 5.5 Scanning Thermal Microscope '

x

105 105 107 112 114 121 121 112 126 127 130 133 133 135 136 140 144 145 147

5.6 Scanning Tunneling Potentiometry and Scanning Noise Microscopy . 5.7 Photon Scanning Tunneling Microscopy and Scanning Plasmon Near-Field Microscopy . 5.8 Near-Field Scanning Optical Microscopy and Spectroscopy .. 5.8.1 Principles of Near-Field Optics . 5.8.2 Optical Probes for Near-Field Optics . 5.8.3 NSOM Operation . 5.8.4 Near-Field Scanning Optical Spectroscopy . 5.8.5 Near-Field Optical Chemical Sensors . 5.8.6 Scanning Exciton Microscopy . 5.8.7 Single-Molecule Detection by Near-Field Optics . 6. STM Studies of Clean Surfaces . 6.1 Metal Surfaces . 6.1.1 Geometric Structures . 6.1.2 Electronic Structures . 6.1.3 Surface Diffusion . 6.2 Elemental Semiconductor Surfaces . 6.2.1 The Si(ll1) Surface .. a) Si(11l)-7 x7 . . b) Si(lII)-2xl 6.2.2 TheSi(OOI)Surface , . a) Geometric Structure . b) Electronic Structure . c) The 2 Xn Structure . 6.2.3 Other Silicon-Surface Structures . 6.2.4 The Ge Surfaces . a) Ge(l11) . b) Ge(OOl) . 6.2.5 The GeSi(l11) Surface . 6.3 Compound Semiconductors and Layered Compounds . 6.3.1 GaAs Surfaces . a) GaAs(l10) . b) GaAs(lOO) . c) GaAs(lll) and GaAs(II I) . d) GaAs-AIGaAs . 6.3.2 Layered Compounds . 6.3.3 Charge-Density Waves in Compound Semiconductors . . . a) CDW Phases of IT-TaS z . .. b) Charge-Density Wave Defects . 6.3.4 High-T c Oxides .

149 151 153 155 156 157 158 161 162 163 165 165 166 168 172

173 173 173 175 177 178 180 182 183 184 184 185 186 187 188 188 190 191 192 193 195 196 200 200

XI

7. Surface Adsorbates and Surface Chemistry . . . . . . . . . . . .. 205 7.1 Adsorption on Metal Surfaces. . . . . . . . . . . . . . . . . . . ., 205 7.1.1 Cu(l10)-O 206 7.1.2 Cu(lOO)-O 208 7.1.3 Dynamics · · · · · · · · · · · 209 7.1.4 Ag(l10)-O , , .. , .. , , 211 7.1.5 Ni(110)-H and Ni(l11)-H ' 213 7.1.6 Sulfur Adsorption ' 215 7. 1.7 Cu( 111 )-S. . . . . . . . . . . . . . . . . . . . . . . . . . . ., 216 · · · · · · · · · · 219 7.1.8 Ni(lII)-S 7.1.9 Cu(l10)-S , 221 7.1.10 Ni(llO)-S ' 224 7.1.11 Mo(OO 1)-S and Re(OOO 1)-S . . . . . . . . . . . . . . . . .. 227 7.1.12 Other Non-metal Adsorbates on Metals . . . . . . . . ., 228 7.1.13 Metallic Adsorbates . . . . . . . . . . . . . . . . . . . . .. 228 7.2 Adsorption on Semiconductor Surfaces , 229 · · · · · · · · · · · 230 7.2.1 Ag/Si(111) 7.2.2 Au/Si(ll1) , 231 7.2.3 Cu/Si(111) , 232 7.2.4 Group-III Metals on Si(111). . . . . . . . . . . . . . . . .. 233 7.2.5 B/Si(111) · · · · · · · · · · 234 7.2.6 Cl/Si(lll) 235 7.2.7 Bi on Si(lOO) and Si(111) Surfaces. . . . . . . . . . . .. 235 7.2.8 Na/Si(111) , 236 7.2.9 Na/GaAs(110) and Cs/GaAs(llO) , 238 7.2.10 Alkali Metals on Si(100)-2Xl , 241 7.3 Molecules, and Molecular Adsorbates. . . . . . . . . . . . . . .. 242 7.3.1 Molecular Crystals . . . . . . . . . . . . . . . . . . . . . ., 242 7.3.2 Chemisorbed Aromatic Molecules in Ultrahigh Vacuum 243 7.3.3 Physisorbed Molecules in Ultrahigh Vacuum. . . . . .. 245 7.3.4 Physisorbed Long-Chain Molecules. . . . . . . . . . . .. 246 7.3.5 Chemisorption of Long-Chain Molecules . . . . . . . .. 249 7.3.6 Fullerenes 252 a) C on GaAs(1lO) , 252 60 b) C 60 on Si(100). . . . . . . . . . . . . . . . . . . . . . .. 254 c) C on Si(111) , 254 60 d) C 60 on MoS 2 (000 1) . . . . . . . . . . . . . . . . . . .. 254 e) C on Cu(111) . . . . . . . . . . . . . . . . . . . . . .. 255 60 f) C on Au(111) and Ag(lll) , 255 60 7.3.7 Langmuir-Blodgett Films , 256 7.4 Observation of Clusters 257 7.4.1 Metal Clusters . . . . . . . . . . . . . . . . . . . . . . . . ., 257 7.4.2 Semiconductor Clusters . . . . . . . . . . . . . . . . . . .. 260 7.5 Nucleation and Growth. . . . . . . . . . . . . . . . . . . . . . . 260 XII

7.5.1 Epitaxial Growth of Metal Films. . . . . . . . . . . . . .. 7.5.2 Growth of Si on Si(OOI) . . . . . . . . . . . . . . . . . . .. 7.6 Chemical Reactions on Metals . . . . . . . . . . . . . . . . . . .. 7.6.1 Reaction on Ni(1lO) . . . . . . . . . . . . . . . . . . . . .. 7.6.2 Reaction on Cu(llO) .. .. .. .. . .. . .. .. .. .... 7.6.3 Chemical Identity with STM . . . . . . . . . . . . . . . .. 7.7 Chemical Reaction on Semiconductors . . . . . . . . . . . . . .. 7.7.1 Reaction of NH 3 with Si( 111)-7 x7 Surfaces , 7.7.2 Reaction of NH 3 with B/Si(lII)-V3 Xv3 Surface .. , 7.7.3 Reaction of NH 3 with Clean Si(OOI) Surface , 7.7.4 Si(lII)-7x70xidation 7.7.5 Si( 100)-2 X1 Oxidation . . . . . . . . . . . . . . . . . . .. 7.7.6 Reaction ofH withSi(111)-7x7 7.7.7 Reaction of Sb4 with Si(100) , 8. Biological Applications I. • • • • • 8.1 Advantages and Problems . . . . . . . . . . . . . . . . . . . . . 8.1.1 Substrates........................... 8.1.2 Fixation of Samples onto Substrates . . . . . . . . . . 8.1. 3 Flexibility of Biological Samples. . . . . . . . . . . . . 8.1.4 Identification and Interpretation of STM Images . . 8.2 Preparation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2.1 Dispersion of Samples on Substrates . . . . . . . . . . 8.2.2 Fixation of Samples . . . . . . . . . . . . . . . . . . . . a) Sample Coatings. . . . . . . . . . . . . . . . . . . . b) Covalently Binding Samples with Strongly Absorbent Groups . . . . . . . . . . c) Binding Samples to the Substrate Covalently. . . 8.2.3 STM Imaging in Acqueous Solutions. . . . . . . . . . 8.2.4 Hopping Technique. . . . . . . . . . . . . . . . . . . . . 8.2.5 STM Directed by an Optical Microscope . . . . . . . 8.3 Nucleic Acids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.3.1 DNA in Air and in Vacuum. . . . . . . . . . . . . . . . 8.3.2 DNA Studies Under Water with an Electrolyte. . . . 8.3.3 DNA-Protein Complex . . . . . . . . . . . . 8.3.4 DNA Bases. . . . . . . . . . . . . . . . . . . . . . . . . . 8.3.5 DNA Sequencing by Scanning-Probe Microscopes . 8.4 Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.4.1 Animo Acids and Peptides . . . . . . . . . . . . . . . . 8.4.2 Structural Proteins . . . . . . . . . . . . . . . . . . . . . 8.4.3 Functional Proteins. . . . . . . . . . . . . . . . . . . . . 8.5 Biological Membranes 8.6 Imaging Cells and Other Applications. . . . . . . . . . . . . . 8.7 Force Spectrum Analysis of Biological Materials. . . . . . .

261 262 265 265 268 269 271 271 273 274 274 276 276 277

•• 279 .. 279 280 .. 280 .. 281 .. 281 .. 282 .. 282 .. 283 .. 283

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

283 284 284 284 286 287 287 289 292 292 294 296 296 298 299 301 .. 302 .. 306 XII!

9. Surface Modification. . . . . . . . . . . . . . . . . . . . .. 9.1 Overview . . . . . . . . . . . . . . . . . . . . . . ., .. '" 9.2 Direct Indentation with the Tunneling Tip. . . . . .. 9.2.1 Modification of Metal Surfaces , . 9.2.2 Modification of Semiconductor Surfaces . 9.3 Nanolithography on Resist Films . 9.4 Nanofabrication in Solution and in Gaseous Environments. 9.4.1 Nanofabrication in Solution . 9.4.2 Nanofabrication in Gaseous Environments . 9.5 Atomic-Scale Manipulation . . . . . . . . . . . . .. 9.5.1 Manipulation of Atoms . a) Xenon Atoms. . . . . . . . . . . . . . . . . . . . . . . . b) Iron Atom . c) Silicon Atom . d) Sulfer Atoms. . .. . . 9.5.2 Manipulation of Molecules and Clusters . a) Carbon Monoxide . . . . . . . . . . . . . . . . . . . . . b) Antimony Molecules . c) Decaborane and Other Organic Molecules. d) H2 0 Molecules. . . . . . . . . . . . . . . ., .,. 9.6 Quantization of Conductance in Nano-Contacts Produced by STM . . . . . . . . . . . . . . . . . . . . .. .., 9.7 Fabrication with Other Scanning-Probe Microscopes 9.7.1 Machining Thin Films . . . . . . . . . . . . . . . . 9.7.2 Charge Storage . . . . . . . . . .. . . 9.7.3 Magnetic Structures and Writing into an Interface 9.8 The Future , .

309 309 310 311 317

321 324 324 325 327 327 327 330 331 333 333 333 333 336 336 338 339 339 340 343 343

References. . . . . . . . . . . . . . . . . . . .

345

Subject Index

367

XIV

.

1. Introduction

In 1981, Binnig and Rohrer [1.1] and their colleagues at the Zurich Research Laboratory of the International Business Machines (IBM) developed a new kind of surface analytical instrument - Scanning Tunneling Microscope (STM). The emergence of STM makes it possible to observe the arrangement of individual atoms on material surfaces, and physical and chemical properties related to the behavior of surface electrons in real space. This technique is revolutionizing surface science and the way we study surface phenomena. G. Binnig and H. Rohrer were awarded the Nobel Prize in Physics in 1986 for their outstanding contribution to science. In the following, we will briefly review the advantages of STM compared with other surface-analysis techniques.

1.1 Advantages of STM Compared with Other Techniques A number of modern instruments for surface structural and chemical analysis [1.2,3] such as the Field Ion Microscope (FIM), the Field Emission Microscope (FEM), Low-Energy Electron Diffraction (LEED), Auger Electron Spectroscopy (AES), Electron Spectroscopy for Chemical Analysis (ESCA), Electron-Probe Microanalysis (EPM), etc. have emerged following the invention of the first Electron Microscope (EM) by E. Ruska and M. Knoll in Berlin in 1931. The development and application of these techniques have played an important role in the field of surface science. However, each of these techniques has certain strengths and limitations. LEED and X-ray diffraction techniques rely on large-scale order, and can at best give averaged information about local and defect structure; a Scanning Electron Microscope (SEM) requires samples with strong corrugation or mass contrast and its resolutions is not high enough to resolve surface atoms; high-resolution Transmission Electron Microscopy (TEM) can in some cases resolve features with atomic dimensions of specially thinned samples. In most cases this can be accomplished only by aligning the electron beam with the rows of atoms in a crystalline lattice. FEM and FIM are only able to probe the two-dimensional geometry of atomic structure on the

surfaces of sharp tips with radii less than 100 nm. In addition, sample preparation is rather complicated. For FIM the samples must be stable in high fields, thus limiting its general usefulness. Some other surface analytical techniques, such as X-ray Photoemission Spectroscopy (XPS), Ultraviolet Photoemission Spectroscopy (UPS) and Electron Energy Loss Spectroscopy (EELS), in fact, can only provide spatially averaged information of electronic structures. Moreover, some of the techniques mentioned above require high-vacuum environment and can only provide indirect results or strongly rely upon model systems for data interpretation. Until the STM was introduced, it still remained a dream to directly observe geometric and electronic surface structures on the atomic level at ambient pressure and at room temperature. Compared with other surface analytical techniques, there are several reasons for the diversity of STM and STM-based technological applications: • STM can achieve atomic-level resolution. The lateral and vertical resolutions can reach 0.1 nm and 0.01 nm, respectively, i.e., individual atoms and molecules can be resolved. Figure 1.1 compares the resolution of STM with that of other kinds of microscopes. The higher vertical resolution of STM relative to other microscopes also offers advantages with regard to qualitative analysis of surface roughness on a nanometer scale . • STM can be performed in different environments, such as vacuum, air, low or high temperature, etc. Samples can even be immersed in water or other solutions under potential control. In most cases, special techniques for sample preparation are not required, and samples remain mostly free of

--,

10

3

HMI

_.:.it

E

I

.s

I

~

os ~ Oi .~

~

I

PC~J

10- 1

10- 1

2

I

10\

10 1 10 3 Lateral scale [nm]

lOS

1.2 From Optical Microscopy to Scanning Tunneling Microscopy Scanning tunneling microscopy is the product of considerable evolution. Microscopy appears to have begun in the 15 th century when simple magnifying glasses were made with which to observe insects. Although tiny objects such as bacteria and cells become visible, the resolution of optical microscopes is limited by the value of the wavelength of visible light. Since the shortest wavelength of visible light is 0.4 /lm, the highest resolution of optical microscopes is only 0.2 /lm. In order to probe atomic structures, a new light source with a shorter wavelength is needed .

lOS

~

damage. With these advantages, STM is especially suitable for in-situ electrochemical studies, biological studies, and the evaluation of sample surfaces under various experimental conditions. • The other unique feature of the STM is its truly local interaction on the atomic scale rather than the averaged properties of the bulk phase or of a large surface area, which allows the study of individual surface adsorbates, surface defects, surface reconstructions, and adsorption-induced surface reconstructions at unprecedented resolution. • Three-dimensional images of the surface and the solid-fluid interface in real space can be obtained in real time, independent of the sample's periodicity. This capability allows in-situ imaging of some dynamical processes taking place on surfaces and at the solid-fluid interface. • Local surface electronic properties such as charge-density waves, the changes of surface barrier and energy gap, as well as spectroscopic images, can be provided by STM performed with related techniques. • STM can be employed for the modification of a surface and for the manipulation of atoms and molecules through tip-sample interactions, opening up the prospects of constructing atomic- or melecular-scale devices.

Fig. I. I. Comparison of the resolution range of STM with that of other microscopes [1.1]. [HM: High-resolution optical Microscope. PCM: Phase Contrast Microscope. (S)TEM: (Scanning) Transmission Electron Microscope. FlM: Field Ion Microscope. REM: Reflection Electron Microscope]

1.2.1 Electron Microscopes The first successful exploration of atomic structures grew out of a basic discovery of quantum mechanics. It is that light and other kinds of energy exhibit characteristics of both waves and particles. In 1927 C.J. Davison and L.H. Germer confirmed experimentally the wave nature of the electron. They also found that a high-energy electron has a shorter wavelength than a 3

low-energy electron. This achievement, together with the establishment of geometrical electron optics, led to the invention of the first electron microscope by E. Ruska and M. Knoll in 1931. Since then several types of electron microscopes have been developed. In SEM an electron probe which has the smallest cross-section of the elctron beam after acceleration, is scanned or rastered over a region of the specimen. The image is displayed on a Cathode-Ray Tube (CRT) deflected in synchronism. The smallest diameter of the electron probe is limited by the minimum-acceptable electron probe current, which is determined by the need to generate an adequate signal-to-noise ratio, and by the spherical aberration of the final probe-forming lens [1.4]. SEM mostly operates in the range E = 10 --;- 20 KeV.I At higher energies, the electron range and the diameter of the electron-diffusion region are larger. Decreasing the electron energy has the advantage that information can be extracted from a volume nearer to the surface, but the diameter of the electron probe increases owing to the decrease of gun brightness. A slowly-moving electron is easily deflected by electrostatic and magnetic fields near the sample surface. Charging effects have to be avoided by coating specimens with a thin conduct ice film, for example, and organic and biological specimens have to be protected from surface distortions by fixation or cryo-techniques. The resolution of SEM is not high enough to resolve atomic structures. In TEM a thin specimen is irradiated with an electron beam of uniform current density emitted from the electron gun; the electron energy is in the range 60 --;- 150 KeV or 200 KeV --;- 3 MeV in the case of intermediate- or high-voltage electron microscopes [1.5]. A two-stage condensor-lens system permits variation of the illumination aperture and an area of the specimen under illumination. The electron-density distribution behind the specimen is imaged with a three- or four-stage lens system onto a fluorescent screen. Electrons interact strongly with atoms by elastic and inelastic scattering. The specimen must therefore be very thin, typically of the order of 5nm --;0.5j.Lm for 100 KeV electrons, depending on the density and elemental composition of the sample and the resolution required. Although TEM has proved to be extremely successful in observing projections of atomic rows and even atomic orbitals in thin crystalline films, it can not resolve surface structures except under very special circumstances. A high-energy electron penetrates deep into the specimen and so reveals little of the surface structure.

1.2.2 Field Ion Microscope In 1951, E.W. Muller made some progress when he invented the Field Ion Microscope (FIM), an instrument that is highly sensitive to surfaces [1.6, 7]. In this technique gas atoms (H2 , He) are field-ionized near the tipshaped sample at a high positive voltage, accelerated by an electric field, then fly to the cathode screen along the direction of electric field lines. Thus a so-called field-ion image corresponding to the atomic arrangements on the tip surface is displayed on the screen. In order to achieve a strong electric field on the surface, the sample must sit on a fine tip, or be prepared in the shape of a tip with radius of curvature ranging from several ten to several hundred Angstroms. This rather complicated sampte-preparation technique, as well as the characteristic of the technique that the sample must be stable against the high electric fields limit the broad application of FIM.

1.2.3 Scanning Tunneling Microscope The principle of operation of the STM makes it possible to avoid the aforementioned difficulties. The main differences between the STM and all other microscopes is that there is no need for lenses and special light or electron sources. Instead the bound electrons already existing in the sample under investigation serve as the exclusive source of radiation [1.8,9]. Conventional STM is based on the control of the tunneling current I through the potential barrier between the surface to be investigated and the probing metal tip. If a small bias voltage is applied between the sample surface and the tip (in the best case, an atomically sharp tip), a tunneling current will flow between the tip and sample when the gap between them is reduced to a few atomic diameters. It takes advantage of the strong dependence of the tunneling probability of electrons on the electrode separation. There it is well known that the tunneling current (at a low bias voltage VT and low temperature) behaves as

z

18~k eff 2kd 10 4 0 d A e-

(1. 1)

where 2k [A-I] = 1.0254>1/2 reV], 4> is the average work function, assumed equal to the mean barrier height between the two electrodes. Aeff = 7r X (l/2L eff )2 is the effective area determining the lateral resolution L eff z 2 X [(R t +d)/k]l/2 which applies when the separation d becomes smaller than the radius R( of the tip. For typical metals (4) z 5eV) the predicted change I The symbol --;- is used throughout the text as a shorthand for "from -to" or "between ". 4

5

(b)

~a)

Vz (Vx• V y) --> z (x, y)

yD

InI (Vx' Vy) --> ,f(p. z (x, y)

Fig. 1.2a, b. Schematic view of two modes of operation in STM [1. 10]. S is the gap between the tip and the sample, I and VT are the tunneling current and bias voltage, respectively, and Vz is the feedback voltage controlling the tip height along the z direction. (a) constant-current mode and (b) constant-height mode

in I by one order of magnitude for the change Lld :::::; 1 A has been verified. If the current is kept constant to within, e.g., 2 %, then the gap d remains constant to within 0.01 A. This fact represents the basis for interpreting the image as simply a contour of constant height above the surface. The STM can be operated in either the constant-current mode or the constant-height mode, as shown in Fig.l.2. In the basic constant-current mode of operation, the tip is scanned across the surface at constant tunneling current, which is maintained at a preset value by continuously adjusting the vertical tip position with the feedback voltage Vz' In the case of an electronically homogeneous surface, constant current essentially means constand d so the topographic height of surface features of a sample can be measured by raster scanning the tip in an x-y scan over the surface and deriving the height of the surface from Vz . The height of the tip z(x, y) as a function of position is read and processed by a computer, and displayed on a screen or a plotter. Alternatively, in the constant-height mode a tip can be scanned rapidly across the surface at nearly constant height and constant voltage V z while the tunneling current is monitored, as shown in Fig. 1.2b. In this case the feedback network is slowed to keep the average tunneling current constant or turned off completely. The rapid variations in the tunneling current due to the tip passing over surface features are recorded and plotted as a function of scan position. Each mode has its own advantages. The basic constant-current mode was originally employed and can be used to track surfaces which are not atomically flat. The height of surface features can be derived from Vz and the sensitivity of the piezoelectric driver element. On the other hand, a disadvantage of this mode is that the finite response time of the feedback network and of the piezoelectric driver set relatively low limits for the scan 6

speed. The constant-height mode allows for much faster imaging of atomically flat surfaces since the feedback loop and the piezoelectric driver do not have to repond to the surface features passing under the tip. Fast imaging is important since it may enable researchers to study dynamic processes on surfaces as well as reducing data-collection time. Fast imaging also minimizes the image distortion due to piezoelectric creep, hysteresis and thermal drifts (Sect.4.2.1). In contrast to the constant-current mode, however, deriving the topographic height information from the recorded variations of the tunneling current in the constant-height mode is not easy because a separate determination of 1/2 is required to calibrate z. In both modes, the tunneling voltage and/or the z position can be modulated to obtain, in addition, information about local spectroscopy and/or the spatially resolved local tunneling barrier height, respectively (Sect.3.2). For a more detailed description of other modes of operation, such as various tracking modes and differential microscopy, the reader is referred to the books [1.11-14].

1.3 Overview

When an atomically sharp tip is scanned across a plane of atoms in the constant-current imaging mode yielding an image with sub-Angstrom corrugations, it is not known exactly from experiments what distance might play the role of d in (1.1). Moreover, since tunneling involves states at the Fermi level, which may themselves have a complex spatial structure, we must expect that the electronic structure of the surface and tip may enter into the equation in a complex way. Equation (1.1), based on analogy with the onedimensional tunneling problem, is rather simple and therefore the full threedimensional tunneling problem as it relates to STM must be considered. These theoretical concepts will be described in Chap. 2. In addition to delineating the atomic topography of a surface STM has made it technically possible for scientists to probe directly the electronic structures of materials at an atomic level by spatially resolved tunneling spectroscopy. Spectroscopy and spectroscopic imaging of STM will be introduced in Chap.3. The lateral resolution of the surface is limited by the sharpness of the tip. What should be the shape of the tip and how is it achieved? How can one avoid mechanical vibrations that move the tip and sample against each other? How can a tip be moved with respect to the sample on a fine scale over long distances? These main instrumental problems and their solutions will be discussed in ChapA. 7

STM has shown that it is possible to control and scan a tip over a conducting surface with Angstrom precision. This same generic principle of STM has been applied to many other novel Scanning-Probe Microscopes (SPM). In Chap.5, the developments in the areas of atomic-force microscopy, lateral-force microscopy, magnetic-force microscopy, bal1istic electron-emission microscopy, electrostatic-force microscopy, scanning ionconductance microscopy, scanning thermal microscopy, scanning tunneling potentionmetry, photon scanning tunneling microscopy, and near-field scanning optical microscopy will be highlighted by results from each area which illustrate the potential of these techniques to provide new information about the physical properties of surfaces on an atomic or nanometer scale. At the early stage of the development and application of STM, mainly surface physicists were engaged since they knew how to built such instruments. In the meantime, STM and related techniques have enjoyed a rapid and sustained growth that is phenomenal for a new branch of science. As of summer 1999, numerous papers and books [1,15-21] on the theories, techniques and applications of STM and SPM have been published, and over 30 companies have manufactured and marketed SPMs and parts of it. Ten international STM conferences have been held since 1986 (Table 1.1). The STM and related techniques have entered many disciplines in physics, chemistry, biology, metrology and materials science. STMs not only allowed the study of surface structures, but also modify surfaces from one micrometer down to the atomic scale. The application of the surface structure of metals, semiconductors, surface adsorption, biological materials and nanofabrication will be outlined in Chaps.6-9. Further details on STM theory and applications can be found in those books and proceedings of STM conferences which have already been mentioned above. Table 1.1. Number of paper presentations at early conferences

Year Date

Location

Country

Number

Ref.

1986 1987 1988 1989 1990 1991 1993 1995 1997 1999

Santiago de Compostela Oxnard. CA Oxford Oarai Baltimore, MD Interlaken Beijing Colorado Hamburg Seoul

Spain USA UK Jap.:m USA Switzerland China USA Gennan)' Korea

59 110 157 213 357 580

12L

July 14-18 July 20-24 July 4-8 July 9-14 July 23-27 August 12-16 August 9-13 July 23-28 July 20-25 July 18-23

464

1.23 1.24 1.25 1.26 1.27 1.28

Fig. 1. 3. Observation of ultra-fine metal panicles in the constant current mode

dT

--..."

/-,

~h I

-_/

//

,_

.----........ ......

~ '~\ partIcle

r--2nm--+--lnm-l

'

_

Substrate

In its brief history, STM has developed into an invaluable and powerful surface and interface analysis technique. However, STM has certain limitations and its operation at the atomic resolution is far from routine. A better knowledge of the role played by the microscopic structure of the probing tip is needed. The size, nature and chemical identity of the tip influence not only the resolution and shape of a STM scan but also the electronic structure to be measured. Although the well-defined tip at the end with one atom can be achieved, a tip with known geometry is not available by a convenient technique, such as electrochemical etching or grounding. The apex of such a tip is also limited by the mechanical stability of th~ tip during the STM scanning process. Moreover, as we can see from the principles of STM, it cannot probe precisely the shape and depth of narrow grooves on the surface in the constant-current mode of operation. This problem is illustrated in Fig. 1.3 for the study of ulrrafine metal particles. The broken line followed by a tip presents a rather narrow separation of particles, and the diameter of the metal particles appears enlarged in the STM topographic Image. Another limitation of the STM technique is its lack of chemical sensitivity although several routes have been proposed to attack this problem [1.29-31 ]. There is already a shift from unreliable instruments to STMs routinely usable by laboratory technicians, allowing researchers to concentrate on the physics of the surface of interest rather than on the instrument itself. The prospect for the future of STM and related SPMs is sure to be as exciting as that of the past decade.

478 826 372

8 9

2. The Tunneling Effect

The tunneling effect originates from the wavelike properties of particles in quantum mechanics. When a particle is incident upon a barrier with a potential energy larger than the kinetic energy of the particle, there is a non-zero probability that it may traverse the forbidden region and reappear on the other side of the barrier; in the classical case the result can be zero (Fig. 2.1). If the lion shown in Fig.2.l is replaced by a wave function of a particle with mass m (Fig.2.2) the tunneling effect becomes more pronounced. In Fig.2.2a, 4>0 is the potential of a simple rectangular barrier and E is the kinetic energy of the particle, and the probability P for the particle to traverse the barrier with thickness d (Fig.2.2b) is given by P cc e- 2kd

(2.1)

,

where k

=

J2m(4)0 - E)/fz 2

.

The physical picture of the tunneling effect is illustrated in Fig.2.2.

Impenetrable barrier

Tunnel effect

Classical mechanics

Quantum mechanics

Fig. 2.1. The difference between classical theory and quantum theory, illustrating tunneling through a potential barrier [2.17) 11

(a)

(b)

:I

1\ !\

0..

~'I

I

I

I

I\!

(b)

.

I

-

p - Chlorobenzoic acid

10

~

6.6

.L: 00 '03 .c

... ......'" co '"

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

f

(b)

cl>3

:...= 2il

Metal B

Fig. 2.9. TRAPSQR barrier used to model the combination aluminum oxide-<Jrganic tunneling barrier. Five parameters have been used, the barrier heights ell I . ell 2 , ell 3 and thicknesses dands [2.6]

(a)

O.lnm Coronene

CyclohexaneCOOH

3.6

oI

)

•..

VA

I

I V.J

V/-O.' 0 case) to get the tunnel current for the tip posiF F tion R. In the LCAO (Linear Combination of Atomic Orbitals) representa-

~////~

\

~/)(E'-E.)/

\

/

vUrY

\

Q

/

r

/

\ \, e·(R)

,\:J

"

e·'(R) /

/

J/

Fig. 2.14. Schematic representation of factors composing the terms of A(R, E, E') [2. 11) 27

26

0

d(A) z

Lx

(0)1

c

~

T' : ' ' ' J t •

,-,

8

~

•

IC

i

I

J

L

I

2.3.2 Tunnel Conductance 7

5

3

i

I

I

(b)., 0.2 1

The derivative of (2.31) with respect to the bias voltage reads

:>

0

dI dV

1\ /]-02;

I

W ,

I

=

...,-0.4 I

I

21re 2 IiA(R, E F -e V, E F ) 2

- 21re Ii

I

J

E -eV F

dE

aA(R, E', E+eV) aE

EF

j

(d)

IE'=E

(2.38)

or dI

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

--

•

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

I

I

~

,

d

dV

=

21re 2 IiA(R, E F , E F +eV)

E -eV

I

Fig. 2.15. (a) Contour plot of constant local density of states(dashed line) and contour of total charge density (solid line). Filled circles indicate the positions of carbon atoms of the top two layers. (b) Potential used for the interaction of tip and surface (schematic). (c) Compress and (d) expansion of graphite for the tip at points A and C of (a), respectively [2.131.

resolution of an STM device reduces with increasing d. In numerical simulation, more and more distant sites are involved in the summation in (2.35) with an increase of d. Then, the LCAO construction of the surface wave function becomes inappropriate. Methods should be developed to calculate the surface wave function in the far-tail region more efficiently. It is worth mentioning here that another important aspect of tip-sample interaction involves forces of various origins, which could lead to variations of the measured corrugations at high tunneling current, and induce photon emissions for some materials. Clark et al.[2.12] carefully examined the tipsample interactions on Cu(lOO) surfaces and concluded that the interaction has atomic site dependence. In addition, the impact of the interacting force on the corrugation measurement is more pronounced on soft material surfaces, such as graphite [2.13] (Fig. 2 .15).

+ 21re

2

aA(R, E', E+eV)

F

1i

JEF

dE

aE' E'=E

(2.39)

If g T , the imaginary part of the Green's function of the tip, does not highly depend on energy, the second term of (2.38) can be neglected. Then the tunnel conductance is given by 21re 2 IiA(R, E F -eV, E ), which stems from F the electronic states of the surface and the tip, where the energy is measured with respect to the Fermi level (Fig. 2 .16). It is in proportion to the surface Density of States (DOS) 9(R, E F -eV). For simple cases, see (2.30),

Sample surface

peR, EF-eV)J----t~ --=--::'~~E~= EF-eV E F .1-:/,,-,-,-,-,..--,

PT (E) P (E)

Fig. 2.16. Offset of the Ferm i level of the surface (E ) and that of the tip (E -eV) [2.111 F F 28

29

dI dV oc

4.0 ,

(2.40)

8(R,E F -eV).

I

(a)

i

I

I

.I"-..

I

30

If a cluster of atoms is used as the model of a protrusion on the tip, the width of the cluster energy levels should be introduced to take into account the embedding effect of the cluster on the remaining part of the tip. Thus the () function in (2.35) is replaced by the Lorentzian function

'

1.0

(2.41)

in the simulation. For a more sophisticated treatment the width ~(Ey) should be determined by the Green's function theory, but for most cases a constant value ~ may be substituted. Kobayashi and Tsukada [2.14] used a W 10 cluster as a model for the (Ill) protrusion of a tungsten tip and simulated STM/STS on a graphite surface. If the width ~ is assumed larger than 1 eV, good overall agreement between the calculated and the observed STM/STS data is obtained. In this case the simulated STM image shows the normal triangular lattice pattern [2.15]. This would mean that the tip has unlocalized electronic states on its protrusion for normal cases. On the other hand, if the width ~ is reduced below 0.1 eV the (dIldV)/(l/V) curve shows rather complicated structures reflecting the discrete levels of the cluster [2.14]. The remarkable point of this model is that various types of the abnormal patterns often observed by STM of graphite are reproduced. Presumably a localized bound states exists on the protrusion of the tip in cases where abnormal STM images are noted. Simultaneous STS measurement should be performed to clarify the nature of the tip state. Recent STS studies revealed characteristic resonance peaks in dIldV versus V curves due to bcc and fcc W (Ill) surfaces, respectively [2.16]. This observation points to the differences in tip geometry effects to image recording and spectroscopic analysis (Fig. 2.17). Similar discussions have been seen in the analysis of imaging atomic structures with an AFM tip in non-contact mode [2.17]. Covalent bonding is proposed to be responsible for enhancing the imaging contrast.

2.3.3 Tunneling Active Orbital at the Tip It is an important problem to clarify what kind of tip orbital dominantly contributes to the tunnel current. Ohnishi and Tsukada [2.18] performed the LDA (Local Density functional Approximation) calculation for the elec-

~

2O

0.0 I ·1.0

~l7r

{)(E+eV-E y) =* (E+eV-E y)2 + ~2

i f'J I

'l

I

!

·0.5 0.0 0.5 sample voltage (V)

I 1.0

15.0

j (b)

~ 10.0 oS

>

J2 '0

5.0

0.0

0.5

1.0

Fig. 2. 17a, b. Experimental differential conductivity for Cu (111). (a) and (b) represent two kinds of spectra that are reproducibly obtained after cleaning the W tip by field emission corresponding to a tip apex with a bcc and a fcc pyramid, respectively [2. 16]

(a)

(b)

Fig. 2. 18. Single-state charge-density contour map for the tunneling active orbital of (a) W 4 (HOMO) and (b) W5 (level just below HOMO) cluster [2. II]

tronic states of the W 4' W 5 and W 8 cluster modes of the tip. The results indicated that tipe states with the large d/ component of the tip's apex atom mainly determine the tunnel current. They are the Highest-Occupied Molecular Orbital (HOMO) for the W 4 cluster and the level just below HOMO for the W 5 cluster. The contour maps of these states are displayed in Fig.2.18. The shape of the tip orbital is sensitively reflected by the STM image. This is, for example, found for the image of C H chemisorbed on a sub6 6 strate, as shown in Fig. 2 .19. The tunnel current is calculated by the level just below HOMO of the W5 cluster. The simulated STM image does not show any distinct shape of individual atoms, but rather the 1r electron cloud swelled over the whole molecule. Contour lines around the neighbouring carbon atoms behave somewhat differently. This is because the orbital of

30 31

,

Fig.2.19. STM current image of C 6 H 6 molecule scanned by a W5 tip. The height of the apex atom of the W5 cluster is fixed at 0.4 nm above the molecular plane [2. 11]

Tip I I I

°1

Surface Fig. 2.20. Geometry of a tip with two protrusions [2.llJ

A(R,E,E') the tip is not axially symmetric, and therefore the two C sites are not equivalent in the whole C6 H6 /W 5 system. With the increase of the cluster size the component of the particular tunneling-active orbital becomes distributed over many levels. The width of these lev~ls governs the features of the STS, as discussed in the previous subsection. For example, if the energy spreading of the tunneling-active orbital is very narrow, the STS spectrum is proportional to the derivative of the surface DOS. On the other hand, if the component of the tunneling-active orbital is distributed over a wide energy region, a simple simulation of (2.40) is appropriate.

2.3.4 Double-Tip and Interference Effects

If the tip has two protrusions contributing equally to the tunnel current, the STM image will be distorted considerably. For distant protrusions, the current image would be the overlap of the two images contributed by the respective protrusions [2.19]. This might correspond to images with ghosts which were often reported in the literature. If the distance between the two protrusions is short, a significant interference effect is expected. To see this we notice that the current contribution A(R, E, E')dE = (E' = E-eV) can be written as

I

=

J

i =1 , 2

OJ

J

dp'V T CP+T)V T (p'+T)Y(Pz;E)y(p;;E)

OJ

X

gS(p+Rj ,p' +R j ;E)g T (p' +Tj,P+Ti ;E')

+

[J

X

dp

O}

J

02

dp'V T (P+T1)V T (P'+T2)Y(Pz;E)y(p;;E)

gS(p+R 1 ,p' +R 2 ;E)gT (p' +T2 ,P+T} ;E') + (1- 2)J .(2.42)

Here, T[, T2 are the relative coordinates to the center of the mass; R} and R2 are the coordinates of these centers-of-mass points with respect to the surface (Fig.2.20). The weight factor used for the definition of the center of the mass is the product of the tip potential VT by J/;p (E-E p) and the decay function y(pz;E) = ex p [- : zY2m IE

I] .

(2.43)

In (2.42) 0i is the spatial region assigned to the protrusion i and gS is defined as gS(p+R j ,p' +Rj ;E)

32

dp

_

gS(p+Rj,p'+Rj;E)

-

y(pz;E)y(p; ;E)

(2.44)

33

The first term of (2.42) is just the simple overlap of the current images contributed by the relevant protrusions. If one takes the zeroth-order terms of the moment expansion [2.20] around the point R j and sums up, we obtain

(a)

(b)

}

r\

'V

V

A(R,E,E') =

L

I

/,

(2.45)

C j (E')8(R j ,E),

\ 1\

i=I,2 as discussed in [2.19]. The above result is just the extension of the theory of Tersoff and Hamann for the noninterfering double-tips case. The second term of (2.42) represents the interference effect. The order of magnitude of this term is estimated as Ainrerference :::: ,,2IvT/2 [gS(R 1 ,R 2 ;E)gT(T2,Tl;E')

+ (1-2)]

" (c)

/,

r\

(d)

.(2.46)

If there is no coherence between the electron wave functions on the two protrusions, the interference term vanishes. This case assumes that the tunneling-active orbital at one of the protrusions does not extend to the other protrusion. _ A numerical simulation of the STM current image of graphite by the antibonding orbital of the H2 molecule, as the tip orbital, distinctively shows the drastic effect of the interference. As depicted in Fig.2.21a, the STM image formed by the tip orbital with an axis in the x direction exhibits an abnormal ridgelike shape running in the y direction. The tip orbital is placed 0.6 om above the graphite surface. Such an unusual image is due to the interference effect of the current components through each tip hydrogen atom, and it is verified by looking at the tilt-angle dependence of the STM image. The images of Fig.2.21b-d correspond to the tip orbital tilt angle 15°,30°,90°, respectively. It is found that the image changes to that of a normal triangular lattice with increasing tilt angle. This is explained by the weakening of the interference, because the current due to the raised hydrogen atom is reduced by the increasing tilt angle. Kobayashi and Tsukada [2.14] demonstrated that the STM image of Fig.2.21a changes drastically when the tip orbital axis is rotated from the x direction parallel to the plane of the graphite surface. This fact is also an evidence of the interference effect, as shown analytically in [2.14]. It is important to note that the STM images (Fig.2.21) are often observed in experiments.

Fig. 2. 21a-d. The STM image of graphite using the antibonding orbital of H as the tunnel2 ing active orbital of the tip. The tilt angle of the axis of the tip orbital is 0 0 . 15 ° . 30 ° and 90° for (a), (b), (c) and (d), respectively [2.11)

34

35

3. Spectroscopy, and Spectroscopic Imaging

In the last chapter, the theoretical concepts of STM were reviewed. The previous description focused on the capability of an STM device to scan a surface and to provide data on its atomic-level topography. This capability is only part of the story; STM also opens up new possibilities in spectroscopy. To first approximation, measurements of the electron density across a surface correspond to the topography of the surface, showing the actual positions of atoms or atomic steps on surfaces. However, this description merely serves as a useful starting point. On an atomic scale, surface atoms are not hard spheres with distinct boundaries. While the STM image usually corresponds to surface corrugations or contours of atoms, STM measurements are actually based on the electron density of states. Electrons exist at specific values of energy, called electronic states. Since the tunneling current reflects the Local Density Of States (LDOS) of the sample's surface, STM can be utilized for atomically-resolved spectroscopy, that is, Scanning Tunneling Spectroscopy (STS). Richer than a mere topographic profile, tunneling spectroscopy provides a wealth of information about which electron energy states are occupied (filled) as well as unoccupied (empty). Detailed information that can be inferred from these measurements includes data on chemical composition, bonding, the energy gap, band-bending effects and adsorption at the surfaces or objects under investigation. The main merits of tunneling spectroscopy performed with an STM device are the following: (i) It is local (here, "local" is equivalent to "atomic ") and can be used to probe the electronic properties ranging from individual adatoms on a surface to spatial properties of vortex states of superconductors; (ii) it can be performed at preselected positions by using the scanning ability of the STM; (iii) it can be performed under well-defined conditions. For example, perfect surfaces with known composition can be examined under ultra-highvacuum conditions; (iv) it can be combined with other methods; and (v) it can provide spectroscopic images.

37

3.1 Concepts of Tunneling Spectroscopies 3.1.1 Solid-State-Barrier Tunneling Historically tunneling spectroscopy with the Metal-Insulator-Metal (MIM) tunneling junctions was first demonstrated by Giaever [3.1]. A common feature of solid-state-barrier tunneling structures is that the application of a bias voltage Y leads to tunneling of electrons with a well-defined range of energies 0 < E < 1 eY. This feature makes possible several versions of classical spectroscopy of the solid electrodes and of the barrier with energy resolution of a few times the thermal energy kT. These versions include spectroscopy of the superconducting state, which probes the details of both the energy-gap structure and the phonon spectrum that produces the pairedelectron system. The second major area of spectroscopy, known as lETS (Inelastic Electron Tunneling Spectroscopy), involves the measurement of inelastic excitations of the electrodes (usually in the normal state) and of the barrier. An example is the threshold for phonon generation in an Esaki diode at a bias of Y = nw/e, which is detected by an accompanying step in dIldY and thus a peak in dI2 Idy2. In a similar fashion, energies of plasmons, of spin waves, of spin-flip transitions of paramagnetic ions. and of a variety of other excitations occurring in or near the barrier in various tunneling structures, have been measured. A related area of particular activity in lETS is the measurement of vibrational frequencies of molecules, including large organic molecules, utilized in a MIM tunneling junction by adsorption to the barrier oxide. Consider an idealized tunneling junction in which the molecules with vibrational level spacing nw are sandwiched between two metal electrodes. If we measure the current as a function of voltage, we will find two components: (i) A steadily increasing current due to elastic electron tunneling, and (ii) a current which has the threshold voltage nw/e and increases steadily thereafter due to inelastic electron tunneling. This threshold is set by the requirement that the electrons must give up the energy nw to excite the molecular vibration. Figure 3.1 shows the total current I which is the sum of the currents through the elastic and inelastic tunneling channels. It has a kink at Y = nw/e which becomes a step in dIldY versus Y, and a peak in dI2/dy2. A plot of dI2/dy2 versus Y is called a tunneling spectrum. Thus, measurements of the spectrum yield a series of peaks appearing at bias voltages simply related to the phonon frequencies by the threshold relation eY = nw. This provides information about the bonding of molecules to solid surfaces and about the interactions or reactions, which adsorbed molecules may undergo. The third class of spectroscopy in junction tunneling is associated with the distribution of electron energy states either in the final electrode or, oc-

....- ...-

Fig. 3.1. The current versus voltage curve has a kink in it when the inelastic electron tunneling channel opens up. This kink becomes a step in the first derivative and a peak in the second derivative

T-r .~~ :

dIldV

o

d'"dV'1 o

I

nw/e

~

nw/e

V

: V

casionally, in the barrier. In classical planar MIM tunneling, the electronic states of a metal electrode may be observed in the current-voltage spectrum characteristic of the tunneling junction. This occurs because the current is determined by the availability of electronic states within the metallic electrodes; specifically bulk states such as the energy band, quantum size effects and surface states with wave functions which are confined mainly to the immediate surface of the metal. A similar dependence of the currentvoltage spectrum characteristic on surface electron structure is expected for the STM geometry. Useful spectroscopic information about energy-band positions in semiconductors and metals can be provided by this class of spectroscopy. Measurements of the current-voltage characteristics of tunneling junctions outlined above have had an important history in solid-state physics. A major disadvantage of conventional planar structures is that the desired spectroscopic information is spatially averaged over the full size of the junction. While junctions as small as 1200 x 3000 A2 have been fabricated, this is still considerably larger than the characteristic length over which the properties of interest might be expected to vary. For the case of superconductors, the relevant length scale is set by the coherence length, which can range from thousands to tens of Angstroms. For resonant or inelastic tunneling, large variations can be expected to occur over a single molecule. Conventional tunneling junctions also suffer from the disadvantage that the interpretation of results can be confused by unwanted effects of the nonideal solid tunneling barrier.

38

39

Fig.3.2. Graph of tunneling currenl vs. distance

4

Fig.3.3. Calculaled vertical lip displacemenl for a Na tip scanning over aNa, S and He alom, respeclively [3.2J

(different bias) and lhe calculaled lunneling currenl 105

(solid lines)

3

.....

"'"

'"

;;; '-

(b)

I

I

2.5

I I

Ca/Na

2.0

.....

1.0

:>

0

~

.':'

~ 0.5

~

"

Q

.......... ,

1.0

0

"""------

-1.0

1.5

0.5

,

0 -2.0

3.0

0 1.0 Energy rei alive 10 Fermi level reV) for Ca

2.0

-0.5 -2.0

'-\

I I

'\J

I I I

-1.0 0 1.0 Bias vollage [V]

2.0

Fig. 3.4. (a) Densily of Slales induced by lhe adsorplion of a Na and a Ca atom on two planar metal eleclrodes, respeclively. The horizontal axis is inverted for lhe Na alom, which is interpreted as the lip. (b) (dIldV)/(I1V) as a funclion of applied bias for the Na-Ca lunnel junction. The full curve is an exaCI calculalion; lhe broken curve is lhe result of a simple model [3.2J

40 41

observed in this spectrum bear a close relationship to the peaks seen in Fig. 3.4a. There are two important implications. First, the tunneling signal contains information that is, to a large degree, representative of the electronic structure of the tunneling electrodes. Secondly, the tip electronic structure is just as prominent in the data as in the sample's electronic structure. This result shows clearly that the microscope does not know which is tip and which is sample, and all wave functions are equal. This is one of the major reasons for the irreproducibility of spectroscopic data. We can conclude from the above discussion that current-voltage spectra may be obtained by varying the tunneling voltage at specific points on the surface. Peaks in d(1nl)/d(1n V) vs V in the low-voltage limit are interpreted as corresponding roughly to resonances in the tip and sample's densities of states, although their exact heights and positions also depend on the gap distance. At higher voltages, however, the current depends nonlinearly on the voltage. Further, electronic-structure effects and inelastic processes may introduce rich structure into the current-voltage characteristics. Alternatively, STM images obtained in the topographic mode at different voltages (described in Chap. I) contain geometric information about the sample (such as the presence of atomic steps), but also contain information on surface electronic structure (such as the presence of localized surface states, band bending, the energy-band gap of semiconductors or superconductors, or work-function variations). Spectroscopic images (or voltage derivatives of images) taken at specific bias voltages may be used to highlight the spatial distribution of particular states. The theoretical results derived from (2.40) for graphite at several bias voltages have subsequently been confirmed experimentally. This theoretical study assumed that the voltage dependence of the transmission coefficient through the barrier can be neglected relative to the voltage dependence of the density of states, and that the density of states at the tip is sufficiently featureless in the low-bias-voltage limit. Of course, the precise relationship between dlldV and the spectra or images depends on the experimental mode employed in taking the data.

3.2 Experimental Modes One can precisely vary the bias voltage and tip-to-sample distance of an STM for measurement at any point. Therefore, an STM device can be used in four related spectroscopy-like modes: (1) Current-voltage spectroscopy (1-V curve), (2) current-separation characteristics, (3) constant-current topography, and (4) current imaging tunneling spectroscopy.

3.2 .1 Current-Voltage Characteristics 1-V curves for fixed separations are obtained by measuring the variation of the tunneling current as a function of voltage at a constant tip-sample separation. This measurement requires the addition of a sample-and-hold circuit to the feedback loop, as discussed in ChapA. This circuit holds the tip position stable for a given time of the measurement and its response time should be fast enough. The feedback loop is interrupted for a few milliseconds by closing the integrator gate without resetting the integrator. Then the tunneling bias voltage is ramped under computer control while the tunneling current is simultaneously measured. In this scheme the sample-tip distance does not change with voltage and the tunneling current is allowed to vanish because the feedback circuit is inactive. The tunneling bias voltage is then reset to its original value and the integrator gate is opened, re-establishing feedback. Typically, a few hundred such current-voltage spectra are averaged to increase the signal-to-noise ratio, and the dlldV curve can be constructed afterwards. Normalizing the dynamic conductance dlldV to the DC conductance I/V can be used to eliminate the dependence of the tunneling signal on the value of the DC tunneling conductance. It is also possible to measure individual I-V spectra at regular intervals during a scan and correlate the I-V characteristics with the scan. Such spatially resolved I-V curves have been used to elucidate the electronic structure of some semiconductor and metal surfaces. For small voltages, < I V, the I-V curves show a linear voltage-dependence characteristic of ohmic behavior. For larger voltage, the exponential dependence dominates the I-V characteristic, and no current density is observed at low voltage for separations larger than 7 A. The implications of this distance dependence on acquiring spectroscopic data are that the data can only be obtained over a limited voltage range for a fixed separation. To achieve a wide dynamic range of tunneling current and conductivity, the tip-sample separation can be varied by adjusting the initial reference tunneling voltage while the feedback is active, in which the tunneling current at low bias voltage is amplified by reducing the separation. Subsequent normalization of the spectrum is performed to remove the effects of varying the tip-sample separation. The I-V curves for the Si(III)-2x I surface [3.4,5] have been acquired in this fashion, i.e. a series of I-V curves is obtained at various fixed tip-sample separations, which are then normalized. Another STM operation mode measures dlldV by adding a small AC component to the DC bias voltage. The frequency w is chosen to be sufficiently high that the feedback cannot respond to it. The tunneling current will contain an in-phase modulation of the frequency w that is the derivative of I with respect to V, at the DC bias voltage V at which the feedback circuit operates. By sweeping V, dlldV can be measured as a function of V. 43

42

Care has to be taken to pick up only the in-phase modulation of the tunneling current with a lock-in amplifier. Displacement currents can give rise to strongly phase-shifted signals. There are three problems with this scheme: (i) The sample-tip separation varies with V, because I is kept constant by the feedback circuit. With increasing voltage and distance, the lateral resolution is degraded; (ii) the DC tunneling resistance is not constant during the measurement, again because I is kept constant. At decreasing bias voltages the DC conductance increases (by decreasing the tip-sample separation) and diverges as the zero bias voltage is approached. Likewise dl/dV diverges. In practice, useful data have not been obtained below 1 V, which for many structures is the most interesting region; (iii) this approach requires that a tunneling current can be maintained over the voltage range of interest. This is not always possible. If there is a band gap, a stable tunneling current can often not be maintained if the bias voltage is set within the gap, and the tip will crash into the sample. Nevertheless, this method has been used to study the electronic structure of surfaces such as Si (111)-7 x7, etc. The third method for improving the measuring procedure is to first set the bias voltage at a relatively large value, typically ±2 V, for inducing a sizeable tunneling current while the tip is stabilized by the active feedback loop. Then the feedback is disabled, and a V-shaped voltage is added to the Z piezo voltage, synchronized with the bias ramp. The tunneling current as a function of the bias voltage is recorded. Using a modulation frequency of about 1 kHz with a lock-in amplifier leads to a measurement time for the spectrum of a few seconds. Thus, this method is only possible for a very stable mechanical system and a very long holding time of the Z position. Generally, the residual thermal or piezoelectric drift rate of the STM should be less than 0.1 A/s. An example of this type of data is the spectroscopic measurements on a monolayer of Sb on the GaAs(l10) surface [3.6).

3.2.2 Current-Separation and Separation-Voltage Characteristics Equation (1.1) may be considered a first approximation for the tunneling current. Making use of the variables in the equation, the local work function or, correctly speaking, the effective tunneling barrier height between the tip and the sample can theoretically be derived from tunneling current I(s) at constant V or from the separation s(V) (we use s instead of d to represent the tip-sample separation hereafter) at constant I. Experimentally, d(1nI)/ds can be measured in STM experiments by a modulation of the gap separation with phase-sensitive detection of the current at the modulation

The modulation signal divided by the current is, neglecting logarithmic terms in s, - d(lnl)/ds

= V + (lIzV .. ~~ •.O~~.'. ~ 1'5";0. .•, w• • ~~~~~~W~". ~ \SJ

..

'·.~·~."dl~ ~~~;,~ .~.,•• ~ .~ ~ fj.~ .""

•. ~ ~ ~ .....-

,.~. •,~I# .~

(a)

.....

'

~

••®.®. •• •••

0

•

Adatom ReSlalom layer Second layer Bulk layer

Fig. 3. 8. (a) A schematic view of the DimerAdatom-Slacking (DAS) model in the 7 X 7 symmelry. (b) Topographic image obtained at +2 V. (c) Adatom Slale al -0.35 V. And (d) dangling-bond state at-0.8 V [3.18] 53

!

~~(;~~~--j o

20

Distance

(c)

.~

.0

• 0

-

E!'

50

75

100

125

Distance (A)

(J)

c

W

Fig. 3. lOa, b. Distribution of the confined electron states of a 36 AAu (Ill) terrace. (a) 3D map of dIldV as a function of the distance perpendicular to the steps and of the vollage appl· ied to the sample(energy). The bias ranges from -0.47 V (bottom) to +0.38 V (top). (b) Individual dIldV line scans for different sample voltages. The dOlled line is a constant-current STM Iine scan, step edge peaks are marked by S [3.26]

monatomic step sites [3.26]. In this case, the steps sites are served as the dimensional restrains and resulted in the energy levels for electronic states (Fig.3.1O).

55

3.4.2 Adsorbate-Covered Surfaces Measurement of I-V curves at constant tip-sample separation has found wide application in spatially-resolved studies of adsorbate-covered surfaces. The most notable cases are the study of the interaction of NH 3 [3.27], Cl [3.28] and 0 [3.29] with Si (111) and the study of localized atomic scale defects. STM images of a clean Si(111)-7x7 surface clearly show two inequivalent types of surface Si adatoms, called "corner" adatoms and "center" adatoms. Upon exposure to NH" individual adatoms disappear from the images as their dangling bonds are passivated by reaction with NH 2 or H. Curves of d(lnI)/d(ln V) were measured on a dense grid, allowing a detailed study of the interaction (Fig. 3.10). Spectroscopic measurements provide further insight into charge transfer and energy-level shifts associated with the reaction. It is found that the so-called rest-atom sites are the most reactive of all. Reaction with H gives rise to significant charge transfer to neighboring adatoms, which react next. Results from the Si(111)-(V3 x V3)AI surface are exhibited in Fig. 3.11. Images in Fig.3.11a,b were obtained with +2 V and -2 V bias, respectively, and show the (V3 XV3) lattice, with some defects. The defects appear as reductions in Fig. 3. 11 a and as elevations in Fig. 3.11 b. Figure 3.11c displays d(lnI)/d(ln V) curves taken on a regular lattice position and on a defect position. In a detailed comparison with theory, Hamers et al. [3.30] demonstrated that the regular lattice positions correspond with an Al atom adsorbed on the top of three-layer Si atoms. The three Al valenceelectron pair with the three Si dangling bonds. These back-bonds are imaged in Fig. 3.11 b. An empty P z orbital protrudes into the vacuum and is imaged in Fig.3.11a. The defects are due to Si atoms substituting for Al atoms. Si has one more valence electron than AI, and the Pz orbital is therefore not empty, giving rise to the extra electronic state at -0.5 eV seen in Fig.3.11c. In the STS measurements of oxygen adsorption on GaAs (110) J .A. Stroscio and co-workers observed both short-range effects due to adsorbateinduced changes in the local density of states as well as long-range effects due to surface band bending. It was noted that the oxygen-GaAs complex displays a positive STM contour when sampling the filled electronic state, while a negative STM contour is observed when sampling the unoccupied states. This striking voltage dependence observed in the STM contours results from the energy dependence of the local density of states. The oxygen data verified the prediction from a theoretical analysis based on a negatively-charged adsorbate. The negative charge can be associated with a filled surface acceptor state on p-type material. The oxygen appears neutral in the

56

--;;;;-"" ~

~

-2

-1

0

1

2

Energy reV]

-2

-I

0

1

2

Energy reV]

Fig. 3.11. (a) Empty-state topography of the Si (111)-7 X7 surface. The spatially resolved tunneling spectra shown below were obtained while in the positions indicated by arrows for the restatom site (A), an adatom next 10 the corner hole (B) and a center adatom (C). (b) Empty-state topography of the Si (111)-7 X7 surface after reaction with NH 3 . Curve A was laken over a reacted restatom, curve B (broken) over a reacted corner adatom, curve B (solid) over an unreacled corner ada 10m and curve C over an unreacted center atom. The reaction leads to readily observable changes in local electronic structure at reacted sites; compare spectra B (left) and B (right, dashed line). In addition, the effect at unreacted sites as a result of a reaction at neighboring sites is apparent; compare spectra C (left) and C (right) [3.27]

images, indicative of the acceptor state being unfilled, which agrees with the lack of band bending observed in spectroscopic measurements.

3.4.3 Superconductivity One of the many promising applications of the combined STM and spectroscope is in the field of superconductivity, especially since there is the possibility of spatially resolving the energy gap in superconductors on the scale of a few Angstroms. Real-space images of variations of the superconducting energy gap have been obtained from Nb foils, NbN films, Nb 2 Sn, Nb 3 Sn, organic su57

perconductor (BEDT-TTFh Cu(NCSh and a number of high-T c oxide superconductors. The comparison of the surface topography and gap distribution permits the detection of possible correlations between local variations in superconductivity and topographic features. One example of the applications of spectroscopic methods to the surface electronic states of superconductors is to clarify the origin of the conductance plane of Bi 2 Sr 2Ca . CU 2 8 , From the results obtained by resonant photoemission (PES), Inverse PhotoEmission Spectroscopy (IPES), X-ray absorption and electron energy loss spectroscopies, it has been proposed that the electronic states at the Fermi level consist of mainly 2p atomic orbitals with x-y (in plane) symmetry. A critical problem in modeling the high-T c mechanism based on these experimental results is which plane in the crystal, BiO or Cu0 2 , provides the 02p orbitals at the Fermi level. Figure 3 .12a shows the normalized differential conductance curve [(dIldV)/(I1V)] of the surface of Bi 2 Sr 2 Ca' CU 2 8 , This curve has a gap of about 0.3 eV at E F , suggesting a nonmetallic nature of the BiO plane (STM and TEM studies of a cleaved surface of a single-crystal sample indicate that the topmost cleaved surface is a BiO plane). Figure 3.12b displays the PES and IPES spectra of the same sample which represent the occupied and unoccupied electronic density of states, respectively, the PES and IPES spectra reveal a clear Fermi edge leading to a substantial density of states at E F , taking into account the effects of energy resolution: 0.2 eV for PES and 0.4 eV for IPES. Considering the differ-

°

°

81

f

~

I

I

AI

AI

:::'0 ;:.

-2

-1

~ 6

--

~

~

°

1

2

1

2

Si

4

°. 2 (c)

·1

0 Energy leV]

3

I (a)

Fig.3.13. (a) Tunneling spectrum of a cleaved surface of a Bi 2 CaSr2 CU2 08 single crystal. The STS spectral intensity (dIldV)/(lIV) represents the local density of states and (dIldV)/(I1V) = 1 corresponds to zero density of states. (b) Photoemission and inverse photoemission spectra in the vicinity of the Ferm i level (E F ) [3.31)

STS

;;-

--

-'

;:'2

> '0

--

§

IL

~)

~

'c::l

-e '"

~

'Vi c:

;":

- l= -1.0

EF

1.0

Energy relative to E F leV]

ence in probing depth between the photoemission (probing a few atomic layers) and the tunneling spectroscopy (probing a topmost atomic layer), the tunneling spectrum presents the electronic density of state of the BiO plane, while PES and IPES spectra probably correspond to the electronic density of states involving both the BiO and the Cu0 2 planes. Thus, Tanaka et al. [3.31] concluded that the BiO planes have a nonmetallic nature and the Fermi liquid states observed by PES and IPES are spread on Cu0 2 planes. A Bi 2 Sr 2 CaCu2 08 crystal may be composed of alternating stacking of metallic (Cu0 2) and nonmetallic (BiO) planes, as sketched schematically in Fig.3.13. This mode is also supported by resonant Raman spectroscopy data. In addition, an energy gap of 20 meV was identified in association with CuO chains in YBa 2 CU 3 7 , The origin was related to either CDW transition or proximity coupled superconductivity [3.32]. Using STM has enabled researchers to find the quasiparticle bound state in type-II superconductor 2H-NbSe 2 [3.33]. It has been further proposed that STM could be used to probe the impurity-induced resonances and the superconducting order parameter symmetry [3.34].

°

Fig.3.12. Si (111)-(''; 3 X V 3)AI: (a) empty-state image; (b) filled-state image; (c) (dIldV)/(IIV) taken on an AI adalOm (top) and a Si adalOm (boltom) [3.30) 58

59

3.4.4 Outlook As outlined above, the capability to measure the spatial variation of the tunneling spectra with the resolution of STM has proved especially fruitful. However, there are still some problems to be resolved in the future.

a) Influence of the Tip The size, shape and chemical identity of the tip influence not only the resolution and shape of a STM scan but also the measured electronic structure. It was shown that the wave functions of well-prepared, clean and stable tips (for instance, prepared by FIM) are apparently sufficiently featureless so as to be indistinguishable in the data. However, such well prepared tips have not been utilized in most STM and STS experiments. In general, the tip DOS does not resemble that of a free-electron metal. To have meaningful STS measurements, the tip DOS must be determined independently. Two methods have been proposed for determining the energy spectrum of a sharp metal tip. For a free-electron metal tip, the Field Emission Spectroscopy (FES) is described by the Young formula [3.35]. A deviation from the formula indicates a deviation from a free-electron metal behavior [3.36]. The other method is known as the marching method, measuring the dynamic conductasnces on the same sample using a free-electron-metal tip and another tip of unknown DOS [3.36].

b) Interpretation of Spectroscopy Results There is no sure and simple way to understand spectroscopic information. For example, the image of the Si (111)-("v'3 XV3) Ag surface contains two maxima in the unit cell, with the maxima arranged in a honeycomb structure. Loenen et al. [3.37] argued on spectroscopic grounds that these have to be Si atoms, but Wilson et al. argued on the basis of essentially identical data that the maxima must be Ag atoms [3.38]. The simple fact suggests, no matter how many I-V curves one measure, that STM is not going to tell us whether the bumps are Ag or Si.

-

.~

'"

U

'0

..--°2

SrO

E

Cu02 Ca

;0

Cu02

'c

SrO Bi ___

~

a.

::J

Bi"--

60

I

Bi ___ Bi

} ) O

2

~

The other note of warning is that, although the (dIJdV)/ (IN) (or d(ln 1)/ d(ln V» corresponds closely to the sample density of states, it is not related directly to the density of states of both sample and tip. Remember that the tunneling current is also determined by the tunneling transmission probability density. If the sample has a large density of states, but these states do not overlap with the tip, these states are inaccessible in a tunneling experiment. One example may be the d band on Ag. In a photoemission experiment, the d orbitals will be the dominant feature in the electron energy distribution curve. However, the d orbitals are too small to be visible in a tunneling experiment. In addition, one has to be careful that the apparent location of an electronic orbital is not the same as the location of the atomic core. In most cases it would be difficult to figure out what is going on without the help of theory [3.39]. In order to fully understand the STM results, Ding et al. [3.40] have investigated this surface using first-principles total-energy calculations. The lowest-energy configuration consists of a top layer of Ag atoms arranged as honeycomb-chained trimers lying above a distorted "missing-top-layer" Si (111) substrate. It is found that the honeycomb structure observed in STM images arises not from the top-layer atomic positions but rather from the electronic charge density of an empty surface band near the Fermi level. The maxima of the electronic distribution for the empty states occur at the center of the Ag trimers, and are situated over the fourth Si layer, in agreement with the result obtained by Wilson and Chiang [3.41]. In this chapter we have reviewed the concept of tunneling spectroscopy and some ways to obtain spectroscopic information with STM. In spite of some limitations mentioned above, scaning tunneling microscopy and spectroscopy will continue to make many important contributions to surface science in the years to come.

_7.

-~ Non·metal

Non-metal

/

~Non-metal

Fig. 3.14. Schematic view of alternating metallic Cu02 and nonmetallic BiO planes derived from Fig. 3. 12 61

4. STM Instrumentation

Even though the principle of a scanning tunneling microscope is not very complicated, many factors must be taken into consideration in the design to ensure a stable and reliable instrument. Optimum functioning of an STM device requires tip-to-sample position control with picometer precision, a rough and fine positioning capability in three dimensions, a scanning speed as high as possible, and also, preferably, simplicity of operation. These requirements have to be satisfied in the presence of building vibrations with up to micrometer-size amplitudes, electric noise, thermal drift, creep and hysteresis of piezoelectric translation elements and other perturbations. Some design rules are discussed in this chapter with emphasis on the vibration isolation, stiffness, electrical control circuitry, tip preparation and computer control system.

4.1 The Vibration Isolation System The frame of the instrument is always subjected to vibrations transmitted from the ground or the air. As indicated in (1.1), the tunneling current depends exponentially on the gap between tip and sample, even the smallest vibrations, such as those caused by sound in air and by people walking around a building, can affect the stability of the instrument. For many materials, especially metals, the atomic corrugations observed in the constantcurrent STM mode will typically be 0.01 nm. Therefore, a good vibrationisolation system is very important for a well-functioning STM, and the changes of the gap distance caused by vibrations must be kept less than 0.001 nm [4.1]. There are two types of disturbance from which an STM device must be isolated: vibration and shock. Vibrations are generally repetitive and continuous, while shock is defined as a transient condition whereby kinetic energy is transferred to a system in a short period of time. Of them, vibration isolation is most critical. Machines running at or near laboratories excite vibrations of the building at frequencies typically between 10 and 100 Hz. The frame, walls and 63

floors of a building undergo mostly shear and bending vibrations which usually resonate at frequencies between 15 and 25 Hz. Vibrations originating from ventilation ducts, transformers and motors are at frequencies between 6 and 65 Hz, while those caused by people walking and working in laboratories typically are in the range of 1 to 3 Hz. Hence, in the process of minimization of the sensitivity of an STM to building vibrations, primary attention has to be given to the frequency range between 1 and 100 Hz. Increasing the inherent resonance frequency of the STM body and employing a vibration damping system are two major ways to isolate vibrations. Viton, metal springs and magnetic eddy-current damping can be utilized for damping components. For an UHV-STM assembly a widely used vibration-isolation scheme which adopts these components is illustrated in Fig.4.1. The vibration isolation consists of two sets of stages which, suspended from springs, nestle within the stainless-steel cylindrical frame of the microscope. The inner stage slips into the outer stage, from which it is suspended by three springs. The inner stage carries the multi-stacked metal plates separated by Viton spacers. The heart of the microscope containing both the scanning tip and the sample sits on the top of the plates. When the

entire microscope resides in vacuum, air resistance is minimum, and the outer and inner stages could, if they were disturbed, bounce up and down almost indefinitely. To stop this motion, permanent magnets mounted on the bottom of the outer stage slide between copper plates attached to the support frame and the inner stage. As each plate slides up and down, the magnetic field induces an eddy current. The interaction between the eddy current and the magnetic field retards the motion of the plate and thereby damps the STM. The Viton spacers between the metal plates are used mainly for the purposes of damping the vibrations propagating along the springs and reducing the effects caused by a large-amplitude shock, although they exhibit pronounced creep. They are substantially incompressible. When strained under compression, their compression stiffness (equivalent to the spring constant) is often large, resulting in a rather high (> 10 -:-100Hz) resonance frequency. However, they have inherent damping of a tenth or a hundredth of the critical damping. Metal springs can have smaller spring constants, resulting in low resonance frequencies (5 -:- 0.5 Hz), but they provide little damping and require additional damping components. The eddy-current damping systems, on the other hand, can easily have varied damping coefficients and be compatible with an UHV system. The eddy-current damping coefficient [g/sJ is equal to 'Y = 9.69' 10- 9 B2 eslp ,

D

Stainless steel

CJ

Magnet

~ Copper

I!'JI

First stage 64

Viton

Fig. 4. l. Schematic diagram of the assembly of a STM instrument and its vibration isolation system.

(4.1)

where B is the magnetic field [GaussJ, e is the thickness (in the direction of B) of the conductor [cm], S is the cross-sectional area [cm2 ], and p is the resistivity [0' cm]. In general, the eddy-current damping with permanent magnets is useful only in the case that the mass and natural frequency of the system are relatively low. Commercially available vibration-isolation systems employing pneumatic springs are often used as the support for a STM. Some of them can have a resonant frequency as low as 0.5 Hz with a variable damping constant. The more rigid the construction of an STM is, the fewer requirements it has for the external vibration-isolation system. Hansma et al. [4.2J described a vibration isolation system that consists of only a concrete block hung from the ceiling with rubber tubes which are at the same time springs and dampers. The very rigid STM unit is placed directly on the concrete block. Vibrations are dissipated by hysteresis losses due to the inherent structural damping of a rigid body. In the construction of a rigid STM body, the resonant frequency of piezo drivers as well as joints tightened by screws, epoxy junctions, three-point contacts and loose spring connectors will have serious effects on the resonant frequency of the constructed STM and the quality factor Q' of the tip-sample junction. 65

The vibration amplitude transfer expression proposed by Kuk and Silverman [4.3] for a spring-viscous damping system under the condition of structural damping P~ » Po (p~ being the resonant frequency of a rigidly constructed STM, and Po is the resonant frequency of a vibration isolation system) is given by

T

=

1 + (~plpo)2 ] 1/2 [ (1-1' 2 /1' 2)2 + (2~plpo)2 [(1-1'2 0

(plp~)2

Ip~2)2 + (plp~Q')2]112 .

(

4.2

)

The first term represents the amplitude-transfer term of a spring's viscous damping system and the second term is the damping-transfer function for a hysteresis loss due to the inherent structural damping of a rigid body. In addition, I' is the external exciting frequency, ~ (= y/y c) is the damping ratio, y is the damping coefficient of the system, and Yc (= 4m7rp o) is the critical damping coefficient. The amplitude transfers calculated for four STMs are shown in Fig. 4.2. The solid line corresponds to a system with Po = 2 Hz, P~ = 2 kHz, ~ = 0.4, and Q' = 10. With a floor-vibration amplitude of a few thousand Angstroms, the gap stability for this system will be better than 1 A. For a very rigid STM assembly (shown by a dotted line in the figure where P~ = 12kHz and Q' = 50), the amplitude transfer is still too large (higher than 1 A. at 200 Hz) for a stable STM junction. With the addition of a vibrationisolation table (po = 1Hz and ~ = 0.4), the amplitude transfer (shown by 10°

"

,,

10- 2 ~

~ '"c::

g "0 '" i= .

10- 4

./' ./"

10- 6

"

E

-

..... .-.-

./

.-

(b) PZT laminate Contraction

f:lW (c) PZT stack assembly . - - - - Insulator

(e) PZT tube

I~+

Fig. 4.3. Several shapes of PZT piezoelectric ceramics 69

ai, expressing the ratio of strain developed along one axis to the electric field strength along the same or a different axis, assuming all external stresses are constant. The first subscript of d indicates the direction of the applied electric field and the second is the direction of the induced strain. The strain coefficient d is negative representing contraction perpendicular to the field; and positive for stain measured along the 3-direction (along which the thickness t is measured) representing expansion parallel to the electric field direction: (4.4)

Lit = d 33 V .

Although there are many ceramic compositions used today, most can be placed into two general categories: hard and soft PZT materials. Typical d coefficients for hard PZT materials are

Stack extension: d 33 Nd V

where Nd is the number of disks. As implied by the equation above, large displacements can be achieved by a small stack as long as each disk is sufficiently thin. As indicated in (4.3) for a strip geometry, the largest extension per one volt is obtained when the ratio elt is maximized. Unfortunately, making very thin strips results in actuators that are compliant and prone to buckling when loaded lengthwise. To improve their strength and rigidity, strips can be bonded side by side with opposing poling directions to form laminated structures (Fig.4.3) with the same extension per one volt as the individual strips. The changes along its length of such a laminated structure can be expressed by Lix

d 33

250.10- 12 m/V, d 31

=

= -

110.10- 12

600'10- 12 m/V, d 31

=

- 270 .10- 12 mlY .

d 31 = - 273.10- 12 mlY .

A bimorph represents a sandwich structure with a thin metal strip bonded in between two PZT strips, as shown in Fig.4.3. When held at one end in an external electric field, bimorphs bend with a displacement at the other end, reaching 100 14m due to the lengthwise expansion or contraction of the PZT elements. However, such devices lack mechanical strength and inherently have low resonant frequencies. Increased mechanical strength is possible by supporting the bimorph at both ends and utilizing the displacement at the center of the strip. In this case, the expression becomes Lix

=

(3/8)d 31 V(elt)2 .

(4.5)

It is apparent from (4.4) that the change in thickness of a PZT material in any of the three directions is only a function of the voltage applied. No matter what thickness, the absolute extension Lit depends only on the voltage V for a given d constant. Therefore, to obtain a larger displacement in the direction in which either t or e is measured, it is necessary to assemble a stack of disk-shaped or strip piezoelectric elements, as sketched in Fig.4.3. The total extension is thereby the sum of the individual elements and given by

70

d 31 Vee/t) .

(4.7)

The choice between a strip or laminate assembly for a particular application will be determined by the need for the load-bearing capability if a thin strip is required to achieve sufficient extension. A tube with electrodes on its inner and outer surfaces is a very common structure for a single piezoelectric element, and the poling direction is right through the wall. Consequently, the axial extension is given by

For PZT-5H d 33 = 593.10- 12 m/V,

=

m/V;

and for soft PZT materials d 33

(4.6)

ve Lie

=

d31 OD _ ID

(4.8)

where e is the length along tube axis, OD and ID are the outer and the inner diameters of the tube, respectively. The radial extension is Llr

=

d 33 V .

(4.9)

A tubular piezoelectric element is required to have higher uniformity of the wall thickness. Otherwise, unequal wall thickness around the perimeter of the tube will cause a bending motion as one side of the tube extends farther than the other.

4.2.2 Three-Dimensional Scanners The 3-dimensional scanner is a device that moves the tunneling tip across the sample surface (x,y) and controls the tip-sample separation (z). BasicalIy, the designs of 3-dimensional scanners made of piezoelectric ceramic materials include the tripod, the single tube and a cross combined with a 71

(3)

Jf

(b)

________l z offset voltage

----r

Teipod

controls tip height

z Electrode

Common

z + YElectrode

_ y Electrode

+ YElectrode

+ y Offset

- x Electrode (hidden)

x Offset

+ x Electrode + x Electrode

x

Fig. 4.5. lIJustrating the voltages applied to the electrodes of the single-tube scanner

x

+ Fig.4.4. Common three-dimensional scanners: (a) tripod, (b) single tube, and (c) a cross combined with a single tube

single tube. The earliest design of a scanner is a tripod of three orthogonal piezo "sticks" (Fig.4.4a). The STM tip is placed on the top of the tripod, and the independent extension and contraction of the three "sticks" make the tip move along the x,y,z directions. The response of the tripod made of three PZT-5H piezo sticks being 13 mm long, 2 mm high and 2 mm wide is generally 1.5 nm per volt in each direction, and the maximum mechanical resonant frequency is 5 kHz. Replacement of the "sticks" by tubes can reduce the cross talk and raises the resonant frequency. As a simple piezoelectric element, the big advantage of tubes over strips is their structural rigidity. A compact, single-tube scanner was first employed in an STM by Binnig and Smith [4.5]. The outer electrode of a PZT tube is divided into four equal quadrants, the inside electrode being not separated. By applying different voltages to opposite quadrants, the tube is made to bend perpendicular to the tube axis. Orthogonal x-y motion is obtained by controlling voltages on quadrants 90 degrees apart, while the contraction of the whole tube along the tube axis by applying a voltage on the inner electrode leads to the movement in the z direction. The other two outer electrodes can be either at ground, or a bias DC voltage can be applied for the purpose of scan-window selection. Figure 4.5 illustrates several schemes for applying voltages onto the electrodes of the tube scanner. The tube scanner exhibits large displacements, a low cross talk and a high resonant frequency. The resonant frequency [Hz] of the single-tube design can be estimated by 72

f r = 1.08·10 5 e- 2

Jro 2 + r2 j

(4.10)

where r i , r a are the radius of the inner and outer wall, respectively, and e is the length of the tube in cm. For a PZT-5H tube of 12.7 mm in length, 6.35 mm in diameter and 0.51 mm in wall thickness, the response in the x, yand z directions is 5 nm/V, and the inherent resonant frequencies parallel and perpendicular to the tube axis are 40 and 8 kHz, respectively. This design has become popular owing to its small size and high resonant frequency. Unequally divided outer electrodes of the tube, unequal wall thickness and the deviation of the tip position from the center of the tube axis will result in nonorthoganl x-y motion. Therefore, only accurately manufactured piezoelectric ceramic tubes can be used for the scanner. There is another design of an x-y scanner employing a cross-shaped piezoelectric ceramic material. The motion in the z direction is accomplished by a piezoelectric tube mounted at the center of the cross. The controlling voltages with the same magnitude but opposite signs are applied to the x,-x and y,-y electrodes, respectively (Fig.4.6). This design with sym-

Fig. 4.6. Illustrating the voltages applied to the electrodes of a cross single-tube scanner 73

metrically arranged piezo blocks has been successful in reducing the effects of thermal drift. The primary design goal is for a scanner to be as rigid as possible for a given scan range. Specifically, the figures of merit are the resonant frquencies of the scanner in each orthogonal direction. It is important to have the resonant frequencies high, not only because they fix the scanning-speed (feedback performance), but also because they determine its rigidity against vibration. Other requirements of a good scanner are high resolution, orthogonality and linearity - the amount of movement should be proportional to the applied voltage. The problems of nonlinearity, hysteresis and creep become more serious for large area scans, where the electric field is large. These effects can be minimized by controlling the total charge applied to the piezoelectric rather than the voltage [4.6], by taking an image of a standard sample to find the amount of distortion and re-map the image with software in a way that removes the distortion, or by using independent position sensors for the x and y axes, such as an optical sensor or a capacitance probe [4.7,8].

age between MF and GP, and are released when the voltage is removed. Elongating and contracting the body of the louse with the appropriate clamping sequence of the feet moves the louse in any direction in steps between 10 nm and I !-,m, and up to 30 steps/so In this way, sample and tip can be approached vibration-free to within the working range of the piezo drivers to avoid accidental contact of tip and sample. The rough drive also serves for separating the sample (ca. I cm) in the cleaning procedures and to compensate for thermal expansions when working at elevated temperatures. Burleigh Instruments has developed a unique piezoelectric linear motor called the Inchworm Motor. The principle of operation is depicted in Fig. 4.8. The Inchworm Motor consists of a piezoelectric ceramic tube and a shaft that closely fits the tube. The outer electrode of the tube is divided perpendicularly to its length into three PZT elements. Two of them serve as clamps and a center one is the PZT actuator that does not touch the shaft. When a voltage is applied to the end element 1 or 3, it is made to clamp tightly to the shaft due to the contraction of the circumference of the tube governed by d 33 , and to release when the voltage is removed. When a voltage is applied to the first PZT element, it clamps the shaft. Then, a vari-

4.2.3 Coarse Sample Positioning The z-piezo range is limited in the STM for which the recognition of individual surface atoms is the key design objective. Therefore, a means of reducing the gap distance between tip and sample to within the working range of the z piezo, known as the coarse-approach system, is required. Such a device also needs to be able to move the sample far enough away from the tip to allow sample transfer. A coarse approach to such a device is called "louse" sketched in Fig. 4.7. It was used in the early STM designed by G. Binnig and H. Rohrer at the IBM ZOrich Laboratory. The louse (L) body consists of a piezoplate with a sample holder on top and resting on three metal feet (MF), separated by high-dielectric-constant insulators from the metal ground plates (GP). The feet are clamped electrostatically to the ground plate by applying a volt-

I

+ 100 -IOOOY

Extend 2

74

I

Cf ==l:=J L..F

I

I

•

I

3 JD § JD ~

Clnmpll

~

I

I

p

Contract 2 «

Unclamp 3 Fig. 4.7. Sketch of the "louse"

I_.,...=::=~=!....

Clnmp I

Unclnmp I

~GP

l~

off

Clnmp 3

2

~

I

I

I

I

I

Fig. 4. 8. The principle of operation of the piezoelectric Inchworm Motor 75

able-rate staircase voltage is applied to the center PZT element causing it to change length in discrete steps of approximately 4 nm each. The staircase may be stopped or reversed at any step. At the end of the staircase a voltage is applied to the third PZT element causing it to grip the shaft. Then, the voltage is removed from the first PZT element, releasing the shaft. The staircase starts downwards until it reaches its lower limit, at which point the first PZT element is activated again, the third PZT element released, and the staircase starts again. If the center PZT element is fixed to a support, the above manipulation will make the shaft move along one direction. This sequence can be repeated any number of times or reversed to make the shaft move in the opposite direction. The total travel distance of the shaft is limited only by the length of the shaft. By employing piezoelectric ceramic materials with different stress coefficients, changing the amplitude and frequency of the voltage applied, the precision and speed of the shaft motion can well be controlled. The Inchworm Motor has successfully been utilized in the sample coarse-approach system of the STM. There are several factors to be considered in selecting a sampleapproach mechanism: reliability, geometry, rigidity and speed. The selection of any kind of sample-positioning device should be based upon the unique requirements of the particular microscope design. Three useful techniques for coarse sample positioning are: •

• •

A walker which moves by piezo expansion and using electrostatic, mechanical or magnetic clamping. Both the louse and the Inchworm Motor belong to this category. A differential screw micrometer pushing on a reduction lever. The differential screw can be driven either manually or by a stepping motor. A lead screw pushing against a differential spring.

Each design has its own advantages: the first design is often used in vacuum chambers, the second in the atmosphere, and the third at low temperature. Lever or differential screw mechanisms have been employed for a fast approach and easy manipulation of the sample holder, which can usually be detached or flipped away from the STM assembly.

4.2.4 STMs for Operation in Various Environments Due to different three-dimensional scanners and sample coarse-approach mechanisms employed in STM, hundreds of types of STMs have been designed to date. The smallest STM is only 1000 x 200 x 8 ~m3 in size; the maximum scanning range can reach 200 ~m. Some of them are designed for operating in an UHV chamber equipped with standard surface characterization tools such as AES, LEED, HM, SEM, etc. (Chap. I). 76

(b)

(3)

Rod connected to linear motion feedthrough Coarse adjust beam (B) .....--'"11'""

Steel ball (B)

I

Slide (S)

I

;'

'II

I

I

I I

Flexible joints Shifter (Sh) connected to linear motion feed through

I

Piezo-tube for z motion

I

t . = ;:::l I

Tunneling tip Cantilever support plate (Csp)

I llr J

-tjf;1J~ UI 'CI

Cantilever Spring loaded screw (Sc) Fig. 4.9. The mechanical adjusting mechanism of the low-temperature UHV AFM/STM

[4.9J

Giessibl et al. [4.9] built an UHV atomic force/scanning AFM/STM which can be operated at 4.2 K. The microscope, of only 20 x20 x70 mm 3 in size, is incorporated into a very small chamber (100ml) which can be evacuated and baked to UHV within a few hours via a specially designed valve. Figure 4.9 illustrates the principle of the AFM/STM system. Figure 4.9a is a sketch of the approach mechanism for the z distance between sample and AFM cantilver. The slide (S) carrying the sample is pressed onto a steel ball (B) and an axle (A) by a spring (Sp) which acts on a ball bearing. The slide is held in position by friction between the slide, the axle (A) and the block that holds the entire instrument. Thus, by turning the axle (A) the slide is rolling gently in the vertical direction. The approach is done as follows: The shifter (Sh) is moved upwards until the lower end of its gap (ca. 4mm wide) touches the axle (A). Then, the entire slide is moved upwards until the sample is about 200 ~m away from the cantilever. Thereafter, the shifter is pulled downwards, setting the axle (A) free and turning the lever (L). By turning the lever, the axle (A) is rotating and thereby further approaching the slide (S) towards the cantilever. The lever and axle are made of I mm steel wire, the length of L is 10 mm, so the motion of the shifter is demagnified by a factor of 10. The flexible joints are made by spot-welded sheet metal. Figure 4.9b depicts a sketch of the approach mechanism for the z distance between the cantilever and the tunneling tip. The tunneling tip sits at the end of a piezo-tube. The microfabricated cantilever is held on its sup77

port plate (Csp) and pivots around a spring-loaded screw (Sc). The coarseadjust beam (B) demagnifies the motion of the rod which is connected to one of the linear feedthroughs. By pulling the linear feedthrough, the coarse-adjust beam pulls the cantilever support plate against the tunneling tip. The cantilever is plated with chrome and gold; the tunneling tip is made of tungsten. The z distance between the lever and sample, and the z distance between the tunneling tip and lever can be controlled mechanically from the outside while the instrument can be operated in UHV and submersed into the liquid-helium dewar. To reduce the effect of thermal drift, the AFM itself is made mainly of Invar. Because the sensitivity (displacement per volt) and thus the full-range motion of the piezoelectric-tube scanner at 4.2 K is only about 25 % compared to that at 300 K, the precision of the mechanical approach has to be even better at liquid-helium temperature than at room temperature. Owing to the small size of the instrument and the use of Invar the cantilever and the tunneling tip remain aligned as the instrument is cooled to liquid-helium temperature. However, stress relief in the tunneling-tip mount during bakeout can move the tunneling tip up to 50 /Lm sideways, so care has to be taken when mounting the tunneling tip. PZT-5A is chosen as piezoelectric material for this design because it has a fairly high Curie temperature (i.e., the temperature at which it becomes unpolarized and therefore loses sensitivity) and thus can be soldered. The tube scanner and the piezoelectric tube that controls the distance between the tunneling tip and cantilever are soldered to metal plates at one end with a lead-free solder. The other end is soldered to a slab of metalcoated alumina. This AFM/STM has been tested by imaging Highly Oriented Pyrolytic Graphite (HOPG) in the STM mode and KBr in the AFM mode at 4.2 K and 300 K in UHV. Atomic resolution has been achieved on HOPG, and both the potassium and the bromium ions on the KBr (00 1) cleavage plane have been resolved. Apart from the criteria listed at the beginning of this section, special requirements have to be satisfied in the design of an STM for in-situ electrochemical studies. First, only the very end of the tunneling tip and the sample can take part in chemical react ions with solut ions. Secondly, unwanted Faradaic current has to be minimized because the large Faradaic leakage current will cause difficulties to the feedback control. In order to reduce the Faradaic leakage current, the ideal tip should have a chemically and electrochemically inert insulation except for the very end of the tip. In the next section, some simple techniques for coating the tip with insulating materials are introduced.

78

Single tube xyz~~

translater

Sample ~ /'0.---"

Fig. 4. 10. Schematic diagram of the STM proposed by Sonnenfeld and Hansma for the operation in solution [4.10)

Stepper motor Insulated sample wire -----..' Vibration isolation: I

MACOR-fJr base

samPle~

Pytex sample holder

~

Pyrex base Teflon base

E

Teflon

~

''''''';00

u

I

Bias wire

I It Piezo-tube

~MACOR

. RTV _ O-ring Insulallon Brass'·:- .•..:.. MACOR Tip

Fig. 4. 11. Schematic diagram of the STM capable for the automatic approach [4.11). The insulated tip is shown

Two designs of the STM that can be operated in solutions are depicted in Figs.4.10 and 11. Both of them use piezoelectric ceramic tubes as the three-dimensional scanner. The later design adopts an automatic approach mechanism, permitting the approach in a glove box under N2 atmosphere.

79

4.3 Tip Preparation A frustrating aspect of STM operation is the reliable formation of tunneling probe tips. The size, shape and chemical identity of the tunneling tip influence not only the resolution and shape of an STM scan, but also the measured electronic structure. Three features of probe tips seem to be most important for reliable STM operation [4.12]. First, blunt macrostructuring leads to high flexural resonant frequencies, thus minimizing phase hysteresis and allowing higher data rates to be achieved. Secondly, the atomic microstructure of the tip is the key to image resolution because the tunneling current depends exponentially on the gap distance. It is necessary for stable operation to have a single site of closest atomic approach, which is well supported, i.e., not a whisker. Anomalous imaging artifacts will appear when simultaneous tunneling occurs through multiple atoms on the tip. This is commonly referred to as double-tip imaging. Finally, tip purity is important so that a series barrier is not present. For example, the effective resistance of a tungsten-oxide layer can easily be much higher than the desired tunneling gap resistance. Therefore, mechanical contact of tip and sample would occur before the required tunneling current can be obtained. A nonmetallic tip may also make the STM tunneling spectrum not represent the true electronic structure of the sample surface. STM tips are typically fabricated from metal wires of tungsten (W), platinum-iridium (Pt-Ir), or gold (Au) and sharpened by mechanical grinding, cutting with a wire cutter or razor blade, "controlled" crashing, field emission/evaporation, ion milling, fracture, or electrochemical etching. W tips which fulfill the requirement of being stiff, have been used to a great extent to image specimens. However, an oxide layer may exist on the surface of the tip and sometimes makes it difficult to acquire STM images. Platinum, although a soft metal, is a material preferred over W because it is inert to oxidation. The addition of Ir to form a Pt/Ir alloy adds stiffness while maintaining a chemical inert material. Pt-Ir tips are widely employed, too, particularly in atmospheric and electrochemical environments.

(a)

(b)

-

OW

:N.OH

~

-Stainle~- ~

/(~l

_steel __-

OW

WO~- flow

Fig.4.12. (a) Schematic diagram of the electrochemical cell showing the tungsten wire (anode) being etched in NaOH. The cathode consists of a stainless-steel cylinder which surrounds the anode. (b) Sketch of the etching mechanism showing the "flow" of the tungstate anion down the sides of wire in solution (4. 13]

potential. Each procedure gives a different tip shape; the AC etched tips have a conical shape and much larger cone angles than the DC etched tips. The DC etched tips in the shape of a hyperboloid, on the other hand, are much sharper than AC etched tips and are preferable for high-resolution STM imaging. Figure 4.12 shows the details of the electrochemical cell used in DC etching, which contains 100 ml of 2M NaOH or KOH. The W wire to be etched is placed in the center of the cell and serves as the anode. It is mounted on a micrometer so that its position relative to the surface of the electrolyte can be adjusted more precisely. The counter electrode (or cathode) consists of a stainless-steel cylinder which surround the anode. When a DC voltage of 13 V is applied to the anode, bubbles are observed emerging on the cathode/solution interface. The overall electrochemical reaction is

W(s)

+ 6e+ 80H-

anode:

W(s)

+ 20H- + 2H 2 0 ~ WO~ + 3H2 (g)

4.3.1 Preparation of Tungsten Tips

80

-- - I - I I

-

cathode: 6H 2 0

The preferred method for preparing tungsten tips is the electrochemical etching method which was developed for preparing samples for FIM and FES (Field Electron Spectroscopy). The electrochemical etching procedure usually involves the anodic dissolution of the metal electrode; tungsten in this case. There are two ways in which this can be done: An AlternatingCurrent (AC) etch or a Direct-Current (DC) etch according to the applied

w

w

+

~

+ 60H+ 4H2 0 + 6e-

3H2 (g)

~ WO~

SRP = - 2.45 V SOP

= + 1.05 V

EO = - 1.43 V .

The above reaction involves the oxidation dissoluation of W to soluble tungstate (WO~-) anions at the anode, and the reduction of water to form bubbles of hydrogen gas and OH- ions at the cathode. EO is the standard elec81

trode potential given by the sum of the Standard Reduction Potential (SRP) for water and the Standard Oxidation Potential (SOP) for tungsten. Actually, the reaction mechanism is much complexer than indicated by the above equations, and the potential required to drive an electrochemical reaction is usually higher than that calculated from standard electrode potentials. The excess potential is called the electrochemical polarization, which is affected by changes in the concentration of the reactants and products, and other mass transfer processes; the rate of the reaction or the current density is an exponential function of the polarization. Several factors affecting the etching process have been studied by [be et al. [4.13]. Due to the surface tension of the aqueous solution, a meniscus is formed around the wire once it is placed into the electrolyte, as sketched in Fig.4.l2b. It is primarily the shape of the meniscus which determines the aspect ratio and overall shape of the tip. The shorter the meniscus is, the smaller the aspect ratio becomes. A low aspect ratio is important in reducing vibration in the tip during scanning. With the reaction going on, the change in the surface area of the wire and in the fluid disturbances may result in the variation of the meniscus height. To avoid oddly shaped tips, the meniscus height should be kept at the same position by adjusting the micrometer during the etching. The current density of the reaction is limited by the surface area of the working electrode (tungsten) and is also dependent on the concentration and activity of OH- ions. The portion of the tip below the meniscus would normally be etched away without the denser tungstate layer flowing down along the sides of the wire to protect the lower end. Therefore, due to the protection of the tungstate layer over the W wire below the meniscus, the DC drop-off method will, in fact, produce two tips simultaneously: the part dropping off to the cell bottom after being etched, and the part above the meniscus. Usually, a quick automatic cutoff circuit is used to cut off the potential to avoid over-etching because any part of the wire (usually the very end of the tip) remaining in the solution will continue top-etch as long as there is an applied voltage. Figure 4.13 illustrates a simple automatic cutoff circuit, where "Load" denotes the electrochemical cell (Fig.4.l2). A variable resistor R b has a significant effect on the potential across the load. The voltage Vb is compared to a reference voltage for the accurate setting of a low cutoff voltage. The cutoff time of the etch circuit has a significant effect on the radius of curvature of the tip: the shorter the cutoff time, the smaller both the radius of curvature and the cone angle of the etched tip; i.e., the faster the cutoff time, the sharper the tip. The etching circuit shown in Fig.4.13 has a minimum cutoff time of 500 ns. If the part of the W wire above the drop-off point is desired, the tip should be raised up quickly and rinsed in distilled water to remove the residual etchant solution after etching. If, on the other hand, the drop-off part is desired, protecting 82

~ + 12 V -=-i--il G

I

Optional capacitor

R ,Vb 50~ + 5V

SW I

~

1.

V'ef

I

Comparator

I

Etch status LEDs Green=ON Red=OFF

~+12V

1

Rref Etch (lOkQ) enable/disable switch

13kQ

Fig. 4.13. Block diagram of the electronic control circuit to minimize the electrochemical reaction cutoff time following dropoff. R Joad denotes the electrochemical cell illustrated in Fig. 4.12 (4.13]

solutions (for example, trichloromethane) should be placed below the electrolyte to avoid further etching by the electrolyte. It was found that the length of the wire in the solution has a direct effect on the tip, with a longer wire causing the stub to drop off much sooner due to the increased weight. When the weight of the stub exceeds the tensile strength of the neck or the drop-off portion, the stub breaks off, leaving a rough surface or a slight recoil at the end of the tip as the result. These tips have a larger radii of curvature than those of a shorter length in the solution. If the tip above the meniscus is desired, the length of the wire in solution in the range of 1--:-3 mm is most appropriate for a 0.25 mm diameter W wire. The electrolyte concentration is another factor affecting the etching process. Because OH- is consumed in the reaction, it is necessary to replace NaOH solution periodically. The drop-off time increases with the decrease of the OH- concentration. Figure 4.14 exhibits the detailed electrochemical cell used in AC etching. In this method, the AC voltage is much lower than the DC voltage, and a shorter drop-off time can be expected. Either the upper portion or the lower portion which drops off, can be adopted as the tip. If the latter is desired, a shielding solution is also needed to protect the tip.

83

Fig. 4.14. Schematic diagram of the electrochemical cell used in AC etching

Tip holder

T shaped stainless steel fixture Tip

2V

AC

Gold wire loop 2 mm in diameter Fig. 4.16. Schematic diagram of a micropolishing setup to etch Pt-Ir tips by volume: 60% saturated CaCI 2 , 36% H2 0, and 4% HCI. Etching conditions: 2 V rms AC against aAu electrode [4. 14) Micropositioner

4.3.2 Preparation of Pt-Ir Tips Mechanical shearing is a common approach for fabricating Pt-Ir tips. In spite of the variation in shape, many experiments have proved that atomic resolution can be reached by using the mechanically fabricated Pt-Ir tips. Additionally, Pt-Ir tips can electrochemically be etched in several solutions: CaCI 2 /H 2 O/HCI, NaCN/NaOH, KCIIH 2 0/HCI, NaCN/KOH, and molten NaNO) /NaCI. Although resolution requirements are usually not as stringent for highly topographic samples, wide-area scans place unique restrictions on tip morphology. For such samples, symmetric, controlled-geometry tips with small radii of curvature and high aspect ratios are necessary to minimize the convolution of the tip shape into the acquired image. For example, Fig.4.15 illustrates the effect of the tip geometry on the imaged profile of grooves 0.75 f-tm wide and 1 f-tm deep. As indicated in the figure above, a tip of 50 nm in radius and a 15 0cone half angle is shown schematically to be too broad to resolve the square bottomed sample feature; while a more accurate profile of the samesize groove is given by using a tip of 50 nm in radius and 50 cone half

(a)

VIS'

---=-=-=,\ \

"n----\

I

\ I I

\

\ \

I

(b)

.----.-----

\

----.\

t~·_ , ,

I

II \

I \

I

\_'

1

1__

J

I I I

Fig. 4. 15a, b. Effect of tip geometry on the measured STM profile of grooves 0.75 f-tm wide and 1 f-tm deep using: (a) a tip with a 50 nm radius of curvature and 15 0 cone hal f angle, (b) a tip with a50 nm radius of curvature and 50 cone half angle 84

angle. From this example it is clear that a proper tip geometry is crucial to obtain the STM image that represents the sample surface. In order to fabricate specially shaped tips suitable for observing narrow and deep grooves on surfaces, Musselman and Russel [4.14] have developed a technique to fabricate Pt-Ir tips having small radii of curvate and high aspect ratios in a two-step process. In the first etching step, a 1.25 cm long piece of 0.2 mm 80:20 Pt/lr wire is etched in bulk etchant solution consisting of saturated CaCI 2 /H 2 0/HCl (60%/36%/4% by volume) against a C rod at 25 V rms AC for about 5 minutes, producing a tip comprised of a rigid structure with a long slender region just prior to the tip end. The second step involves precision micropolishing of the tip in a thin film of etchant held in a Au wire loop (Fig.4.16). With the help of a stereo-microscope and a mechanical micropositioner, the long slender region at the tip end can be thinned by moving it through the film, or sharpened by making brief contact with the film. The tip fabricated in this way can have a small radius of curvature «50nm), a high aspect ratio (8 0 cone half angle) and a smooth surface (Fig.4.17). Figure 4.19 depicts STM images of an Au-coated PMMA lithographic test pattern obtained by using a chemically etched Pt-Ir tip in the two-step process described above and a mechanically cut Pt-Ir tip, an image of which is presented in Fig. 4 .18. The grooves are 0.75 f-tm wide and 1 f-tm deep. The top view and line scan (Fig.4.19a, b) acquired by an electrochemically etched Pt-Ir tip illustrate that the widths of the groove (top and bottom) are approximately equal. In contrast, the top view and line scan (Fig.4.19c, d) acquired using a mechanically cut Pt-Ir tip demonstrate that although the cut tip can reach the groove bottom, providing an accurate measurement of groove depth, the irregular shape and broadness of the cut tip is clearly incorporated into the STM image of the test pattern, and the width at the bottom of the groove is only one third of the actual width of 0.75 f-tm. This striking comparison of STM images acquired from a highly topographic sample using Pt-Ir tips with dramatically different geometries clearly illus85

Fig. 4.17. SEM image of a Pt-Ir tip with a radius of curvature :/; ;~':J~~

:>.....:"-'-',-O-;:."'+-+I------1I>-"1--tr

§

~

High-speed special- purpose digital signal processor

I

Hard drive

I

I

RAM

I

I

80386 microprocessor

z

~L.B

t.~,.

Bias

'.r,·'·.'·,.·.",·.. ,·,.',·".,"'.•'.'.'.·,····,"·,",".· ,.,.,..

!:~~ I

Base support

" .~ j .

Fig. 4.27. Block diagram of the STM control, data acquisition and display system indicating all of the equipment connections 96

97

For more efficient and precise instrumental control, the computer may need more analog inputs to measure some other parameters such as the tunneling bias Vb' the voltages V x ' V y applied to the x-, y-piezoelectric elements, and the offset voltages for selecting the scan window. In this way, data acquired are recorded automatically, and changes of each parameter can be monitored during scanning. It is important to note that connections to the ADCs have to be carefully made in order to avoid electric noise. As discussed above, the x- and y-raster voltages can be supplied by a function generator, but in many cases DACs are used instead. Therefore, at least two D/ A channels are required to generate x and y signals. Furthermore, in order to acquire the STS and work function information, other DACs for controlling the bias voltage Vb between the tip and the sample and a digital input are needed. The latter is used to trigger and turn off the feedback loop many times during the data acquisition process. Of course, offset voltages can also be controlled by DACs. In addition, if automaticapproach devices such as a louse or an Inchworm Motor are employed, it is possible for a computer to control the approach. Some kind of device is always needed to visualize the values of the zpiezoelectric voltage V z, while data are being collected. An analog x-y pIotter was traditionally employed by STM to graphically represent the modulation of Vz' These line scans separated by a fixed distance provide an illusion of three-dimensional surface topography. However, the slow motion of its pen sets a very restrictive limit on the scanning frequency and this kind of presentation provides limited image quality. To fully analyze the scanned-image data and resolve spatial information, gray-scale or top-view images should be generated. For these reasons, digital graphic display devices, such as color graphic controller cards and high-resolution graphic monitors are now widely used. Ideally, graphic display devices are expected to have the resolution of 1024 pixels and 256 levels of gray to enhance image processing.

(3) file management, and (4) some processing tools for postacquisition image processing. The following items are usually included in a parameter setting: an experimental identification name and descriptive text, the number of points per x scan (this number should be a power of two to facilitate the computation of the discrete Fourier transform), the maximum number of x scans (the actual number of scans may be smaller if the experiment ends prematurely), the size, color and position of the image to be displayed, the ranges of the DAC and ADC (if any), and the conversion factors of the digital counts to the real values of the voltages applied to the x-, y- and z-piezoelectric elements so that coordinate data can be given in Angstroms or nanometers. Default values of these parameters can be obtained from a program profile, which allows the investigators to avoid going through several menus. In addition, in order to save storage space and time, data have to be stored in binary format. However, the program should incorporate the possibility of converting STM image files from the binary format to character format. The following techniques can be utilized in the software for immediate data analysis: • Discrete Fourier Transform (DFT) of each scan is performed in order to find its frequency spectrum. This information is relevant in that it permits the microscopist to see whether there is any periodicity in the data, whether there exists any periodic noise that the system is getting, or whether the signal-to-noise ratio is too small. The DFT is computed with the Fast Fourier Transform (FFT) algorithm for a number of data points that is a power of 2. • Root-mean-square noise of the samples per scan is defined as I/NJE[Vz(X)-(V z)]2 ,

4.5.2 Software The computer software that performs instrumental control, data acquisition, and image display is usually written in a high-level language except for some subroutines written in an assembly language for critical tasks in order to get a fast and compact executable module. All the programs should be executed in a user-interactive way through menus. Generally, the program can be organized in four well-defined parts: (1) Setting of parameters that determine the operation of the microscope, (2) STM control, data acquisition and real-time display, 98

where Vz(x) represents the N samples of Vz collected in a scan, (V z) is the mean value of Vz' This value gives information about the reliability of the data. • Real-time display paramters are used to increase the dynamic range of the STM image displayed on the monitor. The image actually displayed in real time is [4.24] V;(x,y)

=

C[Vz(x,y) - (A+Bx)],

(4.18)

where C is a scaling factor, and the parameters A and B are obtained by a 99

least-squares fit of the values Vex, Yo) collected in a test scan. What should be taken into account is that the values v; (x, Yo) should range between a minimum (greater than 0) and a maximum (less than 255) selected by the microscopist to have a reasonably good probability that v; (x, y) does not fall out of the range of possible pixel values. • Real-time digital filtering can be performed within the program on each STM line scan. There are sources of acoustic and vibrational noise that can be filtered very efficiently by using a Finite Impulse Response (FIR) filtering relation [4.25] of the form: x(l) = ax(l) + bx(l-l)

+ cx(I-2) + dx(I-3) + ex(I-4)

,

(4.19)

where a, b, c, d and e represent the coefficients of a 5-tap FIR filter. The real-time filter windows over each data point I in the line scan range from 4 < I < N pixels per line. Software signal averaging is also utilized to improve the Signal-to-Noise Ratio (SNR) of the acquired images. • Since generally the surface normal is not parallel to the tip, the structures we are interested in are in an inclined plane. It is difficult to see details if we look at such pictures in a top-view presentation. In order to correct the microscope image that results due to this sample tilt, a plane must be fit and substracted from the image. An interactive subroutine uses a window (usually of ten previously acquired line scans) to compute a plane that can be written as Zplane

-

ax + {3y + 'Y

are difficult to visualize in the top-view image. A magnified image on the selected area produced by software zoom techniques with interpolation can help reveal detailed features on the surface. Additionally, some other tools which can be included in the software are listed below. a) Histogram Equalization

It is desirable for all surface features to be displayed within all of the available (256, for instance) gray levels. The plane subtraction algorithm described earlier can be used to approximate this condition during real-time image acquisition. During post-acquisition image analysis, the microscopist can expand the dynamic range of the image by performed histogram equalization. This requires the use for a new minimum and maximum pixel values in the range (0--:-- 255) to adjust the contrast of the image. The image pixels can then be adjusted in the following manner: All pixels < minimum are set All pixels

= 0 ;

> maximum are set = 255 ;

Remaining pixels =

(Old pixel value - minimum) x (maximum) (. ..) (4.21) maximum - mlOlmum

where maximum and minimum are the gray-scale limits specified by the user.

(4.20) b) Convolution Filter

where a, (3 and 'Yare computed coefficients. By using the computed plane, slope and offset values are computed to correct the current line scan from the sample tilt.

4.5.3 Image Processing Image processing can include algorithms for systematic data correction, such as a change of scale and a correction of the thermal drifts of the scan origin and of Vz' Two-dimentional fast Fourier transform is very useful for periodic images, and can be applied to a periodicity analysis of the image. Images can be displayed using the same color coding. However, a better matching of the dynamic range of the data to the set of colors can be achieved, in order to optimize the visual interpretation of the images, either based on the image histogram or done directly by the microscopist. Using the cross-sectional cut tool to display the height values of a topography in any direction is valuable for analyzing the detailed surface features which 100

STM images typically contain noise. Space-domain filters are available for smoothing (variable-size window averaging), sharpening (Laplacian filter), and edge detection (gradient filter using Sobel kernels). Although convolution filters do not provide an ideal frequency filter, they are often used as low-pass or high-pass filters. The following algorithm describes the 3 x 3 convolution of image pixels [4.26]:

Updated pixel valuex,y

x+l

y+l

L

L

W(i+x+2,j+y+2) Z(i,j)

i=x-l j=y-l 3 3

L L

(4.22) IWCi,j)1

i=1 j=1 where ZCi,j) are pixel values, and the W(i,j) are the window coefficients. 101

The values in the 3 x 3 matrix of window coefficients are integers, and can be set by the users to implement other filter types. The convolution algorithm requires nine multiplications, nine additions, and one division for each pixel in the image. This process is repeated for each pixel in the N x N image. The edges of the image are handled as special cases since there are no data beyond the edges with which to perform the convolution.

Fig. 4.28a,b. STM image of a Si (111)-7 X7 surface coated with Ag at low coverage. (a) Grey scale rendering of unprocessed data, (b) having been obtained from (a) by statistical differencing [4. 27J

c) Statistical Differencing Statistical differencing is a standard technique which attempts to make the corrugation uniform throughout the image by determining a local corrugation for the input data and empirically choosing a transformation. This method is suitable for boundary enhancement in cases where STM images simultaneously show two different structures. The transformation employed is

o

85 =

(D - A) a

+

A

17

+ "2 + 64

,

(4.23)

where 0 and Dare 8-bit output and input data for a given pixel, respectively. A is a local average, defined over an 8 x 8 array of surrounding pixels, and a is the variance of the data for the given pixel from the local average. The first term is designed to result in a uniform corrugation, the denominator contains an additional term to limit the enhancement in the regions where the corrugation is small. The second term preserves, but reduces, the effect of steps or average height changes. It can be eliminated to equalize the corrugation if desired. The last term simply sets the grey scale value to mid-range for a constant mid-range input image. It is important to note that the choices of the neighborhood size used in computing the local average is important, especially since the lateral extent of the features in the STM image are different for the two domains. The trial-and-error method is usually applied for choosing parameters. Figure 4.28 shows the Si (111) surface processed by means of this method by Wilson and Chiang [4.27]. d) Three-Dimensional Representation STM images can also be displayed as three-dimensional surfaces. Threedimensional perspective images offer additional clarification. especially when complex structures are present. The use of the contours, together with shading is also helpful for rendering images when image display formats have good resolution but few gray levels. The orientation of the surface can be chosen by the users, as well as the point from which the surface is illuminated and the relative proportions of direct and the isotropic diffuse light applied. Figure 4.29 exhibits a three-dimensional representation of the Si (111)-7 x7 surface. 102

Fig. 4.29. Three-dimensional representation of the Si (111)-7 X7 surface

103

50 Other Related Scanning Probe Microscopes

The successful achievement of G. Binnig, H. Rohrer and their colleagues at the IBM Zurich Laboratory initiated a surge of research and engineering activity. This brought about rapid advances in STM technology and led to the development of many other novel scanning probe microscopes, such as the Atomic Force Microscope (AFM), Lateral Force Microscope (LFM), Magnetic Force Microscope (MFM), Ballistic-Electron-Emission Microscope (BEEM), Scanning Ion-Conductance Microscope (SICM), Near-field Scanning Optical Microscope (NSOM), scanning thermal microscope, and Scanning Tunneling Potentiometry (STP). These microscopes take advantage of the remarkable ability to control the spatial position of the tip relative to the sample, and provide new information about the physical properties of surfaces on an atomic or nanometer scale. This chapter is devoted to brief descriptions of the operational principles and some applications of these new types of microscopes.

5.1 Atomic Force Microscope With regard to the operational principle of the STM described in ChapA it is important to note that the "tunneling" phenomenon utilized by STM requires that this instrument be used only for conductors or semiconductors. For a non-conducting material, its surface must be covered with a thin conducting layer. However, the existence of conducting film often lowers the resolution and thus limits the usefulness of the STM. It was chiefly this limitation that prompted the development of Atomic Force Microscopy (AFM) in 1986 [5.1-3], which can be applied to image conductors and non-conductors in air, liquids or vacuum. The operational principle of AFM is explained in Fig.5.!. The cantilever which is extremely sensitive to weak forces is fixed at one end; the other end has a sharp tip which gently contacts the surface of a sample. When the sample is being scanned in x-y direction, because of the ultrasmall repulsive force existing between the tip atoms and the surface atoms of the sample, the cantilever will move up and down in the direction vertical 105

(a)

5.1.1 The Force Sensor

B 1

E

u

1.

I

I

Block (Aluminium)

(b)

A: AFM sample B: AFM diamond tip C: STM tip (Au) D: Cantilever, STM sample E: Modulating piezo F: Viton

251lm1 Diamond tip

B

y

z

·.?S

~

Lever (Au-Foil)

k

TE

Mil}

c:;

6

1

Fig. 5.1. The principle of an AFM [5.1]

to the surface of the sample, corresponding to the contours of the interaction force between the tip and surface atoms of the sample. The topographic images can be obtained either by recording the deflection of the cantilever at each point (variable deflection mode) or by keeping the force constant. using an integral feedback loop and recording the z-movement of the sample (constant-force mode). While the STM yields images related to surface electronic energies near the Fenni level, AFM images are related to surface electronic energies up to the Fenni level. The AFM tip may be kept in direct contact with the surface (contact mode) or it may be vibrated above the surface (non-contact mode). For high-resolution imaging and most routine topographic profiling, the repulsive-force (10- 9 --:- 10- 8 Newton) or contact mode is usually used. In the non-contact mode, the van der Waals force, the magnetic force, or the electrostatic force is detected. The non-contact mode has been employed in other scanning force microscopes which will be reviewed in the next section. An application of STM and AFM of particular interest is the in-situ study in electrochemistry. As discussed in Sects.4.2.4 and 4.3.3, special requirements have to be satisfied in the design of an STM for in-situ electrochemical investigations because the electrolyte is highly conducting, which makes the detection of tunneling current difficult. With the advances in AFM, especially the non-conducting AFM tip (Sect.5.1.1) and the optical-beam-deflection technique (Sect. 5.1.2), the AFM has been used successfully to study the electrode surfaces under potential control in a fluid electrolyte. One example will be given in Sect.5.1.3.

106

An AFM can be designed to look very similar to the STM. The primary difference is that it is convenient to have the sample move rather than the delicate, and sometimes bulky, force sensor (composed of tip and cantilever). Much of the technology for implementing the AFM is now well developed. Techniques for vibration isolation, scanning, sample approach, feedback control and image processing are taken with little modification from the STM. Construction of the force sensing cantilever stylus and the measurements of the deflection of the cantilever still need careful consideration. The force sensor is a crucial component for the AFM, detennining its sensitivity and resolution. While the force sensor senses the forces across a sample, the role of the cantilever is to communicate this infonnation to the outside world. When the AFM is operated in the contact mode, in order to register a measurable deflection with small forces, the cantilever must flex with a relatively low force constant. The data acquisition rate in the AFM is limited by the mechanical resonant frequency of the cantilever. To achieve an imaging bandwidth comparable to that obtainable in the STM, AFM cantilevers should have resonant frequencies > 10 kHz. Fast imaging rates are not just a matter of convenience, because the effects of thermal drift are more pronounced with slow scanning speeds. If the scanning rate is too high or the cantilever resonance is too low, the inertia of the cantilever will cause the stylus tip to exert large forces on steeply sloped protrusions, and prevent the stylus tip from tracking steep downward slopes. The combined requirements of a low spring constant and a high resonant frequency can be met by reducing the mass of the cantilever stylus assembly. High lateral stiffness in the cantilever is desirable to reduce the effects of lateral forces in the AFM. Frictional forces can cause appreciable lateral bending of the cantilever, leading to asssociated image artifacts. Investigations have indicated that choosing a "V" or "X" shape of the lever can yield substantial lateral stiffness. When optical beam deflection is used to measure cantilever deflection, the sensitivity of the detector is inversely proportional to the length of the cantilever, since greater angular bending occurs for a given linear displacement of the end of the cantilever when the length of the cantilever is reduced. Reducing the length of the cantilever is an additional motivation for microfabricating the cantilever stylus. Provision must be made for the operation of the detector used to measure the cantilever deflection. If tunneling is utilized, a metal electrode should be fabricated on the back of the cantilever. If optical methods are applied, a reflective surface is needed. For optical beam deflection, the reflective surface should also be as flat as possible. The microscopic structure of the point of the tip is important for high-reso-

107

lution contact profiling, where the contact area between the tip and sample may involve only a few atoms. For high-resolution topographic imaging, the cantilever stylus used in the AFM should satisfy the following criteria: (1) A low force constant, (2) a high resonant frequency, (3) a high mechanical Q, (4) a high lateral stiffness, (5) short lever length, (6) incorporation of a mirror or electrode for deflection sensing, and (7) a sharp protruding tip. The force sensor used in the first AFM was a gold foil of 0.8 x 0 .25 x 0.025 mm 3 carrying a diamond stylus. The gold foil served as a lever and the diamond fragment attached to one end of the gold foil as the AFM tip. The observations of a ceramic (Al z 03) surface in air demonstrated a lateral resolution of 3 om and a vertical resolution less than 0.1 nm. Thereafter, force sensors have been constructed from thin metal foils (Au, W) or formed from fine metal wires (Au, W). Carbon and fused quartz fibers have also been used as the force sensor. Atomically resolved force maps of graphite were obtained utilizing AFMs equipped with these sensors. It has been shown, however, that microfabricated cantilevers are ideal since they have both low force constants and high mechanical resonance frequencies. The microfabrication technique developed by Albrecht et al. [5.4] (indicated in Fig. 5 .2) has been employed to produce cantilevers from thermally grown SiO z and Si 3 N4 by means of Low Pressure Chemical Vapor Deposition (LPCVD). The fabrication process begins with the deposition of the desired film on both surfaces of a (100)Si wafer. The thickness of the Free cantilever

------>...--

( c) (a)

Top '-----------' VIew

/

se~~i~n~""::':B~:?~4or (lll)plane in Si Fig. 5. 2a-d. Fabrication of thin-film microcantilevers. (a) A thin film of SiOZ or Si 3 N 4 is formed on the surface of a (100) Si wafer and patterned to define the shape of the cantilever and to create open ings on the top and bottom of the wafer. (b) The windows are al igned along (III) planes. (c) Anisotropic etching of the exposed Si with KOH undercuts the cantilever and self-terminates at the (III) planes as shown. (d) A small Si chip is cut from the wafer to serve as a pedestal for mounting the cantilever in the AFM [5.4] 108

Fig.5.3. SEM micrographs of SiOz microcantilevers. (a) Rectangular cantilevers. The shorter of the two has dimensions of 100 X20 XI. 514m3, with a force constant of I N/m and a resonant frequency of 120 kHz. The V-shaped cantilever shown in (b) has increased lateral rigidity which reduces its sensitivity to frictional forces [5.4)

film determines the thickness of the finished cantilever. The film is patterned photolithographically to form openings on the top and bottom of the wafer, with the cantilever protruding into the top opening (Fig.5.2a). The geometry and registry of the top and bottom patterns are chosen so that the edges of the openings lie approximately along common (111) planes in the Si lattice. The EDP (ethylenediamine/pyrocatechol/water) Si etchant selectively etches Si rapidly in the (100) direction, but is inhibited when the region etched away is bound by (Ill) planes. EDP does not attack SiO z appreciably. The bottom side of the cantilever and the surrounding Si is then coated with a thin film of metal for the deflection detector in the AFM. The finished structure is shown in Fig.5.2d, and SEM micrographs of finished SiO z cantilevers are depicted in Fig. 5.3. For imaging atomically flat samples one can simply use the end of a cantilever to act as an effective local tip. For imaging rougher samples, however, one desires a sharp, protruding tip of known shape so that the interaction between the sample and the cantilever can be characterized more precisely. A fabrication-process produced Si 3 N4 cantilevers with an integrated pyramidal tip is presef.ued in Fig.5.4. The first step is the formation of a pyramidal pit on the surface of a (100)Si wafer (Fig.5.4a). The wafer is coated with a suitable masking material, such as a thin film of thermal SiO z , and a small circular or square opening is etched in this film. The exposed Si is etched with an anisotropic etchant. After the pit has been formed, the masking material is removed from the wafer. The Si 3 N4 film from which the cantilevers will be made is then deposited on the wafer surface and conforms to the shape of the pyramidal pit (Fig.5.4b). Annealing in an oxidation furnace follows to prepare the surface of the nitride for the later step of anodic bonding. The Si 3 N4 film is patterned into the shape of a cantilever and aligned so that the end of each lever lies over a pit. All Si 3 N4 109

,,)

181

J,m:::::s
" ,

.,~ll_ .~.,." .'

. . . .~t...

..

.

.~-: ....

",' l..,

.,

'..

..'

'.

~

.. '

.. ~

.

.

"4

-~""-: '\,'""'. ,~~.:.

..

. - ' .~ < ••

90

-'

,.,..

•

~_.' .~. .,,'

-'

.,. ~: • . .,.~

.,:. ~...~'

......

.;;.•._.~..~

.,;,

... t.,..

The achievement of atomic resolution of AFM in the non-contact, especially on semiconductor surfaces under UHV condition, marked a significant progress of AFM probing capability, leading to minimized interactions and making it practical to direct comparison of STM and AFM results (Fig.5.20). Having the advantage of little electric field induced effects, it can be better used to study the point defects and their motions [5.35].

5.2.3 Force Microscope Operating in the Tapping Mode Applications of the non-contact mode operations are limited due to some fundamental drawbacks inherent to the technique. The tip must be vibrated close the sample surface with low energy because van der Waals forces are relatively weak. However, moving the vibrating tip closer to the surface increases the chances of getting the tip stuck in the water layer which covers the surface of all samples exposed to the atmosphere. Trapping the tip-tosample separation, which is typically between 5 and 10 nanometers, defines 126

Fig. 5.20. The nonCOnlact AFM image of Si (111)-7 X7. In [he upper section of the image, the tip was monatomic for about the width of a cell. The unit cell C indicates all 12 adatoms . The unit cells C and E show two adatoms that are not visible in A and B Unit cell A shows an atomic defect. The right of the two central adatoms is misplaced [5.34]

the lateral resolution of a force microscope operated in the non-contact mode. Due primarily to these limitations, the non-contact force microscope has found only limited applications. To overcome the limitations of the non-contact technique. Digital Instruments has developed a new technique for operating the force microscope, the so-called tapping mode. This technique vibrates the cantilever with a larger amplitude than the non-contact mode operation. In the tapping mode, high-aspect-ratio tips with small radii of curvature are used and the vibrating tip contacts the sample surface many times per data point. The cantilever oscillation is damped when the tip contacts the water layer and the sample surface, but the larger vibration amplitude gives the cantilever sufficient energy to overcome the surface tension of the adsorbed water layer. The force imparted onto the sample by the cantilever can be very small because small shifts in the vibration amplitude can be detected. The tapping mode allows delicate samples to be imaged with normal forces on the order of fractions of a nanoNewton and shear forces that are essentially zero. The applied force is significantly lower than the force applied by the contact AFM. The lateral resolution in the tapping mode compared favorably to the contact AFM because high vibration frequencies allow the tip to Contact the sample surface many times before it translates laterally by one tip diameter. Therefore, the tip shape defines the lateral resolution in the tapping mode just as it does with the contact AFM. In practice, this mode 127

can be used to get a topographic image of any sample regardless of its conductivity or mechanical composition. Tapping mode AFM has been widely used in studying surface properties of biological material and polymers, with great successes.

5.2.4 Magnetic Force Microscope The Magnetic Force Microscope (MFM) is essentially a kind of force microscope operating in the non-contact mode except that the tungsten or silicon tip is replaced by a nickel or iron tip which is magnetized along its length. Tips coated with ferromagnetic thin films were also investigated successfully by several groups and have became commercially available recently. Magnetic thin film tips have the substantial advantage of a significantly reduced tip stray field as compared to bulk-wire tips. Another advantage is that their magnetic properties can be controlled by choosing an appropriate coating material. Thus, it is possible to measure selected components of the sample field by coating tips with high coercivity films and suitably magnetizing them in an external field. The magnetic forces measured in the MFM are purely magnetostatic; they arise from the magnetic dipoles in the tip interacting with dipoles in the sample. When the cantilever which is oscillated at its resonant frequency is brought near, a magnetic sample and the tip encounter a magnetic-force gradient, the effective spring constant and, hence, the resonance frequency is shifted. By driving the cantilever above or below the resonant frequency, the oscillation amplitude varies as the resonance shifts. An image of magnetic field gradients is obtained by recording the oscillation amplitude as the tip is scanned over the sample. Much of the reason for this excitement is the fact that MFM is the only magnetic imaging technique that can provide high resolution (l0 --:- 100 nm) with essentially no special sample preparation. This microscope enables us to look at the structure of materials such as magnetic heads that determine the definition, uniformity and strength of the magnetic disks and other media, giving insight into both head performance and quality of the storage medium. Good-quality images can be taken even when the magnetic material is covered with a thin overcoat, an important feature when imaging many technological samples. Figure 5.21 exhibits a pair of images of a magneto-optical disk taken in two passes over each scan line. On the first pass, topographical information is recorded in the tapping mode where the oscillating cantilever lightly taps the surface, as described in the last section. On the second pass, the tip is lifted to a selected separation (typically 20 --:- 200 nm) between the tip and local surface topography. By using the stored topographical data instead of 128

Fig. 5.21. A pair of images of a magneto-<Jptical disk [5.36)

the standard feedback, the separation remains constant without sensing the surface. At this height, cantilever oscillation amplitudes are sensitive to relatively weak but long-range magnetic forces without being influenced by short-range surface interactions. Two-pass measurements are taken for every scan line, producing separate topographic and magnetic field images. The topographical image on the left of Fig. 5 .21 shows grooves that delineate the recording tracks. The right image shows magnetic force gradients as sensed in the lift mode. The gradient image gives a clear picture of the bits, in which magnetization is oriented perpendicular to the sample plane in a direction opposite to the background magnetization. In this experimental disk, the power of the laser used to write these bits was varied, creating magnetic bits of different sizes. The smallest bit shown in this figure is about 90 nm across. A resolution of 10 nm has been obtained on rapidly quenched FeNdB thin films by MFM [5.37]. By using MFM we have successfully imaged domains in very "soft" magnetic materials such as the organic LB films mentioned in Sect.5.2.I. Figure 5.22 depicts a pair of images of LB film 2-(4-hexadecoxyphenyl)4,4,5, 5,-tetramethyl-4, 5-dihydro-l H-imidazolyl-l-oxy-3-oxide. The topographical image on the left obtained in the tapping mode shows a rough surface of the films. The right image reveals that the LB films form quasi-onedimensional band structures composed of magnetic domains on a large scale up to 30 x 30 J.Lm 2 . The results demonstrate that the MFM is also a promising method for detecting weak magnetic signals originating from as thin as 10 molecular layers. In the short period since the first MFM image obtained by Y. Martin and H.K. Wickramasinghe in 1987, the field has grown rapidly. MFM is al129

'. ·'. ·.. .'. .".-. . ,'. ..'.:';','\ \,t1\.t. ..~.'

:.. • : . I

..

. ','

.

.. - '

....• :: . ~

•

.

•

•

.'.'

'. .... -~ t • , ,~... .., . ....,: 1

,

•••.•.

•

'.

i

.,

.I

~'\'.'it f4

-"1

,.\ "

"

~'.

,

~l -

1'.' .

. . . .1 . , . ~

.'

i

'1

. '. '.~'.'. . .f '. . ~ t>i 1 y\ \ " .•'.:' .. ~. .' .. ~\' t ,~~, r, # . . ' ;'.;' t·. • • • • . • . ~.li ,.1q., \~ . .': '·. '" '".'.' .;:',,i 'Il i * ,-.· t-",' ... !.!Uu.V,~ \"\\''',)1 •

'..

.'

.

.

'

I

.'

•.. : :-:""(:;; o

30.0 IJrn 0

't

f l

\'

30.0 IJrn

Fig. 5 _22. AFM and MFM images of a LB film obtained in the same scan area

ready a powerful tool for magnetic imaging, with many scientific and industrial applications. Future advances in MFM will include improved tips, matching of tip-sample characteristics, and determination of field strengths and domain structure from force gradient images. Much effort has been given to improve the resolution of MFM [5.38, 39]. The ultimate goal is set to use the MFM technique to recognize species at the level of individual atoms [5.40].

5.2.5 Electrostatic Force Microscope Similarly to the MFM, the Electrostatic Force Microscope (EFM) also uses a probe vibrating at its resonance frequency. However, in the EFM, the probe has an electric charge. Its vibration amplitude is affected by the electrostatic forces resulting from electrical charges in the sample. The force sensitivity of the EFM suggests that capacitive variations down to 10- 21 F could be measured in a 1 Hz bandwidth. The EFM has been used to map the electrical properties of the sample's surface. For example, the distribution and concentration of dopant atoms in doped silicon plays a critical role in chip performance. A voltage applied across the gap between the probe of the EFM and the sample will mobilize the conduction electrons or holes beneath the probe leaving a charged region that exerts an electrostatic force on the tip. The consequent movements of the tip provide a precise, finescale measure of the charge and hence of the number of mobilized electrons or holes, and the concentration of dopant atoms. 130

Fig.5.23. Images (2.5X2.5J1.m 2 ) of polished sapphire using repulsive contact forces (a) and attractive electrostatic forces (b) [5.4l]

The electrostatic images can also be obtained in another way. In studies of polished sapphire, charge is deposited on the sample by applying a bias of + 10 V to an electrode on the back surface of the sapphire sample for several hours, and allowing it to come to equilibrium. Images were taken by scanning the sample and using feedback (with the opposite polarity than that used for imaging with repulsive contact forces) to maintain a constant electrostatic force on the tip. Figure 5.23 exhibits two images taken over the same area of the sapphire surface with contact forces (Fig.5.23a) and electrostatic forces (Fig.5.23b), respectively. The electrostatic mode images reveal the same large-scale periodicity as with the repulsive mode, but with an increased amplitude. This result implies that the features are more than simple topography, and possibly correspond to regions of charge accumulation. The EFM has also been applied to measure voltages on circuits, to observe discrete steps in the force versus time curve corresponding to the discharge of single charge carriers during the decay process. For example, by monitoring the electrostatic force on the insulating Si 3 N4 films which had been deposited charge carriers [5.421. The recombination was attributed to the thermionic emission of electrons from the tip or sample, depending on the sign of the deposited charge. Therefore, the force microscope has demonstrated its ability to detect individual electrons or to measure currents of the order of 10- 19 A, which may have wide-ranging applications. Another application of the EFM is the observation of ferroelectric domain walls in the ferroelectric-ferroelastic material Gd 2 (Mo0 4 ). The features of the EFM image obtained by measuring electrostatic forces from the 131

5 .3 Ballistic-Electron -Emission Microscopy

Fig.5.24. The lopographic and the electric field images of a cross section of conducting nanowires [5.36)

polarization charge at the sample surface could be explained by modeling the wall as a step function in the electric potential. With a grounded conducting tip, the force microscope operating in the non-contact mode can measure electric field gradients by oscillating the tip near its resonant frequency, depending on which side of the resonance curve is chosen, the oscillation amplitude of the cantilever increases or decreases due to the shift in resonant frequency. By recording the amplitude of the cantilever, an image revealing the strength of the electric field gradient is obtained. Figure 5.24 shows a pair of AFM images depicting a cross section of conducting nanowires embedded in a non-conducting medium. The top surface of this sample exposes the ends of the nanowires. The other side of the sample was attached to a metal sample puck with conductive epoxy and was held at approximately 7 V relative to the cantilever tip. The image on the left is topography and the one on the right is the electric field gradient above the sample. The large spots roughly 200 nm in diameter in the left image are the metallic nanowires. The electric field gradient from these nanowires is shown as measured in the lift mode as described in the last section. Note that several of the nanowires that appear clearly in the topographic image are missing from the electric field image because electrical contact to these nanowires has failed.

The discovery and application of semiconductor materials have resulted in the necessity for a complete understanding of the fundamental characteristics of semiconductor surfaces and interfaces, such as the influence of thinfilm deposition on semiconductor surfaces, the transport properties of interfaces, superlattice carrier mobilities, quantum-well depths, etc. Subsurface interface electronic properties are not directly accessible to conventional surface analytical techniques. Although conventional Schottky-Barrier (SB) characterization methods, including photoemission, photoresponse, current-voltage, and other techniques, can be used for indirect investigation of the properties of interfaces, they are limited by their lack of spatial resolution for probing the variation of SB properties over the interface plane. Ballistic-Electric-Emission Microscopy (BE EM) is the first method for direct spectroscopic investigation and imaging of subsurface interface electronic properties with high spatial resolution.

5.3.1 The Principle of BEEM The BEEM developed by Kaiser and Bell [5.43] utilizes STM in a threeelectrode configuration (Fig.5.25). The sample for BEEM investigation consits of at least two layers separated by an interface of interest, in generaL is a metal/semiconductor SB heterojunction. The STM tunneling tip is positioned near the surface of the heterojunction to emit ballistic electrons into a metal/semiconductor structure via vacuum tunneling. These low-energy electrons have typical attenuation lengths of about 10 nm in the metal

(a)

EF~

----

Base colleccor

Tip (b)

evC//%1 ~I

//1 -":

132

Ijjjjj})j

~)

Jjj j

eV b

jJ---

Fig. 5. 25a, b. Schematic energy-band diagram for lhe three terminal BEEM experiment. The lunneling tip is separaled by a vacuum barrier from lhe base metal. Terminals are applied 10 the !Unneling tip. metal base. and semiconduclOr colleclOr. Collector current Ie is measured bel ween the base and collector. (a) The energy-band diagram for zero !Unneling bias (Y = 0). (b) The energy-band diagram for tunnel ing bias grealer than the barrier voltage (eY > eV b ) [5.43) 133

base, and some of the injected electrons may propagate through a surface layer to the subsurface metal/semiconductor SB interface before scattering. For a base-tip tunnel bias less than the base-collector barrier height Vb' there will be no ballistic-electron current into the collector since the injected ballistic electrons have insufficient energy to surmount the energy barrier. But, as the base-tip bias V is increased above Vb (Fig.5.25b), a dramatic increase in base-collector current Ie occurs. The spatial variation in Ie reveals the differences in electronic structure at the different areas of the interface. Moreover, the Ie-V spectrum provides a direct probe of interface electronic structure, including the important SB height, defect structure at the interface, quantum-mechanical reflection of electrons at the interface, and ballistic-electron transport properties of the base film. The collector current can be described by

j

00

D(E x)

E min

Ie

RI[

00

j

E max f(E)dE t dE x

0 00

(5.l)

!aD(E,l !a[f(E) - f(E+eVl]dE,dE, where It is the tunneling current, D(E x) is the transmission probability for an electron to tunnel through the vacuum barrier, feE) is the Fermi function, R is a measure of attenuation due to scattering in the base layer. R is taken to be an energy-independent constant, since ballistic-electron attenuation lengths in metals are nearly independent of energy for E- E F of less than 2 eV. E max is given by [m{/(m-mt)]'[Ex-Ef+(V-V b)], and E min = EF-e(V-V b). m( is the electron's effective mass parallel to the interface within the semiconductor, and m is the free-electron mass. Equation (5.l) can be fitted to the experimental spectra by adjustment of Vb and R. Most importantly, as a consequence of the parabolic conduction-band minimum and K( conservation, (5.1) predicts that the Ie-V spectrum behaves as (V - V b)2 for the vOltages just above the threshold value Vb' independent of the fitting parameters. There is a critical angle for electron propagation in the base outside of which electrons may not be collected. The critical angle is a dominant effect in transmission and total internal reflection in a base which depends on the incident energy and interface band structure. In addition, it focuses the electrons which provides high spatial resolution of interface properties. As mentioned above, since only electrons with small transverse momenta in the base may be collected, scattering in the base serves mainly to reduce the number of electrons col134

lected, rather than to reduce resolution. For example, for Au on Si or GaAs with mt/m = 0.067, this critical angle is less than 6°, so only electrons within 6° of normal incidence may be collected. For a tunneling voltage just above threshold (i.e., V-Vb is about 0.3V), a lateral resolution of the order of 1 nm is achieved for a 10 nm thick base layer. The resolution obtained by the BEEM experiment is in agreement with this treatment. This demonstrates that critical-angle reflection defines the spectrum shape and spatial resolution of the BEEM spectrom, and is also a dominant effect in interface carrier transport.

5.3.2 The BEEM Experiment There is not much difference between the apparatuses used for BEEM and STM experiments. In order to keep the surface clean and to isolate the vibration from outside it is necessary for some samples to perform the BEEM experiment in vacuum or in inert gas. The criteria to be satisfied in the design of the BEEM are the same as those for the UHV STM. However, in order to measure the collector current Ie' a high-sensitivity (gain :::,,) 011 VIA), low-impedance (l00) current amplifier should be added to the electronics. The typical collector current is less than 100 pA and is weaker than tunneling current by a factor of at least 10. This leads to the stricter requirements for the stability, reproducibility, and high Signal-to-Noise Ratio (SNR) of the BEEM system. The software for data acquisition should have the capability of acquiring the STM topographic and the corresponding BEEM images simultaneously. The Ie signals should be averaged many times at each surface location to improve the spectral SNR. The specific differential resistance Re = (dV/dJ)v=o for the Au/GaAs (l00) SB interface structure has been measured as a function of Vb' It demonstrates that a BEEM signal between 10 -:- 50 pA at room temperature is dominated by noise for a barrier of less than 0.75 V. At 77 K, however, the same signal can be measured for barriers down to 0.2 V. In addition, the current-amplifier noise decreases with increasing impedance of the basecollector junction. Since junction impedance depends on thermally activated processes, for low-energy barrier interface systems it is necessary to perform BEEM measurements at low temperature to obtain large impedance and low-noise spectra. Moreover, reduction in the BEEM system's operation temperature results in reduced smearing of the tunnel tip Fermi distribution and therefore improved spectral energy resolution. Thus the BEEM experiment performed at low temperature gains advantages over room temperature.

135

5.3.3 The Application of BEEM The complexity of Schottky-Barrier (SB) formation phenomena, including the role of interface-defect formation, electrode interdiffusion, and chemical reaction, is expected to induce inhomogeneity into interface structure and electronic properties. For example, the experimental results obtained by other techniques demonstrate that the properties of the Au/GaAs SB interface are strongly affected by interface-defect formation, pronounced interdiffusion and alloy-formation phenomena between the Au and GaAs electrodes. In contrast to Au/GaAs, the Au/Si SB shows simple, reproducible SB characteristics. The topography and the corresponding BEEM images of Au/Si (100) and Au/GaAs (lOa) SB heterojunctions have been obtained. The samples were prepared by evaporation of 1.6-mm diameter Au disk electrodes, 10 nm thick, on chemically etched n-type Si(lOO) and chemically etched n-type GaAs (lOa) wafers. The STM image shows smooth topography at the Au electrode surface for the Au/Si heterostructure. The BEEM image of an Au/Si SB interface also shows homogeneous electronic properties. In contrast to the Au/Si heterostructure, a large degree of interfacial heterogeneity was observed for the Au/GaAs system, with domains of high ballisticelectron transmittance about 2 to 20 om in size. BEEM images of the chemical etched Au/GaAs (lOa) interfaces, prepared whether with or without air exposure before Au deposition, show the same heterogeneous interfacedefect structure. This persistence of the interface heterogeneity indicates that the defect structure is not simply the result of substrate surface contamination. In order to compare SB structures prepared on MBE-grown GaAs with the melt-grown case, and to investigate the role of bulk defects on interface formation, 1-~m-thick GaAs buffer layers were grown on n-GaAs (loa) substrates and an oxide strip grown by chemical treatment prior to deposition of 10 nm-thick Au layers without air exposure. A typical BEEM collector current image of this interface and the corresponding STM topography of the Au surface are exhibited in Fig.5.26. As in the case of the interface systems described above, the BEEM images of this interface again show that only a fraction of the Au/GaAs interface supports ballistic electron transmission. These results indicate that the observed defects cannot be attributed simply to the greater bulk defect density of melt-grown GaAs. It has been known that the formation of the Au/GaAs interface involves dissociation of GaAs and diffusion of Ga and As into the Au electrode. The results of BEEM investigation for this interface indicate that this diffusion process dominates the interface-formation process. Interface heterogeneity observed by BEEM imaging and spectroscopy is attributed to diffusion-induced nonstoichiometry in the form of As-rich precipitates at the interface. 136

Fig. 5.26. STM topographic and BEEM images of a Au/GaAs (100) SB structure prepared by chemical etching of a MBE grown GaAs layer without air exposure before metal deposition. The STM (upper) and BEEM (lower) images were acquired simultaneously. Both images display a 51 X40 nm 2 area. The white calibration bar on the STM image indicates a height of 8 nm. The dark regions of the image are regions of zero detected collector current. Average collector current of the Iight areas is 5 pA [5.44]

The initial GaAs surface stoichiometry is also important in determining the characteristics of the final Au/GaAs interface. The chemical treatment which was used to prepare an oxide-free GaAs surface produces a surface which has been shown to be As-rich. In order to investigate the effects of different initial GaAs surface structures, the Au layers were deposited in UHV on a GaAs substrate grown by MBE without chemical treatment, yielding a Au/GaAs interface prepared completely in situ. Diodes fabricated by this method exhibited an 1-V behaviour dramatically different from samples prepared ex situ. The usual rectifying behavior seen for the Au/GaAs system has drastically been modified, producing an ohmic I-V curve [5.38]. This ohmic behavior persisted in I-V measurements performed at 77 K. These results are interpreted in terms of increased electrode interdiffusion at the interface. Such ohmic behavior was never observed for interfaces prepared on chemically treated GaAs substrates, either melt grown or MBE grown. This drastic difference in diode behavior emphasizes the important role of surface stoichiometry in moderating the diffusion process. In order to inhibit this diffusion, Kaiser et al. [5.44] have grown a diffusion barrier consisting of 2 monolayers (l unit cell) of epitaxial AlAs (loa) on the GaAs buffer layer grown by MBE prior to Au deposition. It is evident that the rectifying behavior has been completely restored in the I-V spectra for the resulting diode structure; the derived SB height is 0.88 eV . STM-topographic and BEEM images are shown in Fig.5.27. The BEEM im137

15

r-:-~~~~-'~~--~~(a)

~:~[2]b) I ' : 2 0 lOt

~" ~1O ~cr5 «

~

5

~

0.5

~20 0-

.

j j

~ ~

40 20

I

0.5

1.0

j,

(b)

0

I

I

1.5

2.0

Tunnel vollage [VI Fig. 5. 31a, b. Experimental (circle symbols) and lheoretical (solid lines) ballislic carrier spectra for the Au/Si (100) SB interface. (a) Experimental ballislic hole (negative lunneling voltage) and ballistic electron (posilive lunneling voltage) spectra measured at 77 K for plype and n-type substrates, respectively. The Au film thicknesses for the p-type and n-type SB structures are 15 nm and 10 nm. respectively. The band gap, valence-band maximum. and conduction band minimum althe subsurface semiconductor interface are clearly seen as thresholds in the hole and electron spectra. The lheoretical spectra are in excellent agreement with the experimental results. (b) Comparison of the ballistic-hole spectrum (dots) with a theoretical fit to the data and evaluated without the requirement of transverse momentum conservation on hole transport at the interface. The drastic discrepancy demonstrates the primary influence of transverse momentum conservation on hole interface transmission for th is system (5.48]

bias, an accurate hole spectroscopy of interface transport and interface valence-band structure is possible. Since the incident hole energy is simply controlled by the bias voltage, direct spectroscopy of interface valence-band structure can be performed. Ballistic-hole spectroscopy can be performed by measuring the hole current transmitted through the interface and reaching the collector as a function of the tunnel bias V applied between the tip and base, Ballistic-hole and ballistic-electron spectra for the Au/Si(lOO) system are displayed in Fig.5.3!. The combined spectra show a region of zero observed collector current bounded by two abrupt thresholds in the current. The threshold for positive tunneling bias (tunneling tip negative with respect to the base) directly indicates the position of the conduction-band minimum. As expected, for negative tunneling bias, the observed collector current is opposite in sign to that observed for positive V. The current provides a spectroscopy of ballistic-hole interface transport. Further, the ballistic-hole spec142

-

I

Tunnel voltage [V] -50 1/ -1.5

80

> 60

-

0

.401-

~

~

10

u

~

~

-~

< W

c

100

I

Wf1 :>

~

=:

I

o

0.5

1.0

Tunnel voltage [V] Fig.5.32a,b. Experimental and theoretical derivative ballistic carrier spectra for the SB structures measured at 77 K. (a) An experimental derivative ball istic hole spectrum (dots) is compared to an experimental derivative ballistic electron spectrum (open circles) for the Au/Si (100) SB interface. The asymmetry between hole and electron spectroscopies is apparent in the experimental spectra and is accurately described by theory (sol id lines). (b) Experimental hole spectrum (dots) of the valence band of the Au/Si (100) SB interface. A theoretical spectrum for a single valence band (dashed line) is in disagreement with experiment. However, a theoretical spectrum for two valence bands (solid line) correctly reproduces the experimenlal result (5.48]

trum threshold directly yields the barrier height formed by the valence-band maximum at the subsurface Schottky barrier. The detailed characteristics of ballistic-hole creation and transport are compared to those for ballistic-electrons in the experimental derivative spectra (Fig. 5 .32a). The marked difference in spectral shape above threshold between the hole and electron derivative spectra reveals a fundamental asymmetry between the collected distributions for electron and hole injection. For ballistic-electron spectroscopy at a positive tunneling bias, the collected electrons originate from the top of the tunneling distribution where the distribution is maximum. The number of electrons created per unit energy remains nearly constant with the bias. However, for ballistic-hole spectroscopy at negative tunneling bias (Fig. 5.32b) the ballistic-hole distribution originates from the bottom of the tunneling distribution where the distribution is at a minimum. Therefore, the number of holes created per unit energy decreases with increasing bias. This fundamental asymmetry between BEEM electron and hole spectroscopy is directly revealed in the experimental derivative spectra for an Au/Si (100) SB interface (Fig.5.32a). The ballistic~hole derivative spectrum displays an abrupt maximum and a sharp 143

decay resulting from the decay in the ballistic-hole distribution with tune 1ing bias above threshold. In contrast, a smooth increase is observed in the ballistic-electron derivative spectrum. It is particularly significant that the accurate understanding of this contrasting behavior presents a sensitive and important test of interface transport spectroscopy and theory.

(a)

5.3.5 Interfacial Modification with BEEM Analogue to STM fabrication on surface, storing information at subsurface region covered by a protective metal film has its unique advantage in application, and also provides an approach to probe interfacial properties. Fernandez et al. [5.49-50] firstly reported their observation that modification in subsurface electronic properties can be induced by ballistic electrons in some specific Au/n-Si system when the energy of injected electrons is several eVs higher than the Schottky Barrier (SB). Their typical modification result was an area of a few hundred angstroms in diameter with decreased BEEM current, surrounded by a ring of enhanced contrast. In addition, protrusions were sometimes observed in the topographic image, indicating slight structural changes in Au film. Since the modification results were strongly dependent on the composition of interfacial impurity layer, the authors attributed the observed variations in BEEM current to be related to the changes at the Au/n-Si interface. As a plausible explanation, they proposed that the ballistic electron current at high bias enhances the interdiffusion of Au and Si at the interface. The created Au-Si intermix layer at the interface thus decreased the ballistic transmission. However, the later discovery of locally enhanced ballistic transmission after a voltage pulse in some cases [5.51-53] has implied that the previously assumed Au-Si alloyment mechanism seems not sufficient. BEES measurements performed inside and outside the modified region have further exhibited that [here exists a huge discrepancy in their natures [5.53]. On the other hand, the dynamic studies [5.52,53] of modification process have revealed a well defined linear relationship between the area of the modified interface region and the applied voltage (Fig.5.33), as well as the duration of the applied pulse, indicating the modification is very likely to be associated with the accumulation of injected electrons. Although the underlying mechanism is not thoroughly understood, yet it is more likely that certain chemical processes might be involved. Further investigations of the compositions of the modified regions should be crucial to understand the principles of the interfacial fabrications.

144

14D

(b)

tID

/I-----~

1m .~

~

'0

m m

ffIJ 6IJ

4D

r

T~~r-----'// l

-"-

aD 0, 0

5

~

~

~

~

duraticn of awIie::l \dta;]e PJlse (soc) Fig. 5.33. (a) The modificat ion outcomes at Au/n ·Si (Ill) interface by a series of the volt~ge pulse~ of -4.0Y and duration from 5 ~o 30 seconds, respectively. The image area is 1200 A X 1200A. The topographic range is 20 A and the BEEM current variation covers a range of2 "730 pA. (b) The modified interface regions increase monoronously with respect ro the duration of applied voltage pulses [5.53]

5.4 Scanning Ion-Conductance Microscope

The Scanning Ion-Conductance Microscope (SICM) is designed to image both the topography of nonconductors (such as membranes) that are immersed in electrolytes, and any ion flow through that sample. As shown in Fig.5.34, the probe of the SICM is an electrolyte-filled micropipette, and the insulating sample is located at the bottom of a reservoir of electrolyte. The probe is lowered toward the sample while the conductance between an 145

Fig. 5. 34. The SICM scans a micropipelle over the can LOurs of a surface by keeping the electrical conductance between an electrode inside the micropipette and an electrode in the reservoir constant by adjusting the vertical height of the probe [5.54]

------

electrode inside the micropipette, and an electrode in the reservoir is monitored. This current serves as the feedback signal for standard scanningprobe microscope electronics. As the tip of the micropipette approaches the surface, the ion conductance decreases because the space through which ions can flow is decreased. The micropipette is then scanned laterally over the sample surface. The SICM topographic images can be obtained by measuring the voltage applied to the z-axis of the piezo-tube translator while holding the conductance constant under a feedback control during a scan. For ion current images, the current flowing into the pipette at each point is monitored as the pipette is scanned over the surface at a constant height. It is also possible to follow the topography with the AC ion current from one electrode in the bath and measure the DC ion current from an electrode below the sample surface. In practice, the ion-conductance signal should be averaged many times at each surface location to improve the signal-to-noise ratio. Just as for other scanning probe microscopes, the sample approach mechanism and the feedback system used for the SICM can be the same as for STM. The critical point is to make a robust micropipette tip with a small inner diameter. The test results reveal that the resolution of the SICM is determined by the inner diameter of the micropipette. In other words, the SICM can resolve features as small as the inner diameter of the micropipette if the noise in the ion conductance signal can be reduced to less than 1 %. The most interesting application for the SICM is not, however, just imaging the topography of surfaces at submicrometer resolution. The SICM can image local ion currents coming out through pores in a surface. Comparison of topographic and ion current images can give a more detailed picture of the type of surface features that correlate with ion channels. Figure 5.35 shows the SICM topographic and ion current images of a Nuclepore membrane filter obtained by Hansma et al. [5.54]. In this model system, ion currents come through the holes in the sample surface. 146

Fig. 5.35. (a) A SICM LOpographic image of the 0.8 J-tm diameter pores in a Nuclepore membrane filter. (b) The same image presented in a plan view. (c) A SICM image of the ion currents emerging through the pores. The imaged area is 7.8 J-tm by 4.5 J-tm for all three images [5.54]

The performance of the SICM has been greatly improved by replacing fragile glass micropipettes by a silicon microfabricated probe [5.55]. The new microfabricated probes are end caps for a 1.5 mm glass capillary. A hollow tip is fabricated in the center of a silicon membrane. An aperture is formed at the apex using microfabrication techniques including photolithography and etching, giving typical minimum diameters of 250 nm. Unlike the glass micropipette, the new tips are compact and hence mechanically robust. The tips are mounted on a flexible membrane that allows the tip to deflect away from the surface in case of a collision with the sample surface. In addition, the microfabricated probes have been designed to have a high mechanical resonant frequency, allowing scan speeds up to 50 times faster than used with glass micropipette probes. The new probes have been used to image the surface topography of a plastic diffraction grating and the ion flow through porous polycarbonate membrane filters [5.55].

5.5 Scanning Thermal Microscope

The scanning thermal microscope can be used to make topographic measurements and to map temperature variations. The tip of a thermal microscope is designed as a tiny thermocouple, for example, tungsten wire with a specially configured tungsten-nickel junction, so that its voltage is propor147

Fig. 5.36. Schematic diagram of a scanning thermal microscope. The probe has a tungsten core, jacketed in nickel but insulated from it everywhere except at the 3D-nm-wide tip [5.56)

Piezoelectric control

Fig.5.37. Surface profile of fixed red blood cells on a glass substrate obtained by the scanning thermal microscope. Each cell is approximately 7 IJ-m in diameter [5.33)

Nickel

Insulator

tional to the temperature (Fig.5.36). If a steady current is passed through this thermocouple junction, it heats up and comes to an equilibrium temperature above the ambient value. When the heated tip approaches the sample surface, its rate of heat loss increases because the solid sample is a much better heat conductor than air. The resulting drop in the voltage across the thermocouple junction depends on the distance from the sample at any point and provides a measurement of the surface topography. The voltage can be used to control tip-sample spacing in much the same way as the tunneling current is used in a STM. In operation, instead of measuring the DC thermoelectric voltage, the tip is vibrated by a few nanometers in the vertical direction, and the ac change in the thermoelectric voltage is used as a monitor of the tip-sample spacing. This renders the system immune to ambient temperature variations caused by room temperature fluctuations and air currents in the vicinity of the probe tip. Figure 5.37 displays an image of the surface of fixed red blood cells obtained with the scanning thermal microscope. The thermocouple tip cannot be made much finer than about 30 nm, which limits the spatial resolution of surface profiles made with the scanning thermal microscope. The tunneling thermometer which can, in principle, have atomic resolution, is based on the fact that as a metal probe tip approaches within 0.5 nm or so from a different metal surface, a potential difference could be measured between the probe tip and the metal surface due to the equalization of the Fermi levels caused by two-way electron tunneling across the gap. In the situation where there is local thermodynamic equilibrium in the gap region, this potential is a measure of the local surface temperature. This technique has been applied to measure the optical absorption spectrum of the gold film [5.57] and to detect differences in the chemical potential signal between molybdenum and sulfur of a cleaved MoS2 sample [5.58]. 148

5.6 Scanning Tunneling Potentiometry and Scanning Noise Microscopy A variant of STM, Scanning Tunneling Potentiometry (STP), allows simultaneous measurement of the topography of surfaces and their electric potential distribution with microscopic resolution, giving an insight into conduction through granular structures, defects, and interfaces. STP requires a minor modification of the STM: two additional electrodes at the sample. A potential difference LlV = V2 - V I is applied across the sample surface (Fig.5.38), while an AC voltage (V TI = VT1 'sinwt) applied across the tunneling gap generates an AC tunneling current which can be used to maintain a constant tunneling gap. VTI is connected to both sample electrodes to generate nearly position-independent tunneling. The DC part of the tunneling current is regulated to zero by shifting the potential of the sample by the amount V R' VR gives the local potential at the scanning position. In operation, an independent control loop whose band (DC to 1kHz) is outside the band of the gap control loop is used to maintain zero DC tunneling current by continuously causing the voltage on the tip to track the voltage on the sample as tip is rastered across its surface. The tip voltage is then equal to the sample voltage at every point on the sample surface. Figure 5.38b shows a schematic diagram of the control system. The distance is regulated for a constant AC tunneling current by means of the lock-in amplifier (LI) in series with the usual logarithmic amplifier (LG) and the control circuity (PI). The STP has been employed to study a goldisland MIM structure [5.59]. The STP is useful for measuring nanometer scale potential variations on devices such as Schottky barriers, pnjunctions and heterostructures. The voltage resolution is typically of the order of a few millivolts. These tech-

149

111

V(x)! (a)

o

v~:

VI

o

x

(b)

Au Electrode 1

Tip

Electrode 2

Vz Fig. 5.38. (a) Schematic of STP. The tip is held at ground potential via a current-to-voltage converter, i.e., V tip = O. The local potential V(x) at the tip is shifted to ground potential. (b) Schematic of sample and feedback circuitry [5.59]

niques have been extended to the EFM which was described in Sect.5.2.5 for measuring potential distributions on insulating surfaces. Another modification of the STM is the scanning noise microscope which can be considered as a STM where no external bias voltage is applied to the tunnel junction [5.60]. Instead of the tunneling current, the meansquare noise voltage from the junction is used to control the tip-sample separation by a feedback loop in the same way as the STM. This technique can be applied to maintain a constant gap resistance when the tip is scanned across the sample because the mean-square noise voltage is proportional to the gap resistance. This technique could have important advantages for special applications such as those in electrochemistry where zero average current is desirable, or for observing delicate samples where extremely small AC currents are required.

5.7 Photon Scanning Tunneling Microscopy and Scanning Plasmon Near-Field Microscopy

Another variant of the STM, the Photon Scanning Tunneling Microscope (PSTM), enables us to probe directly the evanescent field outside the confined-propagating optical fields within the sample, thereby revealing variations in these fields due to topographic changes, the index-of-refraction inhomogeneities, or modal variations within the waveguide. PSTM is the exact analogue of STM. The difference between these two microscope types is that the PSTM works with photon tunneling rather than electron tunneling, and using an optical tip instead of a metal one. Figure 5.35 represents a schematic of the PSTM. In the PSTM the sample is the surface at which total internal reflection of photons occurs unless an optical-fiber probe tip is brought in close vicinity. The optical-fiber tip is drawn to a sharp point and is piezoelectrically positioned, in a manner similar to the tip of an STM, to be moved within the range of the evanescent field. Part of this field is coupled to the fiber tip (Fig. 5.39). The photons are transmitted in the optical fiber to a photomultiplier to produce an electrical signal. Just as in an STM, the tip can be scanned across an optically conducting sample in either of two modes. In the constant-height mode the intensity of the coupled light is detected, and in the constant-current mode the voltage for piezo controlling the height of the fiber tip is measured. By analyzing the signals from each of these

Photomultiplier tube

I

I

Computer control and j mage processing

I Piezoelectric

translator I I

xyz Motion and feedback

I I

I r I I

Elber optic tip

Fiber optic tip Evanescent field .. , .

Totally internally reflected light

Fig. 5.39. SchematicofPSTM [5.61]

150

151

modes, topographic corugations and dielectric variations can be separated with good spatial resolution. In comparison with other optical microscopes, PSTM offers 3-dimensional imaging with subwavelength resolution. It is possible to obtain lateral resolution of one tenth of the wavelength. The resolution normal to the sample is primarily limited by the electronics and is easily one nanometer or less. The PSTM has been employed to study several different waveguide samples, such as a silicon oxynitride planar waveguide, a titanium indiffused lithium niobate channel waveguide, an optical grating, and quartz. In addition, the measurement of photon scanning tunneling spectroscopies on PSTM samples using a split optical fiber may provide important information. Very recently, Specht et al. reported a new form of near-field microscopy-Scanning Plasmon Near-field Microscopy (SPNM) [5.62]. They have achieved lateral resolution of 3 nm (A/200) at optical wavelength. SPNM is based on the interaction of extended Surface Plasmons (SP), as introduced by Ritchie [5.63], with a sharp metal tip which is placed close to the surface of the object. Theoretical considerations indicate that this interaction can be understood mainly in terms of elastic plasmon scattering as well as radiationless energy transfer from the tip to the sample. These processes depend very strongly on the distance between tip and sample and are the key to a high lateral-resolution capability. By raster scanning the tip over the sample surface, nanometer-scale maps of interaction strength (related to the topography of the sample) can be recorded. Unlike most of other near-field optical techniques, SPNM is not based on an aperture (see below), but on metal tips. Taking an example here, Specht et al. utilized a simple, electrochemically etched, tungsten tip (radius of curvature about lOnm) as the near-field optical probe. SPNM uses an ingenious way of extracting an optical signal from the tip. A laser beam incident on the object (for example, a thin silver film evaporated on a glass prism) at a defined angle of total internal reflection excites resonant oscillations of conduction electrons in the silver film which gives a strong evanescent electromagnetic field at the air-silver interface. The plasmons can be recognized by a minimum in the reflected laser intensity which can be understood as destructive interference between light reflection from the silver-air and the silver-glass boundaries. A tungsten tip diving into the evanescent field of the surface piasmons at the air-silver interface reduces the scattering of surface plasmon which radiates in all directions. This also causes a local reappearance of the total internal reflection, which is recorded as a signal for the SPNM as the tip moves over the surface. This change in reflectivity is a measure of the complete tip-surface plasmons interaction, and not just the radiation emitted by the surface plas152

'~Bl a

2Al,

'~'h.

b

10A

.,\~\\\" , ,

, I

,

,

,

\ A11

, I

, ,

,

Fig. 5.40a, b. The STM-stimulatecl photoemission from periodic atomic-scale gratings of one-di.mensional Au atom rows separated by 8.16 A on Au (110) surface. (a) Constantcurrent STM image. (b) Photon map of a stepped area of a (1 X2) reconstructed Au (110) surface measured simultaneously [5.66]

, ,

,

, ,

t~ l~ l ~\ U~ ~ ,

mons into a small solid angle. That is why a relatively large signal is produced from the perturbation by a very small tip. This is an important ingredient in obtaining high lateral resolution. The origin of the surprisingly high resolution of SPNM is not clear. Specht et al. claimed that the elastic plasmon scattering and radiationless energy transfer from the tip to the sample are the main physical processes responsible for the high resolution. However, SPNM does demonstrate higher resolution than any other reported optical microscopy. Additionally, this new technique may permit the detection and spectroscopic identification of single adsorbed molecules. Successful applications of PSTM have revealed the propagation pattern of surface plasmons in silver films [5.64], and the optical modes of silver clusters [5.65]. Furthermore, this technique has provided direct evidence of the electromagnetic interaction between tip and substrate, using Au (110) as an example [5.66] (Fig.5AO).

5.8 Near-Field Scanning Optical Microscopy and Spectroscopy

Another optical microscope with subwavelength resolution is the Near-field Scanning Optical Microscopy (NSOM) [5.67-69]. Optical microscopy and spectroscopy have long been key techniques in medicine, biology, chemistry and materials science. There are a few advantages of optical microscopy and spectroscopy: 153

• Universality. All materials and samples attenuate light and have spectroscopic states. Optical microscopy can be used for observing a wide variety of biological and chemical samples. • Non-destructiveness. Optical microscopy can operate with any transparent fluid medium between the objective and the sample (e.g., air, water or oil) so the sample can be viewed in its native environment. When nonionizing, visible light is used, radiation damage is negligible. Most chemical reactions are not perturbed by long-wavelength light. • Convenience. In most cases, no sample preparation is needed to view specimens with optical microscopy; however, labeling, staining and sectioning are sometimes required. Optical microscopy is inexpensive and simple to operate. Also, optical microscopy is usually safe and precautions are mostly limited to protective eye-glasses. • Real-Time Observation. The speed of optical microscopy is limited only by signal-to-noise ratio considerations, and dynamic processes can be studied with optical microscopy. By using ultrafast light sources, the speed can be extended even into the femtosecond time domain. Thus, biological phenomena, chemical reactions, crystallization etc. can be observed under the microscope as they happen in-situ or in-vivo. • Contrast. There are various contrast mechanisms available in optical microscopy which permit clear imaging of features in a broad range of sampies. These mechanisms, in addition to absorption and scattering, include fluorescence microscopy [5.67-69], polarized-light microscopy, phase-contrast microscopy, and differential-interference microscopy [5.70-73). These are not easily accomplished by other techniques such as electron microscopy or X-ray crystallography. In summary, the conventional optical microscope has numerous advantages which have made it the most popular imaging system. Even with the proliferation of different types of microscopes today, there is no other microscope which can match all the advantages of optical microscopy. However, its primary disadvantage is a fundamental limit to the resolving ability of conventional optical microscopy. The spatial resolution is about A/2, the well-known diffraction limit [5.74], on the order of half a wavelength. This has greatly limited the application of optical microscopy. It was realized as early as 1928 [5.75] that light can be apertured down to much smaller sizes with no obvious theoretical limit. Thus, smaller light sources could be fabricated. The simplest example is the passage of light through a small hole. Whatever light that passes through is confined to the dimension of the aperture in the immediate vicinity outside this aperture, due to the rules governing the "near-field" regime of evanescent waves [5. 75-79]. This principle has been discovered and rediscovered several times, 1928 [5.75], 1956 [5.76], 1972 [5.77] and in the 1980s [5.78,79]. Actually, 154

the photon "scanning probe technique" has preceded all other scanning microscopy. However, only in the 1980s was the principle followed by the optical experiments [5.78,79). More recently came the idea of active subwavelength sized light-sources [5.80-82]. Very recently, these new light sources led to single molecule detection and imaging [5.83-85].

5.8.1 Principles of Near-Field Optics Conventional ("far-field") optical techniques are based on focusing elements such as a lens. This leads to the "diffraction limit" of about AI2 [5.58]. The realization of better resolution by smaller light sources has led to the concept of Near-Field Optics (NFO). The principle underlying this concept is schematically shown in Fig.5A1. The near-field apparatus consists of a near-field light source, sample and far-field detector. To form a subwavelength optical probe, light is directed to an opaque screen containing a small aperture. The radiation emanating through the aperture and into the region beyond the screen is first highly collimated, with dimension equal to the aperture size, which is independent of the wavelength of the light employed. This only occurs in the near-field regime. To generate a high resolution image, a sample has to be placed within the near-field region of the illuminated aperture. The aperture then acts as a subwavelength-size light probe which can be used as a scanning tip to generate an image. Therefore, this optical microscopy is called Near-field Scanning Optical Microscopy (NSOM) [5.67-69]. Unlike Scanning Tunneling Microscopies (STM) or Atomic Force Microscopies (AFM), imaging in NSOM is via the interaction of light with the surface by either a simple refraction/reflection contrast, or by absorption

~ I

\ ..........

~ .. \

,t...

...........

,

g

C'

2000A

D

L: 0: S: D:

AClive or passive light source Opaque malerial Sample (supporl nOI shown) Far field delection syslem (e.g. lens)

Fig. 5.41. Near-field oplics 155

and fluorescence mechanisms. The advantages of NSOM are its non-invasive nature, its ability to look at non-conducting and soft surfaces, and the addition of a spectral dimension, the latter not existing in either STM (room temperature) or AFM. The potential for extracting spectroscopic information from a nanometer-sized area makes it particularly attractive for biomedical research and materials science.

5.8.2 Optical Probes for Near-Field Optics The light source which is the "heart" of the NFO technique has to be: (1) small, (2) intense, (3) durable, and (4) spatially controlled. As is well known, the tip's size determines the resolution, provided that the tip can be scanned close to the sample. In addition, to perform successful microscopy or nano-spectroscopy, the issue of contrast is of the highest importance. There are two major probes used in NFO: metal-coated glass micropipettes and nanofabricated optical-fiber tips [5.67-69]. Optical fiber tips and micropipettes are easily fabricated to sizes of approximately 50 nm, and the smallest nanofabricated optical fiber tip reported to date is about 20 nm [5.68]. The fabrication of miniaturized optical probes has keyed the development and application of NFO in a wide variety of fields. The first step in the probe nanofabrication process is the pulling of micropipette and fiber-optic tips of appropriate size and shape. The second step is the metal coating of such tips. This is followed by crystal (or polymer) growing if active optical or excitonic probes are desired. Here, we illustrate the nanofabrication of optical-fiber tips. Very similar techniques have been applied in the fabrication of micropipettes [5.67,86]. Optical-fiber tips have been used in many areas [5.68,81,82] and can be fabricated either by heating and stretching or by chemical etching. The apparatus for fiber-tip pulling usually consists of a micropipette puller and a CO 2 infrared laser. The CO 2 -laser beam replaces the electric filament in the puller to heat the optical fiber for the pulling process. The laser beam is reflected by a mirror and directed to heat the optical fiber which is fixed on the puller. The details of the pulling setup and procedures can be found in a series of references [5.68,81,82,86,87]. By using appropriate program parameters and laser power, optical fibers can be tapered to subwavelength diameters. After pulling, the optical-fiber tip is coated with aluminum by vapor deposition to form a small aperture. The procedure of vacuum deposition of metals is well known but far from trivial. A specially built highvacuum chamber is employed for coating these pulled fiber tips: only the fiber-tip sides are coated with aluminum, leaving the end face as a transmissive aperture. To make it into a light source, a visible or UV laser beam is coupled to the opposite end of the pulled tip. This probe delivers light very 156

efficiently since most of the radiation is bound to the core until a few micrometers away from the tip. A randomly chosen 0.1 J.Lm optical fiber probe gives 10 12 photons per second [5.86,88]. Using the same puller, glass micropipettes have been pulled with different subwavelength diameters. Both optical fiber tips and micropipettes have been used as optical nanoprobes. Near-field optical nanoprobes can be classified into three different kinds: passive optical probes, such as coated micropipettes or small holes on a screen [5.67,78,79], semi-active light sources, such as optical fiber tips [5.68,69,81], and active light sources, such as nanometer crystal light sources [5.80,82,88]. Compared to a hollow micropipette tip, a nanofabricated optical fiber tip is a "semi-active" photon tip which is orders of magnitude brighter, easily coupled to an optical source and at least as mechanically sturdy as a micropipette. It is interesting to notice that the top of a fiber tip is really very resistant to breakage. The photochemical stability for optical fiber tips is excellent and under very intense illumination it is the heat that damages the aluminum coating at the tip. Both probes have been made around 500 A in diameter without difficulties in applications as light sources.

5.8.3 NSOM Operation The two most important modes of traditional optical microscopy are transmission and luminescence. For NFO's chemical and biological applications these will probably continue to be the most important modes, especially if one includes nano-spectroscopy (see below). Also, various reflection and collection mode NSOM techniques have been devised [5.67-69]. Figure 5.42 gives several arrangements used in NSOM. There are several contrast methods: absorption, refractive index, reflection and fluorescence (luminescence), and not all of them are well understood. One can also count polarization [5.70,71] and spectroscopy as separate modes of contrast. Furthermore, there is a large number of quantum effects, such as energy transfer, energy down-conversion and energy quenching [5.69]. The simplest optical contrast mechanisms in the near-field regime, e.g. refractive index, are not yet well understood [5.93] and thus they are under intensive study. Actually, the microscopic quantum effects are better understood than the mesoscopic (near-field) optical interactions. The most important consideration for sample preparation is sample roughness. It is limited by the probe shape in the most obvious way (the same as for all scanning probes). The sample thickness is an important factor for all transmission (forward scattering) modes of operation, but not for the reflectance (back-scattering) and some "collection" modes. The near-field approach couples the optical resolution with the 157

A

(a) •

\

\ \

\ \

\

(b)

I

I

I I

I

I I I \ I I /

(e)

C=:J .. ~~ ,

(d)

C=:J

I r I I I' , II

U

Fig. 5.42. Experimental implementations of OAs at a controlled, adjustable distance from an object surface: (a) transm iss ion/aperture [5.79], (b) reflection/aperture [5.73], (c) reflection/asperity [5.90], (d) transm iss ion/aperture (collection mode) [5.91, 92)

distance from the probe; the higher the desired resolution, the thinner the required sample. On the other hand, the contrast mode (absorption, refractive index) may limit the thinness of the sample. Even the fluorescence mode may be limited by the thinness of the sample i.e., the absorption cross section. However, this can be overcome by intensity, by auxiliary fluorophores or by quantum mechanisms (energy transfer). Thus, the various luminescence modes appear to be the most promising modes of forward-scattering near-field microscopy. The best resolution to date has been claimed [5.68] to be about 12 nm (with 514nm light). Presumably this was achieved with a 20 nm diameter aperture. A signal of 50 nanowatts has been claimed for an 80 nm aperture [5.68]. Also, NSOM has successfully been applied in single-molecule detection (see below) [5.83-85]. An example would be the study of the domain boundary and the web structures of phospholipid monolayers with SNOM [5.94] (Fig.5.43).

5.8.4 Near-Field Scanning Optical Spectroscopy Near-field Scanning Optical Spectroscopy (NSOS) [5.69,95] is based on NSOM. It basically adds one more dimension, spectroscopy, to NSOM and can be used to obtain spectra of various nanostructures, such as nanocry158

Fig. 5.43. FFM images of DPPC/O. 5 mol % Bodipy-PC monolayer sampled at different surface pressure. (a) 7mN/m; (b) IOmN/m; (c) 30mN/m. (d-f) are th NSOM images of these sample, respectively. The web structure in (e) and (f) reflect the presence of nanoscale crystals with hexagonally packed lipids [5.94)

stals and quantum wells. NSOS inherits all the advantages of NSOM and adds a spectral dimension to the near-field optics technique. The ability to obtain spectroscopic information with a naometer-sized resolution makes NSOS very promising for a wide variety of scientific researches. Examples include the detection of fluorescent labels on biological samples and isolating local nanometer-sized heterogeneities in microscopic samples. People have studied microscopic crystals in order to demonstrate that nanoscopic inhomogeneities can be detected in what might at first appears to be a homogeneous sample [5.95]. The eventual goal is to obtain spectroscopic information with single molecules. The NSOS apparatus is quite similar to that of NSOM [5.69]. In NSOS an optical probe with an emissive aperture that is submicrometer in size is positioned such that the sample is within the near-field region. With piezoelectric control of the fiber tip, the tip can be accurately positioned over a fluorescing region of the sample and a spectrum recorded. Excitation of the sample can be either external with detection through the fiber tip or with the fiber tip itself and subsequent detection of the emitting photons. This means that it is not necessary for the sample to be of any particular thickness or opacity, however it should be a relatively smooth surface. Optical 159

HeCd or Ar+ laser

STM electr.

. ;M: n

Mirror

. Plezo

Computer

Lens

::::: ::::: :.:::

5.8.5 Near-Field Optical Chemical Sensors OMA Near-field microscope

---=--...::::::::::: Fiber tip

Sample Objective---to detector

~

Fluorescence spectrum Fig. 5.44. Schematic of experimental apparatus used for NSOS

probes used in NSOS are the same nanometer-size optical-fiber light sources for sensor preparation [5.68,69]. The experimental apparatus for measuring fluorescence spectra with high spatial resolution is sketched in Fig.5.44. Here a 422 nm line from a HeCd laser is coupled into an optical fiber with a high-precision coupler. The fiber tip is mounted in a hollow tube of piezoelectric material which is positioned by the usual STM control electronics. The sample (deposited on a glass slide) is mounted on the nearfield microscope such that it is perpendicular to the exciting tip and the entire apparatus rests on the base of an inverted frame microscope with a reflected light fluorescence attachment. Excitation of the sample via the fiber tip generates fluorescence which is collected by an objective, filtered (to remove laser light) and collimated before exiting the microscope. The fluorescence is then focused onto an Optical Multichannel Analyzer (OMA), and thus the data are collected and analyzed in a computer. Different samples have been studied. For example, films of a 1.0 wt. % mixture of tetracene in polymethylmethacrylate were prepared by spin-coating a dichloromethan solution on a glass slide [5.95]. Tetracene/PMMA films examined under the fluorescence microscope show microaggregates of tetracene with an average size of about 10 p.m embedded in the polymer. 160

The background fluorescence from the film appears greenish-yellow and is presumed to be from either isolated molecules or crystals which are of a size smaller than what can be resolved with the conventional microscope. What is surprising about the aggregates is that this fluorescence ranges in color from green to yellow to red. Thus, the macroscopic fluorescence spectrum obtained with conventional (far-field) light source excitation is very broad, containing contributions from the background and all colors of aggregates. With NSOS it is then easy to excite a specific aggregate and record individual clusters' fluorescence spectrum.

Micro-spectroscopy has often been utilized for chemical analysis and biological intracellular analysis. For instance, a reagent is introduced into the cell (e.g., with a micropipette under the microscope) and the color or spectrum of the cell changes and provides information about the pH or the calcium content of the cell. More recently, fiber optical chemical sensors have been introduced for such measurements [5.96]. However, their spatial resolution has been limited by the physical size of the optical fiber, typically 100 p.m. A spin-off from NSOM technology is the development of submicrometer, subwavelength Near-field Optical Chemical Sensors (NOCS) [5.82,88]. Their chemical preparation by photopolymerization is based on near-field optical excitation, which limits the size of the produced probe [5.80]. In addition, the sensing occurs in the near-field regime of the optical excitation, thus highly increasing the sensitivity per photon and per sensor molecule. This near-field operation has decreased the volume needed for non-destructive analysis to well below a femtoliters [5.88]. Such a subwavelength pH sensor is schematically shown in Fig.5.39. The first biological application of submicrometer NOCS [5.82,97] was demonstrated for lO-day and l2-day-old rat conceptuses. The NOCS consist of an aluminized fiber tip with a copolymer supertip containing the pH sensitive dye [5.88]. The analysis is based on ratios of fluorescence intensities at different wavelengths of the same spectrum, or on ratios of fluorescence intensities at different wavelengths of two different spectra obtained by two different excitations (ratios of ratios), providing for internal calibration [5.82]. The intra-embryo pH were 7.55 for 10-day rat conceptuses and 7.27 for l2-day rat conceptuses, respectively. These values are in good agreement with the reported results for "homogenized" rat conceptuses samples, where more than 1000 embryos had to be crushed. In contrast, only one single embryo was needed for the pH measurements via miniaturized NOCS. In addition, chemical dynamic alterations in pH of intact rat conceptuses, in response to several variations under their environmental condi161

Fig. 5.45. Schematic drawing of subwavelength optical fiber chern ical and biological sensor

tions, have been measured. This is the first time that such an experiment has been carried out on a single and live rat embryo. The ability of the sensors to measure pH changes, in real time, in the intact rat conceptus, demonstrates their potential application for dynamic analysis in small multicullular organisms and single cells. Compared to conventional devices [5.96], a thousand fold miniaturization of immobilized optical-fiber sensors, a million fold or more samplesize reduction and at least a hundredfold shorter response time have been achieved by combining nanofabricated optical-fiber tips with near-field photopolymerization. Also, the submicrometer sensors have improved the detection limits by a factor of a billion [5.88].

copy. It has been called Scanning Exciton Microscopy (SExM) [5.98]. It has also been called Molecular Exciton Microscopy (MEM) [5.99]. MEM is conceptually quite similar to STM. The excitons "tunnel" from the tip to the sample. However, there is no driving voltage or field. Rather it is the energy-transfer matrix element which controls the transfer efficiency. Its unusual matrix elements allow for the highest sensitivity to distance, higher than that of STM and comparable to that of AFM. In addition. the most striking result of this direct energy transfer is its ultrahigh sensitivity to isolated or single molecular chromophores. The quantumoptics energy transfer is highly efficient within the range of the Forster radius. Thus, a single excitation could be "absorbed" by the sample acceptor. In contrast, based on the Beer-Lambert law, about a billion photons are needed to excite a single acceptor in the absence of other acceptors. Furthermore, as the distance range is limited to about 10 nm for the direct energy transfer, MEM is as much a near-field technique as STM or AFM, i.e., very sensitive in the single digit nanometer range and much less sensitive beyond 10 nm. However, in combination with conventional NSOM, the range can be extended to about 200 nm. Thus MEM is a technique which is able to "zoom in" from macroscopic to nanoscopic distances. Obviously such a "zooming in" enhances the speed of operation. It also allows for a much more universal range of samples, from metal spheres and clusters to soft, in-vivo biological units. In addition, MEM can use fluorophores, metal-clusters, etc. to enhance contrast, sensitivity and resolution with the help of NSOM. It can also be used in conjunction with lateral force feedback, in the same way as NSOM.

5.8.7 Single-Molecule Detection by Near-Field Optics 5.8.6 Scanning Exciton Microscopy The concept of active light sources enables a totally new mode of NSOM, based not on the blocking or absorption of photons but rather on quenching directly the energy quanta that otherwise would have produced photons. For instance, a thin, localized gold film (or cluster) can quench an excitation (or exciton) that would have been the precursor of photons. Furthermore, a single atom or molecule on the sample could quench (i.e., by energy transfer) the excitations located at the tip of the light source. For simplicity, we assume that the active part of the light source is a single atom, molecule or crystalline site, serving as the "tip of the tip". This quenching energy transfer from the excitation source's active part (donor) to the sample's active part (acceptor) mayor may not qualify technically as an NSOM technique. However, it is the best hope, currently, for single-atom or molecule resolution and sensitivity. This technique basically is a quantum optics micros162

Emerging techniques aimed at Single-Molecule Detection (SMD) have potential applications across the physical and biomedical sciences. SMD represents the ultimate goal in chemical analysis and has been proposed as a tool for DNA sequence. Traditionally, molecular structure and dynamics were observed by averaging techniques, such as X-ray crystallography, electron diffraction, and various spectroscopies. On the other hand, electron microscopy and related methods do indeed image single molecules but at a heavy cost to their integrity - observing them in a vacuum and/or under highly perturbative conditions. Recent methods such as STM and AFM come closer to the ideal but there are still some difficulties. These problems are particularly acute for the soft organic/biological molecules. In addition, the observation cannot be performed in-situ or in-vivo, and rarely even invitro. Furthermore, it is impossible or nearly unfeasible to observe the molecule dynamics. Near-field optical microscopy and spectroscopy [5.51- 48] is 163

a new tool providing hopes for highly improved imaging at a single-molecule level [5.83-85]. In a recent study [5.84], individual carbocyanine dye molecules in a submonolayer spread have been imaged with NSOM. About two-dozen isolated dye molecules are imaged within seconds. The imaging resolution is about 50 om, and the molecular location is resolved within about 25 nm in the horizontal plane and 5 nm in the vertical direction. Furthermore, the much smaller molecular transition dipole is a point detector mapping out the electric-field distribution of the near-field light source. In addition to imaging individual dye molecules, we can also obtain information on the orientation of these molecules (via polarization and transition dipole fitting). Another avenue is to obtain spectra characteristic of a single-molecule or molecular aggregate. Also, the mechanism of light-matter interaction may be different in the far- and near-field regimes, leading to different spectral selection rules and, in particular, to an enhanced cross section of light absorption (and thus fluorescence) [5.68,86]. These phenomena are an extra bonus for near-field detection. Similar SMD work has been done on rhodamine-6G molecules. The photophysics and photochemistry of this molecule have been investigated on the single-molecule basis [5.85]. In short, the NSOM approach to SMD permits the determination and localization of single-dye molecules. The advances towards nanometer-resolved microscopy, spectroscopy and chemical sensor probes promise to push chemical analysis much closer to one of its ultimate goals - the non-invasive detection of a single molecule, radical or ion, the determination of its precise coordinates and the characterization of its structural conformation, as well as its internal dynamics and energetics, as a function of time and environmental perturbations.

6. STM Studies of Clean Surfaces

The first and major applications of STM have been in the area of surface physics. Metal and semiconductor surfaces are the seat of many interesting phenomena, both of a practical and fundamental nature. Almost invariably, understanding these requires a detailed knowledge of the atomic structure of the surface. A full array of surface analysis techniques have been used to probe metal and semiconductor surfaces and have shown that in many cases the surface adopts a structure different from a simple termination of the bulle Some surfaces are so complicated or subtle that these techniques have failed to completely determine the structure. The feature that sets Scanning Tunneling Microscopy (STM) apart from all the other structural techniques is the capability to probe not only the geometric but also the electronic structure with atomic-scale resolution in real space. Therefore, STM has already led to unprecedented new insights into the surface structure of metals and semiconductors. In this chapter scanning tunneling microscopy and spectroscopy on clean metal and semiconductor surfaces will be discussed, structures of adsorbate-covered surfaces will be dealt with in Chap. 7.

6.1 Metal Surfaces The metal surface is easily contaminated in air, therefore well controlled Ultra-High Vacuum (UHV) conditions are required for accurate and reliable studies of clean surfaces. When a surface is created by cleavage or annealing, the local energy of the system increases by the surface energy. The surface is either simply strained, or it adopts a completely reconstructed bonding configuration to minimize the surface energy. While large-density corrugations (ca. O.lnm) are often observed on semiconductor surfaces due to the presence of dangling bond, those on (1 x I) metal surfaces are small «O.Olnm), as measured by helium diffraction experiments, unless they are reconstructed or foreign atoms are chemisorbed on them [6.1]. For example, metal corrugations are 50 to 100 times smaller than those observed on the Si (111)-7 x 7 reconstructed surface. Atomic imaging of a 164

165

(1 x I) metal surface, therefore, requires high lateral and vertical resolution. In Sect. 4.3.4, we have discussed the role of the tunneling tip and its particular importance in metal studies.

6.1.1 Geometric Structures The surfaces of Pt, Ir and Au exhibit a variety of surface reconstructions of which the I x 2 structure of the (110) surface has attracted the most attention. Although various models have been proposed on the basis of LEED and scattering experiments, none of these models accurately fits all of the experimental data. One reason might be the nonlocal, averaging character of diffraction experiments. As mentioned previously, local information can be obtained by STM. Thus, this unique technique can be utilized to provide experimental evidence of the basic driving mechanism for the various reconstructions as well as the disorder. For example, STM studies [6.2] of a Au (110) surface have revealed various typical features which consist essentially of clearly separated parallel hills usually running several hundred Angstroms along the [110] direction. Most of the hills are separated by 0.8 nm, thus forming the I x 2 reconstructed ribbons which are further separated by steps and I x 3 channels. This results in considerably stronger disorder along the [00 I] than along the [110] direction. High-resolution images can also reveal an increased density of I x 3 channels and a transition from a I x 4 channel to two I x2 channels. Considering the deep and symmetric I x3 channel and such I x 3 channels together with a sequence of other channels and steps, the following basic driving mechanism for the various reconstructions has been proposed: the reconstructed surface consists of long ribbons of narrow (111) facets along the [110] direction with a maximum of three free rows. Two-row facets give rise to the I x 2 reconstruction of the missing-row type, while three-row facets produce the I x3 reconstruction. Combinations of two- and three-row facets can give other local reconstructions and are the cause of disorder. In other STM studies of the same surface [6.3,4], the topograph shows I x 2 reconstruction with alternate [110] missing rows, in agreement with other experimental results. The observed ordered domain size also agrees with the broadening of the fractional-order spots in LEED measurements. These investigations also provide evidence for significant atomic motion of gold atoms on the Au (110)-1 x 2 reconstructed surface at room temperature. Sequences of images reveal that structural changes are associated with kink sites on (100) microfacets, supporting earlier suggestions that they play an important role in mass transfer. Step edges are formed by closepacked atomic rows in general. In the vicinity of domain boundaries, how166

ever, steps exposing (331) microfacets or steps along the [l00] direction are pinned by domain boundaries. These observations provide local information of direct relevance to nucleation and growth processes of Au(110)-1 x 2 phases. Due to the delocalized character of metal valence electrons the atomic corrugation of metal surfaces observed in STM is found to be much smaller than in the case of semiconductor surfaces. In fact, only a few studies report on the resolution of the individual atoms on a metal surface. Au (111) surfaces have been imaged by STM with atomic resolution [6.5,6]. In the STM study of a Au (111) film prepared by epitaxial evaporte deposition, the atomically resolved images are obtained both in UHV and in air [6.6]. The high resolution and large corrugations are attributed to the existence of a surface state near the Fermi level. Individual atoms on the Au (100)-5 x 20 reconstruction have also been imaged by Kuk et al. [6.7], but they believed that the unusually high spatial resolution and corrugations are the result of foreign atoms picked up by the tip during scanning. Al (111) is another example of atomic resolution on a metal surface. The STM image taken in vacuum shows the individual atoms on a closepacked surface of the nearly free electron metal (Fig.6.1). Following extensive cycles of cleaning and annealing (to 800K), a large fraction of the surface is found to consist of extended, atomically flat terraces of several hundred Angstroms wide, which are separated mostly by atomic steps. Atomically resolved STM images of a bare Cu(1IO)-1 xl surface and a Ni (110) surface have also been investigated, and will be illustrated in the next chapter together with the discussion on gas-induced reconstruction. Although studies on metal surfaces with atomic resolution are scarce, the main characteristics of metal surfces, such as monatomic steps, terraces and flatness can be imaged relatively easily by STM at very high resolution, as in the cases of Ag, Pt, Cr films, Cu-Al, AI-Co-Cu and AI-Cu-Fe alloy

. --..... .".-. -.••••••••• ... .. .. '..•••••• ..••••••••.......••••••....•••• •• ffIr.·'" ...••• ..... ..-". .... .. .. .. . . . -... . ....

3.0

..~ ~",,'

2.0

••

..~

E

c

'

_

~

2

>,

e.o

Fig. 6.4. Energy dispersion of the surface state of Au (III) measured by photoemission (solid line), obtained by STS (circles), and calculated from k = (k II + k F )/2 (dotted line); k F is the wave number at the Fermi level, and k II is the wave number measured by photoemission. (Courtesy of Ph. Avouris)

:'-1 . . .

I'''' '-"" t '_

2 nm 3

-

,. ~'!

4

+240mV -240 mV _.- overage 5 ---- difference

" . . \.

°

Fig. 6.22. STM image of CuO chain in the layer ofYBa2 CU3 7 , (a) Y bias = +240 mY; (b) Y bias = -240 mY; (c) the average of (a) and (b): (d) the difference of (a) and (b). (e) A cross section along the CuO chain marked by the end points of the arrows showing the periodity [6.69J

203

7. Surface Adsorbates and Surface Chemistry

As demonstrated in Chap. 6, STM has had a substantial impact on studies of the atomic and electronic structure of metal and semiconductor surfaces. This technique has also been employed to investigate physical and chemical processes occurring at surfaces, rather than the physical properties of the surfaces themselves. A theoretical model has been developed for the study of tunneling of electrons from the probe to the surface where molecular species had been adsorbed, and with atomic and molecular species intervening between the probe and the surface [7.1]. This theory relies on assessing the changes in the tunneling current caused by variations in the size and structure of surface adsorbates. Some results have been achieved in studies on such issues as the nature and the activities of atoms, molecules and clusters in adsorbate/substrate systems, nucleation and growth of metal films, surface diffusion, as well as the atomic details of surface chemistry. The present chapter deals with the application of STM in this area.

7.1 Adsorption on Metal Surfaces The study of adsorbate-covered metal surfaces is one of the important subjects of surface science. Determination of the atomic positions on surfaces and measurement of the coverage of surface species are crucial in modeling surface structure. A variety of techniques has been used to study adsorbate bonding, surface diffusion, ordering processes and reactions with the surface or other adsorbates. Except for a few methods, a common disadvantage for those conventional techniques is that the structural information is a result that is statistically averaged over large areas, comparing with the dimensions of typical features of the surface structure rather than local properties. STM, on the other hand, has led to unprecedented new structural insight into local features of the surfaces such as the chemisorption of adsorbates on metal surfaces and especially the restructuring of metal surfaces. Its high-resolution capability opens new possibilities for the investigation of a large number of adsorbate-substrate interactions, even for those involving non-periodic structures. Not only are static surface structures rev205

ealed atom by atom but, perhaps even more important, the dynamics of the chemisorption process, i.e. the nucleation and growth of the reconstructed phases, can be studied in real space and time at a rate of several images per second.

7.1. 1 Cu (11 0)-0 The chemisorption of gases on transition-metal surfaces is of particular interest for its role in heterogeneous catalysis. For adsorbates which interact strongly with the substrate, the chemisorption process is often accompanied by the breaking of several nearest-neighbor bonds within the substrate lattice, resulting in a reconstructed surface phase with a substantially altered atomic density in the topmost layer. Oxygen chemisorption on Cu surfaces is a prototype of this category of reconstruction which has provoked considerable experimental and theoretical interest. It is known that molecular oxygen chemisorbs dissociatively on Cu (110), and that LEED patterns show a 2x 1 structure at an oxygen coverage of 0.5 ML (MonoLayer). The halfand integer-order spots of the LEED patterns are of comparable intensity, indicating a surface reconstruction. However, the detailed atomic structure of this Cu (110)-2 x 1 surface is still open for vigorous dispute [7.2]. It appears that such information is of utmost importance also in the understanding of the static surface structure. The two most common structural models proposed in the past are the "missing row" model, where every second [001] row on the surface is absent, and the "buckled row" model, where every second [001] row is shifted outward. By means of STM it is possible to study the microscopic mechanism for nucleation and growth of such adsorbate-induced reconstructions. Figure 7.1 illustrates a series of STM images of a surface for increasing oxygen exposure. Atomic resolution of the bare Cu (110)-1 x I structure is clearly observed in Fig.7.1a. The formation of "added" rows of atoms (interpreted as O-Cu chains) along the [001] direction is initiated when the surface is exposed to oxygen at 373 K for exposures ranging from 0.1 to 1 L (1 L = 10- 6 torr' s). The shortest O-Cu chains appear to be about 6 x 0.36 nm 2 indicating a critical minimum length. At higher exposures (ca. 1--:-2L) resulting in an 0 coverage of 0.1--:-0.2 ML, these "added" rows are found in islands developing a unit mesh, with a periodicity that is doubled in the [110] direction (Fig.7.1b). Typical dimensions for these islands are 10--:-20 om in the [001] direction and 1.5 --:- 2.0 nm in the [110] direction, corresponding to preferential growth in the [001] direction. This structure is consistent with the observation of streaky 2 x 1 LEED pattern for low oxygen exposures, indicating a lack of order in the [l10] direction.

206

Fig. 7.1. (a) Atomically resolved STM topographic image of a 2 X2 nm 2 region of a bare Cu (110)-1 XI surface recorded with V = -0. 35 v and I = 2. 5 nA. (b) STM image (7 X7nm 2 ) showing the formation of "added rows" after exposure with ::::::1 L of oxygen. (c) STM image of the Cu (110)-2 XI-O reconstructed phase at an oxygen exposure of about 10 L recorded with V ::::::-0.8 V and I ::::::0.8 nA. The [00 I] direction coincides with the y-axis in all cases [7.2]

Exposures up to about 10 L lead to an 0 coverage of ::::::0.5 ML where most of the surface is covered with the reconstructed phase. However, several types of defects or irregularities may be observed, as shown in Fig. 7.lc. There is a reconstructed terrace (A) one atomic layer below the top layer, a single chain of atoms (B) between two reconstructed anti-phase domains shifted away from the nearest neighbor chain by an extra [l 10] 2 x 1 lattice parameter, a point defect (C) developed from a vacancy, and a region (D) between in-phase reconstructed areas which shows a very weak corrugation with a periodicity consistent with a c6 x 2 reconstruction. This Cu (11O)-c6 x2 0 structure will be discussed later.

207

7.1.2 Cu (100)-0 In the case of the oxygen-induced reconstruction of a Cu (I 00) surface, the STM images show a phase different from the Cu (110) reconstruction. Prior to oxygen exposure, the single Cu atoms on the Cu(lOO)-1 x 1 surface can be seen in Fig.7.2a. After the crystal is dosed with 1000 L oxygen at 573 K and at a pressure of 2.5 .10- 6 mbar, and followed by an anneal at 573 K for 5 min, the sharp LEED pattern obtained reveals the (2v2xv2)R45° 0 structure. Figure 7.2b depicts an STM image of this surface over an area of 1.5x1.3 nrn 2 . The [001] and [010] directions are equivalent on the Cu(lOO)

surface, and thus the reconstruction is twined into two different domain orientations which appear to be randomly distributed. In this image the chains appear to be grouped in pairs of two, with separations from A to B and from B to Care 0.29 nm and 0.43 nm, respectively. This implies that the structure in the [010] direction repeats itself for every 0.72 nm. The chains of bright spots along the [001] direction can be interpreted as Cu-OCu chains/bonds, equivalent to the observations for the Cu (110)-2 X 1 0 structure. The [010] periodicity of 0.72 nm can be explained by a removal of every fourth row of the Cu atoms.

7.1.3 Dynamics

Fig. 7 _2. Atomically resolved STM image of a 1. 5 X 1. 5 nm 2 region of (a) a bare Cu (100)-1 X 1 surface, and (b) a fully developed Cu (100)(2V2X V2)R45° 0 structure. (c) Atomistic model of the Cu (100)- (2V 2 XV2) R45 ° 0 reconstructed phase. The small and large open circles represent the 0 atoms and the remaining Cu surface atoms, respectively, whereas the gray and black circles represent Cu atoms in the 2nd and 3rd layer. The arrows indicate the missing row of Cu atoms. In (b) and (c) a unit cell is indicated [7.3]

208

While the examples described above provide information on atomic arrangement of surface structure in equilibrium, some studies have been performed on the dynamics or the driving force responsible for structural transformation, which include the nucleation and growth of oxygen-induced Cu(l10) and Cu (100) reconstructions, instabilities inherent in the Si (I 11)-pseudo 5 x 5 Cu reconstructed surface and the formation of disordered regions on the Ge (l11)-c2 x 8 surface. Here, we discuss the Cu-O system, the case for Si and Ge will be considered later. As mentioned in the last chapter, if the STM has a high mechanical resonance frequency, images can be recorded sequentially in about 1 second, and one can follow snapshots of dynamical processes. Apart from being of utmost interest in itself, the dynamic information gained from such STM "movies" is very decisive also for the understanding of the static surface structures. It has been shown that in some cases, although there are strong similarities in the final gas-induced reconstructions, the growth modes may be dramatically different, and that direction information on the number of atoms in the unit cell of a certain reconstruction can be derived from the change in area of particular domains. As described above, when the surface of Cu (II 0) is exposed to oxygen at about 373 K with exposures ranging from 0.1 to 1 L, the 2 x 1 structure is initiated and the reconstruction appears in the form of rows of atoms (interpreted as O-Cu chains) along the [001] direction. In order to investigate the detailed mechanism underlying the formation of the O-Cu chains, the dynamical growth of the reconstructed phase has been studied by imaging a region of several terraces separated by steps. While the crystal is exposed at room temperature to oxygen at a pressure of 1· 10- 8 mbar, it is seen from the sequential STM images that Cu atoms are removed exclusively from edges of the terrace, and that the rate of removal varies at different points along the terrace edge. Simultaneous growth of O-Cu chains, which later agglomerate to reconstructed islands on the terraces, preferentially along 209

the [001] direction, is observed. From the observed great mobility of both single and groups of O-Cu chains, Besenbacher et al. [7.4] concluded that the reconstructed phase of "added row" grows on top of the terraces by nucleation of Cu atoms (coming from step edges) and 0 atoms diffusing onto the surface. The "added-row" model is identical to the previously adopted "missing-row" model at the saturation coverage of 0.5 ML, but the two models differ significantly in terms of mass transport. For the added-row model, the Cu atoms are supplied from step edges, as discussed above, whereas the missing-row model would lead to a mass transport from terraces to step edges. A distinction between the added-row and the missingrow reconstruction types can hardly be obtained by any other technique. The nucleation and growth of the Cu(100)-(2V2XV2)R45° 0 reconstruction phase has also been studied by continuous imaging of two terraces separated by a monoatomic step during oxygen exposure at room temperature at a pressure of 1.5,10- 5 mbar. The dramatic change depicted in the STM topographs reflects the nucleation and growth of small islands on both terraces, whereas the step is essentially intact. This is in contrast to the growth mode for the Cu (110)-2 x 1 phase. The islands, which have a height of 0.18 nm [the interlayer distance for Cu(lOO)], grow preferentially along the [010] and [001] directions. At saturation they cover 25 % of the surface area, as determined from a plot of the height distribution. These observations give the first direct proof that the oxygen-induced reconstruction of the Cu(100) surface is of the missing-row type, with one quarter of the Cu (001) rows in the surface layer "squeezed out", and that these extra atoms nucleate and grow epitaxially in small islands on top of the Cu surface. At oxygen coverages above 0.5 ML a c6x2 phase coexists with the 2x 1 phase on the Cu(110) surface. A dynamical study shows that patches with one layer of Cu missing appear on the surface in the late stages of the formation of the 2 x 1 structure. These patches serve as Cu reservoirs for the added rows in the absence of steps. Further oxygen exposure leads to a build up of the c 6 x 2 structure both on the terrace and within the patches, starting at the edges of the patches. The c6 x2 structure consists of two 0Cu chains for each three [110] 1 x 1 lattice spacings [7.5], as compared to the 2 x 1 structure where there is only one O-Cu chain per two [110] 1 x 1 lattice spacings. The O-Cu chains are connected by Cu atoms, coordinated to every second 0 atom along the chain. These Cu atoms, which are gliding on top of the structure, constitute a c6 x2 "superstructure" with respect to the underlying bare 1 xl Cu surface. The O-Cu chains have been seen as pairs of rows in between the protrusions only in very highly resolved images. The STM movie shows that the c 6 x 2 phase grows isotropically (in contrast to the growth mode for the 2 x 1 phase) and that the protrusions forming the c6 x 2 structure have a high mobility along the [001] direction as single units. During the build up of c6 x2 there is an increase in the sizes 210

of the patches. Since the 2 x 1 structure at this point is already fully developed, this additional Cu supply from the patches indicates an increased density of Cu atoms in the c 6 x 2 phase compared to the 2 x 1 phase. The result of a thorough analysis of the height distributions reveal that the c6x2 unit cell contains 10 Cu atoms corresponding to 5/6 ML. As discussed above, with a high-stability STM one can record STM images sequentially in about one second and thereby can visualize in real time space-dynamical processes on metal and semiconductor surfaces. Such information is very decisive for a full understanding of both the growth mode of a reconstructed phase and the resulting static structure. Furthermore, by analyzing a large number of pictures (with atomic resolution) concerning dynamical processes on surfaces, it is possible to study fundamental atomic quantities like diffusion constants and interaction energies by STM measurements.

7.1.4 Ag (110)-0 Silver is used in the chemical industry as a catalyst for partial oxidation of ethylene. Thus, extensive investigations have been carried out to examine the adsorption characteristics of oxygen on Ag surfaces. However, little is understood of the role of molecular oxygen on the surface, even though it is believed that the epoxidation of ethylene involves molecular oxygen. Hashizume et al. [7.6] have applied STM to investigate the reaction on such a surface and proposed an "added-row" model for the atomic oxygen adsorption on the Ag(110) surface, similar to the case of oxygen adsorption on the Cu (110) surface (Sect. 7 .1.1). After the Ag (110)-1 x 1 surface had been exposed to 5.10- 9 torr oxygen, a series of STM images reveal that the monoatomic steps of the clean Ag (110) surface are not stable at room temperature and diffuse much faster than the STM scanning speed, which results in the zigzag-step-line shape. At an oxygen exposure of 14 L, bright lines ("added rows ") can readily be observed in the direction of [001], which is perpendicular to the Ag atomic row on the Ag (110) surface. The average spacing between "added rows" is 13ao (ao = 2.98;\') at this oxygen exposure. After about 60 L of oxygen exposure the movement of the steps is reduced. A zoomed-in image exhibits the "added row" formation with an average separation of 5.2ao ' With increasing oxygen exposure, the average separation between these "added rows" is reduced (Figs. 7 .3a, b) and after the surface is exposed to a large amount of oxygen, the well-ordered 3 x 1 (at around 1600 L), and eventually (at, or more than, 4000 L) the terminal 2 x 1 surface, consisting of the "added rows", are obtained (Fig.7.3c), in agreement with LEED observations. Note the presence of various separations at the same time (4ao , 211

:.•..:....... -".

... _.... _.:.. _.:..~~:

r:[IIOl

x

[OOll ( (

XX XXX XX I:X XX XX :II:

xi'

X

(I I: X:I I: X X:l I: X X X (' I:? I:X '{ I:X X? (' I:? I: X X I: X X X:l (' I:? I: X ? I: X X? (' I:? I: X X:l I: X X?

I:?

~ I:

I:

:l I:

:t I:

X X.

:l

I:X

:l

X. ~

:I

XX

I: X ? I: :l

XX

I: X ?

I: X X I: :l I:x.

~

...

~ ·00··.··:·.... _~.:.. _.:.. _.... _.:.. _.:

_.: •..

CX

.. ... .

.. ... ..

I:

:lit

•

It

It

X)

.x.

.. .. J:) .• ;

:t I:x. )

XXX I:~ XX ~ X (':1 ( X XX IX XX.l XXXXX :l I: :l I: :t I: :l I: :lit X X X X. X. I:x. X. J... J... J... I: :l I: :l ( J... X X. J... J... J... J... J... X X XX

t:~ ~~

•

~I I:~

2a o Fig. 7. 3a-c. Coverage-dependent STM images showing the reduction of the spacing between. "added rows" as the cqverage increases. The oxygen exposures are (a) 500 L (90' 90A), (b) 2000 L (140 X140A2) and (c) 4300 L (50 XSOA2) (7.6]

3ao and 2ao for the case of Fig7.3a, and 3ao and 2ao for the case of Fig. 7.3b). Based on these STM and LEED observations together with other experimental results, a model has been proposed for the oxygen behaviour on this silver surface, as shown in Fig.7.4. The bright lines observed in the STM images of Fig.7.3 are Ag-O-Ag linear chains running perpendicular to the Ag rows (Fig. 7.4); "added rows" similar to the case of the Cu (110) surface. The "added-row" model has been confirmed by comparing the STM/ STS results with a first-principle theoretical calculation ofy M. Tsukada's group on the Ag (110)-2 x 1-0 surface having "added-row" atomic configuration. Figures 7.5a,b display the bias dependence of simulated images for the 2 x 1 phase "added-row" configuration. In the empty-state image (V := + 1.5V) Ag atoms are imaged and in the filled-state image (V := -1.5V) oxygen atoms are imaged. Figures 7.5c,d show the empty- and filled-state STM image, respectively, observed from the same surface area. The experimental values for the corrugation height (Figs. 7.5c, d and Figs. 7.5e, f, respectively) reveal a good qualitative agreement with the theoretical calculation. A critical terrace width of 100 A was found on the vicinal Ag (110) surface [7.7]. The oxygen causes spontaneous faceting with little nucleation barrier (it should be noted that nuclei growth is key to faceting).

212

3a o

4a o

(a)

X)

X) ~

I:

X)

X) X)

3xl (b)

Fig.7.4. Schematic of the model for (a) the "added rows" associated with the Ag-O-Ag linear chains and (b) the "extra lines" for a 2nd type of oxygen adsorption over the 3 Xl phase of the "added rows". Large open circles, large and small filled circles represent substrate Ag atoms, "added" Ag atoms and adsorbed oxygen atoms, respectively. Small shaded circles show the suggested molecular oxygen adsorption site at low temperature.

7.1.5 Ni(llO)-H and Ni(111)-H The chemisorption of hydrogen on transition-metal surfaces is of particular interest due to its role in catalysis, and as a simple case for the general chemisorption phenomenon. Extensive studies of hydrogen adsorption on Ni (110) have produced conflicting results on the structure of the surface. Six different reconstructions of the Ni (1lO)-H system have been observed by low-energy electron diffraction and helium diffraction: including 2 Xl, 2 x6, c2 x4, c2 x6 and 1 x2 reconstructions below 220 K, and "streaky" 1 x 2 reconstruction at room temperature. However, a STM study [7.8] at room temperature produces the convincing result that more than 80 % of the surface is usually disordered, despite the appearance of the typical streaky 1 x 2 LEED pattern before the STM measurement. Patches of 2 x 1 or 1 x 2 structure are sometimes observed in the disordered areas; whereas 5 x 2 structure, formed by a combination of row pairing and missing [001] rows, are most frequently found in the ordered area. The STM images also show the 5 x 2-ordered domain with an average size of 100 nm along the [110] and 10 nm along the [001] directions. The small domain size along [001], 213

o

(b)

(a)

1.5

v

-1.5 V

Fig. 7.6. Hard-sphere model of part of the Ni (III)-H surface. The 2 X2-2H graphitic superstructure is represented by the black dots. The hydrogen atoms occupy inequivalent hollow sites: above a Ni atom of the second layer or above a hollow site of the second layer

7.1.6 Sulfur Adsorption

('}I~41 Fig.7.5. Theoretical simulation of the (a) empty and (b) filled state STM images of the Ag (110)-2 X 1-0 "added row" configuration and experimental results of (c) empty and (d) filled state STM images (100 XIOOA2). (e) and (f) are the cross section plots of (c) and (d), respectively

and frequent antiphase boundaries and dislocation explain the observed streaky I x2 LEED pattern. In the 5 x2 structure, all the top-layer Ni atoms are paired along the [001] direction by a 0.05 nm lateral displacement to form tetramers, and every fifth [001] row is missing. Although the position of hydrogen is not determined in this experiment, the asymmetric measured heights of the Ni atoms suggest that H atoms may occupy some of the lowsymmetry bridge sites. In the 5 x 2 reconstruction of Ni (110) by adsorbed H at room temperature, all the first-layer Ni atoms are laterally displaced by 0.05 nm along the dimerizing direction, forming tetramers. In the Ni (111)-H system, however, micrographs of the (111) plane show a hexagonal pattern with a dimension of twice the unit structure, and a corrugation of 0.1 nm which increases towards a step. This suggests a p2 x 2-2H monolayer, induced by the high partial pressure of hydrogen [7.9]. Figure 7.6 represents a hardsphere model of part of the surface drawn according to the STM images.

214

It is known that Sulfur (S) forms ordered overlayers on many metal surfaces, often producing several distinct LEED patterns with increasing sulfur coverage. More than 100 LEED patterns have been reported for S chemisorbed on various crystal faces of at least 16 metals. The S-chemisorption system is of technical importance since sulfur is a common ingredient of lubricants. Investigating the structure of sulfied surfaces may help explain the fundamental mechanisms of friction and lubrication. Moreover, even more importantly, the interaction between sulfur and metal surfaces is also a classical problem in the field of heterogeneous catalysis, mainly because it plays an important role in poisoning metallic catalysts. Catalytic activities of most transition metals can drastically be reduced in the presence of S, which is a common impurity in many industrial processes. The reason is that the chemisorption bonding between S and the metal surfaces modifies the electronic or structural properties of the neighboring metallic atoms which are responsible for the adsorption of reactants. If the interaction between S and a metal surface is relatively weak, the structure of the substrate remains unchanged; however, it results in a perturbation all around the adsorption sites which, in turn, deactivates the surface. If the interaction is strong enough to significantly modify the metal-metal bonding, surface reconstruction may occur and produce new superficial structures which are inactive in chemical-reaction processes. In an attempt to understand the poisoning mechanism of metallic catalysts deactivated by S adsorption, extensive studies have been done essentially on copper, nickel, palladium, rhodium and platinum. The S adsorption, on the other hand, is able to modify the selectivity in some catalytic processes; i.e., a S-adsorbed metal catalyst is preferentially active for a particular reaction process. In this respect, a reaction via a Sadsorbed catalyst can mainly produce one pre-selected compound with fewer by-products, and subsequently leads to a higher commercial interest 215

since it enhances the efficiency in related industrial processes. Though a large number of efforts have been undertaken, the mechanism of selectivity modification caused by sulfur adsorption is still far from being understood. There are different models to explain modifications of catalytic selectivity, emphasizing geometry, ligand effect or restructuring. In order to ascertain the fundamental behavior of these S-induced properties, it is essential to characterize the structure and bonding between sulfur and the substrates. Generally speaking, surface structures produced by adsorption are classified into two categories. As mentioned above, when the interaction between adsorbates and a substrate is relatively weak a simple adsorbate layer is produced on the unreconstructed substrate. However, when adsorbates such as hydrogen, carbon, oxygen, sulfur, potassium chemisorb on a metal surface, the surface undergoes reconstructing due to a strong interaction between the adsorbate and the substrate. As a consequence, the atomic structure of the substrate is strongly modified to adopt to the adsorbate-induced chemical environment. Such a rearrangement of atomic positions can result in a slight distortion or displacement of the substrate without invoking significant mass transport (weak reconstruction), or the substrate structure can dramatically change, involving a long-range or local mass transport (strong reconstruction). In the latter case, some energy is needed to break the metal-metal bonding and migrate the metal atoms on the surfaces. This energy requirement can ultimately be overcompensated by the formation of a new adsorbate-metal bonding, and the net lowering of surface energy is the driving force of surface reconstructing. In fact, thermal energy is needed to activate surface reconstructions which are kinetically limited at room temperature; hence those reconstructions occur only at elevated temperature. Often an adsorbate-induced reconstructed structure can exhibit features significantly different from the clean surface, due to a severe change of the substrate.

7.1.7 Cu(lll)-S Different models have been proposed for the adsorption of S on the Cu(111) surface. Among them only the eV7XV7)RI9° structure attracted much interest and was studied by LEED, Surface-Extended X-ray Adsorption Fine Structure (SEXAFS), and X-ray standing-waves technique. In all those models the S coverage was referred from a radioactive-trace measurement, but the Cu coverage was subjectively assumed to fit with the individual proposals under consideration. Moreover, the source of Cu for the formation of the adsorbate-metal layer remained unknown, since only information about the static structure was provided by these techniques.

216

(e)

\~~~ a=9zA b= 11.7A

A

Fig. 7.7. (a) STM image of the zigZ4g struClure (75 X75 z), lhe unil cell is ou11 ined in a zoomed image (b) in a scale of 30 >800 K results in the growth of graphite islands on the Pt (111) surface. The annealed graphite islands exhibit several superstructures with lattice parameters of up to 2.2 nm, which can be explained nicely by the higher-rder commensurability of the graphite and platinum substrate at different relative rotations. STM has been used to distinguish a series of related molecules on Pt (111), including naphthalene, its isomer azulene, and l-methylazulene, 2-methylazulene, 6-methylazulene, 4,8-dimethylazhulene and 4,6,8-trimethylazulene. Typical lower-resolution images are used to measure the relative sticking coefficients and relative diffusion rates of the molecules. The orientations and binding sites of several of the molecules on the surface are also assigned. Tip dependent higher resolution images can show internal structural details on molecular species which have low diffusion rates [7.77].

7.3.4 Physisorbed Long-Chain Molecules As we know, STM allows high-resolution imaging not only in UHV but also at the internal interface between two condensed media, one being a conducting solid and the other a gas, a liquid or a soft solid. Under these con~ 246

ditions various molecules physisorbed to inert substrates have been investigated. The first unambiguous report on high-resolution molecular imaging at the solid-fluid interface was on two liquid crystals, 4-n-octyl-4'-cyanobiphenyl and 4-(trans-4n-pentylcyclohexyl) benzonitrile on HOPG [7.78]. The STM images reveal that the molecular order at the interface is increased over the bulk, and that the adsorbate lattice is oriented relatively to the substrate. Subsequently other liquid crystallHOPG interfaces [7.79] as well as a homologue series of cyanobiphenyls [7.80,81] have been imaged. The latter were prepared as thin films, whose exact thickness is not relevant to STM imaging, since the tip advances in any case all the way through the insulating organic layer. Different packings of molecules on different substrates [7.82,83], and a bias-dependent rearrangement within a molecular layer have also been observed [7.84]. Interesting is the fact that the first monolayer at the interface with graphite could be imaged even if the film forming material was in a crystalline phase. Besides liquid crystalline phases, long-chain alkyl derivatives can also form highly ordered monolayer at the interface between organic solutions and HOPG substrate. It should be emphasized that the interaction energy between a small molecule and a chemically inert substrate is usually too small to immobilize an individual molecule sufficiently at room temperature, therefore, alkyl-derivatization turned out to be a method to immobilize small molecules within a monomolecular layer on graphite and image them by STM. The first image of such a layer was obtained on dotriacontane (C 32 H66 ) [7.85]. Subsequently, similar two-dimensional molecular patterns were obtained on many other alkanes with chain lengths ranging from nonadecane (C I9 H 40 ) through pentacontane (C 50 H I02 ); alcohols such as octadecanol (C 18 H 37 OH), tetracosanol (C24 H 49 OH), and triacontanol (C 30 . H61 XOH); fatty acids including stearic acid (C n H 35 COOH), arachidic acid (C 19 H 39 COOH), and tetracosanoic acid (C 23 H47 COOH) and didodecylbenzene [H 25 C l2 (C 6 H 4 )C 12 H 25 ] [7.86,87]. In-situ STM studies reveal that all of these molecules organize in lamellae with the extended alkyl chains oriented parallel to a lattice axis within the basal plane of graphite. The planes of the carbon skeletons, however, can be oriented either predominantly perpendicularly to or predominantly parallel to the substrate surface, causing the lamella lattice to be either in or near registry with the substrate (alkanes and alcohols) or not in registry (fatty acids and dialkylbenzenes). In Fig. 7.24 the structures of alkanes, alcohols, fatty acids and a dialkylbenzene are compared. Note the superstructure along the lamellae of the flatly lying fatty acid molecules, which does not exist in the alkane lamellae. Since the images contain information on both adsorbate and substrate lattices they are attributed to an incommensurate and commensurate adsorbate, respectively, settling a scientific issue of long standing. In the case of the alcohols and the dialkylbenzene the molecular axes are tilted by +30 or _30 with res0

0

247

{a~;I..

1/1_

... /1:.,. ~ t t

~~;!!'·:·;t~!l%! §~-t4:4:~;~8

"'~ • •: . .... • .. : . •• •

'~

tiifjf~~~J:'.~. ",........ .. ~" 4J ........... ·t' ;:-··:."

" , '.

."

-

4

I

.".

:

•

J1

Z

; l,. .. • .. :t"t~: __ .I":~ . . , . . . . . . . . . . #'.-:./ t":.~! i ft, ~ .... ! r .. ... tr ..r •J•./ t

',' .

i. "/,, __

! I· "'.

"'=1-'-~~~/.-:: f ;

~: ': ! /t:;.t-;. -.. ._~ ~ ·f~,,~

Alkane

..,

Fatty acid

20m

.. .

'.• '. : -.J

f----l

Fig. 1.25. The frontier molecular orbitals of MDW74. (A) LUMO, (B) HOMO and superposition of the lowest four unoccupied orbitals (C). the HOMO and LUMO(D). the highest four occupied orbitals (E) [7.92] c J.

)

; ,J,

E

~l

r

..l

.--'

20m

,,-

.\~ ., . . \, .'-," ....' (.e......"";'" " ); \. - .•" .~ , ·:;.··.·~~~·~.i~:·.~ ,

,. >:."... "'. . "JI .. , :-: ..-. --:;~: .. j":.' ~- .., 1~Il'" ••,. 1".:. . . ,

•. '.:-

• ....."'....4

'1 ••• ' • • • • • • ' ·/('-·.-·---~~40

.

: l' : ;

~'.

....

, I·

'~''''''::'''''. ,.:

-

'~ • f

~".

f".-c

.

.. .. ,.. J 't'l

p'" ••'

"'., ••

~: : ~ ~ ~ . '~~ ~ ..... ~~...::..• •'Il.s ~ •

,.- ...../ .....~,.: t ~ ~ : : ~ . , ,, ..,.I· .' ..' ..... ,- "" .. ,,- .... ~

.

,.'

• -'./..

Alcohol

o'~' ~~

L

'.., -".• ~~

It. • "J'.w .. ).,

" : - '/

"

• 0"

6

...A'. •

1

....

~

Dialkylbenzene

~

Fig. 1. 24a-d. STM images of molecular monolayers at the interface between an organic solution and the basal plane of graphite [1.861. (a) Heptacosane (C n H 56 ); (b) octadecanoic (c) octadecanol (C 1S H 37 0H); (d) didodecylbenzene acid (C 17 H 35 COOH); [H25 C 12 (C 6 H4 )C 12 H 25 ]

pect to an axis normal to the lamella boundaries, giving rise to molecularly well-defined domain boundaries. It was found by in situ AFM studies that the adsorption of anions of the organic acid onto cationic surface is jointly affected by electro-neutrality and steric interactions, and that the adsorption could be reversible [7.88]. STM images of organic molecules are usually interpreted as time averages over their molecular dynamics, which typically occur on the picosecond time scale. However, there are also slower molecularly defined processes occurring on a time scale longer than milliseconds, which can be directly followed by STM. Fast STM image recording allowed the spontaneous switch between the two tilt angles to be observed in the alcohol mono layers on a time scale of a few milliseconds. Similarly, the movement of a domain boundary has also been observed in the didodecylbenzene monolayers, which can be explained by the quasi-simultaneous rotation of at least 248

about a dozen didodecylbenzene molecules around their long axis by 180 0 , together with a lateral shift by half a molecular length [7.89]. The imaging mechanism for adsorbed organic molecules has been a very interesting topic, as it not only involves the sample properties but also the possible interactions between tip and sample, sample-substrate. So far, several proposed mechanisms gave reasonable accounts for the data from different systems. The polarizability of the adsorbed Liquid Crystal (LC) molecules is seen likely to modify the local tunneling barrier. The tunneling current will thus not only reflect the LDOS distributions, but also the properties of the adsorbed dipole layer [7.90]. An attractive interpretation of STM images of LC molecules would be the correlation of tunneling image to the orbitals of molecules. It has been demonstrated for nCB molecules that the electric field under the tip apex is sufficient to induce significant shift of energy levels [7.91]. Furthermore, the STM images may involve several adjacent orbitals close to the Fermi level (Figs. 7.25,26). It could be perceived that validity of the models depends on whether there are orbitals available for tunneling electrons or not. If there exist orbitals involved in the resonant tunneling then the later model applies, otherwise the polarizability associated barrier modification should be the dominant machanism.

7.3.5 Chemisorption of Long-Chain Molecules An important category of chemisorbed molecules are self-assembled species on the surface of noble metals. Self-assembled molecules have been widely perceived as a venue to obtain novel material properties at nanometer scale. The application of STM 249

Fig. 7.26 Molecular mechanics modeling and molecular orbitals in superposition, along with a STM image of MDW74. Molecular model of MDW74 on graphite, with lowest four unoccupied orbitals

~ .

I.

n~ [Il

j

~=

~

~~~

I

i

-=-~

~=

I~~~~~~~~

I

.

roo .". ,

tr: '

m~~~\\\\\\\\\\~~~

~~.' .~~., ~.~:i~V l~l'r"" " , , ' , . "'.

... '. tjft:"f~

, ,\

~

'J

c:

lJ.J

Fig. 7. 27a-f. Constant-',,:\\' \;\.~\ .". . \ .. ', h\\\'\ •

•

\1. ('..

, &

_ .

.', • . \ : ','}.;,

"

i

:'"

't"'. •

.....

'

.

."

.

i

, , \

. .

'. '.'' " ........ ...."*' ...• \fI.... •'

' '. '. ~ '. ". • •i ....... .. ;: .. .:, '. . '. '. " '~""""" '. ~

....

Fig. 7.38. A series images from an STM movie showing the structural change of a reaction between H 2 S and the 0 preadsorbed Cu (l1O)-p2 Xl recorded at H 2 S expsoure 01'3 L (a) 5 L. (b)7 L (c) and 9 L (d) [7. 137]

7.6.2 Reaction on Cu (110) Figure 7.38a-d represent a series of STM images recorded during the surface reaction between H2 S and the 0 preadsorbed Cu (110) [7.137]. A c2 X 2 structure is initially formed on fringes of the gaps (defects) between the p2 X 1-0 islands (Fig. 7.38a). Dynamic observations (Fig. 7 .38b-d) show that the reaction gradually spread to the surroundings with increasing H2 S exposure. The fact indicates that the reaction occurs preferentially on the edges of the p2 X 1-0 domains. As a result of the reaction, small islands appear and develop in size rather than quantity. As mentioned above, the p2 X 1-0 structure is stabilized by forming -0Cu- bonds in the rows along the [001] direction, and the rows are separated by a double lattice parameter along the [101] direction. When H2 S species access the adsorbed 0, the two react to produce water and atomically adsorbed S. While water desorbs from the surface at RT, the metal atoms (0.5ML) can be released from the row structure due to a loss of 0, then diffuse and accumulate to form 1 X l-Cu, on which the S adatoms form the c2 X 2-S structure. Unlike in the Ni (110)-0-S case, whereas a local diffusion 268

occurs to form many small islands and identical patches with the c2 X 2-S structure, the reaction CU(110) produces the c2x2-S islands and patches in large sizes. This different behavior could be ascribed to the considerably large difference of self-diffusion rate between the Cu(110) and the Ni (110). It has been also demonstrated that formaldehyde could be synthesised from methnol on oxygen treated CU(110) surface upon desorption [7.138]. On Cu (110), the high mobility of the Cu atoms at RT is responsible for the appearance of large islands but less in quantity. To the end of the reaction a simple S overlayer forms, and the surface morphology is characterized by the formation of large islands [7.22]. Surface reactions produce not only chemical replacements, but also structural transformation. Direct observations of the resultant structural transformation resulting provide an understanding of the reaction mechanisms from the structural aspect. At atomic scale, the S replaces the adsorbed o and results in c2 X 2-S structures. At a relative macroscale for surface morphology, the Cu atoms released from the p2 X 1 structure accumulate into islands due to the high Cu diffusion rate on the surface. The Ni atoms can form small islands stabilized by adsorbed S atoms at room temperature owing to a low mobility as well as the adsorbed S atoms. Activated by 0 adsorption, the Ni atoms bonded with adsorbed S atoms to form a metastable structure of small islands ascribed to the formation of the reconstructed p4 x l-S structure which otherwise would be obtained by exposing a clean Ni(llO) at elevated temperature. The results suggest how it is possible at the atomic level to activate a surface and thereby open reaction pathways. In this way, the results may have important consequences for the field of surface reactivity in general, and heterogeneous catalysis in particular.

7.6.3 Chemical Identity with STM Although STM. as mentioned in Chaps. 1 and 3, is able to reveal new unique and unprecedented information, extraction of topographic information from STM images is not always straightforward and simple, and it is not a priori possible with the STM to discriminate between the different chemical species on the surface. By reversibly manipulating the chemical identity of the apex of the tunneling tip, Ruan and Besenbacher et al. [7.139] were able to discriminate between the 0 and Ni atoms in low-coordinatred -Ni-Orows on Ni(llO) and, equivalently, between 0 and Cu atoms in -Cu-Orows on Cu (110). Figure 7.39 shows a 2 x 1-0 island on aNi (110)-1 x 1 surface imaged with a clean W tip. The protrusions of the (2 x 1)-0 phase are in line with the [110] rows of the clean surface (Fig.3.32a). Thus with a "clean W" tip, the 0 atoms are imaged. By letting tiny amounts of oxygen 269

7.7 Chemical Reaction on Semiconductors 7.7.1 Reaction of NH 3 with Si (111 )-7 x7 Surfaces

Fig. 7. 39a, b. STM topographs (37 X47 A2) of the added -Ni-O- [OOIJ rows on Ni obtained with a clean W tip (a) and after chern isorption of all 0 atom at the apex of the tip (b). For both images V = IO mV and I = InA. The grids indicate the positions of the Ni atoms in the I XI Ni layer underlying the added -Ni-O- rows. In (a) the protrusions of the added rows are in line with the close-packed rows of the underlying Ni-I XI surface. whereas in (b). they are out of "registry" [7. 139J

into the chamber (0.05 L) or by just scanning the 2 X I-a surface for a while, a sudden uncontrolled tip change may occur, and correlated with the tip change, the protrusions of the 2 X I structure change to be out of registry with the close-packed Ni [110] rows (Fig. 7.39b). R uan and Besenbacher et al. [7.139] suggested that the tip change is associated with the binding of an a atom at the apex of the tip. Thus, with an "0 tip", the Ni atoms are imaged as protrusions. By reversibly manipulating the apex of the tunneling tip to have either a "w atom" or an "0 atom" at the apex, the a and the Ni atoms, respectively, are imaged in the -Ni-O- rows, i.e., chemically different elements such as a and Ni can be discriminated with an STM. Identical conclusions are reached for the Cu (ll 0)-2 XI-a surface. A possible tentative explanation for these observations is to assume that tip-sample interactions may cause a "w tip" to form a stronger chemical bond with the a atoms along the rows than with the metal (Ni, Cu) atoms, whereas an "0 tip" may form a stgronger chemical bond to the Ni or Cu atoms than to the a atoms. By manipulating the chemical identity of the tip apex, the formation of a chemical bond between tip and sample may increase or decrease the density of states at the Fermi level E F since antibonding states, resulting from tip-surface atom interactions, are shifted up or pushed away from E F [7.126].

270

The usefulness of an STM in the study of chemical reactivity on semiconductor surfaces has been demonstrated in dissociative chemisorption of NH 3 on Si(lII)-7X7 [7.140] and on Si(001)-2x I surfaces [7.141]. This was achieved primarily by comparing STM images of clean, welkharacterized surfaces taken before and after dosing with reactive gas. Atomic-scale variations in surface chemical reactivity are observed in the interaction of NH 3 with Si (l11)-7 X 7. Pronounced differences have been found in the chemical reactivity at various sites within the ideal Si (Ill )-7 X7 unit cell. STM topographs allowed the course of the chemical reaction to be mapped on an atom-by-atom basis, and simultaneous tunneling spectroscopy measurements established the relation between local electronic structure and chemical reactivity. Figure 3. Ila exhibit an STM image of the unoccupied states of the clean Si (lll )-7 x7 surface in which the 12 adatoms in the 7 x7 surface unit cell are clearly seen. As we described in the last chapter, STM images of a clean Si(l1l)-7X7 surface clearly show two inequivalent types of surface Si adatoms, called "corner" adatoms and "center" adatoms. Upon exposure to about 1 L of NH 3 , reaction takes place and about half of the adatoms disappear from the images as their dangling bonds are passivated by reaction with NH 2 or H producing surface Si-H and Si-NH 2 groups (Fig.3.11b). Even from a simple inspection of the STM images, important chemical information can be obtained. For example, there are roughly 4 times as many unreacted corner adatoms than center adatoms in Fig.3.llb. The fact that center adatoms are more reactive than corner adatoms would be difficult to determine by conventional surface-science techniques. At a sample bias of +3 V, there is a finite contribution to the local density of states by reaction product states. As a result, product sites can be imaged. The preservation of the 7 X7 reconstruction suggests that the reaction with NH 3 has primarily saturated existing dangling bonds with limited, if any, Si-Si bond breaking. In order to obtain insight into the reasons for the above behavior and to answer questions such as the chemical identity of the products, tunneling' spectroscopy has been used to study the electronic spectra of the clean and NH 3 -exposed Si (l11)-7 X 7 surfaces. The atom-resolved tunneling spectra shown in Fig.3.ll were recorded for the positions indicated by the arrows. The curves A, Band C in Fig. 3 .lla give the spectra over rest-atom, corneradatom and center-adatom sites on the clean surface, respectively. Negative energies correspond to occupied states and positive energies to unoccupied states. The rest-atom spectrum (A) shows a strong occupied states peak at about 0.8 eV below E F . This peak is characteristic of the rest-atom's dangl271

ing bond. The corresponding dangling-bond states of the adatoms appear near 0.5 eV above E F (B and C). However, the center-adatom spectrum (curve C) reveals important differences from that of the corner adatom (curve B). The intensity of the occupied dangling-bond state has decreased while, correspondingly, the intensity of the unoccupied state has increased. The filling of the rest-atom's dangling-bond state and the small occupation of the adatom state suggests an adatom to rest-atom charge-transfer process. In addition, the spectra (curves Band C) indicate that most of this charge is contributed by the center adatoms. This is because center adatoms have two rest-atom neighbors while corner adatoms have only one. The above differences in the occupation of the dangling bond states could be the cause of the observed differences in the reactivity of center and corner adatoms. Figure 3. 11 b (bottom) depicts spectra obtained on a partially reacted surface. The 0.8 eV characteristic state has been eliminated in the spectrum A for a rest-atom site. Adatom spectra (curve B, dashed line) show that the corresponding surface states are also eliminated upon reaction. The elimination of the surface states is the reason for the "disappearance" of the reacted adatoms in Fig.3.11b. More systematic studies which involved spectral maps of large areas of the partially reacted surface demonstrate that rest atoms are more reactive than adatoms, reacting faster than one would predict on the basis of their relative numbers on the 7x7 surface. For example, upon NH 3 exposure, rest atoms react first, and under conditions such as those of Fig.3.11b where about half of the adatoms are still unreacted, but no unreacted rest atoms remain. The spectra B (solid line) and C of unreacted corner and center adatoms on the partially reacted surface (Fig. 3 .11 b) illustrate the effects of the reaction on the electronic structure of still unreacted sites. These two curves reveal that, after the neighboring rest atoms have reacted, the electronic spectra of unreacted corner and center adatoms become virtually indistinguishable. The differences between the two sets of adatomic spectra can be explained in terms of the effects of the reaction on the charge-transfer interactions present on the clean 7 x 7 surface. On the clean surface the center-adatom dangling-bond state is nearly empty (Fig.3.11a, curve C), but on the surface where the rest atoms have reacted (Fig.3.11b, curve C) it shows a much higher occupation with a simultaneous decrease in the intensity of the unoccupied part at +0.5 eV. The above results suggest that during reaction a rest atom to adatom reverse charge transfer takes place which allows the rest atom to transfer the extra charge and react with NH 3 . STM can be utilized to investigate the effect of local structure on the chemical reactivity of the various dangling-bond sites of the 7 x 7 surface. In the DAS model, rest-atom sites involve triply coordinated surface Si atoms. Theoretical calculations find a normal dangling-bond character at these sites and a dihedral angle between bonds close to tetrahedral. Reac. 272

tion at these sites should not produce surface strain. The situation is different at adatom sites which involve considerable strain. Adatoms in this model are members of three four-membered Si rings. These ring structures bring the adatoms close to the Si atoms directly below them in the third atomic layer. This proximity leads to repulsion, and distortion of the structure with adatom dihedral bond angles close to 90°. Another important difference between the two kinds of sites is the occupancy of the respective dangling-bond states. The spectra in Fig. 3.11 a indicate that the rest-atom dangling bonds are fully occupied while the adatom dangling bonds are less than half occupied. The reduced density brought about by the delocalization of the dangling-bond charge at adatom sites and the large deviation from tetrahedral geometry are responsible for the reduced reactivity of adatoms as compared to rest atoms.

7.7.2 Reaction of NH 3 with B/Si (111)-V3 xv3 Surface The other example which demonstrates the chemistry of the Si surface and depends very strongly on the local structure and strain considerations at the reactive site, is the effect of boron on surface chemistry. As we saw earlier, the equilibrium configuration of the B/Si (111)-V3 xV3 surface involves a Si-adatom top layer, with the B dopants below the Si adatoms in substitutional sites. Because of the Si-to-B charge transfer, the Si top layer of the B/Si (l11)-V3 xv3 system has no occupied dangling-bond levels. This drastic change in the dangling-bond level occupancy should be reflected in the chemical properties of the surface. Indeed, it is experimentally found that the top Si layer of this B/Si (111)-V3 xv3 surface has chemical properties very different to those of clean Si surfaces. For example, in contrast to the behavior described above for the reaction of a clean Si (111)-7 x 7 surface with NH 3 , exposures of the B/Si(111)-V3XV3 surface to even a few hundred Langmuirs of NH 3 at room temperature leads to very little reaction [7.51]. This indicates that boron incorporation has a drastic influence on the reactivity of Si adatoms. However, not only is the reaction rate affected by B-doping but the nature of the reaction itself is different: NH 3 adsorbs reversibly on the B-modified v3 x V3 surface by donating N lone-pair electrons to the empty Si dangling-bond state of the adatoms, that is, by a Lewis acid-base reaction. This is in contrast to the dissociative adsorption of NH 3 observed at the adatom sites of the clean 7 x 7 surface. It is a novel shortrange doping effect on chemical reactivity which involves direct chargetransfer interaction between the dopant atom and the surface active (i.e., dangling-bond) site.

273

7.7.3 Reaction of NH 3 with Clean Si (00 1) Surface The chemisorption-induced changes in surface chemical bonding have also been observed, using the interaction of NH 3 with Si(OOI) as a prototypical gas-surface reaction system. The dissociative adsorption of NH 3 on Si(OOI) produces hydrogen atoms which change the local bonding of the Si (001) dimers. These changes allow reacted and unreacted Si (00 I) dimers to be distinguished (Fig. 7 .40), and tunneling spectroscopy is utilized to elucidate the detailed nature of the bonding before and after the reaction. Since the atomic positions for Si remain virtually unchanged by H adsorption, these changes can be directly attributed to the different spatial distributions of SiSi and Si-H bonding states. The STM results on the reacted surface are interpreted in terms of tunneling through localized Si-H bonding orbitals of a Si (00 1)-2 x I monohydride.

A

A

,

7.7.4 Si(I11)-7x7 Oxidation The nature of the initial stages of the oxidation of silicon is a long-standing problem. A large number of studies on this issue has been carried out utilizing a great variety of different techniques. Several configurations have been proposed for the oxygen-containing sites in the early stages of the reaction. They involve oxygen atoms saturating the dangling bonds of top-layer Si atoms, oxygen atoms inserted in back bonds but leaving the dangling bonds intact, or molecular forms of oxygen attached to surface atoms or bridging two surface Si atoms. Avouris et al. [7.142] found that two different oxygen-containing structures are formed in the early stages of Si (111)-7 x 7 oxidation. These structures appear as dark and bright sites, respectively, in STM topographs of the unoccupied states of the sample. At very low exposures «0.2L), the numbers of dark and bright sites are comparable but as the exposure to 02 or N2 increases, the number of dark sites increases steadily, leading eventually to rough surfaces with no obvious order. The number of bright sites does not increase nearly as fast, suggesting that the dark sites are more representative of the main oxidation process. The dark sites remain dark for both positive and negative sample bias voltage, indicating that the adatom dangling bond is saturated by a group that does not have any low-lying unoccupied or occupied states. The different surface site distributions and different growth rates observed with increasing 02 -exposure show them to be two distinct early products. By correlating the spectroscopic results and the results of theoretical calculations one is able to identify the dark sites as adatom sites whose dangling bond has been saturated by an oxygen atom while another oxygen atom has been inserted on one of its back bonds. The bright sites represent adatoms with an intact but emp-

°

274

Fig.7.40a-d. STM topographic images probing the (a) occupied (-2V bias) and (b) unoccupied (+ 1.2 V bias) surface states of clean Si (001) and the (c), (d) occupied (-2 V sample bias) states ofNH 3 -dosed Si (001). In each figure, the arrows point in the (110) direction and point directly at the center of the dimer rows. Each box out I ines a 2 XI unit cell with a dimer centered in the box (a) contains a single "missing-dimer defect". In (d), the position of a single atomic step is indicated by the solid line, and the upper and lower terraces are identified by "U" and "L" [7.141)

275

tied dangling-bond state which is shifted to higher energy and contains an oxygen atom inserted in one of the adatom back bonds. In general, new dark sites tend to nucleate at old dark sites, indicating an electronic-structure perturbation by the oxidation products. The preference of these products for the faulted half of the 7 x 7 unit cell and for corner-adatom sites can be explained in terms of a site-dependent sticking coefficient involving a process analogous to the gas-phase "harpooning" processes. It should be pointed out that the majority of the resulting molecular precursors involve 02 interacting with a single dangling-bond site. The adsorption of 02 on Si (Ill) produces many molecular sites which are highly reactive [7.143].

Hydrogen serves to relax the strained surface bonds primarily by allowing the formation of isolated trihydride species and adatom islands. While reaction with the bulk is thermodynamically feasible, the barrier is too large for the reaction to proceed appreciably under the present conditions. The manner in which this relaxation occurs determines both the nature and the population of different hydrides present on the surface. Hydrogen desorption involves the restructuring and ultimately the breakup of these island and leads to be re-establishment of a 7 x 7 adatom surface. Boland suggested models that fail to account for the presence of these strained bonds due to an inadequate description of the surface chemistry [7.146].

7.7.5 Si(l00)-2x1 Oxidation

7.7.7 Reaction of Sb4 with Si (100)

In contrast to the Si (Ill) -7 x 7 surface which has a metallic Density Of States (DOS), the dangling bonds of the Si dimers on the 2 X 1 surface are paired, leading to the formation of a surface gap and a vanishing DOS near E F . A vouris and Cahill [7.144] studied the initial stages of oxidation of the Si (100)-2 x 1 surface by STM, and observed a reduced DOS near E F . While on the Si(100)-7X7 surface the most reactive sites are the top-layer Si adatom sites, on the Si (111)-2 x 1 surface the majority of the dimer sites are not very reactive compared to defect sites, particularly the C-defect sites. Defects on the Si (100)-2 x 1 surface which have a metallic DOS dominate the reactivity towards 02 in the early stages of the reaction. Among the new sites generated by the exposure to 02 are 1.4 A high bumps on top of the surface. Upon annealing of the 02 -exposed surface or upon 02 exposure at an elevated temperature these bumps form elongated islands. Evidence is presented suggesting that the bumps and islands are probably due to silicon ejected to the surface by the oxidation reaction. The nucleation of oxide clusters are essential in forming multilevel Si islands [7.145].

7.7.6 Reaction of H with Si (111)-7 X7 Little is known about the factors which control the reactivity of hydrogen on the Si(l11)-7x7 surface although the chemistry of hydrogen on this surface has been extensively studied over the years. A combined STM and STS study [7.146] has shown that the chemistry of H on the Si (111)-7 x7 surface is driven by the relaxation of the strained bonds formed by the reconstruction. Such bonds have a reduced activation barrier, and reaction continues until the supply of these weak bonds is exhausted. The limited number of such bonds explains the apparent saturation of hydrogen on this surface. 276

The reaction of Sb 4 with Si(lOO) has been studied with STM. Five distinct types of Sb clusters are observed on Si (l00), each containing four Sb atoms. The final-state cluster consists of two dimers with the dimer bonds perpendicular to the Si dimer bonds in the substrate, consistent with past studies of the equilibrium structure of the Sb layer on Si (100) formed at elevated temperatures. The remaining four types are precursors that form the pathway for the Sb 4 molecules to reach the final state of dissociative chemisorption. These precursor clusters can be converted to the final state of chemisorption either by thermal annealing or through an STM-tip-induced conversion process. Using a new STM method of image-anneal-image cycles, i.e., imaging the same Sb clusters before and after a cycle of thermal annealing at a certain temperature, Mo [7.147] determined the reaction path of the dissociative chemisorption of Sb 4 on Si(lOO). Combined with the measurements of the average population distribution of Sb clusters as a function of thermal treatment, the energy barriers and the prefactors for conversions between different states are obtained. From the conversion rate of precursors to the final state vs temperature, an effective activation energy of 0.5 ±0.1 eV and an effective prefactor of about 10 3 Hz were deduced [7.148,149]. These precursors are found to have no thermal mobility before dissociation, contrary to the popular notion about precursor states. Such results indicate that the reaction of molecules with solid surfaces can be extremely complex. Studying the reaction path of molecules on surfaces with the STM has unique advantages. First, it allows direct identification of different types of coexisting precursors. Second, direct measurements of the conversion rates are less model dependent than the conventional molecular-beam method. The new STM method of image-anneal-image cycles also allows direct observation of the diffusion process while avoiding potential STM tip effects. The method has been used to study the anisotropic diffusion of Sb 277

dimers on Si(100) [7.147]. The energy barrier and the prefactor for the faster diffusion across the substrate-dimer rows are measured. This method offers the same advantages as the FIM method, yet it allows studies of surface diffusion on a much broader range of substrate materials. It is obvious from the above examples that the demonstrated capability of STM to probe the topography and electronic structure of surfaces and adsorbate layers with atomic resolution makes it a powerful tool in the study of surface chemistry, providing unique insight into the mechanisms of surface reactions at the atomic scale.

8. Biological Applications

Shortly after the STM was invented by G. Binnig and H. Rohrer, the application of this new technique to biomolecules was initiated and followed by rapid growth. Many meaningful results have been achieved in the structural investigations of nucleic acids, proteins, biological membranes and supermolecular biosystems. This has demonstrated the great potential of the newly developed technique in the studies of surface structures of biological materials.

8.1 Advantages and Problems As mentioned in Chap. 1, very high resolution can be obtained with STM. Although higher resolution can be achieved by TEM and FIM in the lateral directions, they are limited by the fact that a coated conductive layer on the surface of the sample is required in SEM imaging [8.1]. TEM is only suitable to the study of bulk and interface structures of very thin samples [8.2]. Only the two-dimensional structure of the sample atoms on tips with a diameter of less than 100 nm can be detected by FIM. Additionally, these three kinds of microscopes must be operated in a vacuum environment. Atomically resolved structural information can also be obtained with several other diffraction techniques such as X-ray diffraction, He diffraction and low-energy electron diffraction, but these techniques require crystalline samples, and only provide averaged information rather than local structural information in real space [8.3]. Unlike electron microscopy and diffraction techniques, requiring a vacuum environment and crystal samples, respectively, STM images of the surface structure of crystalline or non-crystalline samples with resolutions ranging from Angstroms to nanometers, can be operated not only in the vacuum, but also in air, at low temperature, at ambient pressure and temperature, and even in solution. Biological samples die in such sterile environments and their structures may be far different from that in the active state. Therefore, to observe directly the structure of biological samples under natural or nearly natural conditions (at ambient pressure and tempera278

279

ture, or in acqueous solutions) remains a dream for biologists. Such a possibility generated the wide interest of biologists in STM. Although these unique advantages imply hopeful prospects for the application of STM to life science, some problems still exist which can be listed as follows:

pies to substrates with strongly adsorbent groups. The second is to reduce the interaction between the sample and the tip, including choosing a smaller reference current to increase the gap between the sample and the tip, or using a hopping technique (which will be described in Sect. 8.2). Salmeron et al. [8.9] argued that the main contribution of coating the sample with a conductive film for STM imaging comes from the fixation rather than the conductivity.

8.1.1 Substrates Highly-Oriented Pyrolitic Graphite (HOPG) had been a convenient substrate commonly used in STM studies on biological molecules before 1991: DNA, for instance. However, some reports [8.4] indicate that features similar to that of DNA and other biomaterials are present in STM images of freshly cleaved HOPG surface onto which no biological samples were deposited. It has been suggested that gold and platinum are also not suitable substrates for use with these materials, no matter how the substrate surface is prepared [8.5]. Unfortunately, choices among other materials which can be employed as substrates are very limited except for HOPG and gold. To improve the applicability of STM to biological materials, one must either find suitable substrates or refine the available methods to distinguish the artifacts of the substrates from the biological molecules themselves. Crystalline gold, which is prepared easily by evaporating gold onto mica heated to a temperature between 600 and 800 K [8.6], is becoming increasingly popular. For AFM study, mica is a convenient substrate commonly used. STM is based on the flow of an electrical current and thus cannot be used to directly image insulating material. It has been speculated theoretically [8.7] and confirmed experimentally [8.8], however, that a very thin film of water (about one monolayer) adsorbed to a surface exhibits a surprisingly high conductivity that is sufficient to allow STM imaging hydrophilic insulators. Biological specimens, such as DNA on mica, have been imaged in humid air by an STM with high resolution recently [8.8].

8.1.2 Fixation of Samples onto Substrates Usually, biological samples must be dispersed on flat and conductive substrates before imaging by STM. Adsorption may be so weak that the interaction between the tip and the sample at such a small distance can cause the movement of the sample on the surface of the substrate with the tip scanning across the surface, thus stable and high-resolution images of adsorbent cannot be easily achieved. Two methods might be used to address this problem. The first is to enhance the adsorption between samples and substrates, including coating samples with conductive films or covalently linking sam280

8.1.3 Flexibility of Biological Samples Most biological samples are not rigid structures and are flexible to various extents. The long chains of peptides, lipids and carbohydrates can be turned or folded. Even DNA, with a structure which is usually considered stable, can form various conformations under different conditions and undergoes structural variances such as twisting and unwinding. During tip scanning across the sample, the tip-sample interaction and the thermal vibration of the sample may lead to changes in the sample surfaces, resulting in a distorted image of the natural structure of samples on the one hand, and low resolution on the other. This problem might be solved by reducing the tipsample interaction. Although placing samples at low temperatures can reduce thermal vibrations; a low temperature is far from the native condition and has some effect on the structures of samples and, moreover, increases the difficulties associated with manipulation.

8.1.4 Identification and Interpretation of STM Images No systematic theories or a simple way for identifying and interpreting the images obtained have been established. It has proved to be a more difficult task to identify and interpret the images of more complicated and irregular biosystems than the images of simple crystal samples. Generally, biological samples have a low conductivity. Although low conductivity of biological materials is not an inpenetrable barrier for STM imaging, the contrast mechanisms for imaging samples with low conductivity are still not clearly understood. This limits image identification and interpretation.

281

8.2 Preparation 8.2.1 Dispersion of Samples on Substrates The simplest method of sample preparation is to deposit a droplet of dilute sample solution on the substrate so that the sample is dispersed and adsorbed naturally after being dried on the substrate. What is to be noted in using this method is that: firstly, the concentration of the sample solution cannot be too high, otherwise the thick deposition of the sample will prevent the STM from providing good images; meanwhile the concentration of salts in the solution cannot be too high, so as to avoid the effect of crystallization of salts from the solution. Usually, the concentration of the solution should be just as high as, or slightly lower than, what is needed to form a monolayer. Secondly, the dispersing abilities of the molecules under investigation should be taken into consideration. In order to promote homogeneous dispersion of sample molecules, a small amount of detergents (such as SDS) or other solvents (such as ethanol) which can reduce surface tension, are added into the solution. Tailored monolayers which either create favorable charge patterns at the interface or offer ligands interacting specifically with molecules in the subphase can serve as intermediate high-affinity substrates. Alternatively, electrochemical methods can be utilized for deposition by applying appropriate potentials between solution and substrate. Biomolecules usually remain attached to the substrate even if the potential is changed within certain limits. With electrochemical STM and AFM the process of adsorption may be monitored in situ (Chaps. 4 and 5). A FIM technique for sample preparation has been applied to depositing samples onto a gold ball. The device employed is shown in Fig. 8.1. A gold ball on which the sample is deposited is held in a threaded pin that is then inserted into the plexiglass cover. This pin is so positioned that the gold ball is inside a small tungsten coil. Deposition of the sample is started by placing the deposition device into the sample solution. Due to surface tension,

sample droplets are pulled up into the coil and in this way the gold sphere is immersed in the solution. After a while the sample will be adsorbed onto the gold sphere [8.10). The adsorbate concentration can be controlled by adjusting the time of adsorption and the concentration of the sample solution. The sample prepared by this method can be observed by STM immediately, or after being rinsed with alcohol or distilled water to remove the undesired salts, or even covered with a layer of conductive film.

8.2.2 Fixation of Samples a) Sample Coatings After the biological sample is dispersed on the substrate it can be coated with conductive films such as metal and carbon. This kind of treatment can provide not only good conductivity but also fixation and rigidness which favor stable STM imaging, but the resolution is limited because of the limitation of particle size of the coating films. For the purpose of reducing the particle size, low-temperature evaporation techniques can be employed to improve the resolution. The results which have been obtained, indicate that the sample-coating method is most suitable to topographical studies of biological samples on a large scale. The topograph of a recA-DNA complex and the whole cell sheath of bacteria, etc. have thus been obtained. To make sure in advance that the preparation of samples fulfills certain quality criteria such as good and homogeneous coverage of the support, imaging metal-coated samples alternatively by STM and TEM or SEM is a good way to examine the quality of the preparation. Such control experiments also help to unambiguously identify the structure under scrutiny. b) Covalently Binding Samples with Strongly Absorbent Groups p-toluenesulfonyl, for example, can serve as an absorbent group on the HOPG surface to link the sample to the substrate. While tris( l-aziridinyl) phosphine oxide (T APO) can be used to anchor DNA to a gold surface, because T APO contains an ethylene imine group which can readily be reacted with a pentose group of DNA on the one hand, it also has a phosphorus oxide group to provide linkage to gold on the other. Therefore, T APO can not only provide a linkage to the a gold surface, but also a conductive pathway. Some highly-resolved images of DNA have been obtained by means of this method (Sect.8.3.1).

Tungsten coil

Fig. 8.1. Devices employed in the FIM sample absorption method [8.10) 282

283

c) Binding Samples to the Substrate Covalently The Langmuir-Blodgett technique has been employed by STM researchers to link samples to substrates (Fig.8.2a). Heckl et al. [8.11] once fixed lipid molecules on the graphite surface for STM observation. By using this technique, individual lipid molecules can be observed. On the basis of the Langmuir-Blodgett technique, Lindsay et al. developed a new technique of anchoring DNA on the graphite surface [8.12]. Graphite is oxidized and then joined with -SH groups which can react with Hg, thus DNA modified by Hg can be linked to the graphite substrate covalently, as shown in Fig.8.2b.

(a)

~

~

SOCl'I -C-CI ~

-C-OH--

I,IO-Decanediol

•

HOCH ,(CH,J,CH,OH

1 ~

KMnO,

-C-0(CH2)9CH20H---

Pyridine

1

2

3

Oxidized graphite surface

o II

Ly,ocaprYIYI' 1 Iccllh,n/DCCI

0

II

-C-0(CHz)9CO~H

1

DMAP

•

0

0

II

II

I

II

I I

CHOH

8.2.3 STM Imaging in Acqueous Solutions

I

5

Z 6

Z

-C-0(CHZ)9C-0-OH

CH zOC(CH2)6 CH 3

4

CH OC(CH ) CH

3

0

II

CHrO-P-OCH2CHzN+Me3

I 0

0

II

CH z -0-P-OCH 2CH zN+ Me 3

I

Lindsay et al. first obtained STM images of DNA in acqueous solutions [8.13]. With regard to the special problems arising when STM is applied under water, they managed to use the following techniques: a tiny cell of ca. 50 ~I was employed to overcome the vaporization of the solution; the scanning probe was covered with glass, leaving only the very end of the tip naked so that the Faradaic current was reduced to about 0.1 nA and the detection of the tunneling current became possible. A tiny gold sphere was utilized as the substrate, and a reference electrode was employed to induce adsorption of DNA molecules to the substrate when a voltage was applied between the reference electrode and the gold sphere.

O' (b)

Step 1:

~

Cr03 -

+

ill ~ U-

C - OH

Step 2:

nC~ U-

OH

+

SOCI 2 - - -

n~ U-

C - C1

Step 3:

n U- ~

C C1 -

.: ~

+

~

:: -C-~-(CHzCH2S-)z-NHz .' H

(NH zCH 2CH 2SH)2 -

Step 4:

8.2.4 Hopping Technique

.:~

~

:: -C-N-(CHzCHzS-)z-NH z + 2-Mercaptoethanol -

Jericho et al. [8.14] developed a new kind of STM technique which was named the hopping technique after its operating principles: when the tip is scanning laterally, it is drawn back from and then reapproaches the sample in the z direction periodically just like "hopping". The height of each "hop" can be kept constant. If the tip "hops" high enough, the lateral force caused by the scanning tip will be reduced and the damage to the sample surface will also be reduced. From the operating principle of the hopping technique shown in Fig. 8.3, it is apparent that "hopping" is achieved by adding a digital modulator into the feedback system of a STM instrument.

-: ~

~

I

H

~

::

C

H

:. C

Ste~5:

II :: -C-~-CHz-CH2-SH+Hg-DNA

~:::: -C-N-CHz-CHz-SH ~

::

-

I

H

0

II

:: -C-NHCH 2CH 2S-Hg-DNA

.

Fig. 8.2a, b. Two methods of binding samples to the substrate covalently [8.11]. (a) Langmuir-Blodgett technique for fixing lipid molecules. (b) DNA fixation technique developed on the base of Langmuir-Blodgett technique [8. 12J

284 285

Pre amp

Tunnel current monitor

8.3 Nucleic Acids

(-I08Y/A)

1,_

CrlflF

IC 4 to STM

z axis

SIQ

I

ChlflF~

Digitalcontrol~-15Y input

R h IkQ

100Q

Fig. 8.3. The principle of the hopping technique [8. 14]

A

Icm

t------i

Fig. 8.4. STM designed to be used in combination with an optical microscope [8.15]. (A: approach screw, S: sample, T: tip, P: piezoelectric tube, and 0: optical microscope)

8.2.5 STM Directed by an Optical Microscope If enough room is left between the tip and the sample, it is possible to use an optical microscope with high magnification to bring the tip to the sample as closely as possible so that the position of the tip can be controlled and the location of the scanned area can be chosen purposely. Figure 8.4 presents a combined instrument consisting of an STM and an optical microscope.

286

Nucleic acids are classified into two major groups: DeoxyriboNucleic Acid (DNA) and RiboNucleic Acid (RNA), of which DNA is the main genetic substance in the life process, with the only exception being viruses where RNA is the main genetic substance. In the field of life science, the structure and functions of DNA are of central interests. Up to now, a lot of results about the structure of nucleic acids have been accumulated mainly from the sources of X-ray diffraction, NMR, rotatory dispersion, circular dichroism analysis and the analysis of the primary structure of nucleic acids, i.e. the sequence. The interest of many researchers is focused on the three-dimensional structures of nucleic acids at the native and active states, and the structural variances which nucleic acids undergo in the process of life, a key to the explanation of the nature of many life phenomena. The emergence of STM and AFM provides scientists with a possibility to observe directly DNA and RNA under natural or almost natural conditions. Another major goal of this research is to assist in sequencing the DNA of the human genome in order to bring people relief from some of the physical and mental disease that is of genetic origin.

8.3.1 DNA in Air and in Vacuum Right-handed DNA with pitches ranging between 2.5 and 3.5 nm and a mean width of 2.0 nm has been observed by STM [8.16] when DNA is fixed on a gold surface with T APO. Furthermore, in the image taken at greater magnification, the details of one helical pitch are shown. The two sections of approximately rectangular shape (2.0 x 1.5nm2 ) form an angle of 40° with molecular axes and clearly represent the shallow minor grooves of the helix, separated by an imperfectly imaged region which corresponds to the deep major groove. The poor imaging of the major groove is probably due to the tunneling current originating from both sides as the tip is trying to enter the deep groove and drawing current from its sides, thus blurring the images. More detailed structure can be resolved in sections of the minor groove. The periodicity of the helix is approximately 3.5 nm, and the width of the minor groove is 1.2-:--1.5 nm. Besides B-DNA as discussed above, other kinds of nucleic acids, such as Z-DNA [8.17], A-DNA [8.18] and single stranded DNA [8.19], have also been imaged with STM on graphite. For example, Fig.8.5a shows the STM image of a double-stranded A-DNA fragment of about 500 bp sampled from B cells in region V of a mouse, and the corresponding model is presented in Fig.5.3b. Four periods of the helix are imaged together with clear 287

Table 8.1. A comparison of X-ray diffraction results with STM images of A-DNA

Helix pitch [run) Minor groove width [nm] Major grrove width [run] Molecular width [run] Phosphate backbone width [nm] Axial nucleotide rise [run] Base-pair angle [0] Helix symmetry

1 ~:~I""; ~ 0.5

I -"=

00

'il

/

..c

.,....

E

8 0.4 ~ 0.3

:a

0.2 NO.1 0

I

,: -_.\.

t

\

.,

-._. ~ .-

, ....... !-

---';/.~\ .. - j l

\

'"

a:l '

' ... : . "-'~i.."

'-':--:--

...

'J

1.0

2.0

3.0

5.0

4.0

Lateral displacement [nm] 0.7 " ---.----~-------,-----.,--- 1

"

. ~iCT

Scanning tunneling microscopy and its applications

Introduction to scanning tunneling microscopy

Atomic Force Microscopy Scanning Tunneling Microscopy

Introduction to Scanning Tunneling Microscopy

Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis

Mono-Cycle photonics and optical scanning tunneling microscopy

Scanning Microscopy for Nanotechnology: Techniques and Applications

Scanning force microscopy

Scanning Probe Microscopy

Scanning Electrochemical Microscopy

Scanning Auger Electron Microscopy

Scanning Force Microscopy of Polymers

Roadmap of Scanning Probe Microscopy

Roadmap of Scanning Probe Microscopy

Scanning Transmission Electron Microscopy: Imaging and Analysis

Scanning Transmission Electron Microscopy: Imaging and Analysis

Scanning transmission electron microscopy : imaging and analysis

Scanning Probe Microscopy in Nanoscience and Nanotechnology

Applied scanning probe methods 5 scanning probe microscopy techniques

Laser Scanning, Theory and Applications

environmental scanning electron microscopy (VP-ESEM)

Applied scanning probe methods 2. Scanning probe microscopy techniques

Applied Scanning Probe Methods VIII: Scanning Probe Microscopy Techniques

Filler-Reinforced Elastomers Scanning Force Microscopy

Scanning Force Microscopy of Polymers (Springer Laboratory)

Exploring Scanning Probe Microscopy with MATHEMATICA

Acoustic Microscopy: Fundamentals and Applications

Methods of Experimental Physics: Scanning Tunneling Microscopy (Experimental Methods in the Physical Sciences)

Exploring Scanning Probe Microscopy with MATHEMATICA

Biological Low-Voltage Scanning Electron Microscopy

Raman Microscopy: Developments and Applications

Scanning tunneling microscopy and its applications

Introduction to scanning tunneling microscopy

Atomic Force Microscopy Scanning Tunneling Microscopy

Introduction to Scanning Tunneling Microscopy

Scanning Tunneling Microscopy in Surface Science, Nanoscience and Catalysis

Mono-Cycle photonics and optical scanning tunneling microscopy

Scanning Microscopy for Nanotechnology: Techniques and Applications

Scanning force microscopy

Scanning Probe Microscopy

Scanning Electrochemical Microscopy

Scanning Auger Electron Microscopy

Scanning Force Microscopy of Polymers

Roadmap of Scanning Probe Microscopy

Roadmap of Scanning Probe Microscopy

Scanning Transmission Electron Microscopy: Imaging and Analysis

Scanning Transmission Electron Microscopy: Imaging and Analysis

Scanning transmission electron microscopy : imaging and analysis

Scanning Probe Microscopy in Nanoscience and Nanotechnology

Applied scanning probe methods 5 scanning probe microscopy techniques

Laser Scanning, Theory and Applications

environmental scanning electron microscopy (VP-ESEM)

Applied scanning probe methods 2. Scanning probe microscopy techniques

Applied Scanning Probe Methods VIII: Scanning Probe Microscopy Techniques

Filler-Reinforced Elastomers Scanning Force Microscopy

Scanning Force Microscopy of Polymers (Springer Laboratory)

Exploring Scanning Probe Microscopy with MATHEMATICA

Acoustic Microscopy: Fundamentals and Applications

Methods of Experimental Physics: Scanning Tunneling Microscopy (Experimental Methods in the Physical Sciences)

Exploring Scanning Probe Microscopy with MATHEMATICA

Biological Low-Voltage Scanning Electron Microscopy

Raman Microscopy: Developments and Applications

Recommend Documents